Paper 6
Paper 6
Paper 6
+ +
(R.3)
Oxygenates:
2nH
2
+ nCO C
n
H
2n
O + (n-1) H
2
O (R.4)
Type, composition and nature of catalyst govern
the selectivity of products from FT process, and hence,
proper selection of catalyst as well as optimization of the
process conditions is of vital importance in FT process.
The carbon number of the product hydrocarbon
(either chain or branched hydrocarbon) is determined by the
process conditions, viz. temperature, pressure and space
velocity. Increase in temperature results in shifts towards the
product with lower carbon number. Several previous authors
have described the effect of these parameters on the yield
and selectivity of the FT process. Dictor and Bel [7],
Donnelly and Scatterfield [8], Mirzaei et al. [9] and
Anderson [10] have studied the effect of temperature on FT
International Journal of Scientific Engineering and Technology (ISSN : 2277-1581)
Volume No.2, Issue No.1, pp : 31-39 1 Jan. 2013
IJSET@2013 Page 32
synthesis. They observed an increase of olefin to paraffin
ratio with increasing temperature for potassium promoted
precipitated iron catalyst. Effect of total pressure on the
product selectivity on alkaline promoted iron catalyst has
been reported by Dry [11]. Donnelly and Scatterfield [8]
found that increasing the H
2
/CO ratio in the reactor feed
results in production of lighter hydrocarbon (with lower
carbon number). The influence of the space velocity (or in
other words, residence time) of the synthesis gas on the
selectivity has been investigated by Bukur et al. [12].
Kuipers et al. [13] have investigated increase in the olefin to
paraffin ratio with increasing space velocity (and hence,
resulting in decrease of the conversion due to smaller
residence time) on a polycrystalline cobalt foil. Similar
studies have also been done by Bukur et al. [12] on a
commercial (Ruhrchemie) supported iron catalyst
(Fe/Cu/K/SiO
2
), and Iglesia et al. [14] on TiO
2
supported
ruthenium catalysts. Bukur et al. [12] have reported no
effect of the space velocity on the molecular weight (i.e. the
carbon number) of the hydrocarbons, while Iglesia et al. [14]
have observed an increase in the average molecular weight
of the products with decrease of the space velocity. The
selectivity to methane and olefins has been reported to
reduce with decreasing space velocity, while the selectivity
towards paraffins remains unchanged.
In this study, we have attempted to optimize the FT
process using commercial cobalt catalyst procured from
Albemarle, Netherlands using a fixed bed reactor. The
optimization variables are pressure and space velocity. The
temperature of the process has been kept constant, as the
effect of temperature on product distribution is well known
[15], in that higher temperatures favor smaller carbon
number hydrocarbons. The physical cause leading to this
effect is higher desorption of tendency of the reactants from
catalyst surface at elevated temperature. This optimization
study will give insight into the variation in product
selectivity in terms of carbon number with optimization
parameters using stoichiometric synthesis gas with ratio of
H
2
to CO as 2:1. Such a study is also important for the best
and effective use of the cobalt catalyst, taking into
consideration high cost of cobalt.
2. Experimental
2.1. Catalyst
The cobalt based catalyst was procured from M/s
Albemarle, Netherlands. The exact composition of the
catalyst was not undisclosed to us under the terms and
conditions of an agreement between Albemarle and the
Indian Institute of Technology Guwahati. However, BET
and TPR study of the catalyst was done in order to
understand the surface and reduction property of the
catalyst, which is a prerequisite for any heterogeneous
catalyst.
2.2 Catalyst characterization
Temperature programmed reduction (TPR) was
performed using Micromeritics TPD/TPR 2720 system to
determine reducibility of cobalt catalyst. The catalyst (0.3 g)
was reduced in a flow of 5% H
2
/Ar (30 cm
3
/min) with a
ramp rate of 10
o
C/min to 1073 K. A thermal conductivity
detector (TCD) was used to measure H
2
consumption. The
detector output was calibrated based upon 100% reducibility
of Ag
2
O powder, and a H
2
O trap was used to remove H
2
O
produced during the reduction. The degree of reduction of
the catalyst was calculated by using data of TPR
measurements as follows [16]:
2
2
H consumption below 723 K
% reduction = 100
Total H consumption in TPR experiment
(1)
2.3 Experimental setup and protocol
Experimental Setup: A schematic diagram of the reactor
and its accessories is shown in Figure 1. A xed bed tubular
reactor (MOC: SS 316, thickness: 2 mm, ID: 6.35 mm or
in.) was used for all reactions. The reactor had two gas
supply lines of 6.35 mm ( in.) ID, each for purge gas
(nitrogen) and synthesis gas (mixture of CO and H
2
). The
synthesis gas supply line was wrapped with thermal tapes so
as to preheat the feed gas to 423 K. The reactor effluent line
(ID 6.35 mm) was also heated in a similar manner to 373 K
to avoid any condensation of the product in the line prior to
backpressure regulator valve. Thereafter, no thermal
wrapping was given to the product lines, thus maintaining
the temperature in them at ambient condition, where higher
hydrocarbon products can condense. The liquid product was
collected in a trap (gasliquid separator) kept at ambient
temperature. Cold water (at 298 K) was circulated in a
jacket around the trap so as to achieve maximum possible
condensation of the hydrocarbon products. The lighter
hydrocarbons and unreacted reactant gases are removed
from the trap and purged through a gas flow meter. The ow
rate of feed synthesis gas was controlled using mass flow
controller (Make: Eureka Ml Flow, Range: 0500 cm
3
/min).
Catalyst (1 g) was placed inside the reactor enclosed
between glass wool plugs. The catalyst was reduced in situ
at 673 K for 12 h under a stream of 50% H
2
and N
2
mixture
(0.1 MPa pressure, 15 ml/min). After 12 h of reduction, the
temperature was reduced to 473 K under the constant flow
of hydrogen. Synthesis gas was introduced at a particular
space velocity (GHSV), and pressure was gradually
increased and set at the desired value. The temperature
inside the reactor was monitored using two thermocouples,
one at the catalyst bed and other at the surface of reactor.
These temperatures were maintained constant by circulating
International Journal of Scientific Engineering and Technology (ISSN : 2277-1581)
Volume No.2, Issue No.1, pp : 31-39 1 Jan. 2013
IJSET@2013 Page 33
cooling water through a thermal jacket (in the form of in.
copper tubing) wrapped around the reactor. Experiments
were performed for five different combinations of reaction
pressure and GHSV. The conditions for each of the five
experiments are listed in Table 1 and designated as protocols
1 to 5.
Materials: All gases used were procured from M/s Vadilal
Gas Ltd. (Gujarat, India, purity: 99.999%). The gas
cylinders containing reaction mixture (H
2
/CO = 67%/33%
v/v) were connected to gas supply lines through twostage
high pressure regulator. The analyses of the gaseous
products in the reactor effluent were performed by an online
gas chromatograph (Make: Thermo Fischer Ceres 800 Plus
system) using PorapackQ column provided with a thermal
conductivity detector (TCD). The liquids products were also
analyzed by the same GC provided with a flame ionization
detector (FID) using BP10.5 column. For characterization
of gas chromatograph, hydrocarbon standards in the carbon
number range of C5-C24 were procured from Sigma-
Aldrich.
Analysis:
1. The CO conversion (%) has been calculated
according to the following equation:
( ) ( )
( )
CO conversion = 100
CO CO
in out
CO
in
n n
n
(2)
2. The molar selectivity toward the individual
components on carbon basis has been calculated according
to the equation:
1
i
i n
i
j
y
m
y
=
=
(3)
y
i
= mole fraction of component i in reactor effluent
3. Results and Discussion
3.1 The reduction profile and the surface properties
of the catalyst
TPR profile of the cobalt oxide catalyst is depicted
in Figure 2. It can be observed that the reduction process of
the catalyst occurs in two distinct stages. The first peak near
573
K is ascribed to the transformation of Co
3
O
4
to CoO,
whereas the second peak near 743
K represents the
transformation of CoO to Co [17,18]. The proposed scheme
for the reduction of cobalt oxides might be written as
follows:
2
H
3 4 2
Co O 3 CoO + H O
(first stage at
573 K) (R.5)
2
3H
2
3 CoO 3 CO + 2 H O
(second stage between 720 and 770 K) (R.6)
Based on calculations according to equation 1, the
reducibility the cobalt catalyst was found to be 33%. For
BET (BrunauerEmmettTeller) surface area determination
of the catalyst, 0.3 g catalyst sample was degassed at 423 K
for 2 h, and then heated at 10 K/min to 423 K and held for 2
h before analysis. The BET surface area of the catalyst was
determined as 190 m
2
/g by N
2
physisorption at 77 K (Make:
Beckcoulter). Despite having large surface area, the lower
reducibility of the catalyst could be due to difficulty in the
reduction of CoO to Co.
3.2 Effect of operation conditions on product profile
To identify and optimize the process condition of
FischerTropsch reaction, effects of pressure and the gas
flow rate (GHSV) on selectivity of the products were
investigated. The results of reactions conducted at different
conditions were adjudged and compared with yardsticks of
overall CO conversion and selectivity for hydrocarbons in
the desired carbon number range.
Effect of total reaction pressure: FTS reaction is essentially
a polymerization type reaction, in which the CH
2
units
formed from adsorbed CO and H
2
reactant gases add up to
form hydrocarbons of different chain length. Thus, as
evident from reactions R.1, the numbers of product moles in
FTS reaction are less than reactant moles resulting in
positive pressuredependence, as per Le Chateliers
principle. An obvious consequence of these effects is that
increase of reaction pressure also promotes the FTS
reaction, leading to an increase in the CO conversion. With
increasing pressure, longer chain hydrocarbons are likely to
form that could possibly condense and saturate the catalyst
pores by liquid product. To assess effect of reaction pressure
on product selectivity, we have carried out a series of
reactions at reaction pressure range of 1 to 3 MPa using
synthesis gas with molar ratio of H
2
/CO as 2:1 and
temperature of 493 10 K. At each reaction pressure, the
reduced catalyst was tested for 10 h. The results are
presented in the Fig. 3 for reaction pressures of 1, 1.5 and 2
MPa, respectively, at GHSV of 1800 h
1
. It is clearly
observed that with the increase in the total pressure, there is
an increase in the carbon number of the product, i.e. longer
chain hydrocarbons that condense at ambient temperature.
This also shows increasing trend in liquefaction of synthesis
gas with rising reaction pressure. At a reaction pressure of 1
MPa, the product molar selectivity is toward C
12
. An
increase in the pressure to 1.5 MPa shifts the product molar
selectivity towards the longer hydrocarbons (in the range
C
16
C
18
), which is evident from the product profiles
depicted in Fig. 3. Further increase in the pressure to 2 MPa
International Journal of Scientific Engineering and Technology (ISSN : 2277-1581)
Volume No.2, Issue No.1, pp : 31-39 1 Jan. 2013
IJSET@2013 Page 34
shows interesting trend in that highest selectivity is seen for
C
24
with moderate selectivities in the range of C
14
C
16
;
while selectivity in the intermediate carbon number range of
C
18
C
22
is, however, lower than both C
14
C
16
and C
24
. At
constant GHSV of 1800 h
1
, the rise in reaction pressure
from 1 to 2 MPa increased CO conversion by approximately
30%.
An explanation for the observed results can be
given as follows: As per the alkyl mechanism Figure 4, the
FT reaction is essentially a chain polymerization type
reaction which is initiated by chemisorption of the carbon
monoxide on the catalyst surface. The carbon monoxide is
believed to have dissociative adsorption on the catalyst
surface. Adsorption of carbon monoxide on the cobalt
catalyst surface has been reported to be on the ontop site of
Co(0001) [19], which is mainly due to interaction between
filled 5orbital and the double degenerated 2 orbital
[20] and the center of the metal dband as shown in Figure
5. Charge is donated from carbon to the metal by the 5
bond and stabilized by back-donation from the metal d-
orbitals to the 2* antibonding molecular orbitals,
weakening the CO bond. Dissociation of chemisorbed CO
results in formation of surface carbon and surface oxygen.
The surface oxygen is removed either by adsorbed hydrogen
(with formation of water) or adsorbed CO (with formation
of CO
2
) [21]. The surface carbon, on the other hand, can
undergo sequential hydrogenation to yield intermediate
methylidyne (CH), methylene (CH
2
) and methyl (CH
3
)
species. In the overall FTS reaction, the CH
3
surface species
plays role of initiator, while CH
2
surface species plays role
of monomer. These species react (either by |hydrogen
abstraction or hydrogen addition) to form long chain
hydrocarbon products, either oolefin or nparaffins. A
higher pressure causes greater adsorption of carbon
monoxide on the catalyst surface, causing greater formation
of surface species (CH, CH
2
and CH
3
), and greater extent of
reaction among them to form hydrocarbon product with
higher carbon number (or a longer chain hydrocarbon).
Hence with increase in the pressure the product selectivity
increases towards the longer hydrocarbons.
Effect of synthesis gas flow rate: Gas flow rate in the
reactions has been measured in terms of the gas hour space
velocity (GHSV). The GHSV influences the residence time
of the gas with the catalyst. To understand the possible
effect of GHSV on the selectivity of the products in terms of
carbon number, we have done experiments at 7 different
GHSVs, viz. 24000, 19200, 14400, 9600, 4800, 3600 and
1800 h
1
. In the GHSV range of 24000 h
1
to 4800 h
1
, no
liquid hydrocarbon products were obtained. This essentially
means that no FTS reaction occurred for these gas hour
space velocities. Even if any FTS reaction occurred, the
yields of gaseous products would have been in too low
concentrations to be detected by the gas chromatograph.
Further reduction of GHSV in the range of 36001800 h
1
,
significant fraction of liquid hydrocarbons (C
5+
) were
observed in the product. The effect of GHSV on the
selectivity of products can be deduced from product profile
depicted in Fig. 3. It could be inferred that an increase in
reaction pressure from 2 to 3 MPa at a constant GHSV of
3600 h
1
, the carbon number range of the product increases
from C
6
C
10
to C
10
C
14
. Similarly, at a constant pressure of
2 MPa, reduction in the GHSV from 3600 to 1800 h
1
results in significant rise in the carbon number of product
from C
6
C
10
to C
14
C
24
. These results show that the product
distribution of the FT reaction in terms of carbon number is
dominated by combined effect of reaction pressure and
GHSV.
An explanation to these results can be given as
follows: With higher residence time, there is longer
interaction between catalyst particles and the gas molecules
leading to higher adsorption of the reactant molecules. This
phenomenon also leads to prolongation of the residence time
of reactants and reaction intermediates on the catalyst
surface causing further hydrogenation and oligomerization
of shortchain hydrocarbons resulting in long chain products
(C
5+
) that condense at ambient conditions. Similar
observations have been made in previous studies by Dictor
and Bell [7], Donnely and Satterfield [8], Bukur et al. [12],
Kuipers et al. [13] and Wang et al. [22] have also reported
decrease in the olefin to paraffin ratio of the product with
decreasing space velocity, indicating readsorption of
olefins and further hydrogenation to paraffins.
4. Conclusion
Present study using commercial cobalt catalyst has
attempted to optimize two important process parameters of
FT synthesis, viz. the total reaction pressure and the
residence time. Interesting trends in the product profile of
the FTS reaction were observed from a set of experiments
with permutation-combination of above-mentioned
parameters. Main goal of our experiments was to deduce the
influence of GHSV and reaction pressure on product profile,
and optimize the reaction conditions for products with
carbon number C
12
or higher, which is the typical range of
hydrocarbons in diesel. Among the range of reaction
pressures and GHSV attempted in this work, reaction
pressure range of 12 MPa with GHSV 1800 h
1
as in
protocol 1, 2 and 3 gives maximum selectivity towards C
14
C
24
. Moving towards a higher reaction pressure of 3 MPa,
with a faster GHSV of 3600 h
-1
or higher as in protocol 5,
the products selectivity shifts in the desired range C
10
C
14
although CO conversion is to some extent lesser than
International Journal of Scientific Engineering and Technology (ISSN : 2277-1581)
Volume No.2, Issue No.1, pp : 31-39 1 Jan. 2013
IJSET@2013 Page 35
protocols 1, 2 and 3. Thus, the optimum set of parameters
for the FT process with commercial catalyst used in this
work is found to be pressure of 3 MPa and GHSV of 3600 h
-
1
. We believe that these results will be useful for design and
optimization of a large scale FT using the same catalyst.
Acknowledgement
This research was supported by Ministry of New and
Renewable Energy, Government of India under grant no.
19/20/2007-R&D/BE dated 06.08.2008.
References
i. Keshav TR, Basu S (2007) Gas-to-liquid
technologies: India's perspective. Fuel Processing
Technology 88:93-500
ii. Hsin Chen W, Chung Chen J, Tsai CD, Jiang TL
(2007) Transient gasification and syngas formation
from coal particles in a fixed-bed reactor.
International journal of energy research. 31(9):895-
911
iii. Tremel A, Gaderer M, Spliethoff H (2012) Small-
scale production of synthetic natural gas by
allothermal biomass gasification. International
Journal of Energy Research. doi:10.1002/er.2933.
iv. Hook M, Aleklett K (2010) A review on coal-to-
liquid fuels and its coal consumption. International
Journal of Energy Research 34(10):848-864
v. Buragohain B, Mahanta P, Moholkar VS (2012)
Performance correlations for biomass gasifiers
using semi-equilibrium non-stoichiometric
thermodynamic models. International Journal of
Energy Research 36(5):590-618
vi. Wei Guo K (2012) Green nanotechnology of trends
in future energy: a review. International Journal of
Energy Research 36(1):1-17
vii. Ranjan A, Moholkar VS (2011) Biobutanol:
science, engineering, and economics. International
Journal of Energy Research 36(3):277-323
viii. Dry ME (2002) The FischerTropsch process:
19502000. Catalysis Today 71: 227-241
ix. Khodakov AY, Chu W, Fongarland P (2007)
Advances in the development of novel cobalt
FischerTropsch catalysts for synthesis of long-
chain hydrocarbons and clean fuels. Chemical
Reviews 107, 1692-1744
x. Trepanier M, Dalai AK, Abatzoglou N (2010)
Synthesis of CNT-supported cobalt nanoparticle
catalysts using a microemulsion technique: Role of
nanoparticle size on reducibility, activity and
selectivity in FischerTropsch reactions. Applied
Catalysis A: General 374:7986
xi. Davis BH (2001) FischerTropsch synthesis:
current mechanism and futuristic needs. Fuel
Processing Technology 71:157-166
xii. Hammer H, Joisten M, Lngen S, Winkler D
(1994) New zeolites in fischer-tropsch synthesis.
International Journal of Energy Research
18(2):223-231
xiii. Baerns M, Guan N, Krting E, Lindner U,
Lohrengel M, Papp H (1994) Catalyst development
for selective conversion of syngas to mainly
aromatic hydrocarbons. International Journal of
Energy Research 18(2):197-204
xiv. Dictor RA, Bell AT (1986) FischerTropsch
synthesis over reduced and unreduced iron oxide
catalysts. Journal of Catalysis 97:121136
xv. Donnelly TJ and Satterfield CN (1989) Product
distributions of the FischerTropsch synthesis on
precipitated iron catalysts. Applied Catalysis A:
General 52:93114
xvi. Mirzaei AA, Babaei AB, Galavy M, Youssefi
A(2010) A silica supported FeCo bimetallic
catalyst prepared by the sol/gel technique:
Operating conditions, catalytic properties and
characterization. Fuel Processing Technology
91:335347
xvii. Anderson RB (1956) Catalysts for the Fischer
Tropsch Synthesis (Vol. 4). New York,Van
Nostrand Reinhold
xviii. Dry ME (1981) The FischerTropsch Synthesis. In:
Anderson JR, Boudart M (eds. )Catalysis Science
and Technology, vol. 1, 160255. New York:
SpringerVerlag
xix. Bukur DB, Patel S A, Lang X (1990) Fixed bed
and slurry reactor studies of FischerTropsch
synthesis on precipitated iron catalyst. Applied
Catalysis A: General 61:329349
xx. Kuipers EW, Scheper C, Wilson JH, Oosterbeek H
(1996) NonASF product distributions due to
secondary reactions during FischerTropsch
synthesis. Journal of Catalysis 158:288300
xxi. Iglesia E, Reyes SC, Madon RJ (1991) Transport
enhanced oolefin readsorption pathways Ru
catalyzed hydrocarbon synthesis. Journal of
Catalysis 129:238256
xxii. Yan Z, Wang Z, Bukur DB, Goodman DW (2009)
FischerTropsch synthesis on a model Co/SiO2
catalyst. Journal of Catalysis 268:196200
xxiii. Kang SH (2010) ZSM-5 supported iron catalysts
for FischerTropsch production of light olefin. Fuel
Processing Technology 91: 399403
International Journal of Scientific Engineering and Technology (ISSN : 2277-1581)
Volume No.2, Issue No.1, pp : 31-39 1 Jan. 2013
IJSET@2013 Page 36
xxiv. Saib A M, Claeys M, van Stenn E (2002) Silica
supported cobalt FischerTropsch catalysts: Effect
of pore diameter of support. Catalysis Today
71:395402
xxv. Song D, Li J (2006) Effect of catalyst pore size on
the catalytic performance of silica supported cobalt
FischerTropsch catalysts. Journal of Molecular
Catalysis A: Chemical 247: 206212
xxvi. Lahtinen J, Vaari J, Kauraala K, Soares EA, van
Hove MA (2000) LEED investigations on
Co(0001): the (33)R30-CO overlayer. Surface
Science 448: 269278
xxvii. van Daelen, M A, Li YS, Newsam JM, van Santen
RA (1994) Transition states for NO and CO
dissociation on Cu(100) and Cu(111) surfaces.
Chemical Physics Letters 226: 100-105
xxviii. Ribeiro MC, Jacobs G, Davis BH, Cronauer DC,
Kropf AJ, Marshall CL (2010) FischerTropsch
synthesis: An in-situ TPR-EXAFS / XANES
investigation of the influence of group I alkali
promoters on the local atomic and electronic
structure of carburized iron/silica catalysts. Journal
of Physical Chemistry C 114:7895-7903
xxix. Pour A N, Shahri SMK, Zamani Yand Zamanian A
(2010) Promoter effect on the CO
2
-H
2
O formation
during Fischer-Tropsch synthesis on iron-based
catalysts. Journal of Natural Gas Chemistry
19:193-197
xxx. Wang ZX, Dong T, Yuan LX, Kan T, Zhu XF,
Torimoto Y, Sadakata M, Li QX (2007)
Characteristics of bio-oil-syngas and its utilization
in Fischer-Tropsch synthesis. Energy & Fuels 21:
2421-2432
xxxi. Pour AN, Housaindokht MR, Tayyari SF, Zarkesh
J (2010) Effect of nano-particle size on product
distribution and kinetic parameters of Fe/Cu/La
catalyst in Fischer-Tropsch synthesis. Journal of
Natural Gas Chemistry 19:107116
International Journal of Scientific Engineering and Technology (ISSN : 2277-1581)
Volume No.2, Issue No.1, pp : 31-39 1 Jan. 2013
IJSET@2013 Page 37
Table 1: FTS reaction data at different Pressure and GHSV
Protocol Pressure
(MPa)
CO Conversion % GHSV(h
1
) Molar
Selectivity
Protocol 1 1 28 1800 C
12
C
16
Protocol 2 1.5 34 1800 C
16
C
20
Protocol 3 2 37 1800 C
14
C
24
Protocol 4 2 27 3600 C
6
C
10
Protocol 5 3 27 3600 C
10
C
14
Figure 1: Experimental setup
International Journal of Scientific Engineering and Technology (ISSN : 2277-1581)
Volume No.2, Issue No.1, pp : 31-39 1 Jan. 2013
IJSET@2013 Page 38
100 200 300 400 500 600 700 800
0.00
0.02
0.04
0.06
0.08
0.10
I
n
t
e
n
s
i
t
y
i
n
a
.
u
.
Temperature (
0
C)
Figure 2: TPR profile of Cobalt catalyst
Figure 3. Product profile and distribution in different experimental protocols
International Journal of Scientific Engineering and Technology (ISSN : 2277-1581)
Volume No.2, Issue No.1, pp : 31-39 1 Jan. 2013
IJSET@2013 Page 39
CO
C
+2H
-H
2
O
+H
CH
CH
2 +H
Initiation:
Monomer
Surface
Carbon
Surface
methylidene
R CH
2 CH
2
R
Chain initiator
CH
2
CH
3
R
Propagation:
Termination/desorption
CH
3
+H
H
2
C CHR
H
3
C CH
2
R
HOH
2
C CH
2
R
o-olefine
n-paraffin
n-alcohol
Surface
methylene
'Alkyl' mechanism
Figure 4. Alkyl mechanism
M
C
O
Valence d-orbitals
2 * t
5o
Figure 5. Molecular orbital diagram of CO to a metal site according to the Blyholder model. (redrawn with modifcations from ref.
[24]).