Molecular Design and Engineering for Affinity Separations of Polar Systems using Isothermal Titration Calorimetry and Molecular Modeling

There is a continuous incentive to improve industrial processes, not only
from an economic and energy point of view, but also from the increasingly
important perspective of green chemistry and bio-based production. The
demand for the production of chemicals based on renewable feedstock and
the design for energy efficiency is therefore increasing. For diluted aqueous
streams ordinary distillation would require evaporation of large amounts of
water and is therefore energy intensive. Also for mixtures of components
with a low relative volatility ordinary distillation is energy intensive and
in the case of azeotropes this separation process may even be impossible.
Furthermore, some components that are applied in industry may even
decompose at elevated temperatures. In these cases, affinity separations
such as liquid-liquid extraction (LLX), extractive distillation (XD) and
adsorption based processes are promising alternatives, as they improve or
enable the separation by adding an affinity agent.

In affinity based separation processes there is a second column, next to the
initial separation column, in which the regeneration of the affinity agent
takes place. This is an important step for the overall feasibility of the
process. In the case of LLX and XD the solvent is (or contains) the affinity
agent and is regenerated in this second column, after which the solvent is
recycled to the first column. Generally there is a trade-off between the LLX
or XD step and the regeneration step in terms of the required energy input
and other costs.

This thesis focuses on two main subjects. The first subject is liquid-liquid
extraction of polar components such as acids from dilute aqueous streams and
the second main subject is the separation of close-boiling mixtures of polar
components by extractive distillation. LLX is already applied in industry for
separation of carboxylic acids, which are polar components that are often
present in highly diluted aqueous streams. Carboxylic acids are building
blocks in the production of many bio-based plastics and other chemicals in
the pharmaceutical, food and chemical industry. As a consequence there is a
strong desire to produce these acids through fermentation, after which they
are present in (dilute) aqueous streams and separation processes are required.
Also in other diluted industrial (waste) streams a variety of valuable polar
components or contaminants is present for which separation processes are
required. Non-aqueous streams of polar components are studied for the second
main subject of this thesis. Also for these components industrial application
of affinity based separation can be of interest. For example in the case of very
low relative volatility or azeotrope formation, XD is a promising technique
instead of ordinary distillation.

Both LLX and XD are solvent based processes. There is a widespread
interest in solvent selection criteria and methods. Initial selection can be
based either on practical considerations, one of the several solvent scales
that have been developed, empirical data and predictions or calculations
based on activity coefficients. However, prediction of solvent effects on the
relative volatility or acid distribution is a challenging task, especially for
the (close-boiling) polar mixtures that are central to this thesis and in
which specific and strong interactions, azeotropes and non-ideality are all
common. Moreover, the feasibility of the regeneration should always be
taken into account, for which more research and data may be required than
only the solvent effects in the initial separation step.

Ideally, solvent selection should be simple, time-efficient and include
considerations on the regeneration. The development of such a procedure
requires fundamental understanding of the mechanisms and phenomena
involved in LLX and XD. The use of isothermal titration calorimetry (ITC),
and molecular modeling (MM) to characterize mechanisms, interactions and
the effect of extractant and solvent composition on liquid-liquid equilibria
(LLE) and vapor-liquid equilibria (VLE) were studied in this thesis. The
thesis consists of two parts that each deal with affinity separations of closeboiling
polar compounds. In the first part (Chapters 4-7) the focus is on
LLX of acids and other polar compounds from aqueous solutions, while
the second part (Chapters 8-9) focuses on extractive distillation processes
of non-aqueous polar mixtures. The methods applied (ITC and MM) are
introduced in the Chapters 2 and 3.

The study on the use of ITC to analyze acid-base interactions relevant
for LLX is described in Chapter 2. The concentrations applied in LLX are
up to a factor thousand higher than those applied in the biotechnology
field where ITC is commonly applied. As complexation constants appeared
to be extractant concentration dependent, a thorough analysis of the
accuracy of ITC applied in the molar concentration domain applicable to
LLX, is presented. Standard deviations were determined for the calculated
thermodynamic parameters. Two different reaction models, based on either
a single reaction or sequential reactions were compared for their applicability
in acid-base interactions in LLX. The sequential reaction model is most
suitable to fit ITC-results for acid-base titration in the system of acetic
acid and trioctylamine (TOA) in toluene. Fitting of the sequential reaction
model was found to be sensitive to the initial values as local minima do occur
as a result of the number of parameters, indicating the importance of good
initial guess values. Chapter 3 describes the method and calculation options
of molecular modeling with Spartan’16 Parallel software. Furthermore, it
includes a validation of the calculation levels applied and the assumptions
related to molecular geometry and the incorporation of water molecules in
the molecular modeling calculations.

Chapter 4 is the first chapter in the part of this thesis focusing on LLX and
reviews solvent developments for LLX of carboxylic acids for three main
categories of solvents, i.e. nitrogen-based extractants, phosphorous based
extractants and ionic liquids. The mechanism of extraction is influenced
by multiple variables, including extractant type, solvent composition and
specific interactions with the diluent. Combining both active and inactive
diluents was shown to be a promising approach. No single parameter can
describe all solvent effects on the LLE. Based on a comparison of different
regeneration strategies for a typical LLX process, a diluent-swing based
process appeared to be the most feasible. For a viable process, focusing on
the change in acid distribution that can be achieved between the extraction
and regeneration step is more important than the absolute distribution ratio
itself. For temperature-swing regeneration processes this change can be
related to the enthalpy of complex formation between acid and extractant.
Therefore a study on the effect of extractant structure and composition on
the enthalpy of complexation (Chapter 6) can strongly contribute to design
of solvents.

Chapter 5 and 6 each focus on the understanding of mechanisms involved
in LLX and study implications thereof on the molecular structure of good
extractants. For these studies ITC was applied - supported by MM - on
various acid-base interactions directly related to LLX. Chapter 5 focuses
on the role of the diluent in the extraction mechanism for complexation
of acetic acid and phenol with extractants TOA, trioctylphosphine oxide
(TOPO) and tributylphosphate (TBP) in various diluents. With the help of
ITC it was shown that increasing the temperature decreases all complexation
constants and - supported by MM - that the diluent affects the mechanism
of complexation, e.g. in the case of protic diluents there is competition
between the acid and diluent, whereas inactive diluents resulted in more
overloading of the amine. The method of ITC was validated with LLE
experiments and the equilibrium model based on ITC data can also be used
to directly predict LLE. The differences between phenol and acetic acid that
were observed in the LLE experiments, and expected on the basis of ITC
and MM, could be described using the BF3-affinity scale for acetic acid and
the hydrogen bond basicity pKBHX scale for phenol.

The effect of molecular structure of the extractant on the enthalpy of
complexation and thereby the temperature sensitivity of the complexation
reaction is the focus point of Chapter 6. ITC and MM were applied to
two cases, i.e. interaction of 4-cyanopyridine with (substituted) phenols
and interaction of acetic acid with tertiary amines. A form of Entropy-
Enthalpy-Compensation (EEC) was shown for extractants with varying
side groups, which means that (small) changes in molecular structure hardly
affect the complexation constant, but affect both the enthalpy and entropy
of complexation. As a consequence, through optimizing the enthalpy of
complexation, the temperature dependency of the equilibrium can be
increased. The enthalpy of complexation itself appeared to increase with
temperature, an effect that was stronger for tertiary amines with longer
carbon chains compared to aminopyridines. For a typical system in LLX
there is an optimum in Gibbs energy of complexation for every value of the
enthalpy of complexation. Chapter 5 and 6 show that combination of ITC
with MM is a strong approach to study the thermodynamics of interactions
in LLX processes, by improving the understanding of the interaction
mechanism and the effect of extractant structure thereon.

Swing processes for solvent regeneration in LLX processes of succinic
acid based on temperature and diluent-swing were compared in the study
described in Chapter 7. Diluent-swing can be performed by either adding an
anti-solvent or evaporating a part of the (active) diluent. Solvents composed
of trioctylamine (TOA) in methyl isobutyl ketone (MIBK) showed the
largest swing in acid distribution. The application of a gaseous anti-solvent
has a high potential as regeneration can be performed by reducing the
pressure, although extra compressors and pumps are required. With the
gaseous ethane a swing in acid distribution was obtained similar to the swing
with the liquid pentane. An evaluation on required investments and energy
input performed using Aspen Plus showed that the ethane-based process is
the most profitable process as the energy required is only 21% (13 MW/kg
pure acid product) of what would be required for evaporation of water with
a four-stage evaporator. The return of the increased capital investments
required for this process requires 110 days of production. Combining both
diluent-swing and temperature-swing was also beneficial as it results in
higher product concentrations.

In XD of close-boiling polar compounds, strong interactions or azeotropes
are common and prediction of VLE data is thus not always straightforward.
Chapter 8 describes the study with the aim to find heuristics for a first
selection of solvents for XD, based on three different mixtures of close-boiling
polar compounds, i.e. a mixture of diethylmethylamine and diisopropylether,
a mixture of valeric acid and its isomer 2-methylbutyric acid, and a mixture
of 2-butanol and 2-butanone. Solvent effects were measured for solvents
selected based on e.g. acidity, structural similarity, steric hindrance, polarity
and hydrogen bonding affinity. For mixtures of very similar components
in terms of acidity, boiling point and structure, stronger interaction with
the solvent is required, although this is limited by the chemical stability.
Therefore, solvents can be initially selected based on ITC or MM. This was
further implemented in Chapter 9 in which a procedure was developed for
solvent screening, combining MM and ITC to predict solvent effects on the
relative volatility. The interaction energies of the mixture components and
solvents were calculated with MM and the heat of mixing was measured with
ITC and related to the solvent effect on the relative volatility. Guidelines for
desired values of these energies and heats were defined and experimentally
validated based on three cases; a) a mixture of octanoic and levulinic acid, b)
the amine-ether mixture of Chapter 8, and c) the alcohol-ketone mixture of
Chapter 8. With this solvent selection procedure the number of solvents, for
which extensive experimental screening on the basis of VLE measurements
is required, is strongly reduced.

Chapter 10 concludes this thesis by evaluating and comparing the results
obtained for both LLX and XD systems. Moderate interactions are favored
for both processes to allow for successful solvent regeneration and to avoid
chemical or thermal instability. For LLX processes, affinity agents that
form stronger interacting complexes with the mixture components based on
proton transfer were successfully applied, whereas this type of interaction
appeared too strong for XD processes where only interactions up to the
strength of hydrogen bonding could successfully be applied. In an outlook
regarding the international climate goals for 2050 opportunities for affinity
fluid separations are discussed and directions for further research and
applications are given.