RES EARCH

Frontiers Science Program Long-Term Fellowship (A.C.); a
Washington Research Foundation Senior Fellowship (A.C.); Howard
Hughes Medical Institute research (A.C.); Howard Hughes Medical
Institute Hanna Gray fellowship grant GT11817 (N.B.). An S10
award (S10OD032290 to D.B., D.V., J.K., and J.Q.) funded the
purchase of a Glacios microscope. SAXS data were collected
at the Advanced Light Source (ALS) SIBYLS beamline on behalf
of US DOE-BER, through the Integrated Diffraction Analysis
Technologies (IDAT) program. Additional support comes from
the NIGMS project ALS-ENABLE (P30 GM124169) and a High-End
Instrumentation Grant (S10OD018483). The Berkeley Center for
Structural Biology is supported in part by the National Institutes
of Health (NIH), National Institute of General Medical Sciences,
and the Howard Hughes Medical Institute. This research used
resources of the ALS, a US DOE Office of Science User Facility under
contract DE-AC02-05CH11231. Some of this work was performed
at the Pacific Northwest Center for Cryo-EM (PNCC), which
was supported by NIH grant U24GM129547 and performed at
the PNCC at Oregon Health & Science University and accessed
through EMSL (grid.436923.9), a DOE Office of Science User
Facility sponsored by the Office of Biological and Environmental the supplementary materials. All the EM maps have been
Research. Molecular graphics and analyses were performed with deposited in the Electron Microscopy Data Bank (accession
UCSF Chimera, developed by the Resource for Biocomputing, codes EMD-25575, EMD-25576, EMD-25577, EMD-25578, EMD-25579,
Visualization, and Informatics at the University of California, and EMD-25580). All code is available from GitHub (https://github.
San Francisco, with support from NIH P41-GM103311. Author com/alexiscourbet/) and is archived on Zenodo (34).
contributions: Conceptualization: A.C. and D.B. Methodology:
A.C., D.B., J.H., and J.K. Software: A.C., Y.H., C.X., S.E.B., SUPPLEMENTARY MATERIALS
G.U., U.N., P.B., D.B., D.S., A.M., N.K., W.S., and N.B. Validation:
science.org/doi/10.1126/science.abm1183
A.C., D.B., J.H., J.K., and Y.H. Formal analysis: A.C., J.H., and
Materials and Methods
N.B. Investigation: A.C., J.H., N.B., Y.-J.P., A.N., D.N., and J.Q.
Figs. S1 to S22
Resources: A.C., D.B., J.H., J.K., and J.Q. Data curation: A.C.,
Tables S1 to S3
J.H., and Y.H. Writing – original draft: A.C. Writing – review &
References (35–59)
editing: A.C., D.B., J.H., Y.H., and J.K. Visualization: A.C., J.H.,
Movies S1 to S5
and N.B. Supervision: D.B. and J.K. Project administration:
Data S1 to S5
A.C. and D.B. Funding acquisition: D.B., J.K., D.V., A.C., and
MDAR Reproducibility Checklist
Y.H. Competing interests: A.C., D.B., J.H., J.K., N.B., and
Y.H. are inventors on a provisional patent application submitted
by the University of Washington for the design, composition,
and function of the proteins created in this study. Data and 26 August 2021; accepted 21 March 2022
materials availability: All data are available in the main text or 10.1126/science.abm1183

◥ As for any compression process, the work

REPORTS associated with electrochemical H2 compres-
sion is minimized by operating isothermally
CATALYSIS (12). In a continuous-flow type of PCER, the
compression ratio and associated entropy dif-
Single-step hydrogen production from NH3, CH4, ference of H2 across the membrane increase
with the extent of separation along the reactor
and biogas in stacked proton ceramic reactors (Fig. 1B and fig. S1). This entropy difference is
expelled as heat (Q = TDS) during the com-
Daniel Clark1, Harald Malerød-Fjeld1, Michael Budd1, Irene Yuste-Tirados1,2, Dustin Beeaff1, pression process, which, if left unbalanced,
Simen Aamodt1, Kevin Nguyen1, Luca Ansaloni3, Thijs Peters3, Per K. Vestre1, Dimitrios K. Pappas1, leads to gradually increasing temperature along
María I. Valls4, Sonia Remiro-Buenamañana4, Truls Norby2, Tor S. Bjørheim1, the reactor and in turn larger electric energy
Jose M. Serra4*, Christian Kjølseth1* consumption per kilogram of compressed H2
(Fig. 1B).
Proton ceramic reactors offer efficient extraction of hydrogen from ammonia, methane, and biogas More efficient isothermal operation can be
by coupling endothermic reforming reactions with heat from electrochemical gas separation and achieved by locally balancing this heat evolu-
compression. Preserving this efficiency in scale-up from cell to stack level poses challenges to tion with a reversible heat sink such as an
the distribution of heat and gas flows and electric current throughout a robust functional endothermic chemical reaction (Fig. 1C) (9).
design. Here, we demonstrate a 36-cell well-balanced reactor stack enabled by a new interconnect However, matching the spatial distribution of
that achieves complete conversion of methane with more than 99% recovery to pressurized heat from compression with the extent of
hydrogen, leaving a concentrated stream of carbon dioxide. Comparable cell performance was also chemical reactions throughout a stacked re-
achieved with ammonia, and the operation was confirmed at pressures exceeding 140 bars. The actor poses one of the main hurdles in scaling
stacking of proton ceramic reactors into practical thermo-electrochemical devices demonstrates PCERs from laboratory to commercial scale.
their potential in efficient hydrogen production. Furthermore, scaled reactors with efficient
current distribution have been hindered by

H
ydrogen can be produced from CH4-rich
streams through steam reforming and
water-gas shift (SMR+WGS, CH4 + 2H2O =
CO 2 + 4H 2 , DrxnH° = 164.7 kJ/mol) or
from the emerging C-free H-carrier NH3
through ammonia dehydrogenation (ADH,
NH3 =1/2N2 + 3/2H2, DrxnH° = 45.9 kJ/mol)
(1–3). In a conventional multistage H2 produc-
tion process, fuel combustion generates the
heat for these endothermic reactions, and sub-
sequent separation and compression occur
downstream by pressure-swing adsorption
1
2
CoorsTek Membrane Sciences AS, 0349 Oslo, Norway.
Department of Chemistry, Centre for Materials Science and
Nanotechnology, University of Oslo, 0316 Oslo, Norway.
3
Department of Sustainable Energy Technology, SINTEF
Industry, 0314 Oslo, Norway. 4Instituto de Tecnologia
Química, Consejo Superior de Investigaciones Científicas-
Universitat Politècnica de València, 46022 Valencia, Spain.
*Corresponding author. Email: jmserra@itq.upv.es (J.M.S.);
ckjolseth@coorstek.com (C.K.)
and mechanical compressors. Efficiencies typ-
ically improve with scale, which favors large,
centralized processes over distributed H2 pro-
duction for energy-carrier applications (4).
H2 can also be separated and compressed
electrochemically with proton (H+)–conducting
ceramic membranes such as Y-doped BaZrO3-
BaCeO3 solid solutions (BZCY), which are
functional and stable over a wide range of
temperatures (300° to 800°C) and chemical
environments (5–10). Proton ceramic electro-
chemical reactors (PCERs) extract pure H2
from gas mixtures by electrolytically pumping
protons across the membrane (Fig. 1A). These
offer process intensification (9) by integrating
reactions such as SMR+WGS or ADH with H2
separation and compression, high energy ef-
ficiencies by supplying heat electrically (3),
and reduced CO2 emissions when that elec-
tricity is renewable (11).
the lack of interconnect materials with high
electrical conductivity and chemical stability
up to 800°C that match the low thermal ex-
pansion coefficient (8 × 10−6 K−1) of the pre-
ferred proton conductor BZCY. Mismatches in
thermal expansion of reactor components can
lead to mechanical stresses and electrical con-
tact failures during thermal cycling.
We present an optimized reactor architec-
ture aided by multiphysics simulations and a
new expansion-matched metal/glass-ceramic
composite interconnect (IC) enabling deploy-
able modular PCER stacks that retain the en-
ergy efficiencies and H2 recoveries of single
cells (9) while achieving a 36-fold increased
H2 production capacity. The reactor is designed
with gas flows that allow internal heat ex-
change from exothermic to endothermic pro-
cesses to minimize auxiliary heat input.
Our PCER can separate H2 by decompres-
sion while recovering electric energy or by

390 22 APRIL 2022 ¥ VOL 376 ISSUE 6591 science.org SCIENCE

 RES EARCH | R E P O R T S

compression through supply of electric energy
(Fig. 1D) at pressures up to 141 bars, illustrat-
ing the range of compression ratios and as-
sociated cell voltages that can be achieved
throughout the PCER stack length. The PCER
stack is a series of six barrels, each with six
single cells connected electrically in parallel
(Fig. 2A and fig. S4A), that uses a newly de-
veloped Ni-based glass-ceramic composite ICs
(13). A conductive washer of the same material
is placed between the end of each membrane
segment and the IC plate (Fig. 2, E and F).
Pure metals with higher thermal expansions,
such as Ni or Cu, are prone to loss of electric
contact during thermal cycling (fig. S5A, Ni
washer) because of an expansion mismatch
with the IC, glass-ceramic, and membrane seg-
ment (fig. S5B). Our expansion-matched Ni/
glass-ceramic composite washer is applied in
a partially heat-treated condition as a sintered
Ni/glass composite rather than a fully heat-
treated Ni/glass-ceramic composite. This means
that the washer can deform under the load
applied during the heating phase of the seal-
ing cycle by virtue of viscous flow in the glassy
matrix phase in the washer and maintain in-
timate surface contact with both components.
The glassy matrix phase wets the ceramic
phases in both the tubular cell support and the
IC and produces a mechanically strong bond.
By the end of the sealing cycle, the glassy mat-
rix phase in the washer crystallizes to produce
a matched expansion Ni/glass-ceramic compo-
site bridge, which retains excellent electrical
continuity between the cell and the IC through-
out subsequent thermal cycling. The adopted
IC material exhibits conductivities >2500 S/cm
at 750°C and thermal expansion coefficients
in a close-to-perfect match with the thermal
expansion of the membrane support (Fig. 2D
and fig. S5B), ensuring efficient current dis-
tribution throughout the stack and mechani-
cal robustness. The IC is chemically stable
under reducing and CO2-rich atmospheres,
but can also be fitted with more oxidation-
resistant metallic components such as Ag
for operation under oxidizing conditions
(fig. S5, C to E). The absence of Cr furthermore
eliminates degradation issues related to for-
mation of resistive Cr2O3 scales or evaporation
of volatile Cr species during long-term opera-
tion at high temperatures.
During operation, individual cells will be
net endothermic or exothermic depending
on the degree of reaction and H2 separation
and compression throughout the stack length
(Figs. 1D and 2B), necessitating internal heat
exchange. To guide the optimal design of the
stack, we adopted a three-dimensional multi-
physics model integrating coupled gas flows,
heat transfer, current distribution, and reac-
tion kinetics for SMR+WGS and ADH that
captures the behavior of the stack from single
cell to stack level (13). Our stack is designed with
a U-bend type of gas flow pattern achieved
by a manifold that distributes the incoming
gas to three of the six gas channels in the
stack while combining the three corresponding
exhaust streams. For the axial type design, the
fast kinetics of SMR and ADH concentrates
the heat consumed by the endothermic reac-
tions to the initial segments of the stack,
whereas the heat caused by compression pri-
marily evolves in the latter segments. This leads
to temperature increase along the reactor
length (Figs. 1A and 3, B and C) which in turn
increases the cell Nernst voltage and com-
pression work (Fig. 3D). Our U-bend design,
however, mitigates this mismatch by spatially
balancing the heat production from compres-
sion with the heat consumption of the reac-
tions enabling a more uniform temperature
profile (Fig. 3A and fig. S6, A and B). This in
turn lowers cell Nernst voltages and thus the
required compression work (Fig. 3F and fig.
S6C). Coupled with the high performing IC,
the U-bend design allows currents (i.e., hydro-
gen fluxes) to self-regulate according to the
local Nernst voltage (fig. S4). For anhydrous
ADH, SMR+WGS and biogas this is particu-
larly evident in the initial segment where the
reaction is concentrated because of fast kine-
tics (figs. S7 to S9). The slower reaction kinetics
of aqueous ADH on the other hand distributes
the reaction over a larger portion of the stack
(fig. S10).
To experimentally demonstrate integration
of reactions beyond SMR+WGS in our PCER
stack (9), single cells were operated with NH3
in both anhydrous and aqueous form (13).
The cells achieve >97% conversion of NH3
even at open-circuit conditions (Fig. 4A) and
near 100% conversions at high H2 recoveries
thus leaving an effluent stream virtually free of
residual NH3. The cells demonstrate compa-
rable performance with anhydrous and aque-
ous NH3, CH4 , and biogas, retaining near
faradaic behavior to above 0.7 A/cm2 (fig. S11),

Fig. 1. H2 separation and compression using

PCERs. (A) Schematic of a proton ceramic
electrochemical continuous-flow reactor illustrating
H2 separation from an H2 + N2 mixture. The local
H2 compression ratio (top axis) increases as pIH2
decreases upon H2 extraction along the reactor length
(assuming a constant pIIH2 ), leading to a corresponding
temperature increase. (B) Compression work for
isothermal and non-isothermal H2 separation from
an N2 + H2 mixture. The local compression ratio
for H2 and the associated entropy difference
 
DSðxÞ ¼ Rln pIIH2 =pIH2 ðxÞ increase along the reactor
coordinate, which leads to an increase in compression
work and heat expelled from the compression process
[wel(x) = Q(x) = TDS(x)]. If left unbalanced, this heat
increases the temperature throughout the reactor,
particularly in the latter parts, resulting in higher
compression work than would be ideal for isothermal
operation. (C) Energy balance and correlated voltages for
thermally balanced operating modes, which include reaction (ER = DHR and UR = DHR/nF), charge transport (EC = iRnF and Uohmic = iR), and compression [Eco = UNernstnF and
 
UNernst = RT/nF ln pIIH2 =pIH2 ] for SMR+WGS at 750°C with an H2 compression ratio of 7.4 and NH3 cracking at 650°C with an H2 compression ratio of 5.4. See also fig. S2
for a decompression mode of operation. (D) Compression ratio as a function of UOhmic + UNernst at 750°C for a maximum pH2 from 8 to 141 bars measured using representative
PCER single cells (fig. S3) at i = 50 mA/cm2, illustrating the different operation modes. The compression ratio range was covered by adjusting the minimum pH2 as well as
gas flows to ensure a low degree of H2 extraction/dilution. The blue region consumes and the red region evolves heat.

SCIENCE science.org 22 APRIL 2022 ¥ VOL 376 ISSUE 6591 391

RES EARCH | R E P O R T S

reflecting the catalytic versatility of the porous

Ni-BZCY support. With CH4, the single cells
even achieved >90% faradaic efficiency up to
7.4 A/cm2 (corresponding to a H2 flux of 47
normal milliliters-per-minute square centi-
meter) (fig. S12) thus doubling the H2 produc-
tion capacity to-date with these materials (9).
The 36-cell PCER stack achieves nearly full
CH4 conversion and high H2 recoveries (>99%;
Fig. 4 and figs. S13 and S14) for CH4 and biogas,
enabling complete equilibrium shift and a CO2
rich effluent stream for facile carbon capture.
The series and parallel design of the stack fa-
cilitates an effective aggregated current of up
to 400 A (i > 0.73 A/cm2) with a H2 production
rate up to 0.34 kg/day from CH4, 0.31 kg/day
from biogas, and 0.34 kg/day from simulated
fully decomposed NH3 streams (Fig. 4D). We
furthermore demonstrate H2 compression to
31 bar with a purity of 99.995% (Fig. 4E) facili-
tated for additional compression and use. The
PCER stack shows promising stability, retaining
a H2 production rate of 2 normal liters per
minute after 1400 hours of operation (fig. S14B).
Fig. 2. PCER stack for electrochemical H2 production. (A) PCER stack (dimensions: height 43 cm, Both the H2 production rate [0.34 versus
diameter 4 cm). (B) Schematic of microthermal heat integration with outward heat flux from the cells. 0.025 kg/day (9)] and active area [584 versus
(C) Schematic of U-bend type of gas flow of the generic molecule AHx reacting to form H2, which is 81 cm2 (14)] greatly surpass those of any re-
electrochemically extracted as H+ through the membrane and recovered as compressed H2 in the outer ported for proton ceramic applications. More-
chamber. (D) Thermal expansion upon cooling of the BZCY/Ni support and the IC, and IC electrical over, the PCER stacks have demonstrated that
conductivity as a function of temperature. (E) Scanning electron micrograph cross section of the interface it is possible to deliver high-pressure H2 at high
between the BZCY/Ni support and IC, connected by a conductive glass-ceramic washer. (F) Schematics purity and a CO2 rich effluent at a hydrogen
of IC and washer assembly. recovery and methane conversion >99%, which

Fig. 3. Multiphysics simulations of PCER stack

thermally balanced operation. Multiphysics
simulations for a stack operating at 750°C external
temperature with 20 bars of total pressure on
both sides of the membrane and a mean
current density of 0.60 A/cm2. Feed: 28.6%
CH4 (0.597 NL/min), 71.4% H2O. Sweep: H2O
(0.18 g/min). (A) Simulated temperature fields in
a U-bend PCER stack architecture, also showing
the gas inlet and outlet flow distribution. (B) Simulated
temperature fields in an axial PCER stack architecture.
(C) Temperature profiles on the reforming side in
the axial and U-bend architecture along the reactor
length. The thermal balancing by heat transfer
between first cell (net endothermic) and last cell
(net exothermic) for U-bend PCER is illustrated by
vertical arrows. (D) Mean compression work and
Nernst voltage for each segment along the reactor
length for the U-bend and axial PCERs. The values
were obtained by integrating the compression work
over each segment divided by the corresponding total
flux or current. (E) Effect of H2 recovery on the
temperature distribution in U-bend and axial PCERs.
(F) Effect of H2 recovery on the mean stack
compression work and Nernst voltage for U-bend
and axial PCERs.

392 22 APRIL 2022 ¥ VOL 376 ISSUE 6591 science.org SCIENCE

 RES EARCH | R E P O R T S

is highly competitive with Pd-based membrane
reformers (table S2). These key performance
indicators build the foundation for highly
energy-efficient hydrogen production at sys-
tem level.
System modeling (13) of a 1 ton/day dis-
tributed H2 production plant adopting our
PCER stack (figs. S15 to S21) reveals that ef-
ficiencies of 91% for CH4 and as high as 95%
for anhydrous NH3 can be achieved by virtue
of microthermal integration and downstream
heat recovery. Furthermore, the PCER delivers
a concentrated and pressurized stream of CO2
when operated on methane or biogas (Fig. 4F)
that can be purified and liquefied by cryogenic
distillation, eliminating the need for complex
downstream absorption-based CO2 capture.
The high degree of process intensification
achieved by our PCER stacks enables a fuel-
flexible energy-efficient alternative to established
technologies for distributed H2 production.
Using a California 2020 electric grid carbon
intensity scenario (82.92 gCO2/MJelec; see table
S3 for references), H2 production with PCERs
using CH4 as fuel would operate at lower em-
issions (75.7 vs. 124.1 gCO2/MJH2) than water
electrolysis powered by grid electricity, even
without CO2 sequestration. With decarboni-
zation of the electric grid, CO2 sequestra-
tion is required for methane reforming to
remain competitive with water electrolysis. In
a California 2050 grid scenario, PCERs can pro-
duce H2 from CH4 with lower CO2-emissions
than water electrolysis (18.7 vs. 26.2 gCO2/MJH2)
when CO2 is sequestrated. PCERs operated
on biogas even offer H2 production with net-
negative carbon emission, as CH4 from a bio-
genic process is considered carbon-neutral.
Calculated scenarios have used the CA GREET
model (15) which includes fugitive methane
emissions from natural gas production that
can be important (16).
To illustrate the practical implications of the
PCER technology, comparable well-to-wheel
emissions for battery electric vehicles (BEVs),
internal combustion engines (ICEs) with diesel
fuel, and H2 fuel-cell electric vehicles (FCEVs)
are provided in figs. S22 and S23 with sensi-
tivity to electric grid carbon intensity. In the
California 2050 scenario, the emissions of
FCEVs (14.6 gCO2/km) using H 2 produced
from CH4 with PCERs including CO2 seques-
tration are 90% lower than those of ICE with
diesel fuel (145.4 gCO2/km) and 26% lower
than FCEVs using H2 from grid-powered water
electrolysis (19.8 gCO2/km). NH3-based H2 can
offer reduced emissions compared with on-
site electrolysis for a wide range of electric grid
carbon intensities, making FCEVs fueled with
NH3-based H2 directly comparable to BEVs in
terms of CO2 emissions (6.3 gCO2/km, a reduc-
tion of 21% compared with BEV in the California
2050 scenario). Here, NH3 is assumed pro-
duced at off-site locations with favorable re-
newable energy resources and transported as
a liquid to the fueling station where efficient
ADH and separation to H2 takes place using
the PCER technology.
The growth of a new energy technology can
be limited by access to raw materials. A de-
tailed examination of raw materials’ usage of
the PCER stack (fig. S24) shows it is composed
of nonprecious, earth-abundant materials, sug-
gesting no material availability setbacks for
scaling.
REFERENCES AND NOTES
1. R. F. Service, Science 345, 610–610 (2014).
2. I. Staffell et al., Energy Environ. Sci. 12, 463–491 (2019).
3. S. T. Wismann et al., Science 364, 756–759 (2019).
4. G. Grashoff, C. Pilkington, C. Corti, Platin. Met. Rev. 27,
157–169 (1983).
5. J. M. Serra, Nat. Energy 4, 178–179 (2019).
6. S. Choi, T. C. Davenport, S. M. Haile, Energy Environ. Sci. 12,
206–215 (2019).
7. C. Duan et al., Nature 557, 217–222 (2018).
8. C. Duan et al., Science 349, 1321–1326 (2015).
9. H. Malerød-Fjeld et al., Nat. Energy 2, 923–931 (2017).
10. H. An et al., Nat. Energy 3, 870–875 (2018).
11. J. L. Barton, Science 368, 1181–1182 (2020).
12. B. L. Kee et al., Membranes (Basel) 9, 77 (2019).
13. Materials and methods are available as supplementary
materials.
14. R. J. Braun et al., ECS Trans. 91, 997–1008 (2019).
15. California Air Resources Board, “CA-GREET3.0 lookup table
pathways: Technical support documentation” (California Air
Resources Board, 2018); https://ww2.arb.ca.gov/sites/
default/files/classic//fuels/lcfs/ca-greet/lut-doc.pdf.
16. R. W. Howarth, M. Z. Jacobson, Energy Sci. Eng. 9, 1676–1687
(2021).

AC KNOWLED GME NTS

Fig. 4. PCER single-cell and stack performance. (A) NH3 conversion as a function of H2 recovery,
measured on a representative single cell (fig. S3) at 650°C and 10 bars (pNH3 = 7.25 bars; pH2O = 2.75 bars),
and aqueous NH3 at 750°C and 10 bars (pNH3 = 3.1 bars; pH2O = 5.8 bars; and pinert = 1.1 bars). Purple
and green lines show the equilibrium conversion for NH3 and aqueous NH3, respectively. (B) and (C) CH4
conversion and yield of CO2 versus H2 recovery, respectively, of PCER stack at 750°C. (D) H2 production
rate as a function of applied current density for the stack with N2/H2 mixture simulating complete NH3
decomposition (750°C, 10 bars), methane (800°C, 15 bars, S/C = 2.5), and biogas (750°C, 20 bars,
S/C = 2.5). Effective current is calculated as current density × PCER stack area (36 × 15 cm2) and applied
current as effective current/6 due to the series and parallel electric architecture. (E and F) H2 purity
(dry basis) versus H2 delivery pressure and differential pressure across the membranes (E) and CO2 purity
versus H2 recovery (F) for SMR+WGS in the stack at 750°C. Reforming side pressure = 25 bars, H2 side
pressure = 25 to 31 bars. Current density = 0.69 A/cm2.
Funding: This work was supported by Norway’s Ministry of
Petroleum and Energy through the Gassnova project CLIMIT grant
618191 in partnership with Engie SA, Equinor, ExxonMobil, Saudi
Aramco, Shell, and TotalEnergies and the Research Council of
Norway NANO2021 project DynaPro grant 296548. Author
contributions: Conceptualization: D.C., H.M.-F., T.P., P.K.V., T.S.B.,
J.M.S., C.K.; Investigation: D.C., H.M.-F., M.B., I.Y.-T., D.B., K.N.,
L.A., T.P., D.K.P., M.I.V., S.R.-B., T.S.B., C.K.; Methodology:
D.C., H.M.-F., M.B., I.Y.-T., S.A., L.A., T.P., P.K.V., D.K.P., S.R-.B.,
C.K.; Resources: M.B., D.B., K.N., M.I.V.; Software: I.Y.-T., S.A.;
Supervision: T.N., T.P., J.M.S., C.K.; Writing – original draft: D.C.,
H.M.-F., M.B., I.Y.-T., S.A., P.K.V., D.K.P., T.N., T.S.B., J.M.S.,
C.K.; Writing – review and editing: D.C., H.M.-F., M.B., I.Y.-T., T.S.B.,
D.B., S.A., L.A., T.P., P.K.V., D.K.P., S.R.-B., T.N., T.S.B., J.M.S.,
C.K. Competing interests: D.C., H.M.-F., M.B., I.Y.-T., D.B., S.A.,
K.N., D.K.P., T.S.B., and C.K. are employed by CoorsTek Membrane
Sciences (CTMS). CTMS has filed relevant patent application
PCT/EP2017/076340. T.N. is a member of the CTMS board.
I.Y.-T.’s doctoral studies at the University of Oslo (UiO) are partially
funded by CTMS. The remaining authors declare no competing
interests. Data and materials availability: All data are available in
the main text or the supplementary materials. Experimental data
are available online at http://hdl.handle.net/10251/181917.
SUPPLEMENTARY MATERIALS
science.org/doi/10.1126/science.abj3951
Materials and Methods
Figs. S1 to S30
Tables S1 to S7
References (17Ð50)
16 June 2021; resubmitted 28 October 2021
Accepted 10 February 2022
10.1126/science.abj3951

SCIENCE science.org 22 APRIL 2022 • VOL 376 ISSUE 6591 393

Copyright 2022 American Association for the Advancement of Science. All rights reserved.