Ceramic microreactors for on-site hydrogen production

Article abstract of DOI:10.1016/j.jcat.2006.04.033

This paper describes the synthesis and characterization of ceramic microreactors composed of inverted beaded silicon carbide (SiC) monoliths with interconnected 0.75-, 2.2-, or 7.2-μm pores as catalyst supports, integrated within high-density alumina reactor housings obtained via an optimized gel-casting procedure. Structural characterization revealed that these tailored macroporous SiC porous monoliths are stable at temperatures up to at least 1200 °C, and have surface areas and porosities as high as 7.4 × 10⁷ m² / m³ and 74%, respectively. Further characterization of the ceramic microreactors using the decomposition of ammonia with Ru as the catalyst at temperatures between 450 and 1000 °C showed that as much as 54 sccm of hydrogen, or 9.8 × 10⁴ sccm H₂ per cm³ of monolith volume, could be obtained from a 36-sccm entering stream of NH₃ at >99.9% conversion at temperatures above 700 °C. Moreover, using SiC as a catalyst support appears to increase the catalytic activity of the Ru catalyst, as evidenced by high turnover frequencies.

Full text of DOI:10.1016/j.jcat.2006.04.033

Journal of Catalysis 241 (2006) 235–242
www.elsevier.com/locate/jcat
Ceramic microreactors for on-site hydrogen production
Christian a, M. Mitchell a, D.-P. Kim b, P.J.A. Kenis a,∗
a
Department of Chemical & Biomolecular Engineering, University of Illinois at Urbana-Champaign, 600 S. Mathews Ave., Urbana, IL 61801, USA
b
Department of Fine Chemical Engineering & Chemistry, Chungnam National University, Daejeon, 305-764, South Korea
Received 19 January 2006; revised 6 April 2006; accepted 8 April 2006
Available online 5 June 2006
Abstract
This paper describes the synthesis and characterization of ceramic microreactors composed of inverted beaded silicon carbide (SiC) monoliths
with interconnected 0.75-, 2.2-, or 7.2-µm pores as catalyst supports, integrated within high-density alumina reactor housings obtained via an
optimized gel-casting procedure. Structural characterization revealed that these tailored macroporous SiC porous monoliths are stable at tempera-
◦
7
2
3
tures up to at least 1200 C, and have surface areas and porosities as high as 7.4 × 10 m /m and 74%, respectively. Further characterization of
◦
the ceramic microreactors using the decomposition of ammonia with Ru as the catalyst at temperatures between 450 and 1000 C showed that as
4
3
much as 54 sccm of hydrogen, or 9.8 × 10 sccm H per cm of monolith volume, could be obtained from a 36-sccm entering stream of NH at
2
3
◦
>99.9% conversion at temperatures above 700 C. Moreover, using SiC as a catalyst support appears to increase the catalytic activity of the Ru
catalyst, as evidenced by high turnover frequencies.
© 2006 Elsevier Inc. All rights reserved.
Keywords: Ceramic microreactor; Silicon carbide; Hydrogen production; Ammonia decomposition; Fuel reforming
1. Introduction
The decomposition of NH3 within a microreactor is a direct
route to producing H2 for PEM fuel cells. The only products
are H2 and N2, and this microreactor can be integrated with a
microscale heat source, such as a microburner [7] or a catalytic
combustor [8], making it advantageous for portable power ap-
plications due to the small overall volume. Compared to steam
reforming of liquid hydrocarbons, the decomposition of NH3
has fewer issues. Steam generation is not necessary, and no
CO is produced; therefore, additional water–gas shift reactors
and preferential oxidation reactors are not needed to avoid CO
poisoning of the fuel cell catalyst [9]. Furthermore, the poten-
tial for sulfur poisoning of the reforming catalyst is avoided,
eliminating the need for sulfur removal units or sulfur-tolerant
catalysts [9]. A limitation still exists because NH3 can poison
the catalyst used in PEM fuel cells; however, an NH3 adsorbent
can be used to reduce the ammonia concentration to ppb levels
on entering a PEM fuel cell [10].
The continuous operation of electronic devices lacking ac-
cess to wired electrical power, such as vehicles and electrical
equipment in remote locations (e.g., construction sites, military
fields), requires electrical power sources of high specific energy.
Rechargeable lithium ion batteries are presently used for many
of these applications, but they have energy densities of only
up to ∼200 W-h/l and thus must be recharged frequently [1].
Higher energy densities can be obtained by using hydrogen gas
in polymer electrolyte membrane (PEM) fuel cells [2–4]. The
transport and storage of compressed hydrogen entails safety is-
sues, however. By producing H2 on-site [5,6] by, for example,
decomposition of ammonia (energy density of ∼4 kW-h/l) or
steam reforming of liquid hydrocarbons such as gasoline (en-
ergy density of ∼9 kW-h/l), fuel cells can provide electrical
power for longer periods and have a higher specific energy than
rechargeable batteries [2].
Several microreactors for on-site production of H2 have been
reported [1,2,11–15]. Channel and posted reactors fabricated
from anodized aluminum have been used for the decomposi-
tion of NH3 up to 650 ◦C [11–13]. Integrated reformer sys-
tems consisting of burner, vaporizer, and reformer units have
*
Corresponding author. Fax: +1 217 333 5052.
E-mail address: kenis@uiuc.edu (P.J.A. Kenis).
0021-9517/$ – see front matter © 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2006.04.033

236
Christian et al. / Journal of Catalysis 241 (2006) 235–242
been fabricated of stainless steel for the steam reforming of
methanol between 260 and 450 ◦C [1,2,14]. In addition, MEMS
fabrication methods have been used to produce silicon-based
microreactors for methanol reforming [15]. Although these mi-
croreactor systems have achieved high conversion of NH3 and
methanol into hydrogen, they are often not stable for continu-
ous operation above 800 ◦C, the minimum temperature required
to eliminate coking of the catalyst during the steam reform-
ing of higher hydrocarbons, such as propane and gasoline [16].
Aluminum melts at 660 ◦C, and silicon and stainless steel ox-
idize in the presence of steam above 800 ◦C and can corrode
to a significant extent. Jensen et al. recently reported the more
promising solution of a microreactor composed of suspended
silicon nitride tubes with wash coats of catalyst to enable am-
monia decomposition up to 825 ◦C [8].
To date, the development of high-surface area catalyst sup-
port structures that are stable above 800 ◦C has been a chal-
lenge, especially when the pressure drop across the support
must be low enough to both reduce parasitic losses due to
pumping of fluids and prevent mechanical failure of the struc-
tures. Recently, we have reported on the fabrication of inverted
beaded monoliths with 7.2-µm interconnected pores, composed
of silicon carbide (SiC) or silicon carbonitride (SiCN) [17],
which are stable up to 1200 ◦C in air and have a ∼2 or-
der of magnitude lower pressure drop than a packed beaded
structure with the same geometric surface area. These mono-
liths also avoid the issues of cracking and channeling of the
reactants, problems commonly encountered with wash coat-
based catalyst layers and packed beds of loose catalyst parti-
cles.
In this paper, we describe the synthesis and characterization
of ceramic microreactors composed of SiC catalytic monoliths
with 0.75-, 2.2-, and 7.2-µm pores integrated in high-density
alumina microreactor housings that we obtained using an op-
timized gel-casting procedure. We report a detailed analysis
of the structural properties of the SiC catalyst supports (sur-
face area, porosity, thermal stability) and of the deposited Ru
metal on these supports (catalyst loading, dispersion, surface
area of active sites). We further characterize these catalytic
monoliths using the decomposition of ammonia into hydro-
gen and nitrogen at temperatures as high as 1000 ◦C, to show
their promise in on-site hydrogen production for PEM fuel
cells.
lutions of polystyrene (PS) spheres (1.1, 3.2, and 10 µm diam-
eter; Polysciences) were centrifuged three times at 10,000 rpm
for 10 min (centrifuge 5415D, Eppendorf), with decanting of
the water and addition of DI water after each centrifugation to
remove surfactant. The final concentrated solution was obtained
by removing half of the water after settling of the solution for
12 h. About 5–50 µL of the concentrated PS sphere solution
was then placed at one end of a microchannel (500 µm wide,
150 µm high, 5–7 mm long) and left for 12 h for the comple-
tion of the packing process, followed by drying under vacuum
at 70 ◦C for 48 h.
SiC porous monoliths were prepared by infiltration of the re-
sulting packed beds of PS spheres in PDMS microchannels with
a mixture of allylhydridopolycarbosilane (SP matrix, Starfire
Systems) and 3–5 wt% of the thermal initiator 1,1-bis(tert-
butylperoxy)-3,3,5-trimethylcyclohexane (92%, Aldrich), fol-
lowed by curing of the precursor mixture at 70 ◦C for 12 h. All
of these steps were carried out in a glove box under an argon at-
mosphere. After the PDMS mold was removed, pyrolysis was
performed at 1200 ◦C for 2 h under an argon atmosphere, as de-
scribed previously [17], to yield SiC porous monoliths typically
350 µm wide, 100 µm high, and 3 mm long after shrinkage, as
shown in Fig. 1.
2.1.2. Catalyst deposition
Ruthenium (Ru) catalyst was deposited on the SiC porous
monoliths by wet impregnation with 14.72 wt% (0.67 M) RuCl3
(Aldrich) in a mixture of 10 vol% DI water in acetone, simi-
lar to a method described previously [11]. The structures were
then dried in a conventional oven at 60 ◦C for 48 h to remove
the remaining water and acetone, followed by calcination in
air at 580 ◦C for 5 h using a tube furnace (HTF5500 series,
Lindberg/Blue M). Finally, the catalyst was reduced in the tube
furnace under an atmosphere of 5% H2 in argon at 550 ◦C for
6 h.
2.2. Characterization of the porous catalytic monoliths
The morphology of the pores in the SiC porous monoliths
was studied using either a Hitachi S-4700 or a JEOL 6060-
LV scanning electron microscope. The surface areas of the
SiC porous monoliths were determined by N2 adsorption–
desorption at 77 K using a BET apparatus (Micromerit-
ics ChemiSorb 2705). Before the BET analysis, the sam-
ples were degassed under vacuum at 100 ◦C for 24 h to
remove any moisture. Mercury intrusion porosimetry (MIP)
analysis was carried out using a Micromeritics AutoPore II
9220 to determine the porosities of the SiC porous mono-
liths.
2. Experimental
2.1. Catalyst preparation
2.1.1. High-surface area porous monoliths (catalyst supports)
preparation
The dispersion of active metal phase of the Ru catalyst
in the SiC porous monoliths was measured using pulsed CO
chemisorption and H2 chemisorption at room temperature us-
ing a Micromeritics ChemiSorb 2705, assuming a 1:1 ratio of
CO or a 1:2 ratio of H2 chemisorbed to exposed Ru metal. The
Ru loading was determined using inductively coupled plasma
spectrometry (OES Optima 2000 DV, Perkin–Elmer).
Poly(dimethylsiloxane) (PDMS) molds with microchannel
patterns were prepared for fabrication of porous monoliths and
for molding of the high-density alumina housings (Section 2.3)
by replica molding of a master obtained through photolithog-
raphy [18]. For the preparation of SiC porous monoliths, we
adopted the micromolding-in-capillaries (MIMIC) method used
previously for the synthesis of porous oxide materials [19]. So-

Christian et al. / Journal of Catalysis 241 (2006) 235–242
237
Fig. 2. Polymer removal and sintering temperature sequence used to obtain
high-density, nonporous alumina housings and lids. The heating and cooling
rates were optimized to avoid cracking and deformation in the final alumina
structures.
Aldrich), and 1-vinyl-2-pyrrolidinone (99%, Aldrich) in a
1.5:1:1.5 ratio by weight, were first dissolved in DI water to
form a 20 wt% aqueous solution of monomers. The dispersant
(Darvan 821A, R.T. Vanderbilt) was then added to the mixture
in the amount of 2 wt% of the total powder used. A staged
addition of powder to the mixture was required to obtain a
ceramic slurry with a high alumina loading (here 50 vol%),
as described previously [22]. The slurry was then separated
from the milling media using a sieve, placed in an ice bath to
prevent solvent evaporation, and degassed in a vacuum desic-
cator for up to 2 h to remove any air bubbles. N,N,Nꢀ,Nꢀ-
tetramethylethylenediamine (TEMED, 99%, Aldrich) and am-
monium persulfate (APS, 99.99%, Aldrich) were added to the
slurry (0.1 vol% of TEMED and 0.1 wt% of APS with respect
to the monomer solution used), followed by stirring and deair-
ing in a vacuum desiccator for 30 min while the slurry was
placed in an ice bath.
The alumina slurry was then poured into PDMS molds
with microchannel patterns (see Section 2.1.1) that were
treated for 12 h with (tridecafluoro-1,1,2,2-tetrahydrooctyl)-
1-trichlorosilane (United Chemical Technologies) after being
cleaned by a plasma sterilizer (PDC-001, Harrick). Typical di-
mensions of the channels in the PDMS mold were 450 µm wide,
170 µm high, and 1.2 or 3.4 mm long; after shrinkage, the fi-
nal channel dimensions were 400 µm wide, 150 µm high, and
1 or 3 mm long. After the slurry completely solidified in the
mold due to catalytic polymerization of monomers, the result-
ing structures, the so-called “green bodies,” were demolded and
dried in a vacuum oven (5831 E series, Napco) at room tem-
perature. The relative humidity inside the oven was initially
maintained above 90% for 2 days, and then reduced by 10%
every other day until 50% was reached. The green bodies were
dried further under ambient conditions for two additional days.
Debinding (i.e., removal of polymer) and sintering of the green
bodies in a high-temperature furnace (1730 HT(c), CM fur-
naces) in air according to an optimized processing procedure
(Fig. 2) yielded high-density, nonporous alumina structures:
a microreactor design with slots for the placement of porous
Fig. 1. SEM micrographs of fracture profiles of inverted beaded SiC monoliths
after heating at 1200 C for 6 h in air with (a) 7.2 µm pores (inset: an optical mi-
crograph of the whole SiC porous monolith); (b) 2.2 µm pores; and (c) 0.75 µm
pores. The black spots are the windows that connect adjacent pores in these
close packed structures. The bits of debris are due to cutting of the samples
before SEM analysis.
◦
2.3. Fabrication of high-density alumina structures
The gel-casting forming method developed by Young et al.
for mesoscale structures [20,21] was adapted here to enable the
fabrication of cm-scale, high-density, nonporous alumina struc-
tures with submillimeter features for reactor housings and lids.
High-purity alumina powder (Baikalox GE1, Baikowski) was
initially deagglomerated using a jar mill (LABMILL-8000, Ad-
vanced Ceramics Research) containing milling media (99.9%
alumina, Union Process) for 48 h. The weight ratio of alu-
mina powder to milling media was 1 to 3.5. The powder was
then mixed with monomer solution and dispersant, and milled
further using the jar mill. The monomers, here methacryl-
amide (98%, Aldrich), N,Nꢀ-methylenebisacrylamide (99%,

238
Christian et al. / Journal of Catalysis 241 (2006) 235–242
(see Fig. 3c) was placed in a tube furnace (HTF5500 series,
Lindberg/Blue M) for final curing of the paste at 150 ◦C for 2 h
in air.
2.5. Testing of the assembled ceramic microreactors
The flow of pure NH3 (anhydrous, S.J. Smith) through
the integrated ceramic microreactor was controlled using a
mass flow controller (1479A MassFlo® Controller, MKS In-
struments), and the temperature of the microreactor was con-
trolled by placing it inside a tube furnace (HTF5500 series,
Lindberg/Blue M). Reactants and products were led into and
out of the ceramic microreactor through alumina tubes at-
tached to stainless steel tubing outside of the tube furnace with
Swagelok® connections. The conversion of NH3 to H2 and
N2 was measured using gas chromatography with a thermal
conductivity detector (TRACE DSQ, Thermo Finnigan). Sep-
arations were performed using a Porapak N column (6 ft ×
1/8 inch, stainless steel, 80/100 mesh; Fisher Scientific) with
helium as the carrier gas. For each flow rate of pure NH3, con-
version data were obtained as a function of temperature by in-
creasing the temperature of the furnace from 450 to 1000 ◦C in
50 or 100 ◦C increments. The average conversion and its stan-
dard deviation at a certain flow rate of NH3 and temperature
were calculated from at least three measurements after steady
state was reached.
2.6. Determination of reaction rate constant
The rate constant at each temperature was determined us-
ing the plug-flow reactor (PFR) design equation for first-order
kinetics with constant reactor temperature and a constant pres-
sure gradient [Eq. (1)] and applying a weighted least squares
method [23]. First-order kinetics are used to describe the de-
composition of NH3 based on earlier findings that at temper-
atures above 500 ◦C, the reaction becomes first order with re-
spect to NH3 and is not inhibited by H2 [24,25],
Fig. 3. (a) Optical and SEM (inset) micrographs of a high-density alumina hous-
ing with five microchannels (400 µm × 150 µm × 1 mm) in negative relief,
obtained via gel-casting. Optical micrographs of (b) alumina lid and (c) assem-
bled ceramic microreactor comprised of an alumina housing structure with five
SiC porous monoliths (see Fig. 1), an alumina lid, and alumina inlet and outlet
tubes.
(ꢀP + Pout
)
kꢀSAactive
(Pout + 0.5ꢀP) {−X − 2ln(1 − X)}
v0
=
= kꢀSAactivey, (1)
monoliths and a lid with inlet holes as shown in Fig. 3. The
polymer was completely removed during the initial stage of the
thermal sequence (∼400–500 ◦C).
where v0 is the inlet volumetric flow rate, kꢀ is the rate constant
in m/s, SAactive is the surface area of active catalytic sites, ꢀP
is the pressure drop across the reactor (as calculated using the
Ergun equation [26]), Pout is the pressure at the outlet of the
reactor, and X is the conversion of NH3. Equation (1) can be
rewritten using y = 1/{−X − 2ln(1 − X)}; to determine the
error in y (i.e., ꢀy) the following propagation of error equation
was used:
2.4. Assembly of integrated ceramic microreactors
After deposition of Ru catalyst, the SiC porous monoliths
were mounted within the high-density alumina housing (see
Fig. 3a) using ceramic paste (Ceramabond 569, Aremco Prod-
ucts). After applying the paste and inserting the monoliths into
the microchannels, the housing was visually inspected under a
stereozoom microscope (Leica, MZ 12.5, up to 30X) to ensure
a good fitting of the monoliths within the channels. The alu-
mina lid (see Fig. 3b) and alumina tubes (0.063 inch o.d. and
0.031 inch i.d.; CoorsTek) were then integrated with the hous-
ing using the same ceramic paste. After air-drying the paste at
room temperature for 3 h, the integrated ceramic microreactor
ꢀy = (∂y/∂X)ꢀX.
(2)
With a weighted least squares analysis, the line of best
fit passing through the origin was determined. The sum S in
Eq. (3) was minimized as a function of the slope A of the line
describing the v0(ꢀP + Pout)/(Pout + 0.5ꢀP) versus y data,
ꢁꢂ
ꢃ
ꢄ
2
ꢀ
(ꢀP + Pout
)
S =
wi v0
− Ayi
,
(3)
(Pout + 0.5ꢀP)
i
i

Christian et al. / Journal of Catalysis 241 (2006) 235–242
239
where wi is equal to 1/(ꢀy)2. Once the value of A was de-
The SiC porous monoliths were coated with Ru catalyst via
wet impregnation, followed by calcination in air and reduction
in H2. Ru was chosen as catalyst because of its high activity
toward the decomposition of NH3 [11,27]. The values for Ru
loading as determined using inductively coupled plasma (ICP)
spectrometry and dispersion of active Ru as obtained from
pulsed CO and H2 chemisorption are listed in Table 1 for the
SiC porous monoliths studied here. The Ru loading and disper-
sion of active Ru (4.5–5.8 wt% and 20–31%, respectively) are
comparable to typical values reported in the literature [27–29].
For each monolith, the surface area of active catalytic sites
per gram of Ru can be derived directly from the Ru disper-
sion data, and the resulting values of 250–400 m2/g are within
the range of values reported in the literature for Ru/SiO2 and
Ru/Al2O3 [27,28]. Although the 2.2-µm SiC porous monoliths
have a geometric surface area 3.2 times larger than that of the
7.2-µm SiC monoliths, the total surface area of active catalytic
sites is only 1.75 times larger, indicating that the coverage of the
SiC surface with Ru catalyst is not as good for the structure with
smaller pore diameters. Repeated impregnation of the structures
with the RuCl3 solution may be required to improve the catalyst
coverage on the SiC monoliths with the smaller pores.
termined, the rate constant was calculated using the relation
kꢀ = A/SAactive
.
3. Results and discussion
3.1. Characterization of the porous catalytic monoliths
For the ammonia decomposition studies that we report here,
we use inverted beaded SiC porous monoliths obtained fol-
lowing a procedure that we reported earlier for SiC and SiCN
structures with 7.2-µm pores only [17]. In the work reported
here, packed beds of PS spheres (1.1, 3.2, and 10 µm in di-
ameter) were used as sacrificial templates for the synthesis of
porous monoliths. Subsequent drying, infiltration with a prece-
ramic polymer, curing, and pyrolysis yielded inverted beaded
SiC porous monoliths with average pore diameters of 0.75, 2.2,
and 7.2 µm, as determined from SEM micrographs (Fig. 1).
We studied the chemical and physical stability of these SiC
structures at high temperatures in an oxidative environment as
described previously [17]. We compared the chemical compo-
sition (i.e., the fraction of silicon present as carbide, oxide, and
oxycarbide) via X-ray photoelectron spectroscopy (XPS) and
the pore morphology via SEM of the structures that were just
prepared by pyrolysis under an argon atmosphere and after sub-
sequent exposure to air at 1200 ◦C for 6 h. As before, XPS
analysis of structures before and after thermal treatment in air
showed a constant oxide content of ∼10–20%, and SEM micro-
graphs showed that the morphology of the pores did not change
on heating in air.
The calculated geometric surface areas of these mono-
liths based on pore diameters of 0.75, 2.2, and 7.2 µm were
5.9 × 106, 2.0 × 106, and 6.2 × 105 m2/m3, respectively.
Using BET analysis, we determined actual surface areas of
7.4 × 107 m2/m3 for the 0.75-µm and 6.4 × 106 m2/m3 for the
2.2-µm SiC porous structures, as listed in Table 1. These values
are, respectively, 12.5 and 3.2 times larger than the correspond-
ing geometric surface areas, clearly indicating the presence of
surface roughness and microporosity in these SiC porous mono-
liths. Furthermore, MIP analysis of the 0.75- and 2.2-µm SiC
porous structures showed a porosity of about 74%, meaning that
the pressure drop across these SiC structures will be lower than
that across the packed catalytic beds for the same geometric sur-
face areas. A reliable value for the surface area of the 7.2-µm
SiC porous structure could not be obtained using BET analysis
because of a low surface area-to-volume ratio.
3.2. Characterization of integrated ceramic microreactors:
ammonia decomposition
3.2.1. Integrated ceramic microreactor assembly
Nondeformed and crack-free cm-scale ceramic structures
with submillimeter feature sizes were obtained by modification
of the drying and sintering steps of the gel-casting procedure
developed previously by Young et al. for fabricating ceramic
structures with much larger feature sizes [20,21]. Fig. 3a shows
both optical and SEM images of a high-density, nonporous
alumina structure consisting of five identical channels approx-
imately 400 µm wide, 150 µm high, and 1 mm long, separated
by 1-mm-thick walls. The Ru-coated SiC porous monoliths de-
scribed in Section 3.1 are mounted in these channels using ce-
ramic paste. A matching alumina lid with inlet and outlet tubes
(Fig. 3b) is then mounted on top with ceramic paste, and an
integrated ceramic microreactor is obtained after curing of the
paste (Fig. 3c).
3.2.2. Ceramic microreactor testing
We characterized the performance of the integrated ceramic
microreactors by studying the decomposition of NH3 as a func-
tion of flow rate and temperature. The microreactor, 0.55 mm3
in monolith volume, was placed in a tube furnace, and pure
Table 1
Catalyst loading and dispersion data for Ru-covered SiC porous monoliths with pore diameters of 7.2, 2.2, and 0.75 µm
Pore diameter
(µm, SEM)
Surface area
(m /m , BET)
Ru loading
(wt%, ICP)
Ru dispersion
(%, chemisorption)
Surface area of active catalytic sites
per gram of Ru (m /g)
2
3
2
b
6a
7.2
2.2
0.75
a
∼10
4.5
5.3
5.8
19.7 (CO)
29.3 (CO)
–
20.7 (H )
250
370
2
6
6.4×10
31.1 (H )
2
7
7.4×10
–
Surface area-to-volume ratio is not sufficiently large for a reliable analysis.
Calculated using the Ru dispersion data obtained with CO chemisorption.
b

240
Christian et al. / Journal of Catalysis 241 (2006) 235–242
(a)
(b)
3
(c)
Fig. 4. NH conversion using Ru-coated SiC porous monoliths (volume = 0.55 mm ) as a function of temperature for different NH flow rates measured from 450
3
3
◦
to 1000 C. The conversion values were obtained for integrated ceramic microreactors comprised of SiC catalytic monoliths with (a) 7.2 µm pores; (b) 2.2 µm pores;
and (c) 0.75 µm pores. All data are fitted with the design equation for first-order kinetics in a PFR, assuming constant temperature and a constant pressure gradient
across the reactor. (- - -) indicates the equilibrium conversion of NH as a function of temperature.
3
NH3 was led into the reactor through stainless steel and alu-
mina tubing after passing through a mass flow controller. The
effluent stream exited the reactor through alumina and stainless
steel tubing and was led through a sampling valve for gas chro-
matography analysis. Fig. 4 shows the conversion of NH3 for
SiC porous monoliths with pore diameters of 7.2, 2.2, and 0.75
µm as a function of temperature over the range of 450–1000 ◦C
for NH3 flow rates of 3–36 sccm. The lines in the graphs are
weighted least squares fits for the conversion data using first-
order kinetics with respect to NH3 in a PFR, assuming constant
temperature and a constant pressure gradient across the reactor
[Eq. (1), Section 2.6]. A maximum temperature drop of only
1.1 ◦C (endothermic reaction) can be estimated using Fourier’s
law of conduction for a maximum NH3 flow rate of 36 sccm at
700 ◦C, justifying the assumption of constant temperature.
The fit of first-order kinetics in a PFR to the conversion data
obtained with the 0.75-µm SiC porous monoliths (Fig. 4c) are
poorer than those for the data obtained from the SiC mono-
liths with the larger pore diameters of 7.2 and 2.2 µm (Figs. 4a
and 4b, respectively). The pressure drop through the 0.75-µm
SiC porous monoliths is much greater than that through the 2.2-
and 7.2-µm SiC monoliths and may deviate more from the cal-
culated value obtained using the Ergun equation. In addition,
the assumption of a constant pressure gradient will no longer
hold for larger pressure drops across SiC monoliths with the
smaller pore diameters, introducing more uncertainty into the
calculated rate constants and resulting in poorer line fitting.
Fig. 4 also shows that the conversion increased with increas-
ing temperature and with longer residence times (lower flow
rates), as expected. Here, for NH3 flow rates of 3–36 sccm,
complete conversion of NH3 (i.e., >99.9% conversion or pro-
duction of 4.5–54 sccm of H2) was reached above 700 ◦C, with
the exact temperature at which full conversion was reached de-
pending on the pore diameters of the SiC monoliths used in
the microreactors. Although the residence time was relatively
low, ranging from 0.9 to 11 ms, high conversion could still be
achieved due to the high operating temperature. A maximum
production of 54 sccm of H2 corresponds to 9.8 × 104 sccm
H2 produced per cm3 of monolith volume. Other microreactors,
operating at temperatures up to 650 ◦C, have been reported to
produce only up to ∼3 × 103 sccm H2 per cm3 of monolith or
channel volume for the steam reforming of methanol [1] and the
decomposition of NH3 [12,13]. For the ceramic microreactors
that we report on here, the amount of H2 produced per cm3 of
overall reactor volume (1.05 cm3) is 51 sccm per cm3, a num-
ber that can be increased significantly by increasing the number
of SiC catalytic monoliths within the reactors and reducing the
wall thickness of the alumina housing.
The integrated ceramic microreactors showed no signs of
failure after exposure to more than 15 thermal cycles of about
8 h each at temperatures as high as 1000 ◦C. In addition, no
deactivation of catalyst was observed after exposing the reac-
tor to 12 sccm NH3 at 800 ◦C and 1000 ◦C for 48 h each. No
conversion (i.e., <0.3%) was measured for an identical inte-
grated ceramic microreactor without Ru catalyst when operated
at 500, 800, and 1000 ◦C over the same range of NH3 flow rates,
indicating that the decomposition of NH3 occurred only on the
surface of the Ru catalyst deposited on the SiC porous mono-
liths. Thus, the rate of homogeneous decomposition of NH3 is
insignificant over the range of flow rates and resulting residence
times studied here. From a ceramic material standpoint, the
microreactor should be able to operate without structural degra-
dation at temperatures up to 1200 ◦C in air; but to minimize the
sintering of metal catalyst (here Ru), the temperature must be
maintained below ∼0.5 times the melting point of metal (Tm
for Ru is 2600 K) [26].
3.2.3. Reaction kineticsꢀanalysis
The rate constants k at each temperature were calculated
using a weighted least squares analysis of the conversion data
based on first-order kinetics with respect to NH3 in a PFR,
assuming constant temperature and a constant pressure gradi-

Christian et al. / Journal of Catalysis 241 (2006) 235–242
241
Table 2
Observed first-order reaction rate constants k at different temperatures for SiC porous monoliths with pore diameters of 7.2 and 2.2 µm
ꢀ
◦
◦
◦
◦
◦
700 C
450 C
500 C
550 C
600 C
ꢀ
−6
−6
−6
−5
−5
−5
−5
−4
−4
k (7.2 µm) in m/s
9.34 × 10
8.90 × 10
0.95
7.84 × 10
3.01 × 10
3.85
4.39 × 10
8.27 × 10
1.88
7.23 × 10
1.27 × 10
1.76
3.55 × 10
2.48 × 10
0.70
ꢀ
−4
k (2.2 µm) in m/s
ꢀ
ꢀ
k (2.2 µm)/k (7.2 µm)
ent across the reactor [23]. The value of kꢀ calculated for the
microreactor with 7.2-µm SiC porous monoliths was 4.39 ×
10−5 m/s at 550 ◦C, and an apparent activation energy (Ea) of
22 2 kcal/mol was calculated from the data acquired below
800 ◦C. This Ea falls within the range of values reported earlier
for the decomposition of NH3 on Ru [10,29,30]. The values of
kꢀ calculated for the microreactors with 2.2- and 0.75-µm SiC
porous monoliths were 8.27 × 10−5 m/s and 5.07 × 10−5 m/s
(estimated based on a 5.8 wt% Ru loading and an assumed
36% Ru dispersion for the 0.75-µm structures), respectively, at
550 ◦C, with activation energies of 19 2 and 14 2 kcal/mol.
Assuming that heat and mass transfer limitations can be ne-
glected, the rate constant kꢀ is only a function of the reaction
temperature. Comparing the values of kꢀ for SiC porous mono-
liths with pore diameters of 2.2 and 7.2 µm, however, shows
that the ratio of kꢀ values for the 2.2 µm structure and for the
7.2 µm structure changed as a function of temperature, rather
than maintaining a constant value of 1 (Table 2). The data in
Table 2 indicate that the kꢀ ratio was approximately 1 at 450 ◦C,
then increased sharply at 500 ◦C, and steadily decreased as the
temperature increased to 650 ◦C, dropping to 0.70 at a temper-
ature of 700 ◦C. A ratio of kꢀ values other than 1 indicates that
at a given temperature, the observed reaction rate is different
for the two catalyst supports, signifying the existence of mass
transfer limitꢀations at that temperature. This trend of the chang-
ing ratio of k values reflects that the effects of mass transfer on
reaction rate predominate at higher temperatures; the diffusion
coefficient of a gas increased with T n, where n is approximately
2, whereas the reaction rate constant had an Arrhenius depen-
dence (i.e., increased exponentially with temperature). At low
temperatures, whꢀere the reactors operate in a kinetically lim-
ited regime, the k ratio should be equal to 1. Here we obtained
a kꢀ ratio of 0.95 at 450 ◦C. As the temperature increases, mass
transfer limitations will first affect the reaction rate for the SiC
structures with a larger pore diameter (7.2 µm), and the kꢀ ratio
will increase from a value of 1, in accordance with our ob-
servations. As the temperature increases further, mass transfer
limitations will control the reaction rates for both reactors, and
the kꢀ ratio will decrease until it is equal to 1. In addition, we
observed a drop in the kꢀ ratio to 0.70 at 700 ◦C, possibly due
to invalidity of the assumption that the pressure drop gradient is
constant along the length of the SiC monoliths or to deviation
of the actual pressure drop from the values calculated using the
Ergun equation, which was derived for a packed-bed structure,
not for an inverted beaded-bed structure as used here.
Fig. 5. Arrhenius plot of integral turnover frequencies for decomposition of
NH in integrated ceramic microreactors comprised of Ru-covered SiC mono-
3
liths with 7.2- and 2.2-µm pores. The lines connecting the data points are added
to guide the eye.
at temperatures above 700 ◦C, followed by a gradual decay to
lower TOF values at lower temperatures. In this lower temper-
ature range (<700 ◦C), the TOFs of the SiC monoliths with
2.2-µm pore diameter were systematically higher than those of
the monoliths with 7.2-µm pores. The TOF values for decompo-
sition of NH3 on the Ru-covered SiC porous monoliths studied
here are about one order of magnitude higher than those re-
ported by others for Ru/Al2O3 and Ru/SiO2 catalysts [10,11].
This may be due to a lower metal–support interaction between
the Ru catalyst and the chemically passive SiC surface [31],
because previous work has shown that SiC has a lower metal–
support interaction with other catalytic metals (Ni, Cu, Co, and
Mo) than with SiO2- and Al2O3-based supports [31,32].
4. Conclusion
We have synthesized and characterized ceramic microreac-
tors for the on-site production of H2. The microreactors consist
of high-surface area inverted beaded SiC monoliths with 0.75-,
2.2-, or 7.2-µm interconnected pores as catalyst supports and
high-density, nonporous alumina reactor housings obtained via
the optimized gel-casting procedure reported here. The porous
SiC monoliths have surface areas as high as 7.4 × 107 m2/m3,
significantly higher than the geometric surface area, thus in-
dicating the presence of surface roughness. The SiC catalyst
supports are also stable at temperatures up to at least 1200 ◦C
and have porosities up to 74%, ensuring low pressure drops
across the microreactors. With an entering stream of 36 sccm
NH3, we produced up to 54 sccm of H2 at >99.9% conver-
sion at temperatures above 700 ◦C. The maximum amount of
9.8 × 104 sccm H2 produced per cm3 of monolith is more than
From the conversion and dispersion data, we derived inte-
gral turnover frequencies (TOFs) for the SiC porous monoliths
with pore diameters of 7.2 and 2.2 µm (Fig. 5). A constant TOF
of about 1000 molecule/(site s) was observed for both reactors

242
Christian et al. / Journal of Catalysis 241 (2006) 235–242
one order of magnitude higher than the hydrogen produced
per cm3 of monolith for steam reforming of methanol or de-
composition of NH3 at temperatures up to 650 ◦C, as reported
previously [1,12,13]. Finally, using SiC as the support for Ru
catalysts seems to have a beneficial effect on its catalytic per-
formance, as evidenced by the high TOFs.
These integrated ceramic microreactors are promising for
the steam reforming of higher hydrocarbons such as propane
or butane. The problem of catalyst coking occurring during
reforming of higher hydrocarbons below 800 ◦C [16] can be
avoided completely by using high-temperature-compatible mi-
croreactors as described herein. This may lead to the develop-
ment of microscale devices for the reforming of liquid hydro-
carbons to produce H2 on-site for use in fuel cells. Work along
these lines is currently in progress. In addition, we are presently
integrating a larger number of SiC monoliths within an alumina
housing of similar size, to increase the hydrogen production
rates per reactor volume.
[2] D.R. Palo, J.D. Holladay, R.T. Rozmiarek, C.E. Guzman-Leong, Y. Wang,
J. Hu, Y.-H. Chin, R.A. Dagle, E.G. Baker, J. Power Sources 108 (2002)
28.
[3] M.L. Perry, T.F. Fuller, J. Electrochem. Soc. 149 (2002) S59.
[4] A.S. Patil, T.G. Dubois, N. Sifer, E. Bostic, K. Gardner, M. Quah, C. Bol-
ton, J. Power Sources 136 (2004) 220.
[5] R.R. Davda, J.A. Dumesic, Chem. Commun. (2004) 36.
[6] P. Ferreira-Aparicio, I. Rodríguez-Ramos, A. Guerrero-Ruiz, Chem. Com-
mun. (2002) 2082.
[7] C.M. Miesse, R.I. Masel, C.D. Jensen, M.A. Shannon, M. Short, AIChE
J. 50 (2004) 3206.
[8] L.R. Arana, S.B. Schaevitz, A.J. Franz, M.A. Schmidt, K.F. Jensen,
J. Microelectromech. Syst. 12 (2003) 600.
[9] J.N. Armor, Appl. Catal. A 176 (1999) 159.
[10] T.V. Choudhary, C. Sivadinarayana, D.W. Goodman, Catal. Lett. 72 (2001)
197.
[11] J.C. Ganley, K.L. Riechmann, E.G. Seebauer, R.I. Masel, J. Catal. 227
(2004) 26.
[12] J.C. Ganley, E.G. Seebauer, R.I. Masel, AIChE J. 50 (2004) 829.
[13] Z. Ni, E.G. Seebauer, R.I. Masel, Ind. Eng. Chem. Res. 44 (2005) 4267.
[14] W. Wiese, B. Emonts, R. Peters, J. Power Sources 84 (1999) 187.
[15] A.V. Pattekar, M.V. Kothare, J. Microelectromech. Syst. 13 (2004) 7.
[16] J.N. Armor, D.J. Martenak, Appl. Catal. A 206 (2001) 231.
[17] I.-K. Sung, Christian, M. Mitchell, D.-P. Kim, P.J.A. Kenis, Adv. Funct.
Mater. 15 (2005) 1336.
Acknowledgments
[18] D.C. Duffy, J.C. McDonald, O.J.A. Schueller, G.M. Whitesides, Anal.
Chem. 70 (1998) 4974.
[19] E. Kim, Y. Xia, G.M. Whitesides, J. Am. Chem. Soc. 118 (1996) 5722.
[20] A.C. Young, O.O. Omatete, M.A. Janney, P.A. Menchhofer, J. Am. Ceram.
Soc. 74 (1991) 612.
[21] O.O. Omatete, M.A. Janney, R.A. Strehlow, Am. Ceram. Soc. Bull. 70
(1991) 1641.
[22] M.A. Janney, O.O. Omatete, C.A. Walls, S.D. Nunn, R.J. Ogle, G. West-
moreland, J. Am. Ceram. Soc. 81 (1998) 581.
[23] J. Neter, W. Wasserman, M.H. Kutner, Applied Linear Statistical Models:
Regression, Analysis of Variance, and Experimental Designs, Richard D.
Irwin, Inc., Boston, 1990.
[24] A.S. Chellappa, C.M. Fischer, W.J. Thomson, Appl. Catal. A 227 (2002)
231.
[25] K. Tamaru, Acc. Chem. Res. 21 (1988) 88.
[26] H.S. Fogler, Elements of Chemical Reaction Engineering, Prentice Hall
PTR, Upper Saddle River, 1999.
This work was supported by the Department of Defense
(DoD) Multidisciplinary University Research Initiative (MURI)
program administered by the Army Research Office (ARO)
under Contract DAAD19-01-1-0582, by the campus research
board of the University of Illinois (UIUC), and by the 2004 Na-
tional Research Lab Project [M 10400000061-04J0000-06110]
administered by the Korean Ministry of Science and Tech-
nology. M.M. gratefully acknowledges the National Science
Foundation for a graduate fellowship. Christian gratefully ac-
knowledges support of an H.G. Drickamer Fellowship from
the Department of Chemical and Biomolecular Engineering at
UIUC. The SEM was carried out by UIUC’s Center for Mi-
croanalysis of Materials, which is partially supported by DOE
grant DEFG02-91-ER45439. Any opinions, findings, and con-
clusions or recommendations expressed in this publication are
those of the authors and do not necessarily reflect the views of
the DoD or the ARO.
[27] X.-K. Li, W.-J. Ji, J. Zhao, S.-J. Wang, C.-T. Au, J. Catal. 236 (2005)
181.
[28] J.C. Ganley, E.G. Seebauer, R.I. Masel, J. Power Sources 137 (2004)
53.
[29] M.C.J. Bradford, P.E. Fanning, M.A. Vannice, J. Catal. 172 (1997) 479.
[30] W. Tsai, W.H. Weinberg, J. Phys. Chem. 91 (1987) 5302.
[31] M.J. Ledoux, C. Pham-Huu, Cattech 5 (2001) 226.
[32] R. Moene, E.P.A.M. Tijsen, M. Makkee, J.A. Moulijn, Appl. Catal. A 184
(1999) 127.
References
[1] J.D. Holladay, E.O. Jones, M. Phelps, J. Hu, J. Power Sources 108 (2002)
21.