Fuel Processing Technology 151 (2016) 19–30

Contents lists available at ScienceDirect

Fuel Processing Technology

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

Research article

A kinetic model for the synthesis of ethanol from syngas and methanol
over an alkali-Co doped molybdenum sulfide catalyst: Model building
and validation at bench scale
M.A. Portillo ⁎, A.L. Villanueva Perales, F. Vidal-Barrero, M. Campoy
Bioenergy Group, Department of Chemical and Environmental Engineering, School of Engineering, University of Seville, Camino de los Descubrimientos s/n, 41092 Seville, Spain

a r t i c l e i n f o a b s t r a c t

Article history: A set of experiments was run in laboratory and bench scale reactors over an alkali-Co doped molybdenum sulfide
Received 3 February 2016 catalyst in order to develop and validate the kinetics of higher alcohol synthesis from syngas/methanol mixtures.
Received in revised form 30 April 2016 The purpose is to generate a kinetic model useful to design a mixed alcohol process where methanol is recycled to
Accepted 18 May 2016
the reactor to increase ethanol production. Experimental data from a laboratory scale reactor were fitted using a
Available online 1 June 2016
power-law model, considering a CO insertion reaction scheme. The model was validated with a group of exper-
Keywords:
iments carried out at a bench scale reactor, consisting of experiments with methanol co-feeding and gas recycle.
Kinetic model Finally, we also studied whether a model would be able to predict the effect of methanol co-feeding if its param-
Ethanol eters were determined by fitting only the experiments where methanol co-feeding was not performed.
Methanol co-feeding © 2016 Elsevier B.V. All rights reserved.
Bench scale validation

1. Introduction on a fractional-factorial experimental design, in which 0, 2 and 4%

mole concentration of methanol in the feed were evaluated. Results
Gasification of biomass to syngas, followed by catalytic conver- indicated that methanol co-feeding enhances CO conversion and
sion of synthesis gas to mixed alcohols is a promising route for the the productivity of ethanol and higher alcohols but also hydrocar-
production of bioethanol [1]. Improvements in catalyst design and bons. Quarderer et al. [6] reported significant improvement in higher
process development need to be achieved to make this route com- alcohol productivity using a K/MoS2 catalyst by injecting methanol
mercially viable [2]. To the best of our knowledge, presently there up to 7.5% mole concentration in the feed. Portillo et al. [7] studied
is only one thermochemical process, proposed by Enerkem company, the effect of methanol addition at different reaction temperatures
operating at commercial scale to produce ethanol, besides methanol (280 to 300 °C) and methanol concentrations in the feed (from 0%
and other chemicals [3]. In this case ethanol is planned to be pro- to 8.5%). The experiments revealed that carbon monoxide conver-
duced by an indirect route using methyl acetate as intermediate sion, ethanol and higher alcohol productivities increase linearly
product. However, ethanol production is not expected until the sec- with methanol content in the feed at any reaction temperature,
ond half of 2016. while hydrocarbon productivity increases exponentially. Therefore,
The reaction mechanism of mixed alcohol synthesis over alkali- there is a trade-off between increasing alcohol productivity and the
doped molybdenum sulfide catalysts was studied by Santiesteban selective conversion of methanol to alcohols. The addition of metha-
et al. [4], who concluded that ethanol is produced from methanol nol has a positive influence in the ethanol selectivity only at low
through a CO insertion mechanism. This means that in a mixed alco- methanol content in the feed so the optimum amount of methanol
hol process, methanol produced in the reactor could be recycled back recycle to the reactor in an ethanol production process with this
to the reactor along with syngas to further increase ethanol produc- kind of catalyst should be determined by an economic analysis [7].
tion. However, the literature where methanol co-feeding is experi- Recently, Rodrigo Suárez et al. [8] also studied the effect of methanol
mentally studied is very scarce. Gunturu [5] ran experiments based co-feeding for various mole concentrations of methanol in the feed
(1.5% to 10%) at typical operating conditions for this kind of catalyst
(71 bar, 340 °C, 6000 NL/h gcat, H2/CO = 1). The results indicated
that under these conditions, methanol recycle is not a viable option
⁎ Corresponding author. to boost higher alcohol production because it results in an increase
E-mail address: mportillo1@us.es (M.A. Portillo). in methane yield, while the effect on higher alcohols was negative

http://dx.doi.org/10.1016/j.fuproc.2016.05.027
0378-3820/© 2016 Elsevier B.V. All rights reserved.
20 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30

or not significant. Therefore, there is still not full agreement about reactor and maintain the activity (sulfidity) of the catalysts. The syn-
the effect of methanol addition. gas, with a molar ratio of H2/CO ~ 1, was stored in a tank using a dual
Experiments with methanol co-feeding may be of interest to gen- stage compressor. The total flow of syngas fed to the reactor was con-
erate a model able to predict how the catalyst may perform as a re- trolled with a mass flow controller (MFC3). In methanol co-feeding
sult of significant methanol recycle in order to design a process experiments, methanol was fed directly to the reactor using a HPLC
that optimizes methanol recirculation. Several kinetic models have pump.
been developed for alkali doped MoS 2 catalysts [9,10,11,12]. Of Before being fed to the reactor, the syngas passed through a car-
these, Gunturu [5] was the first to develop a kinetic model from an bonyl filter to prevent catalyst poisoning by Fe and Ni carbonyls.
experimental design where methanol mole fraction in the feed (0– The reaction was carried out in a high pressure fixed bed reactor
4%) was among the variables considered in the experiments. The ki- (3/4″ Sch 80 and 1500 mm length). A bed of catalyst particles
netic model consisted of a system of Langmuir-Hinshelwood and (100 g) was placed inside the reactor supported with CSi. Five spiral
power-law type reactions to describe the kinetic scheme proposed electrical resistances distributed along the reactor supply heat to
by Santiesteban [4]. Larsson et al. [13] slightly reformulated and maintain the reactor temperature, which was measured by five ther-
refitted the parameters of this kinetic model and validated its predic- mocouples positioned perpendicular to the bed. Each thermocouple-
tion capability against experiments that did not feature methanol co- resistance pair was associated with a PID controller that controls the
feeding. However, Philips et al. [14] used Larsson's kinetic model to temperature in a section of the reactor through a spiral resistance.
study the effect of methanol recycle at different temperatures Downstream of the reactor, the pressure was maintained with a
when designing a thermochemical ethanol production process pressure control valve. The reactor outlet stream was in vapor-
from biomass. Later, Dutta et al. [15] continued the design work of phase and an aliquot was analyzed online using a gas chromatograph
Phillips et al., but in this case they used a kinetic model of promoted detection system (Agilent 7890 A) equipped with a thermal conduc-
MoS2 catalyst developed by The Dow Chemical Company to optimize tivity detector and two flame ionization detectors, which use helium
methanol recycle. Unfortunately, this model is proprietary of Dow as carrier gas.
Chemical and it is not available publicly. The product separation system was comprised of a heat exchang-
In this study, experiments with and without methanol co-feeding er, a condenser, a storage tank for collecting liquid and a pressure
at different operating conditions and reactor scale (laboratory and control valve. The heat exchanger was a double pipe exchanger
bench) were carried out with the aim of generating and validating whose function was to cool and condense the gases down to 30 °C.
a kinetic model able to predict how the catalyst may perform as a re- The unreacted synthesis gas and products circulate inside the inner
sult of significant methanol co-feeding, which is not available in the tube. The flow of cooling water in the outer tube was controlled
literature. with a regulating valve. The condenser further cooled the gases
down to 20 °C with a Peltier cell and separated the gas-phase and
the liquid-phase. To control the pressure in the condenser and pre-
2. Experimental
vent loss of ethanol in the gas-phase, a pressure control valve was
used.
2.1. Supply and characterization of catalyst
The gas-phase stream can be purged or compressed with a single
stage compressor and then be stored in a recycle storage tank. Once
An alkali-Co doped MoS2 catalyst produced by an industrial manu-
there is sufficient gas in the tank, it is analyzed, and can be fed to the re-
facturer based on a patented formulation was used (named as catalyst
actor together with syngas in order to study the effect of the gas recycle
I in the patent) [16]. The contents of Mo, Co and K were determined
to the reactor.
using an ICP Horiba Jobin Yvon to confirm that the measured composi-
tion of the catalyst was that of the patent (Table 1). The BET surface of
2.3. Methodology
the catalyst was 6.56 m2/g.

A reaction scheme for higher alcohol synthesis over sulfide mo-

2.2. Experimental setup lybdenum catalysts was taken from the literature and a power law
model was assumed for the reaction rate of compounds. Experiments
Catalytic studies were carried out at lab and bench scale facilities. at laboratory and bench scale reactors were carried out to obtain ex-
The lab scale experimental setup consisted of a high-pressure fixed perimental data that allow regression and validation of the kinetic
bed reactor (inner diameter: 8.3 mm, length: 330 mm, stainless steel) parameters, respectively. For the fitting, the differential mole bal-
where a bed of catalyst particles (5 g) was placed inside the reactor sup- ance equations in the tubular reactor for the main compounds (car-
ported with silicon carbide. Details of the experimental setup at lab scale bon monoxide, hydrogen, methanol, ethanol, propanol, methane,
have been published elsewhere [7]. carbon dioxide and water) needed to be solved. For that purpose,
Fig. 1 shows a diagram of the experimental setup at bench scale. an initial guess of the kinetic parameters was previously obtained
Syngas (CO + H2), N2 and two mixtures of H2S and H2 were supplied with a simplified method that assumes constant molar flow along
from pressurized cylinders. Nitrogen was used to test leaks and to the reactor so the differential system can be transformed into an al-
flush the system. A mixture of 10% H2S in H2 was used for pretreat- gebraic system, where the kinetic parameters are estimated using
ment of the catalyst, while a mixture of 400 ppm H 2 S in H2 was non-linear regression. Details of the methodology are explained
used to maintain 50 ppm H2S concentration in the feed gas to the below.

2.3.1. Reaction scheme

The chosen reaction scheme is based on the CO insertion mechanism
Table 1
proposed by Santiesteban [4] and Park [9] as it has been successfully
Targeted/measured atomic ratios of catalyst.
used to describe the kinetics of higher alcohols synthesis over MoS2 cat-
Atomic ratios alysts [4,10]:
Targeted Measured
CO þ 2H2 →CH3 OH ðIÞ
K/Mo Co/Mo K/Mo Co/Mo

0.50 1.20 0.42 1.27 CH3 OH þ H2 →CH4 þ H2 O ðIIÞ

 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30 21

Fig. 1. Simplified diagram of the experimental setup at bench scale.

CH3 OH þ CO þ 2H2 →C2 H5 OH þ H2 O ðIIIÞ the WGS reaction calculated as Kws = exp (− ΔG∘WS/(R · T)) where
ΔG∘WS(cal/mol) = − 8514 + 7.71 · T(K) [12], and kx is a parameter that
C2 H5 OH þ CO þ 2H2 →C3 H7 OH þ H2 O ðIVÞ takes into account the non-ideality of gases due to high operating pres-
sures. Partial reaction orders in the WGS reaction have been assumed
CO þ H2 O↔CO2 þ H2 ðVÞ equal to 1 to reduce the number of parameters. The independent vari-
ables are the temperature and the partial pressures of reactants, while
All reactions are assumed irreversible except the water gas shift Ai, Eai, A to J and kx are the parameters of the kinetic model (overall
(WGS) reaction. 21 parameters).

2.3.2. Model equations 2.3.3. Calculation of kinetic parameters using a rigorous method
The power-law model was utilized because it is simple and widely The parameters of the kinetic model were regressed from data of ex-
used in the literature to predict the reaction rate for the higher alcohol periments carried out in the lab reactor with and without methanol co-
synthesis [10,12,17]. In this work, the gross reaction rates of the com- feeding (Section 2.4). By assuming ideal plug flow for the lab reactor, the
pounds are expressed as follows: mass balances of the compounds are:


−Ea1 A dFCO  
r MeOH ¼ A1  exp P CO  P BH2 ð1Þ ¼ ‐ r MetOH þr EtOH þr PrOH þr CO2 ð6Þ
RT dW


−Ea2 C dF H2  
r CH4 ¼ A2  exp P H2  P DMeOH ð2Þ ¼ ‐ 2r MetOH þ2rEtOH þ2r PrOH þrCH4 þr CO2 ð7Þ
RT dW


−Ea3 E dF H2 O
r EtOH ¼ A3  exp P CO  P HF 2  P GMeOH ð3Þ ¼ r EtOH þr PrOH þr CH4 ‐r CO2 ð8Þ
RT dW


−Ea4 H dF MetOH
r PrOH ¼ A4  exp J
P CO  P IH2  P EtOH ð4Þ ¼ r MetOH −r EtOH −r CH4 ð9Þ
RT dW


−Ea5   dF EtOH
¼ r EtOH −r PrOH ð10Þ
rCO2 ¼ A5  exp PCO  PH2 O −ðkx =Kws ÞPCO2 PH2 ð5Þ
RT dW

where Ai is the pre-exponential factor, Eai is the activation energy, A to J dF PrOH

¼ rPrOH ð11Þ
are the partial reaction orders, Kws is the equilibrium constant of dW
22 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30

Table 2
dF CH4
¼ rCH4 ð12Þ Experiments of group 1. Pressure and inlet H2/CO mole ratio were kept constant at 90 bar
dW and 1, respectively. The results of these experiments are shown in Table B.1 of the appen-
dix B

dFCO2 Experiment T (ºC) W/F (kgcat·s/kmol)

¼ rCO2 ð13Þ
dW 1 290 46,670
2 290 57,930
where W is the mass of catalyst and Fi the mole flow of compound i. The 3 290 74,670
4 290 93,330
mass balances are expressed as a function of the five independent gross
5 290 115,820
reaction rates (Eqs. 1–5). 6 290 152,730
The fitting was carried out by minimizing an objective function 7 300 46,670
(Je). According to the maximum likelihood principle, if the errors in 8 300 57,930
the observations are independent, normally distributed, with con- 9 300 74,670
10 300 93,330
stant variance for each dependent variable (homoscedasticity), and 11 300 115,820
the covariance between dependent variables is negligible, the 12 300 152,730
resulting objective function is defined as the sum of the squares of 13 310 46,670
the residuals divided by the variance of the error for each dependent 14 310 57,930
15 310 74,670
variable (Eq. 14) [18,19]. By using this maximum likelihood function,
16 310 93,330
the variance of the fitting parameters is minimized in the regression 17 310 115,820
[20]. 18 310 152,730

XC
1 XN
Je ¼ ð Fij − Fij Þ2 ð14Þ
σ 2 Next, the initial guess of the kinetic parameters was determined
i¼1 i j¼1
with non-linear regression by minimizing the sum of the squares of
the differences between experimental and predicted gross reaction
In the objective function, Fij is the mole flow of compound i at the rates for each compound (Eq. 16).
outlet of the reactor for experiment number j; σ2i is the variance of the
experimental error for compound i, which is estimated as the variance X
N  2

of the residuals of compound i for all experiments; N is the number of Si ¼ r ij −r ij i ¼ 1……5 ð16Þ
experiments and C the number of compounds (carbon monoxide, hy- j¼1
drogen, methanol, ethanol, propanol, methane, carbon dioxide and
water). For Eq. 16, the partial pressures in the kinetic equations were evalu-
For each experiment, the predicted mole flows at the reactor outlet ated at the outlet of the reactor. The kinetics equations for the reaction
were calculated by integrating the differential equations (Eqs. 6 to 13) rates were the same as those used for the rigorous model (Eqs. 1–4) ex-
along the reactor with ode15s routine in Matlab®. The function cept the WGS reaction, which has been simplified (Eq. 17). The reason is
fminsearch of Matlab's Optimization Toolbox, which implements the that by using Eq. 5 for the WGS reaction, or the irreversible version, a
Nelder-Mead algorithm, was used to minimize the error function Je by good fit was not possible and negative values for the frequency factor
changing the parameters of the kinetic model. The variances of the re- were obtained. Therefore, in the simplified method Eq. 17 was used
siduals were updated in each iteration. for obtaining a rough estimation of A5 and Ea5 in Eq. 5.


2.3.4. Estimation of kinetic parameters using a simplified method −Ea5 h c i
To solve the differential mole balance mentioned above, an initial rCO2 ¼ A5  exp PCO  PdH2 O ð17Þ
RT
guess of the kinetic parameters is necessary. This has been obtained
using a simplified method whereby the integration of the differential 2.4. Catalytic studies
equations along the reactor is avoided. The method assumes that the
total molar flow rate (F) remains constant along the reactor. Thus, the The aim of these experiments was to obtain experimental data for
left-hand term of the differential equations, i.e. the net reaction rate, development of the kinetic model. In all experiments, there is neither
can be expressed as follows (Eq. 15): internal nor external diffusion resistance, so the controlling resistance
is the chemical reaction (see appendix A).
dF i dð Fyi Þ Fdðyi Þ dðy Þ
¼ ≈ ¼
i ð15Þ
dW dW dW W
d
F

where yi is the molar composition of component i at the outlet of the

reactor and F the total mole flow at the inlet of the reactor. To calcu-
late the net reaction rate several series of experiments were carried
out at fixed temperature, pressure and inlet H 2/CO ratio while
varying the inlet molar flow rate to generate a range of W/F (see
Section 2.4, Table 2). For each series of experiments at a fixed
temperature the value of each yi was plotted against W/F (see for in-
stance, Fig. 2 for ethanol). For a given W/F the slope of the curve is
the net reaction rate of compound i at the outlet of the reactor.
Thus, for each experiment the net reaction rates of five components
(namely methanol, ethanol, propanol, carbon dioxide and methane)
were calculated from the slope of the curves and then the values of Fig. 2. Plot of outlet mole ethanol concentration versus space velocity for obtaining ethanol
the five gross reaction rates were determined using Eqs. 9–13. reaction rates in the simplified method ( 583 , 573 and 563 K ).
 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30 23

Table 3 Table 5
Experiments of group 2. Factors selected and levels in the experimental design. Table B.2 in Experiments of group 4. The results of the experiments are shown in Table B.3 in the
the appendix B shows all these experiments (20) and their results. appendix B. Pressure was kept constant at 110 bar in all tests.

Experimental factor Levels Mole fraction in reactor feed (%)

Low (−) Middle (0) High (+) T GHSVa CO H2 MetOH CH4 CO2
(°C) (h−1)
Temperature (°C) 290 300 310
Pressure (bar) 80 90 100 A 290 1479 50.77 49.23 0.00 0.00 0.00
GHSV (h−1) 824 1036 1248 293 1479 50.77 49.23 0.00 0.00 0.00
H2/CO 1 1.25 1.5 300 1479 50.77 49.23 0.00 0.00 0.00
304 1479 50.77 49.23 0.00 0.00 0.00
300 1198 50.77 49.23 0.00 0.00 0.00
300 1004 50.76 49.24 0.00 0.00 0.00
Four groups of experiments were performed, groups 1, 2 and 3 at
B 300 1007 41.62 47.75 0.00 3.14 7.49
laboratory scale and group 4 at bench scale. Experiments of group 1 300 1202 41.01 47.67 0.00 3.36 7.97
(Table 2) were performed to apply the simplified method as ex- 305 1007 41.69 47.68 0.00 3.14 7.49
plained before. Experiments of group 2 and 3 were used for the C 293 1475 47.72 46.27 6.02 0.00 0.00
297 1475 47.72 46.27 6.02 0.00 0.00
fitting of the kinetic parameters with the rigorous method. Group 2
300 1475 47.72 46.27 6.02 0.00 0.00
is an experimental design at different conditions of T, P, GHSV and D 293 1472 38.21 44.93 5.34 3.43 8.10
H2/CO to obtain experimental data in a wide range of operating con- 300 1467 37.90 44.79 5.33 3.52 8.46
ditions, typical for this kind of catalyst (Table 3). Group 3 are exper- 305 1457 36.87 44.87 5.30 3.83 9.13
iments where methanol co-feeding was performed and were used to Subgroups: A) Experiments feeding syngas from pressurized cylinders. The gas from the
enhance the prediction capability of the kinetic model for syngas/ reactor was stored in a pressurized tank, in order to be used as gas recycle in experiments
methanol mixtures. Several series of experiments were carried out of group B; B) Syngas from pressurized cylinders and gas from recycle storage tank;
C) Methanol co-feeding experiments. Syngas from pressurized cylinders. The mole con-
at fixed temperature, pressure, inlet H2/CO ratio and space velocity
centration of methanol in the feed was fixed at ~6% because this concentration provided
while varying the concentration of methanol in the feed (Table 4). the largest ethanol selectivity in these experiments; D) Experiments where syngas, recycle
Experiments of group 4 were used to validate the kinetic model. gas and methanol were fed to the reactor.
They were carried out in a bench scale facility with a double purpose: Note: aGSHV is calculated based on CO + H2 flow in the feed.
to verify the kinetic model at a larger scale and at operating condi-
tions similar to an industrial process, that is, considering the recycle
of byproducts such as CO2 and hydrocarbons. The experiments can because the scheme of CO insertion correctly predicts the experi-
be organized into 4 different subgroups depending on the composi- mental results.
tion of the gases fed to the reactor (Table 5). It is necessary to validate the assumptions of the regression model
by residual analyses. The independency of the errors is fulfilled because
experiments were performed randomly. To verify that the errors follow
3. Results and discussion a normal distribution, the Shapiro-Wilk contrast test was used. The con-
trast test considers as null hypothesis (Ho) that the errors are normally
3.1. Kinetic parameters distributed, and as alternative hypothesis (H1) that they do not. The de-
cision rule, expressed in terms of the p-value is: Ho is rejected if p-value
The parameters of the kinetic model calculated using the simplified is less or equal to α (the significance level).
method are shown in Table 6. Experiments 6, 12 and 18 of group 1 As shown in Table 8, the p-value is always higher than 0.05 (chosen
were excluded for the fitting because the assumption of the simplified significance level) except for propanol. Since propanol is a minor com-
method (constant molar flow along the reactor) was largely violated, pound in the experiments, the model is accepted, even though the nor-
so the reaction rate could not be determined properly as the slope of mality assumption for propanol is slightly violated.
the curve yi-W/F. To validate the hypothesis of homoscedasticity, the estimations of
The kinetic parameters from Table 6 were used as initial values the model were plotted against the residuals for each variable (Fig. 4).
for the rigorous method. The initial value of kx in Eq. 5 was assumed This figure demonstrates that the errors are randomly distributed,
1, which would be the case if the behavior of the gases were ideal. with a mean of zero and constant variance. Trends are also not observed,
The parameters of the kinetic model calculated using the rigorous so for each compound the homoscedasticity hypothesis can be
method are shown in Table 7. With these parameters, the error func- accepted.
tion J e was reduced by 67% with respect to the initial parameter Table 9 shows a comparison between the activation energies of
values. higher alcohol synthesis reactions over MoS2 based catalysts reported
Fig. 3 shows the fit between experimental mole flows and those in the literature with those obtained in this work. Perhaps, the most
predicted by the model for carbon monoxide, hydrogen, methanol, striking feature of the kinetic data of this table is the lack of agreement
methane, ethanol, propanol, carbon dioxide and water. The model between the values obtained in the studies. However, the reaction rates
shows a good fit with the experimental results so the chosen reaction are affected by the composition, the structure of the active phase of the
scheme and the assumption of the reaction orders for WGS reaction catalyst and the support and reaction conditions and this may explain
are correct. Christensen et al. [17] recently suggested that alcohol
coupling might be one of the possible routes over MoS2-based cata-
Table 6
lysts but the coupling mechanism is absent or negligible in our case Calculated values of kinetic parameters using the simplified model.

Aa Ea CO H2 MetOH EtOH
Table 4 (kJ/mol) order order order order
Experiments of group 3. Pressure, inlet H2/CO mole ratio and GSHV were kept Methanol 0.0037 79 1.70 0.53 – –
constant at 110 bar, 1 and 1300 h-1, respectively. The results of these experi- Ethanol 0.010 80 1.31 0.14 0.56 –
ments (15) have been published in [7] Propanol 987.85 152 1.35 0.89 – 0.74
Methane 2885.76 115 – 0.06 0.52 –
T (ºC) 280 290 300
CO2 11.85 100 – – – –
Inlet mole methanol concentration (%) 0 2.98 4.68 5.79 8.44 a reactionorders
Units depend on the kinetic expression (kmol/(kg s·bar∑ )).
24 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30

Table 7 Table 8
Calculated values of kinetic parameters using the rigorous method. Hypothesis contrast of normality (Shapiro-Wilk).

Aa Ea CO H2 MetOH EtOH Residuals Shapiro-Wilk

(kJ/mol) order order order order
Statistical p-value
Methanol 0.0036 83.16 1.93 0.44 - -
Carbon monoxide 0.964 0.306
Ethanol 0.014 83.29 1.34 0.24 0.67 -
Hydrogen 0.949 0.103
Propanol 414.96 159.09 1.39 1.22 - 0.75
Methanol 0.958 0.201
Methane 3,152.07 114.58 - 0.06 0.53 -
Methane 0.967 0.368
Carbon dioxide 10.65 57.18 - - - -
Water 0.981 0.801
kx 0.85 Ethanol 0.939 0.051
Propanol 0.912 0.009
Note:
Carbon dioxide 0.968 0.381
a
Units depend on the kinetic expression (kmol/(kg·s·barΣ reaction orders)).

the wide dispersion of results. Overall, the results show that there is a
general agreement that the activation energy tends to increase with 3.2. Validation of the kinetic model
the chain length of alcohols [22]. Moreover, the activation energy of
methane (main hydrocarbon) is approximately equal in the different To validate the kinetic model, its predictive performance was tested
studies and it is always higher than those of alcohols. Therefore, an in- against a group of experiments that were not used for the fitting. These
crease in temperature favors the formation of hydrocarbons but also experiments were carried out in the bench scale facility (Group 4 exper-
higher alcohols. iments, see Section 2.4). Fig. 5 shows the comparison between the molar

Fig. 3. Comparison plots (group 2 and 3 experiments) between the experimental and model flow rates for the different compounds (With methanol co-feeding ▲, Without methanol co-
feeding ●).
 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30 25

Fig. 4. Residuals vs. predicted values of regression variables.

flow of compounds predicted by the model and the experimental data However, we can conclude that the kinetic model developed at lab
at bench scale. The results indicate that the model captured the intrinsic scale can be used to predict the performance of the catalyst at a larger
kinetics and there was not overfitting in determination of the parame- scale because, in general, all the compounds fit reasonably well with
ters. The model slightly underpredicts hydrogen, probably due to the experimental values.
the smaller accuracy in the quantification of hydrogen by gas chroma-
tography at the bench scale plant than at the laboratory scale reactor 3.3. Need for methanol co-feeding experiments
(hydrogen atom balance versus internal standard, respectively). On
the other hand, some experimental values for propanol are slightly A question remains whether a model could correctly predict the ef-
overpredicted, possibly due to uncertainties in the measurements as fect of methanol co-feeding if its parameters were determined by fitting
the amount of propanol produced in the experiments was small. experiments where methanol co-feeding was not performed. To answer

Table 9
Comparison of activation energies (kJ/mol) of higher alcohol synthesis reactions over MoS2 based catalysts reported in the literature with those obtained in this work.

Reference Present work Gunturu et al. [5] Santiesteban et al. [4] Surisetti et al. [23] Christensen et al. [22]

Catalyst K/Co/MoS2 K/Co/MoS2/C Cs/MoS2 Co-Rh-Mo-K/MWCNTs K/Co/MoS2/C

Operating conditions T(°C) 280–320 300–350 235–275 275–340 300–350

P (bar) 70–110 40–70 82.7 55.2–96.5 100
H2/CO 0.75–1.75 0.5–2 0.96 0.5–2 1
Methanol cofeeding Yes Yes No No No
EaMeOH (kJ/mol) 83.1 – 68 35 49
EaEtOH (kJ/mol) 83.3 38 95 57 76
EaPrOH (kJ/mol) 159.1 98 98 94 109
EaCH4(kJ/mol) 115 107 – 112 118
26 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30

this question, the fitting of the kinetic parameters was carried out again of the model to predict the effect of significant methanol content
only with experiments of group 2. Then, they were compared with the in the feed was evaluated. A reaction scheme from the literature (re-
parameters of Table 7, and the relative change in the parameters is actions I-V), based on the CO insertion mechanism, was used to de-
shown in Table 10. The results indicate that there is significant variation scribe the kinetics of the higher alcohols synthesis and a power-law
in some parameters between both cases. With these new parameters model was chosen for the reaction rate of compounds (Eqs. 1–5).
the mean square error on the validation experiments (group 4) has dou- Experiments at laboratory and bench scale reactors, which featured
bled, however, the fit seems acceptable for prediction purposes (Fig. 6). co-feeding of methanol, were carried out to obtain experimental
Therefore, it is possible to predict reasonably the effect of methanol data that allow regression (Table 7) and validation of the kinetic pa-
recirculation with a kinetic model that has been developed without rameters, respectively. Methane, carbon dioxide, methanol, ethanol
experiments where methanol is co-fed, at the expense of a moderate and n-propanol were the main products. For the fitting, the differen-
increase in the uncertainty of the predictions. tial mole balance equations in the reactor for the main compounds
were solved (Eqs. 6–13). An initial guess of the kinetic parameters
4. Conclusions was previously obtained using a simplified method. The hypothesis
of the regression model was confirmed. Validation of the model with
A kinetic model of an alkali-Co doped MoS2 catalyst for the data from the bench scale reactor indicates that the model captured
higher alcohol synthesis reaction was developed and the capability the intrinsic kinetics and there was not overfitting in determination

Fig. 5. Comparison plots between the experimental and model flow rates for the different compounds (feeding Syngas ●, Syngas + recycle X, Syngas + methanol ▲
Syngas + recycle + methanol ■) to validate the kinetic model.
 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30 27

Table 10
Relative change between the calculated values of kinetic parameters using all tests and tests without methanol co-feeding.

A Ea CO order H2 order MetOH order EtOH order

Methanol -1.3% 3.7% 10.1% -15.4% - -

Ethanol 25.7% 4.8% 0.4% 39.3% 7.3% -
Propanol -146.4% 1.0% 5.3% 21.8% - -1.7%
Methane 9.9% -0.8% - -10.8% -15.0% -
Carbon dioxide -11.1% -11.5% - - - -

kx -12.9 %

of the parameters (Fig. 6). Moreover, the kinetic model developed at lab model as selectivity of C4+ alcohols in all the experiments was very low
scale can be used to predict the performance of the catalyst at a larger (below 1%). This situation is expected when operating conditions are
scale. The accuracy of the kinetic model is expected to be high if the re- chosen to favor ethanol selectivity. Finally, it has been concluded that
action conditions are such that formation of C4+ alcohols is negligible a kinetic model could reasonably predict the effect of methanol co-
(for instance, operating at high space velocity where chain growth is feeding if its parameters were determined by fitting experiments
not favored). Formation of C4+ alcohols was not included in the kinetic where methanol co-feeding was not performed.

Fig. 6. Comparison between the experimental and model flow rates for bench experiments using the parameters calculated with experiments where methanol was not co-fed.
28 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30

Acknowledgements A.1. External mass-transfer diffusion

This work has been carried out in the framework of the CENIT The diffusion of the reactants or products between the bulk fluid
Project BIOSOS (CEN-20091040) funded by the Spanish Ministry of and the external surface of the catalyst is known as external resis-
Science and Innovation and Abengoa Bioenergy New Technologies tance. This can be eliminated by increasing the velocity past the par-
(ABNT). Their support is gratefully acknowledged. ticle or using very small particles [21] so the worst case is the lowest
gas velocity and the highest particle size. The particle size selected in
Nomenclature this work is 310 μm. The lowest velocity is 608 h− 1, corresponding to
ρb density of catalyst bed (kg/m3) the case of W/F = 152730 kgcat·s/kmol.
ρc density of catalyst (kg/m3) To evaluate the influence of external diffusion, the Mears criteri-
C total number of compounds on is used [22]. The Mears criterion states that external mass trans-
CAb concentration of reactant A in bulk gas phase (kmol/m3) port limitations are absent, if the following condition is met:
CAs concentration of reactant A at the surface of the catalyst 0

(kmol/m3) −r obs  ρb  dP  nA
C Mears ¼ b0:15 ðA:1Þ
De effective diffusion coefficient in the pores of the catalyst (m2/s) 2  kc  C Ab
dp diameter of catalyst pellets (m)
If CMears ≪0.15 there is no external diffusion limitation and if CMears
F total molar flow (mol/h)
≫0.15 the external diffusion limits the reaction severely [12,22]. The
Fi molar flow of compound i (mol/h)
value of CMears was calculated at high velocity and at three different
kc mass transfer coefficient (m/s)
temperatures (290, 300, 310 °C).
N total number of tests
r'obs and CAb vary along the reactor so the output values have been
nA reaction order of reactant A
used. The CMears values are of the order of 1·10−5 in the worst case,
Pi partial pressures of compounds i (bar)
which is far b 0.15. This confirms that external diffusion is negligible
r radius of the catalyst particle (m)
on the catalyst with the average particle diameter used.
rij reaction rate value of compound i for experiment j (kmol/kg s)
-robs observed reaction rate (kmol/kg s)
A.2. Internal mass-transfer diffusion
W mass of catalyst (g)
yi mole fraction of compound i
The Weisz-Prater criterion (CWP) is used to estimate the internal
mass-transfer resistance in heterogeneous catalytic reactions. If CWP
Subscripts
≪ 1 there is no internal diffusion limitation and if CWP ≫ 1 the inter-
j experiments
nal diffusion limits the reaction severely [12]. This criterion is given
i compounds
as follows:
Superscript 0
−robs  ρc  r2
^ predicted by the model CWP ¼ b1 ðA:2Þ
De  CAs

Appendix A. External and internal mass-transfer diffusion The value of CWP was calculated following the same methodology as
Surisetty and Christensen in their thesis [22,23]. It was considered that
The heterogeneous catalytic reaction is associated with external and the surface concentration of the component A was similar to that of
internal diffusion resistances. Reaction rates in the absence of these in- bulk liquid (CAs = CAb) because there was no resistance to the external
ternal and external diffusion resistances are termed ‘intrinsic kinetics’ transfer. The CWP values are about 10−4, which is far b 1. This confirms
[12]. The absence of these effects in the tests was analyzed to verify that internal diffusion is negligible on the catalyst with the average par-
that the chemical reaction is the controlling resistance. ticle diameter used.

Appendix B. Results of experiments

Table B.1
Results of experiments of group 1.

Experiments T (K) W/F \(kgcat·s/ kmol) mol/h in mol/h out

CO H2 CO H2 MetOH CH4 H2O EtOH PrOH CO2

-3 -4 -5 -3 -5
1 563 46,670 0.166 0.166 0.150 0.143 7.9·10 9.4·10 7.4·10 2.3·10 8.7·10 3.7·10-3
2 563 57,930 0.134 0.134 0.116 0.112 7.2·10-3 1.1·10-3 4.2·10-5 2.5·10-3 9.1·10-5 4.0·10-3
3 563 74,670 0.104 0.104 0.088 0.077 6.6·10-3 1.4·10-3 7.0·10-5 3.0·10-3 1.9·10-4 4.9·10-3
4 563 93,330 0.083 0.083 0.066 0.063 5.1·10-3 1.4·10-3 4.4·10-5 2.7·10-3 1.8·10-4 4.7·10-3
5 563 115,820 0.067 0.067 0.044 0.042 4.9·10-3 2.2·10-3 8.0·10-5 3.1·10-3 3.0·10-4 6.4·10-3
6 563 152,730 0.051 0.051 0.033 0.034 3.1·10-3 1.6·10-3 6.2·10-5 2.5·10-3 2.5·10-4 5.0·10-3
7 573 46,670 0.166 0.166 0.145 0.142 9.2·10-3 1.6·10-3 7.4·10-5 3.3·10-3 1.7·10-4 5.8·10-3
8 573 57,930 0.134 0.134 0.108 0.103 8.1·10-3 2.0·10-3 7.6·10-5 3.7·10-3 2.4·10-4 6.8·10-3
9 573 74,670 0.104 0.104 0.081 0.069 6.8·10-3 2.4·10-3 9.3·10-5 4.2·10-3 3.4·10-4 7.9·10-3
10 573 93,330 0.083 0.083 0.058 0.057 5.2·10-3 2.5·10-3 7.0·10-5 3.7·10-3 3.4·10-4 7.7·10-3
11 573 115,820 0.067 0.067 0.037 0.037 4.0·10-3 3.5·10-3 1.3·10-4 3.8·10-3 5.3·10-4 9.4·10-3
12 573 152,730 0.051 0.051 0.025 0.029 2.7·10-3 2.7·10-3 1.3·10-4 2.9·10-3 4.3·10-4 7.4·10-3
13 583 46,670 0.166 0.166 0.136 0.122 1.0·10-2 3.0·10-3 1.2·10-4 5.0·10-3 3.8·10-4 9.5·10-3
14 583 57,930 0.134 0.134 0.100 0.095 8.3·10-3 3.3·10-3 1.2·10-4 5.0·10-3 4.6·10-4 1.0·10-3
15 583 74,670 0.104 0.104 0.070 0.059 6.6·10-3 4.0·10-3 1.7·10-4 5.4·10-3 6.7·10-4 1.2·10-3
16 583 93,330 0.083 0.083 0.050 0.049 4.9·10-3 3.9·10-3 1.3·10-4 4.6·10-3 6.7·10-4 1.1·10-3
17 583 115,820 0.067 0.067 0.030 0.031 3.5·10-3 4.9·10-3 2.6·10-4 4.3·10-3 9.4·10-4 1.3·10-3
18 583 152,730 0.051 0.051 0.019 0.024 2.5·10-3 4.1·10-3 2.5·10-4 3.1·10-3 7.4·10-4 1.0·10-3

Note: W/F is the mass of catalyst (5 g) divided by the total molar flow. Pressure was kept constant at 90 bar in all tests.
 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30 29

Table B.2
Results of experiments of group 2 (experimental design). The central point (tests 4, 9, 15 and 18) was repeated four times to estimate the experimental error and check catalyst stability.
Load of catalyst is 5 g.

mol/h in mol/h out

GHSV
Experiments T (ºC) P (bar) (h–1) H2/CO CO H2 CO H2 MetOH CH4 H2O EtOH PrOH CO2

1 290 100 1248 1.5 0.083 0.126 0.061 0.097 8.7 · 10–3 2.1 · 10–3 2.0 · 10–4 4.0 · 10–3 3.0 · 10–4 5.9 · 10–3
2 290 80 824 1 0.070 0.070 0.053 0.049 4.6 · 10–3 1.5 · 10–3 1.0 · 10–4 2.6 · 10–3 2.0 · 10–4 4.8 · 10–3
3 310 80 824 1.5 0.056 0.083 0.024 0.047 4.1 · 10–3 5.2 · 10–3 5.0 · 10–4 3.3 · 10–3 5.0 · 10–4 1.0 · 10–2
4 300 90 1036 1.25 0.078 0.096 0.050 0.064 6.3 · 10–3 3.1 · 10–3 2.0 · 10–4 4.1 · 10–3 4.0 · 10–4 8.4 · 10–3
5 310 100 1248 1 0.104 0.104 0.066 0.059 6.4 · 10–3 4.4 · 10–3 3.0 · 10–4 5.6 · 10–3 7.0 · 10–4 1.3 · 10–2
6 290 80 1248 1 0.104 0.104 0.092 0.082 6.2 · 10–3 1.1 · 10–3 1.0 · 10–4 2.3 · 10–3 1.0 · 10–4 3.8 · 10–3
7 310 100 824 1 0.070 0.070 0.033 0.034 3.4 · 10–3 4.9 · 10–3 4.0 · 10–4 4.4 · 10–3 9.0 · 10–4 1.3 · 10–2
8 310 80 1248 1.5 0.083 0.126 0.051 0.083 8.1 · 10–3 4.7 · 10–3 4.0 · 10–4 4.6 · 10–3 5.0 · 10–4 1.1 · 10–2
9 300 90 1036 1.25 0.078 0.096 0.052 0.066 6.4 · 10–3 3.0 · 10–3 2.0 · 10–4 4.1 · 10–3 4.0 · 10–4 8.1 · 10–3
10 290 100 824 1.5 0.056 0.083 0.035 0.062 6.0 · 10–3 2.6 · 10–3 2.0 · 10–4 3.5 · 10–3 3.0 · 10–4 6.8 · 10–3
11 300 90 1460 1.25 0.110 0.137 0.092 0.101 9.0 · 10–3 2.8 · 10–3 2.0 · 10–4 4.2 · 10–3 3.0 · 10–4 8.1 · 10–3
12 300 90 612 1.25 0.046 0.056 0.020 0.031 2.9 · 10–3 3.4 · 10–3 3.0 · 10–4 3.0 · 10–3 4.0 · 10–4 8.2 · 10–3
13 300 90 1036 1.75 0.064 0.113 0.036 0.075 7.0 · 10–3 3.8 · 10–3 3.0 · 10–4 3.8 · 10–3 3.0 · 10–4 8.5 · 10–3
14 320 90 1036 1.25 0.078 0.096 0.034 0.050 4.3 · 10–3 7.3 · 10–3 7.0 · 10–4 4.5 · 10–3 1.0 · 10–3 1.6 · 10–2
15 300 90 1036 1.25 0.078 0.096 0.052 0.066 6.0 · 10–3 3.0 · 10–3 2.0 · 10–4 3.9 · 10–3 4.0 · 10–4 8.3 · 10–3
16 300 110 1036 1.25 0.078 0.096 0.047 0.059 6.4 · 10–3 3.7 · 10–3 3.0 · 10–4 5.2 · 10–3 5.0 · 10–4 1.1 · 10–2
17 300 70 1036 1.25 0.078 0.096 0.059 0.075 6.1 · 10–3 2.2 · 10–3 1.0 · 10–4 3.1 · 10–3 2.0 · 10–4 6.0 · 10–3
18 300 90 1036 1.25 0.078 0.096 0.052 0.067 6.1 · 10–3 3.0 · 10–3 2.0 · 10–4 3.9 · 10–3 3.0 · 10–4 8.2 · 10–3
19 280 90 1036 1.25 0.078 0.096 0.067 0.080 5.3 · 10–3 9.0 · 10–3 1.0 · 10–4 1.9 · 10–3 1.0 · 10–4 3.1 · 10–3
20 300 90 1036 0.75 0.099 0.075 0.077 0.047 4.7 · 10–3 2.0 · 10–3 1.0 · 10–4 3.6 · 10–3 4.0 · 10–4 7.1 · 10–3

Table B.3
Results of group 4 experiments at bench scale.

mol/h in mol/h out

T (ºC) CO H2 MetOH CH4 CO2 CO H2 MetOH CH4 H2O EtOH PrOH CO2

A 290 2.38 2.31 0.00 0.00 0.00 1.880 2.007 0.107 0.043 0.005 0.074 0.006 0.150
293 2.38 2.31 0.00 0.00 0.00 1.902 2.009 0.134 0.034 0.005 0.074 0.005 0.139
300 2.38 2.31 0.00 0.00 0.00 1.768 1.955 0.134 0.046 0.006 0.094 0.008 0.190
304 2.38 2.31 0.00 0.00 0.00 1.634 1.902 0.123 0.062 0.008 0.111 0.011 0.242
300 1.93 1.87 0.00 0.00 0.00 1.313 1.527 0.102 0.054 0.006 0.094 0.009 0.204
300 1.62 1.57 0.00 0.00 0.00 1.018 1.259 0.083 0.078 0.006 0.089 0.010 0.201
B 300 1.49 1.71 0.00 0.11 0.27 1.152 1.500 0.077 0.183 0.008 0.050 0.005 0.366
300 1.76 2.05 0.00 0.15 0.34 1.420 1.821 0.088 0.183 0.011 0.049 0.004 0.433
305 1.49 1.71 0.00 0.11 0.27 1.045 1.446 0.075 0.200 0.010 0.063 0.008 0.398
C 293 2.38 2.31 0.30 0.00 0.00 1.634 1.929 0.201 0.096 0.009 0.137 0.012 0.332
297 2.38 2.31 0.30 0.00 0.00 1.527 2.063 0.196 0.118 0.012 0.152 0.014 0.380
300 2.38 2.31 0.30 0.00 0.00 1.446 2.009 0.185 0.139 0.014 0.158 0.016 0.421
D 293 2.15 2.53 0.30 0.19 0.46 1.661 2.464 0.214 0.210 0.016 0.081 0.006 0.679
300 2.13 2.52 0.30 0.20 0.48 1.554 2.384 0.193 0.229 0.020 0.103 0.009 0.737
305 2.09 2.54 0.30 0.22 0.52 1.473 2.357 0.177 0.247 0.022 0.115 0.012 0.796

Note: Load of catalyst is 100 g.

Pressure was kept constant at 110 bar in all tests.
30 M.A. Portillo et al. / Fuel Processing Technology 151 (2016) 19–30

References gasification-based biorefining in the kraft pulp and paper industry, Volume 2: de-
tailed biorefinery design and performance simulationDecember 2006.
[1] J. He, W.-n. Zhang, Research on ethanol synthesis from syngas, J. Zhejiang Univ. Sci. [14] S. Phillips, A. Aden, J. Jechura, D. Dayton, T. Eggeman, Thermochemical Ethanol via
A 9 (5) (2008) 714–719. Indirect Gasification and Mixed Alcohol Synthesis of Lignocellulosic BiomassGolden
[2] J.J. Spivey, A. Egbebi, Heterogeneous catalytic synthesis of ethanol from biomass- Golden (CO): National Renewable Energy Laboratory 2007 April Report No.: NREL/
derived syngas, Chem. Soc. Rev. 36 (2007) 1514–1528. TP-510-41168, Contract No.: DE-AC36-99-GO10337. Sponsored by the Department
[3] R. Suarez Paris, et al., Catalytic conversion of biomass-derived synthesis gas to fuels, of Energy.
Catalysis: The Royal Society of Chemistry 27 (2015) 62–143. [15] A. Dutta, M. Talmadge, J. Hensley, M. Worley, D. Dudgeon, et al., Process Design and
[4] J.G. Santiesteban, Alcohol Synthesis From Carbon Monoxide and Hydrogen Over Economics for Conversion of Lignocellulosic Biomass to Ethanol: Thermochemical
Mos2-Based Catalysts, Lehigh University, 1989 Ph. D. Dissertation. Pathway by Indirect Gasification and Mixed Alcohol SynthesisGolden (CO): National
[5] A.K. Gunturu, Higher Alcohol Synthesis From Carbon Monoxide and Hydrogen: Ki- Renewable Energy Laboratory 2011 May Report No.: NREL/TP-5100-51400, Con-
netic Studies Over Mos2-Based CatalystsM.S. Ch.E. Thesis. West Virginia University tract No.: DE-AC36-08GO28308. Sponsored by the Department of Energy.
1997. [16] G. Prieto, J.M. Serra, A.M. Feliu, J.L.S. Yagüe, J.C. Bello, R.A. Antolín, Methods for
[6] Quarderer, et al., Preparation of ethanol and higher alcohols from lower carbon obtaining a multimetallic sulfurous catalyst and use thereof in a method for produc-
number alcoholsUnited State Patent number US4825.013 1989. ing higher alcohol by catalytic conversion of synthesis gas, 2012 (Patent n° US 2012/
[7] M.A. Portillo Crespo, A.L. Villanueva Perales, F. Vidal-Barrero, M. Campoy, Effects of 0202898 A1).
methanol co-feeding in ethanol synthesis from syngas using alkali-doped MoS2 cat- [17] J.M. Christensen, P.M. Mortensen, R. Trane, P.A. Jensen, A.D. Jensen, Effects of H2S
alysts, Fuel Process. Technol. 134 (2015) 270–274. and process conditions in the synthesis of mixed alcohols from syngas over alkali-
[8] R.S. París, M. Boutonnet, S. Järås, Effect of methanol addition on higher alcohol synthe- promoted cobalt-molybdenum sulphide, App. Catal. A 366 (2009) 29–43.
sis over modified molybdenum sulfide catalysts, Catal. Commun. 67 (2015) 103–107. [18] Y. Bard, Nonlinear Parameter Estimation, Academic Press, New York, 1974.
[9] T. Park, I. Nam, Y. Kim, Kinetic Analysis of mixed alcohol synthesis from syngas over [19] D. Flávia, Marques-Marinho, et al., Construction of analytical curve fit models for
K/MoS2 catalyst, Ind. Eng. Chem. Res. 36 (1997) 5246–5257. simvastatin using ordinary and weighted least squares methods, J. Braz. Chem.
[10] A.K. Gunturu, E.L. Kugler, J.B. Cropley, D.B.A. Dadyburjor, Kinetic model for the syn- Soc. 24 (9) (2013).
thesis of high-molecular-weight alcohols over a sulfided Co-K-Mo/C catalyst, Ind. [20] C.D. Knightes, C.A. Peters, Statistical analysis of nonlinear parameter estimation for
Eng. Chem. Res. 37 (1998) 2107–2115. monod biodegradation kinetics using bivariate data, Biotechnol. Bioeng. 69 (2)
[11] K.J. Smith, R.G. Herman, K. Klier, Kinetic modeling of higher alcohol synthesis over al- (July 2000).
kali-promoted Cu/ZnO and MoS2 catalysts, Chem. Eng. Sci. 45 (8) (1990) 2639–2646. [21] H.S. Fogler, Elements of Chemical Reaction Engineering, third ed. Prentice Hall PTR,
[12] V.R. Surisetty, A.K. Dalai, J. Kozinski, Intrinsic reaction kinetics of higher alcohol syn- Upper Saddle River, NJ, 1999.
thesis from synthesis gas over a sulfided alkali-promoted co-rh-mo trimetallic cata- [22] J.M. Christensen, Catalytic Synthesis of Long-Chained Alcohols From SyngasThesis
lyst supported on multiwalled carbon nanotubes (mwcnts), Energy Fuel 24 (2010) Technical University of Denmark, 2011.
4130–4137. [23] V.R. Surisetty, Research and Development of Co And Rh-Promoted Alkali-Modified
[13] E.D. Larsson, Final Report to Us Department of Energy and the American Forest & Molybdenum Sulfide Catalysts for Higher Alcohols Synthesis From Synthesis
Paper AssociationContract DE-FC26-04NT42260 A cost-benefit assessment of GasThesis University of Saskatchewan, 2010.