XIN ET AL.
’ NO. 5
April 11, 2012
C2012 American Chemical Society
Revealing Correlation of Valence State
with Nanoporous Structure in Cobalt
Catalyst Nanoparticles by In Situ
Huolin L. Xin,†Elzbieta A. Pach,†Rosa E. Diaz,‡Eric A. Stach,‡Miquel Salmeron,†and Haimei Zheng†,*
†Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, United States, and‡Center for Functional Nanomaterials,
Brookhaven National Laboratory, Upton, New York 11973, United States
of materials can be studied in situ at the
nanometer or atomic scale while chemical
reactions are proceeding. For example,
provides electronic structures of ensembles
of catalyst nanoparticles during catalytic
tron microscopy (TEM) has been developed
to image single nanoparticle growth dy-
gas reactions in real time under relevant
catalytic conditions.4However, there have
which is key in revealing the correlation
betweenthe valence state and morphology
changes of active-metal-containing nano-
particles, that is, during heterogeneous cata-
lysis. With the development of gas environ-
mental TEM5and electron energy loss spec-
ticles can be measured concurrently with
atomic-resolution imaging during the reac-
we study the valence state of nanoporous
cobalt-containing particles and its correlation
with the structural coarsening during the
hydrogen reduction reaction for Fischer?
Tropsch (F?T) synthesis;an industrial reac-
gen and carbon monoxide) to liquid fuels.6,7
The initial reduction of oxide nanoparti-
clesinto metalliccatalysts isacritical step in
F?T synthesis. Among the various active
metal catalysts (Ni, Co, Fe, and Ru), iron
and cobalt are the only catalysts that are
used in commercial F?T reactors, as they
exhibit both low cost and high selectivity,
lyst preparation is the hydrogen reduction
ccompanying recent accomplish-
ments in nanocharacterization, the
electronic structure or morphology
with cobalt being preferred for the synthesis
of heavy hydrocarbons such as jet and diesel
fuels.8?10The preparation and conditioning
of the microstructure and valence state of
the catalyst are essential for achieving the
required durability and catalytic activity.
Because cobalt oxides are typically used
for catalyst formation, a crucial step of cata-
*Address correspondence to
Received for review February 21, 2012
and accepted April 11, 2012.
structure and morphology of
atomic scale while a chemical
for understanding the under-
lying reaction mechanisms
and optimizing a materials
design. This is especially im-
portant in the study of nano-
particle catalysts, yet such experiments have rarely been achieved. Utilizing an environmental
transmission electron microscope equipped with a differentially pumped gas cell, we are able to
reaction conditions. Studies reveal quantitative correlation of the cobalt valence states with the
particles' nanoporous structures. The in situ experiments were performed on nanoporous cobalt
particles coated with silica, while a 15 mTorr hydrogen environment was maintained at various
temperatures (300?600 ?C). When the nanoporous particles were reduced, the valence state
changed from cobalt oxide to metallic cobalt and concurrent structural coarsening was observed.
In situ mapping of the valence state and the corresponding nanoporous structures allows
quantitative analysis necessary for understanding and improving the mass activity and lifetime of
cobalt-based catalysts, for example,forFischer?Tropschsynthesisthatconverts carbonmonoxide
and hydrogen into fuels, and uncovering the catalyst optimization mechanisms.
KEYWORDS: environmental TEM.in situ TEM.cobalt catalysts.
porosity control.Fischer?Tropsch synthesis
XIN ET AL.
’ NO. 5
of oxides at high temperatures. However, during this
process, sintering of the catalyst particles reduces the
to form self-assembled composites.11Previous studies
have shown that partially reduced F?T catalysts can be
as active as a fully reduced one.8,12Therefore, an under-
the underlying nanoporous structure of the catalysts is
critical to yield an optimized catalyst. In this work, we use
the state-of-the-art aberration-corrected environmental
transmission electron microscopy (ETEM) and electron
energy loss spectroscopy (EELS) (see Materials and
Methods for details) to make this connection between
nanoporous structure and the valence state of cobalt/
silica catalysts during hydrogen reduction. Such in situ
correlation provides insights for future optimization of
F?T catalysts, in particular, and more broadly yields
insights into porosity control in the general class of
RESULTS AND DISCUSSION
prepared CoOx/SiO2 nanocomposites dispersed on
a carbon grid. On the basis of the image contrast
and spatially resolved EELS spectra (Figure 1c,d and
FigureS1in SupportingInformation),wecan identify
the cobalt-oxide core and the silica shell. We fur-
ther found that the majority of the core material is
cobalt monoxide with a trace amount of Co3O4
(Figures S2?S4). The spatial variation in the bright-
field image contrast within each nanocomposite par-
likely porous, resulting from the sintering of multiple
CoO crystalline nanoparticles. As shown in Figure S2,
individual particles in the sintered core are separated
by low-angle grain boundaries. However, projection
difficult to distinguish a porous network from a corru-
gated solid structure in a single projection image. In
addition, the observed low-spatial-frequency contrast
modulation can also be a result of diffraction contrast
caused by strain fields. To reliably visualize the 3-D
internal structure without ambiguity, we used annular
dark-field electron tomography.14?16We recorded 73
annular dark-field scanning transmission electron mi-
croscopy (ADF-STEM) images of a single CoOx/SiO2
nanoporous particle from ?72 to 72? with 2? intervals
(see Materials and Methods for more details). The 3-D
structure of the material was reconstructed using the
Figure 1. Morphology and composition of as-prepared CoOx/SiO2nanocomposites. (a) Bright-field TEM image of the
dashed line. (d) Extracted silicon and cobalt concentrations along the line profile in (e). The maximum intensity was
XIN ET AL.
’ NO. 5
simultaneous iterative reconstruction technique (SIRT).14
Figure 2 shows the 3-D rendering of the high-Z CoOx
component in the nanocomposite using isosurfaces
and progressive cross sections. These results clearly
demonstrate in 3-D that the cobalt-oxide core is an
interconnected nanoporous network. It is expected
that this porous structure can facilitate the infiltration
of gas molecules at reaction conditions. Figure 3a
shows the heating and the hydrogen injection trajec-
tories used in in situ experiments (see more details in
Figure S6). First, the sample was heated to 300 ?C in
vacuum and was maintained at the temperature for
an hour to minimize thermal drift of the sample. No
significant changes of the nanoporous core structure
were observed during this stage (Figure 3b(I) and
Figure S7(I)). Subsequently, 15 mTorr H2was injected
into the environmental cell, leading to noticeable
changes of the nanoporous core (Figure 3b(II) and
Figure S7(II)). The structural changes became more
evident when the temperature was raised to 450 ?C
(0.25 h). As shown in Figure 2b(III) and Figure S7(III),
the nanoporous structures coarsened significantly,
although some degree of porosity was still preserved.
Upon increasing the temperature to 500 ?C (1.5 h),
morphological changes lead to particles which have
limited porosity (Figure 3b(IV) and Figure S7(IV)). At
surfaces in silica pockets were observed. During the
sintering process, silica clearly provides a certain
degree of protection for the cobalt cores from inter-
pocket sintering. However, in places with high-density
nanocompsite aggregates, silica from different pockets
can “glue” together to form a micrometer scale compo-
of the core to form a core?shell structure.
Figure 2. Three-dimensional tomographic reconstruction of an as-prepared CoOx/SiO2nanocomposite. The progressing
cross sections and the isosurfaces visualize the internal structures of the porous CoOxcore.
Figure3. In situ observation of structural changes of Co/CoOxcatalysts at different reduction conditions. (a)Heating and
A detailed trajectory is shown in Figure S6.) (b) In situ TEM images of the nanocomposites under the correspond-
ing environmental treatments. Images of large field of view are shown in Supporting Information Figure S8. Scale bars
are 10 nm.
XIN ET AL.
’ NO. 5
To quantify the temperature-dependent structural
evolution during hydrogen reduction, we measured
the effective particle size of the porous network at
different temperatures. Here, the effective particle size
is the average internal feature size inside the core,
which describes a porous network in terms of equiva-
the effective particle size reflects coarsening of the
network and a decrease of the surface-to-volume ratio
at higher reduction temperatures. It demonstrates the
evolution of the core structure from an initial nanopor-
ous network toward solid spheres.
To correlate the structural coarsening with the frac-
of cobalt. Figure 4b shows the near-edge fine struc-
tures of Co L2,3edges (2p63dnf 2p53dnþ1transitions)
recordedat fourconditions(300?Cþ Vacuum,300?Cþ
mostly CoO, the 300 ?C þ Vacuum spectrum can be
approximately assigned as a Co“2þ”fingerprint. Simi-
larly, the 600 ?C þ H2spectrum can be considered ap-
proximately fully reduced, that is, Co“0”fingerprint, since
selected area diffraction (SAD) pattern after reduction
at 600 ?C (Figure S8). The L2,3near-edge fine structures
of cobalt with an average valence state between
can be decomposed into a linear combination of the
two fingerprints.17The corresponding decomposition
coefficient of the Co“0”component reflects the reduc-
tion fraction. Figure 4c shows such decompositions as
a function of reduction temperature. We see that, at
300 ?C þ H2, 40 ( 9% of the material is reduced. At
500 ?C, only a small amount of residual cobalt oxide
(5.5 ( 3%) is present. In addition to demonstrating a
general correlation between the increased reduction
the first direct quantification between morphological
changes and changes in electronic structures. We
further calculated the optimum reduction conditions,
based on the evolution of the surface-to-volume ratio
and the reduction fraction, presented in Figure 4a?c.
This shows that the optimum reduction temperature is
within 390?440 ?C in these experiments (Figure 4d),
which is consistent with the optimal reduction tem-
perature used in commercial F?T plants (∼400 ?C8).
This study reveals the underlying mechanisms quanti-
tatively for the first time by in situ TEM.
It is noted that electron beams can cause knock-on
damage, local heating, and induced coalescence. For
Figure 4. Correlation of valence state with coarsening (via effective particle size) in nanoporous Co/silica catalysts during H2
reduction. (a) Effective particle size as a function of the reduction conditions. (b) Co L2,3edges of the catalysts, determined
from in situ measurements. (c) Reduction fraction as a function of reduction conditions. The dashed lines mark 68%
and (b), and the projected optimum condition.
XIN ET AL.
’ NO. 5
this reason, we have used a nearly parallel beam with
relative low intensity and avoided constant illumina-
tion of the materials by closing the gun valve. Most
to the electron beam show a similar morphology to
that of particles exposed to the beam (Figure S9). This
demonstrates that there are minimal electron beam
modifications in the in situ studies.
After full reduction of cobalt at 600 ?C þ H2, we
switched off the heating and stopped hydrogen flow
after the sample had cooled to room temperature. The
environmental cell was then pumped down to 10?6
Torr, and the reduced sample was left in the cell
overnight. After 14 h, we reinvestigated the reduced
products. We found most of the particles were intact
(Figure S10), but the particles with exposed surfaces
had formed a layer of native oxide (Figure 5a,b).
However, at room temperature, those particles with a
2?4 nm thick silica shell showed a high resistance to
oxidation at ambient conditions. Figure 5 shows ex situ
in situ reduction experiment (also see Figure S10). The
atomic-resolution images suggest that the surface is
of valence state at the catalyst surfaces using atomic-
scale STEM-EELS. It is well-known that the intensity
ratio between the L2and L3edges is sensitive to the
oxidation state of the metal.18?21However, the
traditional L3/L2method requires spectra with rela-
tively good signal-to-noise ratio. To visualize the
valence change of cobalt more reliably at curved
surfaces, we directly integrate the L2and L3intensity
on the background subtracted spectrum (Figure 5d).
To validate our method, we first applied it to the
particle with an oxidized surface, shown in Figures 5b
and S12a. Figure 5e(II) shows the L3/L2ratio as the
electron probe scans from the native oxide layer tothe
metallic part of the particle. As expected, the oxide
layer shows a significantly higher L3/L2ratio than that
of the metallic part. In Figure 5f, we use this method to
probe the particle shown in Figure S12b. The L3/L2line
profile shows there is no statistical significant valence
change from the surface to the center of the particle.
This indicates the silica layer can readily allow gas
molecules such as H2to penetrate through at elevated
blocker at room temperature, protecting cobalt from
oxidation at ambient conditions, even for extended
periods of time.
In conclusion, we performed in situ environmental
TEM study of nanoporous cobalt/silica catalysts rele-
vant to Fischer?Tropsch synthesis. We have demon-
strated quantitatively that H2reduction of the CoOx
into metallic cobalt results in an increase of an effective
arrow. (a) Particle is fully metallic at 600 ?C þ H2. (b) Exposed surface is oxidized 14 h after reduction at room temperature in
room temperature for more than 3 weeks. (d) EELS spectral setup for L3/L2ratio calculation. (e,f) Ex situ measurements of the
L3/L2ratio of across (e) the exposed surface and (f) the silica-coated surface (see Supporting Information Figure S12 for
the ADF-STEM images and the line profile positions). I: Co concentration normalized by the maximum intensity. II: measured
are 4 nm.
XIN ET AL.
’ NO. 5
particle size of the porous structure. An optimum reduc-
tion temperature of about 410 ?C was achieved at the
current reduction environment. Thecorrelation between
the valence state and the structural changes provided
here may be proven to be significant for the design of
catalysts with improved catalytic activity and selectivity.
MATERIALS AND METHODS
Catalytic Nanoparticle Synthesis. The CoOx/Silica nanocompo-
sites were synthesized using a two-step method.13First, Co
nanoparticles (10?12 nm) were prepared using air-free colloi-
self-assembled Co clusters by injection of tetramethylorthosili-
cate (TMOS) and octadecyltrimethoxysilane (C18TMS) into the
Environmental TEM Experiment Procedure. The as-prepared
CoOx/SiO2was dispersed in isopropyl alcohol and spread on a
nonporous amorphous silicon TEM membrane (SIMpore, 5 nm
amorphous Si with 100 μm windows) by micropipetting. The
imaging, and chemical analysis were performed using an FEI
environmental cell Titan 80/300 equipped with apostspecimen
aberration corrector (CEOS) operated at300 kV. This instrument
on the molecular weight of the gas. In the case of hydrogen, a
maximum pressure of 1 Torr can be achieved. The gas pressure
is controlled manually by a simple needle valve and is mon-
itored by a high-accuracy pressure transducer. The ultrahigh
purity [99.9999%] hydrogen used in this experiment was sup-
plied through stainless steel tubes. Prior to the in situ experi-
ment, aberration coefficients were measured using a Zemlin
tableau and corrected until a quarter wavelength semiangle
larger than 20 mrad was achieved. Electron energy loss spectra
were recorded with a Tridiem Gatan imaging filter with an
energy dispersion of 0.3 eV/channel. The EELS spectra were
acquired in selected area TEM imaging mode (10 μm selected
area aperture) with an energy resolution of ∼2 eV. Following
each core-loss spectrum acquisition, the spectrum of the elas-
tically scattered electrons (zero-loss peak)was alsorecorded for
type heating holder was used for in situ heating. The heating
temperature was monitored by a thermocouple attached directly
to the furnace. The heating trajectory of the experiment was
plotted in red in Figure S6. Gas injection pressure trajectory was
plotted in blue in Figure S6. Gas pressure was maintained at 15
mTorr exceptfora short period at 300 ?C,wherethe pressure was
raised to ∼6.5 Torr. The gun valve was closed during this process.
Microscope. All in situ imaging and spectroscopy were per-
formed using a 300 kV image aberration-corrected environ-
mental TEM at Brookhaven National Lab. Ex situ scanning/
200 kV FEI Tecnai at National Center for Electron Microscopy of
Lawrence Berkeley National Lab (LBNL). Some high-resolution
TEM images were taken using a 300 kV image aberration-
corrected environmental TEM at Brookhaven National Lab
(Figure 1b and Supporting Information Figures S2, S7, S9e,
S10a, and S11a). Part of the bright-field TEM images and SAD
patterns were recorded using a 200 kV JEOL 2100 at Materials
Sciences Division of LBNL (Figure 1a and Figures S8 and S11b).
Conditions for annular dark-field STEM imaging in Tecnai are
Schottky field emission gun, 200 kV, 11 mrad semiconvergence
angle, and 26.4?100 mrad collection semiangles. The condi-
tions for EELS in LBNL Tecnai are 11 mrad probe-forming
semiangle, 0?22 mrad collection semiangles, and 0.5?1.5 eV
are LaB6 cathode, 200 kV, and 1.8 Å information transfer. The
conditions for imaging in BNL ETEM are Schottky field emission
gun, 300 kV, quarterwavelength semiangle >20 mrad, and
information transfer up to subangstrom.
Annular Dark-Field Scanning Transmission Electron Tomography. An-
nular dark-field STEM (ADF-STEM) images were recorded
from ?72 to 72? at 2? intervals. Beam current was around 10 pA
acquired for 20 s. No mass loss was observed during the image
acquisition process. The theoretical resolution for this reconstruc-
tion is ∼2.2 nm. However, the resolution can be improved using
the simultaneous iterative reconstruction algorithm.
The acquired tilt series was first aligned using the center of
mass. The fine adjustment was made manually using a Matlab
script package (e?Tomo) written by Robert Hovden at Cornell
University. The 3-D data set was reconstructed by the simulta-
neous iterative reconstruction algorithm implemented in Matlab.
The script was initially written by one of the authors (H.L.X.). It
was modified and integrated into the e?Tomo package by
Robert Hovden. Twenty-five iterations were used for the final
Due to the limited tilt range (?72 to 72?), a wedge of
information is missing in reciprocal space of the reconstruction.
This results in an elongation of the reconstructed features by a
factor 1.27 along the beam incident direction.
Co Valence Determination. The fitting method used in Figure 4
is only accurate when electromagnetic optical condictions are
the dispersion and linearity of the spectrometer varies from
time to time. In Figure 4, the results are self-referenced where
the spectra from the two ends of the reaction are used as
reference spectra as we know the materials can only be
reduced. This method gives an accurate measurement of
the reduction fraction. However, in Figure 5d, we do not have
oxidized references taken at the same condition, and therefore,
the L3/L2method was used.
Conflict of Interest: The authors declare no competing
Acknowledgment. This work was supported by the Officeof
Basic Energy Sciences, Chemical Science Division of the U.S.
DOE under Contrast No. DE-AC02-05CH11231. The in situ en-
Functional Nanomaterials, Brookhaven National Laboratory,
which is supported by the U.S. Department of Energy, Office
of Basic Energy Sciences, under Contract No. DE-AC02-
98CH10886. We performed ex situ TEM experiments at National
Center for Electron Microscopy (NCEM) of the Lawrence Berke-
ley National Laboratory (LBNL), which is supported by the U.S.
Department of Energy (DOE) under Contract No. DE-AC02-
05CH11231. E.A.P. thanks Trevor Ewers and Prof. Paul Alivisatos
for providing guidance and the access to the synthesis laboratory.
H.L.X. thanks Peter Ercius for helping with the tomography setup
and Robert Hovden for the development of the Cornell e?Tomo
Office of Science Early Career Research Program.
Supporting Information Available: Supplementary figures.
This material is available free of charge via the Internet at
REFERENCES AND NOTES
1. Morales, F.; de Groot, F. M. F.; Glatzel, P.; Kleimenov, E.;
Bluhm, H.; Hävecker, M.; Knop-Gericke, A.; Weckhuysen, B. M.
In Situ X-ray Absorption of Co/Mn/TiO2Catalysts for Fischer?
Tropsch Synthesis. J. Phys. Chem. B 2004, 108, 16201–16207.
2. Herranz, T.; Deng, X.; Cabot, A.; Guo, J.; Salmeron, M.
Influence of the Cobalt Particle Size in the Co Hydrogena-
tion Reaction Studied by In Situ X-ray Absorption Spec-
troscopy. J. Phys. Chem. B 2009, 113, 10721–10727.
XIN ET AL. Download full-text
’ NO. 5
3. Zheng, H.; Smith, R. K.; Jun, Y.-w.; Kisielowski, C.; Dahmen,
Nanocrystal GrowthTrajectories.Science 2009,324,1309–
Clausen, B. S.; Topsøe, H. Atom-Resolved Imaging of
Dynamic Shape Changesin Supported Copper Nanocryst-
als. Science 2002, 295, 2053–2055.
5. Sharma, R. Design and Applications of Environmental Cell
Transmission Electron Microscope for In Situ Observations
of Gas?Solid Reactions. Microsc. Microanal. 2001, 7, 494–
Today 2002, 71, 227–241.
7. Huber, G. W.; Iborra, S.; Corma, A. Synthesis of Transporta-
tion Fuels from Biomass: Chemistry, Catalysts, and Engi-
neering. Chem. Rev. 2006, 106, 4044–4098.
8. Dry, M. E. Fischer?Tropsch Synthesis - Industrial. Encyclo-
pedia of Catalysis; John Wiley & Sons, Inc.: New York, 2002.
9. Iglesia, E. Design, Synthesis, and Use of Cobalt-Based
Fischer?Tropsch Synthesis Catalysts. Appl. Catal. A 1997,
10. Brady, R. C., III; Pettit, R. Mechanism of the Fischer?
Tropsch Reaction: The Chain Propagation Step. J. Am.
Chem. Soc. 1981, 103, 1287–1289.
11. Oukaci, R.; Singleton, A. H.; Goodwin, J. G. Comparison of
Patented Cof-T Catalysts Using Fixed-Bed and Slurry Bub-
ble Column Reactors. Appl. Catal. A 1999, 186, 129–144.
12. Dry, M. The Fischer?Tropsch Synthesis. Catal. Sci. Technol.
1981, 1, 159–255.
13. Park, J. C.; Bang, J. U.; Lee, J.; Ko, C. H.; Song, H. Ni@SiO2
Yolk-Shell Nanoreactor Catalysts: High Temperature Sta-
bility and Recyclability. J. Mater. Chem. 2010, 20, 1239–
14. Midgley, P. A.; Weyland, M. 3D Electron Microscopy in the
Physical Sciences: The Development of Z-Contrast and
Eftem Tomography. Ultramicroscopy 2003, 96, 413–431.
15. Li, H.; Xin, H. L.; Muller, D.A.; Estroff,L. A. Visualizing the 3D
Internal Structure of Calcite Single Crystals Grown in
Agarose Hydrogels. Science 2009, 326, 1244–1247.
16. Xin, H. L.; Ercius, P.; Hughes, K. J.; Engstrom, J. R.; Muller,
D. A.Three-DimensionalImaging of PoreStructures inside
Low-Kappa Dielectrics. Appl. Phys. Lett. 2010, 96, 223108-3.
17. Garvie, L. A. J.; Buseck, P. R. Ratios of Ferrous to Ferric Iron
from Nanometre-Sized Areas in Minerals. Nature 1998,
18. Paterson, J. H.; Krivanek, O. L. Elnes of 3D Transition-Metal
Oxides: II. Variations with Oxidation State and Crystal
Structure. Ultramicroscopy 1990, 32, 319–325.
19. Kurata, H.; Colliex, C. Electron-Energy-Loss Core-Edge
Structures in Manganese Oxides. Phys. Rev. B 1993, 48,
20. Rask, J. H.; Miner, B. A.; Buseck, P. R. Determination of
Manganese Oxidation States in Solids by Electron Energy-
Loss Spectroscopy. Ultramicroscopy 1987, 21, 321–326.
21. Zhao, Y.; Feltes, T. E.; Regalbuto, J. R.; Meyer, R. J.; Klie, R. F.
Co and Co Oxides. J. Appl. Phys. 2010, 108, 063704-1–