You are on page 1of 9

Journal of CO2 Utilization 3–4 (2013) 56–64

Contents lists available at ScienceDirect

Journal of CO2 Utilization


journal homepage: www.elsevier.com/locate/jcou

Modeling and kinetic analysis of CO2 hydrogenation using a Mn and K-promoted


Fe catalyst in a fixed-bed reactor
Heather D. Willauer a,*, Ramagopal Ananth b, Matthew T. Olsen c, David M. Drab c,
Dennis R. Hardy d, Frederick W. Williams e
a
Materials Science & Technology Division, Code 6300.2, Naval Research Laboratory, 4555 Overlook Avenue, SW, Washington, DC 20375, United States
b
Chemistry Division, Code 6185, Naval Research Laboratory, 4555 Overlook Avenue, SW, Washington, DC 20375, United States
c
NRC Postdoctoral Research Associate, Materials Science & Technology Division, Code 6300.2, Naval Research Laboratory, 4555 Overlook Avenue, SW, Washington, DC 20375, United
States
d
Nova Research Inc., 1900 Elkin Street, Alexandria, VA 22308, United States
e
Chemistry Division, Code 6104, Naval Research Laboratory, 4555 Overlook Avenue, SW, Washington, DC 20375, United States

A R T I C L E I N F O A B S T R A C T

Article history: Hydrogenation of CO2 to hydrocarbons is investigated over g-alumina-supported modified iron-based
Received 24 April 2013 catalysts under fixed-bed reactor conditions to produce unsaturated hydrocarbons as feedstock
Received in revised form 8 August 2013 chemicals for jet-fuel synthesis. The results are compared to those obtained in a continuously stirred
Accepted 7 October 2013
tank/thermal reactor (CSTR) environment. It is shown that the maximum C2–C5+ yield obtained in the
Available online 13 November 2013
fixed-bed experiments is 49% higher than those obtained by CSTR at lower gas hourly space velocity
(GHSV) of 0.000093 liters per second gram (L/s-g). However, at much higher GHSV (0.0015 L/s-g), C2–
Keywords:
C5+ yield is reduced by 47% in the fixed-bed environment. Modeling and kinetic analysis of the data
CO2 hydrogenation
obtained under fixed-bed reactor conditions is expanded upon by including the formation of methane as
Fixed-bed reactor
Kinetics a by-product and providing kinetic constants. The results show that our Fischer–Tropsch (FT) reaction
Modeling rate is much slower than the reverse water gas shift (RWGS) reaction rate. Therefore the FT reaction
GHSV (gas hourly space velocity) becomes the rate-controlling step in overall conversion of CO2 to olefins. The model was further
expanded to show the potential beneficial effects of water removal from the reactor on CO2 conversion
and product yield.
Published by Elsevier Ltd.

1. Introduction foreign oil supplies, costs, and fuel logistic tails and their
vulnerabilities resulting from fuel delivery. While technologies
The US Department of Defense (DoD) is the single largest buyer currently exist to produce liquid hydrocarbon fuel given sufficient
and consumer of middle-distillate fuel in the world [1]. On average, primary energy resources such as coal and natural gas, most of
the DoD used approximately 9.2 million gallons/day or 3.4 bil- these technologies are not CO2-neutral and they are only practical
lion gallons of fuel in 2010 [1]. Though historical data show the for land-based operations [6,7]. Thus, there is still a need for more
DoD’s average fuel use has steadily decreased from a high of favorable and cost-effective methods to produce hydrocarbon fuel
4.6 billion gallons/year in 2004, the average annual fuel cost has from other resources.
continued to rise [2,3]. The Defense Logistics Agency reports a fuel Carbon dioxide (CO2) is a relatively abundant natural carbon
cost per gallon of $0.60 in 2000. In 2011 this cost rose to over $3.40 resource. As a result of increasing anthropogenic levels and the
per gallon [2,3]. This price does not include the additional costs of foreseeable impact it will have on the environment, research
logistical storage and delivery of the fuel, which in many cases is 2– efforts have focused on capturing CO2 and then sequestering or
3 times the initial price of the fuel [4]. With the global demand for catalytically converting it to energy-rich hydrocarbons [8–13].
fossil fuel increasing and its availability decreasing [5], the DoD is Since the total carbon content of the world’s oceans is roughly
subjected to large price fluctuations. Thus, any production of fuel at 38,000 GtC (gigaton of carbon), it represents a potential resource
or near the point of use would significantly reduce dependence on for carbon feedstock. Over 95% of this carbon is in the form of
dissolved bicarbonate ion (HCO3) [14] which, along with the
carbonate ion (CO32), is responsible for buffering and maintaining
* Corresponding author. Tel.: +1 202 767 2673; fax: +1 202 767 171. the ocean’s pH [14,15]. The sum of these ions and dissolved CO2 gas
E-mail address: Heather.Willauer@nrl.navy.mil (H.D. Willauer). represent the total carbon dioxide concentration [CO2]T of

2212-9820/$ – see front matter . Published by Elsevier Ltd.


http://dx.doi.org/10.1016/j.jcou.2013.10.003
H.D. Willauer et al. / Journal of CO2 Utilization 3–4 (2013) 56–64 57

seawater, as shown in the following equation. remains to find mechanisms to scale down the size of reactor
systems while maintaining sufficient capacity at low costs. To meet
S½CO2 T ¼ ½CO2 ðaqÞ þ ½HCO3  ðaqÞ þ ½CO3 2 ðaqÞ (1) these challenges, companies have developed microchannel units
as modular solutions for off-shore and remote synthetic hydro-
The [CO2]T in seawater on a weight per volume basis is 100 mg/L, carbon fuel production [22,23]. Thus, there has been growing
where approximately 2–3% is dissolved gas and, of the remaining interest to integrate and transition our two-step approach for
CO2, 97–98% is chemically bound as bicarbonate and 1% as producing liquid hydrocarbons from CO2 and hydrogen to reactor
carbonate [14,15]. When compared to the concentration in air systems that have been modeled computationally as microchannel
(370 ppm or 0.7 mg/L), CO2 is approximately 140 times more reactors. In these studies, the effects of space velocity on CO2
concentrated in seawater [6]. This is mainly due to the difference in conversion and product selectivity were evaluated in a fixed-bed
densities of air (0.001204 g/mL) and sea water (1.025 g/mL). tubular reactor, where the results were compared with those
Processes are currently being developed to take advantage of this obtained using similar catalyst compositions in a CSTR. In addition,
higher concentration in seawater as a potential source of feedstock the results were used to modify the multi-step chemical kinetics of
[10,16]. Since CO2 in the atmosphere is in constant equilibrium computational models being designed to predict optimal reactor
with the ocean, it is envisioned that CO2 extracted from seawater size and conditions for hydrocarbon production.
would re-equilibrate from the atmosphere back to the oceans.
Thus, hydrocarbons made from this feedstock using nuclear 2. Experimental
electricity or a renewable energy source such as OTEC (ocean
thermal energy conversion), solar, tidal, or wind could be 2.1. Catalyst preparation
considered as CO2-neutral [3].
One limitation of using CO2 as a feedstock for hydrocarbon Gamma alumina (Matheson Coleman & Bell Manufacturing
production concerns the high energy barrier for polymerization Chemists, Norwood, Ohio 45212) with a surface area and pore
and the need for a source of hydrogen. While reports suggest radius of 218.3 m2/g and 1.2–3.7 nm, respectively, and a particle
electrochemical and photochemical catalytic CO2 conversion size distribution of 80–100 mesh was used as support material. An
processes continue to improve in efficiencies [11], thermal incipient wetness impregnation method was used for catalyst
processes remain one of the few proven methods for producing preparation. In a flask Fe(NO3)9H2O, KMnO4, and KNO3 were
high yields of hydrocarbons greater in carbon number than added to boiling deionized H2O to obtain the desired weight ratios.
methane [12,13,17–20]. To take advantage of this approach, a two- The hot mixed metal nitrate salts solution was then impregnated
step synthesis process that involves converting CO2/H2 to olefins onto the alumina and then dried at 80–90 8C in ambient air with
and subsequent oligomerization over nickel-supported catalyst to continuous agitation to a pasty consistency before additional
jet fuel is being pursued. To date, we have reported CO2 conversion drying overnight at 110 8C. Samples were calcined at 350 8C for
levels as high as 50.2% and an olefin/paraffin ratio of 4.4 using an 87 h, under static air conditions.
optimized bifunctional K/Mn/Fe catalysts impregnated on Al2O3
containing ceria in a continuously stirred tank/thermal reactor 2.2. Characterization
(CSTR) [17].
The CSTR simulates conventional slurry phase reactors that are BET surface areas were measured using a Micromeritics
used for large scale industrial applications [7,21]. The challenge ASAP2010 accelerated surface area and porosimetry system. An

Fig. 1. Schematic of the experimental fixed bed catalytic reactor for making jet fuel.
58 H.D. Willauer et al. / Journal of CO2 Utilization 3–4 (2013) 56–64

appropriate amount (0.11 g) of sample was taken and heated to oil. The gases dissolve in the liquid phase (slurry) under pressure,
120 8C for 17 h under vacuum (50 mTorr). The sample was then diffuse in the liquid phase to the catalyst particle, before coming in
transferred to the adsorption unit, and the N2 adsorption was contact with the active catalyst sites. CSTR bench top reactor
measured at the boiling temperature of nitrogen. processes are proven scalable mechanisms for studying traditional
Fischer–Tropsch (FT) reactions [7,21].
2.3. Fixed-bed reactor synthesis The objective of the present study is to transition CO2
hydrogenation from a CSTR to a fixed-bed process. CSTR and fixed
CO2 hydrogenation reactions were conducted in a 900 stainless bed reactors have fundamentally different flow patterns. In CSTR,
steel tube (3/800 ID, 1/200 OD) FT reactor as shown in Fig. 1. In a typical the liquid and gases are well stirred. In the fixed bed, gas and liquid
experiment, 20 g of a mixture of calcined catalyst and 80–200 mesh flow down the bed of catalyst from the top of the reactor to the
Al2O3 was prepared to obtain different space velocities of catalyst bottom, without any stirring. Further, in the fixed-bed reactor, H2
(Table 1). The mixtures were dried under N2 stream for an hour and CO2 gases are brought in to direct contact with the catalyst
(100 mL/min, 300 8C, 265 psig) followed by in situ reduction using particles dispersed in a support material (alumina) unlike the CSTR
CO at 265 psig and 300 8C for 2 h. Fig. 1 shows three mass flow process. In a CSTR, the gases are bubbled through mineral oil
controllers (Sierra Instruments, USA) which were used to control the (solvent) containing the catalyst. The gases dissolve and diffuse to
flow of all the different gases into the reactor (pictured H2, CO2, N2, the catalyst sites dispersed in the mineral oil so that reaction at
not pictured CO). Immediately after pre-treatment and reduction, catalyst sites can take place [17]. In a fixed-bed, the gases diffuse
hydrogenation of CO2 was conducted at 300 8C, 265 psig, and through the support material to reach the active catalyst sites,
different space velocities of catalyst with a H2/CO2 ratio of 3:1. A where the reactions occur. Clearly, there are differences in the
10 mL/min flow of nitrogen gas was used as a reference gas in each phase behavior of the reactant and product gases between CSTR
reaction. The effluent gases were passed through a cold trap, which and the fixed-bed reactor. Differences in the transport of the gases
was placed in an ice bath to condense the water vapor and any heavy and the catalyst particle characteristics could lead to differences in
liquid hydrocarbons that might have formed in the reactor. The the resulting reaction kinetics. It is well established that a strong
gases leaving the cold trap were passed through a cylinder relationship exists between catalyst composition and reaction
containing 500 g of 3 Å molecular sieves (Molecular Sieve UOP conditions (reactor type, GHSV, reactor temperatures and pres-
Type 3Å, Sigma–Aldrich, Saint Louis, MO 63103 USA) to dry before sures) which leads to changes in kinetic rate equations and the
online analysis by GC as shown in Fig. 1. The effluent gases were kinetic parameters obtained by these equations [19,24]. Using a
analyzed online using a chromatographic analyzer (Agilent Tech- previously designed and tested catalyst composition, we varied the
nologies, Fast RGA analyzer) for gas analysis. Hydrocarbons were reactor conditions, rather than catalyst composition [18]. In this
separated using an HP-Al/S column (Agilent Technologies, 19091P- work, we quantify the effects of gas hourly space velocity (GHSV)
512, 25 mm  320 mm  8 mm) and detected with an FID detector. on CO2 conversion, hydrocarbon product selectivity and yield, and
Fixed gases (H2, CO2, CO, N2) were separated on a Unibead IS column compare with those previously reported in a CSTR for a fixed
(4 ft, 60/80 mesh in UltiMetal) and a 5 Å molecular sieve column catalyst composition.
(8 ft, 60/80 mesh) and detected with a TCD detector. The GC was Data were collected on the conversion of CO2 over a 17 weight
calibrated using a mixture of gases with known molar ratio (MESA percent (wt%) iron-alumina catalyst doped with approximately
Specialty Gas, USA). Time-on-stream (TOS) for the catalyst was 48 h. 8 wt% potassium and 12 wt% manganese for 48 h. CO2 conversion
under eight different GHSV is reported. Initially in the first four
3. Results and discussion studies with the gas ratio of CO2/H2 1:3, the mass of the catalyst,
reactor tube dimensions (length, inner diameter (ID) and outer
Previous results report CO2 conversion levels as high as 41% diameter (OD)) were fixed, and the GHSV was adjusted by
with an olefin/paraffin ratio of 4.2 using an optimized iron-based changing the flow rate of CO2 and H2 into the reactor. The GHSV
supported catalyst (Fe/Mn/K-alumina) in a continuously stirred was changed in the remaining studies by lowering the active
tank/thermal reactor (CSTR) [18]. The incorporation of reverse catalyst amount in the bed. The support material, gamma-alumina,
water–gas-shift (RWGS) activity by the addition of ceria prior to was used to maintain a constant (20 g) mass of material in the
the deposition of the CO2 hydrogenation catalyst increased the CO2 reactor when varying the active catalyst amount.
conversion levels to 50% and the olefin/paraffin ratio to 4.4 [17]. In Table 1 provides the product selectivity, olefin/paraffin ratio,
those studies CO2 hydrogenation reactions were conducted in a and CO2 conversion as a function of GHSV based on the GC analysis.
CSTR slurry reactor, where the catalyst was dispersed in mineral Figs. 2 and 3 specifically show the trends in hydrocarbon

Table 1
Product selectivity, olefin/paraffin ratio, CO2 conversion, and ASF values over K/Mn/Fe catalysts, impregnated on Al2O3 support under fixed-bed conditions after 48 h of run
time.

GHSV Mass Reactor Selectivity (%, Carbon Based) Olefin/ Conversion a a Yield
(L/s-g) catalyst (g) type paraffin CO2 (%) C1–C6+ C1–C5 C2–C5+ (%)
C1 C2–C5+ CO
(methane)

0.0015 10–15 CSTR 26 62.4 11.5 4.2 41.4 0.47 – 25.8

0.000047 17.82 Fixed-bed 22.4 62.3 15.2 4.6 54.8 0.41 0.43 34.1
0.000093 17.82 Fixed-bed 22.4 64.2 13.4 5.6 60.0 0.47 0.57 38.5
0.00019 17.82 Fixed-bed 19.8 58.5 21.7 5.9 53.7 0.50 0.54 31.4
0.00037 17.82 Fixed-bed 16.1 56.0 28.0 5.9 50.6 0.51 0.55 28.3
0.00075 8.91 Fixed-bed 24.4 32.0 43.4 4.8 46.2 0.45 0.46 14.8
0.0015 4.46 Fixed-bed 22.2 30.5 47.3 4.6 44.7 0.48 0.50 13.6
0.0029 2.23 Fixed-bed 21.0 34.6 44.8 4.9 40.7 0.49 0.46 14.1
0.006 1.11 Fixed-bed 17.9 19.0 63.1 5.1 27.8 0.45 0.40 5.3

Bold values are from CSTR and fixed-bed experiments at same GHSV.
H.D. Willauer et al. / Journal of CO2 Utilization 3–4 (2013) 56–64 59

70 Since methane formation is widely considered to be the most


thermodynamically favored product in CO2 hydrogenation [19],
60 the reaction conditions that favored its lowest selectivity are ideal.
Table 1 and Fig. 3 show that, on average over a wide range of GHSV
50 reaction conditions, methane formation was 21% with the lowest
amount produced at a GHSV of 0.00037 L/s-g, and the highest
(% ) CO2, CO

40 amount at 0.0075 L/s-g. The methane selectivity is on average 19%


lower than previously reported using a CSTR process. As the GHSV
was increased, a shift in hydrocarbon selectivity away from C2–
30
C5+ hydrocarbons was seen in favor of much more carbon
monoxide formation whereas CO2 conversion was reduced. The
20
increase in formation of CO has been considered evidence of a two-
step reaction mechanism for formation of hydrocarbons from CO2
10 [20]. The increase in CO might indicate that the FT step for
converting CO to hydrocarbon is slow relative to the RWGS
0 reaction step for converting CO2 to CO. Indeed, the kinetic model
0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 parameters obtained from these experimental data show that the
RWGS to be faster than FT.
GHSV (L/s g)
Yield is calculated by multiplying the C2–C5+ hydrocarbon
Fig. 2. Comparison of experimental data with model predictions for CO2 conversion selectivity values with the CO2 conversion values in Table 1.
and CO selectivity and the effects of varying GHSV. CO2 conversion experimental Table 1 shows that the highest C2–C5+ yield is 38.5% by weight on a
(*), model (*), CSTR (~), and Riedel (!), CO selectivity experimental (&), model
carbon basis. This means 38.5% of CO2 fed into the reactor is
(&), and CSTR (^).
converted to C2–C5+ molecules. The maximum yield of 38.5% in the
fixed-bed reactor is significantly higher than that in the slurry CSTR,
selectivity, unreacted carbon monoxide, and CO2 conversion as a which was 25.8%. It appears that the yield improved by almost 50%
function of GHSV. The product hydrocarbons produced from each by transitioning from CSTR to the fixed-bed reactor. Furthermore,
reaction condition consisted predominantly of methane and light Table 1 shows that the yield increases from 34.1% to 38.5% as GHSV is
olefins. A smaller fraction of the hydrocarbons were found to be increased from 4.7  105 L/s-g to 9.3  105 L/s-g. As GHSV is
paraffins, generally resulting in an olefin/paraffin ratio ranging from increased further, the yield decreases due to decreasing CO2
4.6 to 5.9. It was noticed that there was some liquid hydrocarbon conversion. At a fixed GHSV of 1.5  103 L/s-g, the yield in the
collected in the cold trap along with condensed water. But, the liquid fixed-bed reactor is 13.6%, which is significantly less than the 25.8%
hydrocarbon collected in the cold trap was small and was neglected measured in the CSTR. The precise reason for this trend is not well
in the calculations reported in the present work. understood. However, to explain this trend, three possibilities are
At the lowest GHSV tested in the fixed-bed reactor, CO2 postulated. First, the catalyst composition used in the fixed-bed and
conversion reached 54.8% with an olefin/paraffin ratio of 4.6. CSTR are different though the catalysts were prepared following
Unreacted CO was 15.2% and the formation of unwanted methane similar incipient wetness impregnation methods [18]. On 100 g
was 22.4%. When the GHSV was increased from 4.7  105 L/s-g to Al2O3 basis, the catalyst compositions for K/Mn/Fe/Al2O3 are 7.3/8.2/
9.3  105 L/s-g, CO2 conversion increased from 54.8% to 60% and 11.6/100 and 10.5/12/17/100 by mass for the fixed-bed and CSTR
the hydrocarbon selectivity (without methane) rose to 64.2%. respectively. Clearly, the fixed-bed catalyst has less Fe, K, and Mn
These were the highest conversion levels measured in these compared to the CSTR. Second, the catalyst is dispersed in mineral oil
studies. The olefin to paraffin ratio increased to 5.6 and more CO in the CSTR and in alumina powder in the fixed-bed. The resistances
was consumed in the reaction process. to the transport of species to the catalyst sites could be very different.
Third, the composition of liquid hydrocarbon formed in CSTR and
fixed-bed could be different and each reactor media (mineral oil vs.
80
alumina, respectively) have different effects on the transport of
species to and from the catalyst sites.
Figs. 2 and 3 also provide a comparison of our measurements for
60 CO2 conversion, hydrocarbon and CO yields with those reported in
(% ) CH4, C2-C5 +

the literature for the fixed bed reactor. The figures show that our
data are consistent with those measured by Riedel et al. [19] at low
values of GHSV. Riedel et al. [19] report their data in terms of t g-s/
40 cm3 which is related to GHSV as 1/(1000  t) L/g-s, which is only
nominal because it does not account for differences between the
study of Riedel et al. [19] and ours in the composition of the final
powder and especially in the calculation of ‘‘catalyst mass’’, Mc.
20 Despite these differences in the calculation method for GHSV values
between the present study and that of Riedel et al. [19], Figs. 2 and 3
show that the maximum conversion and maximum hydrocarbon
0 yield reported by Riedel et al. [19] are consistent with the present
study. However, as GHSV is increased CO2 and hydrocarbon yields
0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007
were reported to approach zero by Riedel et al. [19]. In contrast,
GHSV (L/s g) Figs. 2 and 3 show our measured CO2 conversion and HC selectivity
as high as 28% and 18% at a GHSV over 3 times higher than that
Fig. 3. Comparison of experimental data with model predictions for hydrocarbon
(C3H6 or C2–C5+), CH4 selectivities (C basis) and the effects of varying GHSV. C3H6 or
reported by Riedel et al. [19]. The difference in CO2 conversion,
C2–C5+ experimental (&), model (&), CSTR (^), and Riedel experiments (!), CH4 selectivity’s, and yields of species at higher GHSV may be due to
selectivity experimental (*), model (*) and CSTR (~). differences in the catalyst used. Riedel et al. used K/Cu/Fe/Al2O3 in
60 H.D. Willauer et al. / Journal of CO2 Utilization 3–4 (2013) 56–64

the ratio 67/80/752/100. Riedel et al. dilute 2.5 g (Mc) of catalyst with reported reaction [19].
25 g of silica (1 to 10 ratio). So, the final powder has 6.8% Fe, 0.72% Cu,
CO2 þ 4H2 ! CH4 þ 2H2 O (5)
0.61% K in the support consisting of Al2O3 and silica, which is also
equivalent to 0.66/0.8/7.4/100 for K/Cu/Fe/(Al2O3 + silica). We use Another deleterious side reaction is the Boudard reaction
Mn instead of Cu in the ratio 7.3/8.2/11.6/100 for K/Mn/Fe/Al2O3 (Eq. (6)) [19,26,27]
respectively. In a typical run, about 17.8 g (Mc) of the catalyst is
mixed with 2.2 g of alumina to make the final powder that was used 2CO @ C þ CO2 (6)
to form the fixed bed in our tubular reactor. The final powder There is a limited kinetic analysis of Eqs. (2)–(4) involved in
contains 8.1% Fe, 5.7% Mn, 5.1% K in Alumina support by mass as CO2 hydrogenation [19]. While Eqs. (2)–(6) have been reported to
calculated based on the ratios for K/Mn/Fe/Al2O3 7.3/8.2/11.6/100. be involved in CO2 hydrogenation, our gas chromatographic
The final powder we use in the fixed-bed has significantly lower Fe analysis of the product stream shows few major species produced
and K loading in relation to Al2O3 than those of Riedel et al. [19]. under the reaction conditions provided in Table 1. These species
There are also significant differences in catalyst preparation. Riedel include propylene, methane, CO and H2O. There are many other
et al. [19] uses co-precipitated catalyst while we use an impregna- species present but in quantities that are less than 20% of the
tion method. This could lead to differences in the particle size and major species measured in the product stream. Therefore the
distribution. For example, larger particle sizes can lead to slower following reactions (Eqs. (7)–(9)) are considered as the basis for
diffusion of reactants and products formed inside the pores of the development of our one dimensional fixed-bed kinetic reactor
catalyst particle. These results highlight how the changes in catalyst model:
composition and reactor environment (CSTR vs. fixed-bed) from one
study to the next produce significantly different CO2 hydrogenation Reverse water gas shift ðRWGSÞ CO2 þ H2 @ CO þ H2 O (7)
results. Therefore the kinetic analysis of CO2 hydrogenation should
impact the kinetic parameters from one set of experimental FischerTropschðFTÞ CO þ 2H2 ! 1=3C3 H6 þ H2 O (8)
conditions to the next.
FischerTropschðside reactionÞ CO þ 3H2 ! CH4 þ H2 O (9)
The total hydrocarbon product distribution under eight
different GHSV can be described by means of Anderson–Schulz– Since methane formation is a thermodynamically favored
Flory (ASF) growth distribution plots, where ln(WN/N) (WN is the reaction in CO2 hydrogenation, a reaction that describes methane
weight fraction of HC containing N Carbon) is plotted as a function formation was added to the model in Eq. (9). We will show the
of carbon number (N). The chain growth probability, a, is inferred methanation reaction given by Eq. (9) in our computational results.
by the slope [25]. Table 1 shows these values initially increase with This reaction was not shown in the computational results
increasing GHSV and they range from a = 0.41 to 0.51 for C1–C6+. presented in the literature [19]. There are direct hydrogenation
These values are in excellent agreement with those obtained for reactions with CO2 taking place, but these are generally considered
comparable catalysts used in a CSTR (Table 1) [17,18]. The weight to be slower than the hydrogenation reactions with CO [19]. Eq. (7)
fraction of C6+ hydrocarbons is small relative to C1–C5 hydro- was reported to be an equilibrium reaction unlike Eqs. (8) and (9)
carbons, thus analysis of for C1–C5 shows values as high as a = 0.57 in the literature [19]. However, Norval suggests that indeed Eqs. (8)
under the most favorable fixed-bed reactor conditions. These and (9) could be considered reversible to explain the product
values are similar to those reported for iron catalysts used in distribution data in traditional FTS [28]. However we follow
typical FT processes [25], where results indicate the feasibility of a Riedel’s kinetic approach and treat the kinetic constants as
two-step approach to hydrogenation of CO2 to liquid fuel. parameters of the model [19].
Table 1 shows that, while the GHSV does not greatly affect the Next, we developed a one dimensional plug flow, fixed-bed
chain growth probability (0.43–0.57 C1–C5+), it has a significant catalytic, reactor model for the reaction scheme (Eqs. (7)–(9))
impact on CO2 and CO conversion. In particular, their conversion shown above. We considered 1 mole of CO2 and R moles of H2
can be significantly improved by lowering the GHSV and thus present in the system initially. We also consider  moles of CO2 are
increasing the residence time of the reagents CO2, CO, and H2 in converted to CO following reaction (7), y moles of CO are converted
contact with the active catalyst bed. Initially such low GHSV to unsaturated hydrocarbons following reaction (8), and z moles of
appears to be impractical for scale-up since the inverse relation- CO are converted to saturated hydrocarbons including methane
ship between catalyst mass and GHSV necessitates the need for following reaction (9). We also consider a case when w moles of
very large amounts of catalyst to achieve higher conversion. Since water are removed from the reactor at a specified rate, kw. As
extremely low GHSV is impractical, kinetic analysis of CO2 illustrated in Eqs. (7)–(9) water formation is a dominate product of
hydrogenation is used to determine kinetic parameters needed these reactions. From the equations, water should affect the RWGS
to further facilitate CO2 and CO conversion in fixed-bed micro- reaction equilibrium and there is evidence that it also affects the
channel reactor conditions. catalyst active sites in CO2 hydrogenation [29,30]. Even though
The mechanism of CO2 hydrogenation has been proposed to water removal from the reactor in these studies has not been
occur in two steps [19]. The first step (Eq. (2)) is the RWGS, which is attempted experimentally, it is important to understand its effect
endothermic and produces carbon monoxide. This carbon monox- on conversion and selectivity by considering a constant removal
ide is then carried forward in an exothermic FT step (Eq. (3)), rate of water, kw. In these studies water is varied as a parameter of
producing predominantly monounsaturated hydrocarbons the model to determine if experiments should be performed in the
(Eq. (4)). future to examine this effect. In the base case, kw is set to zero (no
water removal) and is set to 0.1 s1 for the case in which water is
nCO2 þ nH2 @ nCO þ nH2 O (2)
removed from the reactor. The concentrations of species can be
nCO þ 2nH2 ! ðCH2 Þn þ nH2 O (3) expressed in Eqs. (10)–(15) as:

CO2 ¼ 1  x (10)
nCO2 þ 3nH2 ! ðCH2 Þn þ 2nH2 O (4)

A thermodynamically favorable side reaction associated with CO ¼ x  y  z (11)


CO2 hydrogenation is the highly competitive methanation
reaction (Eq. (5)), and for many years this was the widely H2 ¼ R  x  2y  3z (12)
H.D. Willauer et al. / Journal of CO2 Utilization 3–4 (2013) 56–64 61

C3 H6 ¼ y (13) experiments. The kinetic constants kSH, kFT, kFTs depend on the
activation energies ESH, EFT, EFTs for the RWGS, FT, and methanation
CH4 ¼ z (14) reactions respectively. The parameters kSH0, kFT0, kFTs0 are the pre-
exponential coefficients for the RWGS, FT, and methanation
H2 O ¼ x þ y þ z  w (15) reactions respectively. The parameters aSHH2 O , bSHCO2 , aFTH2 O ,
bFTCO2 , aFTsH2 O , bFTsCO2 are the adsorption constants for the RWGS,
The gases CO2 and H2 enter the vertical fixed-bed catalytic
FT, and methanation reactions respectively. Eq. (20) represents
tubular reactor of length L at a specified total flow rate, F (m3/s), at
rate of water vapor removal from the reactor. It is assumed to be
standard temperature and pressure (STP) conditions as measured
removed at a rate kw. The value of kw is set to zero in determining
by the mass flow controllers. The reactor contains Mc grams of
the kinetic parameters and comparing model predictions with
impregnated catalyst mixed with (20-Mc) grams of un-impregnat-
experimental data. Only the simulations designed to study the
ed alumina. GHSV is determined as F/Mc. The average velocity of
water removal effect, a non-zero value for kw was used.
the feed stream is given by the following equation.
Substituting Eqs. (10)–(15) into Eqs. (17)–(24) gives final equations
FðT=273:5ÞðP atm =PÞ for x, y, and z. The resulting equations are solved using the
U¼ (16)
2
ðpd =4Þ commercial program COMSOL1. COMSOL1 uses a direct, time-
dependent, solver based on Newton’s method. A multi frontal
where T and P are the reactor temperature and pressure massively parallel sparse (MUMPS) direct solver with automatic
respectively. Patm is the atmospheric pressure and d is the diameter matrix symmetry detection was used.
of the tubular plug flow reactor. The reactor is assumed to be The activation energies (EAi, kJ/mol) in Eqs. (21)–(23) were
maintained at isothermal condition. The small pressure drop reported in the literature [19] and are shown in Table 2. Similarly,
across the reactor length is neglected and isobaric conditions are the adsorption constants aSHH2 O , bSHCO2 , aFTH2 O , bFTCO2 were also
assumed. The residence time, t, is defined as s/U, where s is the axial reported by Riedel et al. [19] and are shown in Table 2. The pre-
distance from the entrance of the reactor. Since the reactor tube exponential coefficient, kSH0, in Eq. (21) and the equilibrium
diameter is much smaller than the length of the reactor, plug flow constant, Keq, in Eq. (24) were also reported in the literature [19].
is assumed, and the radial gradients are neglected. One dimen- The pre-exponential coefficient for the FT reaction, kFT0, reported
sional mass balance equations are given in Eqs. (17)–(20) based on by Riedel et al. [19] was found to over predict the hydrocarbon
Eqs. (7)–(9). concentrations in the outlet stream of the reactor when compared
to the experimentally measured values as analyzed by GC.
dx ðCOÞðH2 OÞ
¼ kSH b ðCO2 ðH2 Þ  c (17) Therefore, we determined the pre-exponential coefficient, kFT0,
dt K eq
in Eq. (22) by empirically fitting the computational results for
specie concentrations to the measured values from our experi-
dy
¼ kFT b ðCOðH2 Þ c (18) ments as shown in Table 1. Table 2 shows the value of kFT0 to be
dt
1.29  103, which turns out to be five orders of magnitude smaller
dz than that reported by Riedel et al. [19] This is somewhat surprising
¼ kFTs b ðCOðH2 Þ c (19) because, the CO2 conversion, selectivities and yields of species CO,
dt
C2–C5+ measured in the present work are higher than those
dw reported by Riedel et al. [19] under similar GHSV, temperature, and
¼ kw (20) pressure conditions (Figs. 2 and 3). Therefore, one might have
dt
expected a higher value for the kinetic coefficient for the FT
Expressions for the kinetic parameters are given by Eqs. (21)–
reaction in our case than that for Riedel et al. The reasons for this
(24). These expressions were derived by Riedel who considered
discrepancy might be explained by the following. It is possible that
inhibition due to water, CO, and CO2 adsorption onto the catalyst
Fe exists in much more concentrated form on an individual catalyst
sites [19].
particle in the study of Riedel et al. [19] compared to ours, even
rcat Rg TkSH0 exp½ESH =ðRg TÞ though the Fe content of the final powder forming the fixed bed is
kSH ¼ (21)
½ðCO þ aSHH2 O ðH2 OÞ þ bSHCO2 ðCO2 Þ comparable between our study (8.1% by mass) and that of Riedel
et al. [19] (6.8% by mass). For example, Riedel et al. [19] uses 2.5 g
rcat Rg TkFT0 exp½EFT =ðRg TÞ of catalyst containing Fe and dilutes it with 25 g of silica particles,
kSH ¼ (22) while we impregnate 17.8 g of catalyst with Fe and dilute it by 2.2 g
½ðCO þ aFTH2 O ðH2 OÞ þ bFTCO2 ðCO2 Þ
of alumina. This might have contributed to the differences in rate
rcat Rg TkFTs0 exp½EFTs =ðRg TÞ constants. The significant differences in catalyst composition and
kFTs ¼ (23) results of CO2 hydrogenation suggest that there should be
½ðCO þ aFTsH2 O ðH2 OÞ þ bFTsCO2 ðCO2 Þ
differences in the kinetic analysis of our data as compared to
Riedel et al. [19].
K eq ¼ 10ð2:0292073=TÞ (24)
As discussed earlier, we include the methanation reaction in our
Here T is reaction temperature, Rg is universal rate constant, rcat is model represented by the kinetic constant kFTs in Eq. (23). The
the catalyst density, which is defined as the mass of impregnated activation energy, EFTs, and pre-exponential coefficient, kFTs0, were
catalyst, Mc/bed volume. The bed volume is 14.54 cm3 in our not reported by Riedel et al. [19]. Therefore, they needed to be

Table 2
Kinetic parameter values used in the model.

Kinetic parameter CO2 shift (SH) FT FTs, methanation

ai, H2O 65 Ref. [19] 33 Ref. [19] 33 Ref. [19]


bi, CO2 7.4 Ref (19) 2.7 Ref. [19] 2.7 Ref. [19]
ki,0, mol/s 1.51  107 Ref. [19] 1.29  103 (present study) 5.16  102 (present study)
EA,i, kJ/mol 55 Ref. [19] 72 Ref. [19] 72 Ref. [19]

i = SH, FT, FTs. Reference (Ref.).


62 H.D. Willauer et al. / Journal of CO2 Utilization 3–4 (2013) 56–64

120 for the gases to travel along the entire length of the reactor. Fig. 4
shows a sudden change in slope around 2 s for all species. This
quick change in slope may be explained by the differences in the
100
(% ) CO2, CO, CH4, C3H6

reaction rates for the RWGS and FT reactions. Fig. 5(a) shows the
RWGS reaction rate reaches its maximum value (3.5  105 s1)
80 initially and decreases to 0.032 s1 within 2 s. In Fig. 5(b) the FT
reaction rate is initially at zero and increases to its peak value
(0.032 s1) in about 2 s. After 2 s, the FT reaction rate begins to
60
decrease smoothly and reaches 4  103 s1 at 18.7 s. After 2 s, the
RWGS reaction rate fluctuates but follows the FT reaction rate
40 curve with time and stays about equal. Therefore, it is clear from
Fig. 5(a) that the RWGS reaction reaches its equilibrium in about
2 s. This corresponds to about 10% of the reactor length. As a result,
20
Fig. 4 shows how the equilibrium during the reaction begins to
shift toward CO2 as CO is being consumed relatively slowly by the
0 FT reactions. After 2 s, the CH4, C3H6, CO2 conversion increase and
0 5 10 15 20 CO selectivity decreases relatively slowly with distance along the
tubular reactor. The quick change in slope at 2 s indicates that the
Time, s
RWGS reaction (Eq. (7)) reaches equilibrium much more quickly
Fig. 4. Model predictions for the distribution of CO, CH4, C3H6 (C basis) selectivities than the FT reaction (Eq. (8)) goes to completion because the
and CO2 conversion with residence time (total flow = 400 mL/min, catalyst mass largest magnitude of RWGS reaction rate is seven orders of
17.82 g, GHSV = 0.00037 L/s-g). magnitude higher than the largest magnitude of FT reaction rate.
The model predictions show that the FT reaction is the rate-
determined for this study. We assumed that the EFTs for the controlling step for the overall conversion of CO2 to olefins.
methanation reaction was equal to EFT. We determined the value of Table 3 lists the selectivities at the end of the tubular reactor for
kFTs0 by empirically fitting the computational results to CH4 different values of GHSV obtained by varying the flow rate and
concentrations measured experimentally in the outlet stream of catalyst mass. Table 3 clearly shows decreased CO2 conversion and
the reactor as shown in Table 2. The pre-exponential coefficients increased CO in the product stream as GHSV is increased. The
reported in Table 2 show a lower than 1 value for the ratio kFTs0/ model predictions suggest that this is mainly due to a small FT
kFT0 = 0.4, which means that our catalyst has lower selectivity to reaction rate relative to the feed rate of the reactants represented
methane compared to propylene. by GHSV. The RWGS reaction rate was found to be very fast relative
The solutions of Eqs. (17)–(20) provide the concentrations of to the feed rate of the reactants. As pointed out in the previous
CO2, CO, H2, C3H6, CH4, and H2O (shown in Fig. 4) along with the section, Boudard reaction and catalyst deactivation could become
RWGS and FT reaction rates (shown in Fig. 5(a) and (b)) for a GHSV important especially at later times of the reactor operation. They
value of 3.7  104 L/s-g (total gas flow of 400 mL/min and active could also affect the CO2 conversion. The deactivation could occur
catalyst mass of 17.82 g), 300 8C temperature and 265 psi. The due to carbon deposition and liquid product formation which could
reaction rates for RWGS and FT shown in Fig. 5(a) and (b) are slow the diffusion into the catalyst particle pores. The deactivation
calculated by evaluating the algebraic expressions on the right kinetics are not well known in general and especially for the
hand side of Eqs. (17) and (18) respectively. Fig. 4 shows the CO2 catalyst at hand [31,32].
conversion, CO, C3H6, and CH4 selectivities on a weight of carbon Fig. 2 shows CO2 conversion predicted by the model similar to
basis rather than the molar basis. This means the percent of carbon those measured in the experiments in the outlet stream from the
in CO2 reactant is converted to the carbon in different product reactor tube. The CO seems to increase from zero steadily with
species. In Figs. 4 and 5(a) and (b), the residence time, t = s/U, really GHSV in the model predictions – while the measurements also
represents the distance, s, along the length of the tubular reactor show an increase in CO with GHSV, CO values remain 20% or more
bed. Fig. 4 shows CO decreases while CO2 conversion and C3H6 and at low values of GHSV. Fig. 3 shows a steady decrease in methane
CH4 increase along the length of the reactor. It takes about 18.7 s and propylene selectivities (on a carbon basis) predicted by the

Fig. 5. (a) RWGS reaction rate as a function of residence time. (b) FT reaction rate as a function of residence time (total flow = 400 mL/min, catalyst mass 17.82 g,
GHSV = 0.00037 L/s-g).
H.D. Willauer et al. / Journal of CO2 Utilization 3–4 (2013) 56–64 63

Table 3
Kinetic model predictions for product selectivity, olefin/paraffin ratio, and CO2 conversion over K/Mn/Fe catalysts, impregnated on Al2O3 support under fixed-bed conditions
after 48 h of run time.

GHSV Mass Total flow rate Selectivity (%, carbon based) C3H6/CO2 Conversion H2O/H2CO2 Yield C3H6
(L/s-g) catalyst (mL/min @STP) consumed CO2 ratio in feed (%) = 100y
(g) (%) = (%) = (%) = (x + y + z)/3
C1 C3H6 CO 100(y/3x) 100(x)
(methane) (%) = 100 (y/x) (%) = 100 (x  y  z)/x
(%) = 100 (z/x)

0.000047 17.82 37.5 28.5 70.5 1.3 23.5 61 40.2 43


0.000093 17.82 75 28 70 2 23.3 54.5 35.8 38.1
0.00019 17.82 150 27.3 68.5 4 22.7 48 31.4 32.9
0.00037 17.82 300 26.5 66 7.5 22 42 26.8 22.7
0.00075 8.9 300 25 62 12.5 20.7 36 22.4 22.3
0.0015 4.46 300 22.3 56 22 18.7 32.5 19.3 18.2
0.0029 2.23 300 18.6 46.7 34 15.6 27.5 15.1 12.8
0.006 1.11 300 14 35 51 11.7 25 12.5 8.7

model consistent with the experimental data. The model seems to the effects of water removal (with kw = 0.1 s1) on CO2 conversion,
over-predict the hydrocarbon (other than methane) selectivity. C3H6 yield, and H2O selectivity. It shows that water reaches peak
Slight differences in the activation energies for our catalyst from value when the RWGS is completed and then begins to decrease
those reported in the literature [19] could be responsible for with distance along the reactor because of the water removal from
relatively small differences observed between the model predic- the reactor. As the water levels decrease, both CO2 conversion and
tions and the experimental data for the species concentrations. C3H6 yield increase with distance along the reactor. Indeed,
Until this point in the paper, the simulations shown do not predicted C3H6 yield increases to 63.3% at the end of the reactor for
include the water removal from the reactor (Eq. (20)). We will now the case when the water is removed, while the yield is only 27.5%
examine the effect of removing water from the reactor. Water without the water removal.
formation is a primary product of the reactions conditions shown
in Eqs. (7)–(9). The water formed under typical FT conditions is 4. Conclusions
known to influence catalyst activity primarily by re-oxidizing the
catalyst [33,34]. This in turn will affect the product selectivity. By In the quest to develop green alternative energy solutions,
removing or reducing the effects of water vapor on the catalyst anthropogenic CO2 is a unique carbon resource for the production
surface, the equilibrium at the catalyst surface should shift of energy-rich hydrocarbons. To take advantage of the higher
favorably to the production of desired intermediates such as concentrations of CO2 in both seawater and industrial flue gas,
olefins over the formation of CO or methane (Eqs. (7)–(9)). modular fixed-bed reactors could offer practical cost-effective
Therefore, we performed simulations by including Eq. (20) into the systems for onsite conversion of CO2 to higher hydrocarbons. To
model. Eq. (20) represents the rate (kw) at which water vapor is benefit from these modular designs in the future, our catalyst
removed from the reactor by some means. This could be achieved process was transitioned into a fixed-bed tubular reactor. The
in practice by condensing the water and removing the liquid water results presented show the effects of GHSV on CO2 conversion and
from the fixed bed reactor. Fig. 6 shows the model predictions for product selectivity under fixed-bed microchannel reactor condi-
tions. The results indicate that the maximum C2–C5+ yield
obtained in these fixed-bed experiments is 49% higher than that
obtained by CSTR. However when the performance of both reactors
100
is compared under similar GHSV conditions, C2–C5+ yield is
reduced by 47% in the fixed-bed conditions. The C2–C5+ yield
(% ) CO2, CH4, C3H6, Water

measured in the fixed bed reactor is consistent with those reported


80
by Riedel et al. under similar GHSV, pressure, and temperature
condition. However at higher GHSV our catalyst appears more
active.
60
There was a small amount of liquid hydrocarbon found in the
product stream that was neglected in the analysis of the data.
Modeling and kinetic analysis of the data indeed show the FT
40
reaction rates are much slower than RWGS reaction rate. As a
result, the RWGS reaches equilibrium within 10% of the reactor
length. Subsequently, the equilibrium slowly shifts toward CO
20
formation at the rate the FT reaction proceeds. Therefore, FT
reaction is the rate-controlling step and leads to the formation
of unreacted CO in the product stream as observed in the
0
experiments. This could be due to catalyst preparation, Fe and K
0 5 10 15 20 loading, physical size and properties. An alternate explanation
Time, s may be found in the composition of hydrocarbon product which
might affect the transport of species to the catalyst sites and the
Fig. 6. Comparison of model prediction for CO2 conversion, C3H6 yield, and water
overall rate of the FT reaction. Furthermore, in this evaluation
with and without water removal from the catalytic tubular reactor (total
flow = 400 mL/min, catalyst mass 17.82 g, GHSV = 0.00037 L/s-g). CO2 conversion we expand on the limited kinetic analysis of CO2 hydrogenation
(*), C3H6 yield (&), H2O (~) with water removal. CO2 conversion (*), C3H6 (&), by including the formation of methane as a by-product in our
H2O (~) without water removal. model and by providing kinetic constants. This model is used to
64 H.D. Willauer et al. / Journal of CO2 Utilization 3–4 (2013) 56–64

show how water removal from the reactor may significantly [13] R.W. Dorner, D.R. Hardy, F.W. Williams, H.D. Willauer, Energy and Environmental
Science 3 (2010) 884–890.
increase the hydrocarbon yield and CO2 conversion. [14] T. Takahashi, W.S. Broecker, A.E. Bainbridge, The Alkalinity and Total Carbon
The modeling and kinetic analysis in these studies will be Dioxide Concentration in the World Oceans. In Carbon Cycle Modelling, Vol. 16;
applied to the further tailor catalyst properties and reactor design SCOPE, John Wiley & Sons, New York, NY, USA, 1981, pp. 271–286.
[15] T. Takahashi, W.S. Broecker, S.R. Werner, A.E. Bainbridge, Carbonate Chemistry of
in efforts to facilitate greater CO2 conversion and olefin selectivity the Surface of the Waters of the World Oceans, in: E.D. Goldberg, Y. Horibe, S.
at higher GHSV in fixed-bed microchannel reactor conditions. Katsuko (Eds.), Isotope Marine Chemistry, Uchida Rokakuho Pub. Co., Tokyo,
Japan, 1980, pp. 291–326.
[16] H.D. Willauer, F. DiMascio, H.R. Hardy, M.K. Lewis, F.W. Williams, Industrial and
Acknowledgments Engineering Chemistry Research 50 (2011) 9822–9876.
[17] R.W. Dorner, D.R. Hardy, F.W. Williams, H.D. Willauer, Catalysis Communications
This work was supported by the Office of Naval Research both 15 (2011) 88–92.
[18] R.W. Dorner, D.R. Hardy, F.W. Williams, H.D. Willauer, Applied Catalysis A:
directly and through the Naval Research Laboratory. The research General 373 (2010) 112–121.
was performed while MTO and DMD held National Research [19] T. Riedel, G. Schaub, K.-W. Jun, K.-W. Lee, Industrial and Engineering Chemistry
Council Research Associateships at the Naval Research Laboratory. Research 40 (2001) 1355–1363.
[20] M.-J. Choi, J.-S. Kim, H.-K. Kim, S.-B. Lee, Y. Kang, K.-W. Lee, Korean Journal of
Chemical Engineering 18 (2001) 646–651.
References [21] Y. Zhang, G. Jacobs, D.E. Sparks, M.E. Dry, B.H. Davis, Catalysis Today 71 (2002)
411–418.
[1] Petroleum Quality Information System Report (PQIS), Defense Logistics Agency [22] S.R. Deshmukh, A-L.Y. Tonkovich, K.T. Jarosch, L. Schrader, S.P. Fitzgerald, D.R.
Energy, DLA Energy-QT, Fort Belvoir Virginia, 2010. Kilanowski, J.J. Lerou, J. Mazanec, Industrial and Engineering Chemistry Research
[2] Defense Logistics Agency: Energy, http://www.energy.dla.mil/DLA_finance_e- 49 (2010) 10883–10888.
nergy/Pages/dlafp03.aspx (accessed 14.11.12). [23] Oil & Gasengineer, February 2010, pp. 11–14, www. Engineerlive.com.
[3] H.D. Willauer, D.R. Hardy, K.R. Schultz, F.W. Williams, Journal of Renewable and [24] R.L. Espinoza, A.P. Steynberg, B. Jager, A.C. Vosloo, Applied Catalysis A 186 (1999)
Sustainable Energy-AIP American Institute of Physics 4 (2012), art. no. 033111-13. 13–26.
[4] Report of the Defense Science Board Task Force on DoD Energy Strategy, More [25] W.-P. Ma, Y.-L. Zhao, Y.-W. Li, Y.-Y. Xu, J.-L. Zhou, Reaction Kinetics and Catalysis
Fight-Less Fuel (Office of the Under Secretary of Defense For Acquisition, Tech- Letters 66 (1999) 217–223.
nology, and Logistics. Washington, DC 20301, February 2008). [26] D.L. Trimm, Catalysis Today 37 (1997) 233–238.
[5] K. Deffeyes, Hubbert’s Peak: The Impending World Oil Shortage, Princeton Univ. [27] H. Habazaki, M. Yamasaki, B.-P. Zang, A. Kaswashima, S. Kohno, T. Takai, K.
Press, Princeton & Oxford, 2001. Hashimoto, Applied Catalysis A: General 172 (1998) 131–140.
[6] T. Coffey, D.R. Hardy, G.E. Besenbruch, K.R. Schultz, L.C. Brown, J.P. Dahlburg, [28] G.W. Norval, The Canadian Journal of Chemical Engineering 86 (2008) 1062–
Defense Horizons 36 (2003) 1–11. 1069.
[7] B.H. Davis, Topics in Catalysis 32 (2005) 143–168. [29] J. Wu, M. Saito, M. Takeuchi, T. Watanabe, Applied Catalysis A 218 (2001) 235–
[8] G.A. Olah, A. Goeppert, G.K.S. Prakash, Beyond Oil and Gas: The Methanol 240.
Economy, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, 2006. [30] M. Saito, T. Fujitani, M. Takeuchi, T. Watanabe, Applied Catalysis A: General 138
[9] V. Nikulshina, C. Gebald, A. Steinfeld, Chemical Engineering Journal 146 (2009) (1996) 311–318.
244–248. [31] J. Barbier, Applied Catalysis A 23 (1986) 225–243.
[10] H.D. Willauer, F. DiMascio, D.R. Hardy, M.K. Lewis, F.W. Williams, Industrial and [32] J.A. Moujlin, A.E. van Diepen, F. Kapteijen, Applied Catalysis A 212 (2001) 3–16.
Engineering Chemistry Research 51 (2012) 11254–11260. [33] M. Bayat, M.R. Rahimpour, Journal of Natural Gas Science and Engineering 9
[11] I. Ganesh, Materials Sciences and Applications 2 (2011) 1407–1415. (2012) 73–85.
[12] F.T. Zangeneh, S. Sahebdelfar, M.T. Ravanchi, Journal of Natural Gas Chemistry 20 [34] R. Guettel, U. Kunz, T. Turek, Chemical Engineering and Technology 31 (2008)
(2011) 219–231. 746–754.

You might also like