Selective recovery of a pyridine derivative from an aqueous waste stream containing acetic acid and succinonitrile with solvent impregnated resins

Article abstract of DOI:10.1016/j.reactfunctpolym.2014.11.007

Solvent impregnated resins (SIRs) were evaluated for the recovery of pyridine derivatives from an aqueous waste-stream containing also acetic acid and succinonitrile. For this purpose, a new solvent was developed, synthesized and impregnated in Amberlite XAD4. Sorption studies were used to determine the capacity, selectivity and the mass-transfer rate. A high capacity of 21 g 4-cyanopyridine (CP) per kg SIR was found, with very high selectivity toward CP over the other solutes of at least 570. A modified Langmuir equation could describe the equilibrium sorption isotherm. Both the linear driving force model and a Fickian diffusion model were evaluated. The Fick-model described both regeneration and loading best. The CP-diffusivity through the solvent phase was estimated at 6.53 · 10^-13 ± 2.5% m² s^-1. The model was validated using fixed-bed column experiments. The R² values for this model ranged between 0.94 at a flow rate of 5 mL/min and 0.99 at a flow rate of 1 mL/min during the loading cycle. Due to mass-transfer limitations the breakthrough profiles were broad and breakthrough occurred after 5 or 23 bed volumes, for flow rates of 5 and 1 mL/min, respectively. Both acetic acid and succinonitrile broke through immediately due to the very high CP-selectivity of the SIR.

Full text of DOI:10.1016/j.reactfunctpolym.2014.11.007

Reactive & Functional Polymers 86 (2015) 67–79
Contents lists available at ScienceDirect
Reactive & Functional Polymers
journal homepage: www.elsevier.com/locate/react
Selective recovery of a pyridine derivative from an aqueous waste
stream containing acetic acid and succinonitrile with solvent
impregnated resins
d
J. Bokhove a, T.J. Visser b, B. Schuur c, , A.B. de Haan
⇑
a Eindhoven University of Technology, Process Systems Engineering Group, Department of Chemical Engineering and Chemistry, PO Box 513, 5600 MB Eindhoven, The Netherlands
b Syncom B.V. Kadijk 3, 9747 AT Groningen, The Netherlands
c University of Twente, Faculty of Science and Technology, Sustainable Process Technology Group, Green Energy Initiative, PO Box 217, 7500 AE Enschede, The Netherlands
d Faculty of Applied Sciences, Department of Chemical Engineering, Delft University of Technology, Julianalaan 136, 2628 BL Delft, The Netherlands
a r t i c l e i n f o
a b s t r a c t
Article history:
Solvent impregnated resins (SIRs) were evaluated for the recovery of pyridine derivatives from an
aqueous waste-stream containing also acetic acid and succinonitrile. For this purpose, a new solvent
was developed, synthesized and impregnated in Amberlite XAD4. Sorption studies were used to
determine the capacity, selectivity and the mass-transfer rate. A high capacity of 21 g 4-cyanopyridine
(CP) per kg SIR was found, with very high selectivity toward CP over the other solutes of at least 570.
A modified Langmuir equation could describe the equilibrium sorption isotherm. Both the linear driving
force model and a Fickian diffusion model were evaluated. The Fick-model described both regeneration
and loading best. The CP-diffusivity through the solvent phase was estimated at 6.53 ꢀ 10ꢁ13 2.5%
m2 sꢁ1. The model was validated using fixed-bed column experiments. The R2 values for this model
ranged between 0.94 at a flow rate of 5 mL/min and 0.99 at a flow rate of 1 mL/min during the loading
cycle. Due to mass-transfer limitations the breakthrough profiles were broad and breakthrough
occurred after 5 or 23 bed volumes, for flow rates of 5 and 1 mL/min, respectively. Both acetic acid
and succinonitrile broke through immediately due to the very high CP-selectivity of the SIR.
Ó 2014 Elsevier B.V. All rights reserved.
Received 2 September 2014
Received in revised form 9 November 2014
Accepted 18 November 2014
Available online 3 December 2014
Keywords:
Solvent impregnated resin
Brominated phenol
Pyridine derivatives
Waste water treatment
Fixed bed
1. Introduction
In this study, a promising alternative for these technologies, the
solvent impregnated resin (SIR) [3] was evaluated for a specific
Trace removal involves the removal of impurities present in low
concentrations from either waste-streams or product streams, and
aims at preventing emission of toxic compounds or at product
purification. The main issue in trace-removal is the high number
of separation stages that are required to achieve the desired
trace-compound(s) removal from the product stream. Typically
outlet concentrations below 5 ppm are desired and especially
when large streams are to be purified, this could lead to large
equipment if the separation is less efficient. Novel, highly efficient
technologies are thus desired, and because traditional separation
technologies like adsorption [1] and extraction [2] can both yield
high capacities and selectivities, they suffer from drawbacks like
difficult regeneration in the case of adsorption, and entrainment
or irreversible emulsification in the case of extraction.
case of wastewater treatment. The wastewater stream that is con-
sidered in this work consists of pyridine derivatives (e.g. 4-cyano-
pyridine, CP), nitriles (e.g. succinonitrile) and oxygenated
compounds (e.g. acetic acid) that are typically formed in the pro-
duction of cyanopyridine, acrylonitrile and pyridine [4–7]. The
aim was to selectively remove the highly water soluble pyridine
derivatives from this stream, since without precautions there is a
chance that they might end up in the wastewater stream, resulting
in highly diluted aqueous waste streams that are complex of nat-
ure. Due to the toxicity and poor biodegradability of some of these
components, the treatment of such wastewater streams is of great
environmental importance [4–7].
Using a SIR-based process [3,8] for water purification is benefi-
cial, because the solvent is immobilized in a macro-porous particle,
and as a result, mixing and settling of the aqueous and organic
phase are no longer required and entrainment and irreversible
emulsification are prevented. In the literature, SIRs have been
developed for the in-situ recovery of products from a fermentation
⇑
Corresponding author.
E-mail addresses: Jeroen.Bokhove@gmail.com (J. Bokhove), t.visser@syncom.nl
(T.J. Visser), b.schuur@utwente.nl (B. Schuur).
http://dx.doi.org/10.1016/j.reactfunctpolym.2014.11.007
1381-5148/Ó 2014 Elsevier B.V. All rights reserved.

68
J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
Nomenclature
Abbreviations
kf
KLDF
KD
L
Mw,cp
n
q
qs,tot
R
r
S
t
aqueous phase resistance toward mass-transfer [sꢁ1
]
CP
4-cyanopyridine
overall mass-transfer coefficient [sꢁ1
distribution coefficient [–]
]
(CP)nDBP complex of 4-cyanopyridine with the reactive solvent
CPH+
DMF
DMS
HAc
H+
protonated 4-cyanopyridine
dimethylformamide
dimethylsulfate
acetic acid
proton
length of the column [m]
molecular weight of 4-cyanopyridine [g moleꢁ1
stoichiometry [–]
]
loading of the SIR [g CP/kg SIR]
solvent loading of the SIR [mL solvent/kg SIR]
radius of the SIR [m]
radial position [m]
selectivity [–]
LDA
DBP
SIR
lithium diisopropylamide
the reactive brominated phenol solvent
solvent impregnated resin
succinonitrile
SN
time [s]
Re
Reynolds
u
V
z
interstitial velocity [m/s]
volume [m3]
normalized position in the bed [–]
Symbols
A
c
CF
D
Dax
dp
eb
cross-sectional area of the column [m2]
concentration [g mꢁ3
capacity factor [L/g]
]
Subscripts
eq
org
aq
bt
equilibrium
effective diffusion coefficient [m2 sꢁ1
]
organic phase
aqueous phase
breakthrough
axial dispersion coefficient [m2 sꢁ1
diameter of the SIR [m]
]
void fraction in the bed [–]
void fraction in the resin [–]
ep
superscript
time
Kacpp
Ka
apparent complexation strength [(L moleꢁ1 1ꢁn
) ]
t
dissociation constant [L moleꢁ1
]
broth [9], the recovery of caprolactam from water [10] and the
removal of several other polar organic compounds from water
[11–17]. The drawback of this technology is that leaching of
solvent may result in a fast depletion of the capacity [18], which
should be minimized by selecting a solvent with a low solubility
in water and a high affinity for the resin.
during the loading of a column and the effluent concentration dur-
ing the regeneration of the column. The models were validated
with experimental fixed-bed column data. Finally, the evaluated
model was used to perform process-simulations to determine the
In our previous study a solvent selection procedure was pre-
sented for the selective removal of CP from an aqueous phase
[19]. In that study, it was found that with phenol based solvents
the highest capacities were obtained, whereas 4-nonylphenol has
a very low solubility (5 ppm) [20] in the aqueous phase, which is
beneficial for SIR processes. By impregnating 4-nonylphenol in
Amberlite XAD4, a resin consisting of polystyrene cross-linked
with divinylbenzene, a highly stable SIR was obtained [21]. The
regeneration could be performed by a pH-swing with hydrochloric
acid at a pH of 1, where the concentration of CP could be increased
from 0.5 g/L in the loading cycle as feed solution to a maximum of
3 g/L in the effluent during the regeneration cycle. However, the
capacity of the SIR containing 4-nonylphenol was limited, and for
this reason a modified phenol with a higher capacity for the target
compound was developed and custom-synthesized. The designed
solvent, consisting of a 1:1 mixture (mole basis) of 3,5-dibromo-
4-(4,6,6-trimethylheptyl)phenol and 3,5,-dibromo-4-(4,8-dim-
ethylnonyl)phenol is presented in Fig. 1. The solvent (abbreviated
by DBP) is a mixture of two molecules varying in their alkyl chain
to reduce the viscosity.
After validating the high capacity of 43 g CP/kg SIR [19], the aim
of the here presented study was to characterize the performance of
a SIR consisting of Amberlite XAD4 resins impregnated with the
brominated solvent for the selective recovery of CP from a mixture
of acetic acid (HAc) and succinonitrile (SN), the molecular struc-
tures of the solutes are given in Fig. 2.
Elements of the study to characterize the performance of the SIR
include the development of a thermodynamic model to describe
the capacity, a mass-transfer model to evaluate the mass-transfer
rate and a fixed bed model to evaluate the breakthrough profiles
Fig. 1. Molecular structure of the solvent, consisting of a 1:1 mixture of two
alkylated dibromophenols.

J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
69
isotherm including the interactions in the solvent phase. The inter-
actions included in the equilibrium isotherm are the hydrogen
bonding interactions between CP and the phenol based solvent.
Here, both nitrogen atoms of CP act as Lewis base and the hydroxyl
functional group of DBP acts as Lewis acid [22]. In a previous study
[23], we established a model describing interactions of the phenol
based solvents with CP through both the cyanide group and the
pyridine nitrogen in the aromatic ring. Additionally, phenols can
form self-associates, and also the oligomers of phenols can com-
plex with CP as depicted in Fig. 3. The model [23] could describe
the experimental data with high accuracy, but needed a large
amount of data to regress the parameters. To simplify the system
in order to minimize computational efforts and the required
data-points, a simplified model was developed. The model is based
on the insights from the previous study. Instead of modeling all
complexations between the phenolic extractant and its oligomers
with CP individually, all interactions were lumped in a single
equilibrium reaction (Eq. (1)). Eq. (1) describes the complexation
reaction between CP and DBP with the formation of ((CP)nDBP)
complexes with average stoichiometry designated by n. From the
previous analysis it can be concluded that multiple phenol
molecules can complex with CP, and therefore n indicates the
overall stoichiometry and equals i + j as in Fig. 3.
Fig. 2. Molecular structures of the solutes: 4-cyanoypridine, acetic acid and
succinonitrile.
length of the mass-transfer zone and to study the radial concentra-
tion profiles inside the SIR during regeneration.
2. Theory
In the following section the model that was developed in this
study is discussed. There are 3 main elements in the model as pre-
sented in Fig. 3:
ꢂ The thermodynamic liquid–liquid equilibrium model, describ-
ing the concentrations of the species in the impregnated organic
solvent phase at equilibrium with the bulk aqueous phase
composition.
nCP þ DBP¡ðCPÞnDBP
The complexation constant Kacpp [(m3 moleꢁ1 1ꢁn
ð1Þ
)
] describes the
thermodynamic equilibrium and is defined by Eq. (2) where all the
concentrations are expressed in [mole mꢁ3]. Since the pyridine
nitrogen and cyanide nitrogen do not have the same reactivity,
and self-association of DBP is not separately included, the
complexation constant is defined as an apparent complexation
constant in which all these effects are lumped. Furthermore, the
physical solubility of CP in the organic phase was neglected and
the stoichiometry was assumed to be independent of the concen-
tration of CP in the organic phase. In Eq. (2) all concentration are
expressed as molar concentration [mole mꢁ3] and the overbar
designates the concentration in the organic phase.
ꢂ The mass-transfer model, describing the diffusion through the
organic phase in the pores of the resin.
ꢂ The fixed-bed model, describing the axial concentration gradi-
ents in the bed.
2.1. Thermodynamic model
The thermodynamic equilibrium of the extraction with a
solvent impregnated resin can be described using an adsorption
Fig. 3. Overview of the model consisting of the three elements. In the equilibrium model Ph is the brominated phenol, m the physical partitioning, KC,i,j the complexation
constant for complexation of oligomers of phenol and CP, i and j indicate the length of an oligomer and the stoichiometry of the complex. In the Mass-transfer model
hypothetical concentration gradients in the SIR are displayed to illustrate how the concentration gradients build up in time inside the pores of the SIR particle.

70
J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
h
i
ꢂ
ꢂ
ꢂ
ꢂ
@cCP
@t
@2cCP
@r2
ðCPÞnDBP
¼ ꢁD ꢀ
ð7Þ
Kacpp
¼
ð2Þ
h
i
CP n ꢀ ½DBPꢃ
r
The SIR is assumed to be spherically symmetrical, and therefore
the concentration gradients in the center of the SIR are equal to
zero as boundary condition (Eq. (8), see also Fig. 3). In a previous
study it was shown that due to the high viscosity of the solvent,
the mass-transfer limitations will be completely in the organic
phase inside the pores of the SIR [21]. For this reason mass-transfer
resistance in the aqueous film was neglected and the organic side
interface concentration was equal to the equilibrium concentration
as defined by Eq. (9).
The thermodynamic equilibrium model is applied in the expres-
sion for the isotherm, given in Eq. (3). The isotherm is defined in a
similar way as reported for the extraction of citric acid by
tri-n-octylamine [17]. It can be considered as a modified Langmuir
isotherm, with as main difference the stoichiometry of the
complexation reaction that is included by the power n.
!
ꢀ
ꢁ
n
Kacpp ꢀ DBP ini ꢀ ½CPꢃ
qeq
¼
ꢀ qS ꢀ Mw;CP
ð3Þ
n
1 þ Kacpp ꢀ ½CPꢃ
ꢂ
ꢂ
@cCP
@t
ꢂ
¼ 0
ð8Þ
ꢂ
In Eq. (3), ½DPBꢃini is the initial phenol concentration [mole mꢁ3],
r¼0
[CP] is the equilibrium aqueous phase concentration of CP
[mole CP mꢁ3 solvent], qS is the solvent loading of the SIR particle
[m3 solvent kgꢁ1 SIR], Mw,CP is the molecular weight of CP
[g moleꢁ1] and qeq is the loading of the SIR [g CP kgꢁ1 SIR]. The
parameters Kacpp and n were regressed to the experimental data.
Next to the capacity, the selectivity of the solvent is of great impor-
tance as the aim is to selectively recover CP from a mixture also
containing SN and HAc. The selectivity was evaluated on the basis
of the capacity factors. The capacity factor (CF [m3 kgꢁ1]) was
defined as the ratio between the equilibrium concentration in
the aqueous phase and the amount adsorbed by the SIR at that
concentration as defined in Eq. (4).
!
n
Kacpp ꢀ ½DBPꢃini ꢀ ½CPꢃ
qjr¼R
¼
ꢀ Mw;CP
ð9Þ
n
1 þ Kacpp ꢀ ½CPꢃ
The concentration of the aqueous phase can be described by Eq.
(10), employing Fick’s first law of diffusion. Where 3/R is the effec-
tive surface area [m2 mꢁ3], VSIR the volume of the solvent impreg-
nated resin [m3] and Vaq the volume of the aqueous phase used in
an experiment [m3].
ꢂ
ꢂ
ꢂ
ꢂ
@cCP
@r
@cCP
@t
3
¼ ꢁ ꢀ D ꢀ
R
VSIR
Vaq
ꢀ
ð10Þ
r¼R
Where the Fick-diffusion model gives the more complete
description of the mass-transfer, it might be possible to reduce
computational efforts by applying a simpler mass-transfer model
like the LDF-model defined by Eq. (11) [26].
qeq
CF ¼
ð4Þ
Ceq;aq
The selectivity (S [–]) of the solvent for CP over HAc is then
defined by Eq. (5), and the selectivity of the solvent for CP over
SN by Eq. (6).
ꢃ
ꢄ
@q
@t
¼ KLDF
ꢀ
qteq ꢁ qt
ð11Þ
CFCP
CFHAc
Eq. (11) relates an overall mass-transfer rate to the difference
SCP;HAc
¼
ð5Þ
ð6Þ
between the equilibrium capacity at time t (qteq [g CP kgꢁ1 SIR]),
the actual capacity at time t (qt [g CP kgꢁ1 SIR]) and the overall
mass-transfer coefficient (KLDF [sꢁ1]). The KLDF depends on the dif-
fusivity of CP through the organic phase and can be estimated by
regressing Eq. (11) to the experimental data obtained with zero-
length column experiments.
CFCP
CFSN
SCP;SN
¼
2.2. Mass-transfer
2.3. Fixed bed column model
The overall rate of mass-transfer in SIRs is typically a function of
the diffusivity of the solute in the organic phase, and by calculation
of the mass-transfer rates the length of the mass-transfer zone in a
fixed bed column can be estimated. Several models in the literature
have been developed based on empirical correlations like the
shrinking core model [15,16], Elvoich equation [24] and models
that describe the combined diffusion and chemical reaction inside
the SIR [11,15,25]. In this study we evaluated both the linear
driving force model (LDF-model) [26] and a model based on Fickian
diffusion [27], in order to evaluate their applicability to describe
both loading and regeneration experiments.
In order to describe the concentration profile in a fixed bed, the
previously derived equations for the equilibrium and the mass-
transfer rate were coupled to the overall mass-balance defined
by Eq. (12). The left-hand side of Eq. (12) contains the none-sta-
tionary terms that describe the time-dependent change in the con-
centration in the aqueous phase and the mass-transfer rate to the
SIR particles, that was either calculated with the Fick model or the
LDF-model. The right-hand side contains the convective and dis-
persive term.
In the model based on Fickian diffusion, the mass-transfer rate
depends on the diffusivity of the solute in the organic phase and
can be described by Fick’s second law of diffusion as defined by
Eq. (7). Here, in contrast to the Maxwell–Stefan approach for
multi-component diffusion [28], it is assumed that the mass-trans-
fer rate can be fully described by a single diffusion coefficient of CP
through the solvent. In Eq. (7), cCP is the concentration of CP in the
solvent [mole CP mꢁ3 solvent], r is the radial position in the SIR
particle [m], t is the time [s] and D is the effective diffusivity
[m2 sꢁ1]. The effective diffusion coefficient is defined as the diffu-
sion coefficient of CP through the solvent corrected for the pore
size and orientation of the resin and is obtained by regressing
Eq. (7) to the data obtained with zero-length column experiments.
@cCP
@t
@q
@t
Qv @cCP
A ꢀ L @z
Dax @2cCP
eb
ꢀ
þ ð1 ꢁ ebÞ ꢀ qSIR
ꢀ
¼ ꢁ
þ
eb
ꢀ
ꢀ
ð12Þ
L2
@z2
In Eq. (12) CCP [g CP Lꢁ1] is the concentration CP in the aqueous
phase, eb [–] is the void fraction of the bed, q [g CP kgꢁ1 SIR] the
loading of the SIR, Qv [m3 sꢁ1] the volumetric flow rate, A [m2]
the cross-sectional area of the column, L [m] the length of the
column, Dax [m2 sꢁ1] the axial dispersion coefficient and z [–] is
the normalized axial coordinate defined as the position in the
column divided by the length of the column.
In the regeneration cycle, the pH of the aqueous phase is
reduced strongly, this leads to protonation of aqueous CP, resulting
in a shift in the equilibrium distribution. As only the neutral form

J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
71
of CP is soluble in the organic phase [29], the pH swing results in a
reduction of the equilibrium capacity, and the SIR particles are
regenerated. The equilibrium of the protonation reaction can be
described by Eq. (13). Because of the instantaneous protonation,
the aqueous phase composition using Eq. (13) is solved simulta-
neously with Eq. (12) to describe the concentration profiles of
the neutral form and the protonated form of CP and at each
position in the fixed bed.
detector. GC–MS apparatus: Agilent 6890 series with Agilent
5973 mass selective detector; Column: Varian FactorFour VF-5MS
(CP9013), 30 m ꢄ 250
60 °C initially, and after 2 min a 20 °C/min ramp until 250 °C. Inj.
temp: 250 °C; Inj vol.: 3 L; Split ratio: 75:1; Detector: MSD EI.
l
m ꢄ 0.25
lm. Temperature program:
l
3.2.2. Synthesis procedure of the brominated phenol
The developed route for the synthesis of compounds 1 and 2 is
depicted in Fig. 4, and starts from 3,5-dibromophenol 3 which was
protected by a bulky trimethylsilyl group. The para-position of
compound 4 was selectively deprotonated by lithium diisopropyl-
amine (LDA) as reported previously [33], and the anion quenched
with dimethylformamide (DMF) to afford aldehyde 5. After initial
unsuccessful attempts to keep the TMS protected group intact by
quenching the reaction with base, pure 2,6-dibromo-4-hydroxy-
benzaldehyde 5 was isolated after quenching with acid. Subse-
quently, compound 5 was treated with dimethyl sulfate (DMS) in
acetone to afford 6 in 95–99% yield (in water the reaction did not
proceed well, even after adding additional portions DMS).
ꢀ
ꢁ
½CPꢃ ꢀ Hþ
ꢀ
ꢁ
Ka
¼
ð13Þ
CPHþ
The pH in the bed is a function of the pH in the feed. At the end
of the loading cycle, the pH is uniform throughout the bed, and
when starting to percolate the column with the hydrochloric
acid solution the concentration profile is described by the axial
dispersion term, as depicted in Eq. (14).
þ
þ
þ
@cH
@t
Qv @cH
A ꢀ L @z
Dax @2cH
eb
ꢀ
¼ ꢁ
þ
eb
ꢀ
ð14Þ
L2
@z2
Grignard reactions on 6 were attempted with alkylmagnesium
iodides prepared from alkyl iodides 9 and 10, and with alkylmag-
nesium bromides prepared from alkyl bromides 11 and 12
(Fig. 5). Product 7 was formed from 9 and 10 in low yield (20–
30%) probably due to the formation of by-products via cross-cou-
pling reactions, starting from 11 and 12, the yield improved to 57%.
Substantial amounts of benzylic alcohol 13, and mono-
bromides 14 and 15 were formed during the 1,2-addition reaction
(Fig. 6). Bromides 11 and 12 were prepared in excellent yield from
their commercially available alcohols by treatment with HBr and
The boundary conditions of Eqs. (12) and (14) were as follows,
the concentration at z = 0 is equal to the feed concentration
(Eq. (15)) and a smooth outlet concentration profile was assumed
(Eq. (16)). In Eqs. (15) and (16), L [m] is the length of the fixed bed.
ꢂ
ꢂ
þ;CP
þ;CP;Feed
cH
¼ cH
ð15Þ
z¼0
ꢂ
ꢂ
ꢂ
ꢂ
þ;CP
@cH
¼ 0
ð16Þ
@z
z¼L
The axial dispersion coefficient was calculated using the
Chung–Wen correlation. In a previous study it was shown that
the axial dispersion in the case of a SIR consisting of Amberlite
XAD4 impregnated with 4-nonylphenol had hardly any effect on
the breakthrough profile [21]. However, it was required to estimate
the concentration gradient of the acid through the column and was
therefore included in the model. The Chung and Wen correlation
[30] is given in Eq. (17), where u is the interstitial velocity
[m sꢁ1], Re is the Reynolds number and dp is the particle diameter
[m]. The concentration gradients of SN and HAc were both
calculated according to Eq. (14) as they were assumed to have a
negligible concentration in the organic phase, an assumption that
was validated experimentally (vide infra).
sulfuric acid [34,35]. Dehydroxylation of compounds
7 was
accomplished by reaction with triethylsilane and borontrifluoride
etherate in quantitative yield. Demethylation with BBr3 gave target
compounds 1 and 2 in respectively 89% and 78% yield after column
chromatography. For additional information regarding the
synthesis of the intermediates see the supplementary information.
3.3. SIR preparation
Amberlite XAD4 was washed first with water and then with
ethanol prior to the impregnation, after washing with ethanol
the resin was dried during 24 h at 80 °C. The dry resin was brought
into contact with the DBP diluted in hexane during 24 h and atmo-
spheric pressure in an incubator at a stirring rate of 100 rpm. After
this time the hexane was removed in a rotary evaporator at 25 °C
and 250 mbar until all hexane was removed. DBP remains inside
the pores of the resin and the loading of the SIR was determined
by measuring the increase in the weight. The resin was character-
ized by measuring the density using an Accupyc 1330 pycnometer
(Micrometrics). The loading of the SIRs was 0.42 mL/g SIR, corre-
sponding with 95% of the maximum porosity to allow the organic
phase to expand without solvent losses and the density of the
impregnated SIR was 1.24 g/mL.
eb
Dax
¼
ꢀ u ꢀ dp
ð17Þ
0:2 þ 0:0011 ꢀ Re0:48
3. Experimental
3.1. Chemicals
Amberlite XAD4, 4-cyanopyridine (>99.9%), hexane (>97%),
ethanol (>99.5%), acetic acid (>99.7%), succinonitrile (>99%), pyri-
dine (>99.9%) were supplied by Sigma–Aldrich, the Netherlands.
3.4. Batch-wise contacting experiments
3.2. Custom synthesized solvent
Equilibrium adsorption measurements were performed by con-
tacting 0.3 gram of the SIR during 3 days in an incubator with a
stirring rate of 300 rpm with aqueous CP-solutions varying in ini-
tial concentration between 0.1 and 10 g/L. The liquid–liquid
extraction experiments were conducted by mixing the bulk liquid
organic phase with aqueous CP-solutions of concentration 0.5–
3.5 g/L, in solvent-to-feed ratios from 0.1 to 1. The two liquid
phases were magnetically stirred during 24 h. After the samples
were equilibrated, the aqueous phase was analyzed using gas chro-
matography, the SIR loading and organic phase composition were
determined on the basis of a mass balance.
3.2.1. General information
Starting materials for the synthesis of 3,5-dibromo-4-(4,6,6-
trimethylheptyl)phenol
nyl)phenol were commercially available and were used without
further purification. 1H NMR spectra were recorded at 300 MHz
on either a VNMRS spectrometer or a MP300 spectrometer (both
from Oxford Instruments), 13C NMR spectra were recorded on the
MP300 spectrometer (at 75 MHz), and multiplicities were distin-
guished using an attached proton test (APT). HPLC-MS apparatus:
Agilent 1100 series with UV detector and HP 1100 MSD mass
and
3,5-dibromo-4-(4,8-dimethylno-

72
J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
Fig. 4. Route toward target compounds 1 and 2.
Fig. 5. Preparation of alkylhalides.
mode with a pH 1 hydrochloric acid solution, while measuring the
concentration with the inline UV-detector. The flow rate with
which the solutions were pumped through the bed was 25 mL/
min at which level it was ensured that the mass-transfer resistance
in the aqueous phase could be neglected and the mass-transfer rate
is only dependent on the diffusion through the organic phase
inside the SIR particle.
Fig. 6. By-products formed in the Grignard reaction.
3.6. Fixed bed column experiments
3.5. Zero-length column experiments
The breakthrough profiles were measured in fixed bed column
experiments. In these experiments a glass column (Omnifit, Eng-
land) was stacked with a bed of SIR particles with a bed height
of 28 cm and a bed diameter of 1.5 cm. The loading and regenera-
tion cycles were performed at flow rates of 1, 2.5 and 5 mL/min
pumped with a Knauer HPLC pump (Knauer GmbH, Germany).
The feed solution consisted of 500 ppm CP, or a mixture of
500 ppm CP, 500 ppm SN and 4.5 g/L HAc. The effluent was ana-
lyzed with a Smartline 2500 inline UV-detector (Knauer GmbH,
Germany) to determine the concentration of CP. The concentration
of SN and HAc were measured by gas chromatography. The
The rate of the uptake of CP by the SIR was determined in a
zero-length column setup (ZL-column) [31,32]. In this setup a thin
layer of SIR particles (2 mm) is placed inside a glass column (Omin-
fit, England) and a solution containing 0.5 or 3 g/L of CP was circu-
lated through the column while measuring the concentration with
a Smartline 2500 inline UV-detector (Knauer GmbH, Germany). In
the experiments the aqueous phase volume used was chosen such
that the aqueous phase concentration reduced with 30% during the
experiment. After equilibrium was attained, the column was rinsed
with demineralized water to remove the CP, and was fed in recycle

J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
73
regeneration procedure was started when the bed was fully satu-
rated and contained the solution used in the loading cycle. The feed
consisted of a pH 1 hydrochloric acid solution. The effluent was
analyzed using gas chromatography during the regeneration cycle.
3.7. Gas chromatography
Gas chromatography was used for the analysis of the equilib-
rium measurements, the effluent of the fixed-bed column experi-
ments during regeneration and the measurement of HAc and SN
in the fixed-bed column experiments. For the equilibrium mea-
surements and the measurement of the concentration of HAc and
SN during the loading cycle with the fixed-bed column experi-
ments, a sample of the aqueous phase was taken and filtered over
a 45 micrometer filter. In the regeneration of the resins in fixed-
bed column experiments the sample was first mixed with a pH
14 sodium hydroxide solution to set the pH to 7. A sample of
1.2 mL was taken and mixed with 0.3 mL of a 0.10 g/L pyridine
solution which was used as internal standard. The sample was then
injected in a Varian CP-3800 gas chromatograph (Varian Inc, the
Netherlands) equipped with a 25 m ꢄ 0.53 mm CP-WAX column
and flame ionization detector. The injected sample volume was
1 lL, the initial column temperature was 50 °C, followed by a ramp
of 20 °C/min to 200 °C, after this ramp the temperature was
directly increased to 240 °C with a ramp of 50 °C/min. Each sample
was injected 3 times; the average relative standard deviation for
the measurement of all compounds were below 0.5%.
3.8. Mathematical modeling
The thermodynamic equilibrium model was programmed in
Matlab, and data regression was done with the global search func-
tion. The mathematical model to calculate the breakthrough curves
was programmed in gProms model builder 3.3.1, and the equations
were solved using the centered finite discretization method.
Regression of the diffusion coefficient was also done with gProms
model builder 3.3.1 using the data regression tool with a constant
relative variance model set at a relative variance of 3%.
Fig. 7. (a) Distribution coefficient measured by liquid–liquid extraction experi-
ments (s) and SIR experiments (h) and modeled with the isotherm (continuous
line). (b) The SIR isotherm regressed with the modified Langmuir isotherm
(continuous line) and the standard Langmuir isotherm (dashed line), the symbols
are the experimental data points.
4. Results and discussion
while the modified Langmuir isotherm with a stoichiometry of
n:1 (CP:DBP) = 0.73:1 and
a complexation constant of 0.052
4.1. Model development
(m3 moleꢁ1 1ꢁn
)
gave a very good fit of the model with a mean
relative error of 2.3%. The capacity reaches up to 70 g/kg at an
aqueous phase concentration of 6 g/L of CP. The value of n is smaller
than 1, indicating that multiple DBP molecules can attach to 1 CP
molecule. This is in accordance with the expectations, because both
the pyridine and the nitrile functionalities are Lewis bases that may
complex with the Lewis acid phenol. Another parameter of interest
was the selectivity, defined as the ratio of the capacity factors of CP
in comparison with SN and HAc. In the adsorption experiment with
a feed consisting of 500 ppm CP, 500 ppm SN and 4.5 g/L HAc no
significant reduction in the concentration of SN and HAc was
measured while the concentration of CP was reduced by 60%. These
results indicate that the selectivity of this SIR is above 500 for CP
with both succionitrile and HAc, given the analytical uncertainty
in the measurement of these compounds.
4.1.1. Equilibrium model
Batch-wise equilibrium experiments were performed to study
the thermodynamic equilibrium of the extraction of 4CP by the
phenolic solvent. Both liquid–liquid extraction experiments to
study the capacity of the solvent and equilibrium adsorption mea-
surements to determine the capacity of the SIR were performed. In
Fig. 7a the distribution coefficient measured in the liquid–liquid
extraction experiments and in the SIR adsorption experiments
are presented and in Fig. 7b the SIR adsorption isotherm is
presented. In both figures the results of the model after data-
regression is also included.
In Fig. 7a it can be observed that the distribution coefficient var-
ies from a maximum value of 290 to a value of 35 over the concen-
tration range from 0.02 g/L to 6 g/L 4-cyanpyridine in the aqueous
phase. It can also be observed that the distribution coefficients
obtained by the liquid–liquid extraction experiments follow the
same trend as those measured by SIR experiments, which results
in the conclusion that the SIR capacity is fully determined by the
capacity of the solvent. In Fig. 7b, it can be observed that the capac-
ity of the SIR follows a favorable isotherm as expected. It is clearly
visible that the standard Langmuir isotherm, assuming a stoichiom-
etry of 1:1 (CP:DBP), was not able to describe the data accurately,
4.1.2. Mass-transfer
ZL-column experiments were conducted to study the mass-
transfer in the SIR. The Fick-diffusion model and the LDF-model
were regressed to the experimental data obtained with an initial
concentration of 0.5 and 3 g/L, resembling the operating conditions
in the fixed bed column in loading and regeneration. After equilib-
rium was reached, the regeneration was performed with an HCl

74
J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
solution of pH = 1. The results are presented in Fig. 8, where c/c0 for
a loading experiment was defined as the ratio of the measured con-
centration over the initial concentration, and for the regeneration
cycle it was defined as the ratio of the measured concentration
over the equilibrium concentration obtained at the end of the
regeneration cycle.
In Fig. 8a and b it can be observed that the LDF-model is not able
to describe all data accurately. Especially the regeneration cycle at
an initial aqueous phase concentration of 0.5 g/L is largely overes-
timated, and during the loading cycle at 0.5 g/L a slight overestima-
tion of the mass-transfer rate was made. At a higher concentration
the differences tend to become smaller. The LDF-model assumes a
mass-transfer coefficient that is independent on the concentration,
the results however show that the mass-transfer rates are depend-
ing on the concentration and as a result the LDF-model could not
describe all experimental data.
The Fick-model, for which the results are presented in Fig. 8c
and d, gives a better description. The loading cycles at both 0.5
and 3 g/L were well described, and the regeneration cycle with
0.5 g/L as initial concentration is perfectly described with Fickian
diffusion. It can however be observed that at an initial concentra-
tion of 3 g/L there is an underestimation of the mass-transfer rate
during the regeneration cycle. The estimated value of the diffusion
coefficient will be used in the next section to compare the model
results in fixed-bed operation. The overall mass-transfer coefficient
for the LDF model was estimated at 2.2 ꢀ 10ꢁ4 sꢁ1 2.2%. The effec-
tive diffusion coefficient of CP in the SIR particle was estimated at
6.53 ꢀ 10ꢁ13 m2 sꢁ1 2.5%. This low value of the diffusion coeffi-
cient can be explained by the high viscosity of the solvent, as also
found for the transfer of CP in a SIR containing 4-nonylphenol as
solvent [21].
4.1.3. Loading cycle
A fixed-bed column stacked with SIR particles was loaded with
CP by displacing a feed with a 500 ppm solution of CP through the
column at varying flow rates. One additional experiment was per-
formed with a mixture of 500 ppm CP, 500 ppm SN and 4.5 g/L HAc
at a flow rate of 5 mL/min. The breakthrough profiles estimated
with the model and experimentally determined are presented in
Fig. 9.
From Fig. 9 it can be observed that by reducing the flow rate, a
narrowing of the breakthrough profiles occurs. This trend was also
observed previously in the case of 4-nonylphenol impregnated in
Amberlite XAD4 with CP as feed [21]. The narrowing of the break-
through profile at lower flow rates can be explained by the reduced
effect of the mass-transfer limitations due to the longer residence
time of the percolated fluid in the column. Breakthrough of the col-
umn was defined as the moment when the outlet concentration
was 1% of the feed concentration and occurred after approximately
5, 10 and 23 bed volumes at a flow rate of 5, 2.5 and 1 mL/min,
respectively. In Fig. 9a it can be observed that HAc and SN break-
through immediately due to the very high selectivity of the SIR.
The breakthrough profiles of HAc and SN were modeled, assuming
the adsorption would be zero and the gradient was described by
axial dispersion only. In Fig. 9a it may be observed that with this
assumption the breakthrough profile of HAc and SN could be
described well.
4.1.4. Regeneration cycle
After the column was fully saturated, it was regenerated with a
pH 1 HCl solution at a flow rate of 1, 2.5 and 5 mL/min. The outlet
concentration of the column was monitored and the results were
compared with the calculated values with the model using both
the Fick and the LDF-model. In Fig. 10 the outlet concentration nor-
malized to the original feed concentration as measured and calcu-
lated with both models are presented after percolating 1 bed
volume through the bed, i.e. when the outlet concentration started
increasing.
In Fig. 10, it can be observed that the outlet concentration is ini-
tially 7 times higher than the original feed concentration. At a pH
of 1, only 14% of CP is in its neutral form which results in a factor
7 reduction in the capacity and hence a factor 7 increase in the
Fig. 8. Results of the zero-length column experiments with the LDF-model for an
initial concentration of 0.5 g/L (A), 3 g/L (B) and with the Fick-model for an initial
concentration of 0.5 g/L (C) and 3 g/L (D). Lines are the model results: regeneration
cycle (black) and the loading cycle (grey). Symbols are the experimental data for the
loading cycle (s) and the regeneration cycle (h).

J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
75
Fig. 9. Breakthrough profiles of SIR fixed-bed experiments: (a) LDF-model and (b)
Fick model. Lines: flow rate 1 mL/min (black), 2.5 mL/min (dark grey) and 5 mL/min
(light grey), experimental results (dashed) and the model results (continuous). The
black dotted line is the modeled breakthrough profile of SN and HAc. Symbols: HAc
(s) and SN (h).
Fig. 10. Effluent concentration during regeneration, modeled with the LDF-model
(a) and Fick model (b). Lines are model results: flow rate 5 mL/min (black), 2.5 mL/
min (dark grey) and 1 mL/min (light grey). Symbols are experimental data: Flow
rate 5 mL/min (s), 2.5 mL/min (h) and 1 mL/min (D).
aqueous phase concentration at equilibrium. After this maximum
was obtained, it gradually reduces with respect to time as the
SIR particles are regenerated. In Fig. 10a the results of the LDF-
model are included, where it can be observed that at 5 and
2.5 mL/min a large underestimation of this outlet concentration
was obtained and only at 1 mL/min the model agrees well with
the experimental data. In Fig. 10b it can be observed that the
Fick-model agrees with the experimental data under all conditions.
The main difference between the Fick-model and the LDF-model is
that the LDF-model assumes a constant mass-transfer coefficient,
while the mass-transfer rate in the Fick-model is inherently
time-dependent.
Table 1
R2-values obtained with the LDF and fick model for the ZL-experiments and fixed bed
experiments in the loading and regeneration cycle.
Fick
Fick
LDF
LDF
Zero-length column experiment
Loading
0.5 g/L
0.94
0.97
0.87
0.98
3 g/L
Regeneration
0.5 g/L
3 g/L
0.99
0.91
0.69
0.84
Fixed bed experiments
Loading
4.1.5. Model performance overview
1 mL/min
2.5 mL/min
5 mL/min
0.98
0.99
0.97
0.98
0.99
0.99
In the previous sub-sections the models were developed to
describe the mass-transfer of CP from the aqueous phase to the
organic phase inside the SIR with the LDF-model and Fick-model.
A comparison of the accuracy of the two models was made, and
the R2-values that were obtained for the ZL-experiments and fixed
bed experiments, for the loading and regeneration cycle are pre-
sented in Table 1.
Regeneration
1 mL/min
2.5 mL/min
5 mL/min
0.99
0.96
0.95
0.98
0.78
0.45
In Table 1 it can be observed that the Fick-model is best appli-
cable for the description of the mass-transfer rates as obtained in
the ZL-column experiments as was previously concluded. The
dependency of the mass-transfer on the concentration could not
be described by the LDF-model. For the fixed-bed column experi-
ments it can be concluded that during the loading cycle, both the

76
J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
Fick-model and LDF-model could describe all data accurately, and
at a flow rate of 5 mL/min the LDF-model was even more accurate
than the Fick-model. At a fixed position in the bed the concentra-
tion changes from initially 0 to the feed concentration in time. This
gradual increase in the concentration resulted in a small depen-
dency of the mass-transfer coefficient with time, and therefore
the LDF-model was able to describe the experimental data with
sufficient accuracy. With the Fick-model however, the initial
mass-transfer rate will be higher due to the steeper concentration
gradients, this effect seems to be overestimated, resulting in a nar-
rower breakthrough profile than experimentally obtained, and a
higher accuracy with the LDF-model. For describing the regenera-
tion of the column the Fick-model is best applicable, as the
R2-values are much higher than the LDF-model. It can clearly be
observed that the LDF-model could not describe the data as func-
tion of flow rate and is only accurate at low flow rates where the
residence time in the column is longer. The time-dependent
mass-transfer rate as estimated with the Fick-model was required
for an adequate description of the experimental data as also
observed with ZL column experiments. Hence, for simulation of
complete loading and unloading cycles, the model making use of
Fickian diffusion can be applied, whereas the Linear Driving Force
model is inapplicable due to failure in simulating the regeneration.
Fig. 11. Length of the mass transfer zone as function of the flow rate. Colors: flow
rate of 5 mL/min (black), 2.5 mL/min (dark grey) and 1 mL/min (light grey). (For
interpretation of the references to colour in this figure legend, the reader is referred
to the web version of this article.)
through time was defined as the time where the outlet concentra-
tion was at its maximum.
4.2. Simulations
In Fig. 12 it can be observed that the concentration profile
develops as a wave through the column. A maximum concentra-
tion is reached, from this maximum the concentration reduces to
the original feed concentration. The profile is set by the gradient
of the HCl concentration through the column, as described by the
axial dispersion. At a flow rate of 1 mL/min, a maximum outlet
concentration of 7 times the original feed concentration was
reached, which is the thermodynamic maximum at these
conditions. This corresponds also with the flattened concentration
gradient near the exit of the column in Fig. 11, which shows that
the aqueous phase composition is in equilibrium with the organic
phase composition. At a flow rate of 5 mL/min this maximum was
not reached, because the residence time in the column was too
short and the mass-transfer limits the regeneration. Therefore a
Simulations were performed with the model making use of
Fickian diffusion to determine the mass-transfer zone length and
to study the evolution of the axial and radial concentration
gradients. These results help in explaining the observed effects like
the great impact of the flow rate on the width of the breakthrough
profile and mass-transfer enhancement during regeneration. In the
experimental section it was established that the mass transfer zone
length (the part of the column where 0.99 > C/CFeed > 0.01) was
larger than the column length of 28 cm. For this reason simulations
were performed using a significantly longer column of 2 m that
would fit the mass-transfer zone and allows to study the effect of
flow rate on the mass-transfer zone length in a larger range of
flow rates.
4.2.1. The loading cycle
maximum concentration of only approximately
original feed concentration is reached.
5 times the
In the simulation, during the loading cycle, the column that is
initially filled with pure water was from t = 0 continuously fed with
a 500 ppm CP solution at varying flow rates. The length of the
mass-transfer zone (MTZ) was calculated at each time and is
plotted in Fig. 11 as function of the breakthrough time, defined
as the ratio of time and the time until breakthrough.
In Fig. 13, the radial concentration profiles are presented as it
develops inside the SIR. Initially the concentration gradient was
In Fig. 11 it can be observed that initially the lengths of the MTZ
do not strongly vary with flow rate, as time progresses the differ-
ences between the lengths of the MTZ increase. With a flow rate of
2.5 and 1 mL/min, the length of the MTZ hardly increases, while at
a flow rate of 5 mL/min the MTZ continuously increases until the
breakthrough. The continuous increase of the MTZ at 5 mL/min is
due to the shorter residence time in the column, also explaining
the greater differences between the model and experimental results
previously determined. The MTZ length varied between 0.4 m up to
1.2 m under the conditions applied in these simulations.
4.2.2. Regeneration
The regeneration cycle of the SIR by a pH swing was simulated
to estimate the axial concentration gradients in the column and the
radial concentration gradients inside the SIR particle. The column
was loaded with a 500 ppm CP solution, and the simulation was
started with a column where the aqueous phase had a concentra-
tion of CP equal to feed concentration and the SIR particles were
fully saturated with CP. In Fig. 12, the axial concentration gradients
during regeneration are presented, where the regeneration break-
Fig. 12. Axial concentration profile in the column during regeneration at 1 mL/min
(black) and 5 mL/min (light grey) at a regeneration breakthrough time of 0.1
(continuous), 0.5 (dashed) and 1 (dotted).

J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
77
[34] and 1-Bromo-3,7-dimethyloctane 12 [35] were prepared
according to literature procedures. 2,6-Dibromo-4-methoxybenz-
aldehyde (6). To 2,6-dibromo-4-hydroxybenzaldehyde (70 g,
0.25 mol) in acetone (600 mL) was added potassium carbonate
(51.8 g, 0.38 mol) and dimethylsulfate (47.3 g, 0.38 mol). The mix-
ture was stirred overnight at room temperature. The color of the
mixture turns from pink to white. The acetone was removed by
rotary evaporation at 50 °C under reduced pressure. Water
(200 mL) was added to the residue and the mixture was stirred
for 15 min and filtered. The pale yellow solid was washed with
water. The solid was dried in vacuo and stripped with toluene
yielding a pale yellow solid (73 g, 0.248 mol, 99%). 1H NMR
(DMSO-d6): o 10.06 (s, 1H), 7.39 (s, 2H), 3.87 (s, 3H). 13C NMR
(75 MHz, CDCl3): 190.34 (CH), 162.91 (C), 127.22 (C), 124.79 (C),
119.93 (CH), 56.39 (CH3). 13C NMR data identical to [33]. A similar
reaction has been performed on 0.1 mol scale. The product was
obtained in 95% isolated yield.
1-(2,6-Dibromo-4-methoxyphenyl)-4,6,6-trimethylheptan-1-ol
(7a). 1-Bromo-3,5,5-trimethylhexane 11 (62.15 g, 0.3 mol) was
Fig. 13. Radial concentration profiles in the SIR particle during regeneration.
added portionwise to
a suspension of Magnesium (7.44 g,
0.3 mol) in Et2O (100 mL) containing a few drops of 1,2-dibromo-
ethane. After start of the Grignard the remainder of the bromide
was added dropwise at such a rate as to maintain reflux. After
addition the solution was heated at reflux for an additional
steep. This results in the fastest mass-transfer due to a higher driv-
ing force for diffusion. As time progresses the gradient flattens and
mass-transfer reduces, these results show a large dependency of
mass-transfer with respect to time. Because of this large depen-
dency, the regeneration process could be correctly described by
the Fick-model, but not with the LDF-model.
30 min and after cooling added dropwise to
a solution of
2,6-dibromo-4-methoxybenzaldehyde (45 g, 0.15 mol) in THF
(600 mL) keeping T < 27 °C using a waterbath. The solution was
allowed to stir at RT for 3 h and poured into 1 N HCl (500 mL). After
separation of the layers, the water phase was extracted with TBME
(2 ꢄ 250 mL). The combined organic layers were washed with
brine (150 mL), dried (Na2SO4), filtered and concentrated in vacuo
to give 68 g of crude oil. Heptanes (100 mL) were added to the oily
residue and a precipitate ((2,6-dibromo-4-methoxyphenyl)metha-
nol, 12.4 g) formed which was filtered off. The target material dis-
solved in heptanes was subjected to column chromatography
(silicagel, 300 g) and eluted with heptanes to remove alkane cross
coupling by-products and subsequently with heptanes/EtOAc 95/5
to 90/10 to give 7a (36 g in two fractions, 21 g pure and 15 g less
pure, containing 10% of unreacted 1-Bromo-3,5,5-trimethylhexane
11. The yellow oily residues were combined (36 g, 85.2 mmol,
57%). 1H NMR (CDCl3): o 7.10 (s, 2H), 5.28 (m, 1H), 3.78 (s, 3H),
2.74 (dd, 1H), 1.97 (m, 2H), 1.58–1.38 (m, 1H), 1.26–1.18 (m,
2H), 1.08–1.00 (m, 2H), 0.93–0.76 (m, 12H).
1-(2,6-dibromo-4-methoxyphenyl)-4,8-dimethylnonan-1-ol (7b).
This compound was synthesized from 1-Bromo-3,7-dimethyloc-
tane analogous to compound 7a. The material dissolved in hep-
tanes was subjected to column chromatography (silicagel, 300 g)
and eluted with heptanes to remove alkane cross coupling by-
products and subsequently with heptanes/EtOAc 95/5 to 90/10 to
give 7b (33.7 g in two fractions, 15.2 g pure and 22 g less pure, con-
taining 10% of 1-(2-bromo-4-methoxyphenyl)-4,8-dimethylnonan-
1-ol. The yellow oily residues were combined (33.7 g, 85.2 mmol,
57%). 1H NMR (CDCl3): o 7.12 (s, 2H), 5.30 (m, 1H), 3.80 (s, 3H),
2.76 (d, 1H, J = 9.8 Hz), 2.1–1.9 (m, 2H), 1.59–1.04 (m, 10H), 0.95
(m, 9H).
5. Conclusions
A solvent impregnated resin (SIR) was developed for the selec-
tive removal of 4-cyanopyridine (CP) from an aqueous waste
stream containing also acetic acid and succinonitrile. The solvent
impregnated resin consisted of Amberlite XAD4 impregnated with
a 1:1 mixture (mole basis) of 3,5-dibromo-4-(4,8-dimethylno-
nyl)phenol and 3,5-dibromo-4-(4,6,6-trimethylheptyl)phenol that
had a capacity of 21 g CP/kg SIR at an aqueous feed concentration
of 500 ppm CP. The selectivity of the solvent impregnated resin
toward CP was above 500. A thermodynamic model was devel-
oped, describing the hydrogen bonding interactions between CP
and the solvent. The model was able to describe the equilibrium
isotherm with high accuracy. The mass-transfer rates were studied
and the diffusion coefficient of CP in this solvent was estimated at
6.53 ꢀ 10ꢁ13 m2 sꢁ1 2.5%. Validation of the model with fixed-bed
column experiments revealed that with a constant diffusion coeffi-
cient the data could be described with sufficient accuracy for the
loading cycle as well as the regeneration cycle using the Fick-
model, R2-values of 0.94 to 0.99 were obtained. The Fick-model
was selected over the linear driving force model, because regener-
ation of the SIR could not be described accurately by the linear
driving force model due to underestimation of the mass-transfer
rates. The axial concentration gradients were simulated and it
was found that mass-transfer is strongly limiting and the mass-
transfer zone lengths varied from 0.4 m to 1.5 m, depending on
the superficial velocity through the bed.
1,3-Dibromo-5-methoxy-2-(4,6,6-trimethylheptyl)benzene (8a).
11-(2,6-Dibromo-4-methoxyphenyl)-4,6,6-trimethylheptan-1-ol
7a (36 g, 85.2 mmol) was dissolved in DCM (500 mL) and cooled to
0 °C. Triethylsilane (38.5 mL, 241 mmol) was added at once fol-
lowed by dropwise addition of BF3.Et2O (15.2 mL, 120.4 mmol)
keeping T < 5 °C and the solution was stirred at this temperature
for 1.5 h. The solution was treated with sat. NaHCO3 (500 mL)
and diluted further with some DCM and stirred for 30 min until
gas evolution ceased. The layers were separated and the organic
phase was washed with brine, dried (Na2SO4), filtered and concen-
trated in vacuo to give 8a (34.2 g, 99%) which was isolated as a
Acknowledgement
This was an ISPT (institute for sustainable process technology)
project.
Appendix A
2,6-Dibromo-4-hydroxybenzaldehyde (5) was prepared in
two steps from 3,5 dibromophenol 3 according to a procedure as
described by Dabrowski [33]. 1-Bromo-3,5,5-trimethylhexane 11

78
J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
yellow oil. The product was used in the next step without further
purification. 1H NMR (300 MHz, CDCl3): o 7.10 (s, 2H), 3.78 (s,
3H), 2.87 (m, 2H), 1.6–0.86 (m, 19H). 13C NMR (75 MHz, CDCl3):
158.20 (C), 133.75 (C), 124.96 (C), 118.31 (CH), 55.89 (CH3),
51.44 (CH2), 39.63 (CH2), 36.58 (CH2), 31.32 (CH2), 30.37 (CH3),
29.34 (CH), 26.46 (CH2), 22.86 (CH3); GC–MS: m/z 404, 406.01,
408. MS calculated for C17H26Br2O: 404.04; 406.03; 408.03.
1,3-Dibromo-2-(4,8-dimethylnonyl)-5-methoxybenzene (8b). This
compound was synthesized analogous to 8a from 1-(2,6-
API ES positive and negative; solution A: 9.65 g ammonium acetate;
2250 mL H2O; 150 mL methanol; 100 mL acetonitrile; solution B:
9.65 g ammonium acetate; 250 mL H2O; 1350 mL methanol;
900 mL acetonitrile. MS calculated for C16H24Br2O: 390.02; MS
(API ES Neg): m/z 389.00 (Mꢁ1).
3,5-Dibromo-4-(4,8-dimethylnonyl)phenol (2). This compound
was synthesized analogous to 1 from 1,3-dibromo-2-(4,8-dim-
ethylnonyl)-5-methoxybenzene 8b. The material was subjected
to column chromatography with heptanes/EtOAc 9/1 to give 2
(28.3 g, 69.6 mmol, 86%) which was isolated as an orange oil
(25.8 g, 63.5 mmol, 78%). 1H NMR (300 MHz, CDCl3): o 7.03 (s,
2H), 4.80 (br s, 1H), 2.84 (m, 2H), 1.60–1.06 (m, 14H), 0.87 (2xd,
9H). HPLC-MS: purity: 96.4% at 215 nm, 94.7% at 288 nm. Column:
dibromo-4-methoxyphenyl)-4,8-dimethylnonan-1-ol
7b.
The
product (34.2 g, quant.) was isolated as a yellow oil and used in
the next step without purification. 1H NMR (CDCl3) o 7.07 (s, 2H),
3.76 (s, 3H), 2.85 (m, 2H), 1.6–1.1 (3xm, 12H), 0.87 (2xd, 9H). 13C
NMR (75 MHz, CDCl3): 158.20 (C), 133.78 (C), 124.94 (C), 118.32
(CH), 55.92 (CH3), 39.60 (CH2), 37.38 (CH2), 37.12 (CH2), 36.55
(CH2), 32.69 (CH), 28.21 (CH), 26.26 (CH2), 24.96 (CH2), 22.97
(CH3), 22.88 (CH3), 19.91 (CH3). GC–MS: m/z 418; 420.1; 422. MS
calculated for C18H28Br2O: 418.05; 420.05; 422.05.
1-Iodo-3,5,5-trimethylhexane (9). 1-Chloro-3,5,5-trimethylhex-
ane (20 g, 0.11 mol) was dissolved in acetone (100 mL) and sodium
iodide (17.0 g, 0.12 mol) was added. The mixture was refluxed for
24 h but the reaction was still incomplete. Therefore additional
sodium iodide (17.0 g, 0.11 mol) was added and the reaction mix-
ture refluxed for another 24 h. The salts were filtered off and
extracted with acetone. The acetone was evaporated. Water
(100 mL) and TBME (200 mL) were added to the residue. After sep-
aration of the layers, the TBME layer was washed with water
(100 mL) and brine (100 mL). After drying (Na2SO4), the organic
phase was concentrated in vacuo to give 9 (27.2 g, 0.1 mol, 90%)
as an oil. 1H NMR (CDCl3) o 3.23–3.11 (m, 2H), 1.88–1.82 (m,
1H), 1.68–1.59 (m, 2H), 1.22–1.05 (m, 2H), 0.92–0.80 (m, 12H).
13C NMR (75 MHz, CDCl3): 50.89 (CH2), 43.42 (CH2), 31.42 (C),
30.60 (CH), 30.29 (CH3), 22.05 (CH3), 5.60 (CH2). GC–MS: purity
96.6%; m/z 254.0. MS calculated for C9H19I: 254.05.
Zorbax SB C18 (2.10 ꢄ 50 mm; 1.8
lm, RRHD 1200 bar); Mobile
phase: Solution A:Solution B = 20:80 (0 min) ? (1 min) ? 0:100
(3 min); Flow: 1.0 ml/min; UV Detection: 215 nm and 288 nm;
Injection volume: 0.2 lL; Mass Detection: API ES positive and neg-
ative; solution A: 9.65 g ammonium acetate; 2250 mL H2O; 150 mL
methanol; 100 mL acetonitrile; solution B: 9.65 g ammonium
acetate; 250 mL H2O; 1350 mL methanol; 900 mL acetonitrile. MS
calculated for C17H26Br2O: 404.04; MS (API ES Neg): m/z 402.95
(Mꢁ1). A small impurity which was identified as 3-Bromo-4-
(4,8-dimethylnonyl)phenol was detected (2.9% at 215 nm and
4.8% at 288 nm; MS calculated for C17H27BrO: 326.12, MS (API ES
Neg): m/z 325.10 (Mꢁ1)).
Appendix B. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.reactfunctpolym.
2014.11.007.
References
1-Iodo-3,7-dimethyloctane (10). This compound was synthesized
analogous to 9 from 1-Chloro-3,7-dimethyloctane. Target material
10 (27.2 g, 89%) was isolated as an oil. 1H NMR (CDCl3) o 3.29–3.12
(m, 2H), 1.93–1.81 (m, 1H), 1.70–1.48 (m, 3H), 1.33–1.10 (m, 6H),
0.88–0.90 (m, 9H). 13C NMR (75 MHz, CDCl3): 41.21 (CH2), 39.42
(CH2), 36.72 (CH2), 34.11 (CH), 28.18 (CH), 24.78 (CH2), 22.96
(CH3), 22.86 (CH3), 19.12 (CH3), 5.77 (CH2). Analytical data is iden-
tical to published data [36]. GC–MS: purity 99.3%; m/z 268.0. MS
calculated for C10H21I: 268.07.
[1] S.W. Wang, A.L. Hines, D.S. Farrier, J. Chem. Eng. Data 24 (1979) 345–347.
[2] S. Kumar, K.L. Wasewar, B.V. Babu, Chem. Eng. Technol. 31 (2008) 1584–1590.
[3] A. Warshawsky, J.L. Cortina, M. Aguilar, K. Jerabek, New developments in
solvent impregnated resins. An overview, in: M. Cox, M. Hidalog, M. Valiente
(Eds.), Solvent Extraction for the 21st Century, Proceedings of ISEC ‘99,
Barcelona, Spain, July 11–16, 1999, 2001, pp. 1267–1272.
[4] Y. Bai, Q. Sun, R. Xing, D. Wen, X. Tang, J. Hazard. Mater. 181 (2010) 916–922.
[5] R.R. Chaudhary, P. Kumar, S. Chand, J. Sci. Ind. Res. 65 (2006) 757–759.
[6] K.V. Padoley, S.N. Mudliar, S.K. Banerjee, S.C. Deshmukh, R.A. Pandey, Chem.
Eng. J. 166 (2010) 1–9.
[7] J.M. Wyatt, C.J. Knowles, Biodegradation 6 (1995) 93–107.
[8] N. Kabay, J.L. Cortina, A. Trochimczuk, M. Streat, React. Funct. Polym. 70 (2010)
484–496.
[9] C. Van Den Berg, F. Boon, M. Roelands, P. Bussmann, E. Goetheer, D. Verdoes, L.
Van Der Wielen, Sep. Purif. Technol. 74 (2010) 318–328.
[10] X. Gong, Y. Lu, G. Luo, Sep. Purif. Technol. 69 (2009) 71–77.
[11] K. Babic, L.G.J. Van Der Ham, A.B. de Haan, Adsorption 14 (2008) 357–366.
[12] B. Burghoff, E.L.V. Goetheer, A.B. de Haan, J. Ind. Eng. Chem. 68 (2008) 1314–
1324.
[13] B. Burghoff, J. Sousa Marques, B.M. van Lankvelt, A.B. de Haan, React. Funct.
Polym. 70 (2010) 41–47.
3,5-Dibromo-4-(4,6,6-trimethylheptyl)phenol (1). 1,3-Dibromo-5-
methoxy-2-(4,6,6-trimethylheptyl)benzene 8a (34.0 g, 83.7 mmol)
was dissolved in DCM (270 mL) and BBr3 (100 mL, 1 M in DCM,
100 mmol) was added. The reaction was not complete after 24 h
of stirring at room temperature. Therefore additional BBr3
(40 mL, 1 M in DCM, 40 mmol) was added. The reaction was com-
plete after 6 additional hours of stirring. The solution was treated
carefully with H2O (300 mL) and stirred for 30 min. The layers
were separated and the DCM layer was washed with sat. NaHCO3
(300 mL), brine (150 mL), dried (Na2SO4), filtered and concentrated
in vacuo to give 34 g of crude oil. The material was subjected to
[14] R.S. Juang, H.L. Chang, Ind. Eng. Chem. Res. 34 (1995) 1294–1301.
[15] A. Kostova, I. Tsibranska, H.J. Bart, Solv. Extr. Ion Exch. 25 (2007) 127–145.
[16] M.O. Ruiz, J.L. Cabezas, I. Escudero, J. Coca, Chem. Eng. Res. Des. 80 (2002) 537–
542.
column chromatography with heptanes/EtOAc 9/1 to give
1
[17] M. Traving, Chem. Eng. Technol. 25 (2002) 997–1003.
(29.1 g, 74.2 mmol, 89%) which was isolated as an orange oil.
HPLC-MS showed a purity of 98% (215 nm) and 97% (288 nm). 1H
NMR (300 MHz, CDCl3): o 7.03 (s, 2H), 4.87 (br s, 1H), 2.83 (m,
2H), 1.6–0.86 (m, chain, 19H). 13C NMR (75 MHz, CDCl3): 154.08
(C), 134.15 (C), 124.98 (C), 119.76 (CH), 51.44 (CH2), 39.63 (CH2),
36.54 (CH2), 31.32 (CH2), 30.39 (CH3), 29.33 (CH), 26.42 (CH2),
22.88 (CH3). HPLC-MS: purity: 98.00% at 215 nm, 97.08% at
[18] D. Muraviev, L. Ghantous, M. Valiente, React. Funct. Polym. 38 (1998) 259–268.
[19] J. Bokhove, B. Schuur, A.B. Haan, Sep. Purif. Technol. 98 (2012) 410–418.
[20] D. Varhanickova, W.Y. Shiu, D. Mackay, J. Chem. Eng. Data 40 (1995) 448–451.
[21] J. Bokhove, B. Schuur, A.B. Haan, React. Funct. Polym. 73 (2013) 595–605.
[22] C. Laurence, J.-F. Gal, Thermodynamic and spectroscopic scales of hydrogen-
bond, in: C. Laurence, J.-F. Gal (Eds.), Lewis Basicity and Affinity Scales: Data
and Measurement, John Wiley & Sons Ltd., Chichester, 2010.
[23] J. Bokhove, B. Schuur, A.B. Haan, Chem. Eng. Sci. 98 (2012) 410–418.
[24] R.S. Juang, T.C. Chou, J. Chem. Eng. Jpn. 29 (1996) 146–151.
[25] B. Burghoff, E. Zondervan, A.B. de Haan, React. Funct. Polym. 69 (2009) 264–
271.
288 nm. Column: Zorbax SB Aq (2.10 ꢄ 50 mm; 1.7
Mobile phase: Solution Solution B = 50:50 (0 min) ?
(1.5 min) ? 0:100 (1.5 min); Flow: 0.60 ml/min; UV Detection:
210 nm and 288 nm; Injection volume: 0.2 L; Mass Detection:
lm, 1200 bar);
A
:
[26] J.D. Seader, E.J. Henley, Adsorption, ion exchange, and chromatography, in: F.
Welter, P. McFadden (Eds.), Separation Process Principles, John Wiley & Sons
Inc., Hoboken, 2006, pp. 548–620.
l

J. Bokhove et al. / Reactive & Functional Polymers 86 (2015) 67–79
79
[27] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, John Wiley &
Sons, 2006.
[32] N. Schouten, L.G.J. Van Der Ham, G.J.W. Euverink, A.B. de Haan, Sep. Purif.
Technol. 68 (2009) 199–207.
[33] M. Da˛browski, J. Kubicka, S. Lulin´ ski, J. Serwatowsk, Tetrahedron Lett. 46
[28] R. Krishna, J.A. Wesselingh, Chem. Eng. Sci. 52 (1997) 861–911.
[29] J. Bokhove, B. Schuur, A.B. Haan, Solvent impregnated resins (SIRS) for the
trace removal of aromatic nitrogen containing compounds from
wastewater streams, in: F.L. Valenzuela, B.A. Moyer (Eds.), 19th
International Solvent Extraction Conference, Gecamin Ltda., Santiago,
2012, Article 151.
(2005) 4175–4178.
[34] J.E. Raymond, A. Bhaskar, T. Goodson III, N. Makiuchi, K. Ogawa, Y. Kobuke, J.
Am. Chem. Soc. 130 (2008) 17212–17213.
[35] R. Kannan, G.S. He, T.-C. Lin, P.N. Prasad, R.A. Vaia, L.-S. Tan, Chem. Mater. 16
(2004) 185–194.
[30] S.F. Chung, C.Y. Wen, AlChE J. 14 (1968) 857–866.
[31] J. Dunnewijk, H. Bosch, A.B. de Haan, Chem. Eng. Sci. 61 (2006) 4813–4826.
[36] For NMR data see: WO2004007475A1 C.Y. Chen, S. Nagumo, H. Akita, Chem.
Pharm. Bull. 44 (11) (1996) 2153–2156.