Sulfur as a selective 'soft' oxidant for catalytic methane conversion probed by experiment and theory

Article abstract of DOI:10.1038/nchem.1527

Developing efficient catalytic processes to convert methane into useful feedstocks relies critically upon devising new coupling processes that use abundant, thermodynamically 'mild' oxidants together with selective catalysts. We report here on elemental sulfur as a promising 'soft' oxidant for selective methane conversion to ethylene over MoS 2, RuS 2, TiS 2, PdS and Pd/ZrO 2 catalysts. Experiments and density functional theory reveal that methane conversion is directly correlated with surface metal-sulfur bond strengths. Surfaces with weakly bound sulfur are more basic and activate methane C-H bonds more readily. In contrast, experimental and theoretical selectivities scale inversely with surface metal-sulfur bond strengths, and surfaces with the strongest metal-sulfur bonds afford the highest ethylene selectivities. High CH 4 /S ratios, short contact times and the provision of a support maximizes the coupling of CH x intermediates and selectivity to ethylene, because these conditions yield surfaces with stronger metal-sulfur bonding (for example, Pd 16 S 7), which suppresses the over-oxidation of methane.

Full text of DOI:10.1038/nchem.1527

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
ARTICLES
PUBLISHED ONLINE: 16 DECEMBER 2012 | DOI: 10.1038/NCHEM.1527
Sulfur as a selective ‘soft’ oxidant for catalytic
methane conversion probed by experiment
and theory
Qingjun Zhu1, Staci L. Wegener1, Chao Xie1, Obioma Uche2, Matthew Neurock2 and Tobin J. Marks1
*
*
Developing efficient catalytic processes to convert methane into useful feedstocks relies critically upon devising new
coupling processes that use abundant, thermodynamically ‘mild’ oxidants together with selective catalysts. We report here
on elemental sulfur as a promising ‘soft’ oxidant for selective methane conversion to ethylene over MoS2, RuS2, TiS2, PdS
and Pd/ZrO2 catalysts. Experiments and density functional theory reveal that methane conversion is directly correlated
with surface metal–sulfur bond strengths. Surfaces with weakly bound sulfur are more basic and activate methane C–H
bonds more readily. In contrast, experimental and theoretical selectivities scale inversely with surface metal–sulfur bond
strengths, and surfaces with the strongest metal–sulfur bonds afford the highest ethylene selectivities. High CH4/S ratios,
short contact times and the provision of a support maximizes the coupling of CHx intermediates and selectivity to
ethylene, because these conditions yield surfaces with stronger metal–sulfur bonding (for example, Pd16S7), which
suppresses the over-oxidation of methane.
he anticipated long-term decline in petroleum reserves catalytic strategies to enhance the selective conversion of
coupled with increasing worldwide energy demands will methane to olefins or to other value-added intermediates with
T
require alternative hydrocarbon feedstocks and conversion lower heat emission are therefore pivotal to developing viable
processes to produce fuels and basic chemical intermediates such methane conversion processes16–23
as ethylene and propene1. Methane, the principal component of
In comparison to O2, the thermodynamic driving force for
natural gas, is proposed as a suitable alternative because of its abun- methane over-oxidation by S2 is significantly less (it is a ‘softer’
dant reserves and high hydrogen:carbon ratio2–4. Although methane oxidant; see Table 1 for a thermodynamic comparison of O2 and S2
steam reforming to give synthesis gas (CO þ H2)5,6—followed by as oxidants)24,25. This result arises primarily from the weaker C–S
catalytic conversion to methanol and other oxygenates7,8 or and H–S bonds and the lower bond enthalpies of the expected inter-
Fischer–Tropsch synthesis9—is widely practised, these routes are mediates and products in comparison with those involving O2. Note
highly capital-intensive. A direct, high-selectivity, one-step route for that the vapour-phase oxidative coupling of methane by sulfur (as
.
the methane-to-olefins process would therefore be highly desirable.
gaseous S2) to produce ethylene is entropically driven, and is exergo-
Significant methane reserves are currently found in ‘stranded’ nic only at temperatures above 950 K (Table 1, reaction 6). The use of
locations, and utilizing these reserves will require strategies for sulfur as the oxidant in methane conversion may alleviate engineering
their efficient conversion to higher-value and/or more transportable issues associated with the significant local temperature excursions
products10. Indeed, for methane to be a viable carbon feedstock,
more selective, less capital-intensive conversion processes are
needed—a true ‘grand challenge’11. To date, approaches explored
Table 1 | Thermodynamic comparison of O2 versus S2 as
oxidants for oxidative methane reactions at 1,073 K and
1,323 K.
for the direct, large-scale chemical transformation of methane to
useful chemicals include aromatization12, oxyclorination13,14 and
oxidative coupling15. However, these processes all have recognized
limitations, including modest selectivities and yields, the require-
ment for corrosive reagents, heat management and temperature
control, and/or dependence on toxic halogenated intermediates.
The seemingly straightforward oxidative transformation of
methane with O2 is in principle an attractive approach15;
however, the yields to desired products are typically limited by
severe over-oxidation pathways. Desired targets such as methanol,
formaldehyde and ethylene are far more reactive with O2 than
methane, and are therefore rapidly converted into more thermo-
dynamically stable products, principally CO2. Furthermore, the
reactions with O2 are highly exothermic, and can increase local
temperatures within reactors by 150–300 8C, thereby presenting
significant heat management and reactor design issues. New
Oxygen oxidation processes
DG8 (kJ mol21
)
(1)
2CH4 þ O2 ꢀ C2H4 þ 2H2O
2CH4 þ ½O2 ꢀ C2H6 þ H2O
CH4 þ ½O2 ꢀ CH3OH
2307 (2313*)
2117 (2104*)
275.3 (262.3*)
2792 (2797*)
21,294 (21,286*)
(2)
(3)
(4)
(5)
CH4 þ 2O2 ꢀ CO2 þ 2H2O
C2H4 þ 3O2 ꢀ 2CO2 þ 2H2O
Sulfur oxidation processes
2CH4 þ S2 ꢀ C2H4 þ 2H2S
2CH4 þ ½S2 ꢀ C2H6 þ H2S
CH4 þ ½S2 ꢀ CH3SH
DG8 (kJ mol21
)
(6)
(7)
(8)
(9)
(10)
24.90 (215.1*)
þ33.9 (þ45.1*)
þ35.5 (þ46.0*)
2124 (2145.3*)
2236 (2235*)
CH4 þ 2S2 ꢀ CS2 þ 2H2S
C2H4 þ 3S2 ꢀ 2CS2 þ 2H2S
*DG value at 1,323 K.
1Department of Chemistry and the Center for Catalysis and Surface Science, Northwestern University, Evanston, Illinois 60208-3113, USA, 2Departments of
Chemical Engineering and Chemistry University of Virginia, Charlottesville, Virginia 22904-4741, USA. e-mail: t-marks@northwestern.edu;
*
Neurock@virginia.edu
104
NATURE CHEMISTRY | VOL 5 | FEBRUARY 2013 | www.nature.com/naturechemistry
© 2013 Macmillan Publishers Limited. All rights reserved.

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
NATURE CHEMISTRY DOI: 10.1038/NCHEM.1527
ARTICLES
hydrocarbons, using argon as the balance gas (see
Supplementary Section A1 for details). Figure 2 presents a
summary of methane conversion and ethylene selectivity data as
a function of temperature over a series of catalysts. Note that
these reactions are very clean and the only detectable carbon-con-
taining products observed in the effluent stream are ethylene and
carbon disulfide. Control experiments included blank reactions in
quartz chip-packed reactor tubes. Passing methane through the
reactor at 1,173 K in the absence of sulfur produces negligible
amounts of ethylene. Although the reactions are carried out at
relatively high temperatures, there is no evidence of significant
coke formation on the catalyst surfaces after reaction, as assayed
by UV–Raman spectroscopy39 and combustion analysis (for a
spent PdS catalyst operated for 4 h at 1,273 K, C ≤ 0.03 wt%),
and the mass balance in these processes is invariably .98%. Both
methane conversion and ethylene selectivity (reported as two
times the moles of ethylene produced/moles of methane reacted)
increase markedly with increasing reaction temperature for all cata-
lysts examined. The selectivity for ethylene is also strongly catalyst-
dependent (Fig. 2b), following the trend PdS . TiS2 . RuS2 .
MoS2. PdS shows the highest ethylene selectivity, nearing 18% at
1,323 K, with methane conversion and ethylene selectivity
CH4
S2
0
C2H4
–5
–10
S2
O2
–15
–20
CS2
C2H4
–400
–600
–800
–1,000
–1,200
–1,400
–1,600
O2
CO2
Reaction coordinate
Figure 1 | Thermodynamic calculations show that using gaseous sulfur (S2)
as a ‘soft’ oxidant can hinder the over-oxidation of methane when
compared with using O2 as the oxidant. Comparison of the reaction
thermodynamics for methane oxidative coupling to ethylene and over-
oxidation by S2 (blue) and O2 (red), at 1,073 K, to undesired by-products.
a
16
that occur when O2 is used. The thermodynamic data show that selec-
tive catalytic oxidative coupling of methane by S2 should be possible if
methane can be activated selectively. Note that the over-oxidation of
methane to CS2 or other species (Table 1, reactions 7–10) is far less
favourable than the analogous over-oxidation by O2 (Table 1, reac-
tions 2–5), as summarized graphically in Fig. 1.
Blank
PdS
14
MoS2
12
10
8
RuS2
TiS2
There is an extensive body of information concerning catalyst
structure and reactivity26–29 for the industrial desulfurization of
hydrocarbons over transition-metal sulfides such as Co/MoS2 and
RuSx catalysts30–35. However, very few studies have reported catalytic
reactions involving elemental sulfur and saturated hydrocar-
bons36,37. Here, we report on the catalytic conversion of methane
to ethylene (as expressed in reaction 6 of Table 1), using elemental
sulfur as an oxidant over transition-metal sulfide catalysts.
Experimental results indicating promising selectivity are used
together with first-principles quantum chemical calculations to
show that both the activation of methane and the selectivity of
the subsequent conversion to ethylene can be linearly correlated
with the catalyst metal–sulfur (M–S) bond strength (defined here
as the energy to remove atomic sulfur as H2S from the sulfide
surface, using H2), and that the two relationships are inversely
related to one another. The sites that are most active for
methane activation involve weakly bound surface sulfur atoms,
which promote hydrogen abstraction, whereas those that are
most selective involve strongly bound surface sulfur atoms, which
facilitate hydrocarbon fragment coupling to C2 products and
suppress the over-oxidation of methane. Higher methane:sulfur
(CH4/S) feed ratios and higher temperatures enhance methane
conversion and ethylene selectivity, and at the same time appear
to reduce the active sulfide catalysts and increase the metal:sulfur
compositions at the surface. This in turn decreases overall conver-
sion while increasing the selectivity to ethylene. These insights
demonstrate that sulfur is a promising reagent for the selective
oxidative transformation of methane and should help guide the
discovery of even more selective catalytic materials. The highest
methane-to-ethylene selectivity is found here for supported PdSx
catalysts operating at relatively high temperatures, high CH4/S
ratios and short contact times.
6
4
2
0
1,050
1,100
1,150
1,200
1,250
1,300
1,350
Temperature (K)
20
15
10
5
b
Blank
PdS
MoS2
RuS2
TiS2
0
1,050
1,100
1,150
1,200
1,250
1,300
1,350
Temperature (K)
Figure 2 | Unsupported metal sulfides are effective catalysts for the
conversion of methane to ethylene using gaseous sulfur as a ‘soft’ oxidant.
a,b, Methane conversion (a) and ethylene selectivity (b) as a function of
temperature over the indicated transition-metal sulfide catalysts. Reaction
conditions for all metal sulfide catalysts: 5% CH4 concentration in Ar;
CH4/S ratio ¼ 5.8; overall WHSV ¼ 30,000 ml g21 h21. Estimated
Results and discussion
Catalytic reactions were carried out in a flow reactor specially uncertainties in data points are +5% of the stated value, and the straight
designed to generate gaseous sulfur38 as S2 for reactions with lines are drawn as a guide to the eye.
NATURE CHEMISTRY | VOL 5 | FEBRUARY 2013 | www.nature.com/naturechemistry
105
© 2013 Macmillan Publishers Limited. All rights reserved.

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
NATURE CHEMISTRY DOI: 10.1038/NCHEM.1527
ARTICLES
favourable surface termination at specific sulfur chemical potentials
and temperatures. As has been shown previously for methane oxi-
dation, the chemical potential sets the composition of the surface
as well as the nature of the active surface sites40,41. The lowest
energy states and compositions for all of these structures were sub-
sequently used to calculate the reaction energies and barriers for
methane activation and the subsequent coupling of hydrocarbon
fragments to form ethylene. These results suggest that under
nearly all experimental conditions, MoS2, RuS2, TiS2 and PdS are
highly sulfided and the active sites comprise S–S pairs. The sole
exception, as will be discussed below, occurs at high temperatures
and high CH4/S ratios where theory and experiment indicate that
PdS is reduced to Pd16S7.
a
b
The rate of methane conversion is controlled by the activation of
the initial CH3–H bond, whereas catalytic selectivity for ethylene is
controlled by the rate of coupling of two bound methylene (CH2–S*)
fragments to form ethylene versus the rate of further C–H bond
activation to ultimately form CS2. The calculated transition states
for CH3–H activation over S–S sites and the coupling of CH2–S*
fragments to form ethylene are shown in Fig. 3 and involve charac-
teristic hydrogen abstraction and reductive C–C formation steps,
respectively. Both reactions are strongly influenced by the M–S
bond strength, which sets the basicity and reactivity of the surface
sulfur sites. Weaker M–S bonds can readily abstract hydrogen and
bind the resulting methyl groups more strongly, thus enhancing
the activation of methane, whereas strong M–S bonds result in
weaker C–S bonds, which enhance C–C coupling and increase the
selectivity to form ethylene. The barriers for both the C–H bond
activation of methane and the subsequent coupling of CH2
surface intermediates to form ethylene appear to be linearly
related to the M–S bond strength, as shown in Fig. 4.
Figure 3 | Ab initio DFT calculations shows the characteristic reactant,
transition state and product structures for the catalytic conversion of
methane to ethylene using gaseous sulfur as the oxidant. a,b, Optimized
reactant, transition state and product structures for the initial C–H activation
of methane over surface S–S pairs (a) and CH2 coupling to form ethylene at
the sulfur edge of the MoS2 surface (b). Similar structures are shown in the
Supplementary Information for these same reactions at S–S sites on model
RuS2, PdS and TiS2 surfaces. Yellow, sulfur; light blue, Pd; white, H; grey, C.
substantially higher with rising temperatures. In contrast, both the
RuS2 and MoS2 catalysts exhibit very low ethylene selectivities—
lower than the selectivity measured in the ‘blank’ experiments
over the entire temperature range—because they efficiently catalyse
over-oxidation to CS2. A minimal ethylene yield is observed
for all catalysts at temperatures below ꢁ1,073 K. The absence
of detectable ethane and methanethiol formation over any of the
catalysts is in agreement with thermodynamic expectations that
the formation of these products is unfavourable under these con-
ditions (Table 1).
CHx coupling can conceivably occur either on the surface of the
catalyst or in the gas phase. The increased selectivity and conversion
with increasing temperature reported in Fig. 2 in principle supports
both surface as well as gas-phase CHx coupling pathways, because
increasing the temperature or conversion will increase CHx
species formation on the surface as well as in the gas phase. The
increase in selectivity that occurs when using less active catalysts,
however, would indicate that, regardless of where the actual CHx
coupling takes place, the key step involves CHx–S* bond weakening
to generate the active intermediates. Although we will explicitly
discuss the coupling of CHx–S* species on the surface, the results
are essentially the same if the coupling occurs in the gas phase, as
the desorption rate of CH·3 radicals from the surface is inversely
related to the CH3–S* bond strength and linearly related to the
M–S bond strengths.
First-principles density functional theory (DFT) calculations
were carried out to help understand the reaction pathways and
mechanisms, the nature of the active catalytic sites and the effects
of the catalyst, as well as the reaction conditions on methane conver-
sion and the selective formation of ethylene. Details concerning the
computational methods used, as well as the results for the initial
activation of methane, the subsequent C–H activation of other
CHx–S* intermediates, the coupling of surface hydrocarbon inter-
mediates (CHx–S*) to form ethylene, and the desorption of CHx
species from the surfaces are discussed in Supplementary Sections
B1, B2. The coupling of CHx intermediates was explored on the
surface as well as in the gas phase.
Detailed structural optimizations were carried out on the close-
packed 111, 100 and 110 surfaces of MoS2, RuS2, TiS2 and PdS
over a range of sulfur coverages to establish the lowest energy struc-
tures of the surfaces at different temperatures and CH4/S ratios. The
results for these model MoS2, RuS2, TiS2 and PdS surfaces (shown
in Supplementary Fig. S5) were used to determine the most
The computed barriers for methane activation over the different
metal sulfides reported in Fig. 4 increase with increasing M–S bond
strength as MoS2 . RuS2 ≈ PdS . TiS2 and directly track the
300
C–H activation
MoS2
Pd16S7
TiS2
250
PdS
RuS2
200
PdS
RuS2
MoS2
TiS2
150
100
Pd16S7
160
CH2 coupling
120
60
80
100
140
M–S bond energy (kJ mol–1
)
Figure 4 | Calculations reveal a correlation between both C–H activation
and CH2 coupling with M–S bond strength. The results from DFT quantum
chemical calculations reveal a linear correlation between both the calculated
activation barriers for methane C–H activation (blue) and the coupling of
methylene (CH2–S*) intermediates to form ethylene (red) with the
corresponding M–S bond strength over S–S site pairs on model MoS2, RuS2,
PdS, TiS2 and Pd16S7 surfaces. The C–H activation and CH2 coupling
barriers over S–S sites on reduced Pd16S7 surfaces are indicated separately
with larger symbols. The best fit lines and the corresponding values of the
correlation coefficient (R) for methane activation (blue) and CH2 coupling
(red) are calculated to be Ea(C–H activation) ¼ 100.27+1.15 kJ mol21 (M–S
bond energy), R ¼ 0.990; and Ea(CH2 coupling) ¼ 367.14+1.60 kJ mol21
(M–S bond energy), R ¼ 0.9958, respectively.
106
NATURE CHEMISTRY | VOL 5 | FEBRUARY 2013 | www.nature.com/naturechemistry
© 2013 Macmillan Publishers Limited. All rights reserved.

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
NATURE CHEMISTRY DOI: 10.1038/NCHEM.1527
ARTICLES
experimental low-temperature conversion results in Fig. 2a.
Although entropy has a significant effect on the reaction rate at
high reaction temperatures, the changes in entropy that result
from changes in the sulfide are very small and can be neglected
without loss of generality. Methane activation is governed by a
hydrogen abstraction, for which the reactant and transition-state
structures for all of the sulfides examined were nearly identical.
Methane conversion is clearly governed by the M–S bond
strength. Sulfides with weaker M–S bonds result in more basic
and reactive sulfur sites that can readily abstract hydrogen in the
transition state and bind the resulting CH3 and hydrogen inter-
mediates. Subsequent C–H activations of the resulting CHx*
species have lower activation barriers and more favourable acti-
vation entropies than those for methane, and are therefore more
reactive. As such, metal sulfides with weaker M–S bonds will not
only readily activate the C–H bond of methane, but will continue
on to activate the C–H bonds of the subsequent CHx fragments
and ultimately mediate over-oxidation to CS2. The calculations
for C–H activation at CH3*–S, CH2*–S and CH*–S sites on
MoS2 yield barriers of 119, 63 and 178 kJ mol21, respectively,
which are significantly lower than the initial 184 kJ mol21 barrier
for methane activation.
a
21
18
15
12
9
CH4 conversion for PdS
C2H4 selectivity for PdS
CH4 conversion for blank
C2H4 selectivity for blank
6
3
0
1
2
3
4
5
6
7
CH4:S
b
10
8
CH4 conversion
C2H4 selectivity
In contrast to these results, computed trends in the methane-to-
ethylene selectivity are inversely related to M–S bond strength, as
also shown in Fig. 4, where the barriers for ethylene formation
decrease as MoS2 . RuS2 . PdS . TiS2. These results are consistent
with the experimental results shown in Fig. 2b at temperatures
below 1,175 K, with the exception that PdS is only slightly more
selective than TiS2. Similar results for selectivity are found if the
CH·3 radicals desorb from the surface and recombine in the gas
phase to form ethane, because the CH·3 desorption barriers are lin-
early correlated with the CH2–S* coupling barriers, as shown in
Supplementary Fig. S12. The barriers to activate the C–H bonds
of CH3–S* intermediates and subsequently couple the CH2–S*
species to form ethylene are both lower than the barriers for CH3·
radical desorption, as shown in Supplementary Fig. S12. This
suggests that CH2–S* coupling on the surface is more favourable
than the desorption and gas-phase coupling of CH·3 radicals.
A more detailed analysis of the C–H bond activation steps versus
6
4
0
10,000
20,000
30,000
WHSV (ml gcat–1 h–1
)
Figure 5 | Markedly altered PdS-catalysed methane conversion and
ethylene selectivity are observed by varying the CH4:S feed ratios and
WHSVs over the catalyst surface. a,b, Effect of CH4/S ratio (reaction
conditions: temperature ¼ 1,223 K, sulfur vapour concentration ¼ 0.85%,
...
the CH2–S* CH2–S* coupling reactions indicates that the M–S
overall WHSV ¼ 30,000 ml g21 h21) (a) and WHSV (reaction conditions:
1,173 K, 5% CH4 concentration in Ar, CH4/S ratio ¼ 5.8) on methane
conversion and ethylene selectivity over a PdS catalyst (b). Estimated
uncertainties in data points are +5% of the stated value, and the straight
lines are drawn as a guide to the eye.
bond becomes sufficiently strong in PdS and TiS2 to begin to
reverse the driving force to over-oxidize to CS2 and instead increases
the likelihood of CH2–S* coupling to form ethylene. On PdS, the
barrier to CH2–S* coupling to form ethylene is 183 kJ mol21
,
which is only 50 kJ mol21 greater than the barrier to activate the
C–H bond of CH2–S* (134 kJ mol21). Note, however, that the
change in entropy for CH2–S* coupling is significantly more favour-
able than that for C–H bond activation. At higher temperatures,
both steps become favourable. On TiS2, the CH2–S* coupling
process is computed to be far more favourable than the C–H acti-
vation of CH2–S* species.
(Supplementary Fig. S3) reveals that these remain unchanged under
all reaction conditions.
The CH4/S ratio is found to be a critical parameter governing
selectivity in the present catalytic conversions over PdS. The
experimental results in Fig. 5a show that over-oxidation of
methane to CS2 is suppressed at high CH4/S ratios, thereby
enhancing ethylene selectivity. In reactant streams with CH4/S
ratios less than 1:1, ethylene formation is suppressed, and CS2 is
the only observed carbon-containing product. The dependence of
selectivity on the CH4/S ratio logically follows because over-
oxidation is stoichiometrically more favourable at high oxidant
concentrations. The increasing ethylene selectivity with increasing
CH4/S over PdS is consistent with the experimental results
presented in Fig. 2b, which show a significant increase in selectivity
with increasing temperature. Both higher temperatures and higher
Further examination of the experimental results in Fig. 2b indi-
cates that although the selectivity to ethylene increases with increas-
ing temperature for all of the sulfides, the results for PdS are
characteristically different because the selectivity increases far
more dramatically with temperature than for the other sulfides.
As can be seen in Fig. 2a, there also appears to be a distinct
change in the relative experimental ordering of methane activation
over the different metal sulfides at temperatures above ꢁ1,150 K.
X-ray diffraction (XRD) data for the PdS catalyst before and after
operation at high temperatures are shown in Supplementary
Fig. S2A and compared with PDF (Powder Diffraction File)
CH4
/S ratios lead to the reduction of PdS to Pd S . The results
database spectra for PdS (Supplementary Fig. S2B) and Pd16S7 clearly show that the more reduced Pd16S7 state is l7ess active but
16
(Supplementary Fig. S2C). These data suggest that the bulk of the
PdS is reduced to Pd16S7 under catalytic reaction conditions.
Similar XRD characterization of the initial and spent MoS2 catalysts
significantly more selective for ethylene. Similar phase changes
between PdO and oxygen-covered Pd metal, and concomitant
decreases in the rate of catalytic methane activation, have been
NATURE CHEMISTRY | VOL 5 | FEBRUARY 2013 | www.nature.com/naturechemistry
107
© 2013 Macmillan Publishers Limited. All rights reserved.