Accepted Manuscript

Title: Effect of <!–<query id="Q1">Please check dochead for

correctness.</query>–>solvent composition on the van’t Hoff
enthalpic curve using amylose
3,5-dichlorophenylcarbamate–based sorbent

Authors: Ang-Yen Lin, Kai-Tse Cheng, Sin-Chang Chen,

Hung-Wei Tsui

PII: S0021-9673(17)31153-6
DOI: http://dx.doi.org/doi:10.1016/j.chroma.2017.08.011
Reference: CHROMA 358756

To appear in: Journal of Chromatography A

Received date: 29-3-2017

Revised date: 19-6-2017
Accepted date: 3-8-2017

Please cite this article as: {http://dx.doi.org/

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 1

Effect of solvent composition on the van’t Hoff enthalpic curve using amylose 3,5-
dichlorophenylcarbamate–based sorbent

Ang-Yen Lin, Kai-Tse Cheng, Sin-Chang Chen, Hung-Wei Tsui*

Department of Chemical Engineering and Biotechnology, National Taipei University of

Technology, 1, Sec. 3, Zhongxiao E. Rd., Taipei 10608 Taiwan

*E-mail: hwtsui@ntut.edu.tw; Phone: 886-2-2771-2171#2526; FAX: 886-2-2731-7117.

Manuscript type: Article

Highlights
 Van’t Hoff enthalpy changes are strongly dependent on modifier
concentration.
 Such dependences of solvent composition are not explained by models in
literature.
 New model with solvent effects can describe the van’t Hoff enthalpies.
 The fractions of complexed solute and occupied adsorption sites control the
enthalpic curve.

Abstract

Van’t Hoff plots have been widely used for investigating the thermodynamic
properties of adsorption processes in various chromatography systems. By measuring the
retention factor k over a certain temperature range, the plot of ln k versus 1/T often yields
0
a straight line with a slope of −∆𝐻vH /𝑅. Although this method provides information on
adsorption enthalpy changes, its theoretical basis does not account for the effect of the
solvent. In this paper, the relationship between apparent enthalpy changes determined
directly from van’t Hoff plots and solvent modifier concentrations is systematically
investigated using three simple solutes—tetrahydrofuran, acetone, and tert-butanol—in
an n-hexane-methyl tert-butyl ether (MTBE) mobile phase with an amylose 3,5-
dichlorophenylcarbamate–based sorbent. The apparent enthalpy changes of solutes are
strongly dependent on MTBE concentration, increasing rapidly with MTBE content at
low concentrations but leveling off after ∆H0 reaches approximately −15 kJ/mol. These
data cannot be explained by the thermodynamic model currently used in the literature. A
 2

new three-equilibrium-constant thermodynamic model is developed herein to account for

solute–sorbent, solvent–sorbent, and solute–solvent interactions. The thermodynamic
parameters of the model are estimated from the apparent enthalpy changes at different
MTBE concentrations. The results reveal that two key dimensionless groups control the
van’t Hoff enthalpic curves: the fractions of solute molecules bound to modifier
molecules and adsorption sites occupied by modifier molecules. As a result, the shapes of
van’t Hoff enthalpic curves reflect the adsorption isotherm of MTBE without
complexation or information regarding solute–MTBE complexation without MTBE
competitive adsorption. The new model is thus demonstrated to be more reliable than the
current model for examining the thermodynamic properties of retention mechanisms.

Keywords:
Thermodynamic retention model, chromatography, polysaccharide-based sorbent, van’t
Hoff plot.

1. INTRODUCTION
Polysaccharide (PS)-based chiral stationary phases (CSPs) have been frequently
used for the analytical, preparative, and production-scale separation of enantiomers using
various liquid-phase separation techniques. The effectiveness and versatility of these
sorbents are due to their high-order structures in which a high proportion of atoms have
well-defined chirality. Although numerous studies have explored the recognition
mechanisms of these sorbents, more research is needed to advance the fundamental
understanding of their interaction mechanisms.
Retention mechanism information has often been obtained by varying
environmental conditions. The mobile phase is commonly considered to modulate the
strength of bimolecular solute–sorbent interactions, and this enables the extraction of
information on the types of noncovalent bonds involved. The immobilized
polysaccharide CSPs have further expanded the application range via their extended
choice of mobile phased, resulting in an enormous flexibility for method development.
These covalently bonded columns have notably increased the available solvent toolkit
beyond the standard alcohol/hexane, moving into the more expansive mid-polarity
solvents for process development [1–3]. The mobile phase used for PS-based sorbents is
often a solution of polar modifier in a hydrocarbon. The modifier molecules bind with the
H-bonding functional groups of the sorbent and possibly also the solute. Retention factors
generally decrease as the concentration of the modifier in the mobile phase is increased.
Numerous studies have reported the effects of the polar modifier on retention behavior
[1,3–13]. The observed trend of retention factor dependence on polar modifier content
has enabled the pinpointing of the involved interaction mechanism.
Several retention models have been developed to investigate the effect of solvent
on retention behavior [5–8,14]. For the retention behavior in a reverse-phase system,
Geng and Regnier proposed a stoichiometric displacement model [5] that incorporated
the competitive adsorption of the modifiers and solute–modifier complexation. The
authors also presented a retention model in which retention behavior is a function of the
 3

number of solvent molecules required to displace the solute from the surface and the
adsorption saturation concentration of a protein is related to the displacing agent
concentration and the stoichiometry of the solvent–solute displacement. A more
sophisticated model was developed by Wang et al. that accounted for the effect of
modifier self-association in addition to solute–solvent complexation and the competitive
adsorption of solute and polar modifier [7,8]. At particularly high isopropanol
concentrations, they discovered that retention behavior was controlled by the number z/n,
where z is the total number of modifier molecules displaced from the solute and sorbent
surfaces upon solute adsorption and n is the average modifier self-aggregation number [8].
The thermodynamic quantities of adsorption equilibrium processes can be
obtained by measuring a retention factor over a certain temperature range. Van’t Hoff
0
plots of ln k versus 1/T often yield a straight line with a slope of −∆𝐻vH /𝑅 (where R is
0
the gas constant and ∆𝐻vH is the change in enthalpy). Rao et al. studied the effects of
different organic modifiers on the separation of (R,S)- and (S,R)-emtricitabine
enantiomers when using Chiralpak IA sorbent [15]. Van’t Hoff plots were drawn to
determine the entropic and enthalpic changes of the adsorptions of the two isomers. For a
0
mobile phase consisting of 35 vol% ethanol in hexane, a separation enthalpy (∆∆𝐻vH ) of
−0.0954 kJ mol−1 was reported. Ma et al. demonstrated the separation of hexa-3,4-diene-
3-ylbenzene enantiomers on Chiralcel OD-3 sorbent using pure heptane as the mobile
phase [16]. Van’t Hoff plots of the retention factor revealed that the retention of the
enantiomers is enthalpy-driven, whereas enantioselectivity is entropy-driven.
Thermodynamic parameters derived from van’t Hoff plots are macroscopic
lumped quantities that do not account for the surface heterogeneity of CSPs. For this
reason, Fornstedt et al. used the van’t Hoff equation with bi-Langmuir adsorption
isotherm to study the thermodynamics of the interactions between propranolol
enantiomers in an acetic acid buffer and the protein cellobiohydrolase I sorbent [17].
Retention has a mixed mechanism involving both chiral and achiral interactions. The
results demonstrated that one of the two Langmuir contributions was equal for both
enantiomers and the other contribution accounted for the enantioselective interactions.
The authors concluded that a mixed retention mechanism must be involved if the
adsorbent is heterogeneous.
Although a plot of ln k versus 1/T provides thermodynamic information on solute
adsorption, the theoretical basis of this method notably does not involve solute–solvent
complexation or solvent adsorption [10]. This assumption may be valid for a normal-
phase system when a pure nonpolar mobile phase is used. Derived enthalpy values have
been reported to be strongly dependent on the composition of the solvent [10]. Moreover,
such thermodynamic analyses have often been performed for molecules with complex
structures. The method is not strictly thermodynamically consistent for systems with
solute–solvent and solvent–sorbent interactions. Instead, these individual contributions
are lumped together macroscopically. Nevertheless, such analysis may still yield useful
global information about the adsorption process.
The mobile phase used in chromatography often contains polar modifier to adjust
the strength of solute–sorbent interactions. The retention mechanism is widely explained
 4

as a mixed mechanism involving solute–solvent, solute–sorbent, and solvent–sorbent

interactions. The detailed thermodynamic investigation of these complex retention
behaviors requires van’t Hoff analysis over a sufficiently wide range of modifier
concentrations and the development of a thermodynamic retention model. The retention
behavior of solutes can be described reasonably well using a combination of
thermodynamic analysis and a retention model, as was previously implemented for
solutes with distinct functional groups on Chiralpak IA sorbent [10]. The objective of this
study was to determine the thermodynamic parameters of the mechanism of solute and
solvent adsorption on Chiralpak IE and to use this information to make some inferences
regarding retention behaviors. The second objective was to illustrate how the conclusions
of studies based on thermodynamic data that have lumped together the solute–sorbent,
solute–solvent, and solvent–sorbent interactions may be inaccurate and determine the
possible extent of these errors. These concerns have not yet been addressed in the
literature, to the best of our knowledge.

2. THEORY
2.1. Thermodynamics of retention behavior
2.1.1. Van’t Hoff plot for solute adsorption
By using the solute movement theory [18], retention factor can be derived as
follows:
concentration of adsorbed solute 𝑞SL 𝜑
𝑘 = concentration of solute in mobile phase = 𝐶 (1)
SL +𝐶SL−SV

where, 𝑞SL 𝜑 (in M) is the concentration of the adsorbed solute, 𝜑 is the ratio of the solid
volume to the liquid volume in the column, 𝐶SL (in M) is the free solute concentration,
and 𝐶SL−SV (in M) is the concentration of the solute-solvent complex.
The adsorption of solute is modeled thermodynamically as reversible reactions
and described by the following equation:
𝑞SL
𝐾SL = 𝐶 (2)
SL ∙𝐶SB

where KSL (in M-1) is the equilibrium constant for solute-sorbent interaction and 𝐶SB (in
mol/L of solid volume) is the free sorbent capacity,

By assuming a negligible solute-solvent interaction, 𝐶SL−SV ≈ 0, the retention

factor k is related to the equilibrium constant of the solute adsorption by combining Eqs.
(1) and (2):

𝑘 = 𝐾SL 𝐶SB 𝜑 (3)

Under conditions of thermodynamic equilibrium, the standard Gibbs free energy
0
change ∆𝐺SL upon solute adsorption is related to k by substituting 𝑘/𝐶SB 𝜑 for 𝐾SL :
 5

0 𝑘
∆𝐺SL = −𝑅𝑇 ln(𝜑𝐶 ) (4)
SB

0 0
where T (in K) is the absolute temperature. With the expansion of ∆𝐺SL in terms of ∆𝑆SL
0
and ∆𝐻SL , Eq. (4) is rearranged in the following form:
0 0
∆𝐻SL 1 ∆𝑆SL
ln 𝑘 = − (𝑇) + + ln(𝜑𝐶𝑆𝐵 ) (5)
𝑅 𝑅

0
The temperature dependence of the retention factor can be used for determining ∆𝐻SL . If
0 0
we suppose that∆𝐻SL and ∆𝑆SL /𝑅 + ln(𝜑𝐶SB ) vary only slightly with temperature over
the temperature range of interest, which is generally assumed, it follows that
0
∆𝐻SL 1 1
ln 𝑘2 − ln 𝑘1 = − ( − ) (6)
𝑅 𝑇2 𝑇1

where 𝑘2 and 𝑘1 are the retention factors at temperature 𝑇1 and 𝑇2 , respectively [19–21].
Although the slope of ln 𝑘 versus 1/T plots provides information on the enthalpy change
of solute adsorption and has been widely used in thermodynamic analysis, it must be
0
borne in mind that Eq. (6) has implicitly assumed a constant ∆𝐻SL over the temperature
range considered and a negligible solvent effect. The latter assumption may be valid for a
normal phase system when a pure nonpolar mobile phase is used.

2.1.2. Van’t Hoff plot in the presence of solvent modifier

To model normal phase retention behavior for a polar modifier–hydrocarbon
mobile phase, Wang et al. developed retention models by considering solvent self-
association, solute–solvent complexation, and competitive adsorption [7,8]. For a
monovalent solute, the retention model without consideration of solvent self-association
behavior can be expressed as
0
ln 𝑘 = − ln(1 + 𝐾SV 𝐶SV ) − ln(1 + 𝐾SL−SV 𝐶SV ) + ln(𝐶SB 𝜑𝐾SL ) (7)

where 𝐾SV and 𝐾SL−SV (all in units of M−1) are the equilibrium constants for the
0
modifier–sorbent and solute–modifier interactions, respectively; 𝐶SB 𝜑 (in M) is the total
sorbent capacity; and 𝐶SV (in M) is the polar modifier concentration in the mobile phase.
In this model, both solute and modifier are assumed to exhibit homogeneous monovalent
H-bonding with the sorbent. Furthermore, one-to-one solute–modifier complexation is
assumed. The dependence of k on CSV is determined using the three equilibrium constants
0
and the parameter 𝐶SB 𝜑.
When the total modifier concentration approaches zero, Eq. (7) is reduced to Eq.
(3) [7]:
0 𝑞
lim
0
(𝑘) = 𝐾SL 𝐶SB 𝜑 ≅ 𝐾SL 𝐶SB 𝜑 = 𝐶SL 𝜑 (8)
𝐶SV →0 SL

By taking the derivative of Eq. (7) with respect to 1/T, the concentration
dependence of the apparent enthalpy change can be derived:
 6

0
𝑑 ln 𝑘 ∆𝐻vH −𝐶𝑆𝑉 𝑑𝐾SV 𝐶𝑆𝑉 𝑑𝐾SL−SV 𝑑 ln 𝐾SL
1 =− = 1+𝐾 ( 1 ) − 1+𝐾 ( 1 )+ 1 (9)
𝑑( ) 𝑅 SV 𝐶SV 𝑑( ) SL−SV 𝐶SV 𝑑( ) 𝑑( )
𝑇 𝑇 𝑇 𝑇

0
where ∆𝐻vH (kJ/mol) is the apparent enthalpy change of adsorption determined from the
0
slope of the logarithm of the retention factor versus 1/T. In Eq. (9), a constant 𝐶SB 𝜑 over
the temperature range of interest is assumed, which is usually valid if the sorbent
structure does not change appreciably.
Using the van’t Hoff equation, Eq. (9) can be rearranged:
0 0 𝑆𝑉 𝐶 𝐾SV 0 𝐶𝑆𝑉 𝐾SL−SV 0
∆𝐻vH = ∆𝐻SL − 1+𝐾 (∆𝐻SV )− (∆𝐻SL−SV ) (10)
SV 𝐶SV 1+𝐾SL−SV 𝐶SV

0 0 0
where ∆𝐻SL , ∆𝐻SV , and ∆𝐻SL−SV (all in units of kJ/mol) are the enthalpy changes of the
solute–sorbent, modifier–sorbent, and solute–modifier interactions, respectively.
An equation for solute–modifier complexation can be derived from the solute
mass balance [7]:
𝐶SL−SV 𝑆𝑉 𝐶 𝐾SL−SV
𝜃𝑐𝑜𝑚𝑝𝑙𝑒𝑥 = 𝐶 = 1+𝐾 (11)
SL−SV +𝐶SL SL−SV 𝐶SV

where 𝜃𝑐𝑜𝑚𝑝𝑙𝑒𝑥 is the fraction of solute molecules that are bound to modifier molecules.
During pulse chromatography experiments, the solute concentration is expected to be
much smaller than the polar modifier concentration, so solute adsorption does not affect
modifier adsorption. The fraction of adsorption sites occupied by modifier molecules
𝜃𝑎𝑑𝑠𝑜𝑟𝑝 follows a Langmuir isotherm:
𝑞SV 𝐾 𝐶SV
𝜃𝑎𝑑𝑠𝑜𝑟𝑝 = 𝑞 = 1+𝐾SV (12)
SV +𝐶SB SV 𝐶SV

where 𝑞SV (in mole/L of solid volume) is the concentration of the adsorbed modifier.
Using Eqs. (11) and (12), Eq. (10) can be rearranged in the following form:
0 0 0 0
∆𝐻vH = ∆𝐻SL − 𝜃𝑎𝑑𝑠𝑜𝑟𝑝 ∆𝐻SV − 𝜃𝑐𝑜𝑚𝑝𝑙𝑒𝑥 ∆𝐻SL−SV (13)
0
The obtained ∆𝐻vH is thus related to the fraction of occupied adsorption sites. When there
is strong modifier interaction or a high modifier concentration, it follows from Eq. (12)
that 𝜃𝑎𝑑𝑠𝑜𝑟𝑝 ≈ 1. Under these conditions, nearly all the sorbent binding sites are occupied
by modifier molecules. When the value of the group 𝐾SV 𝐶𝑆𝑉 is much lower than 1, most
0
binding sites are free. Similarly, ∆𝐻vH is related to the fraction of solutes complexed with
modifier molecules in the same way as that of the modifier adsorption. The overall value
0 0 0
of ∆𝐻vH can vary from ∆𝐻SL at extremely low modifier concentrations to (∆𝐻SL −
0 0
∆𝐻SV − ∆𝐻SL−SV ) at extremely high modifier concentrations. The detailed application of
these models to the data is discussed in Section 4.3.
0
Last but not least, the value of ∆𝐻vH depends on temperature because 𝐾SV , 𝐾SL−SV ,
0 0 0
∆𝐻SL ,∆𝐻SV, and ∆𝐻SL−SV are temperature dependent (see also Supplementary Material).
0
The values of ∆𝐻SL can be estimated from the van’t Hoff plots of solute adsorptions in
 7

0 0
pure n-hexane, and the values of ∆𝐻SV , ∆𝐻SL−SV , 𝐾SV , and 𝐾SL−SV can be obtained by
numerically fitting Eq. (10) to van’t Hoff enthalpic data, assuming that the apparent
enthalpic data estimated from van’t Hoff plots represent the values at an average
temperature of the range of interest.

2.2. Thermodynamics of heterogeneous adsorption surface

Because the actual adsorption of the solute onto Chiralpak IE sorbent may be
heterogeneous, the contribution of each interaction to the overall adsorption enthalpy
should be elucidated. For simplicity, TBA is taken as a solute example for formulating
the model. Because the molecular size of TBA is small, the potential of multivalent
interactions and the differences in the accessibility of the binding sites is minimized.
Hence, for TBA, the surface of Chiralpak IE sorbent is assumed to contain two types of
adsorption sites, NH and C=O binding sites, which are considered heterogeneous.
Generally, this heterogeneous adsorption mechanism can be described reasonably well by
the bi-Langmuir adsorption isotherm:
0 0
𝐶SL 𝐾SL,I 𝐶SB,I 𝐶SL 𝐾SL,II 𝐶SB,II
𝑞SL = 𝑞SL,I + 𝑞SL,II = + (14)
1+𝐶SL 𝐾SL,I 1+𝐶SL 𝐾SL,II

0 0
where 𝐶SB,I and 𝐶SB,II are the monolayer capacities of the NH and C=O adsorption sites,
respectively, and 𝐾SL,I and 𝐾SL,II are the respective equilibrium constants for the solute–
sorbent NH site and solute–sorbent C=O site interactions. Each term of this model
accounts for the contribution of one site type. The model has been used successfully to
describe the stationary phases of the protein cellobiohydrolase I [17].
From Eq. (14), the general expression for the retention factor on a heterogeneous
surface under linear adsorption conditions is derived [17]:

𝑘 = 𝑘I + 𝑘II (15)

The retention factor is the sum of two contributions, 𝑘I and 𝑘II , originating from the NH
and C=O binding sites, respectively.
By taking the derivative of the logarithm of the retention factor with respect to 1/T,
Eq. (15) leads to the following equation:
0
∆𝐻vH 𝑑 ln 𝑘 1 𝑑 ln 𝑘I 𝑑 ln 𝑘II
− = 1 = 𝑘 (𝑘I 1 + 𝑘II 1 ). (16)
𝑅 𝑑( ) 𝑑( ) 𝑑( )
𝑇 𝑇 𝑇

Employing the van’t Hoff equation, Eq. (16) is rearranged in the following form:
0 𝑘I 𝑘II
∆𝐻vH = ∆𝐻I + ∆𝐻II (17)
𝑘 𝑘

where ∆𝐻I and ∆𝐻II are the enthalpy changes of the solute–sorbent NH site and solute–
sorbent C=O site interactions, respectively. The apparent enthalpy change determined
using the van’t Hoff plot is an average value weighted by 𝑘I and 𝑘II .
 8

2.3. Computational methodology based on density functional theory

Density functional theory (DFT) was used to predict the energies of the
interactions involved. The computations were performed using the Becke, three-
parameter, Lee–Yang–Parr hybrid functional, which includes the exact energy from
Hartree–Fock theory and the exchange energy from Becke’s exchange functional. The 6-
311+g(d,p) basis set was used to obtain highly accurate energies. This computational
methodology has provided relatively accurate predictions of molecular geometries and H-
bonding energies [10]. The calculations were performed for (a) methyl tert-butyl ether
(MTBE); (b) tetrahydrofuran (THF); (c) acetone (AC); (d) tert-butanol; and (e) a model
Chiralpak IE side-chain molecule, termed “IE” (Fig. 1). The structures of these molecules
were obtained by searching for the global energy minimum. Intermolecular H-bonding
interactions were simulated by performing computations for MTBE–IE, THF–IE, AC–IE,
TBA–IE, and TBA–MTBE pairs.

3. MATERIALS AND EXPERIMENTAL METHODS

3.1. Materials
A covalently immobilized amylose 3,5-dichlorophenylcarbamate (Chiralpak IE)
analytical column, which was 250 mm long with a 4.6-mm column diameter and 5-μm
particle diameter, was purchased from Daicel Chemical Industries, Ltd. (Tokyo, Japan).
AC (99.9%) and MTBE (99.8%) were purchased from Tedia Company, Inc. (Fairfield,
OH, USA). High-performance liquid chromatography– (HPLC-) grade n-hexane (purity
99%) and tert-butanol (99.77%) were purchased from Fisher Scientific (Loughborough,
UK). Tetrahydrofuran (99.8%) was purchased from Echo Chemical co., Ltd. (Miaoli,
Taiwan), and 1,3,5-tri-tertbutylbenzene (TTBB) was purchased from AK Scientific
(Union City, CA, USA).

3.2. HPLC apparatus and procedures

An HPLC apparatus was employed that comprised a variable wavelength detector
(SPD-10A, Shimadzu Corporation, Kyoto, Japan), differential refractive index detector
(Ruby, ECOM Ltd., Praha, Czech Republic), online degassing unit (DG660B, GL
Science Inc., Tokyo, Japan), solvent delivery system (LC-10AT, Shimadzu Corporation,
Kyoto, Japan), and thermostat-equipped column oven (CO-966, JASCO Co. Ltd., Tokyo,
Japan). The mobile phase was a mixture of MTBE and n-hexane. A flow rate of 1.0
mL/min at temperatures ranging from 25 to 40 °C and a pulse injection volume of 20 μL
were used in all experiments. Retention times were measured at least twice and averaged.
The retention time tref of a nonretained solute, TTBB, was used as the reference time and
was 3.1 min for the experimental conditions used.

4. RESULTS AND DISCUSSION

4.1. Enthalpy changes of solute–sorbent binding
 9

The enthalpy changes obtained from the van’t Hoff plots of solute adsorptions in
pure n-hexane were investigated in the temperature range 25–40 °C; the temperature was
increased in steps of 5 °C (Table 1 and Fig. 2). Because the n-hexane molecule is
nonpolar, any interactions between n-hexane and the sorbent and solutes were expected
to be minor or negligible. As the temperature was increased, the retention factors
0
decreased, indicating an exothermic adsorption process. The values of ∆𝐻SL for solute
0
adsorption were then determined from the respective slopes (−∆𝐻SL /𝑅) of the van’t Hoff
plots of ln k versus 1/T. AC, THF, and MTBE, known as H-bonding acceptors, bind
mainly with sorbent NH functional groups. Although the retention factors of THF were
smaller than those of AC, the enthalpy change (−31 kJ/mol) of THF was almost equal to
that of AC. MTBE may have had the lowest retention factors and enthalpy changes
because its methyl and tert-butyl groups were connected to oxygen atom, which may
have imposed a steric restriction to partially prevent the solute from effective H-bonding.
By contrast, the TBA molecules could act as an H-bonding donor for sorbent C=O groups
and an H-bonding acceptor for sorbent NH groups. However, the enthalpy changes and
retention factors of the TBA molecules were only slightly larger than those of AC,
implying that TBA may bind mainly through one type of H-bond (see Eqs. (15) and (17)).
Because the hydrogen atom in a hydroxyl group may sterically disrupt effective (TBA)
HO ↔ HN (IE) interactions, TBA molecules presumably bind mainly with sorbent C=O
groups.

4.2. Dependence of the apparent enthalpy change on the mobile-phase MTBE

content
The apparent enthalpy changes of THF, AC, and TBA, derived from van’t Hoff
plots, increased as the MTBE concentration was increased from 0 to 5 M (Figs. 3 and 4).
This effect was strong for THF and AC, with the enthalpy change increasing rapidly with
MTBE concentration from approximately −31 to −15 kJ/mol, after which it leveled off.
The similar enthalpic curves of THF and AC suggested similar retention mechanisms
over the studied concentration range. As the MTBE concentration increased, the
competitive adsorption of MTBE became stronger, which weakened the effective solute
0 0
adsorption and lead to a smaller absolute ∆𝐻vH . For TBA, ∆𝐻vH changed more gradually
0
from approximately −32 to −15 kJ/mol than for THF and AC. The dependences of ∆𝐻vH
on MTBE content were related to competitive MTBE adsorption and solute–MTBE
complexation. Because TBA was expected to bind mainly with the sorbent C=O groups,
0
the changes in ∆𝐻vH were presumably caused by solute–MTBE complexation.

4.3. Estimation of the parameters of the thermodynamic retention model from the
HPLC data
4.3.1. Modifier–sorbent interactions
To estimate the thermodynamic parameters and test the validity of the
thermodynamic model, Eqs. (11)–(13) were used to numerically fit the enthalpy data in
 10

Fig. 4. For THF and AC, solute–MTBE interactions were assumed to be negligible. AC,
THF, and MTBE were expected to bind mainly with sorbent NH groups. A Langmuir-
0
isotherm-like equation, in which (−∆𝐻SV ) corresponds to the saturation capacity, for the
van’t Hoff enthalpic curve can be derived:
0
0 0 𝐾SV (−∆𝐻SV )𝐶𝑆𝑉
∆𝐻vH − ∆𝐻SL = (18)
1+𝐾SV 𝐶SV

By taking the reciprocal of Eq. (18), the following equation is obtained:

1 1 1 1
0 −∆𝐻 0 = −𝐾 0 (𝐶 ) − ∆𝐻 0 (19)
∆𝐻vH SL SV ∆𝐻SV 𝑆𝑉 SV

0
𝐾SV and ∆𝐻SV can then be estimated from the slope and intercept of Eq. (19). To apply
0
the model to the solutes with MTBE, the values of ∆𝐻SL —namely −31.00 kJ/mol for
THF and −30.66 kg/mol for AC—must be used as those obtained by applying the van’t
Hoff plots to the THF and AC data using pure n-hexane as the mobile phase. Because the
0
values of 𝐾SV and ∆𝐻SV are expected to be the same for THF and AC, the THF and AC
data were summed and this summation was numerically fitted.
The THF and AC data fitting results are presented in Figs. 4 and 5 and Table 2.
The model and data are in close agreement. The intercept of linear plot of Eq. (19)
yielded an enthalpy change of −20.28 kJ/mol for the MTBE adsorption, which is slightly
smaller than the enthalpy changes obtained from DFT calculations (−23.72 kJ/mol) and
MTBE pulse experiments (−24.46 kJ/mol). The reason for this discrepancy is unclear.
The use of a relatively high MTBE concentration in the mobile phase may have
strengthened MTBE–MTBE dipole interactions, which then weakened the adsorption of
MTBE.
Because the apparent enthalpy changes were obtained in the temperature range
0
25–40 °C, the determined values of 𝐾SV (10.40 M−1) and ∆𝐻SV appear to represent the
value at an average temperature of 32.5 °C. The saturation capacity of MTBE adsorption,
0
𝐶𝑆𝐵 𝜑, which is equal to 𝑘/𝐾SV , was estimated using 𝐾SV and the retention factor data of
MTBE in pure n-hexane (mobile phase) at 32.5 °C. The k value at 32.5 °C was
interpolated from the linear fit of the ln k versus 1/T plot (Fig. 2). The resulting value of
0
𝐶𝑆𝐵 𝜑 was 0.115 M and assumed to be mostly independent of temperature if the sorbent
structure did not change appreciably over the temperature range 25–40 °C. With the value
0
of 𝐶𝑆𝐵 𝜑, the values of 𝐾SV at 25, 30, 35, and 40 °C were estimated and are listed in Table
3. The equilibrium constants decreased from 13.27 to 8.26 M−1 as the temperature
0 0
increased. Furthermore, the thermodynamic parameters ∆𝐺SV and ∆𝑆SV were determined
0
using the values of KSV and ∆𝐻SV and are also presented in Table 3, assuming that the
0
value of ∆𝐻SV changed so little with temperature that they can be treated as constant over
the temperature range considered (see also Supplementary Material). The negative
0
entropy changes indicated that MTBE adsorption is enthalpy-driven. The values of ∆𝑆SV
were approximately −47 J/K·mol and mostly independent of the temperature range
considered.
 11

According to Eq. (18) and Fig. 4, the van’t Hoff enthalpic curves of THF and AC
0
essentially represented the adsorption isotherm of MTBE. Using the values of 𝐶𝑆𝐵 𝜑 and
𝐾SV , the Langmuir isotherms of MTBE adsorption were obtained and are presented in Fig.
6.

4.3.2. Solute–modifier interactions

Because TBA molecules presumably bind mainly with sorbent C=O groups and
0
do not compete with MTBE for NH binding sites, the term 𝜃𝑎𝑑𝑠𝑜𝑟𝑝 ∆𝐻SV in Eq. (13) was
assumed to be negligible for TBA. The model was rearranged to the following form:
0
0 0 𝐾SL−SV (−∆𝐻SL−SV )𝐶𝑆𝑉
∆𝐻vH − ∆𝐻SL = (20)
1+𝐾SL−SV 𝐶SV

Taking the reciprocal of Eq. (20), the model can be rearranged to

1 1 1 1
0 −∆𝐻 0 = −𝐾 0 ( ) − ∆𝐻 0 (21)
∆𝐻vH SL SL−SV ∆𝐻SL−SV 𝐶𝑆𝑉 SL−SV

0 0
If the model is followed, a graph of 1/(∆𝐻vH − ∆𝐻SL ) versus 1/𝐶𝑆𝑉 yields a straight line
0 0
with a slope of 1/ (−𝐾SL−SV ∆𝐻SL−SV ) and an intercept of 1/ (−∆𝐻SL−SV ) . The
corresponding plots are presented in Figs. 4 and 7. The resulting parameters were 3.43
0
M−1 for KSL−SV and −19.81 kJ/mol for ∆𝐻SL−SV (Table 4). From these values, the entropy
change following TBA–MTBE complexation was estimated to be −54.57 J/K·mol. The
determined enthalpy change of TBA–MTBE complexation was slightly larger than the
enthalpy change of −15.12 kJ/mol predicted from the DFT calculation of the O group of a
MTBE molecule H-bonded to the HO group of a TBA molecule. This difference may
have been due to the potential TBA–MTBE interactions in addition to the H-bond
considered and the minor contribution of the competitive adsorption of modifiers and
solute, which was ignored in this study. Similar to the enthalpic curves of THF and AC,
the enthalpic curve of TBA in Fig. 4 represents the dependence on MTBE concentration
of the fraction of TBA molecules bound to MTBE. Using the calculated 𝐾SL−SV , the
complexation isotherm of TBA bound to MTBE was obtained and is illustrated in Fig. 8.
Because the apparent enthalpy changes were obtained from the retention factor
data in the temperature range 25–40 °C, the determined values presented in Table 4
should be considered as the values at an average temperature of 32.5 °C. Nevertheless,
this is the first time, to our knowledge, that the complete thermodynamic parameters of
solute–modifier interactions were deduced entirely using HPLC data.

5. CONCLUSIONS
Van’t Hoff plots have been extensively used to investigate the thermodynamic
properties of adsorption equilibrium processes in various chromatography systems. When
the retention factor is measured over a certain temperature range, the plot of ln k versus
0
1/T often yields a straight line with a slope of −∆𝐻vH /𝑅. Although this method provides
information on adsorption enthalpy changes, the derived thermodynamic parameters are
 12

macroscopic lumped quantities that theoretically do not account for solute–solvent

complexation and solvent adsorption. In this study, a systematic investigation of the
relationship between apparent enthalpy changes determined from van’t Hoff plots and
modifier concentrations was performed using three simple achiral solutes—THF, AC,
and TBA—in an n-hexane-MTBE mobile phase with a commercially available amylose
CSP.
The apparent enthalpy changes of THF, AC, and TBA, derived from van’t Hoff
plots, were strongly dependent on MTBE concentration from 0 to 5 M; the enthalpic
curve changed rapidly with MTBE content but leveled off after ∆H0 reached
approximately −15 kJ/mol. These data could not be explained by the previous
thermodynamic model. A three-equilibrium-constant thermodynamic model for
monovalent solutes involving solute–sorbent, solvent–sorbent, and solute–solvent
interactions was derived herein. Two key dimensionless groups were found to control the
van’t Hoff enthalpic curves: the fraction of solute molecules that are bound to modifier
molecules and the fraction of adsorption sites that are occupied by modifier molecules.
To estimate the thermodynamic parameters, the model was used to numerically fit
the apparent enthalpy data. For THF and AC, solute–MTBE interaction was assumed to
be negligible. AC, THF, and MTBE were expected to bind mainly with sorbent NH
groups. The van’t Hoff enthalpic curves essentially represented the adsorption isotherm
of MTBE. Because the TBA molecules bonded mainly with sorbent C=O groups and did
not compete with MTBE for the NH binding sites, the enthalpic curve of TBA indicated
the dependence of the fraction of complexed TBA molecules. From the fitted parameters
of the enthalpic curves of TBA and of AC and THF, the complete thermodynamic
properties of MTBE adsorption and TBA–MTBE complexation were estimated,
respectively.
The results of this study on monovalent achiral solutes show that the adsorption
and complexation of the polar modifier in the mobile phase must be accounted for in the
van’t Hoff plots used for interpretation of the adsorption of a solute. The above model is
based on the assumptions that there are monovalent interactions of solute-modifier,
solute-sorbent, and modifier-sorbent, and that the binding sites are homogeneous for the
solutes and for modifiers. The thermodynamic model for chiral molecules would be more
complex, and would involve multiple binding sites for solute-sorbent and solute-modifier
interactions. The developed model would be needed to expand to include heterogeneous
binding of chiral solutes. For chiral solutes, one would need to use an isotherm different
from Langmuir’s, and multiple equilibrium constants for binding of chiral solutes with
the sorbent. Moreover, there would be multiple values for the sorbent capacity. Hence,
the model and assumptions may not be valid for most chiral solutes. The ultimate goal
and one motivation of this work, is the application of such models to chiral molecules, to
help understand quantitatively the mechanism of molecular recognition in the presence of
the mobile phase modifier.

Acknowledgments
This study was supported by the National Science Council of Taiwan (MOST 104-2218-
E-027-005-MY2).
 13

References
[1] P. Franco, T. Zhang, Common approaches for efficient method development with
immobilised polysaccharide-derived chiral stationary phases, J. Chromatogr. B. 875
(2008) 48–56.
[2] J.Y. Jin, S.K. Bae, W. Lee, Comparative studies between covalently immobilized
and coated chiral stationary phases based on polysaccharide derivatives for
enantiomer separation of N-protected alpha-amino acids and their ester derivatives,
Chirality. 21 (2009) 871–877.
[3] V.S. Sharp, M.A. Gokey, C.N. Wolfe, G.A. Rener, M.R. Cooper, High performance
liquid chromatographic enantioseparation development and analytical method
characterization of the carboxylate ester of evacetrapib using an immobilized chiral
stationary phase with a non-conventional eluent system, J. Chromatogr. A. 1416
(2015) 83–93.
[4] B. Chankvetadze, Recent developments on polysaccharide-based chiral stationary
phases for liquid-phase separation of enantiomers, J. Chromatogr. A. 1269 (2012)
26–51.
[5] X. Geng, F.E. Regnier, Retention model for proteins in reversed-phase liquid
chromatography, J. Chromatogr. A. 296 (1984) 15–30.
[6] L.R. Snyder, Role of the solvent in liquid-solid chromatography. Review, Anal
Chem. 46 (1974) 1384–1393.
[7] H.-W. Tsui, M.Y. Hwang, L. Ling, E.I. Franses, N.-H.L. Wang, Retention models
and interaction mechanisms of acetone and other carbonyl-containing molecules
with amylose tris[(S)-α-methylbenzylcarbamate] sorbent, J. Chromatogr. A. 1279
(2013) 36–48.
[8] H.-W. Tsui, E.I. Franses, N.-H.L. Wang, Effect of alcohol aggregation on the
retention factors of chiral solutes with an amylose-based sorbent: Modeling and
implications for the adsorption mechanism, J. Chromatogr. A. 1328 (2014) 52–65.
[9] M. Lämmerhofer, Chiral recognition by enantioselective liquid chromatography:
Mechanisms and modern chiral stationary phases, J. Chromatogr. A. 1217 (2010)
814–856.
[10] S.-G. Wu, A.-Y. Lin, H.-Y. Hsieh, H.-W. Tsui, Elucidation of adsorption
mechanisms of solvent molecules with distinct functional groups on amylose
tris(3,5-dimethylphenylcarbamate)-based sorbent, J. Chromatogr. A. 1460 (2016)
123–134.
[11] G.K.E. Scriba, Chiral recognition in separation science – an update, J. Chromatogr.
A. 1467 (2016) 56–78.
[12] H.-W. Tsui, N.-H.L. Wang, E.I. Franses, Chiral Recognition Mechanism of
Acyloin-Containing Chiral Solutes by Amylose Tris[(S)-α-methylbenzylcarbamate],
J. Phys. Chem. B. 117 (2013) 9203–9216.
[13] H.-Y. Hsieh, S.-G. Wu, H.-W. Tsui, Retention models and interaction mechanisms
of benzene and other aromatic molecules with an amylose-based sorbent, J.
Chromatogr. A. 1494 (2017) 55–64.
[14] S. Edward, Solvent composition effects in liquid-solid systems, J. Chromatogr. A.
130 (1977) 23–28.
[15] R.N. Rao, K. Santhakumar, C.G. Naidu, Rat dried blood spot analysis of (R,S)-(−)-
and (S,R)-(+)- enantiomers of emtricitabin on immobilized tris-(3,5-dimethylphenyl
 14

carbamate) amylose silica as a chiral stationary phase, J. Chromatogr. B. 1002 (2015)

160–168.
[16] S. Ma, H.-W. Tsui, E. Spinelli, C.A. Bussaca, E.I. Franses, N.-H.L. Wang, L. Wu, H.
Lee, C. Senanayake, N. Yee, N. Gonella, K. Fandrick, N. Grinberg, Insights into
chromatographic enantiomeric separation of allenes on cellulose carbamate
stationary phase, J. Chromatogr. A. 1362 (2014) 119–128.
[17] T. Fornstedt, P. Sajonz, G. Guiochon, Thermodynamic Study of an Unusual Chiral
Separation. Propranolol Enantiomers on an Immobilized Cellulase, J. Am. Chem.
Soc. 119 (1997) 1254–1264.
[18] P.C. Wankat, Separation Process Engineering: Includes Mass Transfer Analysis, 3
edition, Prentice Hall, Upper Saddle River, NJ, 2011.
[19] Y.-Y. Huang, The temperature dependence of isosteric heat of adsorption on the
heterogeneous surface, J. Catal. 25 (1972) 131–138.
[20] L.E. Drain, J.A. Morrison, Thermodynamic properties of nitrogen and oxygen
adsorbed on rutile, Trans. Faraday Soc. 49 (1953) 654–673.
[21] J.A. Morrison, J.M. Los, L.E. Drain, The heat capacity, integral heat of adsorption
and entropy of argon adsorbed on titanium dioxide, Trans. Faraday Soc. 47 (1951)
1023–1030.

Figure Captions
 15

Fig. 1. Molecular structure of (A) the polymer-repeating unit of the IE polymer, with R
being the side chain and (B) the side-chain model for the DFT calculation.
 16

Fig. 2. Van’t Hoff plots of ln k versus 1/T for the four solutes, using pure n-hexane as the
mobile phase; k is the retention factor.
 17

Fig. 3. HPLC results for elution profiles of THF with Chiralpak IE for 4, 20, and 50 vol.
% of MTBE in n-hexane at 25, 30, 35, and 40 °C.
 18

Fig. 4. Plots of van’t Hoff enthalpy changes versus MTBE concentrations and
thermodynamic model fits to the data for THF, AC, and TBA.
 19

Fig. 5. Fit of the thermodynamic model to the THF and AC van’t Hoff enthalpy data (see
also Eq. (19)).
 20

Fig. 6. Langmuir isotherm plots of MTBE adsorption using the estimated parameters
from Table 3.
 21

Fig. 7. Fit of the thermodynamic model to the TBA van’t Hoff enthalpy data (see also Eq.
(21)).
 22

Fig. 8. Fraction of TBA complexed with MTBE for various MTBE concentrations.
 23

Table 1 Retention factors and van’t Hoff thermodynamic parameters of four solutes using pure n-hexane as the mobile phase. 𝐶 o (1 M)
is the standard state concentration.

Temperature (°C) Van't Hoff plot

0
Solutes ∆𝑆SL /R+ln (φCSB/Co )
0
25 30 35 40 ∆𝐻SL (kJ/mol) (−) R2
Methyl tert-butyl ether 1.522 1.286 1.106 0.947 −24.46 ± 0.25 −9.45 ± 0.10 1.000
Tetrahydrofuran 7.279 6.138 5.058 3.988 −31.00 ± 1.99 −10.50 ± 0.78 0.992
Acetone 13.698 10.750 9.025 7.521 −30.66 ± 1.42 −9.77 ± 0.56 0.996
tert-Butanol 14.682 11.551 9.442 7.872 −32.18 ± 1.11 −10.31 ± 0.44 0.998
 24

Table 2 Fitted parameters in the thermodynamic model for the AC and THF van’t Hoff
enthalpy curves using MTBE as the polar modifier.

Thermodynamic parameters
R2 0.955
Slope (kJ-1mol M) 0.00476 ± 0.00046
Intercept (kJ-1mol) 0.04932 ± 0.00067
-1
KSV (M ) 10.40 ± 1.54
0
∆𝐻SV (kJ/mol) −20.28 ± 0.31a
−24.46 ± 0.25b
−23.72c

a
Estimated from the thermodynamic model developed in this study.
b
Estimated from the van’t Hoff plot using pure n-hexane as the mobile phase.
c
Predicted from DFT calculations.
 25

Table 3 Parameters in the thermodynamic analysis of the AC and THF van’t Hoff
enthalpy curves and the derived thermodynamic properties for MTBE adsorption.

0 0 0 0
T (oC) 𝐶𝑆𝐵 𝜑 (M) KSV (M-1) ∆𝐻SV (kJ/mol) ∆𝐺SV (kJ/mol) ∆𝑆SV (J/K∙mol)
25 13.27 ± 2.69 -6.41 ± 0.50 -46.52 ± 1.97
30 11.22 ± 2.27 -6.09 ± 0.51 -46.80 ± 1.97
0.115 ± 0.023 -20.28 ± 0.31
35 9.64 ± 1.95 -5.81 ± 0.52 -46.97 ± 1.96
40 8.26 ± 1.67 -5.50 ± 0.53 -47.21 ± 1.95
 26

Table 4 Fitted parameters in the thermodynamic model of the TBA van’t Hoff enthalpy curves using MTBE as the polar modifier and
the derived thermodynamic properties for TBA–MTBE complexation.

Thermodynamic parameters
R2 0.959
Slope (kJ-1mol M) 0.0111 ± 0.0008
Intercept (kJ-1mol) 0.0557 ± 0.0036
-1
KSL−SV (M ) 3.43 ± 0.54
0
∆𝐻SL−SV (kJ/mol) −19.81 ± 0.69a
−15.12b
0
∆𝐺SL−SV (kJ/mol) −3.13 ± 0.40
0
∆𝑆SL−SV (J/K∙mol) -54.57 ± 2.59

a
Estimated from the thermodynamic model developed in this study.
b
Predicted from DFT calculations.