You are on page 1of 10

Articles

https://doi.org/10.1038/s41929-019-0395-3

Imaging the dynamics of catalysed surface


reactions by in situ scanning electron microscopy
Cédric Barroo   1,2,3,5*, Zhu-Jun Wang   1,4,5*, Robert Schlögl1 and Marc-Georg Willinger   1,4,5*

Analytical methods that provide direct real-space information about the dynamics of catalysed reactions often require sim-
plified model systems and operate under high-vacuum conditions. There is thus a strong need for the development of meth-
ods that enable observation of active catalysts under relevant working conditions. Here, in situ scanning electron microscopy
is employed to study reaction dynamics and structure–activity correlations on surfaces. High sensitivity to changes in the
work function and surface composition enables the detection of monolayers of adsorbed molecular species on metal sur-
faces, which is used here to visualize catalytic NO2 hydrogenation on platinum. The initiation of reactive behaviours and
propagation of reaction fronts, as well as the spillover of activated species revealed in real-time and across a large pressure
range, demonstrate the power of in situ scanning electron microscopy as a surface science tool in the study of gas-phase- and
temperature-induced processes.

W
hether we want to study the synthesis, reactivity or age- image surface reactions: photoemission electron microscopy
ing of catalytic materials, only experiments performed (PEEM) and low-energy electron microscopy (LEEM) for the study
during the ongoing process are able to capture the of flat, conductive samples (single crystals or foils)9–11 and field ion/
information that is required to establish reliable structure–function emission microscopy (FIM/FEM) for the analysis of tip-shaped
relationships. Catalysts are operating far from thermodynamic equi- samples12,13. PEEM provides submicrometre resolution and a maxi-
librium in an open system that exchanges heat and matter with the mal field of view of a few hundred micrometres, whereas FIM/FEM
surroundings. In their active state they are supposed to repeatably provides rather localized information at nanoscale resolution. Both
facilitate sequences of reaction steps of a catalytic cycle. The combi- techniques can be operated at a maximum pressure of ~10−3 Pa. It
nation of non-equilibrium conditions and inherent cyclic behaviour has only very recently been demonstrated that near-ambient-pres-
(that is, the turnover frequency) of a catalytic reaction can result in sure LEEM can be used to image surfaces under reaction pressures
complex dynamics and the formation of dissipative structures that of up to ~10 Pa (ref. 14). Scanning tunnelling microscopy15,16 and
are relevant to our understanding of catalytic function; catalysts ellipsomicroscopy for surface imaging17 enable measurements at
should thus be studied in their working state. higher pressures, but they have rarely been used for the study of
Conventional surface science methods have been applied as reaction dynamics.
powerful analytical tools for the investigation of fundamental In the field of catalysis, analytical scanning electron microscopy
interactions and processes taking place on catalytically active sur- (SEM) plays an important role in the characterization of precursor
faces; however, local observation of well-defined structures under phases and ex  situ or post-mortem analysis of reaction-induced
high-vacuum conditions can only provide an incomplete picture changes of catalysts. With the development of variable pressure
about the behaviour of catalysts under industrially relevant con- SEM and environmental SEM (ESEM), the observation of mate-
ditions. Extrapolation from quasi- to non-equilibrium conditions rials in a reactive atmosphere and the study of reaction-induced
is complicated by nonlinear and complex collective dynamics. dynamics became possible18,19. It seems, however, that the potential
Consequently, there has been a growing interest to develop in situ of in situ SEM is still not fully unfolded. In addition to its ease of
techniques that allow spatially resolved observations of catalysts in use, the ESEM combines some useful capabilities; for example, it
their reactive state1–5. In this frame, in  situ transmission electron does not require ideally flat model surfaces and thus allows obser-
microscopy has developed to a powerful technique for local inves- vation of samples with complex morphology; it provides a large
tigations of temperature- and gas-phase-induced processes6–8; how- range of magnifications and higher spatial resolution than PEEM
ever, due to the highly localized and even atomic-scale information or LEEM; and it can also be operated at pressures ranging from
provided by in  situ transmission electron microscopy, the ability 10−5 Pa to near-ambient pressure20. Finally, it is not limited to the
to study collective processes and dynamics is somehow limited. study of conductive samples. Lately we have modified the set-up of
Additional in situ methods are thus required to enable a multiscale, a conventional ESEM to use it as in situ tool for the investigation of
context-embedded observation. gas-phase-induced reactions that are relevant for catalysis21, corro-
With a focus on studying the reaction dynamics rather than sion22, material growth and decomposition23–25. Direct observation
the surface dynamics (that is, how the catalytic reaction proceeds of graphene growth by catalytic chemical vapour deposition on
rather than how the catalyst’s surface changes during the reaction), different substrates at temperatures of up to 1,000 °C demonstrated
two main families of techniques have been successfully used to the high sensitivity of the secondary-electron signal to changes in

1
Department of Inorganic Chemistry, Fritz Haber Institute of the Max Planck Society, Berlin, Germany. 2Chemical Physics of Materials and Catalysis, Université
libre de Bruxelles, Brussels, Belgium. 3Interdisciplinary Center for Nonlinear Phenomena and Complex Systems (CENOLI), Université libre de Bruxelles,
Brussels, Belgium. 4Scientific Center for Optical and Electron Microscopy (ScopeM), ETH Zurich, Zurich, Switzerland. 5These authors contributed equally:
Cédric Barroo, Zhu-Jun Wang, Marc-Georg Willinger. *e-mail: cbarroo@ulb.ac.be; zhujun.wang@scopem.ethz.ch; marc.willinger@scopem.ethz.ch

Nature Catalysis | www.nature.com/natcatal


Articles Nature Catalysis

surface coverage26–28. The ability to distinguish individual layers of Here, n* and m* represent the number of n or m free surface sites
atomically thin carbon sheets in few-layer graphene28 suggested and X* represents intermediate X adsorbed on a surface site.
that the instrument should be capable of detecting modulations
of the work function and secondary-electron emission that are Surface sensitivity to gas adsorption. The first set of experiments
caused by changes in surface coverage of a metal substrate with was aimed at confirming the reactivity of a polycrystalline platinum
atomic or molecular species. If that is true, it should then be pos- foil towards NO2 hydrogenation. Titration experiments were per-
sible to visualize reactions between adsorbed species on the surface formed at low pressure; that is, in the range between 10−4 to 10−2 Pa.
of a catalyst. Adsorbed NO2 species and its dissociation products O(ads) increase
The hydrogenation of NO2 was chosen as a test reaction. This the work function of platinum surfaces35,37,38, whereas in the case of
reaction is relevant in the frame of automotive pollution control, NO(ads), the effect of the work function depends on the adsorption
as it was highlighted by recent scandals in the automotive indus- mode39. As NO(ads) undergoes thermal desorption at the work-
try. NO2 hydrogenation was previously studied on a single model ing temperatures40,41, its contributions to contrast modulation are
nanoparticle of platinum by field emission microscopy29–33. The expected to be small. Overall, the adsorption of NO2 thus causes a
selected reaction presents several advantages: NO2 has a high stick- reduction of low-energy secondary-electron emission, which might
ing coefficient34, adsorbed NO2 species are known to increase the be detectable as a decrease in image brightness. As shown in the
work function of platinum35 and nonlinear dynamics can emerge sequence of SEM images that were recorded with the Everhart–
during the reaction without oxidation of the catalyst at low pres- Thornley detector (Fig. 1a–c), the admission of NO2 into the cham-
sures. In FEM studies, the nonlinear behaviours are observed via ber of the ESEM induces an inhomogeneous, wavelike propagation
periodic variations of the brightness signal of the FEM pattern. of darker contrast across the platinum foil. The observed appearance
They are due to periodic variations of the local work functions and propagation of areas with reduced brightness and contrast is
and induced by changes in the adsorbate-layer composition during indicative of an adsorption process. If brightness/contrast changes
the reaction. were due to the increase of the pressure in the ESEM or due to the
Here we show that nonlinear behaviours can also be detected properties of NO2 as an imaging gas (in the case of a gaseous second-
when the reaction is performed on platinum substrates inside the ary-electron detector42) then a homogeneous change in brightness
chamber of an ESEM. Adsorbate induced modulations of the work and contrast would be expected across the whole imaged region.
function can be detected as variations in the secondary-electron When hydrogen is subsequently introduced and the tempera-
signal and used to visualize reaction-induced changes in surface ture increased, the initial brightness and contrast is reestablished,
coverage. Grain-orientation-dependent reactivity and propagation indicating that the adsorbed layer can be reacted off. As shown in
of chemical waves, spillover coupling and formation of complex Fig. 1d–f, the titration process is not homogeneous: it preferentially
spatio-temporal dynamics are observed and presented as a func- initiates at specific surface features and grain boundary regions. The
tion of gas-phase composition, temperature and pressure. The high behaviour presented in Fig. 1d–f strongly suggests the possibility of
flexibility of the SEM combined with the here demonstrated capa- detecting adsorption and desorption processes and thus reactivity
bility as surface science instrument for in situ studies will enable of the platinum surface towards NO2 hydrogenation.
a breakthrough in real-time observation of complex-shaped cata-
lysts across a large range of lateral scales and from high-vacuum to Initiation of spatio-temporal dynamics. The hydrogen partial
near-ambient pressures. pressure was varied at a fixed flow of NO2 once the reactivity of
the surface had been confirmed. The hydrogen concentration acts
Results as a control parameter in these experiments and is increased until
Proposed elementary steps. The reaction of NO2 with H2 is sup- the moment when sustained activity is observed. Reaction-induced
posed to produce water according to the mechanism discussed changes in surface coverage should result in time-variant bright-
in previous publications:31,33,36 ness changes in the secondary-electron image under these condi-
Adsorption of the reactants on the platinum surface generates tions. If these variations seem spatially coherent at scales that can be
NO*, O* and H* species: resolved by the instrument, and at time scales that can be captured
within the achievable temporal resolution, it should be possible to
NO2 þ n* $NO* þ O* þ ðn � 2Þ* ð1Þ observe them. It is indeed known from PEEM and LEEM on well-
defined metal surfaces that reactions between adsorbed species can
be detected as contrast variations over time; they are caused by local
H2 þ 2* $2H* ð2Þ modulations of the work function due to reaction-induced changes
in surface coverage. The propagation of chemical waves is a typical
The NO* can desorb or further dissociate, depending on the cov- feature of 2D excitable systems, in which a local initiation centre
erage and temperature of the surface: (that is, a pacemaker) induces a reaction that further propagates in
the form of chemical waves. Coupled reaction–diffusion dynamics
NO* $NOþ* ð3Þ can give rise to complex spatio-temporal dynamics9. The observa-
tion of spatio-temporal surface dynamics by in  situ SEM would
thus demonstrate that the method provides sufficient sensitivity to
NO* þ m* $N* þ O* þ ðm � 1Þ* ð4Þ detect reactions between adsorbed species on metallic surfaces.
Figure 2 shows SEM images that were recorded while H2 was
slowly added to the NO2 flow using an ultra-high-vacuum (UHV)
N* þ N* ! N2 þ 2* ð5Þ leak valve. During the experiments, the magnification was gen-
erally chosen such that several grains could be monitored at the
Finally, N2 and water are produced and leave the surface, creating same time. The initially low brightness of the surface indicates full
two and three vacant sites, respectively, allowing further reactants to coverage by NO2 and/or O(ads) (Fig. 2a)35,37,38. A gradual increase
adsorb onto the surface and the process to continue. The reaction of the H2 flow combined with simultaneous imaging allows a con-
can thus be considered as autocatalytic:33 trolled transition of the system into its active regime. The onset of
reactivity is first observed as a change in brightness on some grains
2H* þ O* ! H2 O þ 3* ð6Þ and along grain boundaries, indicating that the adsorbed layer is

Nature Catalysis | www.nature.com/natcatal


Nature Catalysis Articles
pNO2 10–2 Pa

a b c

d e f

315 °C 347 °C

Fig. 1 | Surface sensitivity of the ESEM instrument. a–c, A sequence of SEM images showing that the dissociative adsorption of NO2 on a platinum
foil induces a decrease in brightness due to work function changes. Conditions of acquisition: T = 306 °C; initial chamber pressure = 10−5 Pa. NO2 was
introduced until the pressure reached ~10−2 Pa ðpchamber ¼ pNO2 Þ. d–f, Subsequent introduction of H2 and heating from 315 °C to 347 °C induces a reaction.
I
It is detected as increase of brightness and recovery of the initially clean surface. Conditions of acquisition: pNO2  pH2  102 Pa. Scale bar: 100 µm.
I

starting to react off (Fig. 2b). The H2 flow was kept constant from whereas the other wave pattern disappears. This can be seen in
this point onwards. Once NO2 and H2 are delivered at a constant rate Fig. 2d,e, where the active region transferred from B ‘collides’ with
and a well-balanced ratio, the sequence of readsorption, reaction the active region transferred from C, and similarly in Fig. 2f–h
and desorption can happen repetitively and, thus, eventually result where the active region transferred from grain D collides with
in the emergence of spatio-temporal behaviours. As can be seen in the dominant dynamics on grain A. As indicated by the arrows in
Fig. 2 and Supplementary Video 1, only some of the grains show ini- Fig. 2c and as can be observed in Supplementary Video 1, reaction
tial activity (manifested by a change in brightness and appearance of fronts can propagate along specific surface features and along grain
wavelike contrast variations, inactive grains remain in a low-bright- boundaries in the form of chemical waves. Spillover can therefore be
ness state). It is well known that a single crystal surface with a well- transferred across grains through excitability; reaction fronts initi-
defined crystallographic orientation will be active for the reaction ated by spillover from one grain can travel along chemical wave-
only in a restricted parametric region ðT; pNO2 ; pH2 Þ. This paramet- guides or chemical channels through an excitable surface layer on
ric region depends on the orientation Iof the grains, as the surface another grain and induce spillover onto further grains; thus acting
structure influences the sticking coefficients of NO2 and the activa- over extended distances. These observations prove the importance
tion energies for dissociation, reaction, diffusion and/or desorption of spillover effects and intergrain coupling in the emergence of com-
of the involved species. In the case of a polycrystalline sample, it plex dynamics and, most importantly, in the global reactivity of a
is thus expected that not all of the grains are simultaneously reac- catalyst. Due to spillover, the catalytic activity of a platinum foil is
tive under specific experimental conditions; however, because of the not the same as the sum of the catalytic activities of all the facets
structure of the sample as a whole (that is, the presence of intercon- taken separately.
nected grains with different crystallographic orientations), a reac-
tive grain can stimulate reactivity on an initially unreactive grain. Grain-dependent spatio-temporal dynamics. Simultaneous
Spatial coupling is most likely enabled through the diffusional flow imaging of surface dynamics on different grains (as exemplified in
of mainly hydrogen species, which can easily diffuse at the working Fig. 2), coupled with grain orientation analysis by electron back-
temperatures9,43,44. Such a process can be observed in Fig. 2, where scatter diffraction (EBSD), allows us to draw some initial conclu-
grain A seems to be non-reactive (Fig. 2b) but becomes reactive near sions about the reactivity of differently oriented grains. According
the grain boundary by spillover coupling with grain B (see arrows to EBSD, variations in brightness due to the onset of reactivity were
in Fig. 2c). After activity is initiated, grain A develops its own spe- first observed on {110}-type surfaces (such as grain B), the (100)
cific dynamics with an associated periodicity and initiation centre grain on the bottom left and the (111) grain near the top-right cor-
(labelled × in Fig. 2d). Similarly, reactivity of the active grain C (and ner (see Fig. 2i and Supplementary Fig. 1). Although the (100) grain
later of grain D) is transferred to grain A through surface diffusion shows an irregular, flickering contrast with a dominance of the
coupling (Fig. 2d–f). The active areas expand with time and propa- bright surface state, the more open {110}-type facets show a com-
gate into so-far-inactive areas, which seem dark due to dominating plex pattern of propagating reaction–diffusion fronts with a spiral
NO2(ads)/O(ads)35,37,38 coverage. During the gradual activation of pattern. In the case of the (111) grain, the surface is mostly seen
the surface, one can see how different dynamics can emerge for a in the bright state, indicating either a low average NO2 coverage or
single grain, also depending on the neighbouring grains that can a non-reactive state under these conditions. The observed reactive
act as pacemaker (by triggering the light-off as chemical clocks). In oscillating behaviour and the pronounced activity of {011}- and
cases where areas of different dynamics merge or when two chemi- {001}-type grains agree well with earlier FEM observations, accord-
cal waves collide on a single grain, the dominant dynamics expand ing to which the most active facets are of {012}, {011} and {001}

Nature Catalysis | www.nature.com/natcatal


Articles Nature Catalysis

a b c
B

t=0s t = 2 min 52 s t = 5 min 36 s

d e f

t = 7 min 32 s t = 8 min 54 s t = 15 min 26 s

g h i

111

t = 16 min 01 s t = 16 min 04 s
001 101

Fig. 2 | Grain-dependent surface dynamics and spillover. a, Platinum foil under a pure NO2 atmosphere; low brightness indicates coverage of the surface
with NO2/O(ads). b, Grain-dependent change in brightness and onset of reactivity is observed once the H2 flow reaches a critical value. c, Spillover across
grain boundaries induces transfer of activity from active to initially inactive grains (marked with arrows). d, The initially non-reactive grain A becomes
partially reactive. The dynamics of the surface reaction are characterized by a specific periodicity and initiation centre (marked ×). A new initiation centre
(marked by *) results in spillover from grain C to A. e, In a collision of two chemical waves, the dominant dynamics prevail (here the ones originating from ×).
f,g, Spillover from grain D to A with its own initiation centre (marked by +). h, Another collision between two different chemical waves in which the
dominant dynamic overrides the other. i, An EBSD grain orientation map recorded after the in situ experiments, showing colour-coded information about
the orientation of platinum grains. In situ imaging conditions: T = 175 °C; pNO2 : pH2  ≈ 1:10; ptot ≈ 10−2 Pa. Scale bar: 100 µm.
I I

crystallographic orientations, with {113} and {111} being less active of chemical dynamics by in situ SEM is to avoid surface contami-
or inactive30–33. Regarding the activation of initially inactive grains nation during imaging. This requires working under very clean
by spillover, we observe two types of behaviours that are related to conditions and eventually involves long-time annealing of samples
surface reactivity and therefore grain orientation. As can be seen in (see Methods). Furthermore, observation at elevated temperatures
Supplementary Video 1, there are grains where spillover can induce (T ≈ 150–200 °C) is beneficial, as it reduces or even eliminates
a reaction front that propagates across the whole grain. According contamination build-up during imaging. Depending on the type
to the EBSD data, these are grains with surfaces close to a [101] ori- of reaction, surface processes can even induce a self-cleaning.
entation, but with a slight inclination towards [111]. In other cases, Contamination-related limitations therefore need to be evaluated
spillover first activates the region close to the grain boundary and individually for each type of system studied. In the here-presented
reactivity only gradually propagates further into still-unreactive NO2 hydrogenation, contamination was not an issue once the
areas from there. It seems that the rather stable initial coverage is samples were clean. The second point that has to be considered in
gradually replaced by a less stable, more reactive configuration in the study of reaction dynamics is related to electron beam-induced
such a process. The initial coverage was obtained in a pure NO2 artefacts. We have carefully evaluated possible effects of the elec-
atmosphere whereas subsequent adsorption processes occur in the tron beam by changing the local dose rate through changes in the
presence of hydrogen. The higher reactivity could thus be explained scanning resolution, pixel dwell time and beam current. We have
by hydrogen coadsorption and diffusion, as well as the presence of mimicked beam on–off experiments by moving the sample posi-
intermediate species (vide infra). tion and quickly changing the magnifications and observing con-
It is important to address the potential role of the electron beam sistencies across imaged areas. We can conclude from these that
in these experiments before attempting a deeper analysis of the the primary beam hitting the surface does not locally affect the
observations. The first and most crucial point for the investigation observed dynamics. Whether the secondary electrons (which have

Nature Catalysis | www.nature.com/natcatal


Nature Catalysis Articles
a far larger cross-section for the generation of radical species at manifestations of local structural inhomogeneities for which EBSD
the surface or in the gas phase) have an effect is difficult to say and secondary-electron imaging is not sensitive. For example,
because it is not possible to image the surface without them. They local variations in terrace size, step- and kink-density could result
are locally generated, but propagate along different trajectories in slightly increased stability of the initially adsorbed NO2 layer.
through the gas phase towards the detector. Effects related to gas- Alternatively, such unreactive areas could also be caused by poi-
phase interactions would thus be less localized spatially compared soning of the surface due to non-reactive intermediates that do not
with the primary beam. Although we cannot completely exclude undergo thermal desorption, but that can be desorbed by increas-
that the electron beam has an impact on the sticking coefficients, ing the temperature. As shown in Fig. 3a–c, the reaction conditions
adsorption and desorption rates and surface dynamics, the effect have a large effect on reaction dynamics and the observable spatio-
would certainly be very small compared with the effect of the sur- temporal pattern. Slight changes of the NO2/H2 ratio, temperature
face structure, gas composition and temperature: these were the or pressure can lead to activation or deactivation of complete grains
parameters that we varied that, independent of imaging condi- or portions of their surface. In Fig. 3a, for example, grains with a
tions, consistently gave rise to the complex behaviours observed. surface orientation close to [113] (that is, the pink-coloured grains
A comparison with PEEM, which is also based on secondary-elec- in the EBSD map, Fig. 3d) are at least partially active at 172 °C. At a
tron imaging, would certainly be interesting, also to compare pos- slightly lower temperature (164 °C), these grains remain in a dark,
sible roles of different excitation mechanism (ultraviolet light or non-reactive state, reflecting the lower activity of {113}-like surfaces
X-rays versus the primary electron beam in SEM). (Fig. 3b). It is important to note that hysteresis effects and, thus, the
A further source of artefacts is related to the use of a scanning history of experimental conditions, play an important role in the
technique. Surface dynamics can only be captured without alias- analysis of surface reactivity. Such effects are the reason why some
ing artefacts if the scanning speed is at least twice as fast as the fre- grains with a slight inclination from the [110] orientation towards
quency of the spatio-temporal dynamics. Most of the videos were the [111] or [100] orientations (the green-yellow and green-blueish
recorded at the rate of ~1 to ~0.3 frames per second (fps). Some of colours, respectively, in the inverse pole figure map) are active at
the chemical waves and rotating spirals seem to stand still or even a lower temperature (Fig. 3b), but seem inactive and with a dark
move backwards as a consequence of aliasing (see Supplementary surface at a slightly higher temperature (Fig. 3a). Experiments that
Video 1). Furthermore, although the scan of individual lines is fast were performed at an even higher temperature (196 °C) showed that
and thus reproduces the actual intensity profile along the horizon- a reduction of the NO2 sticking coefficient can be compensated by a
tal, fast-scan direction, the scan speed in the perpendicular direc- slightly higher NO2/H2 ratio and that it is possible to find conditions
tion (that is, from line to line) is slower. This leads to a compression under which reactivity is observed on all but the [111]-like surfaces
of waves that are running against the slow-scan direction, and an (Fig. 3c). A comparison of the spatio-temporal patterns shows that
expansion of those running against it. Slow scanning of fast dynam- the lateral spacing between the individual chemical waves on all of
ics thus results in shape distortions of travelling waves and spirals. the grains seems larger at 172 °C (Fig. 3a) than at 164 °C (Fig. 3b).
However, by scanning reduced areas with less pixels at a high scan This is indicative of increased reaction and diffusion rates at higher
rate, it is possible to capture more than 100 frames per second. This temperatures, as also confirmed by comparative experiments that
is sufficiently fast to capture rotating spirals without aliasing arte- were performed at 210 °C and 250 °C under otherwise similar con-
facts (see Supplementary Video 2). Considering that each transition ditions (see Supplementary Fig. 1). An increase of the NO2/H2 ratio,
from dark to bright contrast corresponds to a light-off and, thus, a on the other hand, results in a shortening of the refractory phase
reaction of the adsorbed layer, direct imaging of the surface reactiv- (compare Fig. 3c with Fig. 3a,b). Real-time imaging under different
ity can be used to abstract grain orientation dependent rates (that is, pressure and NO2/H2 conditions proves that the surface dynamics
kinetics by imaging)11. are very sensitive to the reaction conditions and provide a first hint
on the relationships between temperature, pressure and the type of
Spatio-temporal dynamics under steady-state conditions. After dynamics. The experiments summarized in Fig. 3 also confirm that
briefly touching the onset of reactivity and the activation of initially individual grains are active only in a restricted parametric region
inactive grains by spillover, we performed experiments at differ- (T, pNO2 , pH2 ) and that the surface reactivity strongly depends on
ent NO2/H2 ratios and temperatures in the attempt to reveal initial grainI orientation
I and therefore the structure of the platinum sur-
trends in the relationship between reaction conditions and grain- face. We could not find any stationary conditions under which the
dependent spatio-temporal dynamics. Each of the images shown in {001} and {111} surfaces were simultaneously active. However, a
Fig. 3 were recorded under steady-state conditions; that is, when systematic mapping of the phase diagram by in situ SEM is possible
no further areal spread or change in activity could be observed. and will enable detailed studies on the effect of temperature and gas
By subtracting reference images of the initially NO2-covered sur- composition at different pressures. Yet, due to the above-mentioned
face from each recorded image frame, the background originating hysteresis effects, it is advisable to reset the surface by high-tem-
from grain-dependent channelling and work-function contrast, was perature annealing between experiments that are performed under
removed. The remaining contrast variations are then only due to different conditions.
changes in the secondary-electron emission caused by the reaction We want to further point out the high sensitivity of the second-
between adsorbed species on the surface: noise-level darkness thus ary-electron signal to changes in the surface coverage. Figure 4
corresponds to NO2(ads) or O(ads)35,37,38 coverage, while the bright shows an analysis of the variation in the brightness level across
state is due to transient H2O(ads)45 or an empty surface. chemical waves and reveals the brightness difference between active
Under steady-state conditions, the reactivity of most grains is and inactive domains on a grain. The regions that were chosen here
revealed in the form of propagating chemical waves and intrigu- correspond to the above-mentioned cases in which activity gradu-
ing, often disordered, spiral patterns. As already observed during ally extends into so-far-unreactive areas of grains. As can be seen,
the initiation phase (Supplementary Video 1), the highly active chemical waves are reproduced in the secondary-electron images
{100} grains show a low average NO2 coverage and only flicker- with a saw-tooth-like change in brightness between reaction fronts:
ing contrast variations without any detectable coherent structure. a sharp transition from the dark to bright state due to the light-off is
Aside from the active areas, there are grains and domains where followed by a gradual and apparently linear decrease in brightness
the surface remains dark and non-reactive, such as in Fig. 3a,b. The during the refractory phase. The ability to detect different coverage-
presence of non-reactive areas in active grains even under station- induced contrast levels demonstrates the high surface sensitivity
ary conditions is an interesting point, they are most likely to be of the method. As discussed previously, the darker regions are

Nature Catalysis | www.nature.com/natcatal


Articles Nature Catalysis

a b

c d 7,2,9
5,4,9 3,1,10
5,0,1
7,1,11 11,3,13
18,2,4 13,2,14
2,1,12
4,3,9 11,1,12
5,1,12
9,4,12 1,0,12 1,0,2 14,2,17
3,1,4
5,4,9

4,0,5 15,1,15
12,2,14 10,1,11 3,2,9
7,1,12
12,1,17 12,1,17
11,1,12
8,2,13
111 2,1,12
9,1,10 4,3,6
112
4,0,7 4,3,10
113 5,3,12
6,1,8
001 012 101 11,1,11

Fig. 3 | Condition-dependent nonlinear dynamics. a, Platinum foil imaged under the following conditions: T = 172 °C; pNO2 : pH2  ≈ 1:10; ptot = 3.6 × 10−2 Pa.
I temperature
b, As in a, but recorded at slightly lower temperature: T = 164 °C; pNO2 : pH2  ≈ 1:10; ptot = 3.6 × 10−2 Pa. c, Further change in I and reaction
I grain
atmosphere: T = 196 °C; pNO2 : pH2  ≈ 1:8; ptot = 8.7 × 10−3 Pa. d, An EBSD I orientation map of the platinum foil, colour-coded for the crystallographic
direction perpendicular toI the surface.
I The inset of c is a colour legend of the EBSD map. Scale bar: 200 µm.

supposed to be a result of the presence of NO2(ads) or O(ads)35,37,38, such grains is only obtained once the rather stable initial layer is
whereas the brighter regions are due to transient H2O(ads)45 or an removed. Whether, and to which degree a structural response and
empty surface. We are not currently able to say anything about the reconstruction of the platimum surface is involved in the gradual
contribution of intermediate species, and further work is neces- activation of initially inactive grains needs further investigations.
sary to undoubtedly correlate the brightness variations to the sur- Due to the higher spatial resolution of SEM compared with
face species. However, the linear decrease in brightness during the LEEM or PEEM, it should be possible to obtain new insights about
refractory phase is indicative for a gradual building up of surface the role of surface defects and possible causes of destabilization that
coverage. Observations—according to which an increase in NO2 result in spontaneous wavebreaking on spiral arms and the for-
partial pressure results in a shortening of the refractory phase— mation of multiple spiral cores. As can be observed in Fig. 5 and
further corroborate this conclusion. The behaviour in the studied in Supplementary Video 3, spiral cores can either be pinned to a
regime is thus typical for a relaxation-type oscillation. Under fixed surface defect or meander across the surface of a grain during the
reaction conditions, the minimal distance between two successive reaction. This could be due to the presence of several defects that
waves is defined by the propagation speed of the reaction front are in competition with eachother, or the presence of less coherent
through the excitable surface layer and the duration of the refrac- dynamics that are sensitive to fluctuations46. Supplementary Figs. 1
tory phase. Depending on the reactivity of the surface, further and 2 also highlight the role of surface defects during the catalytic
time can pass until the next light-off is triggered, either through reaction (in Supplementary Fig. 2, a propagating wavefront scatters,
a leading centre that acts as pacemaker, or through spillover from splits and later recombines on a specific surface defect at 210 °C;
a neighbouring grain. Here the observed lateral distance between however, in Supplementary Fig. 1 the very same defect acts as lead-
chemical waves lies in the micrometre range, but is strongly ing centre and triggers the reaction at the core of a spiral when the
grain dependent and changes drastically with reaction conditions temperature is increased to 250 °C). Such observations demonstrate
(see discussion of Fig. 3). that in situ SEM will enable studies on the parameter dependence
As can be seen in Fig. 4, the brightness oscillations due to the of the transition from rigidly rotating spiral waves to meander-
reaction are offset with respect to the brightness level of the ini- ing spiral waves and more generally, investigations on the relation
tially adsorbed NO2 layer. The low brightness of the unreactive area between surface features and their role as active sites under specific
confirms the earlier conclusion, according to which a higher cover- reaction conditions.
age and more stable configuration can be established in a pure NO2 After demonstrating the high sensitivity and ability of visu-
atmosphere compared with the coverage that builds up during the alizing travelling chemical waves by in  situ SEM, the next step
refractory phase, in presence of hydrogen and possible intermediate involves observing the same reaction at higher pressures. We have
species. This interpretation fits to the observation that activity on thus conducted experiments at pressures of up to 120 Pa using the

Nature Catalysis | www.nature.com/natcatal


Nature Catalysis Articles
a b c

111

001 101

d e f

Fig. 4 | 3D Visualization of coverage-dependent brightness variations. Analysis of the secondary-electron signal intensity shows that oscillating
reactions happen in a brightness regime that is above the dark level of the initially adsorbed NO2(ads)/O(ads) layer. The activity on the grains shown
is only gradually initiated by the removal of the initial, more stable coverage. a, An area abstracted from Supplementary Video 1. b, A colourized 3D
representation of the contrast levels that were generated for the region indicated by a red square in a. c, The corresponding EBSD map from which the
(9,4,12) orientation of the selected grain was deduced. d–f, Another cut-out from the same video (d) with contrast variations (e) and the EBSD map of
that area (f). The central grain shows a (5,3,12) orientation. Scale bars: 100 µm (d), 20 µm (e).

low-vacuum mode of the ESEM and a gaseous secondary-electron t t+2s


detector (large field detector). As in previous experiments, NO2 was
first introduced into the chamber and H2 was subsequently slowly
added. The reactive behaviour that is observed once the bifurcation
point was reached is presented in Fig. 6 and Supplementary Video 4
for an experiment performed at 13 Pa. Starting from an initially
dark, fully NO2/O(ads)-covered state (Fig. 6a), the initiation of the
reaction occurs simultaneously at sparsely distributed active sites on
the surface of the sample. From there, the reaction fronts propagate
concentrically at a speed of ~1.5 µm s−1 (Fig. 6b–d). This speed is in t+4s t+6s
good agreement with the velocity of reaction–diffusion fronts that
are observed on different systems47–49. Although the reaction front
of the first light-off still propagates, readsorption of NO2 starts at
the centre of the bright circular areas as evidenced by the growing
dark areas (Fig. 6c,d). Readsorption seems to be faster than com-
pared with the reaction of the initial NO2/O(ads) layer as the lateral
spacings between the reaction fronts and the regrown dark areas
decrease with time. In agreement with the low-pressure experi-
ments, the subsequent coverages do not reach the stability of the
initial layer. This is reflected by an increased number of initiation Fig. 5 | Stationary and moving spiral cores. A series of SEM images
centres during the second light-off and fast removal of the read- showing that spiral cores can be pinned (for example, at a structural
sorbed layer, as shown in Fig. 6d,e. The surface then passes through defect) or move across the surface of an active grain. T = 161 °C;
another state of near-full coverage (Fig. 6f) and ignition from a few pNO2 : pH2  ≈ 1:10. Scale bar: 10 µm.
sparsely distributed active centres (Fig. 6e). After that, the system I I
finally switches to a dynamic state that is characterized by prop-
agating chemical waves at much smaller scales (see Fig. 6h,i and final 2D dissipative structure is reached could be a sign for a pos-
more magnified images in Fig. 6j–l). This switching and emergence sible structural modification of the surface or the accumulation and
of a new dynamical behaviour happens ~150 s after the first initia- presence of intermediate species that modulate the surface reactiv-
tion of the reaction. The propagation speed of ~0.12 µm s−1 is about ity. Overall, the observations fit to a propagating oscillating reaction
one-order-of-magnitude lower than the large-scale propagations. that occurs at the surface, phenomenologically similar to the behav-
The passage of the system through an initiation period before the iour observed at low pressure. An interesting difference concerns

Nature Catalysis | www.nature.com/natcatal


Articles Nature Catalysis

a b c d

t=0s t = 14 s t = 27 s t = 38 s

e f g h

t = 68 s t = 95 s t = 122 s t = 174 s

i j k l

t = 196 s t = 252 s t = 288 s t = 323 s

Fig. 6 | Nonlinear dynamics observed at higher pressure. a–l, The sequence shows the emergence of a structured pattern with a typical length scale of
a few micrometres from larger scale propagating waves. Conditions of acquisition: T = 180 °C; Ptot = 13 Pa. The red square in panel i corresponds to the
magnified regions presented in panels j–l. Scale bars: 100 µm (a–i); 20 µm (j–l).

the brightness change during the refractory phase: instead of a grad- adsorption and desorption processes through the associated
ual darkening, the change from the bright to dark state is relatively modulation of the secondary-electron signal intensity. Due to its
abrupt, indicating much faster readsorption at higher pressures. versatility and ability to work across a large range of pressures,
This is reasonable and a possible cause for the different dynamics50. in situ SEM extends the capability of LEEM and PEEM and thus
The results prove the robustness of the NO2 + H2 system, and the enables a further narrowing of the materials and pressure gap.
ability to detect reaction dynamics also with the gaseous second- A large range of magnifications as well as seamless zoom from
ary-electron detector. Such experiments are needed to correlate overview imaging of millimetre-sized areas to high-resolution
the behaviour of the reaction over several orders of magnitude in imaging at nanometre resolution allows context-embedded
pressure, and thus, for a narrowing of the pressure gap. observation across scales and the study of short- and long- range
We have not observed any signs of subsurface oxygen or struc- coupling effects in a very effective way.
tural/morphological modifications of the catalyst surface during The observations made here for the case of NO2 hydrogenation
the reaction thus far, we are therefore not able to draw any conclu- show how real-time imaging in combination with EBSD mapping
sions about adsorbate- or reaction-induced surface reconstructions. enables a direct assessment of the structure–reactivity correlation
With the ESEM it is possible to perform experiments at pressures on a working catalyst. Surface reactivity is detected in the form
of up to 2,000 Pa and, with some modifications, even up to 1 bar of chemical waves that propagate across the surface of individual
(ref. 20). We tested the observability of surface reactivity up to 120 Pa grains of the polycrystalline platinum catalyst. Propagating reac-
(see Supplementary Fig. 3). If reactivity can still be visualized at tion fronts—in combination with the oscillatory behaviour of the
higher pressures, we might be able to study reaction-induced sur- reaction—give rise to complex spatio-temporal dynamics with
face modifications that get more substantial at higher chemical intriguing spiral patterns. Individual reaction fronts are sepa-
potential of the gas phase16,51,52. rated by a refractory phase; during this phase, the surface cover-
Future work will therefore be focused on mapping the kinetic age gradually builds up until the next light-off can be triggered.
phase diagram and the detection of the structural response of the The sensitivity of the secondary-electron signal is sufficient to
platinum surface, as well as on finding possible effects related to a detect variations in coverage and differentiate between contrast
coupling of the dynamics through the gas phase. levels that are associated with more or less stable configurations of
the adsorbate layer. Direct imaging of propagating reaction fronts
Discussion reveals how spillover across grain boundaries can induce activity
In this work, we have demonstrated the capability of in situ SEM on so-far-non-reactive grains. It furthermore shows that scatter-
as a surface science tool that is sufficiently sensitive to image ing of chemical waves at grain boundaries and their propagation

Nature Catalysis | www.nature.com/natcatal


Nature Catalysis Articles
along chemical waveguides can transmit the effect of spillover 6. Takayanagi, K. High resolution surface study by in-situ UHV transmission
across extended distances, thus highlighting the important role electron microscopy. Ultramicroscopy 8, 145–161 (1982).
7. Thomas, J. M. Reflections on the value of electron microscopy in the study of
of spillover for the activity of the whole system. Experiments per- heterogeneous catalysts. Proc. R. Soc. A 473, 20160714 (2017).
formed at different temperature, pressure and NO2/H2 ratios illus- 8. Vendelbo, S. B. et al. Visualization of oscillatory behaviour of Pt nanoparticles
trate the high sensitivity and strong dependence of the reaction catalysing CO oxidation. Nat. Mater. 13, 884–890 (2014).
dynamics on the surface structure and reaction conditions. Similar 9. Nettesheim, S., von Oertzen, A., Rotermund, H. H. & Ertl, G. Reaction
phenomena related to propagating chemical waves and oscillatory diffusion patterns in the catalytic CO-oxidation on Pt(110): front propagation
and spiral waves. J. Chem. Phys. 98, 9977–9985 (1993).
behaviour were observed over five orders of magnitude in pressure, 10. von Boehn, B. & Imbihl, R. Large amplitude excitations traveling along the
demonstrating the robustness of the system and the importance of interface in bistable catalytic methanol oxidation on Rh(110). Phys. Chem.
collective behaviours in the description of the catalytic reactivity of Chem. Phys. 19, 18487–18493 (2017).
surfaces. Nonlinear dynamics are therefore not confined to model 11. Suchorski, Y. & Rupprechter, G. Local reaction kinetics by imaging. Surf. Sci.
surface science studies that are performed under UHV conditions, 643, 52–58 (2016).
12. Gorodetskii, V. V., Block, J. H. & Drachsel, W. Isothermal oscillations of the
but are rather an integral part of catalytic processes and should be hydrogen-oxidation reaction on platinum: investigations in the field electron
accounted for in their description. and field ion microscope. Appl. Surf. Sci. 76–77, 129–135 (1994).
As in situ SEM is rather unconstrained regarding the shape com- 13. Visart de Bocarmé, T. & Kruse, N. Field emission techniques for studying
plexity of a sample and furthermore suitable also for the study of surface reactions: applying them to NO–H2 interaction with Pd tips.
non-conductive samples, this work can be extended to the study Ultramicroscopy 111, 376–380 (2011).
14. Franz, T. et al. Catalytic CO oxidation on Pt under near ambient pressure:
of surface reactivity, spillover and the effect of interfaces on sup- a NAP-LEEM study. Ultramicroscopy 200, 73–78 (2019).
ported metal catalysts. Finally, this work demonstrates how much 15. Sachs, C., Hildebrand, M., Völkening, S., Wintterlin, J. & Ertl, G.
more information one can get from SEM when working under clean Spatiotemporal self-organization in a surface reaction: from the atomic to the
conditions. Regardless of the impressive development of the instru- mesoscopic scale. Science 293, 1635–1638 (2001).
ments, most commercially available scanning electron microscopes 16. van Spronsen, M. A., Frenken, J. W. M. & Groot, I. M. N. Observing the
oxidation of platinum. Nat. Commun. 8, 429 (2017).
still provide relatively poor vacuum levels, preventing observation 17. Dicke, J., Rotermund, H.-H. & Lauterbach, J. Ellipsomicroscopy for surface
of clean and defined surfaces (the same is true for most transmis- imaging: contrast mechanism, enhancement, and application to CO oxidation
sion electron microscopes). In view of the development of the on Pt(110). J. Opt. Soc. Am. A 17, 135–141 (2000).
method, the realization of a UHV-capable environmental SEM— 18. Danilatos, G. D. Review and outline of environmental SEM at present.
such as already proposed by Banbury in 197253—should receive J. Microsc. 162, 391–402 (1990).
19. Millar, G. J., Nelson, M. L. & Uwins, P. J. R. In situ imaging of catalytic
some priority. etching on silver during methanol oxidation conditions by environmental
scanning electron microscopy. J. Catal. 169, 143–156 (1997).
Methods 20. Fitzek, H., Schroettner, H., Wagner, J., Hofer, F. & Rattenberger, J. High-
In situ experiments were performed in a commercial ESEM (FEI Quanta 200F) quality imaging in environmental scanning electron microscopy—optimizing
that is equipped with an oil-free vacuum system, a home-built heating stage and a the pressure limiting system and the secondary electron detection of a
gas feeding unit. Platinum foils and single crystal samples were cleaned ex situ and commercially available ESEM. J. Microsc. 262, 85–91 (2016).
then mounted in the ESEM, where they were first exposed to alternating annealing 21. Greiner, M. T. et al. The oxidation of copper catalysts during ethylene
in hydrogen at 1,000 °C and oxygen at around 800 °C to remove surface impurities epoxidation. Phys. Chem. Chem. Phys. 17, 25073–25089 (2015).
and finally, a long-term annealing in ~10 Pa of H2 at 1,000 °C for several hours. 22. Poitel, S., Wang, Z.-J., Willinger, M.-G. & Hébert, C. In-situ observation of
After annealing, the temperature was decreased and the clean samples exposed to Co–Ce coated metallic interconnect oxidation combined with high-resolution
the reactive gas mixtures. The methodology used was to set a working temperature, post exposure. Anal. ECS Trans. 78, 1615–1632 (2017).
introduce a given amount of NO2 gas into the ESEM chamber and subsequently 23. Hieke, S. W. et al. On pinning–depinning and microkink-flow in solid state
introduce H2 until the emergence of nonlinear behaviours could be observed. The dewetting: insights by in-situ ESEM on Al thin films. Acta Mater. 165,
base pressure of the system lies in the ~10−5 Pa range and observations of surface 153–163 (2019).
dynamics were observed at pressures ranging from 10−4 to 120 Pa. Secondary- 24. Huang, X., Wang, Z.-J., Weinberg, G., Meng, W.-M. & Willinger, M.-G. In situ
electron images were recorded using the Everhart–Thornley detector at pressures scanning electron microscopy observation of growth kinetics and catalyst
up to ~2 × 10−2 Pa and with a large field gaseous secondary-electron detector splitting in vapor–liquid–solid growth of nanowires. Adv. Funct. Mater. 25,
at pressures above 1 Pa. Once surface reactions were observed, all the control 5979–5987 (2015).
parameters (T, pNO2 , pH2 ) and imaging conditions were kept constant to ensure 25. Huang, X. et al. In situ formation of crystallographically oriented
that the variations
I of brightness
I only reflect variations of the surface composition. semiconductor nanowire arrays via selective vaporization for optoelectronic
Surface grain orientation maps were recorded using a FEI Quanta 200F ESEM applications. Adv. Mater. 28, 7603–7612 (2016).
that is equipped with a Hikari EBSD camera. 26. Wang, Z.-J. et al. Direct observation of graphene growth and associated
copper substrate dynamics by in situ scanning electron microscopy.
Data availability ACS Nano 9, 1506–1519 (2015).
All recorded images and data that support the findings of this study are available 27. Blume, R. et al. The influence of intercalated oxygen on the properties of
from the corresponding author on reasonable request. graphene on polycrystalline Cu under various environmental conditions.
Phys. Chem. Chem. Phys. 16, 25989–26003 (2014).
28. Wang, Z.-J. et al. Stacking sequence and interlayer coupling in few-layer
Received: 16 March 2019; Accepted: 1 November 2019;
graphene revealed by in situ imaging. Nat. Commun. 7, 13256 (2016).
Published: xx xx xxxx 29. Voss, C. & Kruse, N. Oscillatory behavior in the catalytic reduction of NO
and NO2 with hydrogen on Pt field emitter tips. Appl. Surf. Sci. 94-95,
References 186–193 (1996).
1. Kuwauchi, Y., Yoshida, H., Akita, T., Haruta, M. & Takeda, S. Intrinsic 30. McEwen, J.-S. et al. Catalytic reduction of NO2 with hydrogen on
catalytic structure of gold nanoparticles supported on TiO2. Angew. Chem. Pt field emitter tips: kinetic instabilities on the nanoscale. Langmuir 26,
Int. Ed. 51, 7729–7733 (2012). 16381–16391 (2010).
2. Roobol, S. B. et al. The reactor AFM: non-contact atomic force microscope 31. Barroo, C., De Decker, Y., Visart de Bocarmé, T. & Kruse, N. Complex
operating under high-pressure and high-temperature catalytic conditions. oscillations patterns during the catalytic hydrogenation of NO2 over platinum
Rev. Sci. Instrum. 86, 033706 (2015). nano-sized crystals. J. Phys. Chem. C. 118, 6839–6846 (2014).
3. Hendriksen, B. L. M. et al. The role of steps in surface catalysis and reaction 32. Barroo, C., De Decker, Y., Visart de Bocarmé, T. & Gaspard, P. Fluctuating
oscillations. Nat. Chem. 2, 730–734 (2010). dynamics of nanoscale chemical oscillations: theory and experiments.
4. Mom, R. V. et al. Simultaneous scanning tunneling microscopy and J. Phys. Chem. Lett. 6, 2189–2193 (2015).
synchrotron X-ray measurements in a gas environment. Ultramicroscopy 182, 33. Barroo, C., De Decker, Y., Visart de Bocarmé, T. & Kruse, N. Emergence of
233–242 (2017). chemical oscillations from nanosized target patterns. Phys. Rev. Lett. 117,
5. Genty, E., Jacobs, L., Visart de Bocarmé, T. & Barroo, C. Dynamic processes 144501 (2016).
on gold-based catalysts followed by environmental microscopies. Catalysts 7, 34. Segner, J., Vielhaber, W. & Ertl, G. Interaction of NO2 with a Pt(111) Surface.
134 (2017). Isr. J. Chem. 22, 375–379 (1982).

Nature Catalysis | www.nature.com/natcatal


Articles Nature Catalysis
35. Parker, D. H., Bartram, M. E. & Koel, B. E. Study of high coverages of atomic 50. Panfilov, A. & Hogeweg, P. Spiral breakup in a modified FitzHugh–Nagumo
oxygen on the Pt(111) surface. Surf. Sci. 217, 489–510 (1989). model. Phys. Lett. A 176, 295–299 (1993).
36. Barroo, C., Voorsluijs, V., Visart de Bocarmé, T., Gaspard, P. & De Decker, Y. 51. Fantauzzi, D., Mueller, J. E., Sabo, L., van Duin, A. C. T. & Jacob, T. Surface
Reconstructing stochastic attractors from nanoscale experiments on a buckling and subsurface oxygen: atomistic insights into the surface oxidation
non-equilibrium reaction. Phys. Chem. Chem. Phys. 20, 21302–21312 (2018). of Pt(111). ChemPhysChem 16, 2797–2802 (2015).
37. Lewis, R. & Gomer, R. Adsorption of oxygen on platinum. Surf. Sci. 12, 52. Rotermund, H. H., Pollman, M. & Kevrekidis, I. G. Pattern formation during
157–176 (1968). the CO-oxidation involving subsurface oxygen. Chaos 12, 157–163 (2002).
38. Dawson, P. T. & Peng, Y. K. The adsorption, desorption, and exchange 53. Banbury, J. R. A versatile ultrahigh vacuum scanning electron microscope.
reactions of oxygen, hydrogen, and water on platinum surfaces: IV. Field J. Phys. E 5, 798–802 (1972).
emission studies on the adsorption of water, hydrogen and the reaction
between hydrogen and adsorbed oxygen. Surf. Sci. 92, 1–13 (1980). Acknowledgements
39. Kiskinova, M., Pirug, G. & Bonzel, H. P. NO adsorption on Pt(111). The authors want to acknowledge Dr Karsten Kunze for recording the EBSD map and
Surf. Sci. 136, 285–295 (1984). ScopeM for the use of the equipment. C.B. thanks the Fonds de la Recherche Scientifique
40. Bartram, M. E., Windham, R. G. & Koel, B. E. The molecular adsorption of (F.R.S.-FNRS) and Wallonie–Bruxelles International (Excellence grant WBI.WORLD)
nitrogen dioxide on Pt(111) studied by temperature programmed desorption for financial support.
and vibrational spectroscopy. Surf. Sci. 184, 57–74 (1987).
41. Huang, W. et al. Decomposition of NO2 on Pt(110): formation of a new
oxygen adsorption state. Surf. Sci. 506, L287–L292 (2002). Author contributions
42. Danilatos, G. D. Theory of the gaseous detector device in the environmental C.B., Z.-J.W. and M.-G.W. designed and performed the experiments and analysed the
scanning electron microscope. Adv. Electron. Electron Phys. 78, 1–102 (1990). data. Z.-J.W. and M.-G.W. modified the ESEM set-up for in situ observations of gas–solid
43. Graham, M. D. et al. Effects of boundaries on pattern formation: catalytic interactions. All authors contributed to the writing of the manuscript, and all authors
oxidation of CO on platinum. Science 264, 80–82 (1994). have given approval to the final version.
44. Asakura, K., Lauterbach, J., Rotermund, H. H. & Ertl, G. Spatiotemporal
concentration patterns associated with the catalytic oxidation of CO and Competing interests
Au Covered Pt(110) Surfaces. J. Chem. Phys. 102, 8175–8184 (1995). The authors declare no competing interests.
45. Heras, J. M. & Viscido, L. Work function changes upon water contamination
of metal surfaces. Appl. Surf. Sci. 4, 238–241 (1980).
46. Karma, A. Meandering transition in two-dimensional excitable media. Additional information
Phys. Rev. Lett. 65, 2824 (1990). Supplementary information is available for this paper at https://doi.org/10.1038/
47. Suchorski, Y. et al. Visualizing catalyst heterogeneity by a multifrequential s41929-019-0395-3.
oscillating reaction. Nat. Commun. 9, 600 (2018). Correspondence and requests for materials should be addressed to C.B., Z.-J.W.
48. Cox, M. P., Ertl, G. & Imbihl, R. Spatial self-organization of surface structure or M.-G.W.
during an oscillating catalytic reaction. Phys. Rev. Lett. 54, 1725–1728 (1985). Reprints and permissions information is available at www.nature.com/reprints.
49. Jakubith, S., Rotermund, H.-H., Engel, W., von Oertzen, A. & Ertl, G.
Spatiotemporal concentration patterns in a surface reaction: propagating Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
and standing waves, rotating spirals, and turbulence. Phys. Rev. Lett. 65, published maps and institutional affiliations.
3013 (1990). © The Author(s), under exclusive licence to Springer Nature Limited 2019

Nature Catalysis | www.nature.com/natcatal

You might also like