Surface speciation and alkane oxidation with highly dispersed Fe(III) sites on silica

Article abstract of DOI:10.1016/j.jcat.2011.01.007

When highly dispersed, supported Fe oxides are selective alkane oxidation catalysts, but new syntheses are required to reliably produce such materials. Here, highly dispersed, supported Fe³⁺ catalysts are prepared via incipient wetness impregnation of SiO₂ with aqueous Fe complexes of ethylenediaminetetraacetic acid (FeEDTA), followed by calcination. With Na ⁺ countercations, UV-visible diffuse reflectance spectra are entirely below 300 nm and H₂ temperature-programmed reduction only shows reduction at ～630 °C for all loadings up to 2.15 wt%, the maximum loading for a single impregnation cycle. These characteristics indicate isolated sites not seen for Fe(NO₃)₃ precursors even at 0.3 wt%. NH₄⁺ countercations lead to amorphous oxide oligomers and a minority species with unusual reducibility at 310 °C. Na⁺ countercations produce 'single-site' behavior in adamantane oxidation using H₂O₂ with a specific turnover frequency of 9.2 ± 0.8 ks^-1, constant for all Fe loadings and approximately 10 times higher than that of other well-dispersed Fe/SiO₂ materials. Similar turnover frequencies are obtained when counting only the highly reducible species on the NH₄⁺-derived catalyst, allowing these sites to be tentatively assigned as small, undercoordinated clusters that are both easily reduced and participate in alkane oxidation, reminiscent of Fe-exchanged MFI zeolites.

Full text of DOI:10.1016/j.jcat.2011.01.007

Journal of Catalysis 279 (2011) 103–110
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Surface speciation and alkane oxidation with highly dispersed Fe(III) sites on silica
⇑
Dario Prieto-Centurion, Justin M. Notestein
Department of Chemical and Biological Engineering, Northwestern University, Evanston, IL 60202, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
When highly dispersed, supported Fe oxides are selective alkane oxidation catalysts, but new syntheses
are required to reliably produce such materials. Here, highly dispersed, supported Fe3+ catalysts are pre-
pared via incipient wetness impregnation of SiO2 with aqueous Fe complexes of ethylenediaminetetra-
acetic acid (FeEDTA), followed by calcination. With Na+ countercations, UV–visible diffuse reflectance
spectra are entirely below 300 nm and H2 temperature-programmed reduction only shows reduction
at ꢀ630 °C for all loadings up to 2.15 wt%, the maximum loading for a single impregnation cycle. These
characteristics indicate isolated sites not seen for Fe(NO3)3 precursors even at 0.3 wt%. NHþ4 counter-
cations lead to amorphous oxide oligomers and a minority species with unusual reducibility at 310 °C.
Na+ countercations produce ‘single-site’ behavior in adamantane oxidation using H2O2 with a specific
turnover frequency of 9.2 0.8 ksꢁ1, constant for all Fe loadings and approximately 10 times higher than
that of other well-dispersed Fe/SiO2 materials. Similar turnover frequencies are obtained when counting
only the highly reducible species on the NHþ4 -derived catalyst, allowing these sites to be tentatively
assigned as small, undercoordinated clusters that are both easily reduced and participate in alkane oxi-
dation, reminiscent of Fe-exchanged MFI zeolites.
Received 20 October 2010
Revised 7 January 2011
Accepted 8 January 2011
Keywords:
Isolated site
EDTA ligand
Single-site catalyst
Supported oxide
C–H bond activation
Ó 2011 Elsevier Inc. All rights reserved.
1. Introduction
mal stability to agglomeration and formation of bulk Fe oxides,
in strong contrast with catalysts derived from simple Fe3+ salts
The selective oxidation of alkanes is an important challenge in
catalysis due to the potential for direct conversion of low-cost
feedstocks, such as natural gas and naphtha, into valuable oxygen-
ated commodity chemicals [1]. Supported Fe catalysts and, in par-
ticular, Fe-substituted MFI zeolites have attracted industrial and
academic attention due to their ability to promote NOx decomposi-
tion [2–6] and selective partial oxidation of hydrocarbons [6–8]. It
is generally accepted that some form of highly dispersed, binuclear
or atomically isolated Fe3+ species involved in the selective pro-
cesses catalyzed by these materials [5,9–13].
such as Fe(NO3)3. The use of this complex as a catalyst precursor
has only been minimally investigated, and no reports of its use
for alkane oxidation are known [17,22,23]. Dissolved FeEDTA com-
plexes have found previous use as absorption additives for NOx
scrubbers [24,25], but the principal application of the ligand con-
tinues to be as a preservative in the food industry [26].
2. Experimental
2.1. Materials and instruments
Several studies have sought to reproduce the structures and
reactivity of these substituted zeolites on different supports [14–
22]. Given that specialized reagents or preparation conditions are
required in some of these preparation methods [15,20], and that
multiple surface structures are often formed [18,19,21], we set
out to develop a route to highly dispersed, supported Fe using com-
monly available reactants and synthesis routes. Here, catalysts are
synthesized by incipient wetness impregnation (IWI) of SiO2 sup-
ports with aqueous, anionic Fe3+ complexes of ethylenediaminetet-
raacetic acid, (EDTA, Fig. 1) followed by oxidative heat treatment.
The nature of the countercation is important, and NaFeEDTA in
particular generates ‘single-site’ catalysts with outstanding ther-
All catalyst syntheses were performed under air at ambient con-
ditions unless specified otherwise. All reagents were obtained from
Sigma–Aldrich. SiO2 supports were obtained from Selecto Scientific
(32–63 l
m particle size, 630 m2 gꢁ1) and dried at 120 °C for 12 h
under ambient pressure before use. Cellulose ester dialysis mem-
brane tubes with a molecular weight cutoff of 100–400 Da were
purchased from SpectraPor and used in the purification of the Fe
complexes.
Electrospray ionization mass spectrometry (ESI-MS) measure-
ments of Fe complexes were taken with a Thermo Finnegan LCQ
instrument. Specific surface areas were calculated, using the BET
equation [27], from N2 absorption isotherms obtained with a
Micromeritics ASAP 2010 apparatus. All materials were degassed
8 h at 5 mTorr and ambient temperature before measurements.
⇑
Corresponding author.
E-mail address: j-notestein@northwestern.edu (J.M. Notestein).
0021-9517/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2011.01.007

104
D. Prieto-Centurion, J.M. Notestein / Journal of Catalysis 279 (2011) 103–110
tion to 4 mL to precipitate out the product. The precipitates were
A
B
O
filtered, redissolved in water, and purified by membrane dialysis.
The purified yellow solids were washed with acetone and dried un-
der dynamic vacuum (20 mTorr) for 12 h. NaFeEDTA and NH4FeED-
TA were also prepared from FeCl3 following identical procedures.
ESI-MS (negative ion mode) shows m/z = 346 as the most promi-
nent species, corresponding to the FeEDTA anion. TGA between
200 °C and 800 °C of NaFeEDTA and NH4FeEDTA gave a mass loss
due to combustion corresponding to 98.2% and 99.9% purity,
respectively, according to the ratio of organic to inorganic compo-
nents (Fig. S1).
O
O
O
O
HO
N
O
HO
N
O
O
N
N
OH
Fe
O
OH
O
O
O
Fig. 1. Structures of (A) H4EDTA ligand and (B) FeEDTA complex.
Elemental analysis was done using a Varian Vista MPX ICP-OES cal-
ibrated with standards of known concentration. Samples for ele-
mental analysis were prepared by dissolving ꢀ20 mg solids in
1 mL HF, followed by dilution to 100 mL in H2O. Diffuse reflectance
UV–visible (DRUV–vis) spectra were recorded with a Shimadzu
3600 UV–visible NIR spectrometer fitted with a Harrick Praying
Mantis diffuse reflection attachment and using polytetrafluoroeth-
ylene (PTFE) as the baseline standard. All materials were thor-
oughly ground before acquiring spectra, and all diffuse
reflectance spectra were converted to pseudo-absorption using
the Kubelka–Munk transform [28,29]. Solution UV–vis measure-
ments were performed with a standard 1-cm pathlength quartz
cell. Thermogravimetric analyses (TGA) were performed in a TA
Instruments Q500 thermogravimetric analyzer with an evolved
gas analysis furnace. Adamantane oxidation reaction products
were quantified by gas chromatography (GC) using a Shimadzu
2010 GC/FID system equipped with a Thermo-Nicolet TR-1 capil-
lary column and compared to standards of known concentration
of adamantane, and the expected products, 1-adamantanol, 2-ada-
mantanol, and 2-adamantanone. Material balances, defined as sum
of these products vs. the conversion of adamantane, were between
85% and 95% in all cases using only these products. The lower val-
ues are obtained with low Fe-loaded catalysts at low conversion
and are thus attributed to the limits of instrumental precision.
For temperature-programmed reduction (TPR) experiments, the
TA Q500 was fitted with a dilute H2 feed (4.5% H2, 4.5% Ar, and 91%
He) and H2O evolution was recorded with a Pfeiffer Thermostar
Q200 mass spectrometer connected to the furnace exhaust by a
fused SiO2 capillary column maintained at 200 °C. The evolved
m/z = 18 signal was calibrated against CuO reduction using Ar as
the internal standard. Before TPR measurements were taken, the
TA Q500 furnace was purged with UHP He for at least 90 min to re-
move O2.
2.3. Catalyst preparation
Fe was deposited on SiO2 by IWI with aqueous solutions of Na-
FeEDTA and NH4FeEDTA. The impregnated supports were stored in
partially covered containers and dried 48 h at room temperature
under ambient conditions, followed by 12 h under dynamic vac-
uum (ꢀ20 mTorr) also at room temperature. The dried materials
subsequently underwent heat treatment under UHP O2 with a
ramp rate of 10 °C minꢁ1 from room temperature to 800 °C,
600 °C or 400 °C. The Fe loading, expressed as Fe surface density,
was varied in the range 0.05–0.65 atom nmꢁ2 for NH4FeEDTA/
SiO2 and 0.05–0.36 atom nmꢁ2 for NaFeEDTA/SiO2 (Table 1). The
higher solubility of NH4FeEDTA (>200 g Lꢁ1) relative to NaFeEDTA
(ꢀ100 g Lꢁ1) accounts for the higher loadings obtained with the
former. The highest loadings were prepared using nearly saturated
solutions and thus may be subject to spontaneous precipitation
Table 1
Characteristics of the catalysts prepared for this study.
Catalyst
Fe loadinga
wt%
l
mol gꢁ1
atom-nmꢁ2
NaFeEDTA/SiO2
0.30
0.58
0.78
0.97
1.23
1.57
1.81
2.15
2.15
2.15
0.91
55
103
140
174
221
282
323
384
384
384
163
0.05
0.10
0.13
0.16
0.21
0.27
0.31
0.36
0.36
0.36
0.16
b
b
b
c
d
e
NH4FeEDTA/SiO2
0.31
0.58
0.75
0.98
1.25
1.57
1.83
2.59
3.25
3.85
1.25
1.25
0.92
56
103
135
175
223
281
331
463
582
689
175
175
165
0.05
0.10
0.13
0.17
0.21
0.27
0.31
0.44
0.55
0.65
0.17
0.17
0.16
b
X-ray photoelectron spectroscopy (XPS) was performed with an
Omicron ESCA probe equipped with EA125 energy analyzer. Photo-
b
b
emission was stimulated by a monochromated Al K
a radiation
(1486.6 eV) with the operating power of 300 W. Survey scan and
high-resolution scan were collected using pass energies of 50 and
25 eV, respectively. Binding energies of spectra were referenced
to the C 1s binding energy set at 284.8 eV. Prior to XPS measure-
ment, the powder samples were degassed in the entry-load cham-
ber for 12 h.
c
d
e
Fe(NO3)3/SiO2
Fe2O3/SiO2
0.67
1.37
2.04
364
244
120
0.34
0.23
0.11
2.2. NaFeEDTA and NH4FeEDTA synthesis
The preparation of monosodium Fe(III)ethylenediaminetetraacetic
acid, or NaFeEDTA, and ammonium Fe(III)ethylenediaminetetraacetic
acid, or NH4FeEDTA, was adapted from that described by Meier and
Heinemann [30]. A two-neck round-bottom flask was charged with
10 mmol H4EDTA, 10 mmol of Fe(NO3)3ꢂ9H2O, 25 mL of H2O and
heated to 60 °C with vigorous stirring under N2 atmosphere. Once
the solids fully dissolved, 40 mmol of NaHCO3 or NH4HCO3 was
slowly added. NaFeEDTA precipitated as a yellow salt within min-
utes, while NH4FeEDTA solution was reduced by rotary evapora-
1.34
240
0.23
a
Fe wt% from amount added during IWI. Molar loadings for EDTA catalysts
calculated from TGA weight losses during heat treatment agree to within 1% of
those calculated from Fe added during IWI. Surface densities calculated from initial
surface area.
b
Indicates Fe loadings confirmed by ICP-OES.
Heat treated to 600 °C.
Heat treated to 400 °C.
FeEDTA prepared from FeCl3, not Fe(NO3)3.
c
d
e

D. Prieto-Centurion, J.M. Notestein / Journal of Catalysis 279 (2011) 103–110
105
under deposition conditions. Under typical deposition conditions,
transmission UV–visible spectra of the solution showed no indica-
tion of dimeric FeEDTA complexes with characteristic absorption
peaks at 475 nm [31]. Transmission electron microscopy and elec-
tron diffraction of the heat-treated materials showed no regions of
high contrast or crystallinity. Transmission FT-IR spectroscopy of
pressed pellets showed no significant differences between the sup-
port and heat-treated materials (Fig. S2). These negative results are
expected from the low Fe loadings of these samples.
EDTA ligand has the intended effect of preventing site aggregation
by limiting metal–metal and metal–surface interactions.
TGA of the as-synthesized Fe-containing materials shows a one-
step decomposition of the EDTA ligand at 240 °C, similar to that of
pure H4EDTA (Fig. S3). The loading of FeEDTA complex is deter-
mined by assuming that any mass loss between 200 °C and
800 °C in excess of that observed for the bare SiO2 support corre-
sponds to ligand combustion leaving NaFeO4 and FeO4 on the sur-
face for NaFeEDTA and NH4FeEDTA, respectively (Fig. S4). The
loadings of selected materials were confirmed by ICP-OES (marked
with b, Table 1).
Control catalysts were prepared through IWI of SiO2 with aque-
ous Fe(NO3)3ꢂH2O and through mechanical mixing of Fe2O3 with
SiO2 (Table 1).
3.2. Diffuse reflectance UV–visible spectrometry
2.4. Catalytic reactions
The UV–vis absorption spectrum of the FeEDTA complex in
aqueous solution shows a characteristic ligand-to-metal charge
transfer (LMCT) peak at 280 nm. A similar peak is also present in
the diffuse reflectance spectra of the as-made catalysts, suggesting
that the complex maintains its structural integrity after impregna-
tion. (Fig. S5) The identity of the countercation, Na+ or NH4þ, does
not alter the shape or location of this peak.
A 50-mL round-bottom flask fitted with a Liebig condenser was
charged with 50 mg of catalyst, 70 mg (0.51 mmol) of adamantane,
and 20 mL of acetonitrile (MeCN). The mixture was heated to 60 °C,
stirred vigorously to dissolve adamantane, and 1 mL aqueous 30%
H2O2 (9.5 mmol) was injected. Aliquots of 0.3 mL were sampled
at fixed time intervals by syringe, passed through Whatman glass
Initial evidence for the high dispersion of the Fe atoms on the
SiO2 support can be obtained by visual inspection of the heat-trea-
ted catalysts – a red tint indicates the presence of Fe2O3 aggregates,
while colorless materials are indicative of well-dispersed, isolated
Fe3+ ions [15]. As Fig. 3 shows, after heat treatment, materials ob-
tained by impregnation of NaFeEDTA remain colorless up to the
maximum Fe loading of 0.36 atom nmꢁ2, whereas those obtained
by impregnation of NH4FeEDTA show a red tint for all loadings.
Catalysts obtained by Fe(NO3)3 IWI are substantially darker at all
loadings.
The DRUV–vis spectra of the heat-treated catalysts corroborate
these initial observations. As generally agreed in the literature
[14–16,18–20], isolated Fe sites absorb below 300 nm, two-dimen-
sional amorphous oligomers absorb in the 300–500 nm range, and
three-dimensional Fe2O3 particles absorb above 500 nm. As shown
in Fig. 4A, the spectra of catalysts obtained by IWI with NaFeEDTA
heat treated to 800 °C are dominated by a single well-defined peak
microfiber syringe filters with 0.7 lm pore size before being ana-
lyzed by GC-FID. The concentration of reaction products, 1-ada-
mantanol (1-ADL), 2-adamantanol (2-ADL), and 2-adamantanone
(2-ADN), was calibrated with respect to authentic standards of
known concentration. Ag powder was added to each GC analysis
vial to decompose any residual H2O2, which under GC inlet condi-
tions potentially could further oxidize the products. Controls
showed no change to product concentrations with this treatment.
3. Results and discussion
3.1. Physical characterization
N2 physisorption of as-prepared NH4FeEDTA and NaFeEDTA
show linear loss of surface area with loading, giving a constant sur-
face area loss of ꢀ0.66 nm2 per Fe complex (Fig. 2), consistent with
the dimensions of the complex given elsewhere [30]. The constant
loss of surface area per unit of FeEDTA indicates that for the load-
ing range under study, the Fe complexes are not forming multilay-
ers. This result suggests that complexing the Fe atoms with the
0.3 wt%
0.6 wt%
1.3 wt%
2.0 wt%
700
600
500
400
300
0
100
200
300
400
500
-1
600
Fe Loading, µmol-gsilica
Fig. 2. Linear loss of catalyst surface area with increasing Fe loading for SiO2 (e)
impregnated with aqueous NaFeEDTA (d) and NH4FeEDTA. (s) The linear loss of
surface area (0.66 nm2 per Fe complex) indicates that FeEDTA complexes do not
form multilayers on the SiO2 surface.
Fig. 3. Catalysts made by IWI of SiO2 with aqueous NaFeEDTA, NH4FeEDTA, and
Fe(NO3)3 after oxidative heat treatment and by physical mixing of SiO2 and Fe2O3.
Materials with highly dispersed, isolated Fe3+ sites are expected to be colorless.

106
D. Prieto-Centurion, J.M. Notestein / Journal of Catalysis 279 (2011) 103–110
A
A
Fe Loading
B
C
200
300
400
500
600
B
Wavelength, nm
Fig. 4. DRUV–vis spectra, after oxidative heat treatment, of catalysts prepared by
IWI of SiO2 with aqueous (A) NaFeEDTA, (B) NH4FeEDTA, and (C) Fe(NO3)3. Fe
loadings are 0.30–2.15 wt%, 0.31–3.25 wt%, and 2.0 wt%, respectively. Note that for
catalysts derived from NaFeEDTA, eight spectra overlay completely. Spectra are
normalized to the most intense feature, but are all of similar absolute intensity.
below 300 nm. Note that eight spectra, for catalysts of Fe loadings
between 0.05 and 0.36 atom nmꢁ2, are overlaid in Fig. 4A. The
remarkable spectral similarity suggests that surface Fe sites in Na-
FeEDTA-derived materials are structurally uniform over the range
of loadings under study. Furthermore, the intense peak below
300 nm can be assigned to the most intense LMCT band of very
highly dispersed, ‘isolated’ Fe3+ in octahedral coordination [16]. Gi-
ven that all spectra were acquired under ambient conditions, the
isolated Fe3+ is expected to be coordinated by six oxygen atoms
combined from the surface and adsorbed H2O [15]. Coordination
by H2O after calcination at high temperature is consistent with a
Lewis acidic site. The spectra of catalysts prepared by impregnation
of NH4FeEDTA possess the LMCT band below 300 nm and a tail
extending beyond 300 nm which, as the Fe loading increases, be-
comes a prominent shoulder at 350 nm (Fig. 4B). The growing tail
and shoulder can be assigned to amorphous Fe3+ oxide oligomers
growing in nuclearity with increasing loading [16]. The much more
prominent shoulder at 370 nm observed in the spectra of the heat-
treated Fe(NO3)3-derived material (Fig. 4C) is also assigned to the
presence of amorphous Fe3+ oligomers, whereas the tail extending
beyond 500 nm signifies the presence of three-dimensional Fe2O3
crystallites [15,16]. The spectrum of mechanically mixed Fe2O3
and SiO2 shows broad absorption bands of comparatively low
intensity at wavelengths higher than 500 nm, indicating that most
of the Fe is in a bulk oxide phase, as expected. Varying the initial
heat treatment temperature did not significantly alter the spectra
of any of these materials.
C
D
100
300
500
700
900
Temperature, 0C
Fig. 5. TPR profiles of catalysts heat treated to 800 °C. Fe loadings increase from
bottom to top within each series. SiO2 impregnated with aqueous (A) NaFeEDTA, (B)
NH4FeEDTA, (C) Fe(NO3)3, and (D) physically mixed Fe2O3 in SiO2. Profiles A show
one reduction event characterized by a Gaussian peak around 630 °C. Profiles B
show two reduction events characterized by Gaussian peaks around 300 °C and
500 °C. Profiles C show a single reduction event characterized by an asymmetric
peak with maxima around 500 °C. Profile D shows the usual Fe2O3 to Fe0 two-step
transition and the expected 2:3 ratio of Fe to evolved H2O. From profiles A, B, and C,
ratios of 1:0.5 of Fe to evolved H2O were calculated, indicating reduction from Fe3+
to Fe2+ and suggesting strong surface–metal interactions.
tively). These ratios indicate single-electron reduction of Fe3+ to
Fe2+ and are consistent with metal–surface interactions that
strongly inhibit the reduction of Fe2+ to Fe0 [15,17,22]. Mechani-
cally mixed Fe2O3 and SiO2 shows the expected 1.52 0.03 H2O-
to-Fe ratio of the bulk oxide [32].
Catalysts prepared with NaFeEDTA show a single reduction
event between 610-650 °C up to 0.36 atom nmꢁ2 – the maximum
Fe loading possible from a single cycle of IWI with this precursor
(Fig. 5A). In contrast, catalysts prepared by the NH4FeEDTA route
show two reduction events, one between 290-320 °C and the other
between 450 and 520 °C (Fig. 5B). The area under both reduction
peaks initially grows proportionally to the loading, but the more
3.3. Temperature-programmed reduction
While DRUV–vis provides information on the presence of differ-
ent oxidic Fe species, their unknown extinction coefficients pre-
clude quantitative observations. However, the degree of
dispersion is also correlated with the redox properties of the sup-
ported metal. Low-nuclearity species are generally considered to
interact strongly with the support and have high reduction tem-
peratures, while the opposite is expected of high nuclearity species
[15]. Fig. 5 shows the TPR profiles, normalized to the total sample
weight, of the heat-treated catalysts after oxidation to 800 °C. For
all catalysts prepared through IWI, TPR shows a similar ratio of
evolved H2O to Fe atoms (0.50 0.04, 0.47 0.03, and 0.49 0.03
for NaFeEDTA, NH4FeEDTA, and Fe(NO3)3-derived catalysts, respec-
reducible one plateaus at ꢀ40
lmol H2O per gram of catalyst,
equivalent to ꢀ80
l
mol Fe per gram of catalyst or ꢀ0.08 Fe-nmꢁ2
(Fig. 6). A broad reduction event at ꢀ500 °C is present in TPR pro-
files of Fe(NO3)3-derived catalysts (Fig. 5C). None of the materials
prepared by IWI show the characteristic two-step reduction profile
of the Fe2O3 powder to Fe0 particles (Fig. 5D) [32]. The reduction
temperatures identified in these catalysts are weak functions of

D. Prieto-Centurion, J.M. Notestein / Journal of Catalysis 279 (2011) 103–110
107
800
600
400
200
0
A
B
C
D
700
710
720
730
740
0
200
400
600
800
Binding Energy, eV
Total Fe Loading, µmol-g-1
Fig. 7. X-ray photoelectron spectra of selected catalysts: 2.15 wt% Fe NaFeEDTA/
SiO2 (A) after 800 °C treatment in O2 and (B) after 800 °C treatment in H2, and
3.25 wt% Fe NH4FeEDTA/SiO2 (C) after 800 °C treatment in O2 and (D) after 800 °C
treatment in H2. Solid lines at 711 and 725 eV mark the binding energy of the 2p3/2
ground state and 2p1/2 excited state, respectively, for Fe3+. The broken lines at 709
and 723 eV mark the binding energy of the 2p3/2 ground state and 2p1/2 excited
Fig. 6. Total vs. specific Fe loadings for NH4FeEDTA catalysts. The loading of the
more reducible Fe3+ oxide species (s) plateaus near 80 mol gꢁ1 or 0.08 Fe-nmꢁ2
whereas that of the less reducible species (d) continues to grow.
l
,
state, respectively, for Fe2+
.
loading and do not show significant variation over the studied
range.
It has been reported in the literature that isolated Fe3+ ions un-
dergo reduction under TPR conditions between 600 °C and 700 °C,
while two-dimensional amorphous oligomers and three-dimen-
sional crystallites reduce around 500 °C and below 400 °C, respec-
tively [15,21]. These trends suggest that the only Fe oxide species
present in the catalysts derived from NaFeEDTA are highly dis-
persed Fe3+ ions, while catalysts derived from NH4FeEDTA form
predominantly amorphous two-dimensional oligomers. The other,
highly reducible sites in NH4FeEDTA-derived catalysts are reminis-
cent of Fe-exchanged MFI zeolites, some of which are known to re-
duce between 300 °C and 400 °C [12,13]. This similarity suggests
that the reducible species be assigned as highly dispersed Fe pres-
ent not as isolated atoms, but rather as very small clusters, similar
to those in exchanged zeolites. The exact structure of these reduc-
ible sites synthesized via NH4FeEDTA, and why they saturate at
711.6 eV and the Fe 2p1/2 excited state at 725.2 eV and 724.8 eV,
respectively. The reduced NaFeEDTA- and NH4FeEDTA-derived
materials show spin-orbit splitting of 2p3/2 ground state at
709.5 eV and 709.7 eV and the Fe 2p1/2 excited state at 723.7 eV
e and 723.2 eV, respectively. The peak positions and energy differ-
ences between Fe 2p3/2 and Fe 2p1/2 are typical of Fe3+ for the oxi-
dized materials and Fe2+ for the reduced ones [37–39]. XPS
measurements were taken after days of exposure to ambient con-
ditions following reduction, and the stability of Fe2+ species for ex-
tended periods is indicative of strong metal–support interactions.
3.5. Oxidation of adamantane
ꢀ80
l
mol Fe per gram of catalyst (ꢀ0.08 Fe-nmꢁ2), is under
The heat-treated materials were tested as catalysts for the oxi-
dation of adamantane by aqueous H2O2 (Fig. 8). Fig. 9 shows turn-
over numbers (TON) for a representative catalyst (1.81 wt% Fe
heat-treated NaFeEDTA/SiO2) over 3 h (11 ks).
investigation.
Alkali metals are known to improve dispersion of transition
metals on SiO2 [33–35]. Here, Na+ cations on the SiO2 surface are
expected to exchange with the silanolic protons upon impregna-
tion [36]. During heat treatment, basic Na+ ions, already tightly
bound to the acidic SiO2 surface, provide an anchor for the weakly
acidic-to-amphoteric Fe3+ oxide which forms upon removal of the
EDTA ligand. Using the EDTA complex, however, is uniquely suc-
cessful in creating isolated sites because it combines the dispersion
enhancement of Na+ with the steric protection of the bulky EDTA
ligand.
Adamantane is a particularly informative substrate for catalytic
oxidations because the ratio of tertiary to secondary oxidation
products (3°/2°) provides insight into the nature of the reaction.
On a per hydrogen basis, the observed ratio of 3°/2° oxidation
products after 30 min was 2.4 0.1 and 2.6 0.1 for catalysts de-
rived from NaFeEDTA and NH4FeEDTA, respectively. Fe(NO3)3-de-
rived catalysts showed similar 3°/2° selectivity. In radical-
mediated reactions, the lower C–H bond dissociation energy on
tertiary carbons produces high 3°/2° selectivity, and some molecu-
lar Fe catalysts show values as high as 10 [40,41]. Fe salts and sup-
ported oxide catalysts typically give values between 3.7 and 2.7
[20,42,43], as observed here, indicating more reactive oxidants
than the molecular catalysts, but less reactive than the largely
3.4. X-ray photoelectron spectroscopy
The quantitative TPR experiments described in Section 2.3
showed that the Fe3+ catalysts prepared by IWI could only be re-
duced to Fe2+ under attainable conditions. Correspondingly, XPS
measurements indicate that only Fe3+ is present in the heat-treated
catalysts, whereas Fe2+ is present in the reduced materials. Fig. 7
depicts the Fe 2p core level of a 0.36 atom-nmꢁ2 Fe/SiO2 derived
from NaFeEDTA (A and B) and a 0.65 atom-nmꢁ2 Fe/SiO2 catalyst
derived from NH4FeEDTA (C and D), after oxidation to 800 °C (A
and C) and after subsequent reduction to 800 °C (B and D). The oxi-
dized materials prepared with NaFeEDTA and NH4FeEDTA show
spin-orbit splitting of the Fe 2p3/2 ground state at 711.5 eV and
OH
Fe3+/SiO2
OH
+
O
+ H2O2
+
60 °C, MeCN
Fig. 8. Adamantane oxidation by H2O2 over Fe3+/SiO2. From left to right, the
reaction products are 1-adamantanol (1-ADL), 2-adamantanol (2-ADL), and 2-
adamantanone (2-ADN).

108
D. Prieto-Centurion, J.M. Notestein / Journal of Catalysis 279 (2011) 103–110
6
5
4
3
2
1
0
60
6
6
4
4
2
2
50
0
0 0.0
0.4
0.0
0.4
0
0..88
40
Loading, atom-nm-2
Loading, atom-nm-2
30
20
10
0
0
2
4
6
8
10
12
0
10
20
30
40
Time, ks
Total Fe, µmol
Fig. 9. Adamantane oxidation products vs. time for 1.81 wt% Fe catalysts prepared
with NaFeEDTA. (s) Total TON, all reactions’ products included. (d) 1-adamantanol.
Fig. 11. Initial reaction rate vs. total Fe in reactor (21 mL solution, 50 mg catalyst) of
heat-treated NH4FeEDTA/SiO2 materials for adamantane oxidation with H2O2. Inset:
TOF in ksꢁ1 vs. Fe loading in atom-nmꢁ2. The catalysts show decreasing activity per
total Fe with increasing loading. (s) Total reaction products. (d) 1-adamantanol.
(ꢃ) Total reaction products over heat-treated Fe(NO3)3/SiO2. (j) Total reaction
products over heat-treated NH4FeEDTA/SiO2 made from FeCl3.
(ꢃ) 2-adamantanone. (
D
) 2-adamantanol.
200
20
15
10
5
50
150
100
50
2200
1155
0
0.0
0.2
0.4
Loading, atom-nm-2
1100
40
5
5
0
00.00
0.05
0.10
-2
0.00
0.05
0.10
Loading, atom-nm
Loading, atom-nm-2
30
20
10
0
0
0
5
10
15
20
Total Fe, µmol
0
1
2
3
4
5
Fig. 10. Initial reaction rate vs. total Fe in reactor (21 mL solution, 50 mg catalyst) of
heat-treated NaFeEDTA/SiO2 materials for adamantane oxidation with H2O2. Inset:
TOF in ksꢁ1 vs. Fe loading in atom-nmꢁ2. Aside from small deviations for the highest
and lowest loadings, the catalysts show the ‘single-site’ behavior expected for
catalysts with uniform, highly dispersed, and equally accessible active sites. (s)
Total reaction products. (d) 1-adamantanol. (ꢃ) Total reaction products over heat-
treated Fe(NO3)3/SiO2. (j) Total reaction products over heat-treated NaFeEDTA/
SiO2 made from FeCl3.
Fe (Tred = 300 oC), µmol
Fig. 12. Initial reaction rate vs. quantity of the more reducible Fe in reactor (21 mL
solution, 50 mg catalyst) for adamantane oxidation with H2O2 over heat-treated
NH4FeEDTA/SiO2 materials. Inset: TOF in ksꢁ1 vs. Fe loading in atom-nmꢁ2. The
more reducible Fe species shows ‘single-site’ behavior. (s) Total reaction products.
(d) 1-adamantanol.
indiscriminate gas-phase OH radicals, which have an expected 3°/
2° selectivity near 2 [44].
fraction of the total Fe atoms being active sites present as defects
in increasingly large oxide sheets or crystallites. Although less ac-
tive than NaFeEDTA catalysts, these materials still consistently
show initial rates twice those of Fe(NO3)3 catalysts. However, if
the initial rate is plotted against the concentration of the more
reducible Fe species (those reducing between 290 °C and 320 °C)
instead of the total Fe concentration, a linear relation reappears
(Fig. 12). Given the low TOF of Fe(NO3)3-derived catalysts, the lim-
iting case can be assumed, where all oxidation reactivity is as-
cribed to these more reducible sites in NH4FeEDTA-derived
catalysts. As such, the renormalized initial TOF for NH4FeEDTA cat-
alysts, 9.3 0.8 ksꢁ1, is comparable to that of the species identified
in NaFeEDTA catalysts, 9.7 0.2 ksꢁ1. Both are an order of magni-
tude larger than adamantane oxidation TOF over highly dispersed
Fe-SiO2 materials reported elsewhere [20].
Catalysts derived from NaFeEDTA show first-order initial oxida-
tion rates with respect to total Fe concentration (Fig. 10), or equiv-
alently, that the initial turnover frequency (TOF) is independent of
the Fe loading (Fig. 10, inset). Constant catalytic TOF is expected
from the likewise-invariant DRUV–vis spectra and TPR results
and suggest that Fe sites possess the same, or very similar, struc-
tures for all loadings studied. Therefore, the NaFeEDTA-derived
materials are well-dispersed, site-isolated catalysts that are verifi-
ably ‘single-site’. In contrast, Fe(NO3)3-derived catalysts show
much lower activity and do not show any increase in absolute rate
with loading (ꢃ symbols, Fig. 10).
Catalysts derived from NH4FeEDTA show decreasing TOF (per
total Fe atoms) with increasing loading (Fig. 11). This common
behavior for supported oxides is consistent with a decreasing

D. Prieto-Centurion, J.M. Notestein / Journal of Catalysis 279 (2011) 103–110
109
The agreement between TOF on the two types of sites, and the
OH2
OH2
OH2
OH2
HO O
generally high oxidation reactivity, gives further support to the
hypothesis that the NH4FeEDTA precursor generates a minority
population of very small clusters, as was argued when analyzing
the TPR data. Such clusters would be easily reduced due to bridging
oxygens, yet also proficient in activating H2O2 due to the presence
of undercoordinated sites. The rigorously isolated species derived
from NaFeEDTA also possess undercoordinated sites capable of
activating H2O2, but are extremely resistant to reduction. In both
cases, the most highly dispersed species on each sample are pro-
posed to be responsible for the observed oxidation catalysts.
Although the pore restrictions do not permit adamantane oxida-
tion with Fe-substituted MFI zeolites, the types of Fe active sites
seen here are also considered to be responsible for high activity
in oxidation with activated oxidants in Fe-substituted MFI zeolites
[12,13].
HOOH
HOOH
Fe
O
Fe
O
H
O
O
O
O
OH2
OH2
OH2
OH2
OH2
OH2
OH2
H2O
H2O
HO O
OH2
Fe
O
Fe
Fe
Fe
H
O
O
O
O
O
O
O
O
O
Fig. 13. (top) Atomically dispersed Fe is proposed to be formed as the sole species
on catalysts derived from NaFeEDTA. (bottom) Small clusters of FeOx are proposed
to be formed as a reactive minority species on catalysts derived from NH4FeEDTA.
Both types of low-nuclearity species are expected to coordinate H2O2 at sites left
vacant by loss of H2O or solvent. Coordinated hydroperoxide is proposed to be the
first intermediate formed on route to a surface species sufficiently oxidizing to
abstract H from adamantane, which then forms the observed products. In contrast
to similar observed oxidation behavior, the bridging O will make the small clusters
substantially more reducible in H2 than are the atomically dispersed species.
3.6. Other catalytic considerations
The effect of the Fe source on the catalysts was tested by pre-
paring catalysts with FeEDTA complexes derived from FeCl3.
Although Clꢁ is largely removed during precursor purification,
using FeCl3 as the Fe source decreases the activity of NaFeEDTA
noticeably, but has less of an effect for NH4FeEDTA-derived cata-
lysts. The NH4FeEDTA complex is more pure, but halogen poison-
ing may also be less significant for the NH4FeEDTA-derived
catalysts because only a minority of the sites is proposed to be ac-
tive in catalysis.
Under reaction conditions, neither isomerization of alcohols nor
oxidation of 2-ADL to 2-ADN were found to occur. This result indi-
cates that a possible reaction mechanism includes the transforma-
tion of adamantane to alkylperoxide intermediates that
subsequently decompose to the product alcohols and ketones. In
such cases, thermal decomposition of residual alkylhydroperoxides
can skew apparent product distributions during GC analysis. Some
reaction aliquots were therefore treated with triphenylphosphine
(PPh3) to reduce alkylhydroperoxides to the corresponding alco-
hols, in accordance with a method from the literature [45]. For
the catalysts under study, treatment with PPh3 did not alter the
product distribution, indicating that alkylhydroperoxide interme-
diates do not reach significant concentrations under reaction con-
ditions, although there is insufficient evidence to completely
discount the formation of metal-alkylperoxide or alkylhydroperox-
ide species as transiently formed intermediates.
lated to a reorganization of the surface FeOx under reaction condi-
tions, which remains under investigation.
4. Conclusions
This study demonstrates a straightforward method to create
well-defined, isolated Fe sites for alkane oxidation by H2O2 via
IWI of SiO2 with aqueous FeEDTA complexes. The Fe dispersion at-
tained by this method is much higher than for IWI with Fe(NO3)3,
as measured by TPR and DRUV–vis spectrophotometry. NaFeEDTA
is particularly effective at dispersing Fe sites, leading to rigorously
‘single-site’ reactivity and spectroscopic behavior, due to the com-
bined action of the Na+ cation and the bulky chelating ligand. Three
types of surface species were identified by DRUV–vis spectropho-
tometry and subsequently quantified by TPR. Two-dimensional
amorphous oligomers and crystallites are present for catalysts de-
rived from Fe(NO3)3; two-dimensional oligomers are also the dom-
inant species on the surface of NH4FeEDTA-derived catalysts. These
surface sites are poorly active in adamantane oxidation by H2O2. A
minority of the sites in NH4FeEDTA-derived catalysts are easily re-
duced and also proposed to be responsible for reactivity in ada-
mantane oxidation, reminiscent of sites assigned as small
clusters in Fe-exchanged MFI zeolites. In contrast, all of the Fe in
NaFeEDTA-derived materials are assigned as isolated sites that
are an order of magnitude more active than those found in other
highly dispersed Fe preparations and share the same apparent
TOF (9–10 ksꢁ1 or 32–36 hꢁ1) as the minority sites in NH4FeED-
TA-derived catalysts. This latter finding suggests that alkane oxida-
tion by H2O2 over Fe catalysts requires undercoordinated Fe found
in isolated sites or small clusters, but is structure insensitive within
this narrow subclass of Fe species. As illustrated in Fig. 13, very
small clusters and atomically dispersed species are both expected
to coordinate H2O2 at sites left vacant by loss of H2O or solvent.
Coordinated hydroperoxide as Fe(III)OOH is proposed as the first
intermediate toward the formation of a surface species sufficiently
oxidizing to abstract H from adamantane. Soluble, non-heme mod-
el complexes suggest the formation of high-valent Fe(IV) or Fe(V)
oxo or oxo-hydroxo species as the oxidizing center [46], but the
limits of the current study permit neither rigorous assignment of
the oxidizing species nor the reaction pathway, analysis of which
is in progress.
Fe leaching into solution under adamantane oxidation condi-
tions was calculated by ICP-OES to be 0.2%, 0.4%, and 0.6% of the
Fe present on the surface for catalysts heat treated at 800 °C,
600 °C, and 400 °C, respectively. Pretreatment at 800 °C was se-
lected for the majority of the catalyst testing primarily due to its
lowest level of leaching; leached Fe concentrations were 0.3 and
3.4 lM for the lowest and highest loaded catalysts, respectively.
If only the Fe3+ in solution were active, its initial TOF would ap-
proach 3500 ksꢁ1 – over 300x that ascribed to surface catalysis.
However, under conditions used in this study, the homogeneous
oxidation of adamantane by H2O2 over 180 l
M Fe3+ from Fe(NO3)3
shows an initial TOF of 4.9 ksꢁ1, which is of the order of the initial
TOF ascribed to the heterogeneous catalysts. Oxidation products
could not be observed when using Fe concentrations comparable
to those expected in solution under reaction conditions. Similarly,
hot filtration after three minutes of reaction completely halts the
reaction in solution, (Fig. S6) which is consistent with the vast
majority of the reaction occurring on the catalyst surface. Likewise,
the loss of a small fraction of surface Fe as leached Fe3+ is unlikely
responsible for the observed decrease in TOF after 1000s – this
would again imply unrealistically high TOF for surface Fe before
they are leached. The change in TOF with time is more likely re-
Acknowledgments
JMN acknowledges the support of Northwestern University and
a 3M Non-Tenured Faculty Grant, and DPC acknowledges the

110
D. Prieto-Centurion, J.M. Notestein / Journal of Catalysis 279 (2011) 103–110
[18] A. Gervasini, C. Messi, P. Carniti, A. Ponti, N. Ravasio, F. Zacheria, J. Catal. 262
(2009) 224–234.
[19] A. Gervasini, C. Messi, A. Ponti, S. Canedese, N. Ravasio, J. Phys. Chem. C 112
(2008) 4635–4642.
[20] C. Nozaki, C.G. Lugmair, A.T. Bell, T.D. Tilley, J. Am. Chem. Soc. 124 (2002)
13194–13203.
[21] A. Parmaliana, F. Arena, F. Frusteri, A. Marinez-Arias, M.L. Granados, J.L. Fierro,
Appl. Catal. A 226 (2002) 163–174.
support of Patrick G. and Shirley W. Ryan through the Ryan Fellow-
ship. XPS was performed in the Keck-II facility of NUANCE Center at
Northwestern University. NUANCE Center is supported by NSF-
NSEC, NSF-MRSEC, Keck Foundation, the State of Illinois, and
Northwestern University. ICP-OES and ESI-MS were performed in
the IMSERC facilities at Northwestern University with the support
of the National Science Foundation under DMR-0114235.
[22] S. Valange, R. Palacio, A. Charmot, J. Barrault, A. Louati, Z. Gabelica, J. Mol. Catal.
A: Chem. 305 (2009) 24–33.
[23] A.J. van Dillen, R.J.A.M. Terorde, D.J. Lensveld, J.W. Geus, K.P. De Jong, J. Catal.
216 (2003) 257–264.
Appendix A. Supplementary material
[24] T.W. Chien, H.T. Hsueh, B.Y. Chu, H. Chu, Process Saf. Environ. Protect. 87
(2009) 300–306.
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.jcat.2011.01.007.
[25] L. Wang, W. Zhao, Z. Wu, Chem. Eng. J. 132 (2007) 227–232.
[26] T.E. Furia, Sequestrants in food, in: T.E. Furia (Ed.), CRC Handbook of Food
Additives, CRC Press, Boca Raton, 1980, pp. 272–273.
[27] S. Brunauer, P.H. Emmett, E. Teller, J. Am. Chem. Soc. 60 (1938) 309–319.
[28] P. Kubelka, F. Munk, Z. Tech. Phys. 12 (1931) 593–601.
[29] F.C. Jentoft, in: B.C. Gates, H. Knozinger, F.C. Jentoft (Eds.), Advances in
Catalysis, Academic Press, Amsterdam, 2009, pp. 141–148.
[30] R. Meier, F.W. Heinemann, Inorg. Chim. Acta 337 (2002) 317–327.
[31] G. McLendon, R.J. Motekaitis, A.E. Martell, Inorg. Chem. 15 (1976) 2306–2308.
[32] M.J. Tiernan, P.A. Barnes, G.M.B. Parkes, J. Phys. Chem. B 105 (2001) 220–228.
[33] M.A. Bañares, N.D. Spencer, M.D. Jones, I.E. Wachs, J. Catal. 146 (1994) 204–
210.
[34] S. Irusta, A.J. Marchi, E.A. Lombardo, E.E. Miro, Catal. Lett. 40 (1996) 9–16.
[35] S. Yang, W. Zhu, Q. Zhang, Y. Wang, J. Catal. 254 (2008) 251–262.
[36] L.H. Allen, E. Matijevíc, L. Meites, J. Inorg. Nucl. Chem. 33 (1971) 1293–1299.
[37] R.S. Prakasham, G.S. Devi, K.R. Laxmi, C.S. Rao, J. Phys. Chem. C 111 (2007)
3842–3847.
[38] T. Herranz, S. Rojas, F.J. Pérez-Alonso, M. Ojeda, P. Terreros, J.L.G. Fierro, Appl.
Catal. A 308 (2006) 19–30.
[39] Y. Joseph, G. Ketteler, C. Kuhrs, W. Ranke, W. Weiss, R. Schlogl, Phys. Chem.
Chem. Phys. 3 (2001) 4141–4153.
[40] A.N. Biswas, P. Das, A. Agarwala, D. Bandyopadhyay, P. Bandyopadhyay, J. Mol.
Catal. A 326 (2010) 94–98.
[41] F. Minisci, F. Fontana, S. Araneo, F. Recupero, S. Banfi, S. Quici, J. Am. Chem. Soc.
117 (1995) 226–232.
References
[1] A.T. Bell, B.C. Gates, D. Ray, Basic research needs: catalysis for energy, in: Basic
Energy Sciences Workshop, US Department of Energy, 2007.
[2] V.I. Sobolev, G.I. Panov, A.S. Kharitonov, V.N. Romannikov, A.M. Volodin, K.G.
Ione, J. Catal. 139 (1993) 435–443.
[3] J. Valyon, W.S. Millman, W.K. Hall, Catal. Lett. 24 (1994) 215–225.
[4] H.-Y. Chen, W.M.H. Sachtler, Catal. Today 42 (1998) 73–83.
[5] R. Joyner, J. Phys. Chem. B 103 (1999) 5963–5976.
[6] Q. Zhu, R.M. van Teeffelen, R.A. van Santen, E.J.M. Hensen, J. Catal. 221 (2004)
575–583.
[7] V.I. Sobolev, K.A. Dubkov, O.V. Panna, G.I. Panov, Catal. Today 24 (1995) 251–
252.
[8] A. Ribera, I.W.C.E. Arends, S. de Vries, J. Pérez-Ramírez, R.A. Sheldon, J. Catal.
195 (2000) 287–297.
[9] L.J. Lobree, I.-C. Hwang, J.A. Reimer, A.T. Bell, J. Catal. 186 (1999) 242–253.
[10] P. Marturano, A. Kogelbauer, R. Prins, J. Catal. 190 (2000) 460–468.
[11] A.L. Yakovlev, G.M. Zhidomirov, R.A. van Santen, J. Phys. Chem. B 105 (2001)
12297–12302.
[12] F. Heinrich, C. Schmidt, E. Löffler, M. Menzel, W. Grünert, J. Catal. 212 (2002)
157–172.
[13] M. Schwidder, M.S. Kumar, K. Klementiev, M.M. Pohl, A. Brückner, W. Grünert,
J. Catal. 231 (2005) 314–330.
[14] F. Arena, F. Frusteri, L. Spadaro, A. Venuto, A. Parmaliana, Stud. Surf. Sci. Catal.
vol. 143 (2000) 1097–1105.
[42] G. Roelfes, M. Lubben, R. Hage, J.L. Que, B.L. Feringa, Chem. Eur. J. 6 (2000)
2152–2159.
[43] B. Singh, J.R. Long, F.F.d. Biani, D. Gatteschi, P. Stavropoulos, J. Am. Chem. Soc.
119 (1997) 7030–7047.
[44] J.K. Kochi, Free radicals, in: J.K. Kochi (Ed.), Reactive Intermediates in Organic
Chemistry, Wiley, New York, 1973.
[45] G.B. Shul’pin, J. Mol. Catal. A: Chem. 189 (2002) 39–66.
[46] A. Bassan, M.R.A. Blomberg, P.E.M. Siegbahn, L. Que, J. Am. Chem. Soc. 124
(2002) 11056–11063.
[15] F. Arena, G. Gatti, G. Martra, S. Coluccia, L. Stievano, L. Spadaro, P. Famulari, A.
Parmaliana, J. Catal. 231 (2005) 365–380.
[16] S. Bordiga, R. Buzzoni, F. Geobaldo, C. Lamberti, E. Giamello, A. Zecchina, G.
Leofanti, G. Petrini, G. Tozzola, G. Vlaic, J. Catal. 158 (1996) 486–501.
[17] Z. Gabelica, A. Charmot, R. Vataj, R. Soulimane, J. Barrault, S. Valange, J. Therm.
Anal. Calorim. 95 (2009) 445–454.