DOI: 10.1126/science.

1257443
, 1326 (2014); 345 Science
et al. Benjamin Rausch
redox mediator in water splitting
Decoupled catalytic hydrogen evolution from a molecular metal oxide
This copy is for your personal, non-commercial use only.
clicking here. colleagues, clients, or customers by
, you can order high-quality copies for your If you wish to distribute this article to others

here. following the guidelines

can be obtained by Permission to republish or repurpose articles or portions of articles

): September 11, 2014 www.sciencemag.org (this information is current as of

The following resources related to this article are available online at
http://www.sciencemag.org/content/345/6202/1326.full.html
version of this article at:
including high-resolution figures, can be found in the online Updated information and services,
http://www.sciencemag.org/content/suppl/2014/09/11/345.6202.1326.DC1.html
can be found at: Supporting Online Material
http://www.sciencemag.org/content/345/6202/1326.full.html#ref-list-1
, 7 of which can be accessed free: cites 38 articles This article
http://www.sciencemag.org/cgi/collection/chemistry
Chemistry
subject collections: This article appears in the following
registered trademark of AAAS.
is a Science 2014 by the American Association for the Advancement of Science; all rights reserved. The title
Copyright American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the Science

o
n

S
e
p
t
e
m
b
e
r

1
2
,

2
0
1
4
w
w
w
.
s
c
i
e
n
c
e
m
a
g
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


o
n

S
e
p
t
e
m
b
e
r

1
2
,

2
0
1
4
w
w
w
.
s
c
i
e
n
c
e
m
a
g
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


o
n

S
e
p
t
e
m
b
e
r

1
2
,

2
0
1
4
w
w
w
.
s
c
i
e
n
c
e
m
a
g
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


o
n

S
e
p
t
e
m
b
e
r

1
2
,

2
0
1
4
w
w
w
.
s
c
i
e
n
c
e
m
a
g
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


o
n

S
e
p
t
e
m
b
e
r

1
2
,

2
0
1
4
w
w
w
.
s
c
i
e
n
c
e
m
a
g
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


o
n

S
e
p
t
e
m
b
e
r

1
2
,

2
0
1
4
w
w
w
.
s
c
i
e
n
c
e
m
a
g
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

(20). In a hollow lattice, the nodes are constrained
only by the shell walls, which has a detrimental
effect on strength and stiffness because load trans-
fer at the nodes occurs via shell wall bending. This,
together with the sharp angles between the tubes,
leads to an uneven distribution of stress and
induces large stress concentrations in the vicin-
ity of the nodes (Fig. 1, B and E). Bending of the
tubes also causes large deflections and addi-
tional ovalization at the nodes, which further
increases the compliance and stress concentra-
tions. In situ experiments and postcompression
analysis revealed that most of the deformation
is localized to the nodes (Figs. 2, Dand E, and 3J),
which implies that improving nodal strength is a
critical factor in enhancing the scaling of strength
and stiffness with density.
We demonstrated the creation of ultralight
hollow ceramic nanolattices that absorb energy,
recover after significant compression, and reach
an untapped strength and stiffness material prop-
erty space. This is achieved using high-strength
ALDalumina engineered into a thin-walled nano-
lattice that is capable of deforming elastically via
shell buckling. The ultralight ceramic nanolattices
represent the concept of materials by design, where
it is possible totransforma stronganddense brittle
ceramic into a strong, ultralight, energy-absorbing,
and recoverable metamaterial. These results
serve to emphasize the critical connectionbetween
material microstructure, hierarchical architecture,
and mechanical properties at relevant length
scales.
REFERENCES AND NOTES
1. D. Jang, L. R. Meza, F. Greer, J. R. Greer, Nat. Mater. 12,
893898 (2013).
2. J. R. Greer, J. T. M. De Hosson, Prog. Mater. Sci. 56, 654724
(2011).
3. X. W. Gu, Z. Wu, Y.-W. Zhang, D. J. Srolovitz, J. R. Greer, Nano
Lett. 13, 57035709 (2013).
4. D. Z. Chen et al., Nano Lett. 13, 44624468 (2013).
5. J. Rys et al., Adv. Eng. Mater. 16, 889896 (2014).
6. L. R. Meza, J. R. Greer, J. Mater. Sci. 49, 24962508 (2014).
7. X. Zheng et al., Science 344, 13731377 (2014).
8. J. Bauer, S. Hengsbach, I. Tesari, R. Schwaiger, O. Kraft,
Proc. Natl. Acad. Sci. U.S.A. 111, 24532458 (2014).
9. A. G. Evans, J. Am. Ceram. Soc. 73, 187206 (1990).
10. I.-W. Chen, L. A. Xue, J. Am. Ceram. Soc. 73, 25852609
(1990).
11. P. F. Becher, J. Am. Ceram. Soc. 74, 255269 (1991).
12. D. C. Hofmann et al., Nature 451, 10851089 (2008).
13. E. Munch et al., Science 322, 15161520 (2008).
14. J. C. Weaver et al., J. Struct. Biol. 158, 93106 (2007).
15. C. E. Hamm et al., Nature 421, 841843 (2003).
16. M. A. Meyers, J. McKittrick, P.-Y. Chen, Science 339, 773779
(2013).
17. L. J. Gibson, M. F. Ashby, Cellular Solids: Structure and
Properties (Cambridge Univ. Press, Cambridge, ed. 2, 1999).
18. V. S. Deshpande, M. F. Ashby, N. A. Fleck, Acta Mater. 49,
10351040 (2001).
19. S. Pellegrino, C. R. Calladine, Int. J. Solids Struct. 22, 409428
(1986).
20. V. S. Deshpande, N. A. Fleck, M. F. Ashby, J. Mech. Phys. Solids
49, 17471769 (2001).
21. T. A. Schaedler et al., Science 334, 962965 (2011).
22. L. C. Montemayor, L. R. Meza, J. R. Greer, Adv. Eng. Mater. 16,
184189 (2014).
23. See the supplementary materials.
24. M. D. Groner, F. H. Fabreguette, J. W. Elam, S. M. George,
Chem. Mater. 16, 639645 (2004).
25. Y. K. Akimov, Instrum. Exp. Tech. 46, 287299 (2003).
26. L. Valdevit, A. J. Jacobsen, J. R. Greer, W. B. Carter, J. Am.
Ceram. Soc. 94, s15s34 (2011).
27. M. Berdova et al., Acta Mater. 66, 370377 (2014).
28. H. G. Allen, P. S. Bulson, Background to Buckling (McGraw-Hill,
Berkshire, UK, 1980).
29. G. Ju, S. Kyriakides, Int. J. Solids Struct. 29, 11431171
(1992).
30. A. Torrents, T. A. Schaedler, A. J. Jacobsen, W. B. Carter,
L. Valdevit, Acta Mater. 60, 35113523 (2012).
31. L. Valdevit, S. W. Godfrey, T. Schaedler, A. J. Jacobsen,
W. B. Carter, J. Mater. Res. 28, 24612473 (2013).
32. K. J. Maloney et al., APL Mater. 1, 022106 (2013).
ACKNOWLEDGMENTS
The authors gratefully acknowledge the financial support from the
Defense Advanced Research Projects Agency under the Materials
with Controlled Microstructure and Architecture program managed
by J. Goldwasser (contract no. W91CRB-10-0305) and to the Institute
for Collaborative Biotechnologies through grant W911NF-09-0001
from the U.S. Army Research Office. The content of the information
does not necessarily reflect the position or the policy of the
government, and no official endorsement should be inferred. The
authors are grateful to the Kavli Nanoscience Institute at Caltech
for the availability of critical cleanroom facilities, and to R. Liontas
and C. Garland for TEM assistance. Part of this work was carried
out in the Lewis Group facilities at Caltech.
SUPPLEMENTARY MATERIALS
www.sciencemag.org/content/345/6202/1322/suppl/DC1
Materials and Methods
Figs. S1 to S6
Table S1
Reference (33)
Movies S1 to S3
12 May 2014; accepted 11 August 2014
10.1126/science.1255908
WATER SPLITTING
Decoupled catalytic hydrogen evolution
from a molecular metal oxide redox
mediator in water splitting
Benjamin Rausch, Mark D. Symes, Greig Chisholm, Leroy Cronin*
The electrolysis of water using renewable energy inputs is being actively pursued as a
route to sustainable hydrogen production. Here we introduce a recyclable redox mediator
(silicotungstic acid) that enables the coupling of low-pressure production of oxygen
via water oxidation to a separate, catalytic hydrogen production step outside the
electrolyzer that requires no post-electrolysis energy input. This approach sidesteps the
production of high-pressure gases inside the electrolytic cell (a major cause of membrane
degradation) and essentially eliminates the hazardous issue of product gas crossover at
the low current densities that characterize renewables-driven water-splitting devices. We
demonstrated that a platinum-catalyzed system can produce pure hydrogen over 30
times faster than state-of-the-art proton exchange membrane electrolyzers at equivalent
platinum loading.
H
ydrogen is vital for the production of com-
modity chemicals such as ammonia and
has great potential as a clean-burning fuel
(1, 2). However, currently around 95%
(~15 trillion mol year
1
) of the worlds sup-
ply of H
2
is obtained by reforming fossil fuels (3),
a process that is both unsustainable and leads to
a net increase in atmospheric CO
2
levels. Of the
alternative methods for H
2
production that are
not linked to fossil resources, the electrolysis of
water stands out as a mature, scalable technol-
ogy for which the only required inputs are water
and energy (in the formof electricity) (4). Hence,
if the energy source is renewable, H
2
can be
produced sustainably from water using electrol-
ysis (5, 6).
Renewable energy inputs tend to be sporadic
and fluctuating, and thus the systems that are
developed to harness this energy and convert it
to H
2
[such as proton exchange membrane elec-
trolyzers (PEMEs) (7), solar-to-fuels systems (8),
and artificial leaves (9)] must be able to deal with
varying energy inputs effectively and have rapid
startup times. At the low power loads that are
characteristic of renewable power sources, the
rate at which H
2
and O
2
are produced may in
fact be slower than the rate at which these gases
permeate the membrane (10). At the very least,
this will severely affect the amount of hydrogen
that can be harvested fromsuch devices (11), and
in extreme cases could give rise to hazardous
O
2
/H
2
mixtures. The PEME is the most mature
technology citedfor renewables-to-hydrogencon-
version, but prevention of such gas crossover at
lowcurrent densities remains a challenge. PEMEs
use nontrivial amounts of precious metal catalysts
and so tend to operate at high current densities
(1 A cm
2
or above) and high pressure, where the
cost of their components can be offset to some
extent (7). However, these optimal conditions may
be hard to maintain in all cases of renewables-
driven electrolysis (for example, in small-scale
facilities), where less-expensive and lower-power
devices would therefore be beneficial. Meanwhile,
the high-pressure and high-current-density condi-
tions under which PEMEs work most effectively
are also not without drawbacks: These condi-
tions can also lead to gas crossover through the
membrane, and the coexistence of H
2
, O
2
and
catalyst particles produces reactive oxygen species
1326 12 SEPTEMBER 2014 VOL 345 ISSUE 6202 sciencemag.org SCIENCE
RESEARCH | REPORTS
WestCHEM, School of Chemistry, University of Glasgow,
University Avenue, Glasgow G12 8QQ, UK.
*Corresponding author. E-mail: [email protected]
(ROS) that degrade the membrane and shortenits
lifetime (12, 13). There is thus a need to develop
newelectrolyzer systems that canprevent product
gases from mixing over a range of current den-
sities and that make more effective use of the
precious metal catalysts they employ, in order to
make renewables-to-hydrogenconversionbothprac-
tically and economically more attractive.
Previously, we introduced the concept of the
electron-coupled proton buffer (ECPB), which
can act to decouple electrolytic H
2
and O
2
pro-
duction, producing these gases at separate times
(14, 15). Here we describe a redox mediator that
can be reversibly reduced in an electrolytic cell
(as water is oxidized at the anode) and then
transferred to a separate chamber for sponta-
neous catalytic H
2
evolution, without the need
for additional energy input after reduction of the
mediator (Fig. 1). This approach leads to a device
architecture for electrolyzers that has several im-
portant advantages. First, it allows the electro-
chemical step to be performed at atmospheric
pressure, while potentially permitting H
2
to be
evolved at elevated pressure in a distinct com-
partment. Second, virtually no H
2
is produced in
the electrolytic cell itself, which (taken with the
feature above) obviates the need to purge H
2
from the anode side of the cell and could signif-
icantly reduce ROS-mediated membrane degra-
dationandthe possibility of explosive gas mixtures
forming at low current densities or upon mem-
brane failure. Third, H
2
evolution from such a
systemis no longer directly coupled to the rate of
water oxidation, and thus the decoupled H
2
pro-
duction step can be performed a rate per milli-
gram of catalyst that is over 30 times faster than
that for state-of-the-art PEMEs (movie S1). Finally,
the hydrogen produced has the potential to have
an inherently low O
2
content, both on account of
its production in a separate chamber from water
oxidation and by virtue of the fact that the re-
duced mediator reacts rapidly with O
2
in solution.
This final point could render the H
2
produced
suitable for applications requiring high-purity H
2
such as fuel cells (16) or the Haber-Bosch process
(17), without the need for post-electrolysis purifi-
cation or built-in recombination catalysts.
The redox mediator investigated in this work
was silicotungstic acid (H
4
[SiW
12
O
40
]), the cyclic
voltammogram (CV) of which on a glassy car-
bon electrode in aqueous solution is shown in
Fig. 2A (black line). H
4
[SiW
12
O
40
] was chosen
for investigation on account of its high solubil-
ity in water (up to 0.5 M), in which solvent it is
a strong acid (18). H
4
[SiW
12
O
40
] has reversible
one-electron redox waves centered at +0.01 V
(wave I) and 0.22 V [wave II; all potentials are
versus the normal hydrogen electrode (NHE)],
the positions of which are critical to the follow-
ing discussion (18). Also shown in Fig. 2A are
reductive scans taken at a similar pH in the ab-
sence of H
4
[SiW
12
O
40
] on carbon and Pt electro-
des (red and green lines, respectively). Given
that the onset of H
2
evolution on Pt occurs at es-
sentially the same potential as the first reduc-
tion of H
4
[SiW
12
O
40
], but that H
2
evolution on
carbon is not appreciable above 0.6 V, we hy-
pothesized that the reduction of H
4
[SiW
12
O
40
]
at a carbon electrode at potentials slightly more
positive than 0.6 V would give the two-electron
reduced form (H
6
[SiW
12
O
40
]) without any com-
peting H
2
evolution. If H
6
[SiW
12
O
40
] were then
exposed to Pt, it should spontaneously evolve H
2
until equilibrium between H
2
and the reduced
mediator was reached, whichFig. 2Asuggests will
correspond to a mixture of H
4
[SiW
12
O
40
] and the
one-electron reduced form, H
5
[SiW
12
O
40
].
To test this hypothesis, an airtight electrolysis
cell was constructed with a Pt mesh or carbon
felt anode (for water oxidation) and a carbon felt
cathode (for H
4
[SiW
12
O
40
] reduction), as shown
in fig. S1. Reduction of the mediator and conco-
mitant water oxidation were performed, and the
composition of the gases in the separated head-
spaces was monitored by gas chromatographic
headspace analysis (GCHA). Full faradaic effi-
ciency for O
2
evolution could be observed (using
SCIENCE sciencemag.org 12 SEPTEMBER 2014 VOL 345 ISSUE 6202 1327
RESEARCH | REPORTS
Fig. 1. A schematic of silicotungstic acidmediated H
2
evolution from water. At the anode, H
2
O
is split into O
2
, protons, and electrons, while the mediator is reversibly reduced and protonated at the
cathode in preference to direct production of H
2
. The reduced H
6
[SiW
12
O
40
] (dark blue) is then
transferred to a separate chamber for H
2
evolution over a suitable catalyst and without additional energy
input after two-electron reduction of the mediator to H
6
[SiW
12
O
40
].
Fig. 2. Performance of silicotungstic acid as an electrochemical mediator for water splitting.
(A) Reductive CVs under Ar and at room temperature: black, H
4
[SiW
12
O
40
] in water (0.5 M, pH 0.5), at a
glassy carbon working electrode (area = 0.071 cm
2
); red, 1 M H
3
PO
4
(pH = 1.0) on a glassy carbon
working electrode; green, 1 M H
3
PO
4
(pH = 1.0) on a Pt disc working electrode (area = 0.031 cm
2
). A Pt-
mesh counterelectrode and Ag/AgCl reference electrode were used at a scan rate of 0.1 V s
1
. (B)
Comparison of the rate of H
2
production possible using electrolysis mediated by silicotungstic acid
(this work) and a selection of state-of-the-art electrolyzers from recent years. Square symbols
indicate data obtained for a mediated system in this paper. Red data (left-hand y axis): the rate of H
2
production per milligram of Pt. Blue data (right-hand y axis): the absolute rate of H
2
production
determined for H
2
production from H
6
[SiW
12
O
40
] (this work, squares) and the various literature
electrolyzer systems. Dashed lines are provided solely as guides to the eye. Code for literature data is as
follows: pentagons, reference (30); stars, reference (26); triangles, reference (31); dots, reference (32).
This comparison does not include the time taken to reduce the mediator to H
6
[SiW
12
O
40
]. For more
details on how these metrics were obtained, see table S1.
Pt anodes), whereas complete two-electron re-
duction of the mediator could be achieved with
only trace H
2
being evolved [see supplemen-
tary materials (SM) section SI-4 and figs. S2 and
S3]. This two-electron reduced H
6
[SiW
12
O
40
]
could then be stored without significant spon-
taneous H
2
evolution (<0.002%loss of H
2
per hour;
fig. S3). Taken together, these data suggest that
O
2
evolution and H
2
evolution can be effectively
decoupled from each other using H
4
[SiW
12
O
40
],
potentially allowing the O
2
produced during
electrolysis to be vented to the atmosphere with-
out the need for additional H
2
removal pro-
cesses (19).
The two-electron reduced mediator could then
be removed from the electrolysis cell and intro-
duced into sealed reaction flasks under an atmo-
sphere of Ar. The addition of various metal foils
to this solution catalyzed H
2
evolution, with Pt
exhibiting the best performance (SM section SI-5
and fig. S4A). Powdered samples of MoS
2
(20, 21)
and Ni
2
P (22) were also found to be effective cat-
alysts for H
2
evolution from H
6
[SiW
12
O
40
] (fig.
S4B). However, by far the greatest rate of H
2
evolution was found when using precious metal
catalysts supported on carbon. Figure 2B shows
that per milligram of Pt used, the rate of H
2
production from H
6
[SiW
12
O
40
] exceeds the rate
of H
2
evolution possible using a state-of-the-art
PEME by a factor of 30 (red data). This more
effective use of the precious metal H
2
evolution
catalyst could be a result of the better disper-
sion of catalyst that is possible when it is not
confined to an electrode.
The kinetics of H
2
evolution from solutions of
H
6
[SiW
12
O
40
] as a function of time and catalyst
are examined in Fig. 3A. Based on the volume of
mediator solution used in these experiments, full
conversion of two-electron reduced H
6
[SiW
12
O
40
]
to one-electron reduced H
5
[SiW
12
O
40
] would be
expected to liberate 122.4 ml of H
2
, whereas com-
plete reversion to H
4
[SiW
12
O
40
] would release
244.7 ml of H
2
(SM section SI-6). In practice,
somewhat more than 122.4 ml of H
2
were liber-
ated in under 30 min with all the catalysts ex-
amined in Fig. 3A, suggesting complete and rapid
transformation of H
6
[SiW
12
O
40
] to H
5
[SiW
12
O
40
],
followed by limited further conversion (10 to
36%) of H
5
[SiW
12
O
40
] to H
4
[SiW
12
O
40
] under these
conditions.
Initial rates were then extrapolated to rates of
H
2
produced per milligramof precious metal per
hour (Table 1 and table S2), giving a maximum
rate of 2861 mmol of H
2
mg
1
hour
1
when using
low loadings of Pt/C. The rate of H
2
evolution
decays from the initial value in Fig. 3B on ac-
count of the process H
6
[SiW
12
O
40
] H
5
[SiW
12
O
40
]
being 80% complete within 30 s for all the Pt/C
loadings shown. Hence, in a continuous flow
system, it should be possible to achieve rates very
similar to the initial rate measured here for as
long as the flowof H
6
[SiW
12
O
40
] is maintained(the
mediator could then be recycled to the cathode
for recharging). Table 1 compares the rate of H
2
production by the mediator-based system with
that achieved by a selection of state-of-the-art
PEMEs from the recent literature (a more ex-
tensive comparison can be found in table S1).
With PEMEs, the rate of H
2
production is nec-
essarily coupled to the rate of water oxidation
occurring at the anode. Ina mediated electrolysis
cell, the rates of water oxidation and mediator
reduction are coupled, but the rate of H
2
produc-
tion depends on the availability of the reduced
mediator. This allows a mediated systemto make
more effective use of the H
2
evolution catalyst,
as illustrated by Table 1. The time required to
reduce the mediator is not included in the
calculations for Table 1: Only the rate of H
2
production (and hence how long it would take
to obtain all the H
2
from the mediator for com-
pression and/or storage) is considered. Details
on mediator reduction can be found in the
1328 12 SEPTEMBER 2014 VOL 345 ISSUE 6202 sciencemag.org SCIENCE
RESEARCH | REPORTS
Fig. 3. Catalytic H
2
evolution from H
6
[SiW
12
O
40
] when in contact with various catalysts. (A) Rate
of H
2
production from a 20-ml sample of 0.5 M H
6
[SiW
12
O
40
] under an Ar atmosphere. (B) Magnification
of the first 2 min of the H
2
evolution process from H
6
[SiW
12
O
40
] in the presence of 50 mg Pt/C (5, 3,
and 1 weight %). Dashed lines indicate the derived initial rates. (C) A typical apparatus configuration for
determining volumes of H
2
evolved when H
6
[SiW
12
O
40
] was exposed to powdered catalysts. Reduced
mediator in the upper flask (blue circle in left-hand photograph) is allowed to flow into the lower chamber
containing catalyst (blue circle in middle photograph), and H
2
is rapidly evolved and collected (as
indicated by the red arrows). A movie showing rapid H
2
evolution (movie S1) and full experimental
details can be found in the supplementary materials (SM section SI-6).
Table 1. Comparison of the rate of H
2
production possible with silicotungstic acidmediated elec-
trolysis and a selection of leading PEMEs from the current literature. Literature values are based
on the highest rate of H
2
production reported in those works. A table with more examples and a full
description of how these metrics were calculated can be found in the supplementary materials.
H
2
evolution catalyst
Amount of
powder/size
of electrode
Total catalyst
used (mg)
H
2
production
rate (mmol
hour
1
mg
1
)
Reference
Pt/C (0.5 mg of Pt cm
2
) 9 100 cm
2
450 22 (33)
Pt/C (0.4 mg of Pt cm
2
) 5 cm
2
2 47 (26)
Pt/C (0.7 mg of Pt cm
2
) 7 cm
2
4.9 53 (31)
Pt/C (0.3 mg of Pt cm
2
) 20 cm
2
6 93 (32)
Pd/C 10% 50 mg 5 143 This work
Rh/C 5% 50 mg 2.5 241 This work
Pt/C 5% 50 mg 2.5 368 This work
Pt/C 3% 50 mg 1.5 423 This work
Pt/C 1% 50 mg 0.5 1275 This work
Pt/C 1% 25 mg 0.25 1336 This work
Pt/C 1% 10 mg 0.1 2861 This work
supplementary materials (SM sections SI-3
and SI-7).
Next, we examined the purity of the H
2
that
was produced by this silicotungstic acidmediated
method. GCHA indicated that the level of
electrolysis-derived O
2
in this H
2
was below de-
tectable limits (T0.08%; SM sections SI-4 and
SI-6). Moreover, if 10% O
2
were deliberately in-
troduced into the headspace of the vessel con-
taining H
6
[SiW
12
O
40
], this extraneous O
2
was
completely removed by reaction with H
6
[SiW
12
O
40
]
(% O
2
in the headspace was only 0.04% after
30 min), ultimately producing water and reoxi-
dized mediator (23) and further guaranteeing
that the H
2
evolved is O
2
-free (SM section SI-9
and fig. S9). This has obvious implications for
electrolyzer safety, because gaseous mixtures of
H
2
and O
2
on the cathode side are now pre-
cluded by the reduced mediators rapid reaction
with O
2
. This reaction is spontaneous and does
not require any precious metalbased recom-
bination catalysts such as those often employed
in PEMEs.
As noted earlier, a primary mode of degrada-
tion of the perfluorinated membranes used in
PEMEs is attack by ROS (12). These ROS form
in the presence of O
2
, H
2
, and precious metals (in-
cluding the catalytic recombination layers that are
designedtoprevent mixtures of O
2
andH
2
forming
in electrolysis product streams). Moreover, recom-
bination of H
2
and O
2
is an exothermic process
that causes local heating, damaging the membrane
through mechanical means; this route is especially
prevalent at Pt sites on the cathode (24, 25). The
use of a mediator can help to mitigate against
membrane degradation in three ways. First, the
amount of H
2
produced in the electrolyzer itself is
vastly diminished, removing the need to purify the
O
2
product stream and preventing ROS forma-
tion on the anode side of the cell. Second, on the
cathode side, the reduced mediator reacts rap-
idly with any O
2
present to produce water, and
any peroxy species that do formwill do so in bulk
solution far from the membrane and will them-
selves rapidly react with reduced mediator to
form water (23). Finally, the catalyst is now iso-
lated in a second chamber and is not in contact
with the membrane, lessening local heating
effects. Hence, using a mediator could potentially
allow increased lifetimes for the membranes
used in such electrolyzers relative to the life span
of similar membranes in PEMEs.
The efficiency of the electrochemical process
to produce O
2
fromwater and H
6
[SiW
12
O
40
] from
H
4
[SiW
12
O
40
] was calculated and compared to
equivalent systems that wouldproduce H
2
and O
2
directly by electrolysis (SM section SI-7 and fig.
S6). In comparison to a systemthat uses a carbon
cathode to reduce protons and a Pt anode to
oxidize water, the mediated systemwas 16%more
efficient, with an overall energy efficiency of 63%.
A standard electrolysis system for direct O
2
and
H
2
production from water, in which both electro-
des are Pt, was found to have an efficiency of 67%
[which agrees well with the efficiency of room-
temperature PEMEs reported in the literature
(26)]. Hence, given the potential for lower load-
ings of precious metal and high initial purity of
the product gases whenusingmediatedelectrolysis
(and other possible technoeconomic advantages;
SM section SI-12), we believe that such systems
will be competitive with PEMEs in terms of cost-
efficiency metrics.
The redox reactions of silicotungstic acid are
summarized in fig. S10A. Starting from fully re-
duced H
6
[SiW
12
O
40
], H
2
evolution in the pres-
ence of a catalyst such as Pt/C is rapid, leading
to the one-electron reduced species H
5
[SiW
12
O
40
].
This process can be reversed by electro-reducing
H
5
[SiW
12
O
40
] at a carbon cathode. Alterna-
tively, starting from the fully oxidized species
H
4
[SiW
12
O
40
], the one-electron reduced species
can be accessed either by electrochemical reduc-
tion or by reaction with H
2
in the presence of a
suitable catalyst such as Pt/C (SM section SI-8
and fig. S8). Likewise, if one-electron reduced
H
5
[SiW
12
O
40
] is placed in a sealed reaction vessel
under Ar in the presence of Pt/C, H
2
evolves
slowly into the headspace, as gauged by GCHA
(fig. S7). This behavior implies that there exists
an equilibriumbetween H
2
and H
4
[SiW
12
O
40
] on
one hand and the one-electron reduced mediator
(H
5
[SiW
12
O
40
]) on the other in the presence of
catalysts such as Pt/C.
Overall faradaic efficiencies for the roundtrip
process were gauged by fully reducing a sample
of H
4
[SiW
12
O
40
] to H
6
[SiW
12
O
40
] with coulometry.
Pt/C was then added to this H
6
[SiW
12
O
40
], and
H
2
was evolved. At the cessation of spontaneous
H
2
evolution, an amount of H
2
corresponding to
68% of the charge passed in reducing H
4
[SiW
12
O
40
]
to H
6
[SiW
12
O
40
] was obtained. In a cyclic system,
any one-electron reduced H
5
[SiW
12
O
40
] could sim-
ply be returned to the electrolyzer for re-reduction
to H
6
[SiW
12
O
40
]. However in this case, once H
2
evolution had ceased, the Pt/C catalyst was re-
moved by filtration under Ar, and the result-
ing Pt-free mediator solution was titrated with
an Fe(III) source in order to oxidize all remain-
ing H
5
[SiW
12
O
40
] to colorless H
4
[SiW
12
O
40
] and
thus ascertain the amount of H
5
[SiW
12
O
40
] still
present at the cessation of H
2
evolution (SM
section SI-10). This value, when combined with
the electrons already accounted for by the amount
of H
2
evolved, gave a faradaic yield in excess
of 98% for the roundtrip H
4
[SiW
12
O
40
]
H
6
[SiW
12
O
40
] H
4
[SiW
12
O
40
].
The stability of the mediator to several cycles
of oxidation and reduction was probed both elec-
trochemically (by comparing the charges passed
in oxidizing the reducing the mediator over a
series of cycles, SM section SI-13) and by com-
paring ultraviolet-visible spectra of fresh and
cycled samples and of reduced samples that were
reoxidized by exposure to air (SM section SI-14).
Figure S11A shows that 98% of the charge passed
in fully reducing the mediator by one electron
could be retrieved by reoxidation over nine full
one-electron reduction-oxidation cycles, with no
apparent degradation of the mediator. Figure
S11B shows the stability of the mediator to four
consecutive cycles of reduction to 80% of the max-
imum for full two-electron reduction, followed
by reoxidation to 20% of this maximum. This
experiment was designed to mimic the condi-
tions under which the mediator would have to
operate in a continuous-flow system. The data
in fig. S11B suggest that there is no decay in the
amount of charge that can be stored in the me-
diator (which would signal irreversible decom-
position) within these bounds over the number
of cycles probed. Similarly, fig. S12, A and B,
show that a sample of silicotungstic acid sub-
jected to 20 consecutive two-electron reduction
and reoxidation cycles has an ultraviolet-visible
spectrum indistinguishable from that of a fresh
sample of silicotungstic acid. Taken together,
these data suggest that the mediator is stable to
redox cycling under these conditions and that
H
4
[SiW
12
O
40
] might thus be suitable as a me-
diator in a continuous-flow system.
Because of the high molecular weight of
H
4
[SiW
12
O
40
], it does not constitute an especially
effective static storage medium for H
2
(or H
2
equivalents). Clearly, lowermolecular-weight me-
diators, or systems capable of storing more elec-
trons, would therefore be at a practical advantage
(15), allowing greater buffering capacity to be built
into the systemand providing more flexibility with
regard to the temporal separation of water oxida-
tion and H
2
generation. We are currently pursuing
the identification of such mediators, and we see
great potential for optimized mediator systems to
be combined with other recent breakthroughs in
catalysis (27, 28) and device design (9, 29), facil-
itating the use of low-power inputs (or those sub-
ject to large fluctuations) inrenewables-to-hydrogen
conversion.
REFERENCES AND NOTES
1. A. Le Goff et al., Science 326, 13841387 (2009).
2. N. Armaroli, V. Balzani, ChemSusChem 4, 2136 (2011).
3. U.S. Department of Energy Hydrogen Analysis Resource
Center, Hydrogen Production, Worldwide Refinery Hydrogen
Production Capacities (2012), http://hydrogen.pnl.gov/
cocoon/morf/hydrogen/article/706.
4. J. D. Holladay, J. Hu, D. L. King, Y. Wang, Catal. Today 139,
244260 (2009).
5. F. M. Toma et al., Nat. Chem. 2, 826831 (2010).
6. D. Gust, T. A. Moore, A. L. Moore, Acc. Chem. Res. 42,
18901898 (2009).
7. M. Carmo, D. L. Fritz, J. Mergel, D. Stolten, Int. J. Hydrogen
Energy 38, 49014934 (2013).
8. Y. Tachibana, L. Vayssieres, J. R. Durrant, Nat. Photonics 6,
511518 (2012).
9. S. Y. Reece et al., Science 334, 645648 (2011).
10. F. Barbir, Sol. Energy 78, 661669 (2005).
11. A. Berger, R. A. Segalman, J. Newman, Environ. Sci. 7,
14681476 (2014).
12. L. Ghassemzadeh, K.-D. Kreuer, J. Maier, K. Mller, J. Phys.
Chem. C 114, 1463514645 (2010).
13. V. Prabhakaran, C. G. Arges, V. Ramani, Proc. Natl. Acad. Sci.
U.S.A. 109, 10291034 (2012).
14. M. D. Symes, L. Cronin, Nat. Chem. 5, 403409 (2013).
15. B. Rausch, M. D. Symes, L. Cronin, J. Am. Chem. Soc. 135,
1365613659 (2013).
16. International Organization for Standardization, ISO 14687-
2:2012, Hydrogen fuel - Product specification - Part 2: Proton
exchange membrane (PEM) fuel cell applications
for road vehicles, www.iso.org/iso/home/store/catalogue_ics/
catalogue_detail_ics.htm?csnumber=55083 (2012).
17. N. Wiberg, E. Wiberg, A. F. Holleman, in Inorganic Chemistry
(Academic Press, San Diego, CA, 2001), vol. 1.
18. B. Keita, L. Nadjo, J. Electroanal. Chem. 217, 287304
(1987).
19. S. A. Grigoriev et al., Int. J. Hydrogen Energy 34, 59865991
(2009).
20. H. I. Karunadasa et al., Science 335, 698702 (2012).
SCIENCE sciencemag.org 12 SEPTEMBER 2014 VOL 345 ISSUE 6202 1329
RESEARCH | REPORTS
21. D. Merki, S. Fierro, H. Vrubel, X. Hu, Chem. Sci. 2, 12621267
(2011).
22. E. J. Popczun et al., J. Am. Chem. Soc. 135, 92679270 (2013).
23. A. Hiskia, E. Papaconstantinou, Inorg. Chem. 31, 163167 (1992).
24. A. B. LaConti, H. Liu, C. Mittelsteadt, R. C. McDonald,
ECS Trans. 1, 199219 (2006).
25. A. S. Aric et al., J. Appl. Electrochem. 43, 107118 (2013).
26. N. Mamaca et al., Appl. Catal. B 111112, 376380 (2012).
27. T. W. Kim, K.-S. Choi, Science 343, 990994 (2014).
28. Z. Han, F. Qiu, R. Eisenberg, P. L. Holland, T. D. Krauss, Science
338, 13211324 (2012).
29. K. S. Joya, Y. F. Joya, K. Ocakoglu, R. van de Krol, Angew.
Chem. Int. Ed. 52, 1042610437 (2013).
30. J. Xu, G. Liu, J. Li, X. Wang, Electrochim. Acta 59, 105112
(2012).
31. P. Millet et al., Int. J. Hydrogen Energy 35, 50435052
(2010).
32. C. Xu, L. Ma, J. Li, W. Zhao, Z. Gan, Int. J. Hydrogen Energy 37,
29852992 (2012).
33. S. Siracusano et al., Int. J. Hydrogen Energy 37, 19391946
(2012).
ACKNOWLEDGMENTS
This work was supported by the Engineeering and Physical
Sciences Research Council (UK). L.C. thanks the Royal Society/
Wolfson Foundation for a Merit Award. M.D.S. thanks the University
of Glasgow for a Lord Kelvin Adam Smith Research Fellowship.
We are grateful to J. Liddell (University of Glasgow) for production
of the H cells used in this work. Supplementary materials
are available, which include full experimental details for
electrochemical procedures, GCHA, and catalytic H
2
evolution,
as well as a movie showing the H
2
evolution step (movie
S1). The advances presented in the work form part of a
patent filing.
SUPPLEMENTARY MATERIALS
www.sciencemag.org/content/345/6202/1326/suppl/DC1
Supplementary Text Sections SI-1 to SI-14
Figs. S1 to S12
Tables S1 and S2
References (3442)
Movie S1
16 June 2014; accepted 6 August 2014
10.1126/science.1257443
BLACK HOLE PHYSICS
Rapid growth of seed black holes
in the early universe by
supra-exponential accretion
Tal Alexander
1
* and Priyamvada Natarajan
2
Mass accretion by black holes (BHs) is typically capped at the Eddington rate, when
radiation's push balances gravity's pull. However, even exponential growth at the
Eddington-limited e-folding time t
E
~ few 0.01 billion years is too slow to grow
stellar-mass BH seeds into the supermassive luminous quasars that are observed when
the universe is 1 billion years old. We propose a dynamical mechanism that can trigger
supra-exponential accretion in the early universe, when a BH seed is bound in a star cluster
fed by the ubiquitous dense cold gas flows. The high gas opacity traps the accretion
radiation, while the low-mass BH's random motions suppress the formation of a slowly
draining accretion disk. Supra-exponential growth can thus explain the puzzling emergence
of supermassive BHs that power luminous quasars so soon after the Big Bang.
O
ptically bright quasars powered by accre-
tion onto black holes (BHs) are now de-
tected at redshifts as high as z ~ 7, when
the universe was 6% of its current age (<1
billion years) (1). Their luminosities imply
supermassive BHs (SMBHs) with the mass of the
BH (M

) 10
9
solar masses (M

) (2). The main

obstacles to assembling such SMBHs so rapidly
are the lowmasses of the hypothesizedinitial seed
BHs, born of first-generation (Pop III) stars, cou-
pled with the maximal growth rate for radiative-
ly efficient accretion, the Eddington limit (35).
Proposed ways to circumvent these limitations
invoke super-Eddington accretion for brief pe-
riods of time (6); the ab initio formation of more-
massive BH seeds (710) from the direct collapse
of self-gravitating pre-galactic gas disks at high
redshifts (1114); and the formation of a very mas-
sive star from runaway stellar mergers in a dense
cluster (15, 16). Discriminating between these
scenarios is challenging, because seed formation
redshifts (z > 10) are observationally inaccessible.
Current data require finely tuned, continuous
early BH growth and massive initial BH seeds
(1720). Recent results from high-resolution sim-
ulations of early star formation at z 15 to 18
exacerbate the problem by indicating that effi-
cient fragmentation and turbulence (2125) lead
to the efficient formation of stellar clusters em-
bedded in the flow, which prevents the forma-
tion of massive seeds (>>10 M

) by limiting the
mass of their potential Pop III progenitor stars.
On the other hand, theoretical and numerical
results on larger scales suggest that ubiquitous
dense cold gas flows (26) stream in along fila-
ments and feed protogalactic cores (27, 28).
Adaptive mesh refinement simulations track the
fate of these sites (collapsed 10
7
M

dark-matter
halos) from1 Mpc scale at z ~ 21 with resolutions
as low as ~ 2 10
10
pc in the central regions.
These simulations find isothermal density cusps
that reach extreme central densities, with an av-
erage density of r

10
16
g cm
3
(10
8
cm
3
for
pure H) on 0.1 pc scales (29). They also reveal a
marginally unstable central gas reservoir of
few10
5
M

in the inner fewparsecs (30), where

the dynamical time scale is ~10
6
years. Although
these simulations are somewhat idealized, we
adopt the properties of this high-density envi-
ronment as the initial conditions for the model
presented here.
We consider a scenario in which a low-mass
Pop III remnant BH remains embedded in a
nuclear star cluster fed by dense cold gas flows
(26) (Fig. 1 and Table 1). The stars and gas are
virialized in the cluster potential, and the BHis
initially a test particle in equipartition with the
stars. Gas within the accretion (capture) radius
of the BH, r
a
= [2c
2
/(c

2
+v

2
)]r
g
, is dynamically
bound to it, where r
g
= G M

/c
2
is the gravita-
tional radius of the BH; c

is the gas sound

speed in the cold flowfar fromthe BH, which is
a measure of the depth of the cluster's gravita-
tional potential; and v

is the BHvelocity relative

to the gas. Gas bound to the BH inside r
a
is not
necessarily accreted by it. Prompt accretion re-
quires gas toflowfromr
a
intothe BHonaplunging
trajectory with low specific angular momentum
j < j
ISO
4r
g
c, through the innermost stable
periapse distance r
ISO
. It is this angular momen-
tum barrier, rather than the Eddington limit,
that is the main obstacle to supra-exponential
growth.
The BH is more massive than a cluster star,
so v

2
< c

2
, and the accretion flow is quasi-
spherical. In the idealized case where the flow
is radial and adiabatic, it is described by the
Bondi solution (31),

M
B
p=

2
p
r
2
a
r

(adia-
batic index G = 4/3 assumed), which can be
written compactly in terms of m = M

/M
i
, where
M
i
is the initial BHmass, as m m
2
=t
B
, with time
scale t
B
= c

3
/ 2
3/2
pG
2
M
i
r

. The stronger-than-
linear dependence of the accretion rate on the
BH mass leads to a solution that diverges supra-
exponentially in a finite time t
B
as m(t) = 1/(1 t/t
B
).
Physical flows, where gravitational energy is re-
leased as radiation, are not strictly adiabatic. As
the mass accretion rate grows, the local luminos-
ity can far exceed the Eddington luminosity L
E
=
4pcG M

/k =

M
E
c
2
(k is the gas opacity), for
which radiation flux pressure balances gravity.
However, radiation produced inside the photon-
trapping radius r
g
~

M=

M
E
r
g
is carried with the
flow into the BH, because the local optical depth
t(r) ~ kr(r)r makes photon diffusion outward
slower than accretion inward (32) [which is a
manifestation of the Ov=c effect of relativistic
beaming (33)]. The luminosity L

that escapes
to infinity from r r
g
translates to a lowered ra-
diative efficiency h
g
= L

/

Mc
2
~ min(r
g
/r
g
,r
g
/r
ISO
),
so it does not exceed ~L
E
, thereby allowing supra-
exponential Bondi mass accretion rates (34). De-
tailed calculations show that L

0.6 L
E
(35).
1330 12 SEPTEMBER 2014 VOL 345 ISSUE 6202 sciencemag.org SCIENCE
RESEARCH | REPORTS
1
Department of Particle Physics and Astrophysics, Weizmann
Institute of Science, Rehovot 76100, Israel.
2
Department of
Astronomy, Yale University, 260 Whitney Avenue, New
Haven, CT 06511, USA.
*Corresponding author. E-mail: [email protected]