You are on page 1of 12

NJC

View Article Online


PAPER View Journal
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

Fischer–Tropsch synthesis over the Fe–Mn/Al2O3


catalyst: modeling and optimization of light
Cite this: DOI: 10.1039/d0nj04077k
olefins using the RSM method
Maryam Arsalanfar

Fe–Mn/Al2O3 catalysts were prepared via a co-precipitation procedure and used for light olefin production
via a CO hydrogenation reaction. The effect of preparation parameters including the precipitation ageing
time, pH and temperature on the catalytic performance has been investigated. The ageing time was
changed from 60 to 300 min, the precipitation pH was changed from 6.5 to 10.5 and the precipitation tem-
perature was investigated in the range of 40–80 1C. Modeling and optimization processes were performed
using response surface methodology (RSM). Polynomial equations for each response were successfully fitted
Received 13th August 2020, to the experimental data. According to the optimization results the best preparation conditions were found
Accepted 29th September 2020 to be aageing time = 160 min, pH = 8.78 and precipitation temperature = 74.16 1C. At these optimized con-
DOI: 10.1039/d0nj04077k ditions the maximum selectivity toward (C2–C4) light olefins and CO conversion% and minimum selectivity
toward methane were achieved. The effect of the preparation parameters was also investigated on the var-
rsc.li/njc ious surface reaction rates. Catalyst characterization was performed using XRD, BET, TGA, DSC and SEM.

1. Introduction can be obtained on Fe/Mn catalysts than on other iron-based


catalysts.14 In order to improve the FTS performance of Fe
The growing concerns about the depletion of crude oil have based catalysts, many studies have been carried out to investi-
boosted global interest in finding alternative feedstocks for gate the effects of many promoters on the FTS performances of
petrochemicals and liquid fuels.1,2 Fischer–Tropsch synthesis iron-based catalysts, especially alkalis, copper, manganese and
(FTS) is a crucial segment of coal/gas/biomass-to-liquid tech- silicon.15–22 The preparation parameters and procedures have a
nology, where a wide range of hydrocarbons are synthesized great influence on the catalytic performance and physico-
from carbon monoxide and hydrogen with water as a bypro- chemical properties of the catalysts for the CO hydrogenation
duct. FT catalysts are typically based on group VIII metals, i.e. reaction.23–29 In previous work we investigated the effect
Fe, Co, Ni and Ru. The most preferred catalysts are iron and of different preparation and operational parameters on the
cobalt based catalysts for industrial scale applications. Due to catalytic performance and structural properties of different
the low cost, high FTS and water–gas-shift (WGS) activity, and catalytic systems.30–33 The aim of the present work is to
feed flexibility (H2/CO = 0.5–2.5), Fe based catalysts are the investigate the effect of various precipitation parameters
preferred catalysts for FTS using low H2/CO ratio syngas derived including the precipitation ageing time, pH and temperature
from coal.3,4 A mixture of two or the most active catalysts is on the catalytic performance and structural characteristics of
often used in the FTS reaction because of the higher activity, the Al2O3 supported Fe–Mn catalyst. This investigation was
selectivity and stability than single component ones by chan- carried out using both experimental and RSM procedures, so
ging the physico-chemical properties of the final catalyst.5,6 The we can observe the effect of different preparation parameters
addition of Mn to Fe or Co catalysts brought about a significant on the catalytic performance independently (by the experi-
increase in light olefin formation and decrease in methane mental method) and also the influence of all of these para-
selectivity.7–10 High C2–C4 selectivity for the iron-rich Fe–Mn meters on the catalyst performance concurrently (via the RSM
solid has been correlated with the iron manganese oxide phase procedure). Beside the main purpose of the present work the
and two carbide phases, while the manganese-rich solid has changing of various surface reaction rates was also investi-
been correlated with two spinel phases and different carbide gated. The optimization of these parameters for C2–C4 light
phases.11–13 It is well known that higher selectivity of alkenes olefins was performed using response surface methodology;
furthermore, the physico-chemical properties of the bimetallic
Department of Chemistry, Amirkabir University of Technology, Hafez Ave., Tehran, catalyst were investigated using various techniques such as
Iran. E-mail: Maryam_galavy@yahoo.com XRD, BET, SEM, TGA and DSC.

This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020 New J. Chem.
View Article Online

Paper NJC

2. Materials and methods


2.1. Catalyst preparation
In order to prepare the 50% Fe–50% Mn/15 wt%Al2O3 catalyst
via the co-precipitation procedure the same molar ratios of
aqueous solutions of Fe(NO3)39H2O and Mn(NO3)24H2O (2 M)
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

were premixed and 15 wt% of Al2O3 (based on the total catalyst


weight) was added to the mixed nitrate solution. The surface
area, pore volume and pore size of Al2O3 were 207 m2 g1,
0.95 m3 g1 and 121 Å, respectively. The obtained solution was
heated at various temperatures from 40 to 80 1C in a round
bottomed flask. The precipitate agent of aqueous Na2CO3
solution (2 M) was added dropwise to the nitrate mixed solution
until the desired pH (6.5–10.5) was achieved. The resulting
suspension was then left in these conditions for the desired
duration (60–300 min). The aged precipitate was filtered and
washed several times with warm distilled water. Then the filtered
precipitate was dried in an oven at 120 1C for 16 h; in this step the
catalyst precursor was obtained. In order to obtain the final oxidic
catalyst the precursor was calcined in a furnace at 600 1C for 6 h.

2.2. Catalyst characterization techniques


2.2.1. X-ray diffraction (XRD). The crystalline phases of
the samples (precursor and calcined) were determined using
the XRD technique. The XRD patterns were collected for the
prepared samples using a Bruker AXS D8 Advance diffracto-
meter (Germany) using monochromatized CuKa radiation at Fig. 1 Schematic representation of the reactor used as a flow diagram.
40 kV and 30 mA, with a scan range of 0 o 2y o 701.
Determination of the present phases was performed by match-
ing the obtained patterns to entries in the Diffractplus version single tube vertical fixed bed micro reactor was employed. This
6.0 indexing software. used reactor was placed inside a tubular furnace (Atbin, Model
2.2.2. BET surface area measurements. The textural properties ATU 150-15). In order to provide a uniform wall temperature
and porosity of the catalysts were measured using a surface along the length of the reactor, an alumina jacket was placed in
area analyzer (Model Nova 2000, Quantachrome Instruments, the space between the tubular reactor and furnace inner wall.
USA) using nitrogen (99.99% purity) as the adsorption gas. The The heating zones ahead of the catalyst bed and the under-
catalyst samples were slowly degassed and heated to 300 1C for neath zone below the catalyst packing were filled with quartz
3 h under a nitrogen atmosphere. The samples were evacuated glass beads. The required amount of the catalyst was diluted
at 196 1C for 66 minutes in order to obtain the BET specific with asbestos and placed among the inert quartz glass beads.
surface areas, pore volume and pore sizes. Three separate thermocouples were used in order to check and
2.2.3. Scanning electron microscopy (SEM). The micro- control the temperature of the (A) preheating zone, (B) catalyst
scopic morphology of the prepared catalysts was observed on bed and (C) underneath zone. The inlet feed gas entered from
a Philips XL30 emission scanning electron microscope instru- ahead of the micro reactor and the obtained products exited
ment (Netherlands) operating at 20 kV. In order to avoid from the bottom of the reactor. The required amount of the
charging troubles before the measurement the samples were catalyst was diluted with asbestos and placed among the inert
coated with gold/palladium for 80 s in 20 s intervals. quartz glass beads. In each evaluation test the desired amount
2.2.4. Thermal gravimetric analysis (TGA) and differential of the catalyst (1.0 g) was in situ reduced under atmospheric
scanning calorimetry (DSC). The weight changes of the catalyst pressure using H2–N2 (H2/N2 = 1, each gas flow = 30 ml min1)
precursor with increasing temperature were investigated using a at 400 1C for 12 h before syngas exposure and then the CO
TGA/DSC simultaneous thermal analyzer (Rheometric Scientific, hydrogenation reaction was performed. In order to achieve
Model 1500+) under a dry air flow. In each measurement a selected steady state conditions after catalyst loading the reactor was
amount of sample (15–20 mg) was loaded and the temperature was operated for 12 h. The reactant gases and obtained products
increased at 10 1C min1 from ambient to 600 1C. were analyzed online using a gas chromatograph (Thermo
ONIX UNICAM PROGC+) equipped with a sample loop, two
2.3. Catalyst evaluation and apparatus Thermal Conductivity Detectors (TCD) and one Flam Ionization
All catalyst evaluation tests were performed in a fixed bed micro Detector (FID) able to perform the analysis of a wide variety of
reactor (Fig. 1); for this purpose a dichotomous stainless steel gaseous hydrocarbon mixtures. First the calibration process of

New J. Chem. This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020
View Article Online

NJC Paper

the gas chromatograph was performed by using pure and


mixed compound capsules and then the catalyst evaluations
were carried out. The following formulas were used in order to
calculate the conversion of carbon monoxide and the selectivity
of the product:
nCO;inlet  nCO;outlet
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

% CO conversion ¼  100 (1)


nCO;inlet

nCH4 ; produced
% CH4 selectivity ¼  100 (2)
nCO;inlet  nCO;outlet

nðC2 C4 Þ; produced


% ðC2  C4 Þ selectivity ¼  100 (3)
nCO;inlet  nCO;outlet

X
4
C5þ selectivity% ¼ 100  selectivity of Cn ð%Þ (4)
1

The effect of the precipitation parameters on the surface


reaction rates was studied using the following equations:
 n  PCO
FCO ¼ n  CCO ¼ (5)
RT
where:
FCO1 = inlet molar flow of CO (mol min1), n1 = volumetric
flow rate of input gas (ml min1)
CCO = concentration of CO (mol ml1), PCO = partial pressure
of CO (bar), T = gas temperature (K), R = universal gas constant
(ml bar mol1 K1).
In order to use the equation of a differential reactor for
Fig. 2 (A) Effect of the ageing time on the catalytic behavior of Fe–Mn/
calculation of the rate of CO consumption in a plug reactor the Al2O3 catalysts. (B) Effect of the ageing time on the surface reaction rates.
plug reactor equation was changed as we explained in our
previous work34,35 and finally eqn (6) was obtained.
 the better catalytic performance for light olefin production.
XCO FCO
rCO ¼ (6) This catalyst has lower selectivity toward the methane product
Wcat
(20%), higher selectivity toward light olefins (46%) and higher
CO conversion (87%). Furthermore, the effect of the ageing
time on the surface reaction rates was investigated and the
3. Results and discussion obtained results are illustrated in Fig. 2B. The reaction rate
measurements show that as the ageing time is increased
In the co-precipitation procedure for catalyst preparation there are from 60 to 180 min the CO consumption rate is increased from
many various precipitation parameters by which the catalytic 2.17  104 to 2.74  104 (mmol g1 min1) and then
performance and catalyst structure are strongly affected. In the with increasing this parameter from 180 to 300 min the CO
present work I attempt to investigate the effect of some precipita- consumption rate is decreased from 2.74  104 to 1.75 
tion parameters as I describe in the next section. For this purpose 104 (mmol g1 min1). The same trend is observed for the
both experimental and RSM procedures are employed. hydrocarbon production rate and the maximum rate value belongs
to an ageing time of 180 min (2.19  104 mmol g1 min1).
3.1. Effect of the co-precipitation parameters The methane production rate has shown an overall reducing
3.1.1. Effect of the precipitation ageing time. In order to trend and with increasing the ageing time from 60 to 300 min
investigate the effect of the ageing time on the catalytic perfor- the methane production rate slightly decreased from 0.694 
mance, a series of catalysts were prepared with different ageing 104 to 0.522  104 (mmol g1 min1).
times from 60 to 300 min and all of these prepared catalysts 3.1.2. Effect of the precipitation pH. For investigation of
were evaluated for the CO hydrogenation reaction under the the effect of the precipitation pH on the catalytic performance
same operational conditions (T = 300 1C, H2/CO = 2/1 and P = 1 of the Fe–Mn catalyst for light olefin production a series of
bar). The catalyst evaluation results for C2–C4 light olefin bimetallic catalysts were prepared with different precipitation pH
production are presented in Fig. 2A; these obtained results from 6.5 to 10.5. These prepared samples were tested under the
reveal that the sample which was aged for 180 min has shown same reaction conditions (T = 300 1C, H2/CO = 2/1 and P = 1 bar)

This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020 New J. Chem.
View Article Online

Paper NJC
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

Fig. 3 (A) Effect of the precipitation pH on the catalytic behavior of Fe–Mn/ Fig. 4 (A) Effect of the precipitation temperature on the catalytic behavior
Al2O3 catalysts. (B) Effect of the precipitation pH on the various surface of Fe–Mn/Al2O3 catalysts. (B) Effect of the precipitation temperature on
reaction rates. the different surface reaction rates.

higher values and the methane selectivity has the lower amount
and the examination results are presented in Fig. 3A. According to concurrently. The effect of the precipitation temperature on the
the obtained results the catalyst with a precipitation pH of 8.5 has surface reaction rates was also investigated and the obtained
shown the better catalytic performance; this catalyst has lower results are illustrated in Fig. 4B. These results show that with
selectivity toward methane, higher selectivity toward light olefins increasing the co-precipitation temperature from 40 to 80 1C
and higher CO conversion concurrently. The effect of the pre- the CO consumption and hydrocarbon production rates have
cipitation pH on the surface reaction rates was investigated and an increasing trend and the maximum values are observed at
the obtained results are illustrated in Fig. 3B. The surface reaction 70 1C. At this temperature the minimum value for methane
rate measurements show that with increasing the co-precipitation production is observed (0.563  104 (mmol g1 min1)).
pH from 6.5 to 10.5 the CO consumption rate passed from a 3.1.4. Stability evaluation. The catalyst stability test was
maximum value at pH = 8.5 (2.77  104 mmol g1 min1). performed under the process conditions of T = 300 1C and
A similar trend is observed for the hydrocarbon production H2/CO = 2/1 under atmospheric pressure for 120 h and the
rate and the maximum value belongs to pH = 8.5 (2.16  obtained results are displayed in Fig. 5.
104 mmol g1 min1). The opposite trend is observed for the
methane production rate and with increasing the pH from 6.5 to
10.5 the rate of methane production passed from a minimum
value at pH = 8.5 (0.609  101 mmol g1 min1).
3.1.3. Effect of the precipitation temperature. The effect of
the precipitation temperature on the catalytic performance of
Fe–Mn catalysts was investigated and for this purpose this
parameter was changed from 40 to 80 1C. These prepared
catalysts were used for the CO hydrogenation reaction under
the same process conditions (T = 300 1C, H2/CO = 2/1 and P = 1 bar)
and the obtained results are displayed in Fig. 4A.
These obtained results revealed that the sample with a
precipitation temperature of 70 1C has presented the better
catalytic behavior for C2–C4 light olefin production; in this case Fig. 5 Effect of time on stream on the catalytic performance of the Fe–Mn/
the CO conversion and light olefin selectivity have shown the Al2O3 catalyst at 300 1C and H2/CO = 2/1 under atmospheric pressure.

New J. Chem. This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020
View Article Online

NJC Paper
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

Fig. 7 TGA and DSC curves of the optimized Fe–Mn/Al2O3 catalyst precursor.

The optimized Fe–Mn/Al2O3 catalyst precursor was characterized


using the TGA and DSC methods and the obtained curves are
Fig. 6 The XRD patterns of the optimized Fe–Mn/Al2O3 catalyst in the
precursor and calcined states (fresh and spent). presented in Fig. 7. In these thermal analyses the temperature is
increased from ambient temperature to 600 1C and in this thermal
range the sample has shown an overall weight loss of 43 wt%.
These obtained results reveal that the supported Fe–Mn According to Fig. 7 the sample shows three weight loss steps
catalyst retains its catalytic performance without decreasing with the temperature increasing; the first weight loss step from
until 48 h and then a slight decrease is observed in the catalyst 40 to 110 1C can be attributed to the evaporation of physically
activity and light olefin production; the reverse trend is adsorbed water which is trapped in the sample tissue (in this
observed for the methane and C2–C4 alkane selectivity and step a weight loss of 3 wt% is observed). The second weight loss
these products are slightly increased. The selectivity of C5+ from 110 to 210 1C is related to decomposition of hydroxy or
products shows little changes during the stability test. In total, basic nitrates; this weight loss step produces water and nitro-
the stability evaluation results show that the supported Fe–Mn gen oxides (a weight loss of 11 wt% is observed in this step).
catalyst preserves a desirable level of performance during the The final weight loss step occurred at around 280–450 1C and it
stability test. was due to the decomposition of hydroxy or basic carbonates,
so the third weight loss step produces water and carbon dioxide
3.2. Catalyst characterization concurrently (during this step a weight loss of 29 wt% is
In order to investigate and determine the phases of the opti- observed). For the catalyst decomposition and weight loss the
mum prepared Fe–Mn/Al2O3 catalyst in the cases of the pre- same trend has been reported for the TGA results.38,39
cursor and calcined (fresh and used), all of these samples were DSC characterization results provide more evidence in
characterized using powder-X-ray diffraction and the obtained agreement with the TGA results; the DSC profile shows three
XRD patterns are illustrated in Fig. 6. The determination of the endothermic peaks, of which the first one appearing at around
catalyst phases was performed and in the precursor case only 40–110 1C is attributed to the removal of physically adsorbed
the MnCO3 phase was detected and presence of amorphous phases water. The second endothermic peak from 140 to 220 1C is
in this sample makes the other phases undetectable. The actual attributed to decomposition of hydroxy or basic nitrate phases
identified phases in the fresh sample were Fe2O3 (monoclinic), and the third endothermic peak from 270 to 400 1C is due
MnO2 (cubic), Mn2O3 (cubic) and AlFeO3 (orthorhombic). In order to decomposition of carbonates. In order to investigate the
to investigate the effect of the CO hydrogenation reaction on the optimized catalyst tissue and morphology the precursor and
catalyst phases the used sample was also characterized using XRD calcined catalysts were characterized using the SEM technique
and various phases of Fe3C (orthorhombic), Fe2C (monoclinic), and the obtained SEM images are shown in Fig. 8. As it can
Fe3O4 (monoclinic), MnAl2O4 (cubic) and MnO (cubic) were identi- be observed, the catalyst has shown different morphologies,
fied in the spent catalyst. It can be observed that the used sample is particle sizes and agglomeration in various cases. The precur-
comprised of oxidic and carbide phases, of which both of them sor sample is comprised of particles with different sizes, shapes
are active; it has been reported that oxidic phases are selective for and aggregation; this sample has a highly dense texture and
olefin production and carbide phases are active in the hydrogena- uneven tissue (Fig. 8a). The calcination process at 600 1C for 6 h
tion of CO.36,37 The catalyst weight loss measurement with increas- leads to major changes in the catalyst morphology and the
ing temperature was carried out using thermal analysis methods. calcined sample is comprised of spherical particles with low

This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020 New J. Chem.
View Article Online

Paper NJC
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

Fig. 8 SEM images of the optimized Fe–Mn/Al2O3 catalyst in the cases of the (A) precursor, (B) fresh calcined and (C) spent sample.

Table 1 BET results of the optimized Fe–Mn/Al2O3 catalyst in different which most of them have undefined geometric shapes; the
cases spent sample has high density agglomerates and contains fine
Specific surface Pore volume particles which are placed between the larger particles (Fig. 8c).
Sample area (m2 g1) (m3 g1) Pore size (Å) The optimized catalyst texture and porosity in various cases
Precursor 161.24 0.27 148.12 (precursor and calcined) were investigated using the BET
Fresh 152.45 0.33 140.46 method. The obtained BET results are reported in Table 1.
Used 149.53 0.31 129.32 These results show that the Fe–Mn/Al2O3 catalyst has different
textural properties and porosity in different cases; it can be
seen that the calcination process and CO hydrogenation
reaction lead to main changes in the specific surface area and
Table 2 Experimental data of the preparation parameters and the
pore sizes but the pore volumes are slightly changed. Table 1
responses values
also shows that the catalyst precursor has the highest specific
Run A (ageing B C CO (CH4) (C2–C4)Q surface area of 161.24 (m2 g1) in comparison with the fresh
no. time/min) (pH) (aPT/1C) conversion% selectivity% selectivity%
(152.45 m2 g1) and used (149.53 m2 g1) samples. According to
1 60 8.5 70 69 32 37 the BET results, the fresh calcined sample has a high specific
2 120 8.5 70 80 22 43
surface area and this characteristic helps the better catalytic
3 180 8.5 70 87 20 46
4 240 8.5 70 78 27 34 performance of the sample.
5 300 8.5 70 72 23 31
6 180 6.5 70 79 40 30
7 180 7.5 70 84 31 35
8
9
180
180
8.5
9.5
70
70
88
83
22
27
48
42
4. Modeling and optimization process
10 180 10.5 70 80 25 37
With consideration of the main effect of the preparation
11 180 8.5 40 67 30 32
12 180 8.5 50 71 33 36 parameters on the catalytic performance the impact of three
13 180 8.5 60 78 31 42 studied variables (ageing time, pH and precipitation tempera-
14 180 8.5 70 85 21 47
ture) was investigated using response surface methodology
15 180 8.5 80 83 25 45
(RSM). In the first section the influence of these parameters
a
PT: precipitation temperature. was investigated using an experimental procedure; in this
method the influence of each factor was investigated
while the other parameters were kept constant. But in the
density agglomerates and the adhesion phenomenon is not RSM procedure the effects of all of these parameters were
observed in this sample; the calcined catalyst has a homo- investigated concurrently and the interaction between these
genous texture and agglomeration (Fig. 8b). parameters could be observed. RSM is a powerful technique
This figure reveals that after the calcination procedure for chemical process optimization and modeling.40,41 The aim
greater particle dispersion is observed, which leads to better of RSM is to obtain a mathematical model which demonstrates
catalytic performance. After the CO hydrogenation reaction the process and explains the main factors and their interac-
the catalyst texture was changed again and the used sample tions. A suitable model is obtained from the most desirable
is comprised of particles with different sizes and shapes, of response by regulating the factors. In our previous work we

Table 3 The coded version of the obtained RSM models

Response Coded model


CO conversion% +78.67 + 0.80  A +7.35  B + 16.92  C  15.28  A2  5.75  B2  3.63  C2  8.92  C3
(C2–C4) selectivity% +41.12  4.20  A + 6.55  B + 7.75  C  11.27  A2  11.74  B2  4.21  C2

New J. Chem. This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020
View Article Online

NJC Paper

Table 4 Analysis of variance (ANOVA) for the CO conversion% response

Term Sum of squares df Mean square F-Value p-Value (prob 4 F) Coefficient estimate Standard error
Model 569.01 7 81.29 19.02 0.0005 — —
Constant — 1 — — — 78.67 1.28
A (ageing time) 1.60 1 1.60 0.37 0.5600 0.80 1.31
B (pH) 4.25 1 4.25 0.99 0.3519 7.35 1.35
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

C (PTa) 154.06 1 154.06 36.04 0.0005 16.92 2.82


A2 321.57 1 321.57 75.23 o0.0001 15.28 1.76
B2 45.56 1 45.56 10.66 0.0138 5.75 1.78
C2 12.37 1 12.37 2.90 0.1325 3.63 2.13
C3 29.56 1 29.56 6.92 0.0339 8.92 3.39
Residual 29.92 7 4.27 — — — —
R2 = 0.950 Radj2 = 0.900 Adeq-precision = 11.965
a
PT = precipitation temperature.

Table 5 Analysis of variance (ANOVA) for the (C2–C4)Q selectivity% response

Term Sum of squares df Mean square F-Value p-Value (prob 4 F) Coefficient estimate Standard error
Model 487.84 6 81.31 6.76 0.0083 — —
Constant — 1 — — — 41.12 1.88
A (ageing time) 44.10 1 44.10 3.67 0.0916 4.20 2.19
B (pH) 100.41 1 100.41 8.35 0.0202 6.55 2.27
C (PTa) 179.30 1 179.30 14.92 0.0048 7.75 2.01
A2 188.37 1 188.37 15.67 0.0042 11.27 2.85
B2 204.44 1 204.44 17.01 0.0033 11.74 2.85
C2 17.57 1 17.57 1.46 0.2612 4.21 3.48
Residual 36.16 8 12.07 — — —
R2 = 0.835 Radj2 = 0.7118 Adeq-precision = 7.031
a
PT = precipitation temperature.

used this method for light olefin production on the Fe–Co–Mn the present work, we used the RSM method for modeling
ternary and also Co–Ce binary catalytic systems.42–44 In and optimizing the Fe–Mn/Al2O3 catalyst and the experimental

Fig. 9 Comparison between the predicted and actual values (A1) and (B1); and comparison between the residual and predicted values (A2) and (B2) for
CO conversion% and (C2–C4)Q selectivity%.

This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020 New J. Chem.
View Article Online

Paper NJC

results were fitted to a full quadratic second order polynomial follows:


equation by applying multiple regression analysis as follows:
CO% ¼ þ 95:86689 þ 0:38871  Ageing Time þ 24:54769
X
n X
n n X
X n
 pH  10:10852  precipitation temperature
Y ¼ b0 þ bi Xi þ bii Xi2 þ bij Xi Xj þ E (7)
i¼1 i¼1 i¼1 j 4 1
þ 1:06124  103  ðAgeing TimeÞ2  1:43810
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

In this equation Y is the response function, Xi and Xj are  ðpHÞ2 þ 0:19164  ðprecipitation temperatureÞ2
independent variables, b0 is the intercept term, bi is the linear
 1:11502  103 ðprecipitation temperatureÞ3
effect of Xi, bii is the quadratic effect of Xi and bij is the two
factor interaction between Xi and Xj. The regression accuracy is (12)
obtained by the following equation:
ðC2  C4 Þ¼ ¼  268:40306 þ 0:24666  Ageing Time
SSE
R2 ¼ 1  (8) þ 51:98206  pH þ 1:65123
SST
 precipitation temperature  7:82387  104
2
R , often called the coefficient of determination, is defined as
the ratio of the sum of squares given by a regression model and  ðAgeing TimeÞ2  2:93424  ðpHÞ2
the ‘‘total’’ sum of squares around the mean.
 0:010532  ðprecipitation temperatureÞ2
The adjusted R2 is a modification of R2. The best way to
define this quantity is: (13)

MSE
Radj 2 ¼ 1  (9)
MST

where MST is the estimated variance of Y and is defined as


follows:

SST
MST ¼ (10)
ðn  1Þ

and MSE is the estimated residual or error variance and is


defined as follows:

SSE
MSE ¼ (11)
ðn  p  1Þ

where p is the number of predictors in the regression


equation. The statistical impact of the regression terms was
checked by Analysis of Variance (ANOVA). For the best
mathematical fitted model a high statistical significance of
P-value r 0.05 is offered by ANOVA; the F-value determines
the statistical significance of the second-order models. In
order to model and investigate the effect of the preparation
parameters on the catalytic behavior using RSM we consid-
ered and investigated two responses of the CO conversion%
and (C2–C4)Q selectivity%. In the RSM modeling we used
three independent variables which are coded as A, B and C
and refer to the ageing time, precipitation pH and precipita-
tion temperature, respectively; the effect of these variables
and their binary interactions was investigated. The obtained
results for these two responses are presented in Table 2.
Before investigation of the influence of the preparation
variables on the responses, using the ‘‘sequence model
sum of squares’’ test, the best models for CO conversion% Fig. 10 Perturbation plot of the (A) CO conversion% and (B) (C2–C4)Q
and(C2–C4)Q selectivity% are polynomial equations with selectivity% responses (A: ageing time, B: pH and C: precipitation
P-value o 0.0001. The obtained empirical equations are as temperature).

New J. Chem. This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020
View Article Online

NJC Paper
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

Fig. 11 3D and contour plots of the effects of the ageing time, precipitation pH and temperature on the CO conversion (%).

The coded versions of the obtained RSM models are also conversion% response (eqn (12)) the maximum effect of the
reported in Table 3. The collection of statistical parameters for variables belongs to the precipitation temperature with a
the responses was obtained sufficiently by ANOVA. coefficient estimate of 16.92. Regarding the light olefin
The obtained ANOVA results for two considered responses selectivity response, the obtained results of the fitted model
(CO conversion% and (C2–C4)= selectivity%) and also corres- (eqn (13)) presented in Table 5 show that the most important
ponding coefficients of R2 are reported in Tables 4 and 5. parameter which has a great effect on the (C2–C4)Q selectivity
For these models, the coefficients, R2-values, F-values and is also the precipitation temperature with a coefficient estimate
p-values are given by the software and therefore one can of 7.75.
confirm the importance of the experimental parameters. The The actual values of the CO conversion% and (C2–C4)Q
correlation coefficients R2 and Radj2 were evaluated in order to selectivity% responses versus the predicted values are plotted
determine the suitability of the model; a high value of R2 is in Fig. 9A1 and B1, respectively. As it can be seen, the suggested
obtained with adding or omitting the terms. The ANOVA results correlation describes the experimental values well. Fig. 9A2
presented in Tables 4 and 5 show that the responses of and B2 show the plots of the residual values versus the
the CO conversion% and (C2–C4)Q selectivity% have R2 values predicted values of the responses obtained from the fitted
of 0.950 and 0.835, respectively. According to the obtained models. These figures reveal that the points are dispersed
results presented in Table 4, for the fitted model of the CO around the horizontal axis.

This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020 New J. Chem.
View Article Online

Paper NJC
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

Fig. 12 3D and contour plots of the effects of the ageing time, precipitation pH and temperature on the (C2–C4) selectivity (%).

A perturbation plot is an important diagrammatic represen- temperature). The perturbation plot of the (C2–C4)Q selectiv-
tation in order to compare the influence of all factors at a ity% (Fig. 10B) shows curvature in all factors and demonstrates
particular point in the design space. In the perturbation curves that this response is sensitive to all factors. The best ways to
the response is plotted by changing only one factor over its describe the relationship between the independent variables
range while the other factors were kept constant. Curvature in a and the effect of the terms on the response are three dimen-
factor demonstrates that the response is sensitive to that factor sional (3D surface) and contour (2D) plots.45 Fig. 11(A–C) show
and a flat line reveals that the response is insensitive to that the 3D surface and contour plots of the CO conversion%. The
factor. A perturbation plot also could be used in order to find impact of the ageing time and precipitation pH on ppH on the
the factors with the greatest effect on the response when there CO conversion percent at a constant precipitation temperature
are more than two factors. of 60 1C is illustrated in Fig. 11A. This figure shows that as the
The perturbations plots of the CO conversion% and ageing time is increased from 60 to 300 min the CO conver-
(C2–C4)Q selectivity% responses are displayed in Fig. 10A and sion% response passed from the maximum value and then
B, respectively. In the perturbation plot for the CO conversion% slightly increased. The obtained results of the fitted model for
response (Fig. 10A) the steep curvature in the three factors CO conversion% reported in Table 4 confirm this observation;
shows that this response is sensitive to all evaluated factors these results show that the ageing time has a lower coefficient
and shows the highest sensitivity to factor C (precipitation than the pH and precipitation temperature and hence the CO

New J. Chem. This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020
View Article Online

NJC Paper

conversion% response is slightly influenced by the ageing time production with almost the same increasing slope. The ANOVA
factor in the evaluated range. Fig. 11B shows the 3D surface results of the (C2–C4)Q selectivity% presented in Table 5 are in
and contour plots of the CO conversion% and the impact of agreement with these observations; the precipitation pH and
the precipitation pH and temperature on this response at a temperature have a coefficient of 6.55 and 7.75, respectively.
constant ageing time of 180 min. As it can be seen, the CO Fig. 12C reveals the impact of the ageing time and tempera-
conversion% is slightly increased as the precipitation pH is ture of precipitation on the (C2–C4)Q selectivity%; as it can be
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

enhanced from 6.5 to 10.5. This response also shows an increas- seen, with increasing the ageing time from 60 to 300 min the
ing trend as the precipitation temperature is increased from 40 to light olefin selectivity is slightly decreased and passed from a
80 1C, but its increasing slope is greater than the rising slope of maximum at around 160 min. This figure also shows that as
the pH plot. The ANOVA results presented in Table 4 also confirm the precipitation temperature is increased from 40 to 80 1C the
these observations; according to these results the coefficient of the (C2–C4)Q selectivity% is greatly increased. This observation is
precipitation temperature is greater than the pH coefficient, so expected from the ANOVA results for the light olefin selectivity
the CO% is mostly affected by the thermal factor. response reported in Table 5; these results show that the
Fig. 11C shows the 3D surface and contour plots of the CO precipitation temperature factor has a higher coefficient than
conversion% and the effect of the ageing time and precipitation the ageing time factor.
temperature on this response at a constant pH of 8.50. This
figure shows that the CO conversion% response is increased 4.1. Determination of the optimum process conditions
with increasing both precipitation factors of the ageing time The aim of the present work is production of light olefins on
and temperature but the increasing slopes for these factors are the supported Fe–Mn catalyst and we try to determine the
different and the precipitation temperature has a greater best preparation conditions with higher CO conversion% and
impact on the CO conversion%. The ANOVA results and coeffi- (C2–C4)Q selectivity and lower methane selectivity concurrently.
cient estimates reported in Table 4 also confirm these observa- The optimization process was performed using the RSM
tions. The 3D surface and contour plots of the (C2–C4)Q method and fitted models; it was found that a pH of 8.78,
selectivity% are presented in Fig. 12(A–C). In Fig. 12A the precipitation temperature of 74.16 1C and ageing time of 160 min
impact of the two parameters of the precipitation pH and are the best preparation conditions with a desirability of 0.879.
ageing time on the (C2–C4)Q selectivity% response is shown. At these optimum conditions the CO conversion%, (C2–C4)Q
This figure reveals that at a constant precipitation temperature selectivity% and CH4 selectivity% are 85.11, 45.77 and 21.00,
of 60 1C, as the ageing time factor is increased the (C2–C4)Q respectively. The desirability chart of the preparation variables
response passed from a maximum at around an ageing time of and three responses is illustrated in Fig. 13.
160 min; in this ageing time range the (C2–C4)Q selectivity% is
slightly affected by the ageing factor. But the pH has a greater
effect on this factor and with increasing this parameter from 5. Conclusion
6.5 to 10.5 the (C2–C4)Q % response is increased with a high
slope. These results and observations were also confirmed by In the present work the effect of various preparation para-
the ANOVA results presented in Table 5 (the pH has a greater meters such as the ageing time, precipitation pH and tempera-
coefficient than the ageing factor). Fig. 12B shows the influence ture on the catalytic behavior and structure of the Fe–Mn/Al2O3
of the precipitation pH and temperature on the (C2–C4)Q catalyst was investigated. According to the usual experimental
selectivity% response at a constant ageing time (180 min). method the co-precipitated bimetallic catalyst shows better
This figure shows that both investigated factors have a great catalytic performance in the case of ageing time = 180 min,
effect on the (C2–C4)Q selectivity% response and increasing pH = 8.5 and a precipitation temperature of 70 1C. The
both of them leads to increasing the light olefin selectivity optimization and modeling processes were performed using
response. According to Fig. 12B both the pH and temperature of the response surface methodology method. In the RSM method
precipitation factors have an enhancing effect on light olefin the effects of the three mentioned preparation parameters
on the CO conversion% and (C2–C4)Q selectivity% responses
were evaluated. Finally, the product distribution was optimized,
in which maximization of the CO conversion% and (C2–C4)Q
selectivity% responses and minimization of the CH4 selectivity%
response occurred concurrently. Using the RSM procedure the
optimum preparation conditions were determined as ageing
time = 160 min, a pH of 8.78 and a precipitation temperature
of 74.16 1C. At these optimum preparation conditions the CO
conversion%, (C2–C4)Q selectivity% and CH4 selectivity% are
85.11, 45.77 and 21.00, respectively, with a desirability of 0.879.
According to the obtained results the optimum conditions deter-
Fig. 13 Desirability chart of the independent parameters and optimized mined via the usual experimental and RSM procedures are
responses. comparable and close to each other.

This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020 New J. Chem.
View Article Online

Paper NJC

Conflicts of interest 21 Y. Yang, H. W. Xiang, L. Tian, H. Wang, C. H. Zhang,


Z. C. Tao, Y. Y. Xu, B. Zhong and Y. W. Li, Appl. Catal., A,
There are no conflicts to declare. 2005, 284, 105.
22 M. Arsalanfar, A. A. Mirzaei, H. R. Bozorgzadeh, A. Samimi
References and R. Ghobadi, J. Ind. Eng. Chem., 2014, 20, 1313.
23 X. Wang, J. Zhang, J. Chen, Q. Ma, S. Fan and T.-S. Zhao,
Published on 30 September 2020. Downloaded by Auckland University of Technology on 10/24/2020 3:20:52 PM.

1 J. Li, Y. Hou, Z. Song, C. Liu, W. Dong, C. Zhang, Y. Yang Chin. J. Chem. Eng., 2018, 26, 761.
and Y. Li, J. Mol. Catal., 2018, 449, 1. 24 G. Jacobs, T. K. Das, Y. Zhang, J. Li, G. Racoillet and
2 M. Oschatz, S. Krause, N. A. Krans, C. Hernández Mejıá, B. H. Davis, Appl. Catal., A, 2002, 233, 263.
S. Kaskel and K. P. de Jong, Chem. Commun., 2017, 53, 10204. 25 G. Jacobs, Y. Ji, B. H. Davis, D. Cronauer, A. J. Kropf and
3 H. M. T. Galvis, J. H. Bitter, C. B. Khare, M. Ruitenbeek, C. L. Marshall, Appl. Catal., A, 2007, 333, 177.
A. I. Dugulan and K. P. De Jong, Science, 2012, 335, 835. 26 S. Iqbal, T. E. Davies, D. J. Morgan, K. Karim, J. S. Hayward,
4 M. Arsalanfar, A. A. Mirzaei, H. R. Bozorgzade and J. K. Bartley, S. H. Taylor and G. J. Hutchings, Catal. Today,
A. Samimi, Phys. Chem. Res., 2014, 2, 179. 2016, 275, 35.
5 K. Eisenacher and A. A. Adesina, Korean J. Chem. Eng., 2000, 27 A. A. Mirzaeia, M. Galavy, A. Beigbabaei and V. Eslamimanesh,
17, 71. J. Iran. Chem. Soc., 2007, 4, 347.
6 A. A. Mirzaei, A. Beig Babaei, M. Galavy and A. Youssefi, Fuel 28 A. A. Mirzaeia, M. Galavy and V. Eslamimanesh, Aust.
Process Technol., 2010, 91, 335. J. Chem., 2008, 61, 144.
7 Y. Yang, H. Xiang, R. Zhang, B. Zhong and Y. Li, Catal. 29 N. Lohitharn and J. G. Goodwin Jr, J. Catal., 2008, 260, 7.
Today, 2005, 106, 170. 30 A. A. Mirzaeia, S. Shahriari and M. Arsalanfar, J. Nat. Gas Sci.
8 S. L. González-Cortés, S. M. A. Rodulfo-Baechler, Eng., 2011, 3, 537.
A. Oliverosa, J. Orozco, B. Fontal, A. J. Mora and 31 M. Arsalanfar, A. A. Mirzaei, H. R. Bozorgzadeh, H. Atashi,
G. Delgado, React. Kinet. Catal. Lett., 2002, 75, 3. S. Shahriari and A. Pourdolat, J. Nat. Gas Sci. Eng., 2012,
9 M. Arsalanfar, A. A. Mirzaei and H. R. Bozorgzadeh, J. Ind. 9, 119.
Eng. Chem., 2013, 19, 478. 32 M. Abdouss, M. Arsalanfar, N. Mirzaei and Y. Zamani, Bull.
10 M. Arsalanfar, A. A. Mirzaei, H. R. Bozorgzadeh and Chem. React. Eng. Catal., 2018, 13, 97.
H. Atashi, J. Ind. Eng. Chem., 2012, 18, 2092. 33 M. Arsalanfar, A. A. Mirzaei and H. R. Bozorgzadeh, J. Nat.
11 G. P. Van der Laan and A. A. C. M. Beenackers, Catal. Rev., Gas Sci. Eng., 2012, 6, 1.
1999, 41, 255. 34 M. Arsalanfar, A. A. Mirzaei, H. Atashi, H. R. Bozorgzadeh,
12 J.-D. Xu, K.-T. Zhu, X.-F. Weng, W.-Z. Weng, C.-J. Huang and S. Vahid and A. Zare, Fuel Process Technol., 2012, 96, 150.
H.-L. Wan, Catal. Today, 2013, 215, 86. 35 A. A. Mirzaei, E. Rezazadeh, M. Arsalanfar, M. Abdouss,
13 J. B. Butt, Carbide phases on iron-based Fischer-Tropsch M. Fatemi and M. Sahebi, RSC Adv., 2015, 5, 95287.
synthesis catalysts—part II: some reaction studies, Catal. 36 H. B. Zhang and G. L. Schrader, J. Catal., 1985, 95, 325.
Lett., 1990, 7, 83. 37 M. D. Shroff, D. S. Kalakkad, K. E. Coulter, S. D. Kohler,
14 H. Schultz and H. Gökcebay, Fischer-Tropsch CO hydro- M. S. Harrington, N. B. Jackson, A. G. Sault and A. K. Datye,
genation as a means for linear olefins production, in J. Catal., 1995, 156, 185.
Catalysis of Organic Reactions, ed. J. R. Kosak, Marcel Dekker, 38 H. R. Azizi, A. A. Mirzaei, R. Sarani and M. Kaykhaii, Iran.
New York, NY, USA, 1984, pp. 153–169. J. Catal., 2019, 9(3), 223.
15 J. F. Li, X. F. Cheng, C. H. Zhang, J. Wang, W. S. Dong, 39 R. Roknabadi, A. A. Mirzaei and M. Mansouri, Adv. Environ.
Y. Yang and Y. W. Li, J. Chem. Technol. Biotechnol., 2017, Biol., 2014, 8(17), 719.
92, 1472. 40 A. Ameen, S. Mondal, S. M. Pudi, N. N. Pandhare and
16 J. F. Li, X. F. Cheng, C. H. Zhang, Q. Chang, J. Wang, P. Biswas, Energy Fuels, 2017, 31, 8521.
X. P. Wang, Z. G. Lv, W. S. Dong, Y. Yang and Y. W. Li, 41 E. Nourafkan, H. Gao, Z. Hu and D. Wen, Chem. Eng. Res.
Appl. Catal., A, 2016, 528, 131. Des., 2017, 125, 367.
17 J. F. Li, X. F. Cheng, C. H. Zhang, Y. Yang and Y. W. Li, 42 M. Akbari, A. A. Mirzaei, H. Atashi and M. Arsalanfar,
J. Mol. Catal. A: Chem., 2015, 396, 174. J. Taiwan Inst. Chem. Eng., 2018, 91, 396.
18 J. F. Li, C. H. Zhang, X. F. Cheng, M. Qing, J. Xu, B. S. Wu, 43 M. Shiva, H. Atashi, A. A. Mirzaei, M. Arsalanfar and A. Zare,
Y. Yang and Y. W. Li, Appl. Catal., A, 2013, 464–465, 10. Fuel, 2014, 123, 205.
19 H. Y. Suo, S. G. Wang, C. H. Zhang, J. Xu, B. S. Wu, Y. Yang, 44 M. Arsalanfar, A. Nouri and M. Abdouss, ChemistrySelect,
H. W. Xiang and Y. W. Li, J. Catal., 2012, 286, 111. 2019, 4, 10632.
20 H. J. Wan, B. S. Wu, C. H. Zhang, H. W. Xiang and Y. W. Li, 45 R. H. Myers, A. I. Khuri and W. H. Carter, Technometrics,
J. Mol. Catal. A: Chem., 2008, 283, 33. 1989, 31, 137.

New J. Chem. This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2020

You might also like