Kinetics and mechanism of 5-hydroxymethylfurfural oxidation and their implications for catalyst development

Article abstract of DOI:10.1016/j.molcata.2013.09.013

The reaction mechanism of 5-hydroxymethylfurfural (HMF) oxidation in neutral aqueous solution with O₂ to 5-hydroxymethyl-2-furancarboxylic acid (HFCA) and 2,5-furandicarboxylic acid (FDCA) was evaluated over a 3 wt% Pt/activated carbon catalyst in a semibatch reactor and confirmed that the mechanism was the same as that determined at high pH. In addition, the reaction kinetics of intermediate HFCA oxidation to FDCA over supported Pt at high pH were investigated. The combination of reaction kinetics and isotopic labeling studies using ¹⁸O-labeled H₂O and O₂ was used to suggest a reaction mechanism in which H₂O inserts oxygen into the product and O₂ scavenges electrons from the metal catalyst. Carbon nanofibers (CNF) containing excess acid or base groups were also used as supports for Pt and Au nanoparticles and evaluated as catalysts in HMF oxidation. Although the CNF-supported samples catalyzed HMF oxidation at rates similar to other carbon-supported Pt and Au catalysts, the CNF support with basic groups improved the ability of supported Au to form FDCA from HMF under mild conditions.

Full text of DOI:10.1016/j.molcata.2013.09.013

Accepted Manuscript
Title: Kinetics and mechanism of 5-hydroxymethylfurfural
oxidation and their implications for catalyst development
Author: Sara E. Davis Angelica Benavidez Robert W.
Gosselink Johannes H. Bitter Krijn P. de Jong Abhaya K.
Datye Robert J. Davis
PII:
S1381-1169(13)00338-5
DOI:
Reference:
http://dx.doi.org/doi:10.1016/j.molcata.2013.09.013
MOLCAA 8867
To appear in:
Journal of Molecular Catalysis A: Chemical
Received date:
Revised date:
Accepted date:
19-4-2013
20-7-2013
11-9-2013
Please cite this article as: S.E. Davis, A. Benavidez, R.W. Gosselink, J.H. Bitter, K.P.
de Jong, A.K. Datye, R.J. Davis, Kinetics and mechanism of 5-hydroxymethylfurfural
oxidation and their implications for catalyst development, Journal of Molecular
Catalysis A: Chemical (2013), http://dx.doi.org/10.1016/j.molcata.2013.09.013
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

1
Title: Kinetics and mechanism of 5-hydroxymethylfurfural oxidation and their implications for
catalyst development
a
b
c
c
Authors: Sara E. Davis, Angelica Benavidez, Robert W. Gosselink, Johannes H. Bitter, Krijn
c
b
a
P. de Jong, Abhaya K Datye, and Robert J. Davis *
a
Department of Chemical Engineering, University of Virginia, 102 Engineers Way,
Charlottesville, VA 22904-4741, USA
b
Department of Chemical and Nuclear Engineering, University of New Mexico, 1 University of
New Mexico, MSC 01 1120, Albuquerque, NM 87131-0001 USA
c
Inorganic Chemistry and Catalysis, Debye Institute for Nanomaterials Science, Utrecht
University, Universiteitweg 99, 3584 CG Utrecht, Netherlands
*
Corresponding Author:
Robert J. Davis
Department of Chemical Engineering
University of Virginia
1
02 Engineers Way
Charlottesville, VA 22904-4741, USA
phone: 434-924-6284
FAX: 434-982-2658
email: rjd4f@virginia.edu
Keywords: Kinetics, heterogeneous catalysis, oxidation, gold catalysis, platinum catalysis
Page 1 of 41

2
Abstract
The reaction mechanism of 5-hydroxymethylfurfural (HMF) oxidation in neutral aqueous
solution with O to 5-hydroxymethyl-2-furancarboxylic acid (HFCA) and 2,5-furandicarboxylic
2
acid (FDCA) was evaluated over a 3 wt% Pt/activated carbon catalyst in a semibatch reactor and
confirmed that the mechanism was the same as that determined at high pH. In addition, the
reaction kinetics of intermediate HFCA oxidation to FDCA over supported Pt at high pH were
1
8
investigated. The combination of reaction kinetics and isotopic labeling studies using O-labeled
H O and O was used to suggest a reaction mechanism in which H O inserts oxygen into the
2
2
2
product and O scavenges electrons from the metal catalyst. Carbon nanofibers (CNF) containing
2
excess acid or base groups were also used as supports for Pt and Au nanoparticles and evaluated
as catalysts in HMF oxidation. Although the CNF-supported samples catalyzed HMF oxidation
at rates similar to other carbon-supported Pt and Au catalysts, the CNF support with basic groups
improved the ability of supported Au to form FDCA from HMF under mild conditions.
1
. Introduction
The development of biomass-derived fuels and chemicals to replace fossil fuels and
petrochemicals continues to be of great interest. One such biomass-derived molecule is 2,5-
furandicarboxylic acid (FDCA), which has been identified as one of the top 12 value-added
chemicals from biomass by the U.S. Department of Energy because of its potential as a
replacement monomer in the production of plastics [1]. The diacid FDCA is structurally similar
to terephthalic acid, the monomer used in PET plastic production. Recently, Avantium
announced large scale production of FDCA-based polymers for bottles and other polymer
applications [2]. Diacid FDCA is an oxidation product of 5-hydroxymethylfurfural (HMF),
which is derived from fructose dehydration.
Page 2 of 41

3
The oxidation of HMF has been investigated under a variety of reaction conditions
utilizing both homogeneous and heterogeneous catalysts [3-15]. In early work, Verdeguer et al.
studied the oxidation of HMF over Pt catalysts with Pb promoters and noted the need for high
pH in the reactant mixture when the oxidation was performed at moderate temperatures [14].
That work also demonstrated hydroxide bases improved reaction rates more than carbonate
bases. The base concentration not only affects the rate of HMF oxidation but also influences the
product distribution, especially when Au catalysts are used. The general scheme of HMF
oxidation is presented in Figure 1.
O
O
O
Diformylfuran
O
O
O
O
O
(
DFF)
O
O
O
OH
HO
HO
OH
O
5
-Hydroxymethylfurfural
HMF)
2
,5-Furan-
O
5-Formyl-2-furan-
carboxylic Acid
(
dicarboxylic Acid
FDCA)
HO
OH
(
(
FCA)
5
-Hydroxymethyl-2-
furancarboxylic Acid
HFCA)
(
Figure 1. The general reaction scheme for 5-hydroxymethylfurfural oxidation
Previous reports indicated that oxidation of HMF at near room temperature (295 K) and
with small quantities of base (2:1 NaOH to HMF or 8% sodium methoxide relative to HMF) over
Au produced a majority of the monoacid, hydroxymethylfurancarboxylic acid (HFCA) [6, 13]. In
one case, simply raising the temperature to 403 K resulted in further oxidation to the diacid
product FDCA [13]. Increasing the amount of base (20:1 NaOH:HMF) and catalyst used also
facilitated the subsequent oxidation of HFCA to FDCA [6, 7]. Similarly, Casanova et al. reported
that the optimized conditions of 403 K, 1000 kPa O , and 4:1 NaOH:HMF produced >99% yield
2
Page 3 of 41

4
FDCA after 8 h over Au/CeO [4]. At the same pressure and solution concentrations but at lower
2
temperature (298 K), HFCA was the major product after complete conversion of HMF.
The literature indicates that the aldehyde group, which is highly reactive in high
concentrations of NaOH, is rapidly oxidized to the acid in the presence of a metal catalyst [6].
-
Since Au is unable to activate the alcohol group, high concentrations of OH are required to
oxidize the alcohol to an aldehyde, which is subsequently oxidized to acid rapidly [6]. In
contrast, supported Pt and Pd catalysts produced a majority of the diacid product, FDCA, at a
relatively low concentration of NaOH (2:1 NaOH:HMF), which is likely the result of higher
alcohol reactivity on those metals compared to Au [6]. The turnover frequency, or TOF, of HMF
oxidation over Pt and Pd catalysts is reportedly lower than that over Au catalysts, however.
The mechanism for HMF oxidation over Au and Pt catalysts at high pH was recently
investigated through the use of isotopically-labeled dioxygen and water [16]. The source of
oxygen insertion was shown to be H O rather than O in all cases. The role of dioxygen was
2
2
proposed to be as electron scavenger, closing the catalytic cycle and allowing the reaction to
proceed.
The high pH reaction conditions typical of HMF oxidation is unfavorable economically
and environmentally, leading to interest in developing reaction conditions that allow for FDCA
production in neutral solution. The use of a solid base support for the metal catalyst is one
potential way to replace the need for homogeneous base in the reaction. For example, some
investigators explored hydrotalcites as solid base supports for Au catalysts in the aqueous
oxidation of HMF to FDCA [8, 17]. Others have reported leaching of hydrotalcite during
2
+
reaction, likely the result of the product acid reacting with the Mg component of the support
Page 4 of 41

5
during HMF oxidation [17-19]. Hydrotalcite therefore appears to serve as a stoichiometric
replacement for the homogeneous base rather than as a true catalytic material. If a solid base
material is to be used to replace homogeneous base, the interaction between basic sites on the
catalyst and product acid FDCA must be considered. In addition, the very low solubility of the
product acids HFCA and FDCA in neutral water must also be addressed, as products may
precipitate and/or inhibit access to the metal catalyst [10, 20].
Lilga et al. reported on the base-free oxidation of HMF to FDCA over a Pt/ZrO catalyst
2
in a flow reactor [10]. Apparently, the oxidation proceeds over Pt in the absence of homogeneous
base at a relatively higher temperature (373 K) than commonly used in previous studies. The
group utilized low concentrations of HMF (0.5 – 3 wt%) to keep the product FDCA within its
solubility limit, ultimately realizing 98% yield of FDCA (373 K, 1035 kPa O2).
Since the pH of the reaction medium drops as the reaction proceeds, a molecular
understanding of the mechanism of base-free HMF oxidation is desired to complement earlier
studies performed under highly basic conditions.
Likewise, the impacts of NaOH concentration, O pressure, and temperature on the rate
2
of transformation as well as the product distribution during HMF oxidation are needed to guide
catalyst development.
Structural characteristics of the catalyst support are also important, especially when
dealing with relatively large molecules typical of biomass transformations. For example, a low
surface area support with large pores may be desirable to enhance the transport of bulky
substrates to the active metal particles located in the interior of the support pellets. Although
carbon has been widely used as a catalyst support, the microporosity of activated carbon can
Page 5 of 41

6
hinder internal mass transfer [21]. Reproducibility issues also plague the synthesis of activated
carbon, which is detrimental to applications in catalysis [21]. To ameliorate some of the transport
problems, non-microporous supports, such as carbon nanofibers (CNF), may be beneficial. An
added benefit of CNFs is that their synthesis is highly reproducible and provides a relatively
uniform product. The CNF materials can be functionalized with various surface groups, forming
both acidic and basic sites, and can be used alone as catalysts or as supports for catalytically-
active metal nanoparticles.
The acidic sites on oxy-functionalized CNF (CNF-ox) prepared by an oxidative treatment
in acid have been studied in detail, as has their influence on the synthesis of supported Pt and Ni
nanoparticles [22-24]. Nitrogen-containing CNFs (CNF-N) are reported to be solid base catalysts
[25-28]. For example, Van Dommele et al. demonstrated the effectiveness of nitrogen-containing
carbon nanotubes as solid base catalysts for the Knoevenagel condensation [26]. Likewise, Gong
et al. studied the electrocatalytic activity of nitrogen-containing carbon nanotubes for oxygen
reduction reactions for fuel cell applications [27]. Moreover, CNF-N materials have been
investigated recently as catalyst supports for several different metals. Liu et al. highlighted the
high dispersion and stability of Pd nanoparticles supported on CNF-N in the hydrodechlorination
of chlorobenzene [28]. The lone electron pair of pyridine species on the functionalized nanofibers
is thought to interact with Pd, enabling a high dispersion of metal nanoparticles and preventing
their aggregation. Similarly, Villa et al. report on Pd and Au-Pd catalysts supported on CNF and
CNF-N [29]. The metals supported on nitrogen-functionalized CNFs had smaller metal particle
sizes and therefore higher overall catalytic activity for liquid-phase alcohol oxidation.
In this contribution, we investigate the reaction mechanism of HMF oxidation over
supported Pt at neutral or lower pH to verify the generality of the mechanism determined
Page 6 of 41

7
previously under basic conditions. Isotopic labeling studies and detailed reaction kinetics provide
the basis for additional catalyst development with acid- and base-functionalized carbon
nanofibers as novel supports for Pt (and Au) particles.
2
. Experimental Section
2
.1 Catalyst Preparation
The carbon nanofibers (CNF) used in this work were grown over a Ni/SiO catalyst. First,
2
a Ni/SiO (5-wt% Ni) was prepared via homogeneous deposition-precipitation using SiO
2
2
(Degussa, Aerosil 200), Ni(II)NO3
●
6H O (Acros, 99% purity), and urea (Acros) [30]. The
2
resulting catalyst was used to grow fibers by flowing CO/H /N at 823 K as previously described
2
2
[
30]. The material containing both CNF and Ni/SiO catalyst was subsequently refluxed three
2
times in 1.0 M KOH solution for 1 h each time to remove SiO , and then washed with deionized
2
H O until the effluent was pH neutral. Next, the material was refluxed twice in 65% HNO for 1
2
3
h each time to remove Ni and introduce oxygen into the CNF. The material was washed again
with deionized H O until the effluent was pH neutral and dried in air at room temperature
2
overnight. This material is referred to as CNF-ox. Toebes et al. used this procedure to produce
CNF with a residual nickel content of <0.5 wt%, and the trace nickel was located inside graphitic
envelopes, making it unavailable for catalysis. [31]
Some of the CNF-ox material was further treated in NH to replace oxygen-containing
3
groups with nitrogen-containing groups [28]. The CNF-ox sample was heated to 1123 K under
3
-1
flowing NH (60 cm STP min ), held under these conditions for 100 minutes, and cooled under
3
flowing NH . The resulting materials are referred to as CNF-N.
3
Page 7 of 41

8
The CNF materials were subsequently used to support Pt and Au nanoparticles. The Pt
catalysts were prepared using an impregnation method [32]. In brief, 0.16 g H PtCl ·6(H O) was
2
6
2
3
dissolved in 30 cm distilled, deionized H O, to which 2.0 g of CNF-ox or CNF-N was added.
2
The mixture was stirred at room temperature for 18 h. The catalysts were filtered from solution,
washed several times with H O, and dried overnight in air at 403 K. The resulting catalysts were
2
3
-1
reduced in 100 cm min flowing H (UHP, Messer gases) at 673 K, cooled in H , exposed to air
2
2
and stored in a refrigerator. The supported Pt catalysts were used without further pretreatment.
The Au catalysts were synthesized using a sol method [33]. First, an Au sol was prepared
3
by dissolving 0.04 g HAuCl4
●
3(H O) (Aldrich) in 1000 cm deionized H O along with a 2 wt%
2
2
polyvinyl alcohol (Acros) solution in a ratio of Au:PVA = 1:1 w:w. The colloid was reduced by
adding 0.1 M aqueous solution of NaBH (Aldrich) dropwise in 3:1 molar ratio of NaBH :Au.
4
4
Following reduction, the sol was ruby red in color. The CNF support was added to the sol and
the mixture was acidified to pH = 2 with H SO . The Au sol-CNF mixture was stirred for 4 h to
2
4
allow the Au to deposit on CNF. The Au/CNF sample was recovered by filtration and washed
with several liters of H O. The filtrate was clear in color, confirming the Au sol had deposited on
2
the support. The catalyst was dried in air at 333 K overnight and subsequently treated in N and
2
3
−1
H (9:1 ratio) flowing at 150 cm min for 4 h at 573 K. The resulting Au catalyst was exposed
2
to air and stored in a refrigerator. As with the Pt catalysts, the Au samples were used without any
additional pretreatment. Metal weight loading (Au or Pt) was determined by ICP analysis
performed by Galbraith Laboratories, Knoxville, TN.
In addition, 3 wt% Pt on activated carbon (Aldrich) was also used in this study. It was
3
-1
reduced in H (UHP, Messer Gas) flowing at 150 cm min for 6 h at 573 K and cooled under
2
Page 8 of 41

9
flowing H . After exposure to air, the catalyst was refrigerated and used without further
2
pretreatment.
2
.2 Catalyst Characterization
2
.2.1 Chemisorption of H2
The dispersion of Pt was determined using a Micromeritics ASAP 2020 automated
-1
adsorption system. The Pt catalysts were heated to 473 K at 4 K min under flowing H (UHP,
2
Messer Gas) and reduced for 1.5 h. The samples were evacuated for 2 h at 473 K and then cooled
to 308 K for analysis. No H uptake was observed on the bare CNF-ox or CNF-N materials.
2
2
.2.2 Physisorption of N2
The surface area of CNF-ox was also determined using a Micromeritics ASAP 2020
automated adsorption system. The CNF-ox material was heated to 473 K and held under vacuum
for 16 h. The material was cooled to 77 K for N adsorption.
2
2
.2.3 Titration
The amount of acid and base sites on the carbon materials, with and without metals, was
determined by titration with NaOH and HCl using a Titralab TIM 845 system [24, 26]. To
evaluate the acid site density, a solution of 0.01 M NaOH and 0.1 M KCl was used as titrant. The
3
solid material (0.05 g) was dispersed in 60 cm of 0.1 M KCl and titrated to pH 5 and pH 7.5,
corresponding to the amount of strongly acidic groups and the total amount of acidic groups,
respectively. To quantify the base site density, a solution of 2.0 mM HCl was used as titrant. To
3
5
0 cm of 0.1 M KCl, 0.1 g of the solid material was added, and the mixture was titrated to pH 7.
Page 9 of 41

10
2
.2.4 Electron microscopy
High-angle annular dark field (HAADF) imaging was performed using a JEOL 2010F
FASTEM field emission gun scanning transmission electron microscope operated at 200kV. The
catalyst samples were suspended in ethanol by grinding in an agate mortar and pestle, and
deposited on a holey carbon support film on Cu TEM grids. The images were recorded and
analyzed using Digital Micrograph software.
2
.3 Oxidation Reactions
The aqueous phase oxidation of HMF (Acros, ≥98% purity) or HFCA (ChemBridge, 95%
3
purity) was carried out in a 50 cm Parr Instrument Company 4592 batch reactor equipped with a
3
glass liner. In all reactions, 7.0 cm of the reactant solution was added to the reactor along with
the appropriate amount of catalyst. The reactor was purged with flowing O (UHP, Messer Gas)
2
and then pressurized to the desired value. A constant pressure was maintained by a continuous
O feed. Samples of the product solution were acquired by removing the top of the reactor,
2
taking the sample, flushing the reactor with O and repressurizing. The maximum O transport
2
2
rate from the gas to the liquid was determined by oxidation of sodium sulfite [34], and the
reactant oxidation rate was kept significantly below this limit when quantitative rates were
measured. The samples from the oxidation reactions were filtered using PTFE 0.2 µm filters and
diluted with deionized H O in a 1:3 sample:water ratio. The analysis was conducted using a
2
Waters e2695 high performance liquid chromatograph (HPLC) at 303 K equipped with refractive
index and UV/vis detectors. The HPLC utilized a Bio-Rad Aminex HPX-87H column and 5mM
3
-1
H SO flowing at 0.5 cm min to perform the separation. The retention times and calibrations
2
4
for observed products were determined by injecting known concentrations.
Page 10 of 41

11
2
.3.1 Mass spectrometry
Mass spectrometry was performed using a Waters Micromass ZQ quadrupole mass
spectrometer (in both electronegative and electropositive modes, depending on the molecule
being ionized) using a direct injection of the product mixture. Prior work in our lab indicated that
exchange of oxygen between the water and the product can occur during HPLC analysis [35].
This exchange process was avoided by direct injection of the product into the mass spectrometer.
Known compounds in aqueous solution were directly infused into the mass spectrometer to
obtain reference spectra. The electronegative mode was best to ionize HFCA, FCA, and FDCA.
The parent peak for HFCA appeared at a mass-to-charge ratio (m/z) of 141 (mass of HFCA, 142,
+
minus 1, since H was removed during ionization). In the case of FDCA, a prominent peak
appeared at m/z of 155, which corresponds to the mass of FDCA minus one proton removed
during ionization. The peak for FCA appeared at 139, which corresponds to the mass of FCA,
+
1
40, minus 1, as 1 H was removed during ionization. The electropositive mode was best to
+
ionize HMF. The peak for HMF appeared at 127, corresponding to the mass of HMF plus 1 H
added during ionization.
3
. Results and Discussion
3
.1 Mechanism of base-free HMF oxidation over commercial Pt/C
In our previous work, the mechanism of HMF oxidation at high pH was based on isotopic
labeling studies over Au and Pt catalysts [16]. In brief, the oxidation of HMF proceeds by the
rapid, reversible hydration of the aldehyde moiety to a geminal diol in solution. This reaction is
greatly accelerated by NaOH. The second step is a hydrogen transfer from the geminal diol on
the catalyst surface to form a carboxylic acid, thus depositing two electrons on the metal. The
Page 11 of 41

12
hydroxymethyl group is subsequently oxidized on the metal surface to an aldehyde, again
depositing two electrons on the catalyst. The aldehyde undergoes reversible hydration to a
geminal diol, as in the first step. Finally, a hydrogen transfer step produces the second carboxylic
acid, again depositing two electrons on the metal surface. The role of O in HMF oxidation is to
2
scavenge electrons from the metal catalyst surface, thus closing the catalytic cycle and allowing
the reaction to continue. The presence of peroxide in solution during the reaction suggests that
O is sequentially reduced to peroxide and eventually to hydroxide, analogous to the previously
2
published mechanism of glycerol oxidation on Au [35]. Hydroxide ions in solution and on the
metal surface facilitate geminal diol formation and hydrogen abstraction from the alcohol
moiety. An investigation into the mechanism of HMF oxidation in neutral solution is presented
here to determine if the pH change influences the mechanism of oxygen insertion.
To confirm that the products of HMF oxidation did not exchange O appreciably with
1
8
liquid water, unlabeled HFCA, FCA, and FDCA were dissolved separately in H O in the
2
presence of Pt/C and heated to reaction temperature for the duration of a reaction experiment.
The same type of control experiment was conducted with HMF, though the duration was 4 h
because 100% conversion of HMF is realized in 4 h during a reaction. All experiments were
performed under N rather than O to minimize oxidation.
2
2
Significant conversion of HFCA, even in an inert environment, was observed in a control
experiment utilizing unlabeled water (60% conversion of HFCA in 18 h, with 83% selectivity to
-1
FCA and 17% selectivity to FDCA). The parent peaks of unlabeled HFCA and FCA (142 g mol
-1
and 140 g mole , respectively) are 2 mass units apart, which would cause a problem in the
18
analysis if O were incorporated. The mass to charge ratio is the same for HFCA containing no
18
18
O and FCA with one O incorporated (m/z = 141); thus, it is impossible to distinguish between
Page 12 of 41

13
the two compounds directly injected into the mass spectrometer. As this is the expected outcome
exchange in the aldehyde group due to reversible hydration in water, but no exchange in the
(
acid group), an exchange experiment utilizing HFCA would not yield useful results and was
therefore not performed. The exchange experiment utilizing FCA resulted in some oxidation,
even under inert atmosphere, but the only product was FDCA. The parent peaks of FDCA (m/z =
1
55) and FCA (m/z = 139) are separated by 16 mass units, so peak overlap was not a concern in
the analysis. Likewise, HMF and its oxidation product, HFCA, are separated by 16 mass units.
The FDCA control experiment did not result in any conversion of FDCA.
1
8
To confirm that the alcohol moiety does not exchange O with water under our reaction
conditions, ethanol was tested as substrate (0.3 M ethanol, 18 h and 348 K; pressure 1034 kPa to
keep substrate in liquid phase). No scrambling of oxygen was observed between ethanol and
18
18
H O; thus, exchange between the hydroxymethyl group of HFCA and H O was not expected.
2
2
An exchange experiment for the dialdehyde product, DFF, was not performed because of the
known reactivity of aldehydes in water.
1
8
The mass spectrum of the mixture from the HMF exchange experiment in H O
2
1
8
displayed a prominent peak at m/z = 129, corresponding to HMF with incorporation of one O.
Thus, scrambling of one oxygen atom in HMF with water must be accounted for when analyzing
the final product. The oxygen atom that was most likely exchanged with water was that of the
highly reactive aldehyde group, presumably the result of reversible hydration.
The spectrum from the FCA control experiment showed a prominent peak at m/z = 141,
18
18
indicating incorporation of one O atom from H O in FCA. An additional smaller peak was
2
1
8
seen at m/z = 143, suggesting the exchange of two O atoms. Thus, the scrambling of at least
Page 13 of 41

14
one oxygen atom, and possibly two, in FCA must be taken into account when analyzing the final
1
8
product FDCA. Very little exchange of oxygen between FDCA and H O was observed.
2
1
8
Although some oxygen scrambling with H O was observed with molecules that have an
2
aldehyde group, some details about the oxidation mechanism can still be gleaned. If oxygen
1
8
insertion occurred through dioxygen, HMF oxidation under O in unlabeled water would be
2
1
8
expected to produce FDCA containing two O atoms (incorporated during the oxidation of the
aldehydes to acids). If no label is found in the product, then oxygen addition utilized H O as an
2
oxygen source.
1
8
The mass spectra of the FDCA products from HMF oxidations under O in unlabeled
2
1
8
water and under unlabeled dioxygen in H O can be seen in Figure 2.
2
Figure 2. Mass spectra (electrospray negative mode) of major product, FDCA, from base-free
1
8
HMF oxidation over Pt/C catalyst. Striped bars indicate experiments conducted in O and
2
16
16
18
H O; solid bars indicate experiments conducted in O and H O. m/z of unlabeled FDCA =
2
2
2
1
55. Reaction conditions: T = 348 K, P = 345 kPa O ; HMF:Pt = 100; 0.15 M HMF. HMF
2
conversion = 100%.
Page 14 of 41

15
1
8
18
As shown in Figure 2, the reaction performed under O did not incorporate O into the
2
1
8
18
diacid product FDCA (m/z 155), whereas the reaction conducted in H O incorporated four O
2
1
8
atoms into FDCA (m/z 163). Incorporation of four labeled O atoms cannot be explained by
oxygen scrambling between reaction intermediates and labeled water. Since we have shown that
FDCA does not exchange O with H O under the conditions of this study, the oxidation of HMF
2
likely proceeds with H O to form FDCA, analogous to the mechanism of HMF oxidation at high
2
pH proposed previously [16].
We propose that the oxidation of HMF in base-free solution proceeds, like the high pH
case, via geminal diol formation due to reversible hydration of aldehyde in water (accounting for
18
two of the O atoms incorporated into FDCA). Oxidation of the geminal diol to carboxylic acid
likely occurs on the catalyst surface, depositing two electrons on the metal surface. Subsequent
activation of the alcohol moiety, on the catalyst surface, to form an aldehyde deposits two more
electrons on the metal surface. The next step is the reversible hydration of aldehyde to geminal
1
8
diol in solution, accounting for the other two O atoms found in the product acid. Finally, the
geminal diol is oxidized on the metal surface, depositing a two more electrons on the metal
surface. Consistent with earlier work, we propose that the role of dioxygen in HMF oxidation in
base-free solution is to scavenge electrons from the metal surface, being reduced to peroxide and
other species, closing the catalytic cycle in the process. It should be noted that a small amount of
HMF can be oxidized to dialdehyde DFF instead of HFCA under base-free conditions, but the
subsequent oxidation of DFF still yields the final product FDCA by analogous steps.
Page 15 of 41

16
3
.2 Reaction kinetics of HMF and HFCA oxidation over commercial Pt/C
With the mechanism of HMF oxidation at low pH established as the same as that at high
pH, the kinetics will provide additional insight into the reaction and will help guide catalyst
development. Though the influence of temperature, substrate concentration, O pressure, and
2
base concentration on the product distribution during HMF oxidation has been widely
investigated and reported in the literature, there are few reports on the effects of these parameters
on reaction rates. To the best of our knowledge, there have not been any reports on the influence
of these parameters on the Pt-catalyzed oxidation rate of HFCA, the main intermediate of HMF
oxidation to FDCA. Generally, the first step of HMF oxidation is oxidation of the aldehyde
moiety to carboxylic acid, HFCA. In contrast, the first step in oxidation of HFCA is oxidation of
the alcohol moiety to aldehyde. Since the mechanism of reaction over Pt is the same at neutral
and high pH, we chose to perform the kinetics studies at high pH to increase the solubility of the
various reaction products. A commercially-available 3 wt% Pt/C catalyst (Aldrich) having a
dispersion of 98% determined by H chemisorption was used to test the influence of the various
2
reaction conditions.
The effect of substrate concentration was investigated over the range of 0.05 M to 0.15
M. All other reaction parameters (295 K, 690 kPa O , 0.30 M NaOH concentration, 0.01 or 0.02
2
g 3 wt% Pt/C depending on substrate) were held constant for each experiment. Initial reactant
concentrations were chosen based on standard conditions in previous work (0.15 M substrate,
0
.30 M NaOH) [6]. To ensure an excess of NaOH was available throughout the reaction, the
range of initial substrate concentrations was kept below 0.15 M. Because the rate of HMF
conversion was greater than that of HFCA, the catalyst loading in the HMF oxidation
experiments was 0.01 g whereas 0.02 g was used in the HFCA experiments. Turnover
Page 16 of 41

17
frequencies derived from initial rates and metal dispersion of the fresh catalyst are plotted versus
initial substrate concentration for both HMF and HFCA oxidation in Figure 3.
0
0
0
0
0
0
0
.12
.10
.08
.06
.04
.02
.00
HMF
HFCA
0
.00
0.05
0.10
0.15
Initial Substrate Concentration (M)
Figure 3. Turnover frequencies of HMF and HFCA oxidation as a function of initial substrate
concentration confirm zero order behavior. Reaction conditions: 0.30 M NaOH, 295 K, 690 kPA
O , 0.01 g 3 wt% Pt/C (HMF) or 0.02 g 3 wt% Pt/C (HFCA)
2
The oxidation rates of both HMF and HFCA were zero order in substrate concentration
over the range investigated. This finding is in agreement with Vinke et al., who investigated
HMF oxidation over Pt/Al O [15]. In general, zero order behavior suggests the Pt catalyst
2
3
surface is highly covered with adsorbed reactant.
Page 17 of 41

18
The influence of initial NaOH concentration was also probed. The initial substrate
concentration 0.15 M was chosen for both HMF and HFCA, and the rest of the conditions were
held constant, as described earlier (295 K, 690 kPa O , 0.01 or 0.02 g Pt/C depending on
2
substrate ). Turnover frequencies are plotted as a function of NaOH concentration in Figure 4.
0
0
0
0
0
0
0
0
0
.16
.14
.12
.10
.08
.06
.04
.02
.00
HMF
HFCA
0
.0
0.5
1.0
1.5
2.0
2.5
Initial NaOH Concentration (M)
Figure 4. Turnover frequencies of HMF and HFCA oxidation as functions of initial NaOH
concentration. Reaction conditions: 0.15 M HMF or HFCA, 295 K, 690 kPA O , 0.01 g 3 wt%
2
Pt/C (HMF) or 0.02 g 3 wt% Pt/C (HFCA)
The rate of HMF oxidation was first order with respect to NaOH concentration in the
range 0.15 M – 0.6 M NaOH. It should be noted that the low end of the concentration range, 0.15
M, represents a 1:1 ratio of NaOH:HMF, and thus formation of acid product will lower the pH
Page 18 of 41

19
during the reaction. Moreover, the oxidation of HMF proceeds through the main intermediate
HFCA. Therefore, the initial oxidation rate, based on the consumption of HMF, is attributed to
the rate of oxidation of the aldehyde moiety to acid, i.e. HMF to HFCA.
The influence of NaOH concentration on the rate of HFCA oxidation over Pt was not the
same as its influence on HMF oxidation. A run utilizing 0.15 M NaOH resulted in very little
conversion of HFCA (5% conversion in 3 h), presumably because the 1:1 stoichiometry of
NaOH:HFCA of this experiment suggests that all of the NaOH was consumed in neutralizing the
acid groups of HFCA. Although the oxidation of HFCA over Pt in absence of base is possible
[10], the temperature for these reactions was not sufficiently high. At higher NaOH
concentrations (above 0.30 M NaOH, 2:1 NaOH:HFCA), the reaction rate was nearly zero order
in NaOH concentration.
The results here indicate that the oxidation of the aldehyde moiety of HMF depends on
NaOH concentration, whereas the oxidation of the alcohol moiety of HFCA does not, above 2:1
NaOH:HFCA. Aldehydes in aqueous solution are recognized to undergo rapid, reversible
hydration to geminal diols and the reaction is accelerated at high pH [36]. Previous work
indicated that the oxidation of HMF proceeds through hydration of aldehyde in solution and
subsequent dehydrogenation of the geminal diol on the catalyst surface to produce carboxylic
acid [16]. This work demonstrates that this step of HMF oxidation follows a first order
-
dependence on NaOH concentration. The higher concentration of OH in solution likely
facilitates the hydration to geminal diol, which subsequently dehydrogenates to the carboxylic
acid on the Pt catalyst. Above a critical concentration (2:1 NaOH:HFCA, or 0.30 M NaOH in
this case), however, there is no dependence of HFCA oxidation rate (the second oxidation step in
Page 19 of 41

20
HMF oxidation to FDCA) on NaOH concentration, suggesting the Pt surface is the key catalytic
component of the reacting system.
The effect of O pressure in the range 690 – 3000 kPa on HMF oxidation rate was
2
investigated previously over a Au/TiO catalyst and reported to be zero order. The relationship
2
between HMF oxidation rate and O pressure, and between HFCA oxidation rate and O pressure
2
2
over Pt, however, has not been probed. Thus, the O pressure was varied between 690 kPa and
2
2
070 kPa and the initial turnover frequencies were evaluated (Figure 5).
0
0
0
0
0
0
0
0
.14
.12
.10
.08
.06
.04
.02
.00
HMF
HFCA
0
500
1000
1500
2000
O Pressure (kPa)
2
Figure 5. Turnover frequencies of HMF and HFCA oxidation as function of O pressure.
2
Reaction conditions: 0.15 M HMF or HFCA, 0.30 M NaOH, 295 K, 0.01 g 3 wt% Pt/C (HMF)
or 0.02 g 3 wt% Pt/C (HFCA)
Page 20 of 41

21
In both cases, the O pressure was increased by 300%, but the rate of oxidation increased
2
by only 20% for HMF oxidation and 10% for HFCA oxidation. Thus, the rates of HMF and
HFCA oxidation were nearly zero order in O over the range of 690 to 2070 kPa. This result is
2
not surprising, given the proposed mechanism for HMF oxidation, in which the role of O is
2
simply to scavenge electrons from the catalyst surface. Recent work by Ide and Davis
investigating diol oxidation over Pt also reports a zero order dependence on dioxygen pressure
over the range 5 – 20 bar [37].
The temperature dependence of both HMF and HFCA oxidation was established and
overall activation energies for both HMF and HFCA oxidation were calculated. The oxidation of
HMF was investigated at temperatures between 273 K and 308 K to keep the oxidation rate
below the known external mass transfer limit in our system. Likewise, oxidation of HFCA was
investigated between 295 K and 323 K. Arrhenius-type plots from HMF and HFCA oxidation
are shown in Figure 6.
Page 21 of 41

22
-1
HMF
-2
-1
Ea,HMF = 29.0 kJ mol
-3
HFCA
-4
-1
Ea,HFCA = 29.6 kJ mol
-5
3.1
3.2
3.3
3.4
3.5
3.6
3.7
-
3
-1
1
/T (x10 K )
Figure 6. Arrhenius plots of HMF and HFCA oxidation. Reaction conditions: 0.15 M HMF or
HFCA, 0.30 M NaOH, 690 kPa O , 0.01 g 3 wt% Pt/C (HMF) or 0.02 g 3 wt% Pt/C (HFCA)
2
-1
The apparent activation energy of HMF oxidation was calculated to be 29.0 kJ mol in
-1
this study, which is lower than 37.2 kJ mol reported by Vinke et al. over Pt/Al O (reaction
2
3
conditions: constant pH = 9.0, p(O ) = 0.2 atm (p(total) 1.0 atm), 1.0 g 5 wt% Pt/Al O , 0.1 M
2
2
3
HMF) [15].
-1
The overall activation energy of HFCA oxidation was 29.6 kJ mol , which is surprisingly
similar to that of HMF oxidation. Ide and Davis also report the activation energy of 1,6-
Page 22 of 41

23
-1
hexanediol oxidation over supported Pt to be 35 kJ mol and that of 1,3-propanediol to be 34 kJ
-1
mol [37].
3
.3 Mechanistic implications
The oxidation of HMF to diacid is comprised of two parts: aldehyde oxidation and
alcohol oxidation. The kinetics of aldehyde oxidation were investigated by using HMF as
substrate. Aldehyde oxidation occurs by rapid, reversible hydration of the aldehyde to a geminal
diol (on the catalyst or in solution), followed by hydrogen transfer from the geminal diol to metal
surface, forming a carboxylic acid. Gaseous dioxygen is reduced by scavenging electrons from
the metal surface and eventually regenerating hydroxide.
The rate of HMF oxidation showed zero order dependence on HMF concentration, which
suggests that the metal surface is saturated in some form of the substrate. We attribute the first
order dependence of the rate on NaOH concentration to the reversible hydration of aldehyde to
geminal diol, which is accelerated by base.
The oxidation of the alcohol group in HFCA was zero order in substrate concentration,
NaOH concentration, and O pressure. This rather counterintuitive finding may be explained by a
2
kinetic model recently proposed by Ide and Davis in their study of alcohol oxidation over Pt at
low pH [37]. In their model, there are two types of non-competing sites on the Pt catalyst: one
accessible to all species involved in the reaction and one accessible only to hydrogen atoms. In
the dehydrogenation of hydrocarbons over transition metals, H atoms preferentially reside on
high coordination 3- and 4-fold sites, which are likely to be inaccessible to larger molecules. We
suspect a similar phenomenon occurs in the oxidation of HMF.
Page 23 of 41

24
The implication of their kinetic mechanism of alcohol oxidation at low pH, which is also
zero order in substrate and O , is that the measured rate is essentially the rate of C-H activation
2
of adsorbed alkoxide to form aldehyde. If C-H activation were the common step being evaluated
in all of these reactions tested (HMF, HFCA, other alcohols), then the similarity of the activation
-1
-1
energy for the reactions studied here (~30 kJ mol ) and elsewhere (~35 kJ mol ) [37] is
expected.
The other implication of this work involves the nature of the metal catalyst. Since Pt can
activate C-H bonds by itself whereas Au is ineffective at that reaction, there is clearly a more
important role of base in the case of Au catalysts compared to Pt catalysts.
3
.4 Carbon nanofibers as supports for Pt and Au nanoparticles
Recognizing the important difference between Pt and Au catalysts as a function of pH,
we decided to test the role of functionalized supports on the oxidation catalysis. As discussed
earlier, recent work has shown hydrotalcite as a support enables Au catalysts to oxidize HMF to
diacid without additional base. However, the hydrotalcite is consumed in the reaction and is not a
true catalytic material. In our work, we chose to use functionalized (acid and base) carbon
nanofibers as novel supports for the reaction.
The oxidation of HMF was therefore investigated over Pt and Au supported on carbon
nanofibers that were functionalized with either oxygen- or nitrogen-containing groups. Table 1
summarizes the physical and chemical properties of the freshly prepared (reduced) materials.
2
The BET surface area for the CNF-ox support materials was found to be 170 m , in agreement
2
-1
with other reports (180-200 m g ) [22, 23].
Page 24 of 41

25
Table 1. Results from characterization of CNF materials
Metal
Surface
Avg
Total
base
sites
Weight
loading Metal
Metal
Avg
Strong acid Total acid
metal
(%)
2.3
Dispersion Diameter Diameter
sites (mmol sites (mmol (mmol
a
b
c
-1 d
-1 d
-1 e
Catalyst
Pt/CNF-N
(%)
100
(nm)
0.7
(nm)
0.9
kg )
kg )
kg )
f
-
-
26.0
-
f
Pt/CNF-ox
Au/CNF-N
0.42
83
1.6
3.7
4.4
-
2.2
5.4
5.7
-
0.0
45
g
0.71
19
-
-
20.7
g
Au/CNF-ox 0.60
18
59
-
134
-
-
CNF-N
CNF-ox
0
0
-
-
8.9
-
-
-
164
333
a
determined via ICP analysis
b
determined via TEM analysis
c
3
2
Σd /d
d
3
determined by dispersing 0.05 g of material in 50 cm 0.1 M KCl solution and titrating with
0
.01 M NaOH and 0.1 M KCl solution. Strong acid sites determined by titration to pH = 5; total
acid sites determined by titration to pH = 7.5
e
3
determined by dispersing 0.1 g of material in 50 cm 0.1 M KCl solution and titrating with 2.0
mM HCl solution
f
determined via H chemisorption
2
g
calculated as inverse of surface average diameter
The final metal loading of the Pt/CNF-ox catalyst was much lower than that of Pt/CNF-N
(0.42 wt% and 2.3 wt %, respectively), although the synthesis procedures were identical and the
intended loading was 3 wt% Pt for both. Expectedly, the presence of acidic or basic surface
groups on the support material affected the uptake of H PtCl (aq) onto CNF. In the case of Au on
2
6
CNF, the metal weight loadings were similar between the two catalysts (0.71 wt% Au/CNF-N
and 0.60 wt % Au/CNF-ox) but both were less than the nominal 1 wt% that was targeted. The
acidic and basic surface functionalization did not have a substantial influence on the uptake of
Au onto CNF materials, presumably because the synthesis procedure involved the adsorption of
Page 25 of 41

26
a pre-reduced Au sol onto the fibers instead of impregnation of unreduced metal precursors as
was the case with Pt.
Results from titration of the acid and base sites are presented in Table 1. The initial pH of
a 0.1 M KCl solution after adding CNF-ox was between 4 and 5, indicating that CNF-ox
possessed acid sites. The materials were titrated to two endpoints, pH 5 and pH 7.5, to count sites
of various acid strength. Strong acid groups, or those titrated to pH 5, are generally attributed to
carboxylic acids [38]. The acid sites titrated between pH 5 and pH 7.5 are considered to be fairly
weak. From the titration of the CNF-ox support material, approximately half of the acid groups
-1
present on the surface are attributed to carboxylic acid groups (333 mmol kg total acid groups
-1
with 164 mmol kg strong acid groups). The addition of either Pt or Au to CNF-ox decreased the
total concentration of acid groups as well as the concentration of strong acid groups. The
-1
-1
Au/CNF-ox sample exposed 134 mmol kg total acid groups with 59 mmol kg attributed to
-1
strong acid groups, while Pt/CNF-ox contained 45 mmol kg total acid groups with no strong
acid groups. Although the loading of Pt (0.42 wt%) was slightly lower than the loading of Au on
CNF-ox supports (0.60 wt % Au), the addition of Pt consumed the acid groups substantially
more than the addition of Au. This result may be related to the much higher dispersion of Pt on
CNF-ox (83%) compared to Au (18%).
The loss of acidic sites that occurred during synthesis of Pt and Au catalysts supported on
CNF-ox may result from the deposition of metal and its reduction at elevated temperatures. The
same trend was noted by Plomp et al. during Pt/CNF-ox synthesis via a homogeneous
deposition-precipitation method [24]. Dihydrogen dissociated on metal nanoparticles might
migrate to the support during the reduction of Pt [26, 32]. Dihydrogen may also assist in the
Page 26 of 41

27
decomposition of oxygen-containing groups [32]. Acidic sites may also anchor metal particles to
the carbon surface, as noted in other works involving Pt/CNF-ox [22, 23].
The low uptake of Pt by CNF-ox during impregnation may be because of the lack of
basic sites. Basic sites on carbon are reportedly the strongest anchoring sites for the H PtCl (aq)
2
6
2
-
utilized in this synthesis method and are responsible for strong adsorption of PtCl6 (aq) on C
32, 39]. Basic sites incorporated in the CNF-N material via ammonia treatment were added at the
[
expense of acidic sites, which is consistent with the results in Table 1 [28]. The basic Pt/CNF-N
material therefore had a higher loading of Pt than the acidic Pt/CNF-ox (2.3 wt% and 0.42 wt%,
2-
respectively), likely because of the greater concentration of strong anchoring sites for PtCl6
aq).
(
The basic nanofibers, CNF-N, were titrated with 2 mM HCl to determine the number of
exposed basic sites. The initial pH of the material suspended in solution was between pH 8.5 and
.1, confirming that they were overall basic. As seen in Table 1, the number of basic sites was
9
-1
-1
higher on the supported catalysts (20.7 mmol kg on Au/CNF-N and 26.0 mmol kg on Pt/CNF-
-1
N) than on bare CNF-N (8.9 mmol kg ). The higher number of basic sites on the supported
catalysts does not indicate that the metal nanoparticles are basic, but rather that the material
likely has acidic and basic sites. This method of titrating acidic and basic sites measures a net
acidity or basicity; i.e., CNF-N materials have more exposed basic sites than acidic sites. The
higher number of basic sites on the supported catalysts may be attributed to the removal of
residual acid groups during reduction of the catalysts by H as described above, or could possibly
2
indicate some anchoring of metal particles on residual acidic sites.
Page 27 of 41

28
Electron microscopy of all supported metal catalysts revealed small, well-dispersed metal
particles. Images of the support materials and all catalysts are seen in Figure 7. In particular, the
Pt catalysts had very small particles (average diameters 0.7 nm for Pt/CNF-N and 1.6 nm for
Pt/CNF-ox) but were larger than the Pt particles on CNF materials. The Au catalysts contained
metal particles of about the same size (average diameters 3.7 nm for Au/CNF-N and 4.4 nm for
Au/CNF-ox). Evidently, the nature of the surface groups did have a large influence on loading
but not on the final metal particle size of supported Pt and Au catalysts.
Page 28 of 41

29
Figure 7. TEM images of [a] CNF-ox support, [b] CNF-N support, [c] Pt/CNF-ox, [d] Pt/CNF-N,
e] Au/CNF-ox, and [f] Au/CNF-N
[
Page 29 of 41