A macrocyclic Pd(II)-Ni(II) complex in Heck and Suzuki reactions

Article abstract of DOI:10.1016/j.ica.2014.11.008

A heterodinuclear Pd(II)-Ni(II) macrocyclic Schiff base complex was applied as a catalyst for Heck and Suzuki cross-coupling reactions. In reaction of 4-bromoanisole with styrene, 4-methoxy-trans-stilbene was formed as the main product with a yield of up to 87%. In the Suzuki coupling of 4-bromoanisole with PhB(OH)₂in a 2-propanol/water solution, up to 96% of the product was obtained. Analyses of the post-reaction mixtures confirmed the high stability of the macrocyclic Pd(II)-Ni(II) complex under the reaction conditions. The partial decomposition of the complex started at the Ni(II) site.

Full text of DOI:10.1016/j.ica.2014.11.008

Inorganica Chimica Acta xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Inorganica Chimica Acta
journal homepage: www.elsevier.com/locate/ica
A macrocyclic Pd(II)–Ni(II) complex in Heck and Suzuki reactions
⇑
´
Ewa Mieczynska, Jerzy Lisowski, Anna M. Trzeciak
Faculty of Chemistry, University of Wrocław, 14 F. Joliot-Curie St., 50-383 Wrocław, Poland
a r t i c l e i n f o
a b s t r a c t
Article history:
A heterodinuclear Pd(II)–Ni(II) macrocyclic Schiff base complex was applied as a catalyst for Heck and
Suzuki cross-coupling reactions. In reaction of 4-bromoanisole with styrene, 4-methoxy-trans-stilbene
was formed as the main product with a yield of up to 87%. In the Suzuki coupling of 4-bromoanisole with
PhB(OH)2 in a 2-propanol/water solution, up to 96% of the product was obtained. Analyses of the post-
reaction mixtures confirmed the high stability of the macrocyclic Pd(II)–Ni(II) complex under the reaction
conditions. The partial decomposition of the complex started at the Ni(II) site.
Received 26 August 2014
Received in revised form 14 November 2014
Accepted 19 November 2014
Available online xxxx
Keywords:
Macrocyclic Pd(II)–Ni(II) complex
Heck reaction
Ó 2014 Published by Elsevier B.V.
Suzuki reaction
ESI-MS
1. Introduction
Another concept was presented by Köhler and co-workers
[33,34]. It was based on the slow and controlled release of palla-
Palladium-catalyzed C–C bond forming reactions have become
a very powerful tool of organic synthesis [1–11]. Organic products
formed in these reactions, such as arylated olefins, unsymmetrical
biaryls, arylalkynes, or conjugated enynes, are industrially impor-
tant substrates for the pharmaceutical and agrochemical indus-
tries. They are also used as natural products, molecular organic
materials and components for optical device production [1–11].
An increasing market for the products of C–C bond forming
reactions motivates searching for new, more efficient catalysts
for such processes, preferably operating under mild conditions.
Different classes of palladium species have been tested in the Heck
and the Suzuki cross-coupling reactions, such as palladacycles [12–
17] and complexes with N-heterocyclic carbenes [18–23]. Good
results have also been obtained for immobilized palladium cata-
lysts [24–27] as well as for Pd(0) nanoparticles [28–30].
The main problem associated with the use of palladium cata-
lysts is the low stability of Pd(0) forms, which easily undergo
agglomeration in the absence of stabilizing agents [31,32]. At the
beginning of the catalytic cycle, Pd(0) is needed for the activation
of the substrate, aryl halide, according to the oxidative addition
pathway. However, at the end of the catalytic process, non-stabi-
lized Pd(0) can agglomerate and finally form inactive palladium
black. To overcome this unfavourable process, strong ligands,
polymers, or inorganic oxides are used.
dium from the precursor to the solution. As a result, the concentra-
tion of palladium in the solution remained very low, and
agglomeration was limited, which allowed a high catalytic activity
even for deactivated substrates [33,34].
While palladium catalysts are first choice in C–C bond forming
reactions, there is growing interest in catalysts based on the much
cheaper and less toxic nickel complexes. In some cases the nickel
based catalysts even surpass palladium catalysts in cross-coupling
reactions [35,36]. In a recent communication a very interesting
effect of cooperative catalysis by palladium–nickel binary nano-
cluster in Suzuki–Miyaura reaction has been reported [37]. In this
case the mixture of Pd and Ni compounds turned out to be much
more effective catalytic system than the catalytic systems based
on the Pd alone or Ni alone.
Inspired by the results obtained in Heck and Suzuki reactions
with the application of a Pd(II)–Pd(II) macrocyclic bimetallic com-
plex [Pd2L](ClO4)2, we decided to study the catalytic activity of a
Ni(II)–Pd(II) analogue [NiPdL](ClO4)2ꢀ3MeCN (Scheme 1) [38,39].
In particular, it was interesting to learn whether the presence of
Ni(II) would influence palladium activity via formation of mixed
Pd–Ni nanoparticles/clusters originating from the same molecular
precursor.
2. Experimental
Macrocyclic complexes [NiPdL](ClO4)2ꢀ3MeCN, [Ni(H2L)](ClO4)2ꢀ
H2O, [Ni2L]Cl2ꢀ2H2O were obtained according to the literature
[38,39].
⇑
Corresponding author. Tel.: +48 71 3757253.
E-mail address: anna.trzeciak@chem.uni.wroc.pl (A.M. Trzeciak).
http://dx.doi.org/10.1016/j.ica.2014.11.008
0020-1693/Ó 2014 Published by Elsevier B.V.
Please cite this article in press as: E. Mieczyn´ ska et al., Inorg. Chim. Acta (2014), http://dx.doi.org/10.1016/j.ica.2014.11.008

´
2
E. Mieczynska et al. / Inorganica Chimica Acta xxx (2014) xxx–xxx
2.1. Heck reaction
3. Results and discussion
The Heck reaction was carried out under an N2 atmosphere
using a standard Schlenk technique. The reagents were introduced
to the Schlenk tube (50 mL) in the following order: catalyst
3.1. Heck reaction
The model Heck reaction of styrene with 4-bromoanisole was
performed according to Scheme 2.
(0.0375–0.225 mol%),
4-bromoanisole
(3 mmol),
styrene
(1.5 mmol), base (Ca(OH)2 or NaHCO3 0.9 mmol), solvent NMP or
DMF (5 mL). The reaction was carried out at 160 °C for 6–26 h.
Afterwards, the reaction mixture was quenched with H2O
(10 mL), and the organic products were separated by extraction
with diethyl ether (3 times, 15, 5, and 5 mL). The products were
analyzed by GC-FID.
The highest conversion of styrene, 95–99%, was obtained after
26 h at 160 °C using 0.0375–0.075 mol% of the catalyst. With a lar-
ger amount of the catalyst, 0.225 mol%, already after 6 h, 95% of
styrene was converted to the products (Table 1).
A good result, 72% conversion, was also achieved in DMF. In this
solvent, the best selectivity was noted with the 1:2 ratio amount-
ing to 71. In all other cases, two products, 4-methoxy-trans-stil-
bene (1) and 1-(4-methoxyphenyl)-1-phenylethene (2), were
formed with a ratio of ca. 10. Similarly, the formation of two prod-
ucts in the Heck reaction of styrene was reported for a macrocyclic
palladium precursor [33,34] and phospha-palladacycle [40].
When [NBu4]Br was added to the reaction in DMF, conversion
decreased to 44% and selectivity also dropped. The inhibiting effect
of ammonium salt was also observed in the Heck reaction of sty-
rene and chloroanisole catalyzed by a macrocyclic Pd(II) complex
[Pd2L](ClO4)2 [33,34]. It was explained by the formation of palla-
dium black in reaction with tributylamine originated from the
decomposition of the salt at 160 °C. The presence of tributylamine
in the reaction mixture was confirmed in our system by GC–MS;
however, we did not obtain confirmation of the formation of palla-
dium black or Pd(0) nanoparticles by TEM.
2.2. Suzuki–Miyaura reaction
The Suzuki–Miyaura reactions were carried out in a 50 mL
Schlenk tube in an air atmosphere. The substrates were weighed
and placed directly in the Schlenk tube: catalyst (0.55–2.1 mol%),
phenylboronic acid (1.5 mmol, 0.183 g), base (KOH or K3PO4
2 mmol), aryl halide (4-bromoanizole or 4-iodoanisole, 1 mmol).
Next, 5 mL of the solvent was added. The reaction mixture was
stirred at 60 or 130 °C for 25.5 or 46.5 h. Afterwards, the reactor
was cooled down, the reaction mixture was quenched with H2O
(3 mL), and the organic products were separated by extraction
with diethyl ether (3 times, 4, 3 and 3 mL). The products were
analyzed by GC-FID.
For recycling experiments the solution remaining after extrac-
tion of organic products was condensed in vacuo and used for
the next experiment with new portion of substrates.
In order to compare activity of [NiPdL](ClO4)2 and [Pd2L](ClO4)2
complexes, one reaction was performed at 140 °C with 0.05 mol%
of catalyst. Using mixed complex we obtained 76% of product 1,
whereas 71% was formed with [Pd2L](ClO4)2 [34]. Considering a
lower content of palladium in the mixed precursor, the obtained
results are quite satisfactory.
2.3. Instruments
When more reactive butyl acrylate was used instead of styrene,
only one product, substituted trans-stilbene, was obtained
(Scheme 3). In this case formation of more stable product is
favoured.
Attempts to perform the Heck coupling with 4-chloroanisole
were unsuccessful and after 26 h the product was not formed.
GC analyses (GC-FID) were performed on Hewlett Packard
8454A instrument. Products were identified by GC–MS using
Hewlett Packard 5890 II instrument with capillary column
ELITE-5MS and stationary phase 5% diphenylpolysiloxane, 95%
dimethylpolysiloxane.
MS spectra were performed on a Bruker Daltonics micrOTOF-Q.
The ions were generated from electrospray ionization source. The
electrospray flow rate was 10 l
l/minꢁ1, which was maintained
3.2. Suzuki reaction
by a syringe pump; spray was directed into a heated glass capillary
at a temperature 200 °C, and a high voltage of 4500 V between the
endplate and the spray needle. Nebulizer vacuum 0.4 Bar; dry gas
flow 4.0 l/min;
The Suzuki reaction was performed according to Scheme 3,
using the palladium–nickel complex [NiPdL](ClO4)2 and the nickel
analogues: [NiH2L](ClO4)2 and [Ni2L1]Cl2, as catalysts. However,
only Pd–Ni was active in the studied reaction.
1H NMR spectra were measured on Bruker 500 MHz
spectrometer.
The optimization of the reaction parameters resulted in the
selection of the best solvents, namely 2-propanol/H2O (1:1) and
ethylene glycol, making it possible to obtain 96% or 80% of the
cross-coupling product (Table 2). The concentration of the catalyst
has an important influence on the reaction course and 1 mol%
appeared the best. Using a smaller amount of the catalyst,
0.55 mol%, 78% of the product was obtained, whereas an increase
of the catalyst amount to 2.1 mol% resulted in a decrease of the
yield to 34%. The modification of the last reaction by the addition
of 1 mmol TBAB made it possible to increase the yield to 71%. Inter-
estingly, in the Suzuki reaction, the presence of [Bu4N]Br favour-
ably influenced the reaction yield, while the same salt inhibited
the Heck reaction (Scheme 4).
2+
N
N
O
O
N
N
Pd
Ni
Attempts to reuse the Pd–Ni catalyst were not very promising,
and in the best case 12% of product was found in the second run
(Table 2).
Two experiments performed under conditions similar to these
used earlier [34] for testing [Pd2L](ClO4)2 complex gave 25–30%
of product after 6 h.
Scheme 1. Macrocyclic Pd(II)–Ni(II) cation ([NiPdL]2+) of the [PdNiL](ClO4)2ꢀ3MeCN
complex.
Please cite this article in press as: E. Mieczyn´ ska et al., Inorg. Chim. Acta (2014), http://dx.doi.org/10.1016/j.ica.2014.11.008

´
3
E. Mieczynska et al. / Inorganica Chimica Acta xxx (2014) xxx–xxx
OMe
OMe
+
Br
+
MeO
2
1
Scheme 2. Heck reaction of styrene with 4-bromoanisole.
Table 1
Results of the Heck reaction of 4-bromoanisole with styrene catalyzed by [NiPdL](ClO4)2.CH3CN.
Time (h)
Solvent
Temp. (°C)
Catalyst amount (mol%)
Conversion (%)
Yield of 1 (%)
Yield of 2 (%)
26
26
6
NMP
NMP
NMP
DMF
DMF
DMF
NMP
NMP
160
160
160
160
160
160
160
140
0.075
0.0375
0.225
0.075
0.075
0.05
99
95
95
72
44
43
90
85
87
81
84
71
39
38
81
76
8
11
8
1
5
5
9
9
6a
6a,b
6c
6c
0.05
0.05
6c
Reaction conditions: styrene (1.5 mmol), 4-bromoanisole (3 mmol), Ca(OH)2 (0.9 mmol), solvent (5 mL).
a
Base: NaHCO3 (0.9 mmol).
1 mmol [Bu4N]Br was added.
Base: CH3COONa.
b
c
MeO
Br
[NiPdL]2+
+
COOBu
MeO
COOBu
38%
Scheme 3. Heck coupling of 4-bromoanisole with butyl acrylate.
Table 2
Results of the Suzuki reaction catalyzed by [NiPdL](ClO4)2.CH3CN.
Substrate
Solvent
Time (h)
Temp. (°C)
Catalyst (mol%)
Conversion (%)
Yield (%)
4-Bromoanisole
4-Iodoanisole
Dioxane
Dioxane
25.5
46.5
25.5
25.5
25.5
25.5
25.5
25.5
25.5
6
130
130
60
60
60
60
60
60
60
1
2.1
1
16
96
80
78
7
96
12
34
71
25
30
16
96
80
78
7
96
12
34
71
25
30
4-Bromoanisole
4-Bromoanisole
4-Bromoanisole
4-Bromoanisole
4-Bromoanisole
4-Bromoanisole
4-Bromoanisole
4-Bromoanisole
4-Bromoanisole
Ethylene glycol
2-Propanol/watera
2-Propanol/waterb
2-Propanol/watera
2-Propanol/waterb
2-Propanol/watera
2-Propanol/watera,c
2-Propanol/waterd
2-Propanol/waterd
0.55
1
2.1
2.1
0.5
1
100
100
6
Reaction conditions: 4-bromoanisole (1 mmol), PhB(OH)2 (1.5 mmol), K3PO4 (2 mmol), 130 °C.
a
KOH (2 mmol).
Recycling (second run with the same catalyst).
[Bu4N]Br (1 mmol) was added.
Na2CO3 (3.2 mmol), 4-bromoanisole (3 mmol), PhB(OH)2 (3.2 mmol).
b
c
d
Br
B(OH)2
+
OMe
MeO
Scheme 4. Suzuki coupling of 4-bromoanisole with phenylboronic acid.
Please cite this article in press as: E. Mieczyn´ ska et al., Inorg. Chim. Acta (2014), http://dx.doi.org/10.1016/j.ica.2014.11.008

´
E. Mieczynska et al. / Inorganica Chimica Acta xxx (2014) xxx–xxx
4
3.3. Mechanistic studies
In addition to the signal of intact [NiPdL]2+, the mass spectra of
the hexane post-reaction extract indicated a weak signal at m/z
equal to 509, corresponding to the [PdLH]+ complex, where the
Ni(II) ion was removed and replaced by a proton. Additionally,
there was a weak signal at m/z equal to 493, corresponding to
the [NaPdL0]+ complex of an open ligand L0, resulting from the
removal of the Ni(II) ion, the hydrolysis of the propylenediamine
sidearm, and the binding of the Na(I) cation originating from
sodium present in the base (Scheme 5). On the other hand we
did not observed in ESI MS spectra signals of analogous [NiLH]+
and [NaNiL0]+ decomposition products. Thus the two metal ions
in the [NiPdL]2+ complex are released at different rates; apparently
the Ni(II) ion is bound less strongly and is removed first. These
results indicate that the catalytic effect is similar to that proposed
Similarly as was observed for the dipalladium complex [Pd2L]2+
,
a substantial amount of [NiPdL]2+ survived the catalytic reactions.
This fact is confirmed by the 1H NMR spectrum of the residue
obtained by the evaporation of the water phase from the reaction
workup. Although the diamagnetic region is obscured by the sig-
nals of residual water and organic products, the characteristic sig-
nals of the paramagnetic complex [NiPdL]2+ at 24.1, 11.9, and
10.0 ppm were observed. The presence of intact [NiPdL]2+, left after
the catalytic reaction, was also confirmed by the UV–Vis spectrum,
which presented a band at 375 nm, and mass spectra that indi-
cated a 284 signal at m/z, corresponding to [NiPdL]2+, with a char-
acteristic isotope pattern (Fig. 1).
Intens.
C24H26N4NiO2Pd,
M
,566.04
284.0
2000
283.0
1500
282.5
285.0
1000
283.5
284.5
282.0
500
285.5
286.0
286.5
287.0
0
4
x10
+MS, 0.0-0.5min #(2-34)
2.0
1.5
1.0
0.5
0.0
284.0
283.0
285.0
282.5
283.5
284.5
282.0
286.0
286
285.5
286.5
287.0
282
283
284
285
287
m/z
Fig. 1. Simulated (top) and experimental (bottom) isotope pattern of [NiPdL]2+ (m/z = 284) found in the sample after Suzuki reaction.
+
+
2+
O
O
O
Na
O
N
N
N
N
O
O
N
N
N
N
O
O
N
N
Pd
Pd
Pd
Ni
H
Scheme 5. Cations [NiPdL]2+, [PdLH]+ and [NaPdL0]+ identified in ESI-MS spectra of the post-reaction mixture.
Please cite this article in press as: E. Mieczyn´ ska et al., Inorg. Chim. Acta (2014), http://dx.doi.org/10.1016/j.ica.2014.11.008

´
5
E. Mieczynska et al. / Inorganica Chimica Acta xxx (2014) xxx–xxx
previously for the dipalladium analogue. It is based on the high sta-
References
bility of the applied Pd(II) catalyst, which very slowly releases min-
ute amounts of Pd to the reaction medium. This slow release
prevents the formation of Pd clusters and palladium black. Indeed,
in our case, Pd(0) nanoparticles and palladium black were not
observed according to TEM results. On the other hand, the ESI
MS data mentioned above suggest that the partial decomposition
of the complex is actually started at the Ni(II) site.
[1] R.F. Heck, Acc. Chem. Res. 12 (1979) 146.
[2] I.P. Beletskaya, A.V. Cheprakov, Chem. Rev. 100 (2000) 3009.
[3] N.J. Whitcombe, K.K. Hii, S.E. Gibson, Tetrahedron 57 (2001). 7449-7449.
[4] A. Molnar (Ed.), Palladium Catalyzed Coupling Reactions, Wiley-VCH Verlag
GmbH & Co. KGaA, 2013.
[5] I. Tsuji, Palladium Reagents and Catalysts. New Perspectives for the 21st
Century, John Wiley & Sons Ltd., 2004.
[6] A. Behr, P. Neubert, Applied Homogeneous Catalysis, Wiley-VCH Verlag GmbH
& Co. KGaA, 2012.
[7] A. Steven, L.E. Overman, Angew. Chem., Int. Ed. 46 (2007) 5488.
[8] Z. Hyder, J. Ruan, J. Xiao, Chem. Eur. J. 14 (2008) 5555.
[9] A.M. Trzeciak, J.J. Ziółkowski, Coord. Chem. Rev. 249 (2005) 2308.
[10] A.M. Trzeciak, J.J. Ziółkowski, Coord. Chem. Rev. 251 (2007) 1281.
[11] R.B. Bedford, C.S.J. Cazin, D. Holder, Coord. Chem. Rev. 248 (2004) 2283.
[12] D. Pugh, A.A. Danopoulos, Coord. Chem. Rev. 211 (2007) 610.
[13] H. Weissman, D. Milstein, Chem. Comm. (1999) 1901.
[14] D.A. Alonso, C. Najera, M.C. Pacheco, Adv. Synth. Catal. 345 (2003) 1146.
[15] J. Dupont, C.S. Consorti, J. Spencer, Chem. Rev. 105 (2005) 2527.
[16] D.A. Albisson, R.B. Bedford, P.N. Scully, Tetrahedron Lett. 39 (1998) 9793.
[17] I. Błaszczyk, A. Gniewek, A.M. Trzeciak, J. Organomet. Chem. 710 (2012) 44.
[18] F. Glorius (Ed.), N-heterocyclic Carbenes in Transition Metal Catalysis,
Springer-Verlag, Berlin, Germany, 2007.
4. Conclusions
A Pd(II)–Ni(II) macrocyclic complex exhibited good catalytic
activity in the Heck coupling of 4-bromoanisole with styrene and
the Suzuki coupling of 4-bromoanisole with phenylboronic acid.
A high conversion of the substrates was obtained in the Heck reac-
tion using MNP or DMF as a solvent. The addition of [Bu4N]Br
caused an inhibiting effect leading to a yield decrease. The perfor-
mance of the Suzuki reaction in a 2-propanol/water mixture made
it possible to obtain 96% of the coupling product. Here, a positive
effect of [Bu4N]Br was observed.
[19] S.P. Nolan (Ed.), N-heterocyclic Carbenes in Synthesis, Wiley-VCH, New York,
2006.
[20] C.M. Crudden, D.P. Allen, Coord. Chem. Rev. 248 (2004) 2247.
[21] W.A. Herrmann, C. Köcher, Angew. Chem., Int. Ed. 36 (1997) 2162.
[22] N. Marion, S.P. Nolan, Acc. Chem. Res. 41 (11) (2008) 1440.
[23] N.M. Scott, S.P. Nolan, Eur. J. Inorg. Chem. (2005) 1815.
[24] A. Astruc, Inorg. Chem. 46 (2007) 121.
The catalytic activity of the Pd(II)–Ni(II) complex can be
explained by the slow release of palladium to the reaction solution
from the stable macrocyclic structure, similarly as was proposed
for the Pd(II)–Pd(II) analogue. Some amount of the complex under-
went decomposition, which started at the Ni(II) site. On the other
hand, the mixed Ni–Pd clusters were probably not formed in the
studied reactions and no increased activity was achieved due to
cooperative effect. The reason for this was the different rate of
release of nickel and palladium from the [NiPdL](ClO4)2ꢀ3MeCN
complex. The presence of the unchanged complex in the post-reac-
tion mixtures can be rationalized by its high stability or by recon-
struction of the structure at the end of the catalytic cycle.
Accordingly, the formation of Pd(0) nanoparticles or palladium
black was efficiently prevented.
[25] T. Borkowski, W. Zawartka, P. Pospiech, U. Mizerska, A.M. Trzeciak, M. Cypryk,
W. Tylus, J. Catal. 282 (2011) 270.
[26] E. Mieczyn´ ska, T. Borkowski, M. Cypryk, P. Pospiech, A.M. Trzeciak, Appl. Catal.
A: General 470 (2014) 24.
[27] A. Gniewek, J.J. Ziółkowski, A.M. Trzeciak, M. Zawadzki, H. Grabowska, J.
Wrzyszcz, J. Catal. 254 (2008) 121.
[28] A. Balanta, C. Godard, C. Claver, Chem. Soc. Rev. 40 (2011) 4973.
[29] M. Perez-Lorenzo, J. Phys. Chem. Lett. 3 (2012) 167.
}
[30] C.-J. Jia, F. Schuth, Phys. Chem. Chem. Phys. 13 (2011) 2457.
[31] J.A. Widergen, R.G. Finke, J. Mol. Catal. A: Chem. 198 (2003) 317.
[32] M.T. Reetz, J.G. de Vries, Chem. Comm. (2004) 1559.
[33] C. Röhlich, K. Köhler, Chem. Eur. J. 16 (2010) 2363.
[34] C. Röhlich, K. Köhler, Adv. Synth. Catal. 352 (2010) 2263.
[35] F.-S. Han, Chem. Soc. Rev. 42 (2013) 5270.
[36] J. Yamaguchi, K. Muto, K. Itami, Eur. J. Org. Chem. (2013) 19.
[37] K. Seth, P. Purohit, A.K. Chakraborti, Org. Lett. 16 (2014) 2334.
[38] P. Starynowicz, J. Lisowski, Chem. Commun. (1999) 769.
[39] J. Lisowski, Inorg. Chim. Acta 285 (1999) 233.
Acknowledgements
Financial support of National ScienceFoundation – Poland, (NCN)
with grant 2012/05/B/ST5/00265 is gratefully acknowledged (AMT).
[40] W.A. Herrmann, V.P.W. Böhm, J. Organomet. Chem. 572 (1999) 141.
Please cite this article in press as: E. Mieczyn´ ska et al., Inorg. Chim. Acta (2014), http://dx.doi.org/10.1016/j.ica.2014.11.008