Development of a transient response technique for heterogeneous catalysis in the liquid phase, Part 1: Applying an electrospray ionization mass spectrometry (ESI-MS) detector

Article abstract of DOI:10.1016/j.jcat.2008.05.002

We have developed a novel, transient response technique for liquid-phase heterogeneous catalytic studies, equipped with an electrospray ionization mass spectrometry (ESI-MS) detector. The technique was successfully applied as an online method for real-time detection of species dissolved in aqueous product streams at the exit of a catalytic reactor. Two test reactions, nitrite reduction with Pt/SiO₂ and glucose oxidation with Pt/CNF/Ni, were used to demonstrate semi-quantitative monitoring of reactants, intermediates, and products. The capability of the novel technique is demonstrated by the fact that the ESI-MS detector is sufficiently sensitive to determine quantitatively extreme small amounts of physisorbing nitrite, down to 0.5% of a monolayer on the Pt surface. Nitrite also reacts with preadsorbed hydrogen, and the quantitative experimental results agree with the formation of both nitrogen and ammonia. The ESI-MS detector is able to distinguish between different components simultaneously, as in the case of glucose oxidation; this is its most significant advantage over existing transient techniques. A limitation of ESI-MS is its inability to detect gases dissolved in liquids, due to the relatively mild ionization process through electrospray. The second part of this article discusses an alternative detector to satisfy this deficiency.

Full text of DOI:10.1016/j.jcat.2008.05.002

Journal of Catalysis 257 (2008) 244–254
Contents lists available at ScienceDirect
Journal of Catalysis
www.elsevier.com/locate/jcat
Development of a transient response technique for heterogeneous catalysis in the
liquid phase, Part 1: Applying an electrospray ionization mass spectrometry
(ESI-MS) detector
D. Radivojevic´ a,1, M. Ruitenbeek b, K. Seshan a, L. Lefferts a,∗
a
+
Catalytic Processes and Materials, Faculty of Science and Technology, Institute of Mechanics, Processes and Control Twente (IMPACT) and MESA , University of Twente, P.O. Box 217,
7500AE Enschede, The Netherlands
b
SABIC Europe R&D, P.O. Box 319, 6160 AH, Geleen, The Netherlands
a r t i c l e i n f o
a b s t r a c t
We have developed
Article history:
a
novel, transient response technique for liquid-phase heterogeneous catalytic
Received 12 February 2008
Revised 29 April 2008
Accepted 1 May 2008
Available online 10 June 2008
studies, equipped with an electrospray ionization mass spectrometry (ESI-MS) detector. The technique
was successfully applied as an online method for real-time detection of species dissolved in aqueous
product streams at the exit of a catalytic reactor. Two test reactions, nitrite reduction with Pt/SiO2 and
glucose oxidation with Pt/CNF/Ni, were used to demonstrate semi-quantitative monitoring of reactants,
intermediates, and products. The capability of the novel technique is demonstrated by the fact that
the ESI-MS detector is sufficiently sensitive to determine quantitatively extreme small amounts of
physisorbing nitrite, down to 0.5% of a monolayer on the Pt surface. Nitrite also reacts with preadsorbed
hydrogen, and the quantitative experimental results agree with the formation of both nitrogen and
ammonia. The ESI-MS detector is able to distinguish between different components simultaneously, as
in the case of glucose oxidation; this is its most significant advantage over existing transient techniques.
A limitation of ESI-MS is its inability to detect gases dissolved in liquids, due to the relatively mild
ionization process through electrospray. The second part of this article discusses an alternative detector
to satisfy this deficiency.
Keywords:
Transient operated reactor
Liquid phase
ESI-MS
Glucose oxidation
Nitrite hydrogenation
© 2008 Elsevier Inc. All rights reserved.
1. Introduction
products (TAP) method, which has been recently used by many
researchers, including Yablonsky et al. [8] and Nijhuis et al. [9].
A similar approach is appropriate for developing and improving
catalysts for application in the liquid phase. However, many of the
experimental techniques developed for gas-phase experiments re-
quire vacuum or low pressure and cannot be applied in the liquid
phase. An example of successful modification of a characterization
technique to enable experiments in the liquid phase is infrared (IR)
spectroscopy in attenuated total reflection (ATR) mode, which was
pioneered by Bürgi et al. [10] and further developed by Ebbesen et
al. [11].
In transient experiments, including chemisorption experiments
in pulse mode or step change mode, there are no major problems
when operating in the liquid phase. The challenge is mostly tech-
nical in nature, because rapid analysis of species at the exit of
the reactor is required for transient studies. In gas-phase transient
experiments, this is normally achieved by online mass spectrome-
try (MS), which allows real-time multicomponent analysis at least
semiquantitatively. The goal of the present study is to develop and
demonstrate a similar technique that can be used for heteroge-
neous catalytic experiments in the liquid phase.
Heterogeneous catalytic reactions in the liquid phase are impor-
tant in such areas as production of bulk chemicals, fine chemicals,
and pharmaceuticals [1,2]. It is well known that in most cases,
production of fine chemicals and pharmaceuticals is less efficient
compared with that of bulk chemicals, due to multistep synthesis
procedures and the use of stoichiometric reduction and oxidation
reactions. Therefore, efficient and selective catalysts are needed to
improve these processes. Many of the targeted molecules are rel-
atively complex and have limited thermal stability, implying that
most of the reactions require solvents.
The development of highly efficient catalysts for gas-phase re-
actions has been assisted by powerful characterization techniques,
in situ characterization [3] and transient techniques for mechanis-
tic studies [4–6]. Transient methods were significantly optimized
by Gleaves et al. [7] when developing the temporary analysis of
Corresponding author. Fax: +31 53 489 4683.
E-mail address: l.lefferts@tnw.utwente.nl (L. Lefferts).
Present address: SABIC Europe R&D, P.O. Box 319, 6160 AH, Geleen, The Nether-
*
1
Previous attempts have been made to perform transient-type
experiments in the liquid phase. A refractive index (RI) detector
lands.
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.05.002

D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
245
was used in a few studies [12,13]. Denayer et al. [12] performed
pulse-type transient experiments to determine the influence of
polarity, pore size, and topology on the adsorption of n-alkanes,
isoalkanes, aromatics, and other organic components on FAU and
MFI zeolites. These authors used a high-performance liquid chro-
matography (HPLC) cartridge filled with FAU or MFI zeolites, com-
bined with an inline differential RI detector. Jonker et al. [13]
used tracer pulse experiments to determine an effective diffusion
coefficient, De, of edible oils and their fatty acid methyl esters
over Ni/SiO2 catalysts. These authors also used an HPLC cartridge
filled with catalyst (5–40 μm) as reactor, combined with an in-
line differential RI detector. Continuous-flow UV–visible (UV–vis)
spectroscopy detectors also have been used [14–16]. Hejtmánek et
al. [14] performed step-type transient experiments to determine
axial dispersion coefficients for tracer/carrier–liquid systems, such
as acetone–water, toluene–methanol, and acetone–ethylene glycol,
using HPLC cartridges filled with 4.5 μm of monodispersed spheri-
cal glass particles. Lin et al. [15] performed tracer studies using liq-
uid chromatography to determine effective intraparticle diffusion
coefficients and adsorption equilibrium constants on silicalite and
alumina of various alcohols in water and of toluene, acetone, and
ethyl acetate dissolved in cyclohexane. Kiraly et al. [16] used an
HPLC cartridge filled with a Pt/Al2O3 catalyst in transient pulse ex-
periments to determine Pd dispersion at the solid–liquid interface.
Gao et al. [17] used a flow-through liquid cell for Fourier trans-
form infrared spectroscopy (FTIR) analysis, connected to an HPLC
cartridge filled with 55-μm Pt/Al2O3 catalyst particles to study the
hydrogenation of acetophenone.
oped the technique for application in the aqueous phase, knowing
that many other processes in organic solvents could benefit as
well.
2. Experimental
2.1. Materials
HPLC silica (Microsorb from Varian, BET surface area 198 m2/g;
mean particle size 5 μm) was used as the support for Pt/SiO2. The
fixed-bed reactor was created by filling an HPLC cartridge (4.6 ×
−3
10
m i.d., 0.1 m long) with Pt/SiO2 catalyst. The catalyst bed
was fixed in the cartridge with two metallic 2-μm frits at the inlet
and the outlet. CNF/Ni foam was prepared as described elsewhere
[22,28]. The CNF loading was 30 wt%, and the total surface area
was 50 m2/g (BET, Micromeritics). Pt/CNF supported on Ni foam
−3
−3
m
was obtained in cylindrical pieces (4.55×10 m wide, 5×10
long).
Tetra ammonium–platinum hydroxide (NH3)4Pt(OH)2 (99.9%,
Aldrich) was used as the platinum precursor for the SiO2 sup-
ported catalyst, whereas Pt(C5H7O2)2 (99%, Aldrich) was used
as the platinum precursor for CNF. Ultra-pure LC-MS grade wa-
ter (Biosolve) was used to prepare (NH3)4Pt(OH)2, NaNO2 (99%,
Aldrich), glucose (99%, Aldrich), and gluconic acid (50% aqueous
solution, Aldrich) solutions. Toluene (99%, Merck) was used to pre-
pare solutions of Pt(C5H7O2). Argon (Air Products BIP, 99.9999%)
was further purified using an oxygen trap (OxyTrap, Alltech) to de-
crease the oxygen content to ꢀ10 ppb. Hydrogen (99.999%) and
oxygen (99.999%) were purchased from Indugas.
All of these examples have in common that the equipment con-
sists of (i) a liquid feed section equipped with a device to generate
pulses or steps, (ii) a catalytic reactor based mainly on HPLC car-
tridges filled with catalyst, and (iii) a detector. The most significant
difference is the choice of detector. Commonly used detectors (e.g.,
RI, UV–vis [18]) have the disadvantage of not being able to dis-
tinguish between different compounds. This is a serious limitation
when more then one component is present in the mixture leaving
the reactor. Consequently, we have developed alternative methods
for detecting different species leaving the reactor simultaneously,
based on electrospray ionization mass spectrometry (ESI-MS). Mass
spectrometry is a fast technique that in principle is able to detect
multiple components simultaneously [19], although this is not triv-
ial. Mass spectrometers with electrospray ionization are available
commercially as part of HPLC-MS systems equipped with a liquid-
phase–MS interface [20]. To the best of our knowledge, a transient
response technique for liquid-phase heterogeneous catalytic stud-
ies equipped with an ESI-MS as detector has not been described
previously.
2.2. Catalyst preparation and characterization
The Pt/SiO2 catalyst was prepared by an ion-exchange method
[21]. The platinum loadings of fresh and spent catalysts were de-
termined with X-ray fluorescence spectroscopy (XRF; Phillips PW
1480 spectrometer). The Pt/CNF/Ni catalyst was created by wet im-
pregnation in the following sequence. First, 0.9 g of CNF/Ni was
contacted overnight with 60 mL of Pt(C5H2O7)2 in toluene. Then
◦
the CNF/Ni pieces were dried for 3 h at 60 C in a rotary evapo-
◦
rator, followed by calcination at 270 C for 2 h in 20% O2 in N2
and reduction in 20% H2 in N2 for 1 h. The flow rate of gases
during treatment was 100 ml min−1. The Pt concentration of the
Pt/CNF/Ni catalyst could not be determined with XRF because the
foam did not fit into the XRF spectrometer; however, the amount
of platinum introduced should result in 3% by weight.
Platinum dispersions of the Pt/SiO2 and Pt/CNF/Ni catalysts
were measured with a pulse-flow chemisorption apparatus (Chemi-
Sorb 2750, Micromeritics). The platinum metal dispersion was cal-
culated assuming a H:Pts ratio of unity [29]. Particle sizes of fresh
Pt/SiO2 and Pt/CNF/Ni catalysts were cross-checked with trans-
mission electron microscopy (TEM; Phillips CM 30 microscope,
300 kV); typically, the sizes of 100 platinum particles were de-
termined in each sample.
The pressure drop in the catalytic reactor has been minimized
in this study to generate sufficient flexibility to optimize liquid
flow rates. The first approach was to use a catalyst based on
monodispersed silica support particles (Microsorb silica) [21]. The
second approach was to use carbon nanofibers (CNFs) supported
on Ni (CNF/Ni) [22] foam as a catalyst support with extreme high
porosity.
To develop and demonstrate our technique, we selected two
catalytic reaction systems in aqueous phase studied previously by
our group: (i) catalytic reduction of nitrite (NO2 ) over a Pt/SiO2
2.3. Transient response experimental setup
−
Fig. 1 presents a schematic representation of the equipment for
the experiments. The setup consisted of (i) a feed section (contain-
ers, selection valve and pump) with a switching valve to introduce
a well-shaped step/pulse function in the reactant concentration;
(ii) a catalytic reactor, (i.e., packed bed or filled with foam); and
(iii) an ESI-MS detector connected directly to the reactor outlet.
catalyst and (ii) catalytic oxidation of glucose over Pt supported
on carbon. Nitrite hydrogenation [23] is a relevant reaction for
both nitrate-to-nitrogen denitrification of drinking water [24] and
nitrate hydrogenation to hydroxyl-amine, an intermediate in the
production of caprolactam [1]. Oxidative conversion of glucose is
important in view of the anticipated shift toward renewable feed-
stocks for the chemical industry. Glucose can be upgraded to a
more valuable compound, such as gluconic acid, a building block
for production of chelating agents [25–27]. Therefore, we devel-
2.3.1. Feed section with pulse/step introducing device
Gas-tight liquid containers (three homemade 1.8-L vessels made
of Pyrex glass) were connected to a pulse-free HPLC binary pump

246
D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
Table 1
Instrumental parameters for ESI-MS analysis
Capillary voltage (kV)
Desolvation gas temperature ( C)
Nebulising gas flow rate (L min
Water with analyte flow rate (mL min
4.0/3.8a
350
12
0.5
0.5
◦
−1
)
−1
)
−1
Water flow rate (mL min
)
a
Capillary voltage used in experiments with glucose.
formation of gas-phase ions from charged droplets. Parameters
that can affect the ESI process are solvent composition, temper-
ature, spray voltage, solvent flow rate, and nebulizing gas flow
rate [32]. ESI is usually applied to analyze high-molecular-weight
and polar molecules, like organic acids, which easily form ions.
Unfortunately, compounds that are difficult to ionize (e.g., nitro-
gen and other dissolved gases) cannot be detected with ESI-MS. In
the present study, ESI was operated in negative mode, and anal-
ysis was performed in the mass range of 10 and 500 m/z. The
signal-to-noise ratio was optimized for both cases via the choice
of parameters shown in Table 1. Additives usually applied to en-
hance the ionization process [33] were not used in this study, to
prevent any interaction of these additives with the catalyst.
Fig. 1. Scheme of the experimental set-up.
ꢀꢀ
(G1312A, Agilent) with 1/8 tubings (Swagelock), either directly
(container 3, containing pure water) or through an HPLC valve 1
(G1160A, Agilent) (containers 1 and 2, containing dissolved gases
and/or with reacting species). The containers were stirred electro-
magnetically. Gas was supplied to the liquid containers through
2.4. Catalytic experiments
glass frits positioned close to the stirrer. The HPLC pump yielded
−1
flow rates in the range 0.2–0.5 ml min
( 1%). Pulse- or step-
Before the experiments, vessels 1 and 3 were filled with ap-
proximately 1 L of pure water and vessel 2 was filled with 1 L of
aqueous solution containing the reacting species (glucose or nitrite
type changes in the reactant concentrations were introduced by
switching HPLC valve 2 (G1158A, Agilent). The system was auto-
mated and controlled with Chemstation software, version b 02.01
(Agilent) [20].
−3
ions). An aqueous solution of NaOH (1×10 mol) was used to ad-
just pH of the glucose solution to 9 0.3, to prevent inhibition of
the catalysts with carboxylic acids [34]. The liquids in vessels 2 and
3 were flushed with argon to remove any dissolved gases (60 min,
200 ml min−1), whereas in vessel 1 water was saturated with oxy-
2.3.2. Reactors
Two types of catalytic reactors—a Pt/SiO2 fixed-bed reactor and
Pt/CNF/Ni supported on Ni foam—were used. To prevent channel-
ing of liquid, the cartridge was filled with the Pt/SiO2 catalyst
very carefully, following standard procedures for creating HPLC
columns [30]. The CNF foam was fitted tightly fitted into the car-
tridge, and the distance to the wall was in the same range as the
◦
gen or hydrogen (30 min, 200 ml min−1, 22 C). Experiments were
−1
◦
performed with at a flow rate of 0.5 ml min
at 60 C.
Pulse titration experiments with aqueous nitrite solution over
Pt/SiO2 were performed in the following sequence. First, water
saturated with hydrogen (from vessel 1) was flowed for 60 min
through the reactor to preadsorb hydrogen. The amount of hy-
drogen introduced was well in excess of that needed to reach
full coverage of the platinum surface. Then pure water (from ves-
sel 3) was flowed through the reactor for 30 min to remove dis-
solved hydrogen from the solution and physisorbed hydrogen from
Pt/SiO2. The results given in Part 2 of this paper (Fig. 2b) demon-
strate that 30 min is more than enough time to obtain complete
−4
pore diameter in the foam (1–5×10 m). The volume of the voids
within the reactors was estimated by weighing the reactors before
and after filling them with water. The estimated volume was used
to calculate residence time and was always within an error of 5%.
The dynamic behavior of the reactors (filled with silica support or
with CNF/Ni support without Pt) was determined by pulsing nitrite
−4
solution (concentration 2.2 × 10
mol dm−3, 50 ml for 100 min)
removal. Finally, the preadsorbed hydrogen was contacted with
−5
or glucose solution (concentration 5.6 × 10
mol dm−3, 12.5 mL
−4
pulses of nitrite solution (concentration 2.2×10
mol dm−3, 5 ml
for 25 min). An earlier study found that silica leached when al-
kaline glucose solution was flowed [30]; therefore, we used Pt
supported on CNF/Ni for the experiments with glucose.
for 10 min); pulses were repeated every 20 min, with pure water
flowed between pulses.
Another sequence was applied to study the interaction of
Pt/CNF/Ni with aqueous glucose solution. First, water saturated
with oxygen (from vessel 1) was flowed for 30 min (flow,
The catalytic reactor was placed in an oven made of alu-
minum blocks with two integrated heating fingers (G4A78, Wat-
low/Kurval). The temperature of the oven was kept uniform at
◦
0.5 ml min−1) at 60 C through the reactor to preadsorb oxygen.
◦
◦
60 C ( 0.5 C) by a temperature controller (Eurotherm 2132).
Then pure water (from vessel 1) was flowed through then re-
actor for 60 min to remove dissolved oxygen from the solution
and physisorbed oxygen from Pt/CNF/Ni. Finally, preadsorbed oxy-
2.3.3. Analyzer
Qualitative and quantitative analysis of ionic species present in
the effluent from the reactor was performed with an online mass
spectrometry detector (MSD) (SL/G1956B, Agilent) equipped with
an electrospray ionization (ESI) liquid-phase interface (G1948A, Ag-
ilent). The liquid chromatography part of the equipment was re-
moved, and the effluent from the reactor was directly injected into
the electron-spray ionization chamber.
gen was contacted with pulses of glucose solution (concentration
mol dm−3, 30 ml for 60 min). Pulses were repeated af-
−5
5.6 × 10
ter 20 min, with pure water flowed between pulses. Calibrations
for quantitative interpretation are described in next section.
3. Results and discussion
The ESI process involves three main steps [31]: (i) production
of charged droplets at the ES capillary tip; (ii) shrinkage of the
charged droplets by solvent evaporation, leading to very small
charged droplets capable of producing gas-phase ions; and (iii)
We first present and discuss the results of catalyst characteriza-
tion, then explore pressure drop, response time of the detector, and
the hydrodynamic behavior of the reactors. We then describe de-
tection and calibration of species with ESI-MS. Finally, we present

D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
247
−
−4
−3
Fig. 2. Response of (a) by-pass and (b) reactor with silica to the pulse of nitrite (50 mL during 100 min with flow rate 0.5 ml min−1, cM (NO2 ) = 2.2 × 10
mol dm
,
◦
T = 60 C). Response of (c) by-pass, (d) column with silica and (e) column with CNF to a pulse of glucose (12.5 ml during 25 min with flow rate 0.5 ml min−1, cM (C6H12 O6)
−5
◦
= 5.6 × 10
mol dm−3, T = 60 C).
Table 2
dicating that the assumption of the Pt loading of Pt/CNF/Ni is
reasonable. In the case of Pt/SiO2, the calculated dispersion based
on TEM was slightly lower than that according to chemisorption,
indicating that this catalyst may have contained some very small
particles (<2 nm) that could not be clearly observed with TEM.
Catalyst loadings and dispersions determined in gas phase
Catalyst
Loading
(wt%)
H/Pt
Amount of Pt
H/Pt–TEMa
Bulk (μmol/g)
Surface (μmol)b
Pt/SiO2
Pt/CNF
1.0
0.05
0.65
0.25
51
154
10.0
11.0
0.55
0.29
3.0c
1.08
a
Calculated based on: D(TEM) =
[29].
3.2. Pressure drop, response time of the detector, and hydrodynamic
behavior of the reactors
d(nm)
b
c
Amount of surface platinum atoms available in the reactor.
The Pt loading assumed based on preparation procedure.
−1
Flowing 0.5 ml min
over the Pt/SiO2 packed bed resulted in
the results of transient catalytic experiments, probing the reactiv-
ity of preadsorbed hydrogen with nitrite and preadsorbed oxygen
with glucose.
a typical pressure drop of 40 bar; 2–3 bars is typically due to the
two metal frits. After 1 week of operation, the pressure drop in-
creased slightly, from 40 to 43 bars, and remained constant there-
after. This effect may be due to further compaction of the bed.
For the same flow rate, the pressure drop over the foam increased
from 2 bars initially to 12 bars after 2 weeks of use. During the
subsequent 2 weeks of use, the pressure drop remained constant.
Inspection of the used reactor showed black CNF fines deposited
on the frit at the outlet. After the frit was cleaned, the pressure
drop recovered completely, with no new fines or pressure drop
were observed. Thus, a small part of the CNFs on Ni foam were
not well attached and could be removed by flowing compressed
air. This observation agrees well with the results of Jarrah et al.
[28,35] who, when describing the preparation of CNFs on Ni foam,
reported that a small fraction of the CNFs was loosely attached,
3.1. Catalyst preparation and characterization
The platinum loading of Pt/SiO2 (1 0.05 wt%) was in agree-
ment with expectations based on the preparation procedure used
(Table 2). Pt/CNF/Ni could not be analyzed with XRF; therefore,
the platinum loading according the preparation procedure (3 wt%)
was assumed. No contamination could be detected with XRF on all
catalysts before or after the experiments. Table 2 also reports the
platinum dispersion measured with hydrogen chemisorption (65%
for Pt/SiO2 and 25% for Pt/CNF/Ni). The platinum dispersions es-
timated from the average metal particle size obtained from TEM
analysis closely agreed with the hydrogen chemisorption data, in-

248
D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
−3
−
−4
◦
−1
Fig. 3. Spectrum obtained during analysis of 2 × 10
mol dm
nitrite (NO2 ) aqueous solution at 60 C and liquid flow rate of 0.5 ml min .
−
◦
−1
.
Fig. 4. Calibration curve for nitrite (ꢀ—m/zl 15 [Na(NO2)2] ) obtained at 60 C and liquid flow rate of 0.5 ml min
whereas the remaining CNFs was strongly bonded to the Ni foam
support.
adsorbed reversibly. The shape of the desorption curve indicates
two types of adsorbed glucose molecules. Weakly adsorbed glu-
cose desorbed first, reaching a stable concentration in the outlet at
a lower level when strongly adsorbed glucose desorbed. Whether
this was due to inhomogeneity of the silica surface or lateral inter-
action between glucose molecules remains unclear.
Fig. 2e shows the response of a reactor with CNF/Ni to a glucose
pulse. Stabilization of the signal was attained after 20 min (Fig. 2e),
much longer than the residence time of 3 min in the free volume
of the CNF/Ni reactor. Furthermore, the shape of the response sig-
nal indicates significant back-mixing in the CNF/Ni foam. Therefore,
longer pulses were required in this case (see Section 3.5).
The response time of the detector (i.e., the time required for
the signal to reach a level of 95% of the stable value) was on the
order of 20 s, as observed in the blank experiment by-passing the
reactor for both nitrite ions (Fig. 2a) and glucose (Fig. 2c). Fig. 2
presents the complete response to the pulse.
Fig. 2b shows the response to a nitrite pulse with the SiO2 fixed
bed. A delay of 1.45 min was observed, in good agreement with
the residence time in the reactor (1.40 min) based on the free vol-
ume in the reactor. The breakthrough curve from the reactor with
silica was sharp, and the signal stabilized within 20 s (Fig. 2b).
The slight decrease in signal during the plateau, as well as the
small step after 40 min, were due to process disorder and were
not significant. Results from the same experiment with glucose
and SiO2 are shown in Fig. 2d. Surprisingly, a delay of 6.10 min
was seen at both the front and back ends of the pulse, whereas
the residence time remained 1.4 min. The difference between the
step response in the blank experiment (Fig. 2c) and in the experi-
ment with silica (Fig. 2d) indicates that about 0.1 μmol of glucose
3.3. Calibration
Fig. 3 shows a typical MS spectrum for a sodium–nitrite so-
lution revealing significant peaks at m/z 46, 115, and 184. Mass
−
m/z 46 corresponds to the NO2 anion, whereas m/z 115 and
184 are agglomerates of nitrite and sodium ions [36], that is,
−
−
+ −
+ −
) , respectively. The most in-
([2NO2 ]Na
)
and ([3NO2 ]2Na

D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
249
−3
−5
◦
−1
Fig. 5. Glucose spectrum for 5.6 × 10
mol dm
glucose solution at 60 C and liquid flow rate of 0.5 ml min .
−
◦
−1
.
Fig. 6. Calibration curve for glucose (m/z 215 − 2H2O·[C6H11 O6] ) obtained at 60 C and liquid flow rate of 0.5 ml min
−
tense peak (i.e., m/z 115) can be used for quantitative determi-
nation, because intensity increases regularly with concentration
(Fig. 4); surprisingly, the intensities of m/z 46 and m/z 184 de-
creased with increasing concentration. The calibration line was
calculated using linear regression (R2 = 0.977). This indicates that
the m/z values generally must be selected carefully based on cal-
ibration experiments. The concentration of nitrite could be ac-
curately determined in the concentration window between 2 ×
molecules of water (2H2O·[C6H11 O6] ) and is the most suitable
for quantitative determinations, because the calibration curve for
this peak increased linearly with increasing glucose concentration
(Fig. 6). The calibration line was calculated using linear regression
(R2 = 0.988). The peak at m/z 179 corresponded to [C6H11 O6]
−
(i.e., deprotonated glucose molecule without water); this signal
could not be used here, because it did not increase regularly with
increasing glucose concentrations. We cannot rule out the pres-
ence of gluconic acid contamination because of the peak at m/z
195; however, this signal also could be due to fragmentation of
glucose during ionization. The limit of detection for glucose was
−3
−4
−5
10
mol dm
(detection limit with signal/noise >5) and up to
about 5 × 10
mol dm−3; at higher concentrations, the MS sig-
nal showed little further increase (Fig. 4). This is a saturation
effect [37] caused by a decrease of the ionization efficiency. Am-
monia is, along with nitrogen, a product from the hydrogenation
of nitrite. Unfortunately, ammonia could not be detected due to
the low molecular mass. On the other hand, we verified that the
presence of ammonia did not affect the calibration of nitrite.
Fig. 5 shows a typical spectrum for glucose. The most signifi-
cant peaks can be seen at m/z 215 and 179. The peak at m/z 215
corresponds to the deprotonated glucose ion with two associated
5.6×10 mol dm−3, and the glucose concentration could be accu-
−6
−4
rately determined up to 1 × 10
mol dm−3, before the saturation
effect occurred (Fig. 6).
Calibration also was done for gluconic acid (Fig. 7); the most
significant signal observed was at m/z 195, corresponding to de-
−
protonated gluconic acid [C6H11 O7] anion. The calibration line
was calculated using linear regression (R2 = 0.952). Gluconic acid
−6
could be accurately determined between 1 × 10
up to 1 ×

250
D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
−
◦
−1
.
Fig. 7. Calibration curve for gluconic acid (m/z 195 − [C6H11 O7] ) obtained at 60 C and liquid flow rate of 0.5 ml min
10
10
mol dm−3. Experiments with mixtures of glucose (5.6 ×
in our experiment (i.e., 2.4 μmol) lies between these boundary
values, indicating formation of both ammonia and nitrogen. Un-
fortunately, these species could not be detected by MS because of
their low molecular weight and stability compared with the rel-
atively soft ESI ionization method. Nevertheless, the sensitivity of
the MS detector allows quantitative determination of nitrite con-
version caused by hydrogen coverage as low as 3%.
−4
−3
mol dm ) and gluconic acid (5.1 × 10
mol dm−3) revealed
−5
−5
some charge competition; that is, the glucose concentration was
underestimated, whereas the gluconic acid concentration was over-
estimated. The error in both cases was a maximum of 20%, which
we currently accept as the limit of experimental inaccuracy at this
stage of development.
Calibrations were repeated every 2 weeks, resulting in repro-
ducible results with 10% error. Applying the ESI-MS detector in the
continuous-flow mode caused frequent blockage of the nebulizing
needle, due to deposition of dissolved species on the wall of the
needle (especially during the analysis of nitrite). This was man-
ifested by the pressure rise in the ESI (2–3 bars after 1 month
of operation). This effect was prevented by daily cleaning and re-
placement of the nebulizing needle when pressure started to rise.
Fig. 8b shows the details of the responses to the nitrite pulse
after all of the hydrogen was consumed. The striking difference
in the shapes of the responses for Pt/SiO2 and the support only
clearly indicates a reversible and significant interaction of dis-
solved nitrite with the platinum surface. The step response indi-
cates that about 0.05 μmol of nitrite adsorbed. This must have
occurred on the Pt surface, because the results shown in Fig. 8b
demonstrate the difference between the catalyst and the support
only. The step-down response was less clear, in the sense that
the amount of nitrite desorbed seemed smaller than the amount
adsorbed; probably desorption occurred over a long period, after
which the resulting nitrite signal could not longer be distinguished
from the baseline. A strong interaction of nitrite with Pt was re-
ported by Ebbesen et al. [39] in an ATR-IR study on adsorption of
3.4. Catalytic reduction of nitrite species over Pt/SiO2
Fig. 8a shows the experimental result of pulsing nitrite for
both the prereduced Pt/SiO2 catalyst and for the silica support
only (which served as a blank experiment). Complete consumption
of nitrite occurred during the first pulse on the reduced Pt/SiO2.
A breakthrough occurred during the next three pulses. Finally, the
surface area of the peaks became identical to that seen in the
blank experiment, indicating no further hydrogen consumption.
The total amount of nitrite converted was 2.4 μmol.
−
NO2 on Pd/Al2O3 and Pt/SiO2. Along with strong adsorption, our
findings also indicate that reversible adsorption occurred as well.
The amount was equivalent to 0.5% of a monolayer on platinum.
−
We speculate that NO2 interacted exclusively with a small frac-
tion of specific sites at the surface of the Pt particles (e.g., corner
sites). The remarkable sensitivity of the detector should be recog-
nized at this point, allowing quantitative determination of nitrite
adsorption of <0.01 monolayer (ML).
Selective hydrogenation of nitrite with hydrogen preadsorbed
on platinum under the present experimental conditions resulted in
formation of nitrogen and ammonia, according the following equa-
tions [38]:
We observed a lower conversion of nitrite (i.e., 1.8 μmol) in
subsequent NO2 titration experiments, which implies a loss of
−
−
2NO2 + 6Ha = N2 + 2OH + 2H2O
(1)
−
platinum surface area. Furthermore, XRF analysis of the Pt/SiO2
catalyst after three titration experiments showed a decrease in Pt
loading (from 1.0 to 0.87 wt%). Hydrogen chemisorption of the
Pt/SiO2 catalyst after the titration experiments revealed a decrease
in Pt dispersion from 0.65 to 0.58. These effects are possibly due to
leaching of the smallest metal particles during the titration experi-
ments; this would account for both the loss of Pt and the decrease
in dispersion. Similar effects also have been reported by Douidah
et al. [40] for small silica-supported platinum particles in aqueous
media under hydrogen atmosphere.
and
−
+
−
NO2 + 6Ha = NH4 + 2OH
.
(2)
The total amount of hydrogen available, adsorbed on the Pt cat-
alyst, was 10 μmol, assuming one hydrogen atom on every sur-
face Pt atom (Table 2). Accordingly, the amount of nitrite that
could be converted if ammonia were the sole product [Eq. (2)] is
1.7 μmol (maximal consumption of Ha). If nitrogen were the sole
product [Eq. (1)], then the amount of converted nitrite would be
3.3 μmol (minimal consumption of Ha). The conversion observed

D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
251
(a)
(b)
−3
−
−4
◦
Fig. 8. (a) Titration of pre-adsorbed hydrogen on Pt/SiO2 using pulses with nitrite aqueous solution (10 min long, 0.5 ml min−1, cM = 2 × 10
mol dm
NO2 , T = 60 C)
over: (A) Reactor filled with silica and (B) reactor filled with Pt/SiO2 catalyst with pre-adsorbed hydrogen. (b) Titration of pre-adsorbed hydrogen on Pt/SiO2 using pulses
−3
−
−4
◦
with nitrite aqueous solution (10 min long, 0.5 ml min−1, cM = 2 × 10
mol dm
NO2 , T = 60 C) detail from pulse 7: (A) Reactor filled with silica and (B) reactor filled
with Pt/SiO2 catalyst with pre-adsorbed hydrogen.
In Part 2 of this paper we describe the development and appli-
cation of a detector similar to membrane inlet mass spectrometry
(MIMS) for online and real-time analysis of dissolved gases like
N2, to compensate for the ESI-MS detector’s unsuitability in this
respect.
193, and 209 that were not found in the blank experiment, cor-
responding to fragments of side or intermediate products dur-
ing glucose oxidation. These peaks can be assigned to glucose
−
−
dialdehyde ([C6H9O6] at m/z 177) glucuronic acid ([C6H9O7]
−
at m/z 193), and glucaric acid ([C6H9O8] at m/z 209), respec-
tively [41].
Fig. 9b shows the results of repetitive pulsing with glucose on
the preoxidized Pt/CNF/Ni catalyst. The left side of the figure com-
pares the amount of remaining glucose (B) with the amount of
glucose in the blank experiment in the absence of Pt (CNF-only,
curve A); curve C shows the amount of gluconic acid formed.
Clearly, glucose was converted to gluconic acid, and both the con-
version and yield of gluconic acid decreased on repetitive pulsing,
approaching zero on the fifth pulse. A total of 8.2 μmol of glu-
cose was converted, whereas 5.6 μmol of gluconic acid was formed.
The right side of Fig. 9b shows the response signal for the species
3.5. Catalytic oxidation of glucose over a Pt/CNF/Ni supported on Ni
foam
Fig. 9a shows the MS spectrum of the reactor effluent when
a glucose solution was pulsed over Pt/CNF/Ni foam containing
preadsorbed oxygen. A major part of the glucose was clearly con-
verted, as demonstrated by the much lower peaks at m/z 215 and
m/z 179 compared with those in the blank experiment shown in
Fig. 5. The most significant peak in Fig. 9a, at m/z 195, was due
to gluconic acid. Furthermore, peaks could be seen at m/z 177,

252
D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
(a)
(b)
−1
Fig. 9. (a) Spectrum from the first pulse during titration of pre-adsorbed oxygen on Pt/CNF catalyst with glucose from aqueous solution (pulse of 30 ml, 0.5 ml min
,
−3
−5
◦
cM = 5.6 × 10
mol dm , T = 60 C). (b) Titration of oxygen pre-adsorbed on Pt/CNF supported on Ni-foam catalyst with the pulses of glucose aqueous solution (pulse of
−3
−5
◦
30 ml, 0.5 ml min−1, cM = 5.6 × 10
mol dm , T = 60 C) reactor with CNF/Ni (A) glucose; reactor with Pt/CNF/Ni (B) glucose; (C) gluconic acid; (D) glucose dialdehyde;
(E) glucuronic acid; (F) glucaric acid.

D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
253
needed to decrease the level of conversion to obtain the reaction
order in adsorbed oxygen.
The novel technique introduced here performs better than ex-
isting transient techniques. The most significant advantage is the
ESI-MS detector’s ability to distinguish between different compo-
nents, as was clearly demonstrated in the case of glucose oxida-
tion. In contrast, RI and UV–vis detectors can follow only single
compounds. In principle, ATR-IR also can distinguish between dif-
ferent compounds; however, IR absorption peaks must be well
separated, which generally is not the case, and sensitivity is signif-
icantly lower. The response time of the detector allows detection
in the range of 20 s, which is significantly faster then typical re-
sponses of the RI (range, 40 s for flows of 0.2–3 mL min−1) or FTIR
(range, 5 min for flows of 3–7 mL min−1), underscoring the advan-
tages of ESI-MS. The typical concentrations detected in our tran-
Fig. 10. Glucose oxidation reaction rate as function of the maximum estimate of the
oxygen surface coverage (θ).
sient experiments (10−6–10
mol dm−3) were significantly lower
−4
than those that can be detected by RI (10−4–10
mol dm−3) or
mol dm−3). The experiment with nitrite detected ni-
−2
−2
by FTIR (10
with m/z 177, 193, and 209. Glucose dialdehyde and glucuronic
acid clearly were formed; however, it should be noted that intensi-
ties of these signals were much lower than those for gluconic acid.
Intensities decrease during repetitive pulsing. The peculiar differ-
ences in the shape of the first peak indicate that glucuronic acid
was formed especially at very high oxygen coverage. It is not pos-
sible to discuss the shapes of the peaks in more detail, because
chromatographic effects of the support are likely, as demonstrated
by the fact that glucose adsorbs reversibly on silica. The mech-
anistic relevance of this observation is beyond the scope of this
paper. The intensity of the signal due to glucaric acid was even
lower, but the intensity remained significant, at least in the first
few pulses.
trite that visibly interacted with 0.01 ML Pt, demonstrating the
advantages of the ESI-MS detector. Clearly, application of ESI-MS
detection improved the sensitivity of the pulse-experiment signifi-
cantly.
In summary, in both the experiments with glucose (converting
adsorbed oxygen) and those with nitrite (converting adsorbed hy-
drogen), our technique allowed the detection of adsorbed species
on Pt surface areas as small as 0.01 m2. This remarkable sensitivity
is very suitable for experiments using small amounts of catalysts
in general, and thus our technique is promising for microreactors.
We have demonstrated this for operation in the aqueous phase. In
principle, application in the organic phase also is possible; how-
ever, the ESI-MS detector would need to be optimized for each
specific application.
The main reaction during selective oxidation of glucose can be
described by the following equation:
4. Conclusion
+ 1/2O2 =
(3)
We have reported a novel, transient response technique for
liquid-phase heterogeneous catalytic studies that was applied suc-
cessfully for the first time. Our method is based on an online
method for simultaneous detection of species dissolved in an aque-
ous stream at the exit of a catalytic reactor. ESI-MS was used for
detection of dissolved molecules that readily ionize, as well as dis-
solved ions. Two test reactions, nitrite reduction with Pt/SiO2 and
glucose oxidation with Pt/CNF/Ni, were used to demonstrate semi-
quantitative monitoring of reactants, intermediates, and products.
The ESI-MS detector was demonstrated to be sufficiently sensitive
to quantitatively determine very small amounts of physisorbing ni-
trite, down to 0.5% of a monolayer on the Pt surface. Nitrite also
reacts with preadsorbed hydrogen, and the quantitative experi-
mental results agree with the finding that both nitrogen and am-
monia were formed. The ESI-MS detector is able to distinguish be-
tween different components simultaneously, as was demonstrated
for glucose oxidation, providing its most significant advantage over
existing transient techniques.
The amount of oxygen available in the reactor was 10 μmol Oa
(Oa:Pts = 1), and thus a maximum of 10 μmol of glucose could
be converted to gluconic acid. The amount of glucose converted
in our experiment (i.e., 8.2 μmol) was slightly less than this, and
the amount of gluconic acid formed (i.e., 5.6 μmol) was even
lower. The discrepancy between glucose converted and gluconic
acid formed (i.e., 8.2 vs 5.6 μmol) is due, at least in part, to ioniza-
tion suppression and charge competition effects between glucose
and gluconic acid during ionization in ESI. In other words, calibra-
tion is influenced by the presence of the other species. Conversion
to other products also may contribute somewhat, given the fact
that traces of other compounds were detected, as discussed ear-
lier.
Fig. 10 shows the relationship between the relative average re-
action rate of glucose oxidation in each pulse (calculated based on
the conversion level in each pulse) and the oxygen surface cover-
age. It is assumed that Oa:Pts is equal to unity initially and that
each glucose molecule reacts with only one preadsorbed oxygen
atom. It should be noted that this procedure overestimates the
oxygen coverage, because deep oxidation products also are formed
(Fig. 9b). Unfortunately, quantifying these products is not possible;
therefore, the oxygen surface coverages shown in Fig. 10 are just
estimates, and the lowest oxygen coverage is probably <0.5 ML
and may well be close to 0. Nevertheless, Fig. 10 shows decreasing
reaction rates with decreasing oxygen coverage. This result demon-
strates that in principle, the effect of O surface coverage on the
reaction rate can be determined; however, a higher flow rate is
Acknowledgments
The authors thank L. Vrielink for XRF and BET analysis, M.
Smithers for TEM measurements, A. Hovestad for assistance with
reactor preparation, and J. Spies and B. Geerdink for technical as-
sistance. Financial support for the project (TPC 5694) by STW, The
Netherlands is kindly acknowledged. This work was performed un-
der the auspices of NIOK.

254
D. Radivojevi´c et al. / Journal of Catalysis 257 (2008) 244–254
References
[21] D. Radivojevic´, K. Seshan, L. Lefferts, Appl. Catal. A 301 (2006) 51.
[22] N.A. Jarrah, J.G. van Ommen, L. Lefferts, J. Catal. 239 (2006) 460.
[23] S. Ebbesen, Spectroscopy under the surface—In-situ ATR-IR studies of heteroge-
neous catalysis in water, Ph.D. thesis, University of Twente, 2007.
[24] A. Pintar, Catal. Today 77 (2003) 451.
[1] C.G.M. van de Moesdijk, The catalytic reduction of nitrate and nitric oxide to
hydroxylamine: Kinetics and mechanism, Ph.D. thesis, University of Eindhoven,
1979.
[25] M. Besson, P. Gallezot, Catal. Today 57 (2000) 127.
[26] H. Saito, O. Shinji, F. Shigeo, US Patent 4 843 173 (1989), to Kawaken Fine
Chemical Corp. Ltd and Kao Corp.
[2] J.H. Clark, Pure Appl. Chem. 73 (2001) 103.
[3] J.W. Niemantsverdriet, Spectroscopy in Catalysis: An Introduction, Wiley–VCH,
2000, p. 5.
[27] K. Deller, H. Krause, E. Peldszus, B. Despeyroux, US Patent 5 132 452 (1992), to
Degussa.
[28] N.A.K.A. Jarrah, F. Li, J.G. van Ommen, L. Lefferts, J. Mater. Chem. 15 (2005)
1946.
[29] C.H. Bartholomew, R.J. Farrauto, Fundamentals of Industrial Catalytic Processes,
Wiley–Interscience, 2006, p. 146.
[30] K.K. Unger, J. Chromatogr. Libr. 16 (1994) 53.
[31] P. Kebarle, Y. Ho, in: R.B. Cole (Ed.), Electrospray Ionization Mass Spectro-
metry—Fundamentals, Instrumentation and Applications, J. Wiley & Sons, Inc.,
1997, p. 6.
[32] Y. Cai, M.C. Concha, J.S. Murray, R.B. Cole, J Am. Soc. Mass. Spectrom. 13 (2002)
1360.
[4] H. Kobayashi, M. Kobayashi, Cat Rev. Sci. Eng. 10 (1974) 139.
[5] C.O. Bennet, M.B. Cutlip, C.C. Yang, Chem. Eng. Sci. 27 (1972) 2255.
[6] A.T. Bell, L.L. Hegedus (Eds.), ACS Symp. Ser., vol. 178, ACS, 1982, pp. 1–308.
[7] J.T. Gleaves, J.R. Ebner, T.C. Kuechler, Cat. Rev. Sci. Eng. 30 (1988) 49.
[8] G.S. Yablonsky, M. Olea, G.B. Marin, J. Catal. 216 (2003) 120.
[9] T.A. Nijhuis, L.J.P. van den Broeke, M.J.G. Linders, J.M. van de Graaf, F. Kapteijn,
M. Makkee, J.A. Mouljin, Chem. Eng. Sci. 54 (1999) 4423.
[10] T. Bürgi, A. Baiker, J. Phys. Chem. B 106 (2002) 10649.
[11] S.D. Ebbesen, B.L. Mojet, L. Lefferts, J. Catal. 246 (2007) 66.
[12] J.F. Denayer, A. Bouyermaouen, G.V. Baron, Ind. Eng. Chem. Res. 37 (1998) 3691.
[13] G.H. Jonker, Hydrogenation of edible oils and fats, Ph.D. thesis, Rijksuniversiteit
Groningen, 1999.
[33] S. Gao, Z.-P. Zhang, H.T. Karnes, J. Chromatogr. B 825 (2005) 98.
[34] T. Mallat, A. Baiker, Catal. Today 19 (1994) 247.
[35] N.A.K.A. Jarah, Microstructured catalyst support based on carbon nano-fibers
(CNFs), Ph.D. thesis, University of Twente, 2004.
[14] V. Hejtmánek, P. Schneider, Chem. Eng. Sci. 49 (1994) 2575.
[15] Y. Lin, Y.H. Ma, Ind. Eng. Chem. Res. 28 (1989) 622.
[16] Z. Kiraly, A. Mastalir, A. Csaszar, H. Demir, D. Uner, G.H. Findenegg, J. Catal. 245
(2007) 267.
[17] F. Gao, K.P. Ng, C. Li, K.I. Krummel, A.D. Allian, M. Garland, J. Catal. 237 (2006)
49.
[36] M.J. van Stipdonk, D.R. Justes, C.M. Force, E.A. Schweikert, Anal. Chem. 72
(2000) 2469.
[18] D.K. Lloyd, in: W.J. Lough, I.W. Wainer (Eds.), High Performance Liquid Chro-
matography: Fundamental Principles and Practice, Blackie Academic & Profes-
sional, 1996, pp. 120–133.
[19] D.C. Harris, Quantitative Chemical Analysis, W.H. Freeman and Co., 2003, p. 517.
[20] http://www.agilent.com.
[37] C.G. Enke, Anal. Chem. 69 (1997) 4885.
[38] M. D’Arino, F. Pinna, G. Strukul, Appl. Catal. B 53 (2004) 161.
[39] S.D. Ebbesen, B.L. Mojet, L. Lefferts, Langmuir 24 (2008) 869.
[40] A. Douidah, P. Marécot, J. Barbier, Appl. Catal. A 225 (2002) 11.
[41] S. Hermans, M. Devillers, Appl. Catal. A 235 (2002) 253.