The Feasibility of Electrochemical Ammonia Synthesis in Molten LiCl–KCl Eutectics

Article abstract of DOI:10.1002/anie.201909831

Molten LiCl and related eutectic electrolytes are known to permit direct electrochemical reduction of N₂ to N^3? with high efficiency. It had been proposed that this could be coupled with H₂ oxidation in an electrolytic cell to produce NH₃ at ambient pressure. Here, this proposal is tested in a LiCl–KCl–Li₃N cell and is found not to be the case, as the previous assumption of the direct electrochemical oxidation of N^3? to NH₃ is grossly over-simplified. We find that Li₃N added to the molten electrolyte promotes the spontaneous and simultaneous chemical disproportionation of H₂ (H oxidation state 0) into H^? (H oxidation state ?1) and H⁺ in the form of NH^2?/NH₂^?/NH₃ (H oxidation state +1) in the absence of applied current, resulting in non-Faradaic release of NH₃. It is further observed that NH^2? and NH₂^? possess their own redox chemistry. However, these spontaneous reactions allow us to propose an alternative, truly catalytic cycle. By adding LiH, rather than Li₃N, N₂ can be reduced to N^3? while stoichiometric amounts of H^? are oxidised to H₂. The H₂ can then react spontaneously with N^3? to form NH₃, regenerating H^? and closing the catalytic cycle. Initial tests show a peak NH₃ synthesis rate of 2.4×10^?8 mol cm^?2 s^?1 at a maximum current efficiency of 4.2 %. Isotopic labelling with ¹⁵N₂ confirms the resulting NH₃ is from catalytic N₂ reduction.

Full text of DOI:10.1002/anie.201909831

Angewandte
Research Articles
Chemie
International Edition: DOI: 10.1002/anie.201909831
German Edition: DOI: 10.1002/ange.201909831
Nitrogen Reduction Hot Paper
The Feasibility of Electrochemical Ammonia Synthesis in Molten LiCl–
KCl Eutectics
Ian J. McPherson, Tim Sudmeier, Joshua P. Fellowes, Ian Wilkinson, Tim Hughes, and
S. C. Edman Tsang*
Abstract: Molten LiCl and related eutectic electrolytes are
process while developing new technologies that are compat-
ible with renewable energy resources. Simultaneously, am-
monia has begun to gain attention as a green energy vector,
particularly for high capacity storage in isolated locations.
With an energy density similar to methanol, zero carbon
emission on combustion and widespread storage and distri-
bution already demonstrated by the agricultural sector, it is an
attractive candidate. However, the challenge remains the de-
carbonisation of its production.
3
ꢀ
known to permit direct electrochemical reduction of N to N
2
with high efficiency. It had been proposed that this could be
coupled with H oxidation in an electrolytic cell to produce
2
NH3 at ambient pressure. Here, this proposal is tested in
a LiCl–KCl–Li N cell and is found not to be the case, as the
3
previous assumption of the direct electrochemical oxidation of
3
ꢀ
N
to NH is grossly over-simplified. We find that Li N added
3 3
to the molten electrolyte promotes the spontaneous and
simultaneous chemical disproportionation of H (H oxidation
Currently most ammonia is produced using the Haber–
Bosch process, employing high pressures (200 bar) and an Fe
catalyst to obtain equilibrium conversion of N and H into
2
ꢀ
+
state 0) into H (H oxidation state ꢀ1) and H in the form of
2
ꢀ
ꢀ
2
NH /NH /NH (H oxidation state + 1) in the absence of
3
2
2
[5]
applied current, resulting in non-Faradaic release of NH . It is
NH3. These complex plants generally operate on a very
large scale, running continuously at high pressure and
deriving the required H2 from steam reforming of natural
gas. However, this model is not easily converted to use
renewable energy, whose wider geographic dispersion and
inherent intermittency disfavour the use of large, continuous,
centralised reactors. In their place a new generation of
distributed, small scale, easily switched plants are envisaged
whose activity can be ramped up and down to match the
3
2
ꢀ
ꢀ
further observed that NH and NH2 possess their own redox
chemistry. However, these spontaneous reactions allow us to
propose an alternative, truly catalytic cycle. By adding LiH,
3
ꢀ
rather than Li N, N can be reduced to N while stoichiometric
3
2
ꢀ
amounts of H are oxidised to H . The H can then react
2
2
3
ꢀ
ꢀ
spontaneously with N to form NH , regenerating H and
3
closing the catalytic cycle. Initial tests show a peak NH
3
ꢀ8
ꢀ2 ꢀ1
synthesis rate of 2.4 ꢀ 10 molcm
s
at a maximum current
1
5
[6]
efficiency of 4.2%. Isotopic labelling with N confirms the
renewable energy supply. Holding back this vision are the
2
resulting NH is from catalytic N reduction.
difficulties in designing small yet efficient and productive
3
2
reactors. The kinetics of N reduction are very slow, requiring
2
Introduction
the Haber–Bosch process to operate at high temperatures.
Using temperature to increase rate carries a thermodynamic
penalty, however, as the equilibrium constant, K, decreases
significantly from room temperature (K(300 K) = 766) to
Synthetic ammonia plays a central role in modern
civilisation, accounting for as much as half of the fixed
nitrogen in the terrestrial nitrogen cycle. While production
on this huge scale is necessary to meet the food demands of
a growing population, it comes with a correspondingly large
and increasing energy requirement, consuming over 1% of
[
1]
ꢀ3
higher temperatures (K(700 K) = 9.8 ꢀ 10 , see calculation in
[
5]
section S2 in the Supporting Information (SI)). This is
compensated in the Haber-Bosch process with high reactant
pressures, which itself has significant financial and energetic
costs.
[
2,3]
the global energy supply.
Furthermore, since this energy is
currently supplied from fossil fuels, there are also significant
In contrast to thermal reactors, electrochemical cells
permit the position of equilibrium to be controlled via the cell
potential. If suitable redox reactions can be found, judicious
control over potential enables the requirement for high
pressures to be relaxed, in addition to the other benefits of
electrochemical cells such as being able to control activity
directly via applied voltage or current, rather than via
parameters such as temperature or pressure which take much
longer to change. Several electrochemical cells have been
proposed for ammonia synthesis, with those based on molten
salt electrolytes reporting the highest rates and current
CO emissions from this process, estimated at over 1% of
2
[
4]
global CO emitted in 2012. As a result, widespread efforts
2
are underway to decrease the energy consumption of this
[
*] Dr. I. J. McPherson, T. Sudmeier, J. P. Fellowes, Prof. S. C. E Tsang
Department of Chemistry, University of Oxford, Inorganic Chemistry
Laboratory, South Parks Road, Oxford, OX1 3QR (UK)
E-mail: edman.tsang@chem.ox.ac.uk
Dr. I. Wilkinson, Dr. T. Hughes
CT NTF, Siemens plc
Wharf Road, Oxford, OX29 4BP (UK)
[7–12]
efficiencies.
In particular, it has been reported that molten
LiCl, and related eutectics, possess the fascinating ability to
Supporting information and the ORCID identification number(s) for
the author(s) of this article can be found under https://doi.org/10.
3
ꢀ
3ꢀ
stabilise the N ion and establish the reversible N /N redox
2
[13–16]
1002/anie.201909831.
couple at ambient pressure.
Ito and co-workers proposed
&
&&&
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
These are not the final page numbers!

Angewandte
Research Articles
Chemie
that this could be utilised in an ammonia synthesis process, by Results and Discussion
coupling the reduction of N [Eq. (1)] at a cathode with its
2
[10]
oxidation in the presence of H [Eq. (2)] at an anode.
Ammonia Synthesis from Li N
2
3
[10]
To confirm that the cell described by Ito and co-workers
N2 þ 6 eꢀ ! 2 N3ꢀ
ð1Þ
ð2Þ
is truly capable of catalytic ammonia synthesis, we repro-
duced their setup in which N and H bubbled through Ni
2
2
foam electrodes into molten LiCl–KCl eutectic containing
3ꢀ
þ 3 H2 ! 2 NH3 þ 6 eꢀ
+
2
N
Li N (0.5 mol%). A potential of 0.7 V vs. Li /Li was applied
3
to the H /Ni foam working electrode in a three-electrode
2
setup, with a Ag/AgCl electrode as the reference and the N2/
Ni foam as the counter electrode. The current, evolved NH3
They subsequently demonstrated that NH was rapidly
3
3
ꢀ
evolved from cells containing N when H , H O and several
and N /Ni electrode potential were then sampled over the
2
2
2
[
10,17–20]
other H donors were added and potential was applied.
course of 8 hours (Figure 1) before the current became
negligible. Over the first hour the rate of ammonia evolution
Mechanistic investigation of the reaction with H found that
2
ꢀ
9
ꢀ2 ꢀ1
the rate of NH3 evolution was independent of applied
is 3 ꢀ 10 molcm s , identical to the rate reported previ-
[21]
[10]
potential, but had a first order dependence on H2. How-
ever, to date the ability of such cells to catalytically produce
ously after 40 minutes. However, after one hour the rate
begins to rise dramatically, peaking after 5.5 hours at 1.5 ꢀ
ꢀ
6
ꢀ2 ꢀ1
NH has yet to be carefully confirmed, with only initial rates
10 molcm
s
before falling. The delay in peak NH3
3
3
ꢀ
[22]
and non-catalytic conversion of N presented.
evolution is suggested to be due to the strong absorption of
+
Figure 1. Ammonia evolution under anodic potential control. H electrode potential: +0.7 V vs. Li /Li. Electrolyte: LiCl–KCl–Li N (0.5 mol%),
7
2
3
ꢀ1
ꢀ1
ꢀ1
23 K. Ni foam electrodes. Flow rates: H (working electrode) 5 mLmin , N (counter electrode) 45 mLmin , Ar (headspace) 350 mLmin .
2 2
A) Ammonia evolution rate. Inset: 0–3 hours, rate on log scale. B) Current density (left), N electrode potential (right). C) Cumulative current
2
efficiency. D: Cumulative Li N conversion into NH .
3
3
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
&&&&
These are not the final page numbers!

Angewandte
Research Articles
Chemie
ꢀ
1
Figure 2. Ammonia evolution at open circuit. Electrolyte: LiCl–KCl–Li N (0.5 mol%), 723 K. Ni felt electrodes. Flow rates: H , 5 mLmin , N2,
1
3
2
ꢀ1
ꢀ1
0 mLmin , Ar (headspace) 250 mLmin . A) Open circuit potential (left) and ammonia evolution rate (right). B) Conversion of initial Li N.
3
C) Equilibrium potential dependence of LiCl–KCl–Li N (0.5 mol%) on H partial pressure before and after reaction with H . Measured against Ag/
3
2
2
AgCl reference electrode. D) Infrared difference spectrum in NꢀH stretching region of frozen LiCl–KCl-L-i N after reaction with H . Li NH and
3
2
2
LiNH shown for reference.
2
3
ꢀ
NH by the melt in the presence of N , as reported previously
The yield of ammonia (2.5 mmol) was equivalent to
3
3
ꢀ
ꢀ
and attributed to the reaction of NH with N to form NH2
conversion of only 25 mol% of the initial Li N (10 mmol,
3
3
2
ꢀ
[22]
and NH species [Eq. (3)].
Figure 1D), making it impossible to determine if any N2
reduction occurred at all (only yields greater than
1
00 mol% would unambiguously indicate catalysis). Further-
3
ꢀ
þ NH3 ! NH2ꢀ þ NH2ꢀ
N
ð3Þ
more, we found that the current efficiency exceeded 100%,
that is, the overall charge passed was insufficient to account
for the ammonia produced (Figure 1C), revealing that the
primary ammonia evolution process here is not Faradaic as
originally attributed. To determine if there is any Faradaic
component to ammonia evolution in this system, a control
experiment was carried out under open circuit conditions
(Figure 2). The initial rates are lower than seen at 0.7 V vs.
Interestingly the current seems to be inversely correlated
with the rate, starting high but falling at the start of ammonia
evolution, eventually becoming negative after 3.5 hours (Fig-
ure 1B, left axis). This behaviour correlates with the N2
[
23]
electrode potential, which starts off more negative than the
+
H electrode (as expected) at + 0.5 V vs. Li /Li but rises
2
+
steadily to become more positive than it, reversing the
polarity of the cell (Figure 1B, right axis). This reversal
suggests that the composition of the electrolyte near the
anode is becoming harder to oxidise, possibly a result of the
depletion of N and accumulation of NH and NH2
described above.
Li /Li and there is a longer delay before any increase is
observed, however after 4 hours the rate does dramatically
increase, reaching a peak rate again at 5.5 hours of 3.6 ꢀ
ꢀ6
ꢀ2 ꢀ1
10 molcm
s
before falling to negligible rates. The final
3
ꢀ
2ꢀ
ꢀ
yield is again equivalent to only 30% conversion of the Li N
initially present. This behaviour is very similar to the case with
3
&
&&&
www.angewandte.org
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
These are not the final page numbers!

Angewandte
Research Articles
Chemie
applied current, although a slower onset is observed, and
shows that the ammonia evolved under applied potential is
not produced in a Faradaic reaction and is actually the result
of a spontaneous chemical reaction between the initial
dissolved Li N and H [Eq. (4)]. This contrasts to the original
3
2
claim of electrocatalytic N reduction for ammonia produc-
2
tion in this molten medium. Note that no ammonia was
evolved in a separate control experiment where current was
applied to the LiCl/KCl/Li N electrolyte without supplying
3
3
ꢀ
H , confirming that N is reacting with H and not residual
2
2
water.
3ꢀ
þ 3 H2 ! NH3 þ 3 Hꢀ
N
ð4Þ
The spontaneous chemical reaction is actually very similar
to the desired electrochemical oxidation reaction [Eq. (2)],
3
ꢀ
ꢀ
except the electrons from N are used to reduce H to H ,
2
rather than being transferred to the electrode. This reaction is
spontaneous in the solid state (as the Li salts), although only
ꢀ
1
[24]
slightly (D G = ꢀ1.15 kJmol ),
and has been explored
r
extensively for its role in H2 storage, via the reversible
formation of Li NH and LiNH . It is therefore proposed
[25]
2
2
that the reaction occurs stepwise here too, via the formation
2
ꢀ
ꢀ
of NH and NH2 species [Eq. (4a)–(4c), and/or via the
ꢀ
bimolecular decomposition of NH2 [Eq. (5)].
3
ꢀ
þ H2 ! NH2ꢀ þ Hꢀ
N
ð4aÞ
ð4bÞ
ð4cÞ
ð5Þ
NH þ H2 ! NH2ꢀ þ Hꢀ
2ꢀ
ꢀ
ꢀ
NH2 þ H2 ! NH þ H
3
ꢀ
2
NH2 ! NH2ꢀ þ NH3
Figure 3. Cyclic voltammetry of the proposed N,H species. First
Evidence that these reactions occur in the molten state is
(dotted/solid lines) and second (dashed lines) cycles shown. Arrows
provided by two direct observations: first, the significant H2
partial pressure dependence of the open circuit potential; and
secondly, the appearance of IR peaks in the NꢀH stretching
3
indicate starting potential and direction of scan. A,B) Li N (0.5 mol%);
C) LiNH (0.2 mol%); D) LiNH (1.2 mol%); E,F) Li NH (1 mol%);
2
2
2
ꢀ1
G,H) LiH (1 mol%). A,C,E,G) Scan rate: n=2 Vs
B,D,F,H) n=100 mVs
;
region following hydrogenation. Figure 2C shows the open
ꢀ1
.
circuit potential of a LiCl–KCl–Li N (0.5 mol%) melt,
3
measured at a Ni wire, recorded as a function of P(H ) both
2
evidence for Equation (4a)–(4c) is provided by IR spectros-
copy of the melt frozen midway through reaction, where
just after H was introduced, and after 20% conversion of the
2
Li N to NH3 by H2 bubbling. There is a strong positive
3
peaks are observed in the region expected for both LiNH and
correlation between P(H ) and the potential, indicating not
2
2
[
26]
Li NH (Figure 2D). This is also supported by recent in situ
only that H is involved in a potential determining equilib-
2
2
infrared experiments in which NꢀH stretches were observed
rium, but that it is the oxidised partner. This is consistent with
the potential determining equilibrium being that between H2
[27]
when H was bubbled through Li N.
2
3
ꢀ
The stepwise change in equilibrium potential of the H2
and H [Eq. (6)], giving rise to Nernst equation [Eq. (7)],
0
ꢀ
electrode observed during the open circuit control measure-
ment, and its change in polarity during the constant potential
where E ’ is the formal potential of the H /H couple, P(H ) is
2
2
ꢀ
ꢀ
the partial pressure of H and c(H ) the mole fraction of H .
2
0
.7 V experiment, further suggest that the intermediates
H2 þ 2 e Ð 2 Hꢀ
ð6Þ
ð7Þ
formed are themselves redox active. The redox properties of
all proposed components of the electrolyte were therefore
characterized to understand their interactions and determine
if a truly catalytic path to ammonia synthesis in this system is
indeed possible.
0
0
ꢀ
ꢀ 2
E ¼ E ðH =H Þ þ ðRT=2 FÞ ln PðH Þ=cðH Þ
2
2
This is further supported by the translation of the curve
towards more negative potentials as the reactions proceeds,
ꢀ
consistent with an increase in H mole fraction. Evaluation of
the slope in terms of Equation (7) suggests the number of
electrons, n = 2.5, close to the expected value of 2. Additional
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
&&&&
These are not the final page numbers!

Angewandte
Research Articles
Chemie
3
ꢀ
3ꢀ
Voltammetric Analysis of Redox Active Species
on the earlier assignment of the N /N couple, N would not
2
stable above 0.19 V and a net oxidation would be observed
instead.
The reactions of the proposed NH species were charac-
terized using cyclic voltammetry, which provides character-
istic potentials for the onset of redox reactions as well as
NH þ 2 e Ð N3ꢀ þ Hꢀ
2ꢀ
ð10Þ
insight into the reaction kinetics. In LiCl–KCl–Li N
3
2
ꢀ
It is therefore suggested that NH is specifically adsorbed
at high potential during the scan, such that current observed
at a2/c2 arises from reductive desorption [Eq. (11)]. This can
be followed by decomposition to form N(ad) and H(ad)
(
0.5 mol%) a series of five reversible redox couples is
ꢀ1
observed at fast scan rates (2 Vs ), labelled a1/c1 to a5/c5
Figure 3A). The a1/c1 couple (Emid = 0.19 V, DE = 0.12) is
(
p
3
ꢀ
assigned to the six electron reduction of N to N [Eq. (1)]
2
[
Eq. (12)], which themselves can be reduced [Eq. (13), (14)],
presumably at the same potentials as c1 and c3, respectively.
The production of H from NH during a2/c2 explains why the
current for the non-N based a3/c3 couple is of similar
magnitude to the N based processes.
having the lowest midpoint potential of all the couples,
slightly more negative than the reported formal potential
2
ꢀ
0
[28]
(
E ’ = 0.382 V). A similar reduction peak appears when N2
is bubbled over a Ni wire in pure LiCl–KCl electrolyte
(
(
Figure S2), while a slow scan of a Ni foam electrode in Li N
0.5 mol%) shows a linear current-potential relationship also
3
NHðadÞ þ 2 e Ð NH ꢀ
NHðadÞ Ð NðadÞ þ HðadÞ
NðadÞ þ 3 e Ð N3ꢀ
2
ð11Þ
ð12Þ
ð13Þ
ð14Þ
with 0.19 Vas the x intercept (Figure S3). Examination of the
blank voltammogram reveals a single reduction peak around
the position of c4, attributed to the one electron reduction of
ꢀ
[29]
residual H O to H2 + OH .
The magnitude of this
2
reduction current, along with a suitable estimate of the
electrode area, can be used to deduce the amount of water in
the melt, which is estimated to be 0.03 mol%. This non-
HðadÞ þ e Ð Hꢀ
3
ꢀ
Reduction of N(ad) is evident in the slower scan rate scans
in both NH and NH2 as a second reduction peak at the c1
position (Figure 3D,F). The much greater significance of the
a1/c1 peaks in both NH2 scans is tentatively attributed to the
more facile loss of H2 from NH2 than NH , resulting in
a lower coverage of H(ad) and thereby favouring reductive
negligible amount of water may react with some of the N to
generate the same protonated species as that formed via
2
ꢀ
ꢀ
reaction with H [Eq. (8), (9)], explaining the presence of
2
ꢀ
further peaks in the voltammetry. To confirm that these
species are indeed formed, voltammograms were also col-
lected of Li NH, LiNH and LiH.
ꢀ
2ꢀ
2
2
3
ꢀ
desorption of N(ad) as N over reaction with H(ad) and
2
ꢀ
3
ꢀ
þ H O Ð NH þ OHꢀ
2ꢀ
N
ð8Þ
ð9Þ
desorption as NH .
2
Following the same approach, the a5/c5 couple (Emid
.19 V, DE = 0.20) can be assigned to the adsorption/desorp-
p
tion of the NH2 ion [Eq. (15)], with similar decomposition
into N(ad) and H(ad) possible [Eq. (16)].
=
NH þ H O Ð NH2 þ OHꢀ
2
ꢀ
ꢀ
2
1
ꢀ
ꢀ
1
The voltammogram of LiNH2 at 2 Vs (Figure 3C)
consists of four pairs of peaks, aligning well with the a1/c1,
a2/c2, a3/c3 and a5/c5 couples seen in Li N. At the slower scan
ꢀ
3
NH ðadÞ þ e Ð NH
ð15Þ
ð16Þ
2
2
ꢀ
1
rate of 100 mVs (Figure 3D) only a1/c1 is reversible, with c2
appearing much more significant than a2 and the a5/c5 couple
replaced by a noisy limiting oxidation current. In comparison,
NH
ðadÞ Ð NðadÞ þ 2 HðadÞ
2
the voltammograms of Li NH only shows a single clear couple
The difference in N:H stoichiometry can also be used to
2
(
Figure 3E), aligning with a2/c2, which also appears to
explain the different behaviour seen at high potential. While
two NH(ad) molecules must decompose (in adjacent sites) in
become irreversible at the slower scan rate (Figure 3F). The
scan also shows evidence for a high potential shoulder on a2
matching the potential of the a3 peak. The onset of a new
oxidation process is also observed at potentials more positive
than 1.5 V. The occurrence of the a2/c2 couple in both NH
and NH2 is unsurprising, given that even in the absence of
H , when LiNH was added to LiCl–KCl at 723 K around
order to evolve H , giving the rate equation a quadratic
2
2
ꢀ
dependence on NH concentration, and the requirement for
ꢀ
neighbouring free surface sites, only one NH
decompose to release H
vacant sites, making the process much faster. Thus, for the
same concentration, oxidation to H is likely to be much faster
has to
2
2
ꢀ
, possibly without requiring adjacent
2
ꢀ
2
2
2
ꢀ
2ꢀ
2
5 mol% was immediately evolved as NH (data not shown),
for NH
2
than for NH . This matches well with the presence
3
ꢀ
presumably via the reaction shown in Equation (5), suggest-
ing a corresponding 25 mol% is converted into NH . This
of a noisy, limiting current for NH
2
oxidation but only a high
2
ꢀ
2ꢀ
potential current peak for NH oxidation.
suggests that measurements of LiNH are actually measure-
2
Finally, the a3/c3 couple (E = 0.73 V, DE = 0.20) is
ꢀ
2ꢀ
mid
p
ments of both NH2 and NH , here in the apparent ratio 2:1.
Focusing on the assignment of the a2/c2 couple (Emid
.51 V, DE = 0.12), there are limited options for reactions of
ꢀ
readily assigned to the H /H couple [Eq. (6)], based on the
2
=
voltammetry of LiH (Figure 3G,H) and the previously
[
30]
0
reported midpoint potential (0.755 V at 673 K). Interest-
p
2
ꢀ
ꢀ
NH which could explain the irreversible reduction behav-
iour seen at slow scan rates in Figure 3D. One possible
reduction would form N and H [Eq. (10)], however, based
ingly at slow scan rates (Figure 3H) H also exhibits a noisy
ꢀ
limiting oxidation current, similar to NH2 , further support-
3
ꢀ
ꢀ
ing the assignment of the limiting current to H evolution.
2
&
&&&
www.angewandte.org
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
These are not the final page numbers!

Angewandte
Research Articles
Chemie
Ammonia Synthesis from LiH
From our analysis of the original LiCl–KCl–Li N system
3
we conclude that the addition of Li N provides a non-catalytic
3
route to ammonia evolution, casting doubt about the feasi-
bility of the original cell. However, our analysis also suggests
that a true, alternative pathway to catalytic ammonia
production does exist. If the cell can be made to reduce N
2
3
ꢀ
to N , then this should spontaneously react with H gas to
2
ꢀ
form ammonia and H ions. The catalytic cycle could then be
closed by oxidizing the H at the anode. The desired cell
ꢀ
reaction would then be represented by Equation (1), (4) and
(
6), and is shown schematically in Figure 4. Note that it might
Figure 4. Catalytic cycle proposed for electrochemical NH synthesis in
molten LiCl–KCl–LiH. Species added to the reaction are shown in blue,
3
Figure 5. NH synthesis results based on N reduction in a LiCl–KCl–
3
2
ꢀ
1
LiH (1 mol%) electrolyte. T=723 K, Flow rates: H =5 mLmin
,
2
remaining species in black are generated in situ, sole product NH is
3
ꢀ1
ꢀ1
N =45 mLmin , Ar headspace purge=350 mLmin . Geometric
2
shown in green. Red dashed lines indicate opportunities for unproduc-
tive side reactions.
2
+
electrode area=0.126 cm . H electrode potential= +1.0 V vs. Li /Li.
2
has decreased (Figure 5 and Section S3, SI), presumably along
ꢀ
ꢀ
appear more obvious to couple oxidation of H (g) than
with the H ion concentration, suggesting that H may also
2
ꢀ
H (sol) in this reaction, however its standard potential in
trap NH via the reverse reaction of Equation (4c). The rate
of NH3 evolution peaks at 2.8 ꢀ 10 molcm
2.2 hours online, which is among the highest reported rate
for any electrochemical ammonia synthesis. The overall
current efficiency, based on the total charge passed and the
total ammonia produced, was 4.2%, while the average molar
3
0
+
+
ꢀ8
ꢀ2 ꢀ1
LiCl–KCl is highly positive (E ’(H /H ) =+ 2.610 V vs. Li /
s
after
2
calc
[
31,32]
0
Li),
almost the same as the Ni electrode itself (E ’ =
[
32]
+
2.615 V),
leading to much higher electrolysis voltages
and the risk of corroding the electrode (see S6, Supporting
Information).
We therefore setup the LiH cell to confirm that catalytic
ammonia production is feasible. The operating conditions
were modified slightly (1 mol% LiH in place of 0.5 mol%
conversion of N was 0.011 mol%. This can be compared to
2
the maximum conversion of 0.006 mol% expected for purely
thermal equilibrium of the gas mixture over the course of the
experiment (calculations in Supporting Information, Ta-
ble S3).
Li N, 1.0 V anode potential, same gas electrodes and flow
3
rates) and the cell tested for NH synthesis activity over the
3
course of 3 hours (Figure 5, S3 Supporting Information). Note
To confirm that NH evolved is a result of N reduction in
3
2
ꢀ
that 1.0 V was chosen to favour productive H oxidation over
our system rather than from an extraneous source, an isotope
labelling experiment was performed (Figure 6). All condi-
tions were maintained as in the previous LiH experiment,
ꢀ
unproductive reaction of NH2 , based on the voltammetry
(
H
Figure 3), and LiH was added from the beginning to facilitate
ꢀ
oxidation before the homogeneous melt reactions
with the exception of the N source, which was replaced with
2
1
5
[
Eq. (4a)–(4c)] reached steady state.
N for approximately 1 hour (the contents of one cylinder),
2
1
4
Initially very large current densities were observed, but
after which the source was switched back to N2 for the
these decreased over the course of one hour. Similar to the
remaining time. Very similar current densities are observed in
3
ꢀ
15
N
case, NH evolution is not observed until after the current
the N experiment (Figure 6A), along with a similar (but
3
2
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
&&&&
These are not the final page numbers!

Angewandte
Research Articles
Chemie
1
4
product distribution. The presence of NH at early times
3
1
4
presumably corresponds to reduction of residual N in the
2
reactor or gas tubing (despite evacuating and refilling the
system with Ar multiple times). Furthermore, usually the
system is purged with N before an experiment, allowing the
2
melt to equilibrate before applying the voltage, however to
1
5
conserve N , Ar was used to purge and the gas flow was only
2
1
5
switched to N2 just before applying the voltage. This
difference may explain the slight delay in the peak rate
appearing.
The significant (1900 s) delay between N flow ending
and the evolution of NH provides an estimate of the time
1
5
2
1
5
3
3
ꢀ
taken for the nascent N to react with 3H to form NH and
2
3
leave the melt. Interestingly, there is no corresponding sharp
1
4
peak for NH . This indicates that exchange of H between
3
ꢀ
2
ꢀ
ꢀ
NH and NH2 is faster than the reaction between NH2 and
H , such that while it takes a long time for successive reactions
2
1
5
3ꢀ
15
14 3ꢀ
of
N
to generate NH , later when
N
forms it can
3
1
5
ꢀ
15
2ꢀ
rapidly equilibrate with the existing NH2 and NH
species, permitting NH evolution almost immediately after
1
4
3
switching gases.
It is clear from the behaviour of the current that the cell
reactions has not yet reached the steady state as summarised
in Figure 4. The reason for this becomes clear when the
conditions required for the spontaneous solution reaction-
s (4a)–(4c) to occur are considered. The H disproportionation
reactions proposed to occur in the electrolyte can be
separated into two half reactions:
3
ꢀ
þ 6 eꢀ
2
3
N
þ 3 H
2
! 2 NH
3
ð2Þ
ð6Þ
H2 þ 6 eꢀ ! 6 Hꢀ
Figure 6. Isotope labelling study to confirm presence of N reduction.
To be spontaneous the potential of the disproportionation
2
1
5
ꢀ
Reaction conditions as in Figure 6, except N was used for the first
reaction must be positive, that is, E(H /H )ꢀE(NH /H ) > 0.
2
2
3
2
hour. A) Current (black line) NH production rate determined by ISE
3
This difference is calculated over a range of P(NH ) and
3
1
5
14
(
coloured points) and distribution of NH and NH produced at
ꢀ
ꢀ
3
3
c(H ) values in Figure 7. It can be seen that the H
concentration must decrease from 1 mol% before sponta-
different times during reaction (based on the absolute values found by
integrating the respective multiplet, shown in panel B, below).
1
neous reaction to form NH becomes possible, explaining the
B) H NMR spectra of the trap solutions. Spectra shown offset for
3
clarity.
1
slightly extended) delay in NH evolution. H NMR spectros-
3
copy was used to distinguish between the isotopes of NH3
present in the trap solution, based on the different coupling
1
15
1
14
between H and N (I = = ) and N (I = 1) as previously
2
[
33]
described. The NMR spectra (Figure 6B) show only the
characteristic triplet of NH in the first three time points,
1
4
3
followed by the sudden appearance of a doublet attributed to
1
5
NH , corresponding with the peak in the rate. The presence
3
1
5
of NH confirms that N is being reduced in the reactor and
3
2
thus this represents the first demonstration of truly catalytic
ammonia synthesis in LiCl–KCl eutectic, that is, without
[
10,17,21]
starting from Li N in the melt, as in previous studies.
3
Performing the isotope experiment in this way not only
3ꢀ
Figure 7. Reaction conditions where the hydrogenation of N is
provides confirmation that N is being reduced in our system,
2
spontaneous (red) or non-spontaneous (blue). Defined as red region
where E(H /H )ꢀE(NH /H )>0. T=723 K, c(N )=10 , P(H )/
ꢀ
3ꢀ
ꢀ4
but enables the origin of the decay in current and rate over
2
3
2
2
1
5
14
0
time to be examined via the evolution of the NH / NH
P =0.0125.
3
3
&
&&&
www.angewandte.org
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
These are not the final page numbers!

Angewandte
Research Articles
Chemie
ꢀ
delay in NH evolution observed. Should P(NH ) or c(H )
(ThermoFisher Scientific) to determine the rate of ammonia evolu-
3
3
1
5
tion. Use of N (98% + , CK Isotopes Ltd.) was used to confirm N
then rise significantly during operation the melt reactions
become unfavourable and NH evolution stops. To operate at
steady state the cell must therefore be operated under closely
controlled conditions. It appears from experiments in Figure 5
and 6 that conditions to achieve this steady state have not yet
2
2
1
5
reduction via the presence of NH in the acid trap, as detected by
3
3
1
H NMR spectroscopy. Further experimental details can be found in
supporting information (S1 and Tables S1–S4, SI).
been found, explaining the drop in NH production over time.
3
Acknowledgements
In addition, corrosion or passivation of the Ni foam electrodes
could affect the rate. Both electrodes appear black after the
reaction (Figure S5), although there is no sign of NiO in the
X-ray diffraction patterns of pulverised electrodes (Fig-
This work was funded by Siemens PLC. T.S. acknowledges an
EPSRC CASE studentship and support from a Royal Com-
mission for the Exhibition of 1851 fellowship. N. Vranje sˇ , A.
Garzon Gonzalez, T. Adams, L. Hill and S. Davies (University
of Oxford) and R. Selway (Inspired Engineering Design Ltd.)
are thanked for the design and construction of equipment.
ure S6), just soluble NiCl . The next step in the development
2
of this cell is therefore the optimisation of the reaction
conditions to maximise production of NH while maintaining
3
low concentrations of NH to remain in the spontaneous
3
regime. One way to achieve this may be by the introduction of
porous membranes to separate the electrolyte at the anode,
from the bulk and from that at the cathode, minimising
Conflict of interest
[34]
crossover, as suggested recently.
The authors declare no conflict of interest.
Keywords: ammonia · electrochemistry ·
heterogeneous catalysis · nitrogen · renewable resources
Conclusion
From this work, we have shown that the previous
assumption of direct electrochemical reaction of N and H2
2
in molten LiCl–KCl to form NH with high efficiency is not
3
correct. Detailed investigation into the electrochemical
mechanism has revealed the existence of a number of rapid
3
ꢀ
non-electrochemical reactions of N with H to form NH3
2
[
1] P. M. Vitousek, J. D. Aber, R. W. Howarth, G. E. Likens, P. A.
Matson, D. W. Schindler, W. H. Schlesinger, D. G. Tilman,
Ecological Appl. 1997, 7, 737 – 750.
2
ꢀ
ꢀ
along with NH and NH2 , which are shown to be electro-
chemically active by cyclic voltammetry. This insight led us to
propose a new, truly catalytic reaction Scheme. Based on this
we demonstrate electrochemical NH synthesis from N and
[2] International Energy Agency, Tracking Industrial Energy Effi-
ciency and CO2 Emissions, International Energy Agency, 2007.
3
2
[
3] World Energy Outlook 2007 (Ed.: International Energy Agen-
cy), OECD, Paris, 2007.
4] T. Brown, “Ammonia production causes 1% of total global GHG
emissions”; can be found under https://ammoniaindustry.com/
ammonia-production-causes-1-percent-of-total-global-ghg-
emissions, 2016..
H at some of the highest rates yet reported, and confirm this
2
1
5
by N labelling. The understanding provided by this work
2
[
should enable more extensive optimisation of this cell and
realisation of longer term sustained electrochemical ammonia
synthesis.
[
[
[
[
[
5] M. Appl, Ammonia: Principles and Industrial Practice, Wiley-
VCH, Weinheim, 1999.
6] E. Morgan, J. Manwell, J. McGowan, Renewable Energy 2014,
Experimental Section
72, 51 – 61.
7] I. J. McPherson, T. Sudmeier, J. Fellowes, S. C. E. Tsang, Dalton
Trans. 2019, 48, 1562 – 1568.
8] S. Giddey, S. P. S. Badwal, A. Kulkarni, Int. J. Hydrogen Energy
Electrochemical experiments were carried out in a stainless steel cell
mounted inside an electric furnace (Figure S1). An alumina crucible
suspended from the top of the cell held the LiCl/KCl mixture
2013, 38, 14576 – 14594.
(59:41 mol%, 99% Alfa Aesar), which was extensively dried before
9] V. Kyriakou, I. Garagounis, E. Vasileiou, A. Vourros, M.
use. Gas electrodes were formed of rolled Ni foam inserted into
alumina tubes, with Ni wire connections. The reference electrode was
Ag/AgCl, prepared by adding AgCl (99.9%, Alfa Aesar) to LiCl-KCl
Stoukides, Catal. Today 2017, 286, 2 – 13.
[
[
[
10] T. Murakami, T. Nishikiori, T. Nohira, Y. Ito, J. Am. Chem. Soc.
2003, 125, 334 – 335.
(0.05 mol%) in a Pyrex tube with one end closed, and contacting the
11] S. Licht, B. Cui, B. Wang, F.-F. Li, J. Lau, S. Liu, Science 2014,
345, 637 – 640.
12] J. M. McEnaney, A. R. Singh, J. A. Schwalbe, J. Kibsgaard, J. C.
Lin, M. Cargnello, T. F. Jaramillo, J. K. Nørskov, Energy
Environ. Sci. 2017, 10, 1621 – 1630.
salt with a Ag wire (99.9%, Alfa Aesar). A new electrode was used
for each measurement. All potentials are converted into the Li /Li
scale by correlation with the potential of zero current in cyclic
voltammograms of a Ni wire after sweeping to Li reduction
potentials. The cell was operated under potentiostatic control using
either a Vertex or CompactStat potentiostat (Ivium Technologies) at
+
+
[
[
[
[
13] A. Bonomi, M. Hadate, C. Gentaz, J. Electrochem. Soc. 1977,
1
24, 982 – 986.
14] A. Bonomi, M. Hadate, F. Breda, J. Electrochem. Soc. 1979, 126,
48 – 251.
15] T. Goto, M. Tada, Y. Ito, J. Electrochem. Soc. 1997, 144, 2271 –
276.
16] Y. Ito, T. Goto, J. Nucl. Mater. 2005, 344, 128 – 135.
7
23 K. Mass flow controllers supplied N (99.998%, BOC) and H
2
2
(99.99%, BOC) to the respective electrodes, while the cell headspace
2
was purged with Ar (99.998%, BOC). Outlet gas passed was through
an acid trap containing dilute H SO (0.01m, 50 mL). The acid trap
was changed periodically and analysed with an ion selective electrode
2
4
2
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
&&&&
These are not the final page numbers!

Angewandte
Research Articles
Chemie
[
[
[
[
[
[
[
17] T. Murakami, T. Nohira, T. Goto, Y. H. Ogata, Y. Ito, Electro-
[27] N. Serizawa, K. Takei, T. Nishikiori, Y. Katayama, Y. Ito,
Electrochemistry 2018, 86, 88 – 91.
[28] T. Goto, Y. Ito, Electrochim. Acta 1998, 43, 3379 – 3384.
[29] S. H. White, Ionic Liquids, Plenum Press, New York, 1981,
p. 450.
[30] T. Nohira, Y. Ito, J. Electrochem. Soc. 2002, 149, E159 – E165.
[31] H. A. Laitinen, J. A. Plambeck, J. Am. Chem. Soc. 1965, 87,
1202 – 1206.
[32] H. C. Gaur, H. L. Jindal, Electrochim. Acta 1968, 13, 835 – 842.
[33] F. Zhou, L. Miguel Azofra, M. Ali, M. Kar, A. N. Simonov, C.
McDonnell-Worth, C. Sun, X. Zhang, D. R. MacFarlane, Energy
Environ. Sci. 2017, 10, 2516 – 2520.
chim. Acta 2005, 50, 5423 – 5426.
18] T. Murakami, T. Nohira, Y. Araki, T. Goto, R. Hagiwara, Y. H.
Ogata, Electrochem. Solid-State Lett. 2007, 10, E4 – E6.
19] T. Murakami, T. Nohira, Y. H. Ogata, Y. Ito, Electrochem. Solid-
State Lett. 2005, 8, D12 – D14.
20] T. Murakami, T. Nohira, Y. H. Ogata, Y. Ito, J. Electrochem. Soc.
2
005, 152, D109 – D112.
21] T. Murakami, T. Nishikiori, T. Nohira, Y. Ito, J. Electrochem. Soc.
005, 152, D75 – D78.
2
22] N. Serizawa, H. Miyashiro, K. Takei, T. Ikezumi, T. Nishikiori, Y.
Ito, J. Electrochem. Soc. 2012, 159, E87 – E91.
23] Note that without the involvement of the electrode, the normal-
isation used is irrelevant and the units of the spontaneous rate
should therefore be considered arbitrary.
[34] Y. Ito, T. Nishikiori, H. Tsujimura, Faraday Discuss. 2016, 190,
307 – 326.
[
24] C. W. Bale, E. Bꢁlise, P. Chartrand, S. A. Decterov, G. Eriksson,
A. E. Gheribi, K. Hack, I. H. Jung, Y. B. Kang, J. Melancon,
et al., Calphad 2016, 54, 35 – 53.
Manuscript received: August 3, 2019
[
25] P. Chen, Z. Xiong, J. Luo, J. Lin, K. L. Tan, Nature 2002, 420,
Revised manuscript received: September 11, 2019
Accepted manuscript online: September 19, 2019
302 – 304.
[
26] Y. Kojima, Y. Kawai, J. Alloys Compd. 2005, 395, 236 – 239.
Version of record online: && &&
,
&&&&
&
&&&
www.angewandte.org
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
These are not the final page numbers!

Angewandte
Research Articles
Chemie
Research Articles
Nitrogen Reduction
The key challenges and opportunities of
using molten LiCl eutectics as media for
the direct electrochemical reduction of N2
to ammonia are identified. By adding LiH,
rather than Li N, N can be reduced to
I. J. McPherson, T. Sudmeier,
J. P. Fellowes, I. Wilkinson, T. Hughes,
S. C. E Tsang*
&&&& — &&&&
3
2
3
ꢀ
ꢀ
N
while stoichiometric amounts of H
The Feasibility of Electrochemical
Ammonia Synthesis in Molten LiCl–KCl
Eutectics
are oxidised to H . The H can then react
2
2
3
ꢀ
spontaneously with N to form NH3,
ꢀ
regenerating H and closing the catalytic
cycle.
Angew. Chem. Int. Ed. 2019, 58, 2 – 11
ꢀ 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
&&&&
These are not the final page numbers!