Citation
Optimization based design and analysis of tailor-made ionic liquids

Material Information

Title:
Optimization based design and analysis of tailor-made ionic liquids
Creator:
Mehrkesh, Amirhossein ( author )
Language:
English
Physical Description:
1 electronic file (242 pages). : ;

Subjects

Subjects / Keywords:
Ionic solutions ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Abstract:
Solvents comprise two thirds of all industrial emissions. Traditional organic solvents easily reach the atmosphere as they have high vapor pressure and are linked to a host of negative environmental effects including climate change, urban air-quality and human illness. Room temperature ionic liquids (RTIL), on the other hand, have low vapor pressure and are not flammable or explosive, thereby resulting in fewer environmental burdens and health hazards. However, their life cycle environmental impacts as well as freshwater ecotoxicity implications are poorly understood. RTILs are molten salts that exist as liquids at relatively low temperatures and have unique properties. Ionic liquids consist of a large organic cation and charge-delocalized inorganic or organic anion of smaller size and asymmetric shape. The organic cation can undergo unlimited structural variations through modification of the alkyl groups attached to the side chain of the base cation skeleton and most of these structural variations are feasible, from chemical synthesis point of view, due to the easy nature of preparation of their components. Functionally, ionic liquids can be tuned to impart specific desired properties by switching anions/cations or by incorporating functionalities into the cations/anions. It is estimated that theoretically more than a trillion ionic liquid structures can be formed. Due to their tunable nature, these molten salts have the potential to be used as solvents for variety of applications. ( ,,,, )
Abstract:
This dissertation presents a computer aided IL design (CAILD) methodology with an aim to design optimal task-specific ionic liquid structures for different applications. We utilize group-contribution based ionic liquid property prediction models within a mathematical programming framework to reverse engineer functional ionic liquid structures. The CAILD model is then utilized to design optimal ionic liquids for solar energy storage, as a solvent for aromatic-aliphatic separation, and as an absorbent for carbon capture process. Using the developed CAILD model, we were able to computationally design new ionic liquid structures with physical and solvent properties that are potentially superior to commonly used ILs. The accuracy of the developed model was back tested and verified using available experimental data of common ILs. However, we would like to note that the computational design results from this dissertation needs to be experimentally validated.
Abstract:
This dissertation also developed ecotoxicity characterization factors for few common ILs. The developed characterization factors (CFs), can be used in future studies to perform holistic (cradle-to-grave) life cycle assessments on processes using ILs to understand their environmental and ecological impacts.
Thesis:
Thesis (Ph.D.)--UNiversity of Colorado Denver.
Bibliography:
Includes bibliographic references
System Details:
System requirements: Adobe Reader.
General Note:
Department of Civil Engineering
Statement of Responsibility:
by Amirhossein Mehrkesh.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
951978439 ( OCLC )
ocn951978439

Downloads

This item has the following downloads:


Full Text
OPTIMIZATION-BASED DESIGN AND ANALYSIS OF TAILOR-MADE IONIC
LIQUIDS
By
AMIRHOSSEIN MEHRKESH
B.S. Isfahan University of Technology, 2006
M.S. University of Isfahan, 2009
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Civil Engineering
2015


This thesis for the Doctor of Philosophy degree by
Amirhossein Mehrkesh
has been approved for the
Civil Engineering Program
by
Arunprakash T. Karunanithi, Advisor
Kannan Premnath, Chair
Azadeh Bolhari
Indrani Pal
Fernando Rosario-Ortiz
21 November 2015


Mehrkesh, Amirhossein (PhD, Civil Engineering)
Optimization-based Design and Analysis of Tailor-made Ionic Liquids
Thesis directed by Associate Professor Arunprakash T. Karunanithi
ABSTRACT
Solvents comprise two thirds of all industrial emissions. Traditional organic solvents easily
reach the atmosphere as they have high vapor pressure and are linked to a host of negative
environmental effects including climate change, urban air-quality and human illness. Room
temperature ionic liquids (RTIL), on the other hand, have low vapor pressure and are not
flammable or explosive, thereby resulting in fewer environmental burdens and health
hazards. However, their life cycle environmental impacts as well as freshwater ecotoxicity
implications are poorly understood. RTILs are molten salts that exist as liquids at relatively
low temperatures and have unique properties. Ionic liquids consist of a large organic cation
and charge-delocalized inorganic or organic anion of smaller size and asymmetric shape. The
organic cation can undergo unlimited structural variations through modification of the alkyl
groups attached to the side chain of the base cation skeleton and most of these structural
variations are feasible, from chemical synthesis point of view, due to the easy nature of
preparation of their components. Functionally, ionic liquids can be tuned to impart specific
desired properties by switching anions/cations or by incorporating functionalities into the
cations/anions. It is estimated that theoretically more than a trillion ionic liquid structures can
be formed. Due to their tunable nature, these molten salts have the potential to be used as
solvents for variety of applications.
This dissertation presents a computer aided IL design (CAILD) methodology with an aim to
design optimal task-specific ionic liquid structures for different applications. We utilize
group-contribution based ionic liquid property prediction models within a mathematical


programming framework to reverse engineer functional ionic liquid structures. The CAILD
model is then utilized to design optimal ionic liquids for solar energy storage, as a solvent for
aromatic-aliphatic separation, and as an absorbent for carbon capture process. Using the
developed CAILD model, we were able to computationally design new ionic liquid structures
with physical and solvent properties that are potentially superior to commonly used ILs. The
accuracy of the developed model was back tested and verified using available experimental
data of common ILs. However, we would like to note that the computational design results
from this dissertation needs to be experimentally validated.
This dissertation also developed ecotoxicity characterization factors for few common ILs.
The developed characterization factors (CFs), can be used in future studies to perform
holistic (cradle-to-grave) life cycle assessments on processes using ILs to understand their
environmental and ecological impacts.
The form of and content of this abstract are approved. I recommend its publication.
Approved: Arunprakash Karunanithi
IV


DEDICATION
This dissertation is dedicated to my brilliant and outrageously loving and supporting wife,
Anna and to my always encouraging, ever faithful mother, Soraya, and to the memory of my
late father, Eskandar, who taught me how to live a peaceful and happy life, the person who
will be missed forever.
v


ACKNOWLEDMENTS
I am grateful to have had Dr. Arunprakash Karunanithi as my advisor. Without his
knowledge, guidance, support, and enthusiasm towards this research, I would not have been
able to complete this dissertation. He has taught me to be optimistic, persistent and confident
in the work I am doing. He provided me all the tools needed to accomplish this research from
financial support, to computer software to grants for attending conferences.
I also would like to thank the faculty, staff and my friends (fellow graduate students in our
research group) whom I interacted with during my graduate program at the University of
Colorado Denver.
I also want to extend my acknowledgment to Dr. Azadeh Bolhari, Dr. Mike Tang and Mr.
Eric Ziegler who helped me in editing and proofreading. I also would like to thank all my
committee members for their constructive suggestions and feedback.
vi


TABLE OF CONTENTS
Chapter
1 Introduction.......................................................................1
1.1 Computer-aided ionic liquid design (CAILD)......................................5
1.2 Current applications............................................................9
1.3 Environmental impacts of ionic liquids.........................................10
1.4 Ionic liquids safety...........................................................11
2 Forward Problem, Prediction of Melting Point and Viscosity of ILs..................13
2.1 Introduction...................................................................13
2.2 Methods........................................................................16
2.3 Results and discussion.........................................................19
2.3.1 Melting point..............................................................21
2.3.2 Viscosity..................................................................24
3 Reverse Problem; Computer-aided Design of Ionic Liquids............................27
3.1 Introduction...................................................................27
3.2 Computer-aided ionic liquid design (CAILD).....................................28
3.2.1 Mathematical framework.....................................................31
3.2.2 Ionic liquid structural constraints........................................32
3.2.3 Physical property constraints..............................................38
3.2.4 Solution property constraints.............................................39
3.2.5 Solution of the underlying MINLP...........................................42
3.3 Proof of concept examples......................................................45
3.3.1 Electrolytes...............................................................46
3.3.2 Heat transfer fluids.......................................................54
3.3.3 Toluene-heptane separati on................................................59
3.3.4 Naphthalene solubility.....................................................65
4 Application 1: Design of Ionic Liquids for Thermal Energy Storage..................70
4.1 Introducti on..................................................................70
4.2 Formulation of the design problem..............................................73
4.2.1 Ionic liquid structural constraints........................................76
4.2.2 Ionic liquid property prediction...........................................79
4.2.3 CAILD model solution.......................................................83
4.3 Results and discussion.........................................................86
vii


5 Application 2: Design of Ionic Liquids for Aromatic-Aliphatic Separation.......96
5.1 Introduction.................................................................96
5.2 Computer-aided ionic liquid design (CAILD)...................................98
5.2.1 Forward problem.........................................................99
5.2.2 Reverse problem........................................................105
5.3 Results.....................................................................110
6 Application 3: Design of Ionic Liquids for CO2 Capture.........................115
6.1 Introduction................................................................115
6.2 Forward problem.............................................................116
6.3 Reverse problem.............................................................121
6.3.1 Case study.............................................................121
6.4 Results.....................................................................124
7 Life Cycle Environmental Implications of Ionic Liquids..........................128
7.1 Life Cycle Perspectives on Aquatic Ecotoxicity of Common Ionic Liquids.....128
7.2 Methods.....................................................................131
7.2.1 Goal, Scope, System Boundary...........................................131
7.2.2 Life Cycle Inventory of Ionic Liquid Production and Data Sources.......131
7.2.3 Fresh Water Ecotoxicity Impacts of Ionic Liquid Production.............132
7.2.4 Development of Ecotoxicity Characterization Factors for Ionic Liquids.133
7.2.5 Fresh Water Ecotoxicity Impact of Direct Release of Ionic Liquids......139
7.2.6 Uncertainty............................................................140
7.3 Results and Discussion......................................................141
7.4 Life cycle assessment of energetic ionic liquids............................148
7.4.1 Process and Energetic requirements for triazolium nitrate and TNT synthesis 152
7.4.2 Life cycle assessment (LCA) of energetic ionic salts...................157
7.4.3 Results and discussion.................................................160
8 Conclusions and Future Work.....................................................169
8.1 Limitations and Recommendations.............................................172
8.2 Contributions...............................................................175
8.3 Future work on ionic liquid applications....................................177
References.........................................................................182
Appendix
A. A comprehensive list of IL structural groups used in CAILD....................208
B. IL structural Groups used in CAILD for Aro/Ali separation.....................211
viii


C. UNIFAC parameters.................................................................212
D. IL structural Groups used in CAILD for CO2 capture................................214
E. Life Cycle Inventory of Ionic Liquids Production..................................215
ix


LIST OF TABLES
Table
2-1. Model characteristics in prediction of melting points (calculated on validation
data...................................................................23
2- 2. Model characteristics in prediction of viscosity (calculated on validation data)
.......................................................................26
3- 1. Cation and alkyl side chain groups valences..............................35
3-2. Values of yl and ngkl for 1,3-diethylimidazolium tetrafluoroborate......36
3-3. The basis set used for ionic liquid design...............................45
3-4. Group contributions for parameters Ax and B),............................47
3-5. Group contributions for parameters A, B and V............................48
3-6. ac values for different cations..........................................49
3-7. Group Contributions for Ionic Liquids Melting Point.....................49
3-8. Decomposition approach: Subproblem Results...............................51
3-9. Design Results of the optimal IL, 1-methylimidazolium [Tf2N].............51
3-10. Experimentally Measured Electrical Conductivities of Ionic Liquids......53
3-11. Group contributions for parameters Ak and Bk.............................54
3-12. Decomposition approach: Subproblem Results...............................56
3-13. Design Results of the Optimal IL, 1-ethyl-3-methylimidazolium [BF4]......56
3-14. Experimental thermal conductivity data...................................58
x


3-15. Decomposition Approach: Subproblem Results.............................61
3-16. Design Results of the Optimal IL, 1-methylpyridinium [BF4]'............62
3-17. Experimentally Measured Selectivity values for Aromatic/Aliphatic
Separations...........................................................64
3-18. Decomposition Approach: Subproblem Results.............................66
3- 19. Physical properties of 1-butyl-3-ethylimidazolium [TfzN]'..............67
4- 1. Ionic liquid building blocks (groups) considered for thermal fluid design .... 74
4-2. Thermo-physical properties of VP-1, Hitec XL and [3-hydroxy-
Imidazolium]+ [BF4]'..................................................87
4-3. Effect of anion variation on the thermal storage properties of ionic liquids .. 88
4-4. Effect of number of CH2 groups on q, Tm and Tapp [ILs with BF4' anion].90
4-5. The effect of variation of functional groups (FG) connected to the
cation head group on q, Tm and Tapp...................................92
4-6. Comparison of COSMO-predicted Cp and p of selected ILs with the
corresponding exp. data...............................................94
4- 7. Comparison between COSMO-predicted and GC-predicted values for Cp and
p, and q of the optimal IL............................................95
5- 1. Experimental infinite dilution activity coefficients (yoo) vs. Cosmo-
predicted values ....................................................103
5-2. Experimental solubility data vs. CAILD predicted data.................104
5-3. CAILD results for the optimal ILs and furfural at T=330 K.............Ill
5-4. Physical properties and solvency power of optimal ILs and furfural....112
5-5. A schematic of the structure of optimal ILs...........................113
xi


6-1. A comparison between experimental and COSMO-based values of Henrys
constant.................................................................119
6-2. Experimental and UNIFAC predicted values of CO2 solubility in different ILs
120
6-3. Name, symbol and structure of the optimal ILs.............................125
6- 4. Pure (physical) and mixture properties of the optimal ILs.................126
7- 1. Toxicity values of selected Ionic Liquids.................................135
7-2. Environmental Properties of the studied Ionic Liquids.....................138
7-3. USEtox based effect factors, fate factors, exposure factors, and
characterization factors for different ionic liquids.....................141
7-4. Breakdown of energy and material related ecotoxicity impacts of IL........143
7-5. Breakdown of freshwater ecotoxicity impacts of ILs associated with use phase
release..................................................................146
7-6. Impact of ionic salt and TNT (functional unit: 1 MJ energy content).......161
7-7. Sensitivity analysis......................................................165
7-8. Environmental Impact for scenario 1.......................................167
7-9. Environmental Impact for scenario 2.......................................167
xii


LIST OF FIGURES
Figure
1- 1. A schematic of an ionic liquid, [Bmim] BF4'................................2
2- 1. A correlation between van der Waals (VdW) and experimental Radii..........20
2-2. Actual vs. model predicted melting points for training and test data sets.22
2-3. Goodness of the model for predicting melting points of the selected ionic
liquids.................................................................23
2-4. Actual vs. model predicted viscosities for training and test data sets....25
2- 5. Goodness of the model for predicting viscosities of the selected ionic liquids
.........................................................................25
3 -1. CAILD framework...........................................................29
3- 2. Conceptual description of CAILD...........................................30
3-3. A general Scheme of two feasible cation structures........................37
3-4. Examples of feasible and non-feasible Ionic liquids.......................38
3-5. Different types of group interactions involved in solutions containing ionic. 39
3-6. 1-methylimidazolium [TfzN]"...............................................51
3-7. l-ethyl-3-methylimidazolium [BF4]'........................................56
3-8. 1-methylpyridinium [BF4]'.................................................62
3- 9. 1-butyl-3-ethylimidazolium [TfzN]'........................................67
4- 1. A schematic of the structure of optimal IL with highest thermal storage
capacity.................................................................86
xiii


5- 1. A schematic of the single stage extraction process.....................106
6- 1. A schematic of single stages CO2 absorption-desorption processes..122
7- 1. Ecotoxicity impacts related to production and use phase release of ILs.......
142
7-2. a) 1,2,3 triaolzium nitrate; b) TNT.....................151
7-3. Material and energy flows associated with the life cycle tree for producing the
ionic salt 1,2,3 triazolium nitrate......................................156
7-4. Comparison of scaled impacts of ionic salt and TNT (functional unit of 1MJ
energy content): GWP (Global Warming Potential), AP (Acidification
Potential), EP (Eutrophication Potential), HH (Human Health).............163
7-5. Sensitivity analysis of the scaled impacts of ionic salt and TNT (functional
unit of 1 MJ): GWP (Global Warming Potential), AP (Acidification
Potential), EP (Eutrophication Potential), HH (Human Health).............165
xiv


Chapter 1: Introduction
Ionics liquids (ILs) are an emerging new class of chemicals that show tremendous promise in
creating customized designer compounds (solvents, electrolytes, energy storage media ...),
which can be used for new applications or to replace current materials that lack flexibility or
dont meet ecological safety concerns.1 Ionic liquids are normally comprised of a large
organic cation with positive charge and a charge-delocalized organic or inorganic anion of
smaller size (can be monoatomic such as Cl) and asymmetrical shape. The molecules
possess a strong positive and negative charge which lends to its name as an ionic liquid. First
ionic liquid triethylammonium nitrate was discovered more than a century ago.
Compared to the case of naturally occurring ionic salts (e.g. Na+Cf), the larger size of cations
and anions in ionic liquids will result in distribution of a small charge (+1 or -1) over a much
larger surface area. This fact, along with the asymmetric nature of cations and anions, explain
the lower melting points of ionic liquids. Ionic liquids (ILs) are salts that normally melt at
100C or less.145 A schematic of an ionic liquid, 1-Butyl-3-methylimidazolium
tetrafluoroborate, [Bmim] BF4' is shown in Figure 1-1. The cation head group, imidazolium,
with positive charge (+1) is shown in green, side chain groups attached to the cation-base,
butyl and methyl are shown in yellow and anion with negative charge (-1) is shown in brown.
1


head group
H,
3
X X
anion
side chain(s)
Figure 1-1: A schematic of an ionic liquid, [Bmim]+ BF4"
Interest in ionic liquids has continued to build in the academia and industry due to their
interesting tunable properties and potential to provide environmentally friendly alternative to
volatile organic compounds (VOCs) currently used in chemical/industrial processes. The
properties of ionic liquids (pure physical properties such as viscosity and mixture properties
such as solvency power) vary enormously as a function of their molecular structure, i.e. the
type of cation-base, anion and number/type of side chain alkyl/functional groups present in
the structure. ILs also offer a wide window of liquid state making them attractive as liquid
solvents since they normally have high boiling points and very low vapor pressures. ILs are
rarely flammable or explosive, thereby presenting fewer environmental burdens and health
hazards.
During the past few years, considerable effort has been devoted to identifying and
understanding ILs that have superior properties. The desirable properties of ionic liquids
include:
2


Negligible vapor pressure
Ability to dissolve organic, inorganic, and polymeric materials
High thermal stability (i.e. they do not decompose over a large temperature range)
Versatile and customizable for task specific applications
Nonvolatile and rarely flammable or explosive
Strong regeneration properties that allow for their reuse and recycle
Room temperature ionic liquids (RTILs) can be fluid at temperatures as low as -96C
Liquid phase temperature range from -96C to 300C; thermally stable up to 200C
Moderate to high electrical conductivity
These properties of ionic liquids (ILs) make them attractive as potential alternatives to
current chemical compounds. Ionic liquids can be tailor-made for different applications (task-
specific ionic liquids) by varying the building blocks of ionic liquids (i.e. cations, anions or
groups attached to the cation-base). Thus Ionic liquids present a fantastic opportunity to step
away from the status quo of utilizing volatile organic compounds, which are caustic to the
environment, as solvents in current chemical processes. As the continued push towards
environmentally conscious decisions at all level of industry continues, ionic liquids have the
necessary properties and customizability to deliver better alternatives with reduced
environmental impacts.
3


Ionic liquids can easily be removed from water. The ability to separate solvents from water is
a critical property as industrial water pollution is increasingly becoming a major issue.
Tolerance towards ecologically unsound industrial practices are diminishing on a global scale
as the anthropogenic effects on the planet becomes more evident every year.
Ionic liquids have the potential to reduce the overall cradle to grave environmental/ecological
impacts of current processes by offsetting upstream pollutant release from energy use and by-
products which are manufactured through the VOC creation process. Ionic liquids have the
potential of lower unusable by-products to designed product ratio. Their regenerative
properties further their green profile as their potential for reuse and recycle exceeds those of
currently used VOCs.
As regulatory oversight of emission of chemicals released into water and the atmosphere
continues to tighten, finding alternatives will reduce the economic burden on industry.
Industry has a strong concern towards negative externalities that result from their economic
activity. In addition, adoption of environmentally friendly technologies can lead to greater
acceptance from general public and can reduce the opportunity of public outcry or protest. It
is far superior to develop alternatives now, than to wait for a forced decision.
Design of alternate ionic liquids can be made easy if we know how different structural
groups present in them will influence the properties of interest. For example, if we are
interested in an ionic liquid with high solvency power towards a specific chemical, we need
to know which cations or anions contribute to higher values of the solvency power towards
that compound. After identifying the best cations and anions, addition of functional/alkyl
4


groups to the side chain of the cation-base can further help us fine tune the desirable
properties (e.g. relatively lower values of melting point or low viscosity). In the next section,
we discuss how computer models can help us expedite the process of selecting optimal ionic
liquids for different applications. A computer-based ionic liquid design model can reduce the
enormous number of experiments needed to find the optimal candidate thereby saving time
and money.
1.1 Computer-aided ionic liquid design (CAILD)
The vast number of combinations of ionic liquids (ILs) is what provides their versatility and
customization properties (estimated to be as many as 1014 ionic liquids feasible). Ionic liquids
are still considered as a new generation of chemicals, which are garnering attention from
academia and industry. Therefore, there is limited information on the properties of less
common ionic liquids in chemical libraries and databases. Without the necessary
information, random synthesis of ionic liquids and testing of their properties is costly and
time consuming. Computer-aided molecular design (CAMD) is a promising approach that
has been used for molecular systems to design compounds (e.g. solvents) for a variety of
applications.6'12 CAMD method integrates property prediction models and optimization
algorithms to reverse engineer molecular structures with unique properties of interest. Due to
the fact that ionic liquids are made of replaceable building blocks (structural groups), we
believe that a similar approach is even more relevant for designing tailor-made ionic
liquids.13
5


Use of computer-aided models to design optimal compounds, reduces the costs and allows
engineers to model a multitude of potential candidates for a specific application. We propose
that CAMD approach can be adapted specifically towards ionic liquid design [Computer-
aided Ionic Liquid Design (CAILD)] where we take into account cations, anions and
functional groups attached to cation core.13'16 In this dissertation, we successfully show that
the CAILD model is capable of creating ionic liquids with optimal desired properties for
different applications (e.g. an ionic liquid with high thermal storage capacity (rj = pCp) can
be a good candidate for a solar thermal storage process).
Based on the above discussion, it is clear that in order to find optimal ionic liquids for
different applications using a computer-aided design framework, we need to know how
different structural groups in an ionic liquid will contribute towards the properties of interest.
For example, lets consider a situation where we want to design a good solvent to remove
toluene from a multicomponent mixture. If we know that imidazolium cation-base usually
results in higher values of solvency power towards toluene, compared to the other cation-
bases, then an ionic liquid with imidazolium cation should always be chosen as the optimal
ionic liquid unless it violates other physical properties or process constraints (such as the
selected ionic liquid has an unacceptably high melting point or viscosity values).
In order for us to be able to design an optimal ionic liquid for an application of interest, we
need to make sure that first it is a theoretically feasible chemical structure (chemical
feasibility constraints) and secondly the designed ionic liquid meets other process criteria
necessary for it to be used in large industrial scales. Therefore, the optimization framework
6


needs an objective function (the value of the property of interest) that needs to be minimized
or maximized along with a set of constraints that should be satisfied to guaranty that the
designed ionic liquid is a feasible candidate. A good example of the type of constraints
needed in a CAILD model relates to the design of an ionic liquid which is liquid at room
temperature. Here a constraint of Tm<25 C should be enforced within the optimization
model.
Based on the above discussion, it is clear that we need models capable of predicting different
properties of ionic liquids based on the type and number of structural groups present in
them. These models commonly referred to as group contribution (GC) models have been
-to some extent- developed for ionic liquids. The GC models can be used within a CAILD
framework to enable prediction of physical properties of the ionic liquids during the design
process. Without comprehensive group contribution models capable of predicting different
properties of ionic liquids, it is not possible to utilize the power of CAILD models to their
full extent. In other words, CAILD models are most useful when they have the capability of
exploring all possible combinations of ILs towards finding the optimal candidate for a given
application and this is not possible unless we have group contribution models covering all
cations, anions and side chain groups.
Therefore, when group contribution models for certain properties or contribution parameters
for certain structural groups are not available one needs to develop these from scratch. This is
where unavailability of experimental data on the properties of interest could be problematic
7


since accurate group contribution models cannot be developed without sufficient amount of
experimental data.
It would be important to point out that for certain cases first order group contribution models
for of ionic liquids are not able to predict certain properties accurately. The first order group
contribution models simply consider one value for the contribution of a particular group
irrespective of where that group is located within the ionic liquid (e.g. they do not distinguish
between a CH2 group directly attached to the aromatic carbon and a CH2 group attached to
other aliphatic side chain groups). In situations where group contribution models do not work
properly, other approaches such as computational chemistry based correlative models or
Quantitative Structure Property Models can be utilized to predict the physical properties of
ionic liquids.
In this research, we utilized COSMO-RS (Conductor like Screening MOdel
for Real Solvents), a quantum chemistry-based equilibrium thermodynamics model with the
purpose of predicting the chemical potentials of compounds in the liquid phase, to predict
pure properties (e.g. melting point or viscosity) or mixture properties (e.g. activity
coefficients and solubility) of ionic liquids. When an optimal ionic liquid is designed for a
specific application using the CAILD model predictions based on COSMO-RS model can
serve to validate the results and show us whether the design solution is suitable for use in
large industrial scales.
Further, computational chemistry models can provide a strong foundation of information to
build libraries of data to draw upon for future research and development. Modeling reduces
8


the overall cost by eliminating ionic liquids whose properties do not meet the desired
application.
1.2 Current applications
During the past few years, ILs have been studied for variety of applications. Some of them
are listed below:2,525-30
Battery Technologies
Advanced fuel cell concepts
Dye sensitized solar cells (DSSC)
Thermo-electrical cells
Supercapacitors
Hydrogen generation through water splitting
Carbon (CO2) capture
Nuclear fuel processing
Solar (thermal) energy storage
BASFs commercial investigation of ILs reveals that the compounds have strong potential as
solvents that can provide efficiency improvements in a several applications including:
9


in chemical reactions and separation processes
as hydraulic fluid and lubricant
as polymer additives (antistatic)
in metal deposition processes
in dissolving and processing cellulose
as electrolytes in electronic devices
1.3 Environmental impacts of ionic liquids
The non-volatile nature of ionic liquids greatly limits the impact on air quality by reducing or
completely eliminating their direct emissions to the atmosphere. For this reason ionic liquids
are often considered as inherently green/environmentally benign solvents with the potential
to completely replace traditional volatile organic solvents in several applications. However
toxicological studies have shown that some ionic liquids are very toxic towards freshwater
organisms or human cell lines,31'34 but due to their immense variety, ionic liquids can be
designed/tuned to be environmentally benign. In order to conclude that ionic liquids (ILs)
are benign alternatives to molecular solvents, their environmental impacts need to be
analyzed in a holistic manner. Life cycle assessment (LCA), which is a technique for
assessing the environmental aspects associated with different steps in the production of a
product, can be performed on ionic liquids like any other chemical compound to evaluate
their true greenness. It is worth mentioning that, even though life cycle analysis of ionic
10


liquids could be very beneficial for the informed selection of these compounds, it is quite
challenging. The challenge arises from the fact that: ionic liquids are not yet produced in the
large commercial scales. Consequently, no primary data is available on material/energy
consumption and direct environmental discharges during their production. On the other hand
there is very little data on the environmental fate, transport and toxicity of ionic liquids in the
literature.
As environmental impact studies continue to reveal the negative impacts that VOC
compounds have on the environment, chemist and engineers are striving to develop
alternatives that reduce the overall ecological impact of current chemical processes. Although
the ionic liquid field is developing rapidly it is important to consider the environmental,
ecological, and human health impacts at the design stage for their successful use and long-
term acceptance. Currently, there is very little understanding of the environmental impacts of
producing ionic liquids as well as their impacts on fresh water ecotoxicity once they are
released to the environment.
1.4 Ionic liquids safety
The low volatility/negligible vapor pressure of ionic liquids eliminates an important pathway
for their release into the environment. The diversity of the ionic liquids' variants available
makes the process of selecting the ones that meet the defined safety requirements easier. A
study shows that ultrasound waves can convert a solution of imidazolium-based ionic liquids
with acetic acid and hydrogen peroxide (H2O2) to less harmful compounds.36 Despite the fact
that ionic liquids mostly have negligible vapor pressure, few of them have shown
11


combustible properties and therefore should be handled carefully.37 A brief exposure of some
ionic liquids (~5 seconds) to a flame torch can ignite them.
12


Chapter 2: Forward Problem, Prediction of Melting Point and Viscosity of ILs
In this chapter, we present two empirical correlations to predict the melting point and
viscosity of ILs in a way that does not require experimental input or complex simulations, but
rely on inputs from simple calculations based on standard quantum chemistry (QC). To
develop these correlations, we used data related to size, shape, and electrostatic properties of
cations and anions that constitutes the ionic liquids.
2.1 Introduction
As it can be interpreted from their name, ionic liquids are composed of ions, a cation and an
anion, but their properties can significantly vary from their relatives, salts, in two main ways.
First, the properties of salts can be mostly attributed to their ionic nature since strong ionic
bonds hold the particles together. Ionic salts are mostly made of small monoatomic ions,
which are in the close vicinity of each other in their crystal network. Since the lattice energy
of a crystalline compound is proportionally related to the inverse of the distance between the
two components, the ionic bonds of salts are very strong, which contributes to properties
such as very high melting point and high viscosity. On the other hand, ionic liquids are made
of larger multiatomic cations and anions that result in weaker ionic bonds compared to that of
salts. This explains the considerably lower melting point (many of them are in liquid state at
room temperature) and viscosity of ionic liquids. Secondly, contrary to salts, ionic liquids do
not occur naturally in the environment and must be artificially synthetized.
13


The multiatomic nature of cations and anions in ionic liquids presents a great opportunity for
researchers to fine tune ILs properties and tailor them for different applications. In ionic
liquids, mainly cations and occasionally anions are composed of several alkyl side chain
groups (CH2, CH3...) and functional groups (OH, NH2, COOH...). A vast number of different
ionic liquids (an estimated number of 1014 ionic liquids)1 can be potentially synthesized
through distinct combinations of different cation-bases, alkyl groups, functional groups
(attached to cations or anions), and anions. Careful evaluation of experimental data from
literature on the physical and thermodynamic properties of ILs shows that substituting one
type of functional group or anion with a different type can drastically alter the property of
interest, such as its solvency power towards a specific compound. Such behavior and trends
can be seen in all different categories of ionic liquids.
Despite the fact that the ionic bonds in ILs are relatively weak, their properties can still be
attributed to their ionic nature as even a weak ionic bond is still much stronger than other
types of intermolecular forces. Studies show that after ionic forces, hydrogen bonds between
ionic particles (cations and anions) are the most important contributor to physical properties
of ILs. Even though there are, potentially millions of different ILs that are possible, to date
only a few hundred of them have been actually synthetized. It is not humanly possible to
synthesize every feasible ionic liquid; therefore, we need to customize and intelligently
design them before synthesizing for task-specific applications. Computer-aided optimization
frameworks can help us design optimal ionic liquids suitable for a wide range of applications
from an extraction solvent to thermal energy storage. Ionic Liquids are generally salts that
are liquid below 100C. Therefore, not all ILs are in the liquid state at room temperature.
14


When it comes to the melting point of ILs, those with significantly lower melting points or
room temperature ionic liquids [RTILs] (Tm<25C), are of great interest to researchers
seeking new application for ILs. The reason for the desirability of low melting point ionic
liquids is the fact that ILs are being considered as separation solvents for selective dissolution
of gaseous (e.g. CO2), liquid (e.g. toluene), and solid (e.g. cellulose) solutes. They are also
widely considered as liquid solvents to promote chemical reactions.39,40 From a practical
view point, for an IL to be used as an industrial solvent it needs to be transported (pumped)
across multiple unit operations and therefore it must be in the liquid phase.
Another significant barrier towards commercialization of IL based applications is their high
viscosity that occurs due to their ionic nature (existence of strong ionic bonds) making them
difficult to transport. It is necessary to have powerful pumping equipment and efficient
process equipment to handle viscous fluids. Therefore, looking for and customizing ionic
liquids that have relatively low viscosity and melting point will greatly aid in
commercialization of ionic liquids.
Studies related to the ionic materials show that strong ionic bonds are mainly responsible for
holding charged particles together. Crystal lattices of ionic materials (e.g. ionic liquids) are
made of cations and anions held together by electrostatic attraction. The ionic force between
charged particles is directly proportional to the charge of each particle and inversely to the
distance between the two ions. The larger the cations and anions are, the weaker the ionic
bonds between them would be. This is due to the fact that by increasing the distance between
two ions the electrostatic attraction, which holds them together, will be reduced.
15


2.2 Methods
Widely available information related to melting point and viscosity of salts explicitly shows
that there is a meaningful relationship between the magnitude of the two discussed properties
and the lattice energy of ionic bonds. Generally, in an ionic compound, size (volume and
area), shape (sphericity), molecular weight, and dielectric constant of ions play an important
role in determining the strength of the ionic bond of the compound. Normally, the larger size
of the positive and negative ions (cations and anions) results in longer distances between the
ions in the crystal making the ionic bond weaker. On the other hand, the shape of the ions is
also important as they are better packed together when they are more symmetrical in shape. It
has been suggested that asymmetry of ions in an ionic compound, most likely, will decrease
the melting point since ions are more loosely connected and can be separated from each other
more easily (by applying lower amount of energy). When we compared the size of cations
and anions of a variety of ionic liquids with their melting points and viscosities, we were able
to observe that in the case of ionic liquids the relationship is much more complex. We came
across ionic liquids, which violated the above discussed trends where certain ILs with
relatively larger cations and anions did not necessarily have lower melting point or viscosity
compared to smaller ILs. One reason for this is the fact that ionic bonds are not the only
intermolecular forces responsible for holding the particles together and other types of forces
such as hydrogen bonding and polar-polar forces also come into play. Therefore we
developed new correlations to predict melting point and viscosity of ionic liquids using
information related to the size, shape and electrostatic properties of their ions. In order to
account for deviations related to the above discussed trends, in addition to the three
16


descriptors, we included several other quantum chemical descriptors to refine the correlations
with an aim to cover wide variety of ionic liquids. Another issue was that for many ionic
liquids, there were multiple experimental values reported for the two physical properties.
This inconsistency was especially observed in the case of melting point, primarily due to the
fact that the process of synthesis and existence of impurities affects this property. We
avoided considering ILs that had inconsistent experimental values during the development of
the correlation. Ionic liquids selected for this study were all 1:1 (one cation and one anion)
with delocalized charges. These types of ILs are normally able to avoid crystallization and
form glasses compounds far below room temperature.41
Currently, the most widely used approach to predict the melting point of ILs is quantitative
structure-property relationship (QSPR) methods, mostly combined with artificial neural
networks (ANNs).41 In this approach, there is a reasonably good correlation between actual
and predicted melting points within a standard deviation of less than 10C.41 The limited
availability of experimental data on physical properties of ILs is the main drawback of
constructing good QSPR models. In recent years, simulations with molecular dynamics (MD)
have evolved to study the behavior of ILs. The quality of these simulations strongly depends
on the employed force fields. Several groups have tuned them specifically for ILs, while
others have modified previously existing ones. For example, Alavi and Thompson have used
MD simulations to predict the melting temperature of [C2MIm]+PF6. The demanding
simulation indicated a melting point that was approximately 43C too high.41 Maginn used a
similar model for the two polymorphs of [C4MIm]+CF and obtained a T|lls that was between
17


20 and 55C too high.41 The drawbacks of all MD simulations are the high load of
computational calculations and need to know the crystal structure.
In order to collect quantum chemistry data of cations and anions, we used TURBOMOLE,
which is a powerful, general-purpose Quantum Chemistry (QC) program, which can be used
for ab-initio electronic structure calculations.41 This software allows accurate prediction of
cluster structures, conformational energies, excited states, and dipoles that can be used in a
broad variety of applications. When a chemical compound, in our case a cation or anion, is
simulated using TURBOMOLE it can be exported as a Cosmo file, which can be later used
in COSMOtherm software. COSMOtherm is a universal tool, which combines quantum
chemistry (QC) and thermodynamics to calculate properties of liquids.41 This tool is able to
calculate the chemical potential of different molecules (in pure or mixed forms) at different
temperatures. In contrast to other available methods, COSMOtherm is able to predict
thermodynamic properties of compounds as a function of concentration and temperature by
equations, which are thermodynamically consistent.41
Previously, computational methods such as BP86/SV(P) optimization approaches were
carried out within the TURBOMOLE program package through the resolution of identity
(RI) approximation. The imported/created geometries were then used for further optimization
with the TZVP basis set. Next, when program converged a file with .cosmo format was
exported, which later was used in COSMOtherm software for further calculations. At the
next step, the computational chemistry data associated with cations and anions in the selected
ionic liquids required for the development of aforementioned empirical correlations were
18


collected. The data obtained from COSMOtherm software was later used to develop the
empirical correlations for the prediction of melting point and viscosity of ILs. In the
COSMOtherm software BP_TZVP_C30_1401 dataset was chosen for all of the calculations.
2.3 Results and discussion
Molar mass of cation (MWc), molar mass of anion (MWa), volume of anion [A] (VoU),
volume of cation [A] (Vole), Area of anion [A] (AreaA), Area of cation [A] (Areac),
dielectric energy of cation (Dic), dielectric energy of anion (DiA), symmetrical value of ions
k o
(a), density [] (p), radii of cation [A] (Rc) and radii of anion [A] (RA), and the average
distance between one cation and one anion in the network [A] (Rt) were used as descriptors
to develop empirical correlations to predict melting point and viscosity of ILs. The accurate
measurement of volume, area and ionic radii of ions is only possible through imaging
approaches such as X-ray diffraction. In the case of unavailability of X-ray data correlative
or approximation approaches (e.g. van der Waals model) can be used to estimate the
characteristics related to size and shape of particles. This data is very sparse for ILs and since
we are interested in the development of a universal correlation covering a wide range of IL
structures we have to rely on predictive data. Typical cations and anions in ILs, do not
usually have spherical shapes, so it is necessary to estimate their ionic radii through
correlations for further use. To develop a predictive approach to estimate approximate values
for radii of the cations and anions, we selected ions for which experimental ionic radii data
were available. Next, the radii of cations and anions were estimated using the van der Waals
model in which all cations and anions were assumed to have spherical shape. Volume of
19


corresponding ions were estimated from COSMOtherm software and were converted to the
radii through eqn. (2-1).
R
Calc
= -x-x Vol1/3
(2-1)
A linear correlation between the experimental radii of cations and anions (for which X-ray
values were available) and the corresponding values of their van der Waals radii is displayed
in Figure 2-1. As it can be seen from Figure 2-1, the actual and model predicted values for
the radii of ions are perfectly correlated through a linear relationship with R =0.98622 and
the corresponding equation shown in eqn. (2-2).
Rpred 0.48601 X Rcalc ~ 0.01725
(2-2)
T3
CO
cm
M
o
0
CO
>
0
E
0
Q.
X
Lll
3.5
3
2.5
2
1.5
1
0.5
0
y=0.48601x-0.01725
R2 = 0.98622
3 4 5
Vdw Radii (A0)
Figure 2-1: A correlation between van der Waals (VdW) and experimental Radii
20


Although, there were only few experimental data available for use in this correlation, since it
covers a large range of VdW ionic radii from 3.8 A to 7.5 A, we used this correlation to
predict the unknown values of ionic radii for the rest of cations and anions considered in this
study. The distance between anions and cations in the studied ILs, Rt, were estimated as the
sum of ionic radii of cations and anions present in the crystal network of ILs, The
symmetrical value of ionic liquids, o, were calculated using the sphericity of cations and
anions. Sphericity is a measure of how spherical an object is and can be calculated using the
formula shown in eqn. (2-3).
Sph =
1 I
7T3(6Vp)3
AP
(2-3)
where, Vp and Ap are the volume and surface area of the particle, respectively.
Further, the symmetrical value of ionic liquids were calculated through the following
equation, eqn. (2-4).
a = AjSphc SphA (2-4)
where, Sphc and SphA are the sphericity of cations and anions, respectively.
2.3.1 Melting point
Experimental data on the melting point of several ILs covering different categories (different
type of cation head groups, and anions), were gathered from the literature.42'59 A multivariate
correlation with several inputs based on quantum chemistry parameters gathered from
21


COSMOtherm, and the output parameter, experimental values of Tm, were performed
utilizing Eureqa software, a powerful data analysis tool developed by Nutonian, Inc.
In the case of predicting the melting points of ionic liquids, 37 points of data on the actual
melting point values of ILs were used, out of which 17 data points (i.e. 45% of the data
points) were chosen solely for validation set, thereby not participating in the training process
and are depicted as the points in green color in Figure 2-2.
The experiential data of the melting point of ILs and the multivariate trend line, representing
the empirical correlation developed to predict the melting point, are shown in Figure 2-2. As
it can be seen the trend line is capable of predicting the melting point of the selected ILs.
0 5 10 15 20 25 30 35
Figure 2-2: Actual vs. model predicted melting points for training and test data sets
At the next level, a comparison between the experimental (observed) values of melting points
and their corresponding values, predicted by the correlative model, are shown in Figure 2-3.
22


390
380
370
360
350
340
330
320
| 310
% 300
a
290
230
270
260
250
240
230
11
size 23*r3jn)
si* 23(vaHdat<#i



210 220 230 240 250 260 270 280 290 300 310 320 330 340 350 360 370 380 390
Predicted
Figure 2-3: Goodness of the model for predicting melting points of the selected ionic
liquids
The best mutivariative correlation (possessing highest R ) developed by the software was
used to predict the melting point of ILs as shown in eqn. (2-5).
Tm = a (Ra) + b (p) + c (Kc)(cr) d e(MWc) f(DiA) (2-5)
a=11.38, b=0.05413, c=196.6, d=434.3, e=0.649, f=1661
Table 2-1 lists the characteristics related to the selected predictive empirical correlation.
Table 2-1: Model characteristics in prediction of melting points (calculated on validation
data
Parameter Value
R2 (Goodness of Fit) 0.8534
Correlation Coefficient 0.9588
Maximum Error 33.7208
23


Parameter Value
Mean Absolute Error 14.8818
Maximum Relative Error 11.9865
Mean Relative Error (%) 4.6378
2.3.2 Viscosity
In the next step, to develop an empirical correlation for predicting viscosity of ILs,
experimental data on the viscosity of several different ILs at different temperatures, were
gathered from the literature.60'75 Once more, a multivariate correlation with several quantum
chemistry descriptors along with temperature, as inputs parameters, and Ln (viscosity), as the
output parameter, was developed using Eureqa software.
In the case of predicting the viscosity, we used 78 points of data on the actual viscosity of ILs
at different temperatures, out of which 23 data points were chosen to be in our validation set,
thereby not participating in the training process and are depicted as the points with green
coloring in Figure 2-4.
The experiential data for the viscosity of ILs and the multivariate trend line, representing the
empirical correlation developed to predict the viscosity, are shown in Figure 2-4. As it can be
seen the trend line is capable of predicting the viscosity of selected ILs.
24



ln_Vis(train)
I n_Vi s fva I i d ati o n'
aiie29 ----------------
60 70 80
Figure 2-4: Actual vs. model predicted viscosities for training and test data sets
Once more, a comparison between the experimental and model predicted values of viscosity
are depicted in Figure 2-5.
4
3.5
3
CD
£
CD
S 2.5
o
2
1.5



*
I*



1.5 2 2.5
Predicted
3.5 4
Figure 2-5: Goodness of the model for predicting viscosities of the selected ionic liquids
25


2
The best correlation (possessing highest R ) to predict the viscosity of ILs is shown in eqn.
(2-6).
ln(vis) = a (a) + b (Rt) + c (VolA) d e (AreaA) f (T) g (DiA) h (Dic) (2-6)
a=16.513, b=2.2179, c=0.00892, d=15.0073, e=0.02686, f=0.02975, g=15.8297, h=48.1367
Table 2-2 lists the characteristics related to the empirical model used to predict the viscosity
of ILs as a function of QC parameters and temperature.
Table 2-2: Model characteristics in prediction of viscosity (calculated on validation data)
Parameter Value
R2 (Goodness of Fit) 0.9633
Correlation Coefficient 0.9823
Maximum Error 0.3430
Mean Absolute Error 0.08895
Maximum Relative Error (%) 11.3247
Mean Relative Error (%) 3.366
26


Chapter 3: Reverse Problem; Computer-aided Design of Ionic Liquids
In this chapter a general computer-aided IL design (CAILD) framework along with 4 case
studies used to evaluate the ability of the model to select optimal ionic liquids for different
applications, are presented.
3.1 Introduction
There exists a large library of anions and cations.7677 Similar to organic compounds, where
the atoms carbon, hydrogen and oxygen can be combined to form thousands of alternative
molecular structures ionic liquids can be formed through any combination of cations, anions,
and alkyl groups attached to the cation core leading to several structural possibilities
(estimated to be as many as 10 combinations). This is due to the fact that ILs are
composed of organic cations and these organic compounds can have unlimited structural
variations due to the easy nature of preparation of many components. Moreover, synthesis
of a wide range of ionic liquids is relatively straightforward. This presents a great
opportunity to engineer ionic liquids that have specific properties. Task specific ionic liquids
can be designed for a particular application by controlling the physicochemical properties by
judicious selection/modification of the cation, the anion, and/or the alkyl chains attached to
the cation. This also presents an unusual challenge, where synthesizing, screening, and
81
testing the limitless possibilities becomes an impossible task.
This is where in silico methods could act as a valuable tool for discovering new ionic liquids
with tailored properties. Up until now, the majority of ionic liquid computational studies are
27


based on ab initio methods such as molecular dynamics, and quantum chemical
calculations. These methods are extremely important and offer useful insights as they are
able to predict properties without performing costly experiments. However, as in the case of
experimental studies, one needs to perform several individual simulations which again is
impractical due to the long simulation times required for statistical averaging. Both
molecular dynamic simulations and experimentation are necessary steps in the selection of
task specific ionic liquids. These are important steps to be applied at the final stages of ionic
liquid selection. The missing piece is a method for fast exploration, design and identification
of a subset of promising candidates, from the millions of ionic liquid alternatives that are
available.
Computer-aided molecular design (CAMD), is a promising approach that has been widely
applied for molecular systems to design organic solvents for a variety of applications.86'92 It
integrates property prediction models and optimization algorithms to reverse engineer
molecular structures with unique properties. We believe that this approach is even more
needed for the design of ionic liquids due to the numerous ionic combinations that are
possible.
3.2 Computer-aided ionic liquid design (CAILD)
In this study we present an overarching framework that aims to identify ionic liquids that
exhibit certain desirable behavior. Here, the identity of the compound (in this case an ionic
liquid) is not known a priori, but we can specify the properties that the compound (i.e. ionic
liquid) needs to have.86 This approach termed as computer-aided ionic liquid design
28


(CAILD) can be defined as given a set of ionic liquid functional groups (i.e. base Cations:
Im+, Py+, NH4+ ...anions: Cf, Tf2N, BF4'...side chain groups: CH3, =0, -S-, OH....) and
a specified set of target properties (e.g. melting point, electrical conductivity, viscosity,
solubility...) we can find an ionic liquid structure that matches these properties.

Given IL Groups, Find Properties (Structure-Property Models)
*-
Forward Problems (e.g. GC)
Problems (CAILD)
*---------------
Given Desired Properties, find optimal IL structure
( \
Properties
e.g. Tb. Tns. p. p. £
V____________________________________________
r Fragments (groups) N
e.g. IM*. TftN', CHb. CH2_
V /
Reverse
Figure 3-1: CAILD framework
Intuitively, CAILD can be thought of as a reverse problem of structure (or group) based
property prediction as shown in Figure 3-1. In the forward problem (property prediction) we
know the ionic liquid structure and are interested in its properties. In the reverse problem
(CAILD) we know the target property values (or ranges) and are interested in feasible ionic
liquid structures. To implement CAILD, we need: 1) a framework to fragment ionic liquids
into groups; 2) combination and feasibility rules to identify chemically feasible ionic liquids;
3) structure (or group) based models for property prediction; and 4) an optimization
framework to search through millions of available alternatives. Mathematical programming
29


approaches provide a useful mechanism to solve such CAILD problems. Different methods
to solve such problems include generate and test type approaches (e.g. Harper et al.) or
optimization based approaches (e.g. Sahinidis et al.)94 The solution to the underlying
Mixed Integer Non-Linear Programming (MINLP) model results in the optimal molecular
structure for a given application. The objective function is usually an important property
related to the design problem, while the constraints relate to structural feasibility, pure
component properties, solution (mixture) properties and equilibrium relationships. A
conceptual representation of this approach is shown in Figure 3-2.
Figure 3-2: Conceptual description of CAILD
Section 3.2.1 will focus on describing the general mathematical framework of the proposed
approach; section 3.2.2 deals with structural constraints, providing an in-depth mathematical
treatment of feasibility, complexity and bonding rules required to design chemically feasible
ionic liquids; section 3.2.3 deals with physical property constraints where group contribution
based structure-property models are discussed for predicting ionic liquid physical (pure
30


component) properties; section 3.2.4 focuses on solution property constraints that utilize the
functional group concept based models, such as UNIFAC, for calculation of solution
(mixture) properties through activity coefficients; section 3.2.5 focuses on operations
research methods for solving the proposed optimization model.
3.2.1 Mathematical framework
The generic mathematical formulation of the CAILD model as an optimization problem is
shown in eqns. (3-1) to (3-11). This formulation takes the form of a mixed integer non-linear
programming (MINLP) model.
max f(c,a,y,ng,x) (3-1)
h^c, a,y,ng) = 0 (3-2)
h2(c,a,y,ng) < 0 (3-3)
g2{c,a,y,ng) < 0 (3-4)
d1(c, a,y,ng,x) = 0 (3-5)
d2(c,a,y,ng,x) < 0 (3-6)
c e Rm (3-7)
a G Rn (3-8)
y £f (3-9)
31


ng G Rq
(3-10)
x G Rr (3-11)
where /ix is a set of structural feasibility and complexity equality constraints, h2 is a set of
structural feasibility and complexity inequality constraints, g2 is a set of pure component
physical property inequality constraints, d1 is a set of equality design constraints, d2 is a set
of solution (mixture) property inequality design constraints, c is a w-dimensional vector of
binary variables denoting cation base groups, a is a ^-dimensional vector of binary variables
denoting anion, y is a //-dimensional vector of binary variables denoting the alky side
chains, ng is a //-dimensional vector of integer variables representing number of groups in
the alkyl side chains, and x is a r-dimensional vector of continuous variables representing
compositions, flow rates etc.
3.2.2 Ionic liquid structural constraints
The designed ionic liquids need to satisfy certain rules to ensure chemical feasibility. These
rules, termed as structural constraints, include feasibility rules such as the octet rule, the
bonding rule and complexity rules. Similar rules have been previously developed for
molecular compounds.9596 Eqns. (3-12) to (3-24) represent a comprehensive set of
constraints that were developed to ensure design of ionic liquid candidates that are
chemically feasible.
hecCt = 1 (3-12)
32


2yea G 1
(3-13)
M n' II M m Cl 'ci Cl (3-14)
T,iec(2-vcdCi+T,i=1TlkeG(.2-vGkl')ylngki = 2 (3-15)
I.keGyingki (2-vGkl) = 1 (3-16)
'Li=i'LkeGyingki < nuG (3-17)
Ikecyingia < nuGl (3-18)
HkEG*yin9kl ^ G (3-19)
Y>kEG*yin9ki ^ G (3-20)
H kEG* yin9kl = G (3-21)
Zf=i ZkEG^yingki < G (3-22)
Ef=i ZkEG^yingki > G (3-23)
Zf=i ZkEG^yingki = G (3-24)
where q is a vector of binary variables representing the cations and a,- is a vector of binary
variables representing the anions. yt is a vector of binary variables representing the alkyl
chains l. ngki is a vector of integer variables representing the number of groups of type k in
the alkyl side chain l. vci vGkl are vectors of group valencies of the cations and alkyl groups,
respectively. G is the set of all alkyl groups available for the cation side chains, eqns. (3-12)
33


and (3-13) ensure a maximum of one cation base and one anion respectively for each IL
candidate, eqn. (3-14) fixes the number of alkyl side chains attached to the cation based on
available free valence of the cation base. Modified octet rule: The implementation of the
modified octet rule, eqn. (3-15), ensures that any designed cation is structurally feasible and
that each valence in all structural groups of the cation is satisfied with a covalent bond. Note
that this formulation has already accounted for the positive charge associated with the cation,
eqn. (3-16) ensures that the octet rule is implemented for each side chain / to ensure that the
valences in the individual chains are satisfied with a covalent bond. Cation size: The size of
the cation is controlled by introducing an upper bound on maximum number of groups (rig)
that are allowed in the cation, eqn. (3-17). Alky chain size: The size of the alkyl chains are
controlled by introducing an upper bound on the maximum number of groups that can be
present in each alkyl side chain, eqn. (3-18). Eqns. (3-19), (3-20) and (3-21) can be utilized
to place restrictions on number of occurrences (ti, t2 and t?) of a certain group, G*, in each
side chain /. In other words, eqn. (3-19) can be used to make sure that a certain main group
such as aldehyde or alcohol not being present more than a certain number of times in each
side chain of the cation and eqn. (3-20) can be applied when we want a certain group to be
present at least t2 times and eqn. (3-21) can be used when an exact number of occurrence of a
certain group is desired e.g. when we want to have exactly one aldehyde group in a certain
side chain in the cation. Eqns. (3-22), (3-23) and (3-24) can be utilized to place restrictions
on number of occurrences 04, E and tf,) of a certain group G*\ in the cation, which can be
calculated as summation of number of occurrences of the particular group in all the side
chains in the cation. The purpose of eqns. (3-22) to (3-24) is exactly similar to that of eqns.
34


(3-19) through (3-21) with the difference of placing restrictions on the number of
occurrences of a particular group in whole cation (summation of all of the side chains)
instead of only one side chain.
The cation related structural feasibility constraint, eqn. (3-14), is explained using the generic
cation dialkylimidazolium (shown in Figure 3-3a) as an example. According to the proposed
formulation the Valence for this cation is 2 (i.e.uc; = 2), as there are 2 alkyl side chains (Ri
and R2) that are allowed. Similarly, the Valence of a trialkylimidazolium (shown in Figure 3-
3b) is 3. For dialkylimidazolium, the right hand side of constraint 3, eqn. (3-14), will
translate into £iec Ciuci = (1)(2) = 2, which will fix the left hand side of constraint 3 as
Y,i=i Ji = 2 i. e. [y1 = 1] & [y3 = 1], Therefore the vector y, will take the following values
[1 0 1 0 0 0] with the two ones representing the presence of 2 alky side chains at positions 1
and 3. Next, with the use of few feasible and infeasible examples shown in Figure 3-4, we
explain how the whole set of feasibility constraints, eqns. (3-12) to (3-16), work. Figure 3-4a
shows a feasible ionic liquid, 1,3-diethylimidazolium tetrafluoroborate. The cation Valence
and alkyl group valences related to this ionic liquid are listed in Table 3-1.
Table 3-1: Cation and alkyl side chain groups valences
Cation groups (c) i vci
1 -alkyl -3 -alkyl -Im 1 2
Alkyl groups (k) / vGkl
ch3 1 1
CFF 1 2
CH, 3 1
CTF 3 2
35


The values related to the vectors yt and ngkifor this ionic liquid are listed in Table 3-2.
Table 3-2: Values of yf and ngki for 1,3-diethylimidazolium tetrafluoroborate
/ Vi k n9kl
1 1 ch3 1
ch2 1
ch3 1
3 1 ch2 1
In this case there is only one cation base (eqn. (3-12) is satisfied) and one anion (eqn. (3-13)
is satisfied). The number of side chains are 2 which equates to the valence of cation (eqn. (3-
14) is satisfied). Now, left hand side (LHS) of eqn. (3-15) translates into (2-2)(l)+[(2-
2)(l)(l)+(2-l)(l)(l)]+[(2-2)(l)(l)+(2-l)(l)(l)]=2 which is equal to right hand side (RHS) of
the equation (eqn. (3-15) is satisfied). For both of the side chain positions (/) 1 and3, LHS of
eqn. (3-16) translates into [(1)(1)(2-1)+(1)(1)(2-2)]=1 which is equal to RHS of the equation
(eqn. (3-16) is satisfied). Eqns. (3-17) through (3-24) are only used to control the cation size
and place restrictions on the type and number of occurrences of select groups. As such these
are not feasibility constraints but user specified structural design constraints. Figure 3-4b
shows an infeasible imidazolium-based ionic liquid. In this case there is only one cation base
(eqn. (3-12) is satisfied) and one anion (eqn. (3-13) is satisfied). The number of side chains
are 2, which equates to the valance of cation (eqn. (3-14) is satisfied). The LHS of constraint
4, eqn. (3-15), translates to (2-2)(l)+[(2-2)(l)(l)+(2-2)(l)(l)]+[(2-l)(l)(l)+(2-l)(l)(l)]=2,
36


which equates to the RHS of the equation and hence constraint 4, related to the whole cation,
is satisfied. However, constraint 5, eqn. (3-16), related to the side chains are violated as
follows: for side chain 1=1 containing 2 ethyl groups, eqn. (3-16), translates to, [(1)(1)(2-
2)+(l)(l)(2-2)]=£ 1, and for side chain 1=3 containing 2 methyl groups eqn. (3-16) translates
to, [(1)(1)(2-1)+(1)(1)(2-1)]=£ 1. Therefore, the structure shown in Figure 3-4b is infeasible.
For the structure shown in Figure 3-4c, eqns. (3-12), (3-13) and (3-14) are satisfied. The
LHS of constraint 4, eqn. (3-15), translates to (2-2)(l)+[(2-2)(l)(2)+(2-l)(l)(l)]+[(2-
1)(1)(2)]=3, which does not equate to the RHS of the equation (i.e. 2) and hence constraint 4,
related to the whole cation, is violated. Constraint 5, eqn. (3-16), related to the side chains
translates to the following: For side chain position 1 (i.e. 7=1), LHS of eqn. (3-16) is
[(1 )(2)(2-2)+( 1)(1 )(2-1 )]=1, which equates to the RHS of the equation (i.e. 1), but for side
chain position 3 (i.e. 1=3) the same equation translates to [(1)(2)(2-1)]=2 which does not
equate to the RHS of the equation (i.e. 1) and hence constraint 5, related to side chain 3, is
violated. Therefore, the structure shown in Figure 3-4c violates two of the feasibility
constraints and hence is infeasible.
*3
'1 b)
Figure 3-3: A general Scheme of two feasible cation structures
37


ch2- ch3
i
CHj CH3
CH.-CH
F / s rN' F Y F^ CHjCHj- CHj
i b-F F yrT F O <- Y fA-f
a) CH3-CH3 h CHj CHj F ^
Figure 3-4: Examples of feasible and non-feasible Ionic liquids
3.2.3 Physical property constraints
Ionic liquid structures play a key role in determining their unique physical properties.
Physical property constraints utilize structure-property models which provide insights into
the relationship between molecular structures and their properties. The particular type of
structure-property relationships suited for CAILD are group contribution (GC) models. As
discussed before, GC models for physical properties are based on functional group additive
principle. The ionic liquid is fragmented into characteristic groups and the property of
interest is predicted as an additive function of the number of occurrence of a given group
times its contribution to the pure component property. The contribution parameters of
different groups are derived by correlating experimental data to a group additive expression.
These models exist for several IL physical properties such as viscosity97, density98'101,
melting point102, electrical conductivity103, thermal conductivity103, heat capacities104,
solubility parameter105 and toxicity.106
38


3.2.4 Solution property constraints
Thermodynamic properties of non-ideal solutions are important for evaluating intermolecular
interactions between multiple components (both ionic and molecular) present in a mixture.
These thermodynamic properties are essential to evaluate the potential of ionic liquids as
solvents for reaction (solid solubility and liquid miscibility) and the separation of fluid
mixtures (liquid-liquid extraction and gas-liquid absorption). An essential requirement is the
ability to predict excess Gibbs free energy (activity coefficients) of systems involving ionic
liquids which enable prediction of equilibrium concentrations. These constraints are not only
a function of binary/integer structure variables but also relate to the compositions of the
various components of the mixture. The proposed CAILD framework requires models for the
prediction of activity coefficients that are based on solution of groups concept. The basic
hypothesis of the solution of groups concept is that interactions between molecules can be
approximated as interactions between functional groups. To illustrate this concept, the
different interactions amongst the groups of a mixture of an ionic liquid, [Mim] TfzN, and
CH3OH is shown in Figure 3-5.
><
OH

w
gro u p gro u p i nteracti on
ion-ion interaction
ion-group interaction
Figure 3-5: Different types of group interactions involved in solutions containing ionic
39


The number of distinct cation head groups, anion groups and alkyl groups are much less in
comparison to the number of distinct ionic liquids that can be generated from them.
Therefore, a relatively small number of group interaction parameters are required to represent
all possible ionic liquids. UNIFAC (UNIversal quasi-chemical Functional-group Activity
Coefficients) is a widely used group contribution model to predict phase equilibrium in non-
electrolyte systems. The UNIFAC model combines the concept of functional groups with a
model for the activity coefficient based on UNIQUAC (Universal Quasi Chemical). The
activity coefficient has a combinatorial contribution (due to differences in size and shape of
molecules) and a residual contribution (due to energetic interactions).
In y; = lnyf + lnyf (3-25)
The group volume (R) and surface area (Q) parameters of the combinatorial part are
calculated as summation of group parameters (volume Rk and surface area Q/:) while binary
group interaction parameters (amn and anm) are required for the calculation of the residual
component. The UNIFAC approach was originally used for non-electrolyte systems, however
in recent studies several research groups have utilized this approach for ionic liquids by
careful representation of ionic groups and/or incorporating assumptions that factor the ionic
nature of the groups. In order to apply the UNIFAC model, in its current form, to ionic
liquids Wang et a/.108 and Lei et a/.109 treated ionic groups as a single non-dissociate neutral
entity. For example, the ionic liquid [Bmim] BF4', was decomposed into two CH3 groups,
three CH2 groups and one [Im] BF4' group. Using the above representation Lei et a/.109 have
added 12 new ionic groups (e.g. [Im] PF6) to the existing UNIFAC table. Ionic liquid
40


groups are included in the modified UNIFAC (Dortmund) model.110 Most recently, Roughton
et al.105 have characterized the ionic groups in the same way as proposed in our proposed
CAILD formulation; i.e. as separate cation base, anion and alkyl groups. The underlying
assumption is that the ionic groups can be treated separately and the interactions between the
ionic groups can be assumed to be zero due to the strong interaction and weak dissociation
between ion pairs.105 More detailed treatment of UNIFAC approach for ionic liquids can be
found in Roughton et a/.105, and Wang et al.108, Lei et al109
Liquid-Liquid equilibrium
Designing industrial scale liquid-liquid separation systems using ionic liquid requires
modeling equilibrium relationships. In a non-ideal liquid mixture, species which have limited
mutual solubility in the given liquid phase exhibit positive deviations from Raoult's Law. The
quantitative measure of non-ideality is the liquid activity coefficient y, which is a function of
composition and temperature. If we identify the two liquid phases as l\ and l{, their
respective mole fractions in the two phases are related by the equilibrium condition as
follows:
Yhx*u = yhAx 2,i (3-26)
Where, yll,i ar|d yl2J are activity coefficients of component i in the liquid phases 1 and 2
respectively, and xl t and x2;x are mole fractions of component i in the two phases.
Solid-Liquid equilibrium
41


Estimation of equilibrium saturation concentrations of solid-liquid systems is essential to
model processes that involve solute dissolution, and crystallization. The liquid phase activity
coefficient predictions discussed previously and the pure component properties of solute
(AfUSH, Tm), can be utilized for these calculations.
where AfUSH Tm and T represent enthalpy of fusion (J/mol), melting point (K) and
temperature (K), respectively. Yiat-. represents activity coefficient of solute at saturation and
x{at is the solubility of solute.
3.2.5 Solution of the underlying MINLP
The presented CAILD model is a non-convex, mixed integer non-linear programming
(MINLP) problem, involving large number of integer and binary variables. Consideration of
mixture properties through the UNIFAC model results in non-linearity and most of the binary
design variables (structural) participate in the non-linear terms. Combinatorial complexity is
an inherent issue in CAMD MINLP models due to the nature of the search space. The most
direct approach for solving the underlying MINLP model is complete enumeration. Generate
and test methods fall under this category. Solution to the MINLP model can also be
achieved through mathematical programming using deterministic (e.g. branch and bound111,
branch and reduce ) and stochastic optimization methods (e.g. simulated annealing ,
genetic algorithms114, and tabu search90). Approaches that combine features from both
(3-27)
42


domains, such as decomposition methods, have also been previously developed.94,115 Achenie
et at.X6 provide a detailed description of various solution techniques in the context of
molecular design problems. In this section we focus on solving the CAILD framework
utilizing two different methods: the decomposition methodology (includes generate and test
algorithms) and genetic algorithm based optimization. Our purpose is to demonstrate that
different types of solution approaches can be used towards a solution of the proposed CAILD
formulation. Main details about the two approaches are provided below while in depth
analysis can be found elsewhere.115,116
Decomposition Method
In this approach the CAILD-MINLP model is decomposed into an ordered set of
subproblems where each subproblem requires only the solution of a subset of constraints
from the original set. As each subproblem is being solved large numbers of infeasible
candidates are eliminated leading to a final smaller subproblem. The first subproblem usually
consists of the structural constraints and it equates to enumeration. The second subproblem
consists of pure component (physical) property constraints while the third subproblem
consist of mixture property constraints. These three subproblems taken together equate to
generate and test methods. The ionic liquid candidates that pass through all of the above
subproblems are the only ones that will be considered in the final optimization subproblem
that involves the objective function, equilibrium relationships and process models (if
considered in the design problem). Most often, the solution to the final subproblem can be
achieved by solving a set of non-linear programming (NLP) problems.
43


Genetic Algorithm
Genetic algorithm (GA) is a method that can be used to solve optimization problems based
on the natural selection process that mimics biological evolution. It can be applied to solve
problems that are not well suited for standard optimization algorithms, including problems in
which the objective function is discontinuous, non-differentiable, stochastic, or highly
nonlinear.116 Unlike traditional search and optimization methods, GAs perform a guided
stochastic search where improved solutions are achieved by sampling areas of the search
space that have a higher probability for good solutions.116 The optimization process starts
with a collection of chromosomes (candidates). The fitter candidates are selected as parents
and allowed to exchange or alter their genetic information, through crossover and mutation
operations, with an aim to create more fitter off springs. At every iteration new populations
of off springs are created to replace the existing population. This process of evolution is
repeated for a pre-determined number of generations or until the solution is found.116 In GA,
the selection of fitter parents for next generation is based on their fitness values as
determined by a fitness function. The fitness function is usually very closely related to the
original objective function of the search problem (in all of the case studies presented in this
study, the fitness function was identical to the objective function). The GA solution of
CAILD model was implemented in the MATLAB environment with most parameters taking
the default values. Specifically, for all the case study problems, the population size was fixed
at 20 and the initial population was generated randomly. The crossover fraction was fixed at
0.8 while the mutation probability was fixed at 0.2. We allowed two candidates with the best
44


fitness values (elite candidates) in the current generation to automatically survive to the next
generation.
3.3 Proof of concept examples
In this section several case studies have been presented to illustrate the usefulness of the
proposed approach. Table 3-3 lists the entire set of groups and their respective valences, from
which the basis sets for the four case studies were derived. Note that this basis set covers
only a small set of cations, anions and functional groups for which group contribution
parameters are currently available for the properties of interest. However, the design
approach itself is universal in nature and upon availability of group contribution models and
parameters can be easily extended (for example, to all groups in Appendix A) to cover the
entire spectrum of possible ionic liquids. The maximum value allowed for the number of
groups were fixed at 6 for each side chain and 12 for the whole cation.
Table 3-3: The basis set used for ionic liquid design
Cations Valence Anions Groups Valence
Im 2 BFJ ch3 1
Mim 1 pf6- ch2 2
Py 2 Tf2N-
Mpy 1 cr
45


3.3.1 Electrolytes
This case study demonstrates the design of an ionic liquid that has high electrical
conductivity. Electrical conductivity measures the ability of a material to conduct electric
current. It is an important property for the development of electrochemical devices such as
high energy batteries. Other design requirements include the following: the electrolyte (i.e.
ionic liquid) needs to be a room temperature ionic liquid (RTIL); and it should have
reasonably low viscosity.
The electrical conductivity of ionic liquids can be estimated using a Vogel-Tamman-Fulcher
(VTF) type equation shown in eqn. (3-28).
In X = In Ax +
Bx
(T-Toti
(3-28)
where Ax, and Bx, are adjustable parameters that can be obtained through group contribution
expressions, eqn. (3-29) and eqn. (3-30), as proposed by Gardas et a/.103, and Tok has the
value of 165K for all considered IL types.
Ax = Zi=ini ax (3-29)
& II M if* 3 C3- (3-30)
where rit is the number of groups of type i and k is the total number of groups considered.
Table 3-4 shows the group contribution parameters used.
46


Table 3-4:
Group contributions for parameters A% and B,
Species a. bx(K)
Im 77.8 -501.5
Mim 77.9 -537.6
Py 69.6 -544.9
Mpy 69.7 -581.0
BF4 85.8 -129.4
pf6 117.3 -278.6
Tf2N" 10.1 -46.4
ch3 0.1 -36.1
ch2 0.1 -36.1
117
The viscosity of ionic liquids is calculated using an Orrick-Erbar-type approach. In this
method, viscosity (rj, in cP) can be predicated as a function of density (p, in g cm-3)
molecular weight(M), temperature (T) and parameters A and B through the use of eqn. (3-
31).
In = A + - (3-31)
pM T v 103
103
We employ the group contribution technique proposed by Gardas et al. to estimate the
parameters A and B as follows
A = J$=1niAVit (3-32)
B = J$=1niBVii (3-33)
47


where nt is the number of occurrences of group i (cation, anion and functional groups) and
Av i and Bv i are the contributions of group / to the parameters A and B respectively. The
ionic liquid densities are estimated using the below formula.
_ M
P ~ NV(a+bT+cP)
(3-34)
3 1
where p the density in kg m' M is the molecular weight in kg mof N is the Avogadro
number, V is the molar volume in A3, T is the temperature in K and P is the pressure in MPa.
Based on the data provided in Gardas et al. we developed group contribution parameters
for molar volume with expressions similar to eqns. (3-32) and (3-33). The values of
coefficients a, b and c are 0.8005 0.0002 6.652 x 10-4 0.007 x 10-4 K_1
and5.919 X 10-4 0.024 X 10-4 MPa-1. Table 3-5 shows the group contribution
parameters used in this model.
Table 3-5: Group contributions for parameters A, B and V
Species AV Av Bv
Im 84 8.04 1257.1
Mim 119 7.3 1507.1
Pv 111 7.61 1453.6
Mpy 146 6.87 1703.6
BF2 73 -18.08 1192.4
PFfi 109 -20.49 2099.8
Tf>N" 248 -17.39 510.0
cr 47 -27.63 5457.7
CFF 35 -0.74 250.0
CFF 28 -0.63 250.4
48


The melting point of ionic liquids is estimated by a group contribution approach using eqn.
118
(3-35) as proposed by Aguirre et al.
Si
a+crc+CTc
(3-35)
where nt is the number of occurrences of group i (cation, anion and functional groups) and
Tm i is the contribution of group i to the melting point, a and c are constants with values of
0.1 and 0.012 respectively. rc which is related to the cation flexibility, is estimated using
eqn. (3-36) and oc is a cation symmetry parameter having values shown in Table 3-6.
tc = lk(n(CH2)k 1) (3-36)
Table 3-6: ac values for different cations
Type of cation ac value
R = R (Im, Pyr, Pip) 0.265
R =£ R (Im, Pyr, Pip) 0.317
R or R dimethyl amino in(C(4)Py) 0.265
The contribution, Tm of different groups are listed in Table 3-7.
Table 3-7: Group Contributions for Ionic Liquids Melting Point
Group Tm,i
Im 107.99
Mini 109.88
Py 117.212
Mpy 119.102
49


Group Tm,i
bf4 -0.479
pf6 16.746
Tf2N" -0.966
cr 35.852
ch3 -1.463
ch2 -1.463
The CAILD design problem expressed in mathematical form is shown in eqns. (3-37) to (3-
40).
Objective function
fobj = max (1)
(3-37)
Constraints
Ionic liquid Structural Feasibility (3-38)
rj < 65 cP (3-39)
Tm < 298.15 K (3-40)
Results: The design statistics for this problem are summarized in Table 3-8. A total of 138
feasible IL structures were enumerated in subproblem 1. Out of these, 26 ILs satisfied the
physical property constraints (viscosity and melting point) in subproblem 2. There were no
mixture properties considered (subproblem 3) and the solution to final subproblem
(subproblem 4) resulted in the optimal ionic liquid structure (1-methylimidazolium bis-
trifluoromethylsulfonyl imide) with the highest electrical conductivity (shown in Figure 3-6).
The properties of the designed ionic liquid are listed in Table 3-9.
50


Table 3-8: Decomposition approach: Subproblem Results
Subproblem 1: Number of ionic liquids (ILs) generated, 138_____
Subproblem 2: Number of ILs satisfying pure component properties, 26
Subproblem 3: No mixture properties____________________________
Subproblem 4: Optimal IL, [Mim]+[Tf2N]~________________________
The same design problem, eqns. (3-37) to (3-40) was solved using the genetic algorithm
toolkit in MATLAB and the program picked the exact same structure (shown in Figure 3-6)
as the optimal solution.
_,ch3
rr~ N + 0 0
O f3c-s-n-s-cf3
Figure 3-6: 1-methylimidazolium [TfzN]
Table 3-9: Design Results of the optimal IL, 1-methylimidazolium [Tf2N]
Properties Value
Melting point (K) 270.15
Viscosity (cP) 21.293
Electrical Conductivity (S/m1) @ 25C 1.0956
Analysis: In this section we focus on the validation of design results through careful
consideration and analysis of available experimental data. Table 3-10 lists the available
experimental electrical conductivity, \ (S/m) at 25 C for 7 different ionic liquids that are
51


based on the cation, anion and side groups considered in this case study (Im, Py, PF6, BF4,
Tf2N, CH3 and CH2). Unfortunately, we could not find the electrical conductivity data for the
designed IL (1-methylimidazolium bis (trifluoromethylsulfonyl) imide). Therefore, we
perform a qualitative IL structure-property trend analysis to validate the design results.
The electrical conductivity values have the following trend: \cAmim\ Tf2N~ > ^\c4mim\ bf4 >
/t\c4mim\ PFs Since, all of the above ionic liquids have the same cation (C4mim) but different
anions (Tf2N, PF6" and BF4') we can infer that electrical conductivities of ionic liquids with
Tf2N' anions are greater than those with PF6 and BF4 anions (for same cation and alkyl
groups). The design result is consistent with this observation as the optimal structure has
Tf2N anion. Similarly, by comparing the electrical conductivities of ionic liquids having the
same anion (Tf2N) we can see that [C2mim] Tf2N' > [C4mim] Tf2N' > [Cr,mim] Tf2N'.
Therefore, we can conclude that increasing the number of alkyl groups on the cation side
chain decreases the electrical conductivity. The design result (Figure 3-6) is also consistent
with this observation as there is only one methyl group (minimum needed to satisfy the
cation Valence) present in the cation side chain. Next, we compare the X of [C4mim] Tf2N'
and [C4Py] Tf2N' which have the same anion and different cations. We found experimental
electrical conductivity data for [C4Py] Tf2N' but there was no data available for [C4mPy]
Tf2N'. Since we already know that addition of alkyl groups to the cation base will decrease
the electrical conductivity, we can infer that X [C4mPy] Tf2N' < 0.33 S/m (i.e. X [C4Py] Tf2N'
) which in turn is much less than X [C4mim] Tf2lSr (0.406 S/m). Therefore, we can conclude
that electrical conductivity of ionic liquids with imidazolium based cations are greater than
Pyridinium-based cations (for same anion and alkyl groups). The design result is consistent
52


with this observation as the optimal structure selected had an imidazolium cation. Overall,
the model results are in full agreement with the observed trends from experiments, thereby
validating the proposed approach.
Table 3-10: Experimentally Measured Electrical Conductivities of Ionic Liquids
Ionic liquid T(C) X (S/m) Ref
H,CW^ PF6 25 0.146 [119]
/=\ Tf2N 3 V ^ CH? 25 0.912 [119]
/=\ Tf2N H3C W1n s^n 25 0.406 [119]
H C Tf2N 25 0.218 [119]
f=\9 BF4 h3c^'n^n-ch3 25 0.59 [120]
H r M Vi BF4 H3CWn-^n~CH3 25 0.35 [120]
O Tf2N N CH3 25 0.33 [121]
53


3.3.2 Heat transfer fluids
Ionic liquids show great promise as heat transfer fluids and heat storage medium. High
thermal conductivity is an important property for such applications. Thermal conductivity
measures the ability of a material to conduct heat. Thermal conductivity of ionic liquids is
weakly dependent on temperature and could be fitted with the following linear correlation.
k = Ak BkT (3-41)
where, k and T are the thermal conductivity in Wm4K4 and temperature in K respectively.
We utilize a method96 that employs group contribution approach to estimate the parameters
Ak and Bk.
^ = I?=i ntak (3-42)
Bk = Ylt=1nibk (3-43)
Table 3-11 shows the group contribution parameters that were used. Pyridinium and methyl
pyridinium have not been considered in this part since their group contributions were not
found.103
Table 3-11: Group contributions for parameters Ak and Bk
Species ak bk (K1)
Im 0.1272 0.000000104
Mim 0.1314 0.00000787
bf4 0.0874 0.00008828
pf6 0.0173 0.000009088
54


Tf2N" 0.0039 0.00002325
cr 0.0166 0.00001
ch3 0.0042 0.000007768
ch2 0.0010 0.000002586
Melting point and viscosity of ionic liquids were calculated through the same methods
proposed in case study 1. The CAILD design problem expressed as an optimization model is
shown in eqns. (3-44) to (3-47).
Objective function
fobj = max (k) (3-44)
Constraints
Ionic liquid Structural Feasibility (3-45)
7j < 65 cP (3-46)
Tm < 298.15 k (3-47)
Results: Decomposition approach: The design statistics for this problem are summarized in
Table 3-12. A total of 92 feasible IL structures were enumerated in subproblem 1. Out of
these, 15 ILs satisfied the physical property constraints (viscosity and melting point) in
subproblem 2. There were no mixture properties considered (subproblem 3) and the solution
to final subproblem (subproblem 4) resulted in the optimal ionic liquid structure (1-ethyl-3-
methylimidazolium tetrafluoroborate) with the highest thermal conductivity (shown in Figure
3-7). The properties of the designed ionic liquid are listed in Table 3-13.
55


Table 3-12: Decomposition approach: Subproblem Results
Subproblem 1: Number of ionic liquids (ILs) generated, 92
Subproblem 2: Number of ILs satisfying pure component properties, 15
Subproblem 3: No mixture properties
Subproblem 4: Optimal IL, 1-Ethyl-3-methylimidazolium tetrafluoroborate
The same design problem, eqns. (3-44) to (3-47) was solved using the genetic algorithm
toolkit in MATLAB and the program picked the exact same structure (shown in Figure 3-7)
as the optimal solution.
bf4-
Figure3-7: l-ethyl-3-methylimidazolium [BF4]
Table 3-13: Design Results of the Optimal IL, 1-ethyl-3-methylimidazolium [BF4]
Properties Value
Melting point (K) 291.71
Viscosity (cP) 60.75
Thermal Conductivity (Wm 'K1) @ 25C 0.193
56


Analysis: Table 3-14 shows experimental thermal conductivity, k [Wm^K'1] for 10 different
ionic liquids that are based on the cation, anion and side groups considered in this case study
(Im, PF6", BF4', Tf2N', CH3 and CH2). The designed ionic liquid (Figure 3-7) is same as the
IL with the highest thermal conductivity value in Table 3-14 ([C2mim] BF4'). This partially
validates the results. However, for a more holistic assessment we perform a qualitative ionic
liquid structure-property trend analysis to determine whether the designed results are
consistent with observed data. By comparing the thermal conductivity values (Table 3-14),
we note/C[C4mim]> ^[c4mim]pf6 ^ c4mim\Tf2N Since all of these ionic liquids have
the same cation (C4mim) but different anions (Tf2N, PF6" and BF4'), we can infer that
thermal conductivities of ionic liquids with BF4' anion are greater than those with PF6" and
Tf2N' anions. The design result is consistent with this observation as the optimal structure has
BF4' anion. The only base cation considered in this design problem is imidazolium. By
comparing the k values of different ionic liquids with the same base cation and same anion,
but different side groups (i.e. [C4mim] PF6 vs [Cemim] PF6 vs [Cgmim] PF6, and [C2mim]
Tf2N' vs [C4mim] Tf2N' vs [Cemim] Tf2N' vs [Cgmim] Tf2N' vs [Ciomim] Tf2N', we can see
that the contribution of alkyl side chain groups are not as high as that of anion, and there is
no uniform trend that is observed in relation to varying number of alkyl side chain groups.
Therefore, the optimal numbers of alkyl side chain groups relate to other requirements such
as the IL needing to be a liquid (i.e. Tm< 25 C for RTILs) and have relatively low viscosity.
Overall we can conclude that the design results are consistent with the observed experimental
structure-property trends of thermal conductivity data.
57


Table 3-14: Experimental thermal conductivity data
Ionic liquid T (K) k (WmV) Ref
PF6 315 0.145 [122]
H3Cv^v^Ns^n*ch PFe 315 0.146 [122]
/ ^ H3C\/VyVX/NsiJ'N^CM PF6 315 0.145 [122]
t \ H3C^N^N-Ch3 BF4 315 0.1968 [123]
H3C^/rvOSJ-CH3 Bp4 315 0.1847 [123]
H3CV'rC^N?CH3 Tf2N 315 0.1294 [124]
H3C^/hOSJ-CH3 Tf2N 315 0.1264 [124]
/ \ @ Vwx/Nx^N Tf2N 315 0.1263 [124]
H3CwwVN"ch Tf2N 315 0.12715 [124]
/ \ Tf2N 315 0.1299 [124]
58


3.3.3 Toluene-heptane separation
A common use of solvents in industrial processes is as a separating agent to isolate two liquid
components. This case study relates to the design of optimal ionic liquid to separate toluene
(aromatic) and n-heptane (aliphatic). Sulfolane (C4H8O2S) is a molecular solvent that is
commonly used for this purpose. The design objective is to find an ionic liquid that can
improve performance in comparison to sulfolane. One key requirement is to select an ionic
liquid with as low viscosity as possible since, viscous solvents are not ideal from the stand
point of industrial equipment design. The other requirement is to ensure that the designed
solvent is a room temperature ionic liquid (RTIL) as the process requires a liquid solvent. A
constraint on melting point, eqn. (3-56), is necessary to ensure design of RTILs only. Melting
point and viscosity of ionic liquids were calculated through the same methods proposed in
case study 1. A good separation solvent should have a high value for selectivity, eqn. (3-48),
and solvent power, eqn. (3-49), and low value for solvent loss, eqn. (3-50).
Selectivity: /? = (3-48)
Ya,s
Solvent power: SP = (3-49)
Ya,s
Solvent loss: SL = (3-50)
Ys,b
The three properties are a function of infinite dilution activity coefficients of the n-
heptane/toluene/IL solution. The activity coefficients are calculated using the UNIFAC
model (discussed in section 2.4) and the interaction parameters for ionic liquids were taken
59


from Roughton et a/.105 Another important consideration is that, the addition of IL to the
binary liquid mixture should result in the creation of two liquid phases. The appearance of
new phases in a multi-component system can be checked through the implementation of
necessary and sufficient conditions for phase stability. These conditions for a ternary system,
are shown in eqns. (3-51) and (3-52) were derived by Bernard et al. (1967). The activity
coefficients were again calculated using the UNIFAC method as discussed before.

dlny2\
dx2 )
dlnY2
3 dx3
< 0
(3-51)
J_ + dlny2\ + f J_ + dlny3\ dlny2 dlny3 ^ ^
,x2 dx2 J \x3 dx3 J dx3 dx2
(3-52)
The CAILD design problem expressed as an optimization model is shown in eqns. (3-53) to
(3-58).
It is worth mentioning that currently cost data is not available for ionic liquids as they are for
the most part not commercially produced and it is also difficult to utilize cost information
within a computer-aided molecular design framework. Hence cost was not considered for
minimization.
Objective function
fobj = max (P)
(3-53)
Constraints:
Ionic liquid Structural Feasibility
(3-54)
60


r| <65 cP
(3-55)
Tm < 298.15 k (3-56)
SL < 0.0065 (3-57)
SP > 0.3719 (3-58)
Results: The design statistics for this problem are summarized in Table 3-15. A total of 185
feasible IL structures were enumerated in subproblem 1 (structural constraints). Out of these,
27 ILs satisfied the physical property constraints (viscosity and melting point) in subproblem
2. Out of these, 1 ionic liquid satisfied the mixture property constraints, eqns. (3-57) and (3-
58). The optimal ionic liquid structure (1-methylpyridinium tetrafluoroborate) with the
highest selectivity is shown in Figure 3-8. The properties of the designed ionic liquid are
listed in Table 3-16. Finally, we verified whether the designed ionic liquid created two
phases when added to a hypothetical binary mixture consisting of 70% n-heptane (aliphatic)
and 30% toluene (aromatic). This was accomplished by solving eqns. (3-51) and (3-52) for a
range of ternary compositions (keeping n-heptane to toluene ratio constant). We identified
that a phase split occurs at solvent composition range of 0.4 to 0.9.
Table 3-15: Decomposition Approach: Subproblem Results
Subproblem 1: Number of ionic liquids (ILs) generated, 185
Subproblem 2: Number of ILs satisfying pure component properties, 27
Subproblem 3: Number of ILs satisfying mixture properties, 1
Subproblem 4: Optimal candidate, [Mpy]+[BF4]'
61


The structure of the optimal ionic liquid that satisfies all the constraints and has the
maximum selectivity (P) is shown in Figure 3-8. The properties of the designed ionic liquid
are listed in Table 3-16.
CH,

B~F
F
Figure 3-8: 1-methylpyridinium [BF4]"
Table 3-16: Design Results of the Optimal IL, 1-methylpyridinium [BF4]
Properties 1-methylpyridinium [BF4]" [BF4] Sulfolane
Melting point (K) 294.8 300.65
Viscosity (cP) 55.084 10.07
SL 0.006471 0.0065 (q)
SP 0.67193 0.3719 (t2)
P 87.262 6.8023 (t0
Analysis: Table 3-17 shows experimental selectivity values for the separation of aromatics
from an aromatic/aliphatic mixture using different ionic liquids. By comparing the
selectivity values for separation of benzene from benzene/heptane mixture we see that
P[hmim][BF4\ > P[hmim][PF6]- Since both of these ionic liquids have the same cation (hmim)
62


but different anions (PFr," and BF4'), we can infer that selectivity values of ionic liquid with
BF4' anions are greater than those with PF6" anions. By comparing the selectivity values of
[bmim] Tf2N' and [bmim] PF6", (same cation and different anions), for separation of
toluene/heptane mixture we can infer that selectivity values with PF6" anion are greater than
selectivity values with Tf2N' anion. Therefore we can conclude that among the anions used in
the design problem (Tf2N~, PF6~ and BF4~\ ionic liquids having BF4 anion should have the
highest selectivity towards aromatic compounds. The design result is consistent with this
observation as the optimal ionic liquid has BF4' anion. Similarly, by comparing the
selectivity values for separation of toluene/heptane mixture we can see that P [mmim] ten > P
[emim] ten > P [bmim] ten- This shows that increasing number of alkyl groups on the cation side
chain decreases selectivity. The design result (Figure 3-8) is consistent with this observation
also as there is only one methyl group (minimum needed to satisfy the cation Valence)
present in the cation side chain. With respect to cation, several studies have reported that
Pyridinium-based cations have higher selectivity than imidazolium-base cations for
aliphatic/aromatic separation. This is also consistent with our design results as the optimal IL
had Pyridinium-based cation. This trend analysis qualitatively validates the CAILD
methodology as well as the group interaction parameters (e.g. UNIFAC parameters provided
in Roughton et al.) used in the model.
63


Table 3-17: Experimentally Measured Selectivity values for Aromatic/Aliphatic
Separations
Solvent Separation T(C) P (Selectivity)
/ \ H3C ^ CH3 ltziN Toluene/heptane 40 29.8
H3C\/N^NCH Tf2N Toluene/heptane 40 22.2
H3CVvV0Ch3 Tf2N Toluene/heptane 40 16.7
H3CV^N^NCh PF6 Toluene/heptane 40 21.3
H3Cv\Vn^n^ PFe Benzene/heptane 25 8.20
H3C VvVNs^N -ch Bp4 Benzene/heptane 25 8.40
CH, A \ CH3 Toluene/heptane 40 32.8
64


These qualitative trends from experiments are consistent with our CAILD results. The
optimal design has a BF4' anion and has minimal number of alkyl side groups (note that the
selected methylpyridinium base cation needs at least one CH3 group; i.e. the minimum
possible number of groups to satisfy the valence requirement).
3.3.4 Naphthalene solubility
In this case study we consider the design of an ionic liquid solvent for the dissolution of
organic compound naphthalene. A molecular solvent having high solubility for naphthalene
and commonly used for its dissolution is chloroform. The measured solubility of naphthalene
in chloroform is 0.473 mole fraction.126 Our objective is to find an ionic liquid that has higher
solubility for naphthalene than chloroform. In addition, the ionic liquid needs to be an RTIL,
and have a reasonably low viscosity. The melting point and viscosity are calculated using the
same models described in the case study 1. The expression shown in eqn. (3-61) is invoked to
ensure solid-liquid phase equilibrium conditions, in order to determine the saturation
concentration of solute (i.e. solubility).115 The CAILD design problem expressed as an
optimization model is shown in eqns. (3-59) to (3-63).
Objective function
fobj = max (xfat)
(3-59)
Constraints
Ionic liquid Structural Feasibility
(3-60)
65


(3-61)
In x1 (l + In Yiat = 0
Tm V T '
7] < 65 cP (3-62)
Tm < 298.15 k (3-63)
Results: The design statistics for this problem are summarized in Table 3-18. A total of 185
feasible IL structures were enumerated in subproblem 1. Out of these, 27 ILs satisfied the
physical property constraints (viscosity and melting point) in subproblem 2. The optimal
ionic liquid structure (1-Butyl-3-ethylimidazolium [Tf2N]) with the highest solubility is
shown in Figure 3-9. The properties of the designed ionic liquid are listed in Table 3-19.
Table 3-18: Decomposition Approach: Subproblem Results
Subproblem 1: Number of ionic liquids (ILs) generated, 185
Subproblem 2: Number of ILs satisfying pure component properties, 27
Subproblem 3: Number of ILs satisfying mixture properties, 27
Subproblem 4: Optimal candidate, 1-Butyl-3-ethylimidazolium [Tf2N]
The structure of the optimal ionic liquid that satisfies the constraints is shown in Figure 3-9.
The optimal properties of the designed ionic liquid are shown in Table 3-19.
66


ch3
0 0
F3C-S-N-S-CF3
II II
0 0
Figure 3-9: l-butyl-3-ethylimidazolium [Tf2N]"
Table 3-19: Physical properties of 1-butyl-3-ethylimidazolium [Tf2N]
Properties l-Butyl-3-ethyl imidazolium [Tf2N] Chloroform
Melting point (K) 222.78 209.65
Viscosity (cP) 55.59 0.542
Naphthalene solubility @ 25 C 0.5069 0.473
An important component of green chemistry relates to the solvent medium in which synthetic
transformations are carried out. Traditional volatile organic solvents which act as
common reaction media for several chemical processes are linked to a host of negative
environmental and health effects including climate change, urban air-quality and human
illness. Jessop states that one of the major challenges in the search for environmentally
benign solvents is to ensure availability of green solvents as replacements for non-green
solvents of any kind. He uses the Kamlet-Taft plots to show that current list of green solvents
67


populate only a small region of the entire spectrum of solvents needed for various
applications and argues that large unpopulated areas of this diagram mean that future process
chemists and engineers need solvents having certain desirable properties and are green. Ionic
liquids offer great potential to satisfy this need.
This study presents an overarching framework that can be utilized to design optimal ionic
liquids for a given application through the theoretical/computational consideration of all
possible combinations. Currently, the few ionic liquid structure-property models that are
available can be applied to a small subset of all available ionic liquid types. However, for this
method to be fully effective, we need group contribution models and parameters that span the
entire spectrum of ionic liquids. It is indeed possible to overcome this challenge as one needs
property values of only few representative compounds in each class of ionic liquid (for
example, covering the groups shown in Appendix A) to regress the contributions of the
various groups. We propose that future research should focus on experimental property
measurements and data collection of ionic liquids that cover a diverse set of cations, anions
and functional groups. The second challenge is the lack of ionic liquid structure-property
models (i.e. solution to the forward problem) for various thermo-physical properties of
interest. There, is a great need to develop structure-property models of pure-component
physical properties and thermodynamic solution (mixture) properties for a comprehensive set
of ionic groups. The third challenge would relate to the accuracy of the group contribution
models. However, as discussed earlier, since the primary aim of the CAILD method would
be to narrow down to a small set of ionic liquids from the millions of available alternatives,
reasonably accurate predictive models are sufficient. The final IL can be selected by ab initio
68


computational chemistry calculation or experimental verifications of these small set of
designed compounds.
Progress towards designing ionic liquids through proposed CAILD, framework, will not only
contribute towards our understanding of the relationship between cation-anion structures and
ionic liquid properties, but will also provide a mechanism to engineer new environmentally
benign ionic liquids for critical applications.
69


Chapter 4: Application 1: Design of Ionic Liquids for Thermal Energy Storage
In this chapter, we present a computer-aided framework to design task-specific ionic liquids
(ILs) for solar energy storage, using structure-property models and optimization methods.
Thermal energy storage density (capacity) was used as a measure of the ability of an IL to
store thermal (solar) energy.
4.1 Introduction
Advancements in solar trough and solar tower technologies have enabled concentration of
thermal energy to the extent that it can be used to drive traditional steam cycles providing an
alternative to fossil fuel use. Therefore, harvesting solar energy using arrays of parabolic
trough collectors will enable generation of electricity at a large scale. In solar power plants,
thermal energy storage (TES) is necessary to extend production periods of low or no
sunlight. An important component of TES systems is thermal fluid which is needed to
transfer and store heat for relatively short periods. A parabolic trough system typically
consists of a series of collectors that are big mirror-like reflectors used to concentrate solar
energy. When the solar radiation is received by these collectors the reflected light is
concentrated at the center of the collector. A heat transfer fluid (HTF) passed through tubes
present at the center of the collector absorbs the accumulated heat. The collected thermal
energy is then transferred from the HTF to a storage medium or is stored in a reservoir using
the heat capacity of the HTF itself. The storage media can then release the thermal energy
when needed for further conversion to electrical energy. Thermal energy storage (TES)
therefore makes solar energy a more reliable and economic alternative source of energy.
70


Fluids that have high potential to store heat energy such as thermal oil (e.g. VP-1), or
nitrate salts (e.g. HITEC-XL) are suited for thermal energy storage applications. However,
nitrate salts have melting points (freezing point) above 200C while mineral oils have upper
temperature limit of 300C thus limiting their use to a narrow temperature range and
thereby reducing the overall efficiency of the process. Ionic liquids (and salts) have
properties that are ideal for thermal storage applications. These attractive properties include
high heat capacity, high decomposition temperature and relatively high density at operating
conditions. Ionic liquids (ILs) are a new generation of materials that that have a wide range
of applications.1 Similar to salts ILs are composed of ions but have much lower melting
points. Several ILs are in liquid state at room temperature (commonly referred as room
temperature ionic liquids [RTILs]). ILs consist of an organic cation (a cation base with alkyl
side chain) and a charge-delocalized inorganic or organic anion. They usually possess
good thermal stability (i.e. high decomposition temperature) making them appropriate for
processes operating at high temperatures. Ionic liquids can be customized through
appropriate selection of cations, anions and alkyl side chains. Therefore, ILs can be tuned to
impart specific functionalities for a given application by changing cation/anions/side chain
groups.
In this study, we focus on the computational design of optimal ionic liquids with high
thermal storage density for solar energy storage applications. The key requirements of a
thermal storage media include high thermal storage capacity (p. Cp [~r^]), high thermal
stability and a wide liquid range. Therefore the properties of ionic liquid that need to be
optimized for thermal storage applications include: density, heat capacity, thermal
71


decomposition temperature and melting point. Heat capacity measurements of diverse ionic
liquids reveal wide ranging values. The melting point of ionic liquids can be easily
adjusted by the choice of cation, anion or the groups attached to the cation side chains.
Thermal stability of ionic liquids has been previously studied and it has been reported that
many of them have high thermal decomposition temperatures (-400 C). In order for us to
find the optimal ionic liquid structure having desired melting point and decomposition
temperature as well as maximum thermal storage capacity, thousands of different ionic
liquids need to be tested. Since this is not feasible experimentally, a computer-aided
approach is suggested in this study.
Computer-aided molecular design (CAMD) is a promising technique that has been widely
used to design compounds for different applications. Gani and co-workers initially
conceptualized this method for screening solvents based on UNIFAC group contribution
approach. CAMD usually integrates structure based property prediction models (e.g. group
contribution models) and optimization algorithms to design molecular compounds with
desired properties. More recently, this approach has been extended to the design of ionic
liquids.140141142143144 A comprehensive framework for computer-aided ionic liquid design
(CAILD) was recently published by our group. Key to the successful development and use
of CAILD methods is the availability of predictive models for the properties of interest. In
this work we present a new CAILD model to design novel ionic liquids as thermal fluids for
solar energy storage applications. This CAILD model utilizes existing group contribution
methods to predict physical and thermal properties of ionic liquids. By considering a
72


structurally diverse set of building blocks we are able to demonstrate that new and novel
structural variants of ionic liquids can be tailored specifically for this application.
4.2 Formulation of the design problem
In this section, we focus on presenting a computer-aided ionic liquid design (CAILD) model
to find (design) optimal ionic liquid structures with high thermal storage capacity, reasonably
low melting point and high decomposition temperature. In this method a variety of cation
head groups, cation side chain groups (including certain functional groups), and anions were
selected as ionic liquid building blocks. Typical CAILD approach requires solution to the
forward problem (i.e. property prediction) as well as the reverse problem (i.e. structure
generation). In a mathematical programming based CAILD approach the physical properties
of ionic liquids are estimated using structure based predictive models such as group
contribution (GC) models (forward problem) and optimal ionic liquid structures are
generated by solving a mixed-integer non-linear programming (MINLP) formulation of the
design problem (reverse problem). This study utilizes GC methods from
literature145146,147148a148b to predict the physical properties density, heat capacity, melting
point and decomposition temperature. The CAILD framework proposed in Karunanithi and
Mehrkesh was utilized to formulate the thermal storage fluid (TSF) design problem as an
MINLP model. Structural constraints, eqns. (4-1) to (4-5), were included to design feasible
ionic liquid structures. These constraints are a sub-set of a comprehensive set of structural
142
constraints presented in Karunanithi and Mehrkesh.
73


Linear physical property constraints based on GC predication, eqns. (4-11) to (4-15), were
integrated with the structural feasibility constraints. The objective of the design problem was
to identify the optimal ionic liquid structure that has the highest thermal storage capacity.
Therefore, the objective function was formulated to maximize the product of density and heat
capacity of the ionic liquid. The solar thermal storage process is typically carried out at
temperatures of around 300C. Therefore, the thermal storage fluid -in this case the designed
ionic liquid- should be operable at temperatures slightly above 300C. To ensure that the
ionic liquid does not decompose during the process, we enforce a constraint on thermal
decomposition temperature to be above 400C, eqn. (4-24). The temperature of ionic liquid
after energy exchange should be higher than its melting point. To ensure this, a constraint on
melting point to be above 140C is imposed, eqn. (4-25). The temperature window between
melting point and decomposition temperature is the range at which the process can operate.
The basis set (the structural building blocks) considered for this problem included 5 cation
head groups, 9 anions and 5 side chain groups (alkyl and functional groups) which are shown
in Table 4-1. This selection was based on groups for which group contribution parameters
were available for all the properties of interest.
Table 4-1: Ionic liquid building blocks (groups) considered for thermal fluid design
Cation Anions Groups
Imidazolium / ^ R1 ^ R2 Tetrafluoroborate [BF4]" Methylene (CH2)
74


Cation
Pyridinium

N.
R
1
Anions
Hexafluorophosphate [PF6]"
Groups
Methyl (CH3)
Ammonium
I1
/f'K
R, \ R4
bis(trifluoromethylsulfonyl) imide
O O
F3C-S-N-S-CF3
M II
o o
Benzyl
Phosphonium
R,
R3 \
'R
R
Chloride
cr
Methoxy
^ ^CHo
Pyrrolidinium
Bromide
Br~
T rifluoromethanesulfonate
0
S-00
W
o
Benzoate
Nitrate
0
I!
LCK OJ
Hydroxyl (-OH)
75


Cation Anions Groups
Acetate
0
I H3C^^O'
Objective function
fob] = max (p.Cp)
Constraints
1) Structural (feasibility) Constraints
2) Tm < 140 C
3) Td > 400 C
4.2.1 Ionic liquid structural constraints
A key requirement in computational design of ionic liquids is the ability to guarantee
solutions that are theoretically feasible chemical structures. Further, one would have to
account for practical considerations such as limits on the size or presence/absence of certain
groups etc. that will lead to design candidates that are workable from a synthesis view
point. In order to incorporate these considerations the proposed method requires that the
designed compound satisfies certain rules, broadly termed as structural constraints, that
includes chemical feasibility rules such as octet rule, bonding rule and complexity rules as
76


well as rules that restrict/determine the size and the constituents of possible solutions. More
142
details about these structural constraints can be found in Karunanithi and Mehrkesh. The
specific structural constraints from Karunanithi and Mehrkesh that are invoked for this
problem are discussed below.
The first two constraints, eqns. (4-1) and (4-2), ensures the selection of only one cation and
one anion from the basis set.
M m II (4-1)
(4-2)
Where C is a set of all cation head groups considered (i.e. imidazolium, pyridinium,
ammonium, phosphonium and pyrrolidinium); and A is a set of all anions considered. The
next set of feasibility constraints deal with valence requirements of cations which enables us
to add appropriate side chain groups. Per our definition, anion is considered as a single group
and does not require constraints related to addition of groups and therefore valence
constraints are relevant only for the cation part. The three equations below make sure that
octet rule for the cation as a whole as well as for each side chain (branch) is not violated.
Ifii yi=hecCiVci (4-3)
Iiec(2 vci)Ci + 2=i Zfcec(2 - = 2 (4-4)
Zfcec Jingki (2 vGkl) = 1 (4-5)
77


Where we fixed n to be 2, 1, 4, 4, and 2 for imidazolium, pyridinium, ammonium,
phosphonium and pyrrolidinium respectively. This was based on the fact that for
imidazolium, pyridinium and pyrrolidinium the side chains are commonly connected to only
the nitrogen atoms in the cation ring. However, note that if we want to broaden the design
problem we can allow side chains to be attached to the carbon atoms in the cation rings by
adjusting these constraints. More specific details and definitions about these constraints can
be found in Karunanithi and Mehrkesh. The next set of constraints deal with restricting the
size of cation as well as putting limits on the presence/absence of certain groups. Limits on
the total number of groups that can occur on each side chain as well as the whole cation was
fixed using eqns. (4-6) and (4-7). These two constraints make sure that we do not design a
very large cation, which cannot be practically synthesized. Limits on the number of
functional groups with valence 1 (i.e. OH, OCH3, and benzyl) that can occur in the whole
cation was fixed using eqn. (4-8).
I?=i lkecyingki < (n* X 16) (4-6)
IkecVingu < 16 (4-7)
liLilkeG^ingn < t (4-8)
Where, G is a set of all groups considered (i.e. CH2, CH3, OH, OCH3, and benzyl) and G1 is
a subset of G which consist of functional groups OH, OCH3, and benzyl. Eqn. (4-8) is
invoked three times for each k G G1 while t1 for each of these three equations was assigned
78


to be 2,1,2,2,2 for imidazolium, pyridinium, ammonium, phosphonium and pyrrolidinium
head groups respectively.
4.2.2 Ionic liquid property prediction
This section describes in detail the physical properties of ionic liquids that are relevant for
thermal storage applications and the structure based property prediction models and
correlations that are utilized to predict these properties within the proposed computer-aided
ionic liquid design (CAILD) framework.
Thermal storage density
Thermal storage density (t] [~r^:]) of an ionic liquid is defined as a product of its heat
capacity and density.
r\ = p.Cp (4-9)
Since both density and heat capacity are a function of temperature, thermal storage densities
of ionic liquids vary with temperature, t] is the most critical design parameter for thermal
fluids as higher value of t] will result in lower volume of thermal fluid requirement.
Heat Capacity
The heat capacities of ionic liquids were predicted using the approach developed by
Valderrama et al. In this method the authors used the concept of mass connectivity index
(MCI) to build a structure based predictive model. Molecular connectivity was first
79


147
introduced by Randic and has been used then by several authors for property prediction.
As suggested by Valderrama et al. the MCI concept can be used to quantify the extent of
branching in ionic liquids thereby enabling us to predict IL heat capacities better. This index
considers the mass of structural groups as well as the type of connections between them as
following:
A =
(4-10)
where, m, and mj are the mass of neighboring groups i and j in a molecule. In this expression,
the sequence of groups is important, as i and j are two distinct groups and hence the
connection m j in, is different from mjm;. Valderrama el al. showed that the MCI approach
is capable of predicting the heat capacity as a function of temperature for a variety of ionic
liquids with an acceptable level of accuracy. To predict the ionic liquid heat capacity, a
reference value of Cpo at reference condition T0 is used as follows
CP = CP0+A\p(T-T0) + q(T-T0)2] (4-11)
Here p and q are constants specified for each ionic liquid and have been correlated from
experimental data of ionic liquid heat capacities. The overall Cp. as a temperature dependent
variable, can be estimated as a function of molar mass of cation and anion, molar volume of
ionic liquid and mass connectivity index as follows
Cp = a, + bVm + cA + dr] + A[e(T Tg) + f (T2 T^)] (4-12)
where, T0 is the reference temperature, 298.15 K, a = 15.80, b = 1.663, c = 28.01, d = -7.350,
80


5 3
e = 0.2530, f = -1.372x10' ,Vm is the molar volume (cm /mol), X is the connectivity index
(MCI), and rj = of ionic liquid.
Density
Density of ionic liquids were predicted using the approach presented by Valderrama et al.145
In this approach the density of each ionic liquid at a given temperature was estimated as a
secondary property using its critical properties (Tc and Vc) and normal boiling point (Tb) as
primary properties through the following expression, eqn. (4-13):
P =
AlnB
B
}(
T-Tb
Tc~Tb
)
(4-13)
where, A = a + ,B = (- + -)VCS, a= 0.3411, b= 2.0443, c=0.5386, d = 0.0393, 5=
Vc yVcMJC
1.0476.
Critical properties and normal boiling point of the ionic liquids were estimated using the
group contribution method proposed by Valderrama et al.146
Melting Point
An ionic liquid with low melting point would be desirable since at all times the operating
temperature should be kept above the melting point of the thermal storage fluid to avoid solid
formation in the system. Alternately, when the melting point is too high a high process
temperature needs to be maintained which will decrease the system efficiency by reducing
81


the rate of heat transfer (sensible heat). An appropriate thermal storage fluid should possess a
melting point lower than 140 C. To calculate the melting point of ionic liquids a group
contribution method proposed by Lazzus148b was used. In this approach two different sets of
contribution for melting point were used: 1) the contribution of cation head group (cation
base) and alkyl groups/functional groups attached to the side chains of the cation head
groups; 2) the contribution of groups associated with the anion. Cation head groups are
tabulated as whole (e.g. imidazolium or pyridinium) but side chain groups and anions are
split into smaller structural fragments. The melting point of any given ionic liquid can be
calculated by the summation of contribution of cations, anions and side chain groups as
following:
Tm(k) = 288.7 + £ rijAtCj + £ njAtaj (4-14)
where rij is the number of occurrence of group i, Atq is the contribution of cation to the
melting point and Ata,j is the contribution of anion to the melting point of the given ionic
liquid.
Thermal Decomposition Temperature
The final property of interest is thermal decomposition temperature. This property is an
estimate of the highest temperature at which the ionic liquid will remain in the associated
state (non-decomposed). It is important for thermal storage as it will determine the maximum
applicable temperature at which the thermal fluid can be utilized. To predict the
82


decomposition temperature of ionic liquids a group contribution method proposed by
Lazzus148a was utilized.
This method is similar to the melting point prediction described above.
Td(k) = 663.85 + £njAtq + £rijAtaj (4-15)
where rij is the number of occurrence of group i, Atq is the contribution of cation to the
decomposition temperature and Ata,j is the contribution of anion to any given ionic liquid.
4.2.3 CAILD model solution
The complete CAILD-MINLP model is shown below:
Objective function
fob] = max (p.Cp)
Constraints
Z iec ci = 1 (4-16)
ljeAai = l (4-17)
ifiiyi = hecCiVd (4-18)
E;ec(2 vci)Ci + 2=i Zfcec(2 Vckdyingu = 2 (4-19)
Zfcec Yingki (2 vGkl) = 1 (4-20)
83


liilkecyingu < (n* X 16)
(4-21)
Ikecyingu < 16 (4-22)
lkEG1 yingki < h (4-23)
Td(k) = 663.85 + £ rijAtq + £ njAtaj > 400 C (4-24)
Tm(k) = 288.7 + £ rijAtCj + £ njAtaj < 140 C (4-25)
p = + (£){li!i£}(l^A) (4-26)
^ B K7JL B }KTc-TbJ v y
4 = a + (4-27)
5 = (^ + ^/ (4-28)
a = 0.3411, b = 2.0443, c = 0.5386, d = 0.0393,6 = 1.0476
Vc(cm3/mol) = 6.75 + 'ZniAVci (4-29)
Tb(K) = 198.2 + Ei nf AT*, (4-30)
7(7W = b/[A+BZiniATc a:iniATc)2] (4'31)
A = 0.5703,B = 1.012
Cp a + £>1^ + cA + dr] + 2.[e(7 7q) + / (T2 7)^)] (4-32)
84


T0 = 298.15 K,a = 15.80,b = 1.663,c = 28.01,d = -7.350,e = 0.2530,/ = -1.372 X 10-5
(4-33)
(4-34)
Myyani
(4-35)
The above computer-aided ionic liquid design (CAILD) model is a Mixed Integer Non-
Linear Programming (MINLP) formulation. MINLP models combine combinatorial aspects
with nonlinearities and are more difficult to solve than mixed integer programming (MIP)
and non-linear programming (NLP) problems. The most precise approach to solve this
MINLP model would be to fully enumerate each possible combination within the entire
search space. The number of possible combinations increase exponentially with number of
groups considered resulting in combinatorial explosion requiring large computational times.
Other multi-level approaches have been proposed to address this issue and avoid complete
enumeration.149,150 Several deterministic optimization based methods have been employed to
solve CAMD-MINLP models.137,138,151 Different stochastic methods have also been used to
solve CAMD problems.152,153,154 Property clustering technique used within a reverse problem
formulation is yet another approach that has been successfully used to solve CAMD
problems.155,156 In this work a genetic algorithm (GA) based solution approach is utilized to
solve this optimization model. GA is a stochastic approach that mimics natures process of
biological evolution (principle of natural selection) and has been previously used to
successfully solve complex optimization (minimization) problems of different formulations.
85


GA can be an attractive alternate solution approach for problems that are not well suited for
standard optimization algorithms such as problems with discontinuous, stochastic, or highly
nonlinear objective functions. The optimization process usually starts with a random
collection of initial candidates from which the fitter candidates are selected as parents and
can exchange their genetic information, through crossover and mutation operations. At each
iteration, a new population of fitter candidate structures is generated to replace the existing
population and this process is repeated for a pre-specified number of iterations or until the
pre-defined value of tolerance for objective function is met. The GA solution of CAILD-
MINLP problem was implemented in the MATLAB environment with most parameters set at
their default values. A large population size which was generated randomly by the program
was used to start the search process. The crossover fraction was fixed at 0.8, whereas a
uniform mutation probability with rate value of 0.01 was used.
4.3 Results and discussion
The optimal ionic liquid consists of a hydroxyl-functionalized imidazolium cation and
tetrafluoroborate anion whose structure is shown in Figure 4-1.
F
L
B------F
I
F
Figure 4-1: A schematic of the structure of optimal IL with highest thermal storage
capacity
HO
/=\
-N^NH* F.
86


Full Text

PAGE 1

OPTIMIZATION BASED DESIGN AND ANALYSIS OF TAILOR MADE IONIC LIQUIDS By AMIRHOSSEIN MEHRKESH B.S. Isfahan University of Technology 2006 M.S. University of Isfahan, 2009 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Civil Engineering 2015

PAGE 2

ii This thesis for the Doctor of Philosophy degree by Amirhossein Mehrkesh h as been approved for the Civil Engineering Program by Arunprakash T. Karunanithi, Advisor Kannan Premnath Chair Azadeh Bolhari Indrani Pal Fernando Rosario Ortiz 2 1 November 2015

PAGE 3

iii Mehrkesh Amirhossein (PhD Civil Engineering ) Optimization based Design and Analysis of Tailor made Ionic Liquids Thesis directed by Associate Professor Arunprakash T. Karunanithi ABSTRACT Solvents comprise two thirds of all industrial emissions. Traditional organic solvents easily reach the atmosphere as they have high vapor pressure and are linked to a host of negative environmental effects including climate change, urban air quality and human illness. Room temperature ionic liquids (RTIL), on the other hand, have low vapor pressure and are not flammable or explosive, thereby resulting in fewer environmental burdens and health hazards. However, their life cycle environmental impacts as well as freshwater ecotoxicity implications are poorly understood. RTILs are molten salts that exist as liquids at relatively low temperatures and have unique properties. Ionic liquids consist of a large organic cation and charge delocalized inorganic or organic anion of smaller size and asymmetric shape. The organic cation can undergo unlimited structural variations through modification of the alkyl groups attached to the side chain of the base cation skeleton and most of these structural variations are feasible, from chemical synthesis point of view, due to the easy nature of preparation of their component s. Functionally, ionic liquids can be tuned to impart specific desired properties by switching anions/cations or by incorporating functionalities into the cations/anions. It is estimated that theoretically more than a trillion ionic liquid structures can b e formed. Due to their tunable nature, these molten salts have the potential to be used as solvents for variety of applications. This dissertation presents a computer aided IL design (CAILD) methodology with an aim to design optimal task specific ionic li quid structures for different applications. We utilize group contribution based ionic liquid property prediction models within a mathematical

PAGE 4

iv programming framework to reverse engineer functional ionic liquid structures. The CAILD model is then u tilized to design optimal ionic liquids for solar energy storage, as a solvent for aromatic aliphatic separation and as an absorbent for carbon capture process. Using the developed CAILD model, we were able to computationally design new ionic liquid structures w ith physical and solvent properties that are potentially superior to commonly used ILs. The accuracy of the developed model was back tested and verified using available experimental data of common ILs. However, we would like to note that the computational design results from this dissertation needs to be experimentally validated. This dissertation also developed ecotoxic i ty characterization factors for few common ILs. The developed characterization factors (CFs), can be used in future studies to perform h olistic (cradle to grave) life cycle assessment s on processes using ILs to understand their environmental and ecological impacts. The form of and content of this abstract are approved. I recommend its publication. Approved: Arunprakash Karunanithi

PAGE 5

v DEDICATION This dissertation is dedicated to my brilliant and outrageously loving and supporting wife, Anna and to my always encouraging, ever faithful mother, Soraya, and to the memory of my late father, Eskandar who t a ught me how to live a peaceful and happ y life the person who will be missed forever.

PAGE 6

vi ACKNOWLEDMENTS I am grateful to have had Dr. Arunprakash Karunanithi as my advisor. Without h is knowledge, guidance support and enthusiasm towards this research, I would not have been able to complete this dissertation H e has taught me to be optimistic, persistent and confident in the work I am doing He provided me all the tools needed to accomplish this research from financial support, to computer software to grants for attending co nferences. I also would like to thank the faculty, staff and my friends ( fellow graduate students in our research group) whom I interacted with during my graduate program at the University of Colorado Denver. I also want to exten d my acknowledgment to Dr. Azadeh Bolhari, Dr. Mike Tang and Mr. Eric Ziegler who helped me in editing and proofreading. I also would like to thank all my committee members for their constructive suggestions and feedback

PAGE 7

vii TABLE OF CONTENTS Chapter Introduction ................................ ................................ ................................ ........................... 1 1 1.1 Computer aided ionic liquid design (CAILD) ................................ .............................. 5 1.2 Current applications ................................ ................................ ................................ ...... 9 1.3 Environmental impacts of ionic liquids ................................ ................................ ...... 10 1.4 Ionic liquids safety ................................ ................................ ................................ ...... 11 Forward Problem, Prediction of Melting Point and Viscosity of ILs ................................ 13 2 2.1 Introduction ................................ ................................ ................................ ................. 13 2.2 Methods ................................ ................................ ................................ ....................... 16 2.3 Results and discussion ................................ ................................ ................................ 19 2.3.1 Melting point ................................ ................................ ................................ ....... 21 2.3.2 Viscosity ................................ ................................ ................................ ............. 24 Reverse Problem; Computer aided Design of Ionic Liquids ................................ .............. 27 3 3.1 Introduction ................................ ................................ ................................ ................. 27 3.2 Computer aided ionic liquid design (CAILD) ................................ ............................ 28 3.2.1 Mathematical framework ................................ ................................ .................... 31 3.2.2 Ionic liquid structural constraints ................................ ................................ ........ 32 3.2.3 Physical property constraints ................................ ................................ .............. 38 3.2.4 Solution property constraints ................................ ................................ ................. 39 3.2.5 Solution of the underlying MINLP ................................ ................................ ..... 42 3.3 Proof of concept examples ................................ ................................ .......................... 45 3.3.1 Electrolytes ................................ ................................ ................................ ......... 46 3.3.2 Heat transfer fluids ................................ ................................ .............................. 54 3.3.3 Toluene heptane separation ................................ ................................ ................ 59 3.3.4 Naphthalene solubility ................................ ................................ ........................ 65 Application 1: Design of Ionic Liquids for Thermal Energy Storage ................................ 70 4 4.1 Introduction ................................ ................................ ................................ ................. 70 4.2 Formulation of the design problem ................................ ................................ ............. 73 4.2.1 Ionic liquid structural constraints ................................ ................................ ........ 76 4.2.2 Ionic liquid property prediction ................................ ................................ .......... 79 4.2.3 CAILD model solution ................................ ................................ ....................... 83 4.3 Results and discussion ................................ ................................ ................................ 86

PAGE 8

viii Application 2: Design of Ionic Liquids for Aromatic Aliphatic Separation ...................... 96 5 5.1 Introduction ................................ ................................ ................................ ................. 96 5.2 Computer aided ionic liquid design (CAILD) ................................ ............................ 98 5.2.1 Forward problem ................................ ................................ ................................ 99 5.2.2 Reverse problem ................................ ................................ ............................... 105 5.3 Results ................................ ................................ ................................ ....................... 110 Application 3: Design of Ionic Liquids for CO 2 Capture ................................ ................. 115 6 6.1 Introduction ................................ ................................ ................................ ............... 115 6.2 Forward problem ................................ ................................ ................................ ....... 116 6.3 Reverse problem ................................ ................................ ................................ ....... 121 6.3.1 Case study ................................ ................................ ................................ ......... 121 6.4 Results ................................ ................................ ................................ ....................... 124 Life Cycle Environmental Implications of Ionic Liquids ................................ ................. 128 7 7.1 Life Cycle Perspectives on Aquatic Ecotoxicity of Common Ionic Liquids ............ 128 7.2 Methods ................................ ................................ ................................ .................... 131 7.2.1 Goal, Scope, System Boundary ................................ ................................ ........ 131 7.2.2 Life Cycle Inventory of Ionic Liquid Production and Data Sources ................ 131 7.2.3 Fresh Water Ecoto xicity Impacts of Ionic Liquid Production .......................... 132 7.2.4 Development of Ecotoxicity Characterization Factors for Ionic Liquids ......... 133 7.2.5 Fresh Water Ecotoxicity Impact of Direct Release of Ionic Liquids ................ 139 7.2.6 Uncertainty ................................ ................................ ................................ ........ 140 7.3 Results and Discussion ................................ ................................ ............................. 141 7.4 Life cycle assessment of energetic ionic liquids ................................ ....................... 148 7.4.1 Process and Energetic requirements for triazolium nitrate and TNT synthesis 152 7.4.2 Life cycle assessment (LCA) of energetic ionic salts ................................ ....... 157 7.4.3 Results and discussion ................................ ................................ ...................... 160 Conclusions and Future Work ................................ ................................ .......................... 169 8 8.1 Limitations and Recommendations ................................ ................................ ........... 172 8.2 Contributions ................................ ................................ ................................ ............. 175 8.3 Future work on ionic liquid applications ................................ ................................ .. 177 Refere n ces ................................ ................................ ................................ ............................. 182 Appendix A A comprehensive list of IL structural groups used in CAILD ................................ ......... 208 B IL structural Groups used in CAILD for Aro/Ali separation ................................ ........... 211

PAGE 9

ix C UNIFAC parameters ................................ ................................ ................................ ........ 212 D IL structural Groups used in CAILD for CO 2 capture ................................ ..................... 214 E Life Cycle Inventory of Ionic Liquids Production ................................ ........................... 215

PAGE 10

x LIST OF TABLES Table 2 1 Model characteristics in prediction of melting points (calculated on validation data ................................ ................................ ................................ .................. 23 2 2 Model characteristics in prediction of viscosity (calculated on validation data) ................................ ................................ ................................ ......................... 26 3 1 Cation and alkyl side chain groups valence s ................................ .................. 35 3 2 Values of and for 1,3 diethylimidazolium tetrafluoroborate ........... 36 3 3 The basis set used for ionic liquid design ................................ ....................... 45 3 4 Group contributions for parameters A and B ................................ ............... 47 3 5 Group contributions for parameters A, B and V ................................ ............. 48 3 6 values for different cations ................................ ................................ ........ 49 3 7 ................................ ... 49 3 8 Decomposition approach: Subproblem Results ................................ .............. 51 3 9 Design Results of the optimal IL, 1 methylimidazolium [Tf 2 N] .................... 51 3 10 Experimentally Measured Electrical Conductivities of Ionic Liquids ............ 53 3 11 Group contributions for parameters A k and B k ................................ ............... 54 3 12 Decomposition approach: Subproblem Results ................................ .............. 56 3 13 Design Results of the Optimal IL, 1 ethyl 3 methylimidazolium [BF 4 ] ........ 56 3 14 Experimental thermal conductivity data ................................ ......................... 58

PAGE 11

xi 3 15 Decomposition Approach: Subproblem Results ................................ ............. 61 3 16 Design Results of the Optimal IL, 1 methylpyridinium [BF 4 ] ...................... 62 3 17 Experimentally Measured Selectivity values for Aromatic/Aliphatic Separations ................................ ................................ ................................ ...... 64 3 18 Decomposition Approach: Subproblem Results ................................ ............. 66 3 19 Physical properties of 1 butyl 3 ethylimidazolium [Tf 2 N] ............................ 67 4 1 Ionic liquid building blocks (groups) considered for thermal fluid design .... 74 4 2 Thermo physical properties of VP hydroxy Imidazolium] + [BF 4 ] ................................ ................................ ...................... 87 4 3 Effect of anion variation on the thermal storage properties of ionic liquids .. 88 4 4 Effect of number of CH 2 m and T app [ILs with BF 4 anion] ...... 90 4 5 The effect of variation of functional groups (FG) connected to the m and T app ................................ ............................... 92 4 6 Comparison of COSMO predicted C p corresponding exp. data ................................ ................................ .................. 94 4 7 Comparison between COSMO predicted and GC predicted values for C p and ................................ ................................ ............... 95 5 1 Experimental infinite dilution activity coefficients ( ) vs. Cosmo predicted values ................................ ................................ ............................ 103 5 2 Experimental solubility data vs. CAILD predicted data ............................... 104 5 3 CAILD results for the optimal ILs and furfural at T=330 K ........................ 111 5 4 Physical properties and solvency power of optimal ILs and furfural ........... 112 5 5 A schematic of the structure of optimal ILs ................................ .................. 113

PAGE 12

xii 6 1 A comparison between experimental and COSMO constant ................................ ................................ ................................ ......... 119 6 2 Experimental and UNIFAC predicted values of CO 2 solubility in different ILs 120 6 3 Name, symbol and structure of the optimal ILs ................................ ............ 125 6 4 Pure (physical) and mixture properties of the optimal ILs ........................... 126 7 1 Toxicity values of selected Ionic Liquids ................................ ..................... 135 7 2 Environmental Properties of the studied Ionic Liquids ................................ 138 7 3 USEtox based effect factors, fate factors, exposure factors, and characterization factors for different ionic liquids ................................ ....... 141 7 4 Breakdown of energy and material related ecotoxicity impacts of IL .......... 143 7 5 Breakdown of freshwater ecotoxicity impacts of ILs associated with use phase release ................................ ................................ ................................ ........... 146 7 6 Impact of ionic salt and TNT (functional unit: 1 MJ energy content) .......... 161 7 7 Sensitivity analysis ................................ ................................ ........................ 165 7 8 Environmental Impact for scenario 1 ................................ ............................ 167 7 9 Environmental Impact for scenario 2 ................................ ............................ 167

PAGE 13

xiii LIST OF FIGURES Figure 1 1 A schematic of an ionic liquid, [Bmim] BF 4 ................................ ................... 2 2 1 A correlation between van der Waals (VdW) and experimental Radii .......... 20 2 2 Actual vs. model predicted melting points for training and test data sets ...... 22 2 3 Goodness of the model for predicting melting points of the selected ionic liquids ................................ ................................ ................................ .............. 23 2 4 Actual vs. model predicted viscosities for training and test data sets ............. 25 2 5 Goodness of the model for predicting viscositi es of the selected ionic liquids ................................ ................................ ................................ ......................... 25 3 1 CAILD framework ................................ ................................ .......................... 29 3 2 Conceptual description of CAILD ................................ ................................ .. 30 3 3 A general Scheme of two feasible cation structures ................................ ....... 37 3 4 Examples of feasible and non feasible Ionic liquids ................................ ...... 38 3 5 Different types of group interactions involved in solutions containing ionic 39 3 6 1 methylimidazolium [Tf 2 N] ................................ ................................ ......... 51 3 7 1 ethyl 3 methylimidazolium [BF 4 ] ................................ .............................. 56 3 8 1 methylpyridinium [BF 4 ] ................................ ................................ .............. 62 3 9 1 butyl 3 ethylimidazolium [Tf 2 N] ................................ ................................ 6 7 4 1 A schematic of the structure of optimal IL with highest thermal storage capacity ................................ ................................ ................................ ........... 86

PAGE 14

xiv 5 1 A schematic of the single stage extraction process ................................ ....... 106 6 1 A schematic of single stages C O 2 absorption desorption processes ............. 122 7 1 Ecotoxicity impacts related to production and use phase release of ILs ............ 142 7 2 a) 1,2,3 triaolzium nitrate; b) TNT ................................ ................................ 151 7 3 Material and energy flows associated with the life cycle tree for producing the ionic salt 1,2,3 triazolium nitrate ................................ ................................ .. 156 7 4 Comparison of scaled impacts of ionic salt and TNT (functional unit of 1MJ energy content): GWP (Global Warming Potential), AP (Acidification Potential), EP (Eutrophication Potential), HH (Human Health). .................. 163 7 5 Sensitivity analysis of the scaled impacts of ionic salt and TNT (functional unit of 1 MJ): GWP (Global Warming Potential), AP (Acidification Potential), EP (Eutrophication Potential), HH (Human Health). .................. 165

PAGE 15

1 Introduction Chapter 1: Ionics liquids (ILs) are an emerging new c lass of chemicals that show tremendous promise in creating customized designer compounds (solvents, electrolytes, energy storage media which can be used for new applications or to replace current materials that lack flexibility or 1 Ionic liquids are normally comprised of a large organic cation with positive charge and a charge delocalized organic o r in organic anion of smaller size (can be monoatomic such as Cl ) and asymmetrical shape. 2 The molecules possess a strong positive and negative charge which lends to its name as an ionic liquid. F irst ionic liquid triethylammonium nitrate was discove red more than a century ago. 3 Compared to the case of naturally occurring ionic salts (e.g. Na + Cl ), t he larger size of cations and anions in ionic liquids will result in distribution of a small charge (+1 or 1) over a much larger surface area This fact along with the asymmetri c nature of cations and anions, explain the lower melting points of ionic liquids. Ionic liquids (ILs) are salts that normally melt at 100C or less. 1,4,5 A schematic of an ionic liquid, 1 Butyl 3 methylimidazolium tetrafluorobora te, [Bmim] BF 4 is shown in Figure 1 1. The cation head group, imidazolium, with positive charge (+1) is shown in green, side chain groups attached to the cation base, butyl and methyl are shown in yellow and anion with negative charge ( 1) is shown in bro wn.

PAGE 16

2 Figure 1 1 : A schematic of an ionic liquid, [Bmim] + BF 4 Interest in ionic liquids has continued to build in the academia and industry due to their interesting tunable properties and potential to provide environmentally friendly alternative to volatile organic compounds (VOCs) currently used in chemical/industrial processes. The properties of ionic liquids (pure physical properties such as viscosity and mixture properties such as solve ncy power) vary enormously as a function of their molecular structure i.e. the type of cation base, anion and number/type of side chain alkyl/functional groups present in the structure ILs also offer a wide window of liquid state making them attractive a s liquid solvents since they normally have high boiling points and very low vapor pressures. ILs are rarely flammable or explosive, thereby presenting fewer environmental burdens and health hazards. During the past few years considerable effort has been devoted to identifying and understanding ILs that have superior properties The desirable properties of ionic liquids include:

PAGE 17

3 Negligible vapor pressure Ability to dissolve organic, inorganic, and polymeric materials High thermal stability (i.e. they do not decompose over a large temperature range ) Versatile and customizable for task specific applications Nonvolatile and rarely flammable or explosive Strong regeneration properties that allow for their reuse and recycl e Room temperature ionic liquids (RTILs) can be fluid at temperatures as low as 96 C Liquid phase temperature range from 96 C to 300 C ; thermally stable up to 200 C Moderate to high electrical conductivity These properties of ionic liquids (ILs) make them attractive as potential alternatives to current chemical compounds Ionic liquids can be tailor made for different applications (task specific ionic liquids) by varying the building blocks of ionic liquids (i.e. cations, anions or groups attached to t he cation base). Thus Ionic liquids present a fantastic opportunity to step away from the status quo of utilizing volatile organic compounds, which are caustic to the environment, as solvents in current chemical processes. As the continued push towards env ironmental ly conscious decisions at all level of industry continues, ionic liquid s have the necessary properties and customizability to deliver better alternatives with reduce d environmental impact s

PAGE 18

4 Ionic liquids can easily be removed from water. The abi lity to separate solvents from water is a critical property as industrial water pollution is increasingly becoming a major issue Tolerance towards ecologically unsound industrial practices are diminishing on a global scale as the anthropogenic effects on the planet become s more evident every year. Ionic liquids have the potential to reduce the overall cradle to grave environmental/ ecological impacts of current processes by offsetting upstream pollutant release from energy use and by products which are man ufactured through the VOC creation process. Ionic liquids have the potential of lower unusable by products to designed product ratio. Their regenerative properties further their green profile as their potential for reuse and recycl e exceeds those of curren tly used VOCs. As r egulatory oversight of emission of chemicals released into water and the atmosphere continues to tighten, finding alternat ives will reduce the economic burden on industry. Industry has a strong concern towards negative externalities that result from their economic activity. In addition, adoption of environmentally friendly technologies can lead to greater acceptance from general public and can reduce the opportunity of public outcry or protest It is far superio r to develop alternatives now, than to wait for a forced decision. Design of alternate ionic liquids can be made easy if we know how different structural group s present in them will influence the properties of interest. F or example, if we are interested i n an ionic liquid with high solvency power to wards a specific chemical, we need to know which cations or anions contribute to higher values of the solvency power towards that compound. After identifying the best cations and anions, addition of functional /a lkyl

PAGE 19

5 groups to the side chain of the cation base can further help us fine tune the desirable properties (e.g. relatively lower values of melting point or low viscosity). In the next section we discuss how computer models can help us expedite the process of selecting optimal ionic liquids for different applications. A computer based ionic liquid design model can reduce the enormous number of experiments needed to find the optimal candidate thereby saving time and money 1.1 Computer a ided i onic l iquid d esign (CAILD) The vast number of combinations of ionic liquids (ILs) is what provides their versatility and customization properties (estimated to be as many as 10 14 ionic liquids feasible) I onic liquids a re still considered as a new generation of chemical s, which are garnering attention from academia and industry Therefore, there is l imited information on the properties of less common ionic liquids in chemical libraries and databases. Without the necessary information, random synthesis of ionic liquids and testing of their properties is costly and time consuming. Computer aided molecular design (CAMD) is a promising approach that has been used for mo lecular systems to design compounds (e.g. solvents) for a variety of applications. 6 12 CAMD method integrates property prediction models and optimization algorithms to reverse engineer molecular structures with unique properties of interest Due to the fac t that ionic liquids are made of replaceable building blocks (structural groups), w e believe that a similar approach is even more relevant for designing tailor made ionic liquids 13

PAGE 20

6 Use of computer aided models to design optimal compounds, reduces the cost s and allows engineers to model a multitude of potential candidates for a specific application. We propose that CAMD approach can be adapted specifically towards ionic liquid design [ Computer aided Ionic Liquid Design (CAILD) ] w h ere we take into account cations, anions and functional groups attached to cation core 13 16 In this dissertation, we successfully show that the CAILD model is capable of creating ionic liquids with optimal desired properties for different application s (e.g. an ionic liquid with h igh thermal storage capacity ( ) can be a good candidate for a solar thermal storage process ). Based on the above discussion it is clear that in order to find optimal ionic liquids for different applications using a computer aided design framewor k, we need to know how different structural groups in an ionic liquid will contribute to wards the properties of interest. For example, consider a situation where we want to design a good solvent to remove toluene from a multicomponent mixture. I f we kn ow that imidazolium cation base usually results in higher values of solvency power towards toluene compared to the other cation bases, then an ionic liquid with imidazolium cation should always be chosen as the optimal ionic liquid unless it violates ot her physical properties or process constraints (such as the selected ionic liquid has an unacceptably high melting point or viscosity values). In order for us to be able to design an optimal ionic liquid for an application of interest, we need to make sur e that first it is a theoretically feasible chemical structure (chemical feasibility constraints ) and secondly the designed ionic liquid meet s other process criteria necessary for it to be used in large industrial scales. Therefore, the optimization framew ork

PAGE 21

7 needs an objective function (the value of the property of interest) that needs to be minimize d or maximize d along with a set of constraints that should be satisfied to guaranty that the designed ionic liquid is a feasible candidate. A good example of the type of constraints needed in a CAILD model relates to the design of an ionic liquid which is liquid at room temperature Here a constraint of T m be enforced within the optimization model Based on the above discussion it is clear that we need models capable of predict ing different properties of ionic liquids based on the type and number of structural groups present in them. 17 24 These models commonly referred to as group contribution (GC) models have been to some extent developed for ionic liquids. The GC models can be used within a CAILD framework to enabl e predict ion of physical properties of the ionic liquids during the design process. Without comprehensive group contribution models capable of predicting different properties of ionic liquids, it is not possible to utilize the power of CAILD models to their full extent. In other words CAILD mo dels are most useful when they have the capability of exploring all possible combinations of ILs to wards find ing the optimal candidate for a given application and this is not possible unless we have group contribution models covering all cations, anions an d side chain groups. Therefore, when group contribution models for certain properties or contribution parameters for certain structural groups are not available one need s to develop these from scratch This is where unavailability of experimental data on the properties of interest could be problematic

PAGE 22

8 since accurate group contribution models cannot be developed without sufficient amount of experimental data. It would be important to point out that for certain cases first order group contribution models for of ionic liquids are not able to predict certain properties accurately The first order group contribution models s imply consider one value for the contribution of a particular group irrespective of where that group is located with in the ionic liquid ( e.g. they do not distinguish between a CH 2 group directly attached to the aromatic carbon and a CH 2 group attached to other aliphatic side chain groups) In situations where group contribution models do not work properly, other approaches such as computational chemistry based correlative models or Quantitative Structure Property Models can be utilized to predict the physical properties of ionic liquids. In this research, we utilized COSMO RS (Conductor like Screening MOdel for Real Solvents), a quantum chemistry based equilibrium thermodynamics model with the purpose of predicting the chemical potentials of compounds in the liquid phase, to predict pure properties (e.g. melting point or viscosity) or mixture properties (e.g. activity coefficients and solub ility) of ionic liquids. When an optimal ionic liquid is designed for a specific application using the CAILD model predictions based on COSMO RS model can serve to validate the results and show us whether the design solution is suitable for use in large industrial scales. Further, computational chemistry models can provide a strong foundation of information to build libraries of data to draw upon for future research and development. Modeling reduces

PAGE 23

9 the overall cost by eliminating ionic li quids whose properties do not meet the desired application 1.2 Current a pplications During the past few years ILs have been studied for variety of applications. Some of them are listed below : 2 5, 25 30 Battery Technologies Advanced fuel cell concepts Dye sensitized solar cells (DSSC) Thermo electrical cells Supercapacitors Hydrogen generation through water splitting Carbon (CO 2 ) capture Nuclear fuel processing Solar (thermal) energy storage BASF c ommercial investigation of ILs reveals that the compounds have strong potential as solvents that can provide efficiency improvements in a several applications including:

PAGE 24

10 in chemical reactions and separation processes as hydraulic fluid and lubricant as polymer additives (antistatic) in metal deposition processes in dissolving and processing cellulose as electrolytes in electronic devices 1.3 Environmental i mpacts of i onic l iquids The non volatile nature of ionic liquids greatly limits the impact on air quality by reducing or completely elimin ating the ir direct emissions to the atmosphere. For this reason ionic liquids are often considered as inherently green /environmentally benign solvents with the potential to completely replace traditional volatile organic solvents in several applications. H owever t oxicological studies have shown that s ome ionic liquids are very toxic towards freshwater organisms or human cell lines 31 34 but due to their immense variety ionic liquids can be designed/tuned to be environmentally benign. 35 In order to conclude that ionic liquids (ILs) are benign alternatives to molecular solvents, their environmental impacts need to be analyzed in a holistic manner. Life cycle assessment (LCA) which is a technique for assessing the environmental aspects ass ociated with different steps in the production of a product, can be performed on ionic liquids like any other chemical compound to evaluate their true greenness. It is worth mentioning that, even though life cycle analysis of ionic

PAGE 25

11 liquids could be very be neficial for the informed selection of the se compounds it is quite challenging. The challenge arise s from th e fact that : ionic liquids are not yet produced in the large commercial scales Consequently, no primary data is available on material/energy consu mption and direct environmental discharges during their production On the other hand there is very little data on the environmental fate transport and toxicity of ionic liquids i n the literature. As environmental impact studies continue to reveal the negative impacts that VOC compounds have on the environment, chemist and engineers are striving to develop alternatives that reduce the overall ecological impact of current chemical process es Although the ionic liquid field is developing rapidly it is important to consider the environmental, ecological, and human health impacts at the design stage for their successful use and long term acceptance. Currently, there is very little understanding of the en vironmental impacts of producing ionic liquids as well as their impacts on fresh water ecotoxicity once they are released to the environment 1.4 Ionic liquids safety The low volatility/negligible vapor pressure of ionic liquids eliminates an important pa thway for their release into the environment. The diversity of the i onic liquids' variants available make s the process of selecting the ones that meet the defined safety requirements easier A study shows that u ltrasound waves can convert a solution of imidazolium based ionic liquid s with acetic acid and hy drogen peroxide (H 2 O 2 ) to less harmful compounds. 36 Despite the fact that ionic liquids mostly have negligible vapor pressure few of them have shown

PAGE 26

12 combustible properties and therefore should be handled careful ly 37 A b rief exposure of some ionic liquids ( ~ 5 seconds) to a flame torch can ignite them.

PAGE 27

13 Forward Problem, Prediction of Melting Point and Viscosity of ILs Chapter 2: In this chapter, we present two empirical correlations to predict the melting point and viscosity of ILs in a way that does not require experimental input or complex simulations, but rely on inputs from simple calculations based on standard quantum chemistry (QC). To develop these co rrelations, we used data related to size, shape, and electrostatic properties of cations and anions that constitutes the ionic liquid s 2. 1 Introduction As it can be interpreted from their name, ionic liquids are composed of ions, a cation and an anion, bu t their properties can significantly vary from their relatives, salts, in two main ways. First, the properties of salts can be mostly attributed to their ionic nature since strong ionic bonds hold the particles together. Ionic salts are mostly made of smal l monoatomic ions, which are in the close vicinity of each other in their crystal network. Since the lattice energy of a crystalline compound is proportionally related to the inverse of the distance between the two components, the ionic bonds of salts are very strong which contributes to properties such as very high melting point and high viscosity. On the other hand, ionic liquids are made of larger multiatomic cations and anions that result in weaker ionic bonds compared to that of salts. This explains t he considerably lower melting point (many of them are in liquid state at room temperature) and viscosity of ionic liquids. Secondly, contrary to salts, ionic liquids do not occur naturally in the environment and must be artificially synthetized.

PAGE 28

14 The mul tiatomic nature of cations and anions in ionic liquids presents a great opportunity for liquids, mainly cations and occasionally anions are composed of several al kyl side chain groups (CH 2 CH 3 2 COOH...). A vast number of different ionic liquids (an estimated number of 10 14 ionic liquids) 1 can be potentially synthesized through distinct combinations of different cation bases, alkyl groups, functional groups (attached to cations or anions ), and anions. Careful evaluation of experimental data from literature on the physical and thermodynamic properties of ILs shows that substituting one type of functional group or anion with a differen t type can drastically alter the property of interest, such as its solvency power towards a specific compound. Such behavior and trends can be seen in all different categories of ionic liquids. Despite the fact that the ionic bonds in ILs are relatively we ak, their properties can still be attributed to their ionic nature as even a weak ionic bond is still much stronger than other types of intermolecular forces. Studies show that after ionic forces, hydrogen bonds between ionic particles (cations and anions) are the most important contributor to physical properties of ILs. Even though there are, potentially millions of different ILs that are possible, to date only a few hundred of them have been actually synthetized. It is not humanly possible to synthesize every feasible ionic liquid; therefore, we need to customize and intelligently design them before synthesizing for task specific applications. Computer aided optimization frameworks can help us design optimal ionic liquids suitable for a wide range of appl ications from an extraction solvent to thermal energy storage 38 Ionic Liquids are generally salts that are liquid below 100C. Therefore, not all ILs are in the liquid state at room temperature.

PAGE 29

15 When it comes to the melting point of ILs, those with significantly lower melting points or m <25C), are of great interest to researchers seeking new application for ILs. The reason for the desirability of low melting point ionic liquids is the fact that ILs are bein g considered as separation solvents for selective dissolution of gaseous (e.g. CO 2 ), liquid (e.g. toluene), and solid (e.g. cellulose) solutes. They are also widely considered as liquid solvents to promote chemical reactions. 39,40 From a practical view poi nt, for an IL to be used as an industrial solvent it needs to be transported (pumped) across multiple unit operations and therefore it must be in the liquid phase. Another significant barrier towards commercialization of IL based application s is their high viscosity that occurs due to their ionic nature (existence of strong ionic bonds) making them difficult to transport. It is necessary to have powerful pumping equipment and efficient process equipment to handle viscous fluids. Therefore, looking for a nd customizing ionic liquids that have relatively low viscosity and melting point will greatly aid in commercialization of ionic liquids. Studies related to the ionic materials show that strong ionic bonds are mainly responsible for holding charged partic les together. Crystal lattices of ionic materials (e.g. ionic liquids) are made of cations and anions held together by electrostatic attraction. The ionic force between charged particles is directly proportional to the charge of each particle and inversely to the distance between the two ions. The larger the cations and anions are, the weaker the ionic bonds between them would be. This is due to the fact that by increasing the distance between two ions the electrostatic attraction, which holds them together will be reduced.

PAGE 30

16 2. 2 Methods Widely available information related to melting point and viscosity of salts explicitly shows that there is a meaningful relationship between the magnitude of the two discussed properties and the lattice energy of ionic bond s. Generally, in an ionic compound, size (volume and area), shape (sphericity), molecular weight, and dielectric constant of ions play an important role in determining the strength of the ionic bond of the compound. Normally, the larger size of the positiv e and negative ions (cations and anions) results in longer distances between the ions in the crystal making the ionic bond weaker. On the other hand, the shape of the ions is also important as they are better packed together when they are more symmetrical in shape. It has been suggested that asymmetry of ions in an ionic compound, most likely, will decrease the melting point since ions are more loosely connected and can be separated from each other more easily (by applying lower amount of energy). When we c ompared the size of cations and anions of a variety of ionic liquids with their melting points and viscosities, we were able to observe that in the case of ionic liquids the relationship is much more complex. We came across ionic liquids, which violated th e above discussed trends where certain ILs with relatively larger cations and anions did not necessarily have lower melting point or viscosity compared to smaller ILs. One reason for this is the fact that ionic bonds are not the only intermolecular forces responsible for holding the particles together and other types of forces such as hydrogen bonding and polar polar forces also come into play. Therefore we developed new correlations to predict melting point and viscosity of ionic liquids using information related to the size, shape and electrostatic properties of their ions. In order to account for deviations related to the above discussed trends in addition to the three

PAGE 31

17 descriptors we included several other quantum chemical descriptors to refine the corr elations with an aim to cover wide variety of ionic liquids. Another issue was that for many ionic liquids, there were multiple experimental values reported for the two physical properties. This inconsistency was especially observed in the case of melting point primarily due to the fact that the process of synthesis and existence of impurities affect s this property We avoided considering ILs that had inconsistent experimental values during the development of the correlation. Ionic liquids selected for thi s study were all 1:1 (one cation and one anion) with delocalized charges. These types of ILs are normally able to avoid crystallization and form glasses compounds far below room temperature. 41 Currently, the most widely used approach to predict the meltin g point of ILs is quantitative structure property relationship (QSPR) methods, mostly combined with artificial neural networks (ANNs) 41 In this approach, there is a reasonably good correlation between actual and predicted melting points within a standard deviation of less than 10C 41 The limited availability of experimental data on physical properties of ILs is the main drawback of constructing good QSPR models. In recent years, simulations with molecular dynamics (MD) have evolved to study the behavior o f ILs. The quality of these simulations strongly depends on the employed force fields. Several groups have tuned them specifically for ILs, while others have modified previously existing ones. For example, Alavi and Thompson have used MD simulations to pre dict the melting temperature of [C 2 MIm] + PF 6 The demanding simulation indicated a melting point that was approximately 43C too high. 41 Maginn used a similar model for the two polymorphs of [C 4 MIm] + Cl and obtained a T fus that was between

PAGE 32

18 20 and 55C too high. 41 The drawbacks of all MD simulations are the high load of computational calculations and need to know the crystal structure. In order to collect quantum chemistry data of cations and anions, we used TURBOMOLE which is a powerful, general purpose Q uantum Chemistry (QC) program, which can be used for ab initio electronic structure calculations. 41 This software allows accurate prediction of cluster structures, conformational energies, excited states, and dipoles that can be used in a broad variety of applications. When a chemical compound, in our case a cation or anion, is simulated using TURBOMOLE it can be exported as a Cosmo file, which can be later used in COSMOtherm software. COSMOtherm is a universal tool, which combines quantum chemistry (QC) an d thermodynamics to calculate properties of liquids. 41 This tool is able to calculate the chemical potential of different molecules (in pure or mixed forms) at different temperatures. In contrast to other available methods, COSMOtherm is able to predict thermodynamic properties of compounds as a function of concentration and temperature by equations, which are thermodynamically consistent. 41 Previously, computational methods such as BP86/SV(P) optimization approaches were carried out within the TURBOMOLE program package through the resolution of identity (RI) approximation. The imported/created geometries were then used for further optimization with the TZVP basis set. Next, when program converged a file with .cosmo format was exported, which later was us ed in COSMOtherm software for further calculations. At the next step, the computational chemistry data associated with cations and anions in the selected ionic liquids required for the development of aforementioned empirical correlations were

PAGE 33

19 collected. Th e data obtained from COSMOtherm software was later used to develop the empirical correlations for the prediction of melting point and viscosity of ILs. In the COSMOtherm software BP_TZVP_C30_1401 dataset was chosen for all of the calculations. 2. 3 Result s and d iscussion Molar mass of cation (MW C ), molar mass of anion (MW A ), volume of anion [ ] 3 (Vol A ), volume of cation [A ] 3 (Vol C ), Area of anion [ ] 2 (Area A ), Area of cation [ ] 2 (Area C ), dielectric energy of cation (Di C ), dielectric energy of anion (Di A ), symmetrical value of ions ] (R C ) and radii of anion [ ] (R A ), and the average distance between one cation and one anion in the network [ ] (R t ) were used as descriptors to develop empirical correlations to predict melting point and viscosity of ILs. The accurate measurement of volume, area and ionic radii of ions is only possible through imaging approaches such as X ray diffraction. In the case of unavailability of X ray data correlative or approximation approaches (e.g. van der Waals model) can be used to estimate the characteristics related to size and shape of particles. This data is very sparse for ILs and since we are interested in the devel opment of a universal correlation covering a wide range of IL structures we have to rely on predictive data. Typical cations and anions in ILs, do not usually have spherical shapes, so it is necessary to estimate their ionic radii through correlations for further use. To develop a predictive approach to estimate approximate values for radii of the cations and anions, we selected ions for which experimental ionic radii data were available. Next, the radii of cations and anions were estimated using the van de r Waals model in which all cations and anions were assumed to have spherical shape. Volume of

PAGE 34

20 corresponding ions were estimated from COSMOtherm software and were converted to the radii through eqn ( 2 1 ) ( 2 1 ) A linear correlat ion between the experimental radii of cations and anions (for which X ray values were available) and the corresponding values of their van der Waals radii is displayed in Fig ure 2 1. As it can be seen from Fig ure 2 1, the actual and model predicted values for the radii of ions are perfectly correlated through a linear relationship with R 2 =0.98622 and the corresponding equation shown in eqn ( 2 2 ) ( 2 2 ) Figure 2 1 : A correlation between van der Waals (VdW) and experimental Radii y = 0.48601x 0.01725 R = 0.98622 0 0.5 1 1.5 2 2.5 3 3.5 4 0 1 2 3 4 5 6 7 8 Expermental value of Radii (A 0 ) Vdw Radii (A 0 )

PAGE 35

21 Although, there were only few experimental data available for use in this correlation, since it covers a large range of VdW ionic radii from 3.8 to 7.5 A we used this correlation to predict the unknown values of ionic radii for the rest of cations and anions considered in this study. The distance between anions and cations in the studied ILs, Rt, were estimated as the sum of ionic radii of cations and ani ons present in the crystal network of ILs. The anions. Sphericity is a measure of how spherical an object is and can be calculated using the formula shown in eqn (2 3). ( 2 3 ) w here V p and A p are the volume and surface area of the particle, respectively. Further, the symmetrical value of ionic liquids were calculated through the following equation, eqn (2 4 ) ( 2 4) w here Sph C and Sph A are the sphericity of cations and anions, respectively. 2. 3.1 Melting point Experimental data on the melting point of several ILs covering different categories (different type of cation head groups, and anions), were gathered from the literature. 42 59 A multivariate correlation with several inputs based on quantum chemistry parameters gathered from

PAGE 36

22 COSMOtherm, and the output parameter, experimental values of T m were performed utilizing Eureqa software, a powerful data analysis tool developed by Nutonian, Inc. In the case of predicting the melting points of ionic liquids, 37 points of data on the actual melting point values of ILs were used, out of which 17 data points (i.e. 45% of the data points) were chosen solely for validation set, thereby not participating in the training process and are depicted as the points in green color in Fig ure 2 2. The experiential data of the melting point of ILs and the multivariate trend line, representing the empirical correlation developed to predict the melting point, are shown in Fig ure 2 2. As it can be seen the trend line is capable of predicting the melting point of the selected IL s. Figure 2 2 : Actual vs. model predicted melting points for training and test data sets At the next level, a comparison between the experimental (observed) values of melting points and their correspond ing values, predicted by the correlative model, are shown in Fig ure 2 3.

PAGE 37

23 Figure 2 3 : Goodness of the model for predicting melting points of the selected ionic liquids The best muti variative correlation (possessing highest R 2 ) developed by the software was used to predict the melting point of ILs as shown in eqn (2 5 ). ( 2 5 ) a=11.38, b=0.05413, c=196.6, d=434.3, e=0.649, f=1661 Table 2 1 lists the characteristics related to the selected predictive empirical correlation. Table 2 1 : Model characteristics in prediction of melting points (calculated on validation data Parameter Value R 2 (Goodness of Fit) 0.8534 Correlation Coefficient 0.9588 Maximum Error 33.7208

PAGE 38

24 Parameter Value Mean Absolute Error 14.8818 Maximum Relative Error (%) 11.9865 Mean Relative Error (%) 4.6378 2. 3.2 Viscosity In the next step, to develop an empirical correlation for predicting viscosity of ILs, experimental data on the viscosity of several different ILs at different temperatures, were gathered from the literature. 60 75 Once more, a multivariate correlation with several quantum chemistry descriptors alo ng with temperature, as inputs parameters, and Ln (viscosity), as the output parameter, was developed using Eureqa software. In the case of predicting the viscosity, we used 78 points of data on the actual viscosity of ILs at different temperatures, out of which 23 data points were chosen to be in our validation set, thereby not participating in the training process and are depicted as the points with green coloring in Fig ure 2 4. The experiential data for the viscosity of ILs and the multivariate trend line, representing the empirical correlation developed to predict the viscosity, are shown in Fig ure 2 4. As it can be seen the trend line is capable of predicting the viscosity of s elected ILs.

PAGE 39

25 Figure 2 4 : Actual vs. model predicted viscosities for training and test data sets Once more, a comparison between the experimental and model predicted values of viscosity are depicted in Fig ure 2 5. Figure 2 5 : Goodness of the model for predicting viscosities of the selected ionic liquids

PAGE 40

26 The best correlation (possessing highest R 2 ) to predict the viscosity of ILs is shown in eqn (2 6 ) ( 2 6 ) a=16.513, b=2.2179, c=0.00892, d=15.0073, e=0.02686, f=0.02975, g=15.8297, h=48.1367 Table 2 2 lists the characteristics related to the empirical model used to pre dict the viscosity of ILs as a function of QC parameters and temperature. Table 2 2 : Model characteristics in prediction of viscosity (calculated on validation data) Parameter Value R 2 (Goodness of Fit) 0.9633 Correlation Coefficient 0.982 3 Maximum Error 0.3430 Mean Absolute Error 0.08895 Maximum Relative Error (%) 11.3247 Mean Relative Error (%) 3.366

PAGE 41

27 Reverse Problem; Computer aided D esign of I onic L iquids Chapter 3: In this chapter a general computer aided IL design (CAILD) framework along with 4 case studies used to evaluate the ability of the model to select optimal ionic liquids for different applications are presented 3 .1 Introduction There exists a large libra ry of anions and cations 76,77 Similar to organic compounds, where the atoms carbon, hydrogen and oxygen can be combined to form thousands of alternative molecular structures ionic liquids can be formed through any combination of cations, anions, and alky l groups attached to the cation core leading to several structural possibilities (estimated to be as many as 10 14 combinations) 78,79 This is due to the fact that ILs are composed of organic cations and these organic compounds can have unlimited structural variations due to the easy nature of preparation of many components 80 Moreover, synthesis of a wide range of ionic liquids is relatively straightforward. This presents a great opportunity to engineer ionic liquids that have specific properties. Task speci fic ionic liquids can be designed for a particular application by controlling the physicochemical properties by judicious selection/modification of the cation, the anion, and/or the alkyl chains attached to the cation. This also presents an unusual challe nge, where synthesizing, screening, and testing the limitless possibilities becomes an impossible task 81 This is where in silico methods could act as a valuable tool for discovering new ionic liquids with tailored properties. Up until now, the majority of ionic liquid computational studies are

PAGE 42

28 based on ab initio methods such as molecular dynamics, and quantum chemical calculations 82 85 These methods are extremely important and offer useful insights as they are able to predict properties without performing costly experiments. However, as in the case of experiment al studies one needs to perform several individual simulations which again is impractical due to the long simulation times required for statistical averaging. Both molecular dynamic simu lations and experiment ation are necessary steps in the selection of task specific ionic liquids. T hese are important steps to be applied at the final stages of ionic liquid selection. The missing piece is a method for fast exploration, design and identific ation of a subset of promising candidates, from the millions of ionic liquid alternatives that are available. Computer aided molecular design (CAMD), is a promising approach that has been widely applied for molecular systems to design organic solvents for a variety of applications 86 92 It integrates property prediction models and optimization algorithms to reverse engineer molecular structures with unique properties. We believe that this approach is even more needed for the design of ionic liquids due to t he numerous ionic combinations that are possible. 3 .2 Computer a ided i onic l iquid d esign (CAILD) In this study we present an overarching framework that aims to identify ionic liquids that exhibit certain desirable behavior. Here, the identity of the compound (in this case an ionic liquid) is not known a priori but we can specify the properties that the compound ( i.e. ionic liquid) needs to have 86 This approach termed as computer aided ionic liquid design

PAGE 43

29 Im + Py + NH 4 + Tf 2 N BF 4 3 =O, S a specified set of target properties (e.g. melting point, electrical conductivity, viscosity, we can Figure 3 1 : CAILD framework Intuitively, CAILD can be thought of as a reverse problem of structure (or group) based property prediction as shown in Figure 3 1 In the forward problem (property prediction) we know the ionic liquid structure and are interested in its properties. In the reverse problem (CAILD) we know the target property values (or ranges) and are interested in feasible ionic liquid structures. To implement CAILD, we need: 1) a framework to fragment ionic liquids into groups; 2) combination and feasibility rules to identify chemically feasible ionic liquids; 3) structure (or group) based models for property prediction; and 4) an optimization framew ork to search through millions of available alternatives. M athematical programming

PAGE 44

30 approach es provide a useful mechanism to solve such CAILD problems. Different methods to solve such problems includ e generate and test type approach es (e.g. Harper et al. ) 93 or optimization based approaches (e.g. Sahinidis et al. ) 94 The solution to the underlying Mixed Integer Non Linear Programming (MINLP) model result s in the optimal molecular structure for a given application. The objective function is usually an imp ortant property related to the design problem, while the constraints relate to structural feasibility, pure component properties, solution (mixture) properties and equilibrium relationships. A conceptual representation of this approach is shown in Figure 3 2. Figure 3 2 : Conceptual description of CAILD Section 3. 2.1 will focus on describing the general mathematical framework of the proposed approach; section 3. 2.2 deals with structural constraints, provi ding an in depth mathematical treatment of feasibility, complexity and bonding rules required to design chemically feasible ionic liquids; section 3. 2.3 deals with physical property constraints where group contribution based structure property models are d iscussed for predicting ionic liquid physical (pure

PAGE 45

31 component) properties; section 3. 2.4 focuses on solution property constraints that utilize the functional group concept based models, such as UNIFAC, for calculation of solution (mixture) properties throu gh activity coefficients; section 3. 2.5 focuses on operations research methods for solving the proposed optimization model. 3. 2.1 Mathematical framework The generic mathematical formulation of the CAILD model as an optimization problem is shown in eqn s. (3 1 ) to (3 11 ) This formulation takes the form of a mixed integer non linear programming (MINLP) model. ( 3 1) ( 3 2) (3 3 ) ( 3 4) ( 3 5) ( 3 6) ( 3 7) ( 3 8) ( 3 9)

PAGE 46

32 ( 3 10) ( 3 11) where is a set of structural feasibility and complexity equality constraints, is a set of structural feasibility and complexity inequality constraints, is a set of pure component physical property inequality constraints, is a set of equality design constraints, is a set of solution (mixture) property inequality design constraints, is a m dimensional vector of binary variables denoting cation base groups, is a n dimensional vector of binary variables denoting anion, is a u dimensional vector of binary variables denoting the alky side chains, is a q dimensional vector of integer variables representing number of groups in the alkyl side chains, and is a r dimensional vector of continuous variables representing compositions, flow rates etc. 3 .2.2 Ionic liquid structural constraints The designed ionic liquids need to satisfy certain rules to ensure chemical feasibility. These rules, termed as structural constraints, include feasibility rules such as the octet rule, the bonding rule and complexity rules. Similar rules have been previously developed for molecular compounds 95,96 Eq ns. (3 12 ) to (3 24 ) represent a comprehensive set of constraints that were developed to ensure design of ionic liquid candidates that are chemically feasible. ( 3 12)

PAGE 47

33 (3 13) (3 14) (3 15) (3 16) (3 1 7 ) (3 1 8 ) (3 19) (3 20) (3 21) (3 22) (3 23) (3 24) where is a vector of binary variables representing the cations and is a vector of binary variables representing the anions. is a vector of binary variables representing the alkyl chains l is a vector of integer variables representing the number of groups of type k in the alkyl side chain l are vectors of group valencies of the cations and alkyl gr oups respectively. G is the set of all alkyl groups available for the cation side chains. eqn s. (3 12 )

PAGE 48

34 and (3 13 ) ensure a maximum of one cation base and one anion respectively for each IL candidate. eqn (3 14 ) fixes the number of alkyl side chains attac hed to the cation based on available free valence of the cation base. Modified octet rule : The implementation of the modified octet rule eqn (3 15) ensures that any designed cation is structurally feasible and that each valence in all structural groups of the cation is satisfied with a covalent bond. Note that this formulation has already accounted for the positive charge associated with the cation. eqn (3 16 ) ensures that the octet rule is implemented for each side chain l to ensure that the valence s in the individual chains are satisfied with a covalent bond. Cation size : The size of the cation is controlled by introducing an upper bound on maximum number of groups ( ) that are allowed in the cation eqn (3 17 ). Alky cha in size : The size of the alkyl chains are controlled by introducing an upper bound on the maximum number of groups that can be present in each alkyl side chain eqn (3 18) Eqns (3 19 ) (3 20 ) and (3 21 ) can be utilized to place restrictions on number of occurrences (t 1 t 2 and t 3 ) of a certain group, in each side chain l In other words, eqn (3 19 ) can be used to make sure that a certain main group such as aldehyde or alcohol not being present more than a certain number of times in each side chain of th e cation and eqn (3 20 ) can be appl ied when we want a certain group to be present at least t 2 times and eqn (3 21 ) can be used when an exact number of occurrence of a certain group is desired e .g. when we want to have exactly one aldehyde group in a certain side chain in the cation. Eqns (3 22 ) (3 23 ) and (3 24 ) can be utilized to place restrictions on number of occurrences (t 4 t 5 and t 6 ) of a certain group in the cation, which can be calculated as summation of number of occurrences of the particular group in all the side chains in the cation. The purpose of eqn s. (3 22 ) to (3 24 ) is exactly similar to that of eqn s.

PAGE 49

35 (3 19 ) through (3 21 ) with the difference of placing restrictions on th e number of occurrences of a particular group in whole cation (summation of all of the side chains) instead of only one side chain. The cation related structural feasibility constraint eqn (3 14) is explained using the generic cation dialkylimidazoli um (shown in Figure 3 3a) as an example. According to the proposed formulation the Valence for this cation is 2 (i.e. ), as there are 2 alkyl side chains (R 1 and R 2 ) that are allowed. Similarly, the Valence of a trialkylimidazolium (shown in Figure 3 3b) is 3. For dialkylimidazolium, the right hand side of constraint 3 eqn (3 14) will translate into which will fix the left hand side of constraint 3 as Therefore the vector y will take the f ollowing values [1 0 1 0 0 0] with the two ones representing the presence of 2 alky side chains at positions 1 and 3. Next, with the use of few feasible and infeasible examples shown in Figure 3 4, we explain how the whole set of feasibility constraints e qn s. (3 12 ) to (3 16) work. Figure 3 4a shows a feasible ionic liquid, 1,3 diethylimidazolium tetrafluoroborate. The cation Valence and alkyl group valence s related to this ionic liquid are listed in Table 3 1. Table 3 1 : Cation and alkyl side chain groups valence s Cation groups (c) i 1 alkyl 3 alkyl Im 1 2 Alkyl groups (k) l CH 3 1 1 CH 2 1 2 CH 3 3 1 CH 2 3 2

PAGE 50

36 The values related to the vectors and for this ionic liquid are listed in Table 3 2. Table 3 2 : Values of and for 1,3 diethylimidazolium tetrafluoroborate l k 1 1 CH 3 1 CH 2 1 3 1 CH 3 1 CH 2 1 In this case there is only one cation base ( e qn (3 12 ) is satisfied) and one anion ( e qn (3 13 ) is satisfied). The number of side chains are 2 which equates to the valence of cation ( e qn (3 14 ) is satisfied). Now, left hand side (LHS) of e qn (3 15 ) translates into (2 2)(1)+[(2 2)(1)(1)+(2 1)(1)(1)]+[(2 2)(1)(1)+(2 1)(1)(1)]=2 which is equal to right hand side (RHS) of the equation ( e qn (3 15 ) is satisfied). For both of the side chain positions ( l ) 1 and3, LHS of e qn (3 16 ) translates into [(1)(1)( 2 1)+(1)(1)(2 2)]=1 which is equal to RHS of the equation ( e qn (3 16 ) is satisfied). Eqn s. (3 17 ) through (3 24 ) are only used to control the cation size and place restrictions on the type and number of occurrences of select groups. As such these are not feasibility constraints but user specified structural design constraints. Figure 3 4b shows an infeasible imidazolium based ionic liquid. In this case there is only one cation base ( e qn (3 12 ) is satisfied) and one anion ( e qn (3 13 ) is satisfied). The number of side chains are 2 which equates to the valance of cation ( e qn (3 14 ) is satisfied). The LHS of constraint 4 e qn (3 15) translates to (2 2)(1)+[(2 2)(1)(1)+(2 2)(1)(1)]+[(2 1)(1)(1)+(2 1)(1)(1)]=2,

PAGE 51

37 which equates to the RHS of the equation and hence constraint 4, related to the whole cation, is satisfied. However, constraint 5 e qn (3 16) related to the side chains are violated as follows: for side chain l =1 containing 2 ethyl groups e qn (3 16 ), translates to, [(1)(1)(2 2)+(1)(1)(2 2)] and for side chain l =3 containing 2 methyl groups e qn (3 16 ) translates to, [(1)(1)(2 1)+(1)(1)(2 1)] Therefore, the structure shown in Figure 3 4b is infeasible. For the structure shown in Figure 3 4c, e q n s (3 12 ) (3 13 ) and (3 14 ) are satisfied. The LHS of constraint 4 e qn (3 15 ), translates to (2 2)(1)+[(2 2)(1)(2)+(2 1)(1)(1)]+[(2 1)(1)(2)]=3, which does not equate to the RHS of the equation (i.e. 2) and hence constraint 4, related to the whole cat ion, is violated. Constraint 5 e qn (3 16), related to the side chains translates to the following: For side chain position 1 (i.e. l =1), LHS of e qn (3 16 ) is [(1)(2)(2 2)+(1)(1)(2 1)]=1, which equates to the RHS of the equation (i.e. 1), but for side chain position 3 (i.e. l =3) the same equation translates to [(1)(2)(2 1)]=2 which does not equate to the RHS of the equation (i.e. 1) and hence constraint 5, r elated to side chain 3, is violated. Therefore, the structure shown in Figure 3 4c violates two of the feasibility constraints and hence is infeasible. Figure 3 3 : A general Scheme of two feasible catio n structures

PAGE 52

38 Figure 3 4 : Examples of feasible and non feasible Ionic liquids 3 2.3 Physical property constraints Ionic liquid structures play a key role in determining their unique physical properties. Physical property constraints utilize structure property models which provide insights into the relationship between molecular structures and their properties. The particular type of structure property relationships suited for CAILD are group contribution (GC) models. As discussed before, GC models for physical properties are based on functional group additive principle. The ionic liquid is fragmented into characteristic groups and the property of interest is predicted as an additive function of the number of occurrence of a given group times its contribution to the pure component property. The contribution parameters of different groups are derived by correlating experimental data to a group additive expression. These models exist for several IL physical p roperties such as viscosity 97 density 98 101 melting point 102 electrical conductivity 103 thermal conductivity 103 heat capacities 104 solubility parameter 105 and toxicity 106

PAGE 53

39 3 .2. 4 Solution property constraints Thermodynamic properties of non ideal solutions are important for evaluating intermolecular interactions between multiple components (both ionic and molecular) present in a mixture. These thermodynamic properties are essential to evaluate the potential of ionic liquids as solvents for reaction (solid solubility and liquid miscibility) and the separation of fluid mixtures (liquid liquid extraction and gas liquid absorption). An essential requirement is the ability to predict excess Gibbs free energy (activit y coefficients) of systems involving ionic liquids which enable prediction of equilibrium concentrations. These constraints are not only a function of binary/integer structure variables but also relate to the compositions of the various components of the m ixture. The proposed CAILD framework requires models for the prediction of activity coefficient s that are based on hypothesis of the solution of group concept is that interactions between molecules can be approxima ted as interactions between functional groups. To illustrate this concept, t he different interaction s amongst the groups of a mixture of an ionic liquid, [ M i m] Tf 2 N and CH 3 OH is shown in Figure 3 5. Figure 3 5 : Different types of group interactions involved in solutions containing ionic

PAGE 54

40 The number of distinct cation head groups, anion groups and alkyl groups are much less in comparison to the number of distinct ionic liquids that can be generated from them. Therefore, a relatively small number of group interaction parameters are required to represent all possible ionic liquids. UNIFAC 107 (UNIversal quasi chemical Functional group Activity Coefficients) is a widely used group contri bution model to predict phase equilibrium in non electrolyte systems. The UNIFAC model combines the concept of functional groups with a model for the activity coefficient based on UNIQUAC (Universal Quasi Chemical). The activity coefficient has a combinato rial contribution (due to differences in size and shape of molecules) and a residual contribution (due to energetic interactions). (3 25) The group volume ( R ) and surface area ( Q ) parameters of the combinatorial part are calculated as summation of group parameters (volume R k and surface area Q k ) while binary group interaction parameters (a mn and a nm ) are required for the calculation of the residual component. The UNIFAC approach w as originally used for non electrolyte systems, however in recent studies several research groups have utilized this approach for ionic liquids by careful representation of ionic groups and/or incorporating assumptions that factor the ionic nature of the g roups. In order to apply the UNIFAC model, in its current form, to ionic liquids Wang et al. 108 and Lei et al. 109 treated ionic groups as a single non dissociate neutral entity. For example, the ionic liquid [Bm i m] BF 4 was decomposed into two CH 3 groups, three CH 2 groups and one [Im] BF 4 group. Using the above representation Lei et al. 109 have added 12 new ionic groups (e.g. [Im] PF 6 ) to the existing UNIFAC table. Ionic liquid

PAGE 55

41 groups are included in the modified UNIFAC (Dortmund) model. 110 Most recently, Roughton et al. 105 have characterized the ionic groups in the same way as proposed in our proposed CAILD formulation ; i.e. as separate cation base, anion and alkyl groups. The underlying assumption is that the ionic groups can be treated separat ely and the interactions between the ionic groups can be assumed to be zero due to the strong interaction and weak dissociation between ion pairs. 105 More detailed treatment of UNIFAC approach for ionic liquids can be found in Roughton et al 105 and Wang et al 108 Lei et al 109 Liquid Liquid equilibrium Designing industrial scale liquid liquid separation systems using ionic liquid requires modeling equilibrium relationships. In a non ideal liquid mixture, species which have limited mutual solubility in the given liquid phase exhibit positive deviations from Raoult's Law. The quantitative measure of non ideality is the liquid activity coefficient which is a function of composition and temperature. If we identify the two liquid phases as l 1 and ` l 2 ', their respective mole fractions in the two phases are related by the equilibrium condition as follows: (3 26) Where, and are activity coefficients of component i in the liquid phases 1 and 2 respectively, and and are mole fractions of component i in the two phases. Solid Liquid equilibrium

PAGE 56

42 Estimation of equilibrium saturation concentrations of solid liquid systems is essential to model processes that involve solute dissolution, and crystallization. The liquid phase activity coefficient predictions discussed previously and the pure component properties of solute ( ), can be utilized for these calculations. (3 27) where T m and T represent enthalpy of fusion (J/mol), melting point (K) and temperature (K), respectively. represents activity coefficient of solute at saturation and is the solubilit y of solute. 3. 2.5 Solution of the underlying MINLP The presented CAILD model is a non convex, mixed integer non linear programming (MINLP) problem, involving large number of integer and binary variables. Consideration of mixture properties through the UNIFAC model results in non linearity and most of the binary design variables (structural) participate in the non linear terms. Combinatorial complexity is an inherent issue in CAMD MINLP models due to the nature of the search space. The most direct approach for solving the underlying MINLP model is complete enumeration. Generate and test methods fall under this category. 93 Solution to the MINLP model can also be achieved through mathematical programming usi ng deterministic ( e.g. branch and bound 111 branch and reduce 112 ) and stochastic optimization methods ( e.g. simulated annealing 113 genetic algorithms 114 and tabu search 90 ). Approaches that combine features from both

PAGE 57

43 domains, such as decomposition methods have also been previously developed. 94,115 Achenie et al. 86 provide a detailed description of various solution techniques in the context of molecular design problems. In this section we focus on solving the CAILD framework utilizing two different methods : the decomposition methodology (includes generate and test algorithm s ) and genetic algorithm based optimization. Our purpose is to demonstrate that different types of solution approaches can be used towards a solution of the proposed CAILD formulation. Ma in details about the two approaches are provided below while in depth analysis can be found elsewhere 115,116 Decomposition Method In this approach the CAILD MINLP model is decomposed into an ordered set of subproblems where each subproblem requires only the solution of a subset of constraints from the original set. As each subproblem is being solved large numbers of infeasible candid ates are eliminated leading to a final smaller subproblem. The first subproblem usually consists of the structural constraints and it equates to enumeration. The second subproblem consists of pure component (physical) property constraints while the third s ubproblem consist of mixture property constraints. These three subproblems taken together equate to generate and test methods. The ionic liquid candidates that pass through all of the above subproblems are the only ones that will be considered in the final optimization subproblem that involves the objective function, equilibrium relationships and process models (if considered in the design problem). Most often, the solution to the final subproblem can be achieved by solving a set of non linear programming ( NLP) problems.

PAGE 58

44 Genetic Algorithm Genetic algorithm (GA) is a method that can be used to solve optimization problems based on the natural selection process that mimics biological evolution. It can be applied to solve problems that are not well suited for st andard optimization algorithms, including problems in which the objective function is discontinuous, non differentiable, stochastic, or highly nonlinear 116 Unlike traditional search and optimization methods, GAs perform a guided stochastic search where improved solutions are achieved by sampling areas of the search space that have a higher probability for good solutions 116 The optimization process starts and a llowed to exchange or alter their genetic information, through crossover and mutation operations, with an aim to create more fitter off springs. At every iteration new populations of off springs are created to replace the existing population. This process of evolution is repeated for a pre determined number of generations or un til the solution is found 116 In GA, the selection of fitter parents for next generation is based on their fitness values as determined by a fitness function. The fitness function is usually very closely related to the original objective function of the search problem (in all of the case studies presented in this study the fitness function was identical to the objective function). The GA solution of CAILD model was implemented in the MATLAB environment with most parameters taking the default values. Specifically, for all the case study problems, the population size was fixed at 20 and the initial population was generated randomly. The crossover fraction was fixed at 0.8 while the mutat ion probability was fixed at 0.2. We allowed two candidates with the best

PAGE 59

45 fitness values (elite candidates) in the current generation to automatically survive to the next generation. 3.3 Proof of concept examples In this section several case studies have been presented to illustrate the usefulness of the proposed approach. Table 3 3 lists the entire set of groups and their respective valence s, from which the basis sets for the four case studies were derived. Note t hat this basis set covers only a small set of cations, anions and functional groups for which group contribution parameters are currently available for the properties of interest. However, the design approach itself is universal in nature and upon availabi lity of group contribution models and parameters can be easily extended (for example, to all groups in Appendix A ) to cover the entire spectrum of possible ionic liquids. The maximum value allowed for the number of groups were fixed at 6 for each side chai n and 12 for the whole cation. Table 3 3 : The basis set used for ionic liquid design Cations Valence Anions Groups Valence Im 2 CH 3 1 Mi m 1 CH 2 2 Py 2 Mp y 1

PAGE 60

46 3.3.1 Electrolytes This case study demonstrates the design of an ionic liquid that has high electrical conductivity. Electrical conductivity measures the ability of a material to conduct electric current. It is an important property for the development of electr ochemical devices such as high energy batteries. Other design requirements include the following: the electrolyte (i.e. ionic liquid) needs to be a room temperature ionic liquid (RTIL); and it should have reasonably low viscosity. The electrical conductiv ity of ionic liquids can be estimated using a Vogel Tamman Fulcher ( VTF ) type equation shown in e qn (3 28 ). ( 3 28) where A and B are adjustable parameters that can be obtained through group contribution expressions e qn (3 29 ) and e qn (3 30) as proposed by Gardas et al. 103 and T 0k has the value of 165K for all considered IL types. (3 29) (3 30) where n i is the number of groups of type i and k is the total number of g roups considered Table 3 4 shows the group contribution parameters used.

PAGE 61

47 Table 3 4 : Group contributions for parameters A and B Species a b (K) Im 77.8 501.5 Mi m 77.9 537.6 Py 69.6 544.9 Mp y 69.7 581.0 BF 4 85.8 129.4 PF 6 117.3 278.6 Tf 2 N 10.1 46.4 CH 3 0.1 36.1 CH 2 0.1 36.1 The viscosity of ionic liquids is calculated using an Orrick Erbar type approach. 117 In this method, viscosity can be predicated as a function of density molecular weight temperature and parameters A and B through the use of e qn (3 31 ) (3 31) We employ the group contribution technique proposed by Gardas et al 103 to estimate the parameters A and B as follows ( 3 32) ( 3 33)

PAGE 62

48 where is the number of occurrences of group i (cation, anion and functional groups) and and are the c ontributions of group i to the parameters A and B respectively. The ionic liquid densities are estimated using the below formula. ( 3 34) where the density in kg m 3 M is the molecular weight in k g mol 1 N is the Avogadro number, V is the molar volume in T is the temperature in K and P is the pressure in MPa. Based on the data provided in Gardas et al 103 we developed group contribution parameters for molar volume with expressions similar to e qn s. (3 32 ) and (3 33 ) The values of coefficients a, b and c are and Table 3 5 shows the group contribution parameters used in this model. Table 3 5 : Group contributions for parameters A, B and V Species A v B v Im 84 8.04 1257.1 Mi m 119 7.3 1507.1 Py 111 7.61 1453.6 Mp y 146 6.87 1703.6 BF 4 73 18.08 1192.4 PF 6 109 20.49 2099.8 Tf 2 N 248 17.39 510.0 Cl 47 27.63 5457.7 CH 3 35 0.74 250.0 CH 2 28 0.63 250.4

PAGE 63

49 The melting point of ionic liquids is estimated by a group contribution approach using e qn (3 35 ) as proposed by Aguirre et al 118 ( 3 35) where is the number of occurrences of group i (cation, anion and functional groups) and is the contribution of group i to the melting point, a and c are constants with values of 0.1 and 0.012 respectively. which is related to the cation flexibility, is estimated using e qn (3 36 ) and is a cation symmetry parameter having values shown in Table 3 6. ( 3 36) Table 3 6 : values for different cations Type of cation value 0.265 0.317 0.265 The contribution, T m,i, of different groups are listed in Table 3 7. Table 3 7 : Group T m,i Im 107.99 Mim 10 9.88 Py 117.212 Mp y 119.102

PAGE 64

50 Group T m,i BF 4 0.479 PF 6 16.746 Tf 2 N 0.966 Cl 35.852 CH 3 1.463 CH 2 1.463 The CAILD design problem expressed in mathematical form is shown in e qn s (3 37 ) to (3 40 ) Objective function f obj = max ( ) ( 3 37) Constraints Ionic liquid Structural Feasibility ( 3 38) ( 3 39) ( 3 40) Results: The design statistics for this problem are summarized in T able 3 8. A total of 138 feasible IL structures were enumerated in subproblem 1. Out of these, 26 ILs satisfied the physical property constraints (viscosity and melting point) in subproblem 2. There were no mixture properties considered (subproblem 3) and the solution to final subproblem (subproblem 4) resulted in the optimal ionic liquid structure ( 1 methylimidazolium bis trifluoromethylsulfonyl imide ) with the highest electrical conductivity (shown in Figure 3 6). The properties of the designed ionic liquid are listed in Table 3 9.

PAGE 65

51 Table 3 8 : Decomposition approach: Subproblem Results Subproblem 1: Number of ionic liquids (ILs) generated, 138 Subproblem 2: Number of ILs satisfying pure component properties, 26 Subproblem 3: No mixture properties Subproblem 4: Optimal IL, [Mim] + [Tf 2 N] The same design problem e qn s. (3 37 ) to (3 40) was solved using the genetic algorithm toolkit in MATLAB and the program picked the exact same structure (shown in Figure 3 6) as the optimal solution. Figure 3 6 : 1 methylimidazolium [Tf 2 N] Table 3 9 : Design Results of the optimal IL, 1 methylimidazolium [Tf 2 N] Properties V alue Melting point (K) 270.15 Viscosity (cP) 21.293 Electrical Conductivity (S / m 1 ) @ 25 o C 1.0956 Analysis : In this section we focus on the validation of design results through careful consideration and analysis of available experimental data. Table 3 10 lists the available

PAGE 66

52 based on the cation, anion and side groups considered in this case study (Im, Py, PF 6 BF 4 Tf 2 N, CH 3 and CH 2 ). Unfortunately, we could no t find the electrical conductivity data for the designed IL ( 1 methylimidazolium bis (trifluoromethylsulfonyl) imide). Therefore, we perform a qualitative IL structure property trend analysis to validate the design results. The electrical conductivity val ues have the following trend: Since, all of the above ionic liquids have the same cation (C 4 mim) but different anions (Tf 2 N PF 6 and BF 4 ) we can infer that electrical conductivities of ionic liquids with Tf 2 N anions are greater than those with PF 6 and BF 4 anions (for same cation and alkyl groups). The design result is consistent with this observation as the optimal structure has Tf 2 N anion. Similarly, by comparing the electrical conduct ivities of ionic liquids having the same anion (Tf 2 N ) we can see that [C 2 mim] Tf 2 N > [C 4 mim] Tf 2 N > [C 6 mim] Tf 2 N Therefore, we can conclude that increasing the number of alkyl groups on the cation side chain decreases the electrical conductivity. The design result (Figure 3 6 ) is also consistent with this observation as there is only one methyl group (minimum needed to satisfy the cation Valence 4 mim] Tf 2 N and [C 4 Py] Tf 2 N which have the same anion and different cations. We found experimental electrical conductivity data for [C 4 Py] Tf 2 N but there was no data available for [C 4 mPy] Tf 2 N Since we already know that addition of alkyl groups to the cation base will decrease t 4 mPy] Tf 2 N 4 Py] Tf 2 N 4 mim] Tf 2 N (0.406 S/m). Therefore, we can conclude that electrical conductivity of ionic liquids with imidazolium based cations are greater than Pyridinium based cations (for same anion and alkyl groups). The design result is consistent

PAGE 67

53 with this observation as the optimal structure selected had an imidazolium cation. Overall, the model results are in full agreement with the observed trends from experiments, thereby validating the proposed approach. Table 3 10 : Experimentally Measured Electrical Conductivities of Ionic Liquids Ionic liquid T (C) Ref PF 6 25 0.146 [ 119 ] Tf 2 N 25 0.912 [ 119 ] Tf 2 N 25 0.406 [ 119 ] Tf 2 N 25 0.218 [ 119 ] BF 4 25 0.59 [ 120 ] BF 4 25 0.35 [ 120 ] Tf 2 N 25 0.33 [ 121 ]

PAGE 68

54 3.3.2 Heat transfer fluids Ionic liquids show great promise as heat transfer fluids and heat storage medium. High thermal conductivity is an important property for such applications. Thermal conductivity measures the ability of a material to conduct heat. Thermal conductivity of ionic liquids is weakly depe ndent on temperature and could be fitted with the following linear correlation. ( 3 41) where, and T are the thermal conductivity in Wm 1 K 1 and temperature in K respectively. We utilize a method 96 that employs group contribution approach to estimate the parameters A k and B k ( 3 42) ( 3 43) Table 3 11 shows the group contribution parameters that were used. Pyridinium and methyl pyridinium have not been cons idered in this part since their group contribution s w ere not found. 103 Table 3 11 : Group contributions for parameters A k and B k Species a k b k (K 1 ) Im 0.1272 0.000000104 Mi m 0. 1314 0.00000 787 BF 4 0.0874 0.00008828 PF 6 0.0173 0.000009088

PAGE 69

55 Tf 2 N 0.0039 0.00002325 Cl 0.0166 0.00001 CH 3 0.0042 0.000007768 CH 2 0.0010 0.000002586 Melting point and viscosity of ionic liquids were calculated through the same methods proposed in case study 1. The CAILD design problem expressed as an optimization model is shown in e qn s (3 44 ) to (3 47 ) Objective f unction f obj = max ( ) ( 3 44) Constraints Ionic liquid Structural Feasibility ( 3 45) ( 3 46) ( 3 47) Results: Decomposition approach: The design statistics for this problem are summarized in T able 3 12. A total of 92 feasible IL structures were enumerated in subproblem 1. Out of these, 15 ILs satisfied the physical property constraints (viscosity and melting point) in subproblem 2. There were no mixture properties considered (subproblem 3) and the solution to final subproblem (subproblem 4) resulted in the optimal ionic liquid structure ( 1 e thyl 3 methylimid azolium tetrafluoroborate ) with the highest thermal conductivity (shown in Figure 3 7). The properties of the designed ionic liquid are listed in Table 3 13.

PAGE 70

56 Table 3 12 : Decomposition approach: Subproblem Results Subproblem 1: Number of ionic liquids (ILs) generated, 92 Subproblem 2: Number of ILs satisfying pure component properties, 15 Subproblem 3: No mixture properties Subproblem 4: Optimal IL, 1 Ethyl 3 methylimidazolium tetrafluoroborate The same design problem e qn s. (3 44 ) to (3 47) was solved using the genetic algorithm toolkit in MATLAB and the program picked the exact same structure (shown in Figure 3 7) as the optimal solution. Figure 3 7 : 1 ethyl 3 methylimidazolium [BF 4 ] Table 3 13 : Design Results of the Optimal IL, 1 ethyl 3 methylimidazolium [BF 4 ] Properties Value Melting point (K) 291.71 Viscosity (cP) 60.75 Thermal Conductivity (Wm 1 K 1 ) @ 25 o C 0.193

PAGE 71

57 Analysis : Table 3 14 shows experimental thermal conductivity, k [Wm 1 K 1 ] for 10 different ionic liquids that are based on the cation, anion and side groups considered in this case study (Im, PF 6 BF 4 Tf 2 N CH 3 and CH 2 ). The designed ionic liquid (Figure 3 7) is same as the IL with the highest thermal conductivity value in T able 3 14 ([C 2 mim] BF 4 ). This partially validates the results. However, for a more holistic assessment we perform a qualitative ionic liquid struct ure property trend analysis to determine whether the designed results are consistent with observed data. By comparing the thermal conductivity values (Table 3 14), we note Since all of these ionic liquids have the same cation (C 4 mim) but different anions (Tf 2 N PF 6 and BF 4 ), we can infer that thermal conductivities of ionic liquids with BF 4 anion are greater than those with PF 6 and Tf 2 N anions. The design result is consistent with this observation as the optimal structure has BF 4 anion. The only base cation considered in this design problem is imidazolium. By comparing the k values of different ionic liquids with the same base cation and same anion, but different side groups (i.e. [C 4 mim] PF 6 vs [C 6 mim] PF 6 vs [C 8 mim] PF 6 and [C 2 mim] Tf 2 N vs [C 4 mim] Tf 2 N vs [C 6 mim] Tf 2 N vs [C 8 mim] Tf 2 N vs [C 10 mim] Tf 2 N we can see that the contribution of alkyl side chain groups are not as high as that of anion, and there is no uniform trend that is observed in relation to varying number of alkyl side chain groups. Therefore the optimal numbers of alkyl side chain groups relate to other requirements such as the IL needing to be a liquid (i.e. T m < 25 C for RTILs) and have relatively lo w viscosity. Overall we can conclude that the design results are consistent with the observed experimental structure property trends of thermal conductivity data.

PAGE 72

58 Table 3 14 : Experimental thermal conductivity data Ionic liquid T (K) k (Wm 1 K 1 ) Ref PF 6 315 0.145 [122] PF 6 315 0.146 [122] PF 6 315 0.145 [122] BF 4 315 0.1968 [123] BF 4 315 0.1847 [123] Tf 2 N 315 0.1294 [124] Tf 2 N 315 0.1264 [124] Tf 2 N 315 0.1263 [124] Tf 2 N 315 0.12715 [124] Tf 2 N 315 0.1299 [124]

PAGE 73

59 3.3.3 Toluene h eptane separation A common use of solvents in industrial processes is as a separating agent to isolate two liquid components. This case study relates to the design of optimal ionic liquid to separate toluene (aromatic) and n heptane (aliphatic). Sulfolane (C 4 H 8 O 2 S) is a molecular solvent that is commonly used for this purpose. The design objective is to f ind an ionic liquid that can improve performance in comparison to sulfolane. One key requirement is to select an ionic liquid with as low viscosity as possible since, viscous solvents are not ideal from the stand point of industrial equipment design. The o ther requirement is to ensure that the designed solvent is a room temperature ionic liquid (RTIL) as the process requires a liquid solvent. A constraint on melting point e qn (3 56 ) is necessary to ensure design of RTILs only. Melting point and viscosity of ionic liquids were calculated through the same methods proposed in case study 1. A good separation solvent should have a high value for selectivity e qn (3 48) and solvent power e qn (3 49), and low value for solvent loss e qn (3 50). Selectivity: ( 3 48) Solvent power: ( 3 49) Solvent loss: ( 3 50) The three properties are a function of infinite dilution activity coefficients of the n he ptane /toluene/IL solution. The activ ity coefficients are calculated using the UNIFAC model (discussed in section 2.4) and the interaction parameters for ionic liquids were taken

PAGE 74

60 from Roughton et al. 105 Another important consideration is that, the addition of IL to the binary liquid mixture s hould result in the creation of two liquid phases. The appearance of new phases in a multi component system can be checked through the implementation of necessary and sufficient conditions for phase stability. These conditions for a ternary system, are sho wn in e qn s. (3 51 ) and (3 52 ) were derived by Bernard et al. (1967). The activity coefficients were again calculated using the UNIFAC method as discussed before. ( 3 51) ( 3 52) The CAILD design problem expressed as an optimization model is shown in e qn s (3 53 ) to (3 5 8 ) It is worth mentioning that currently cost data is not available for ionic liquids as they are for the most part not commercially produced and it is also difficult to utilize cost information within a computer aided molecular design framework. Hence cost was not considered for minimization. Objective function f obj ) ( 3 53) Constraints: Ionic liquid Structural Feasibility ( 3 54)

PAGE 75

61 ( 3 55) ( 3 56) ( 3 57) ( 3 58) Results : T he design statistics for this problem are summarized in T able 3 15. A total of 185 feasible IL structures were enumerated in subproblem 1 (structural constraints). Out of these, 27 ILs satisfied the physical property constraints (viscosity and melting poin t) in subproblem 2. Out of these, 1 ionic liquid satisfied the mixture property constraints e qn s (3 57 ) and (3 58). The optimal ionic liquid structure ( 1 methylpyridinium tetrafluoroborate ) with the highest selectivity is shown in Figure 3 8. The properties of the designed ionic liquid are listed in Table 3 16. Finally, we verified whether the designed ionic liquid created two phases when added to a hypothetical binary mixture consisting of 70% n heptane (aliphatic) and 30% toluene (aromatic ). This was accomplished by solving e qn s. (3 5 1 ) and (3 5 2 ) for a range of ternary compositions (keeping n heptane to toluene ratio constant). We identified that a phase split occurs at solvent composition range of 0.4 to 0.9. Table 3 15 : Decomposition Approach: Subproblem Results Subproblem 1: Number of ionic liquids (ILs) generated, 185 Subproblem 2: Number of ILs satisfying pure component properties, 27 Subproblem 3: Number of ILs satisfying mixture properties, 1 Subproblem 4: Optimal candidate, [Mp y ] + [BF 4 ]

PAGE 76

62 The structure of the optimal ionic liquid that satisfies all the constraints and has the 3 8. The properties of the designed ionic liquid are listed in Table 3 16. Figure 3 8 : 1 methylpyridinium [BF 4 ] Table 3 16 : Design Results of the Optimal IL, 1 methylpyridinium [BF 4 ] Properties 1 methylpyridinium [BF 4 ] [BF 4 ] Sulfolane Melting point (K) 294.8 300.65 Viscosity (cP) 55.084 10.07 SL 0.006471 0.0065 (t 1 ) SP 0.67193 0.3719 (t 2 ) 87.262 6.8023 (t 3 ) Analysis: Table 3 17 shows experimental selectivity values for the separation of aromatics from an aromatic/aliphatic mixture using different ionic liquids. 125 By comparing the selectivit y values for separation of benzene from benzene/heptane mixture we see that Since both of these ionic liquids have the same cation (hmim)

PAGE 77

63 but different anions (PF 6 and BF 4 ), we can infer that selectivit y values of ionic liquid with BF 4 anions are greater than those with PF 6 anions. By comparing the selectivit y values of [bmim] Tf 2 N and [bmim] PF 6 (same cation and different anions), for separation of toluene/heptane mixture we can infer that selectivit y values with PF 6 anion are greater than selectivit y values with Tf 2 N anion. Therefore we can conclude that among the anions used in the design problem ( Tf 2 N PF 6 and BF 4 ), ionic liquids having BF 4 anion should have the highest selectivity towards aromatic compounds. The design result is consistent with this observation as the optimal ionic liquid has BF 4 anion. Similarly, by comparing the selectivit y values for [mmim] Tf2N [emim] Tf2N [bmim] Tf2N This shows that increasing number of alkyl groups on the cation side chain decreases selectivity The design result (Figure 3 8) is consistent with this observation also as there is only one methyl group (minimum needed to satisfy the cation Valence ) present in the cation side chain. With respect to cation, several studies have reported that Pyridinium based cations have higher selectivity than i midazolium base cations for aliphatic/aromatic separation. This is also consistent with our design results as the optimal IL had Pyridinium based cation. This trend analysis qualitatively validates the CAILD methodology as well as the group interaction parameters (e.g. UNIFAC parameters provided in Roughton et al .) used in the model.

PAGE 78

64 Table 3 17 : Experimentally Measured Selectivit y values for Aromatic/Aliphatic Separations Solvent Separation T (C) Tf 2 N Toluene/heptane 40 29.8 Tf 2 N Toluene/heptane 40 22.2 Tf 2 N Toluene/heptane 40 16.7 PF 6 Toluene/heptane 40 21.3 PF 6 Benzene/heptane 25 8.20 BF 4 Benzene/heptane 25 8.40 BF 4 Toluene/heptane 40 32.8

PAGE 79

65 These qualitative trends from experiments are consistent with our CAILD results. The optimal design has a BF 4 anion and has minimal number of alkyl side groups (note that the selected methyl pyridinium base cation needs at least one CH 3 group ; i.e. the minimum possible number of groups to satisfy the valence requirement). 3.3.4 Naphthalene solubility In this case study we consider the design of an ionic liquid solvent for the dissolution of organic compound naphthalene. A molecular solvent having high solubility for naphthalene and commonly used for its disso lution is chloroform. The measured solubility of naphthalene in chloroform is 0.473 mole fraction 126 Our objective is to find an ionic liquid that has higher solubility for naphthalene than chloroform. In addition, the ionic liquid needs to be an RTIL, an d have a reasonably low viscosity. The melting point and viscosity are calculated using the same models described in the case study 1. The expression shown in e qn (3 61 ) is invoked to ensure solid liquid phase equilibrium conditions, in order to determine the saturation concentration of solute (i.e. solubility) 115 The CAILD design problem expressed as an optimization model is shown in e q ns (3 59 ) to (3 63 ) Objective function f obj = max ( ) ` ( 3 59) Constraints Ionic liquid Structural Feasibility ( 3 60)

PAGE 80

66 ( 3 61) ( 3 62) ( 3 63) Results : The design statistics for this problem are summarized in T able 3 18. A total of 185 feasible IL structures were enumerated in subproblem 1. Out of these, 27 ILs satisfied the physical property constraints (viscosity and melting point) in subproblem 2. The optimal ionic liquid structure (1 Butyl 3 ethylimidazolium [Tf2N] ) with the highest solubility is shown in Figure 3 9. The properties of the designed ionic liquid are listed in Table 3 19. Table 3 18 : Decomposition Approach: Subproblem Results Subproblem 1: Number of ioni c liquids (ILs) generated, 185 Subproblem 2: Number of ILs satisfying pure component properties, 27 Subproblem 3: Number of ILs satisfying mixture properties, 27 Subproblem 4: Optimal candidate, 1 Butyl 3 ethylimidazolium [Tf 2 N] The structure of the optimal ionic liquid that satisfies the constraints is shown in Figure 3 9. The optimal properties of the designed ionic liquid are shown in Table 3 19.

PAGE 81

67 Figure 3 9 : 1 butyl 3 ethylimidazolium [Tf2N] Table 3 19 : Physical properties of 1 butyl 3 ethylimidazolium [Tf 2 N] Properties 1 Butyl 3 ethyl imidazolium [Tf 2 N] Chloroform Melting point (K) 222.78 209.65 Viscosity (cP) 55.59 0.542 Naphthalene solubility @ 25 o C 0.5069 0.473 An important component of green chemistry relates to the solvent medium in which synthetic transformations are carried out 127 Traditional volatile organic solvents which act as common reaction media for several chemical processes are linked to a host of negative environmental and health effects including climate change, urban air quality and human illness. Jessop 128 states that one of the major challenges in the search for environment ally benign solvents is to ensure availability of green solvents as replacements for non green solvents of any kind. He uses the Kamlet Taft plots to show that current list of green solvents

PAGE 82

68 populate only a small region of the entire spectrum of solvents n eeded for various applications and argues that large unpopulated areas of this diagram mean that future process chemists and engineers need solvents having certain desirable properties and are green. Ionic liquids offer great potential to satisfy this need This study presents an overarching framework that can be utilized to design optimal ionic liquids for a given application through the theoretical/computational consideration of all possible combinations. Currently, the few ionic liquid structure property models that are available can be applied to a small subset of all available ionic liquid types. However, for this method to be fully effective, we need group contribution models and parameters that span the entire spectrum of ionic liquids. It is indeed p ossible to overcome this challenge as one needs property values of only few representative compounds in each class of ionic liquid (for example, covering the groups shown in A ppendix A ) to regress the contributions of the various groups. We propose that fu ture research should focus on experimental property measurements and data collection of ionic liquids that cover a diverse set of cations, anions and functional groups. The second challenge is the lack of ionic liquid structure property models (i.e. solut ion to the forward problem) for various thermo physical properties of interest. There, is a great need to develop structure property models of pure component physical properties and thermodynamic solution (mixture) properties for a comprehensive set of ion ic groups. The third challenge would relate to the accuracy of the group contribution models. However, as discussed earlier, since the primary aim of the CAILD method would be to narrow down to a small set of ionic liquids from the millions of available al ternatives, reasonably accurate predictive models are sufficient. The final IL can be selected by ab initio

PAGE 83

69 computational chemistry calculation or experimental verifications of these small set of designed compounds. Progress towards designing ionic liquid s through proposed CAILD, framework, will not only contribute towards our understanding of the relationship between cation anion structures and ionic liquid properties, but will also provide a mechanism to engineer new environmentally benign ionic liquids for critical applications.

PAGE 84

70 Application 1: Design of Ionic Liquids for Thermal Energy Storage Chapter 4: In this chapter, we present a computer aided framework to design task specific ionic liquids (ILs) for solar energy storage, using structure property models and optimization methods. Thermal energy storage density (capacity) was used as a measure of t he ability of an IL to store thermal (solar) energy. 4 .1 Introduction Advancements in solar trough and solar tower technologies have enabled concentration of thermal energy to the extent that it can be used to drive traditional steam cycles providing an a lternative to fossil fuel use 129 Therefore, harvesting solar energy using arrays of parabolic trough collectors will enable generation of electricity at a large scale. In solar power plants, thermal energy storage (TES) is necessary to extend production p eriods of low or no sunlight. An important component of TES systems is thermal fluid which is needed to transfer and store heat for relatively short periods. A parabolic trough system typically consists of a series of collectors that are big mirror like re flectors used to concentrate solar energy. When the solar radiation is received by these collectors the reflected light is concentrated at the center of the collector. A heat transfer fluid (HTF) passed through tubes present at the center of the collector absorbs the accumulated heat. The collected thermal energy is then transferred from the HTF to a storage medium or is stored in a reservoir using the heat capacity of the HTF itself 130 The storage media can then release the thermal energy when needed for further conversion to electrical energy. Thermal energy storage (TES) therefore makes solar energy a more reliable and economic alternative source of energy.

PAGE 85

71 Fluids that have high potential to store heat energy such as thermal oil (e.g. VP 1 TM ), or nitr ate salts (e.g. HITEC XL TM ) are suited for thermal energy storage applications. However, nitrate salts have melting points (freezing point) above 200C while mineral oils have upper temperature limit of 300C 130 thus limiting their use to a narrow temperature range and thereby reducing the overall efficiency of the process. Ionic liquids (and salts) have properties that are ideal for thermal storage applications. These attractive properties include high heat capacity, high decomposition temperature and relatively high density at operating conditions. Ionic liquids (ILs) are a new generation of materials that that have a wide range of applications 1 Similar to salts ILs are composed of ions but have much lower melting points. Several ILs are in liqui d state at room temperature (commonly referred as room temperature ionic liquids [RTILs]). ILs consist of an organic cation (a cation base with alkyl side chain) and a charge delocalized inorganic or organic anion 131 They usually possess good thermal stab ility (i.e. high decomposition temperature) making them appropriate for processes operating at high temperatures. Ionic liquids can be customized through appropriate selection of cations, anions and alkyl side chains. Therefore, ILs can be tuned to impart specific functionalities for a given application by changing cation/anions/side chain groups. In this study, we focus on the computational design of optimal ionic liquids with high thermal storage density for solar energy storage applications. The key req uirements of a thermal storage media include high thermal storage capacity ( high thermal stability 129 and a wide liquid range. Therefore the properties of ionic liquid that need to be optimized for thermal storage applications includ e: density, heat capacity, thermal

PAGE 86

72 decomposition temperature and melting point. Heat capacity measurements of diverse ionic liquids reveal wide ranging values 132 The melting point of ionic liquids can be easily adjusted by the choice of cation, anion or t he groups attached to the cation side chains. Thermal stability of ionic liquids has been previously studied and it has been reported that many of them have high thermal decomposition temperatures (~400 C). In order for us to find the optimal ionic liquid structure having desired melting point and decomposition temperature as well as maximum thermal storage capacity, thousands of different ionic liquids need to be tested. Since this is not feasible experimentally, a computer aided approach is suggested in this study Computer aided molecular design (CAMD) is a promising technique that has been widely used to design compounds for different applications 133 139 Gani and co workers initially conceptualized this method for screening solvents based on UNIFAC gr oup contribution approach 133 CAMD usually integrates structure based property prediction models (e.g. group contribution models) and optimization algorithms to design molecular compounds with desired properties. More recently, this approach has been extended to the design of ionic liquids 140,141,142,143,144 A comprehensive framework for computer aided ionic liquid design (CAILD) was recently published by our group 142 Key to the successful development and use of CAILD methods is the availability of p redictive models for the properties of interest. In this work we present a new CAILD model to design novel ionic liquids as thermal fluids for solar energy storage applications. This CAILD model utilizes existing group contribution methods to predict physi cal and thermal properties of ionic liquids. By considering a

PAGE 87

73 structurally diverse set of building blocks we are able to demonstrate that new and novel structural variants of ionic liquids can be tailored specifically for this application. 4.2 Formulation of the design problem In this section, we focus on presenting a computer aided ionic liquid design (CAILD) model to find (design) optimal ionic liquid structures with high thermal storage capacity, reasonably low melting point and high decomposition temper ature. In this method a variety of cation head groups, cation side chain groups (including certain functional groups), and anions were selected as ionic liquid building blocks. Typical CAILD approach requires solution to the forward problem (i.e. property prediction) as well as the reverse problem (i.e. structure generation). In a mathematical programming based CAILD approach the physical properties of ionic liquids are estimated using structure based predictive models such as group contribution (GC) model s (forward problem) and optimal ionic liquid structures are generated by solving a mixed integer non linear programming (MINLP) formulation of the design problem (reverse problem). This study utilizes GC methods from literature 145,146,147,148a,148b to pred ict the physical properties density, heat capacity, melting point and decomposition temperature. The CAILD framework proposed in Karunanithi and Mehrkesh 142 was utilized to formulate the thermal storage fluid (TSF) design problem as an MINLP model. Structu ral constraints e qn s (4 1 ) to (4 5) were included to design feasible ionic liquid structures. These constraints are a sub set of a comprehensive set of structural constraints presented in Karunanithi and Mehrkesh 142

PAGE 88

74 Linear physical property constraints based on GC predication e qn s (4 11 ) to (4 15) were integrated with the structural feasibility constraints. The objective of the design problem was to identify the optimal ionic liquid structure that has the highest thermal storage capacity. Therefore, the objective function was formulated to maximize the product of density and heat capacity of the ionic liquid. The solar thermal storage process is typically carried out at temperatures of around 300 Therefore, the thermal storage fluid in t his case the designed ionic liquid should be operable at temperatures slightly above 300 To ensure that the ionic liquid does not decompose during the process, we enforce a constraint on thermal decomposition temperature to be above 400 e qn (4 24). T he temperature of ionic liquid after energy exchange should be higher than its melting point. To ensure this, a constraint on melting point to be above 140 is imposed e qn (4 25). The temperature window between melting point and decomposition temperature is the range at which the process can operate. The basis set (the structural building blocks) considered for this problem included 5 cation head groups, 9 anions and 5 side chain groups (alkyl and functional groups) which are shown in Table 4 1. This sele ction was based on groups for which group contribution parameters were available for all the properties of interest. Table 4 1 : Ionic l iquid b uilding b locks (groups) c onsidered for t hermal f luid d esign Cation Anions Groups Imidazolium Tetrafluoroborate [BF 4 ] Methylene ( CH 2 )

PAGE 89

75 Cation Anions Groups Pyridinium Hexafluorophosphate [PF 6 ] Methyl ( CH 3 ) Ammonium bis(trifluoromethylsulfonyl) imide Benzyl Phosphonium Chloride Methoxy Pyrrolidinium Bromide Hydroxyl ( OH ) Trifluoromethanesulfonate Benzoate Nitrate

PAGE 90

76 Cation Anions Groups Acetate Objective function ) Constraints 1) Structural (feasibility) Constraints 2) 3) 4. 2.1 Ionic l iquid s tructural c onstraints A key requirement in computational design of ionic liquids is the ability to guarantee solutions that are theoretically feasible chemical structures. Further, one would have to account for practical considerations such as limits on the size or presence/a bsence of certain groups etc. that will lead to design candidates that are workable from a synthesis view point. In order to incorporate these considerations the proposed method requires that the designed compound satisfies certain rules, broadly termed as structural constraints, that includes chemical feasibility rules such as octet rule, bonding rule and complexity rules as

PAGE 91

77 well as rules that restrict/determine the size and the constituents of possible solutions. More details about these structural cons traints can be found in Karunanithi and Mehrkesh 142 The specific structural constraints from Karunanithi and Mehrkesh 142 that are invoked for this problem are discussed below The first two constraints eqn s. (4 1 ) and (4 2) ensures the selection of only one cation and one anion from the basis set. ( 4 1 ) ( 4 2 ) Where C is a set of all cation head groups considered (i.e. i midazolium, p yridinium, a mmonium, p hosphonium and p yrrolidinium); and A is a set of all anions considered. The next set of feasibility constraints deal with valence requirements of cations which enables us to add appropriate side chain groups. Per our definition, anion is considered as a single group and does not require constraints related to addition of groups and therefore valence constraints are relevant only for the cation part. The three equations below make sure that octet rule for the cation as a whole as well as for each side chain (branch) is not violated. ( 4 3) ( 4 4 ) ( 4 5)

PAGE 92

78 Where we fixed n to be 2, 1, 4, 4, and 2 for imidazolium, pyridinium, ammonium, p hosphonium and p yrrolidinium respectively. This was based on the fact that for imidazolium, pyridinium and pyrrolidinium the side chains are commonly connected to only the nitrogen atoms in the cation ring. However, note that if we want to broaden the design problem we can allow side chains to be a ttached to the carbon atoms in the cation rings by adjusting these constraints. More specific details and definitions about these constraints can be found in Karunanithi and Mehrkesh 142 The next set of constraints deal with restricting the size of cation as well as putting limits on the presence/absence of certain groups. Limits on the total number of groups that can occur on each side chain as well as the whole cation was fixed using e qn s (4 6 ) and (4 7 ) These two constraints make sure that we do not design a very large cation, which cannot be practically synthesized. Limits on the number of functional groups with valence 1 (i.e. OH, OCH 3 and benzyl) that can occur in the whole cation was fixed using e q n (4 8 ). ( 4 6) ( 4 7) ( 4 8) Where, G is a set of all groups considered (i.e. CH 2 CH 3 OH, OCH 3 and benzyl) and is a subset of G which consist of functional groups OH, OCH 3 and benzyl. Eqn (4 8 ) is invoked three times for each while for each of these three equations was assigned

PAGE 93

79 to be 2,1,2,2,2 for i midazolium, p yridinium, a mmonium, p hosphonium and p yrrolidinium head groups respectively. 4. 2.2 Ionic liquid p roperty prediction This se ction describes in detail the physical properties of ionic liquids that are relevant for thermal storage applications and the structure based property prediction models and correlations that are utilized to predict these properties within the proposed comp uter aided ionic liquid design (CAILD) framework Thermal storage density Thermal storage density ( of an ionic liquid is defined as a product of its heat capacity and density. ( 4 9) Since both density and heat capacity are a function of temperature, thermal storage densities of ionic liquids vary with temperature. is the most critical design parameter for thermal fluids as higher value of will result in lower volume of thermal f luid requirement. Heat Capacity The heat capacities of ionic liquids were predicted using the approach developed by Valderrama et al. 147 In this method the authors used the concept of mass connectivity index (MCI) to build a structure based predictive mod el. Molecular connectivity was first

PAGE 94

80 introduced by Randic 147 and has been used then by several authors for property prediction. As suggested by Valderrama et al. 147 the MCI concept can be used to quantify the extent of branching in ionic liquids thereby enabling us to predict IL heat capacities better. This index considers the mass of structural groups as well as the type of connections between them as following: ( 4 10) where, m i and m j are the mass of neighboring groups i and j in a molecule. In this expression, the sequence of groups is important, as i and j are two distinct groups and hence the connection m i m j is different from m j m i Valderrama et al 147 showed that the MCI approach i s capable of predicting the heat capacity as a function of temperature for a variety of ionic liquids with an acceptable level of accuracy. To predict the ionic liquid heat capacity, a reference value of C P0 at reference condition is used as follows ( 4 11) experimental data of ionic liquid heat capacities. The overall C p, as a temperature dependent variable, can be estimated as a function of molar mass of cation and anion, molar volume of ionic liquid and mass connectivity index as follows ( 4 12) where, T 0 is the reference temperature, 298.15 K, a = 15.80, b = 1.663, c = 28.01, d = 7.350,

PAGE 95

81 e = 0.2530, f = 1.37210 5 ,V m is the molar volume (cm 3 (MCI), and of ionic liquid. Density Density of ionic liquids were predicted using the approach presented by Valderrama et al. 145 In this approach the density of each ionic liquid at a given temperature was estimated as a secondary property using its critical properties (T c and V c ) and normal boiling point (T b ) as primary properties through the following expression e qn (4 13): ( 4 13) where, = 1.0476. Critical properties and normal boiling point of the ionic liquids were estimated using the group contribution method proposed by Valderrama et al. 146 Melting Point An ionic liquid with low melting point would be desirable since at all times the operating temperature should be kept above the melting point of the thermal storage fluid to avoid solid formation in the system. Alternately, when the melting point is too hi gh a high process temperature needs to be maintained which will decrease the system efficiency by reducing

PAGE 96

82 the rate of heat transfer (sensible heat). An appropriate thermal storage fluid should possess a melting point lower than 140 C. To calculate the me lting point of ionic liquids a group contribution method proposed by Lazzus 148b was used. In this approach two different sets of contribution for melting point were used: 1) the contribution of cation head group (cation base) and alkyl groups/functional gr oups attached to the side chains of the cation head groups; 2) the contribution of groups associated with the anion. Cation head groups are tabulated as whole (e.g. imidazolium or pyridinium) but side chain groups and anions are split into smaller structur al fragments. The melting point of any given ionic liquid can be calculated by the summation of contribution of cations, anions and side chain groups as following: ( 4 14 ) where is the number of occur rence of group i, is the contribution of cation to the melting point and is the contribution of anion to the melting point of the given ionic liquid. Thermal Decomposition Temperature The final property of interest is thermal decomposition temperature. This property is an estimate of the highest temperature at which the ionic liquid will remain in the associated state (non decomposed). It is important for thermal storage as it will determine the maximum applicable temperature at which the t hermal fluid can be utilized. To predict the

PAGE 97

83 decomposition temperature of ionic liquids a group contribution method proposed by Lazzus 148a was utilized. This method is similar to the melting point prediction described above. ( 4 15 ) where is the number of occurrence of group i, is the contribution of cation to the decomposition temperature and is the contribution of anion to any given ionic liquid. 4.2.3 CAILD m odel s olution The complete CAILD MINLP model is shown below: Objective function ) Constraints ( 4 16) ( 4 17) ( 4 18) ( 4 19) ( 4 20)

PAGE 98

84 ( 4 21) ( 4 22) ( 4 23) ( 4 24) ( 4 25) ( 4 26) ( 4 27) ( 4 28) ( 4 29) ( 4 30) ( 4 31) .012 ( 4 32)

PAGE 99

85 ( 4 33) ( 4 34) ( 4 35) The above computer aided ionic liquid design (CAILD) model is a Mixed Integer Non Linear Programming (MINLP) formulation. MINLP models combine combinatorial aspect s with nonlinearities and are more difficult to solve than mixed integer programming (MIP) and non linear programming (NLP) problems. The most precise approach to solve this MINLP model would be to fully enumerate each possible combination within the entir e search space. The number of possible combinations increase exponentially with number of groups considered resulting in combinatorial explosion requiring large computational times. Other multi level approaches have been proposed to address this issue and avoid complete enumeration 149,150 Several deterministic optimization based methods have been employed to solve CAMD MINLP models 137,138,151 Different stochastic methods have also been used to solve CAMD problems 152,153,154 Property clustering technique used within a reverse problem formulation is yet another approach that has been successfully used to solve CAMD problems 155,156 In this work a genetic algorithm (GA) based solution approach is utilized to solve this optimization model. GA is a stochastic biological evolution (principle of natural selection) and has been previously used to successfully solve complex optimization (minimization) problems of different formulations.

PAGE 100

86 GA can be an attractive alternate solu tion approach for problems that are not well suited for standard optimization algorithms such as problems with discontinuous, stochastic, or highly nonlinear objective functions 157 The optimization process usually starts with a random collection of initia can exchange their genetic information, through crossover and mutation operations. At each iteration, a new population of fitter candidate structures is generated to replace the ex isting population and this process is repeated for a pre specified number of iterations or until the pre defined value of tolerance for objective function is met 157 The GA solution of CAILD MINLP problem was implemented in the MATLAB environment with most parameters set at their default values. A large population size which was generated randomly by the program was used to start the search process. The crossover fraction was fixed at 0.8, whereas a uniform mutation probability with rate value of 0.01 was u sed. 4.3 Results and d iscussion The optimal ionic liquid consists of a hydroxyl functionalized imidazolium cation and tetrafluoroborate anion whose structure is shown in Fig ure 4 1 Figure 4 1 : A schematic of the structure of optimal IL with highest thermal storage capacity

PAGE 101

87 In Table 4 2, we list the physicochemical properties of the optimal IL along with two common commercial products (VP 1 TM a heat transfer oil and Hitec XL TM a molten salt) used as thermal storage medium 130 The designed ionic liquid has a higher thermal storage capacity than VP 1 TM oil and has about the same value of Hitec molten salt. However, the estimated viscosity of the designed ionic liquid is an order of magnitude lower than that of the molten salt making it a better alternative. The lower viscosity is extremely important as the overall commercial feasibility of the process depends on pumping energy and low viscous TES medium would result in less operating costs. Overall, the design results suggest that the optimal ionic liquid has better thermal storage properties than existing commercial products. Table 4 2 : Thermo physi cal properties of VP hydroxy Imidazolium] + [BF 4 ] Properties (@ T) VP 1 TM Hitec XL TM [3 hydroxy Imidazolium] + [BF 4 ] Melting Point ( 13 120 129.8 Decomposition Temp ( 400 500 578.1 Density (kg/m 3 ) 815 (300 1992 (300 1477 Heat Capacity C p (J/Kg K) 2319 (300 1447 (300 1947 (300 3 K) 1.9 (300 2.9 (300 2.88 (300 Viscosity (cP) 0.2 (300 6.27 (300 0.88 (300

PAGE 102

88 We performed a comprehensive set of analysis to study in more detail the relationship between ionic liquid structural components and their thermal storage properties. First, we considered each category (head group) of cation separately and identified the o ptimal structures and studied the effect of different anions considered in this study in terms of their influence on the TES properties (Table 4 3). It is clear tha t among all the anions considered, [BF 4 ] always resulted in the highest thermal storage capacity irrespective of the cation type. This finding is also consistent with our optimal IL candidate, which too had [BF 4 ] anion. Further, from these tables it is a lso clear that imidazolium cation always resulted in the highest thermal storage capacity, which is consistent with our optimal IL candidate that had an imidazolium cation Table 4 3 : Effect of anion variation on the thermal storage properties of ionic liquids Cation Anion BF 4 PF 6 Tf 2 N Cl Br Density (kg/m 3 ) 1477.5 1548.9 1697.7 1415.6 1788.3 C p (J/kg K) 1947.36 1640.3 1527.9 1843.4 1453.4 p (MJ/m 3 K) 2.88 2.54 2.59 2.61 2.6 T m (K) 402.9 422.1 350.0 438.1 436.3 T app (K) 851.3 857.3 882.0 721.8 757.9 Cation

PAGE 103

89 Anion BF 4 PF 6 Tf 2 N Cl Br Density (kg/m 3 ) 1181.5 1260.5 1414.7 1101.4 1286.7 C p (J/kg K) 2256.9 1982.2 1798.5 2226.5 1934.5 p (MJ/m 3 K) 2.67 2.50 2.54 2.45 2.49 T m (K) 409.7 428.9 356.7 444.9 443.05 T app (K) 765.6 771.5 796.3 636.1 672.2 Cation Anion BF 4 PF 6 Tf 2 N Cl Br Density (kg/m 3 ) 1227.6 1309.4 1474.4 1156.5 1363.1 Cp (J/kg K) 2198.1 1919.5 1743.6 2155.8 1852.8 3 K) 2.70 2.51 2.57 2.49 2.52 Tm (K) 386.4 405.7 333.5 421.6 419.8 Tapp (K) 828.6 834.6 859.3 699.2 735.3 Cation Anion BF 4 PF 6 Tf 2 N Cl Br Density (kg/m 3 ) 1241.7 1317.4 1470.0 1174.0 1367.4 Cp (J/kg K) 2036.7 1812.1 1675.7 1960.7 1724.6 3 K) 2.53 2.39 2.46 2.3 2.36 Tm (K) 378.0 397.2 325.1 413.2 411.3 Tapp (K) 850.8 856.8 881.5 721.3 757.4 Cation Anion BF 4 PF 6 Tf 2 N Cl Br Density (kg/m 3 ) 1022.8 1075.9 1190.8 965.9 1063.5 Cp (J/kg K) 2538.3 2340.9 2127.8 2502.6 2337.3 3 K) 2.59 2.52 2.53 2.42 2.49 Tm (K) 410.4 429.6 357.4 445.6 443.7 Tapp (K) 691.6 697.6 722.3 562.2 598.3

PAGE 104

90 Next, in order to study the effect of methylene groups present in the cation side chain on the TES properties of ILs we fixed the optimal anion as [BF 4 ] For each category of cation, a variant of the optimal structure (in terms of functional groups) was identified and the number of methylene group was varied from 0 to 4. The results are shown in Table 4 4. It is clear from Table 4 4 that addition of methy lene group decreases the thermal storage capacity, melting point and decomposition temperature consistently across all categories. This shows the tradeoff that needs to be made, as higher thermal storage capacity and decomposition temperature are preferred while lower melting point is required. Therefore, in general, absence of CH 2 CH 2 groups (e.g. 11 CH 2 groups in the case of pyrrolidinium) to meet the tight melting point constraint. These findings are also consistent with the designed IL, which did not have any methylene groups. Table 4 4 : Effect of number of CH 2 m and T app [ILs with BF 4 anion] Number of CH 2 0 2 4 Cation (MJ/m 3 K) T m (K) T app (K) Cation T m T app Cation T m T app Imidazolium 2.88 402.9 851.3 2.77 395.4 843.2 2.71 387.9 835.1

PAGE 105

91 Number of CH 2 0 2 4 Cation (MJ/m 3 K) T m (K) T app (K) Cation T m T app Cation T m T app Pyridinium 2.80 428.5 785.8 2.73 420.9 777.7 2.68 413.4 769.6 Ammonium 2.70 386.4 828.6 2.66 378.9 820.5 2.64 371.4 812.4 Phosphonium 2.53 378.0 850.8 2.52 370.4 842.7 2.50 362.9 834.6 Pyrrolidinium 2.67 440.4 724.0 2.65 432.9 715.9 2.63 425.4 707.8 Next, to study the influence of the functional group (FG=hydroxyl or benzyl or ether or methyl) present in the cation side chain on the TES properties of ILs we fixed the optimal anion as [BF 4 ] For each category of cation, a variant of the optimal structure was identified and the functiona l group (position marked as FG in Table 4 5) was varied. From Table 4 5 we see that cations functionalized with benzyl group always had the highest thermal storage

PAGE 106

92 capacity followed by hydroxyl functionalized cations. However, we also see that benzyl funct ionalized cations have significantly lower thermal decomposition temperature in comparison to hydroxyl functionalized cations thereby mostly violating the decomposition temperature requirements (constraints). Further, benzyl functionalized cations usually had a slightly higher melting point than hydroxyl functionalized cations. These findings are consistent with our optimal IL, which consisted of a hydroxyl functionalized imidazolium cation. Table 4 5 : The effect of variation of functional groups (FG) connected to the m and T app Functional group (FG) CH 3 Benzyl OCH 3 OH Cation T m T app T m T app T m T app T m T app Imidazolium 2.76 413.5 661.6 2.90 411.8 613.1 2.79 403.0 534.3 2.88 402.9 851.3 Pyridinium 2.61 420.3 575.9 2.74 418.6 527.4 2.63 409.8 448.6 2.67 409.7 765.6

PAGE 107

93 Functional group (FG) CH 3 Benzyl OCH 3 OH Cation T m T app T m T app T m T app T m T app Ammonium 2.63 397.1 638.9 2.77 395.4 590.5 2.65 386.6 511.7 2.70 386.4 828.6 Phosphonium 2.48 388.6 661.1 2.64 386.9 612.6 2.50 378.1 533.8 2.53 378.0 850.8 Pyrrolidinium 2.57 421.0 501.9 2.62 419.3 453.50 2.58 410.5 374.7 2.60 410.4 691.6 Since the hydroxyl functionalized optimal IL is, a novel and non traditional IL there were no experimental data available for validation purposes. Therefore, we relied on QSPR p ) based on quantum chemical calculation inputs implemented within COSMOtherm software to cross check our results. The optimal ionic liquid (3 hydroxy imidazolium [BF 4 ] ) was simulated in Turbomole, which was linked to COSMOtherm software. In the first step to evaluate the accuracy of QSPR correlation in

PAGE 108

94 COSMOtherm we compared exper imental heat capacity (C p common imidazolium based ionic liquids with corresponding prediction from COSMOtherm QSPR models. The comparison results are tabulated in the Table 4 6. As it can be seen from Table 4 6 the QSPR/COSMO therm predictions were very close to the experimental data with an average relative error of 1.49%. Next, Table 4 7 shows the comparison of the physical properties of the optimal ionic liquid (3 hydroxy imidazolium [BF 4 ] ) predicted by QSPR/COSMOtherm appr oach and the group contribution models used in this study. As it can be seen from T able 4 7, the predictions are close which further validates our design results. Table 4 6 : Comparison of COSMO predicted C p a corresponding exp. data BF 4 PF 6 Tf 2 N C p C p C p Exp COS Exp COS Exp COS Exp COS Exp COS Exp COS [emim ] 308.1 313.29 1285.5 1300.65 --350.43 1480.5 1496.59 505 490.17 1517 1545.94 [bmim] 364.8 368.91 1201.6 1205.37 397.6 406.05 1367.7 1376.93 564 545.79 1436.1 1451.68 [hmim] 426.1 424.21 1145.5 1139.33 419.25 431.35 1292.7 1291.14 623 601.08 1372.0 1377.78 Avg. Relative error (%) 1.49 % Max Relative error (%) 3.52%

PAGE 109

95 Table 4 7 : Comparison between COSMO predicted and GC predicted values for C p and Method Properties COSMO RS Group Contribution Abs Relative Difference (%) Cp (J.mol 1 .K 1 ) 240.98 231.42 3.97 3 ) 1600.34 1624.1 1.48 3 K) 2.23 2.174 2.54

PAGE 110

96 Application 2: Design of Ionic Liquids for Aromatic A liphatic Chapter 5: Separation In this chapter, a computer aided design framework with the capability of predicting ionic liquid selective dissolution power towards aromatic compounds is presented COSMO RS method was first evaluated for its accuracy for the prediction of the solubility of liquid solutes in ionic liquids and then these prediction were utilized within an optimization framework to fit the binary interaction and group parameters for UNIFAC model Next, the fitted parameters along with the UNIFAC model was integrated within the CAILD framework to design optimal ionic liquids for this separation problem. 5.1 Introduction In order for a chemical compound to be considered as a green substitute of a common dverse impact these solvents might have on the environment. 158 A liquid liquid extraction process is normally used when other methods of separation (e.g. distillation) are not feasible. Aromatic/Aliphatic separation is one of the well known industrial proc esses in which a chemical (solvent) is utilized to withdraw/remove aromatic compounds from a multi component mixture of aromatic and non aromatic compounds. Current industrial plants, which use organic solvents such as furfural or sulfolane as the solvent in the extraction process, show several adverse environmental impacts. 159 162 Thus, substitution of commonly used solvents with their ionic liquid counterparts can help address some of these deficiencies Several studies have shown that when ILs are added to a binary mixture of one aromatic and one non aromatic

PAGE 111

97 compound, they mostly follow a type II ternary system, meaning that they are technically able to process feed streams containing aromatic compounds at different compositions. 163 On the other hand, ionic liquids are known to have the potential to act as solvents in separation processes where a low concentration of a solute is available, and thereby, can be economically beneficial when compared to their conventional organic counterparts. 164 Ionic liquids have also shown a lower tendency towards dissol ution in hydrocarbon compounds, leading to lower solvent losses during the regeneration step 165 In order for an ionic liquid (IL) to be considered as a solvent for an aromatic/aliphatic separat ion process it needs to show high values for selective extraction capacities towards aromatic compounds. 166 As discussed in our previous work, 166 a large inventory of anions, cations, and functional groups exist or can be synthetized. 167,168 ILs can be f ormed through different combinations of cations, anions, and alky side chain groups (attached to the cation base) leading to a vast number of feasible ionic liquid structures (estimated to be as many as 10 14 combinations). 169,170 This presents a very good opportunity to tailor ILs for specific applications. Task specific ILs can potentially be designed for different applications. The design process is usually accomplished through controlling the properties of ILs by informed selection or modification of the cation, the anion, and/or the alkyl chains connected to the cation base. While this can be very exciting to potentially screen a large number of feasible ionic liquids, it can also present an unusual challenge, since synthesizing, screening, and testing t he limitless number of ionic liquids can become a daunting task. 171

PAGE 112

98 As mentioned before, the ILs can be tuned to possess the properties of interest. Similarly to what has been done before for molecular compounds, we can tailor ILs to increase their desired properties (such as solvency power or thermal storage capacity), while decreasing their undesired features ( e.g. high viscosity or high toxicity) at the same ti me. Although ILs have many interesting properties, they have not been commercialized yet. One of t he reason for this could be that ILs possess high viscosities and relatively high melting points. These drawbacks must be addressed in order to make ILs mor e appealing to industrial stakeholders. Computer aided molecular design (CAMD) has been a promising approach for years now and has been widely used to design organic (molecular) solvents for different applications. 172 178 A CAMD framework integrates model s used for property prediction with optimization algorithms to design molecular structures with desired properties. In our previous work 166 we successfully showed that a similar approach can be applied to the case of designing new I L s through the numerous possible combinations of cations, anions, and side chain groups. 5.2 Computer a ided i onic l iquid d esign (CAILD) For this study, the CAILD framework that we developed 166 is used again to identify ionic liquids (ILs) that possess desirable properties for aromatic/aliphatic separation In the CAILD approach a set of IL structural groups (i.e., cations bases such as: Imidazolium (Im + ) Pyridinium (Py + ) ammonium (NH 4 + ) anions such as: Tf 2 N BF 4 and side chain groups such as: CH2, CH 3 OH etc.) are considered as building blocks and are combined together in

PAGE 113

99 different ways to design ILs which have specific desired properties of interest (e.g., melting point, viscosity, solubility To design an optimal candidate (ILs) as an extraction solvent for toluene/n heptane separation process, a comprehensive approach consisting of five stages was used. The presented CAILD model in this study is a nonconvex, Mixed Integer Non linear Programming (MINLP) model involving a large number of integer and binary va riables. Consideration of mixture properties through the UNIFAC model results in nonlinearity and most of the binary design variables (structural) participate in the nonlinear terms. Combinatorial complexity is an inherent issue in CAMD MINLP models due to the nature of the search space. In this study we focus on solving the CAILD framework utilizing GA based optimization. Genetic algorithm (GA) GA is a stochastic method used to solve optimization problems based on the natural selection process mimicking t he evolution phenomenon which occur in biological systems. GA can be used to solve problems that are not well suited for standard optimization algorithms. 179 In the presented case study, the fitness function of the GA is identical to the objective function. 5.2.1 Forward problem In order to evaluate the intermolecular interactions between different components present in a mixture (both ionic and molecular), the thermodynamic properties of non ideal solutions need s to be calculated. An essential requirement to predict the equilibrium conditions of a

PAGE 114

100 system involving ionic liquids is to calculate the activity coefficients of compounds in that system. These activity coefficients can be used to predict the equilibrium conditi ons of the system and are not only a function of binary (group group) interactions, but also relate to the compositions of the chemical components in the mixture. Therefore the proposed CAILD framework requires a model for predicting activity coefficient that is based on the concept of solution of groups. The basic assumption related to the prediction of activity coefficients of compounds in a mixture using solution of group approach is that the interactions between different molecules can be approximated as the summation of interactions between the groups present in those molecules. T he total number of cation head groups, anions, and alkyl side chain groups are much les s than the number of distinct ILs which can be generated from their combinations. Therefore, a relatively small number of group interaction parameters are required (are enough) to represent all feasible ILs. UNIFAC 180 model is a widely used group contribut ion (GC) method to predict the phase equilibrium conditions in nonelectrolyte systems. The UNIFAC method makes use of the concept of functional groups (their contribution and the number of their occurrences in each compound) to predict the activity coeffic ients. The activity coefficient calculated using the UNIFAC model is the summation of two terms: a combinatorial term that considers the differences in size and shape of groups and a residual term that accounts for the energetic interactions of different groups. The volume ( R ) and surface area ( Q ) parameters related to the combinatorial part of the activity coefficients of each compound are calculated as the summation of group parameters

PAGE 115

101 ( R k for volume and Q k for surface area), while binary interaction parameters ( a mn and a nm ) between different functional groups present in the components of the system are required to calculate the residual part. The UNIFAC model was originally developed for nonelectrolyte systems, however, recen tly several groups have utilized this method for system s containing ILs by careful representation of ionic groups in the system. In order to apply the UNIFAC method to systems with ILs Wang et al. 181 and Lei et al. 182 considered ionic liquids as single non dissociated neutral entities. IL s as separate functional groups are included in the modified UNIFAC. 183 Recently, Roughton et al. 184 characterized ionic groups in the same way as proposed in our developed CAILD framework ; i.e. as separate cation bases, anions, and side chain groups. The assumption here is that the cation and anion can be treated separately and the interaction between them is assumed to be equal to zero. 184 Due to paucity of experimental data on activity coefficients of chemical compoun ds dissolved in wide variety of ILs, fitting the relevant UNIFAC parameters for systems involving ILs is not possible For the time being, the main limitation on the use of computer aided models to design optimal ionic liquids for multi component separatio n processes can be attributed to the lack of enough experiential data on activity coefficients E ven though millions of different ILs can be theoretically feasible, due to the gaps in the experimental data and non availability of property prediction models a vast majority of ionic liquids cannot be considered within a computer aided design fra mework In order to overcome this limitation, we use a quantum chemistry (QC) approach to simulate the chemical compounds and ionic liquids (based on energy minimization), to predict their activity coefficients in the mixture. COMSO RS (COnductor like Screening M Odel for Real Solvents), is a quantum chemistry

PAGE 116

102 based equilibrium thermodynamics model that can predict the chemical potentials of compounds in the liquid phase. COSMO RS is able to calculate the distribution of charge compound in the solution. The che mical potentials calculated through COSMO RS approach are the basis for the calculation of other thermodynamic equilibrium properties such as activity coefficients solubility and free energy of solvation. The advantage of this method is that it was developed as a general prediction method that does not require any system specific adjustments. S ince COSMO require functional group parameters, meaning that we do not require experimental data to predict activity coefficients. C OSMO RS has shown to be a promising tool for a solvent screening task in which the most powerful solvent for a specific liquid liquid extraction process can be selected. During the past few years many researchers have used COMSO RS to calculate the solubi lity and/or activity coefficients of different chemical compounds in mixtures containing ILs. In order to study the ability of COSMO RS method to accurately predict activity coefficients of chemical compounds in ILs a comparison was made between experime ntal values of and COSMO RS predictions These results along with the absolute error (%) and the temperature at which the activity coefficients were measured/estimated are tabulated in Table 5 1. Results in Table 5 1 show that in most cases, COSMO RS has been able to predict the activity coefficients (i.e. solubility) of chemical compounds in their mixtures with ILs to a reasonable degree of accuracy.

PAGE 117

103 Table 5 1 : Experimental infinite dilution activity c oefficients ( ) vs. Cosmo predicted values IL T (k) Err. (%) T(k) Err. (%) T(k) P66614 Cl 308.15 318.15 328.15 0.408 0.403 0.401 0.432 0.430 0.429 5.84 6.78 7.23 308.15 318.15 328.15 0.766 0.753 0.746 0.695 0.688 0.681 9.23 8.65 8.75 308.15 318.15 328.15 Bmim Tf 2 N 303.15 313.15 323.15 0.881 0.892 0.903 1.180 1.138 1.099 33.98 27.63 21.75 303.15 313.15 323.15 14.2 13.5 12.7 14.294 12.445 10.911 0.66 7.81 14.09 303.15 313.15 323.15 BmPyr Tf 2 N 303.15 313.15 323.15 333.15 0.84 0.86 0.88 0.89 1.121 1.081 1.045 1.011 33.46 25.73 18.70 13.55 303.15 313.15 323.15 333.15 13.8 13.3 12.1 11 10.027 8.821 7.814 6.968 27.34 33.68 35.42 36.65 303.15 313.15 323.15 333.15 H mim Tf 2 N 301.65 312.25 333.25 343.75 0.779 0.776 0.777 0.769 0.977 0.945 0.888 0.863 25.37 21.74 14.31 12.22 298.15 313.15 333.15 8.330 7.710 6.580 8.766 7.340 5.911 5.23 4.79 10.16 298.15 313.15 333.15 BmPy TfO 298.15 318.15 338.15 358.15 1.41 1.46 1.5 1.54 1.697 1.593 1.497 1.410 20.37 9.11 0.20 8.43 Average error (%) 16.41 Based on our results we suggest that we c an use the activity coefficients predicted by COSMO RS method as surrogates for missing experimental data. These predicted activity coefficients were used within an optimization framework (minimizing the relative error) to fit both group parameters (R and Q) and the binary interaction parameters (a ij and a ji ) of the UNIFAC model. The advantage of this approach is that t hese fitted parameters cover a wider range of binary coefficients related to multitude of cations, anion, and functional groups compared to wha t would have been possible using experimental data. A list of cations, anions, and side chain group s considered in this study are listed in Appendix B The UNIFAC parameters (group and binary interaction parameters) fitted through the discussed

PAGE 118

104 procedure (listed in Appendix C) was used in a computer aided design (CAILD) framework to explore a wide variety of ionic liquids (~29,000) with an aim to design the most optimal solvent for the liquid liquid extraction process. These UNIFAC parameters for imidazoli um and pyridinium cation head groups (which appeared in all the 8 optimal ionic liquids), along with all considered anions and functional groups are listed in Appendi ces C 1 and C 2 respectively The accuracy of the fitted UNIFAC model was tested using ava ilable experimental data on the solubility of toluene and n heptane in different ionic liquids and the results along with the relative errors are tabulated in Table 5 2. Table 5 2 : Experimental solubility data vs. CAILD predicted data Ionic Liquid Solute Ref 1 hexylpyridinium BF 4 Toluene 298.15 0.421 0.03 0.49458 17.48 185 1 hexylpyridinium BF 4 Toluene 318.15 0.441 0.03 0.518175 17.500 185 1 hexylpyridinium BF 4 Toluene 338.15 0.453 0.03 0.542791 19.82 185 1 hexyl 3 methylimidazolium Tf 2 N Toluene 298.15 0.487 0 0.35888241 26.31 186 1 ethyl 3 methylpyridinium Tf 2 N Toluene 293.15 0.295 0.012 0.3237049 9.73 187 1 ethyl 3 methylpyridinium Tf 2 N Toluene 303.15 0.298 0.012 0.3269968 9.76 187 1,3 dimethylimidazolium Tf 2 N Toluene 313.2 0.5891 0.521006 11.56 188 1 pentyl 3 methylimidazolium Tf 2 N Toluene 298.15 0.577 0.013 0.2745038 52.42 189 1 octyl 3 methylimidazolium PF 6 Heptane 313.2 0.0095 0.0014 0.0080512 15.25 190 1,3 dimethylimidazolium Tf 2 N Heptane 313.2 0.0178 0.002 0.0123246 30.76 188 1 hexyl 3 methylimidazolium T f 2 N Heptane 313.2 0.1375 0.0061 0.154986 12.72 188 1 ethyl 3 methylpyridinium T f 2 N Heptane 298.15 0.042 0.014 0.0336166 19.96 191

PAGE 119

105 5.2.2 Reverse p roblem Naser and Fournier 192 used a strategy for the computer aided design of molecular compounds in which the features related to the process model were also included (as the optimization constraints) This framework was used to find an optimal solvent (i.e. an ionic liquid) for the extraction process. One of the advantage of this approach is that it enabl es us to predict the compositions of the three components (toluene, n heptane and IL) in the raffinate and extract phases as well. In order to illustrate the use of the CAILD method using the UNIFAC parameters developed in the Forward problem, we solve a cas e study for an aromatic/aliphatic separation process. Case Study: Toluene/n heptane separation Recovering toluene, an aromatic compound, from a toluene/n heptane mixture is desired. Due to the fact that boiling points of toluene and n heptane are in the close vicinity of each other, the distillation process is not a feasible method for this separation and instead a solvent extraction process (liquid liquid extraction) is commonly used The feed was assumed to be a mixture of toluene (A) and heptane (B) wi th mole fraction of toluene, and heptane, being 0.27 (24 wt% ) and 0.73 (76 wt%), respectively with the molar flow rate of F=1000 mole/hr. This data is based on a real case scenario adopted from an industrial plant producing lubricating bas e oil through the removal of petroleum based aromatics from the lube oil cut. The solvent was assumed to be a pure ionic liquid that is to be designed and pure furfural for comparison (common solvent for aromatic/aliphatic separation) with the molar flow r ate of 2450 mole/hr. A schematic of the single stage extraction compartment is shown in F igure 5 1.

PAGE 120

106 Figure 5 1 : A schematic of the single stage extraction process Detailed Formulation of the Problem In this optimization framework we seek to minimize the extraction of n heptane from the feed stream while extracting as much toluene as possible s ince our goal is to maximize the concentration of n heptane in the raffinate phase. In order to find a proper objective function, a performance index (PI) was used through the division of the yield of extraction of toluene by yield of extraction of heptane squared (since the lower amount of the yield of extraction of n heptane is more important). These two yield of extractions were defined as the fractions of toluene and n heptane in the feed stream which was separated from it during the solvent extraction process. Objective function = Max ( ( 5 1) Feasibility constraints (5 2)

PAGE 121

107 (5 3) (5 4) (5 5) (5 6) (5 7) Solution property constraints ( 5 8) (5 9) (5 10) (5 11) (5 12) (5 13) (5 14) (5 15) (5 16)

PAGE 122

108 (5 17) (5 18) As discussed before, t he optimal ILs designed through our CAILD framework must satisfy certain rules to ensure the chemical feasibility of the proposed compounds. These rule s also termed as structural constraints, include feasibility rules (e.g. the octet rule), the bonding rule, as well as other complexity rules. Eq ns. (5 2 ) to (5 6 ) represent a set of constraints developed to ensure that the designed ILs are chemically feas ible which are adopted from our previous work. 166 Eq n. (5 2 ) ensures that only one cation from the vector c i representing the entire inventory of cations, is chosen. In the same way, eq n. (5 3 ) ensures that only one anion from the vector is chosen. is a vector of binary variables (0,1) dealing with the existence (1) or non existence (0) of vacant valence s on the cation base. For example means that there is an open/vacant position on the atom number 1 (location 1) of the catio n base to which a side chain group can be attached. is a vector of integer variables representing the number of groups of type k in the alkyl side chain l is the Valence of the selected cation (i.e. the number of open positions a cation base has for side chain s ) and is the Valence of the functional or alkyl side chain groups (e.g. CH 2 has a Valence of 2, so it can be connected to two other groups ; CH 3 has a Valence o f 1 and can be a terminal group attached to the end of a chain or to the cation base directly). Eq. 4 fixes the number of alkyl side chains attached to the vacant valence s of the cation base. The implementation of the modified octet rule (Eq. 5) ensures th at any designed cation is

PAGE 123

109 structurally feasible meaning that each Valence in all structural groups of the cation is filled with a covalent bond. Eq. 6 ensures that the octet rule is implemented for each side chain ( l ) in the sense that the valence s of eac h individual chains are satisfied with a covalent bond. Eq. 7 restricts the size of the side chain groups connected to the cation base and in this case has an upper bound of 15 for the number of CH 2 groups present in each side chain. E qns. (5 8 ) to (5 10 ) deal with the properties of interest of the designed solvent to show how well the optimal candidate can satisfy our needs. As aforementioned we need to design a solvent for toluene/n heptane separation which has the solvent properties better than those of furfural. In order to design the ionic liquid three solvent properties were considered : solvent loss, selectivity and solvent power. These solvent properties were calculated for a system consisting of toluene/n heptane and furfural (as the extraction sol vent) using the infinite dilution activity coefficients of different binary mixtures of the compounds as shown in e qns. (5 8 ) to (5 10 ) Since the designed IL (optimal candidate) needs to be a better solvent than furfural it must have a lower solvent loss and higher solvent power (towards toluene) and selectivity ( higher tendency towards dissolution of toluene rather than n heptane). In this problem, the final/equilibrium compositions (mole fractions) of the three components in the raffinate ( ) a nd extract ( ) phases as well as the molar flow rate of the raffinate (R) and extract phases (E) were unknown. In the problem, the feed and solvent flow rates and the compositions of the three compounds in these two input streams are given in m ole s /hr and mole fraction, respectively. E qns. (5 11 ) to (5 13 ) ensure equilibrium condition s and hence the concentration of components in the two phases ( r affinate and

PAGE 124

1 10 extract) do not change. Eq ns. (5 14 ) to (5 16 ) deals with mass balance of each component (toluene, n heptane and IL) around the system. Eq ns. (5 17 ) and (5 18 ) are necessary to ensure that the summation of the mole fractions of the three compo nents equal to unity in both t phases ( r affinate and extrac t), showing that no normalizations are required. It is worth mentioning that due to systematic errors and uncertainties that arose from the models used to predict the solubility and activity coefficients of components it was impossible to satisfy the equi v alence s of e qns. (5 11 ) to (5 18 ) with 100% accuracy. A ( tolerance was applied to the se constraints to enable the optimization program to converge. 5. 3 Results The CAILD program was run several times to ensure that it converged to a global optima rather than a local optima. A high number for the population size input parameter in the GA ensured convergence of the program to the same result for every run We used the optimization model to identify 8 ionic liquids (ILs) with optim al solvent properties. The design results related to the solubility of toluene and n heptane in optimal ionic liquids along with the molar flow rate s of raffinate and extract phases for each case are tabulated in Table 5 3. The results related to the same extraction process using furfural as a solvent was also calculated and listed on the last row of Table 5 3 for comparison purposes All of the calculations were made at T=330 K which is a typical temperature for an aromatic/aliphatic separation (liquid li quid extraction) process.

PAGE 125

111 Table 5 3 : CAILD results for the optimal ILs and furfural at T=330 K Raffinate Extract Compound No. of moles x (mole fraction) No. of moles x (mole fraction) E (mole/hr) R (mole/hr) Toluene n heptane 1 (ethoxyethyl)pyridinium AlCl 4 117.69357 701.24674 4.59405 0.14291 0.85151 0.00558 150.80643 31.25326 2453.4059 0.05722 0.01186 0.94092 2635.465 823.534 Toluene n heptane 1 (benzyl) 3 methylpyridinium AlCl 4 93.01245 686.16365 5.34647 0.11856 0.87463 0.00681 178.08755 41.63635 2449.6535 0.06672 0.01560 0.92549 2669.377 784.522 Toluene n heptane 1 propyl 3 methylpyridinium AlCl 4 107.35496 690.78479 12.95947 0.13236 0.85166 0.01598 159.14504 42.61521 2439.0405 0.06026 0.01655 0.92360 2640.801 811.099 Toluene n heptane 1 ethyl 3 methylpyridinium AlCl 4 106.12357 686.83145 16.01235 0.13118 0.84902 0.01979 163.37643 42.56855 2440.9876 0.06278 0.01608 0.92219 2646.932 808.967 Toluene n heptane 1 (3 ethoxypropyl)pyridinium AlCl 4 108.87295 689.43320 11.07183 0.13451 0.85181 0.01368 160.02705 43.06680 2436.4282 0.06063 0.01632 0.92416 2639.522 809.377 Toluene n heptane 1 benzyl 3 methyl imidazolium AlCl 4 92.64559 678.73373 8.54488 0.11879 0.87026 0.01096 177.7544 51.7663 2442.955 0.06651 0.01954 0.91412 2672.476 779.924 Toluene n heptane 1 (methybenzyl) 3 propylimidazolium PF 6 99.420 674.886 6.828 0.12728 0.86398 0.00874 174.079 58.3139 2448.172 0.06494 0.02095 0.91330 2680.56 781.13 Toluene n heptane 1 methylbenzyl 3 methoxymethylbenzyl methylsulfate 123.469 683.141 0.0751 0.15306 0.84685 0.000092 145.33 1 49.25 9 2447.925 0.05645 0.01864 0.92636 2642.51 806.68 Toluene n heptane Furfural 55.262 488.500 72.951 0.08961 0.79210 0.11829 214.738 241.499 2377.050 0.07579 0.08478 0.83897 2833.29 616.71 As discussed before, it is critical for the designed ILs to have relatively low viscosities and melting points to enable its use as a solvent for extraction in an industrial setting The initial selection of anions considered (as part of the basis set) in this study was based on the fact that they contribute to wards lower values of melting points making it unlikely that any IL designed will be in solid state The viscosities and melting points of the 8 optimal ILs, listed in Table 5 3, were calculated post design using the correlations developed in chapter 2 and

PAGE 126

112 presented in the previous study 193 These predictions along with solvency power calculated through the CAILD model are listed in Table 5 4. Table 5 4 : Physical properties and solvency power of optimal ILs and furfural Compound Vis_Solvent (cP) Tm_Solvent (K) Yield of Ext_hep (%) Yield of Ext_tol (%) PI toluene n heptane 1 (ethoxyethyl)pyridinium AlCl 4 15.6798 262.9530 4.2813 55.8542 304.727 toluene n heptane 1 (benzyl) 3 methylpyridinium AlCl 4 20.6084 271.5854 5.7036 65.9584 202.754 toluene n heptane 1 propyl 3 methylpyridinium AlCl 4 11.3945 258.1737 5.8377 58.9426 172.960 toluene n heptane 1 ethyl 3 methylpyridinium AlCl 4 11.3151 262.9873 5.8313 60.5098 177.948 toluene n heptane 1 (3 ethoxypropyl)pyridinium AlCl 4 19.7611 275.2263 5.8996 59.2693 170.290 toluene n heptane 1 benzyl 3 methyl imidazolium AlCl 4 21.3088 270.2397 7.0913 65.8350 130.920 toluene n heptane 1 benzyl 3 propylimidazolium PF 6 52.9785 312.9393 7.9882 64.4739 101.038 toluene n heptane 1 methylbenzyl 3 ethoxymethylbenzyl imidazolium methylsulfate 105.0221 304.6925 6.7478 53.8262 118.215 toluene n heptane furfural 1.0630 236.0000 33.0822 79.5326 7.2670

PAGE 127

113 The chemical structure of the 8 optimal ionic liquids which were designed through the CAILD model are listed in Table 5 5. Table 5 5 : A schematic of the structure of optimal ILs Ionic Liquid (IL) Structure 1 (ethoxyethyl)pyridinium Tetrachloroaluminate 1 (benzyl) 3 methylpyridinium Tetrachloroaluminate 1 propyl 3 methylpyridinium Tetrachloroaluminate 1 ethyl 3 methylpyridinium Tetrachloroaluminate 1 (3 ethoxypropyl)pyridinium Tetrachloroaluminate 1 benzyl 3 methyl imidazolium Tetrachloroaluminate

PAGE 128

114 Ionic Liquid (IL) Structure 1 benzyl 3 propyl imidazolium Hexafluorophosphate 1 methylbenzyl 3 methoxymethylbenzyl imidazolium Methyl sulfate

PAGE 129

115 Application 3: Design of Ionic Liquids for CO 2 Capture Chapter 6: 6.1 Introduction The rapidly growing challenge of global warming (GW), which is believed to be caused by human activates resulting in the release of greenhouse gases (e.g. CO 2 CH 4 ) into the atmosphere, needs to be dealt with in a timely manner. During the past decade, researchers working in this area are divided into two distinct groups; one group has been focusing on the substitution of fossil fuels with their renewable counterparts (e.g. solar energy, wind energy, and the other group has been working on carbon (mainly CO 2 ) ca pture storage, and utilization (CCS U ). Absorption, adsorption, and membrane separation are among the chemical processes being considered for CO 2 capture process. 194 Among the aforementioned processes, absorption (chemical or physical) with a re generabl e solvent, appears to be one of the most promising methodologies for CO 2 capture. 195 Monoethanolamine (MEA) a polar solvent has shown the ability to be a good solvent for CO 2 capture 196 Even though amine based solvents normally possess relatively high s olvency power towards CO 2, their relatively high vapor pressure make s the solvent regeneration step inefficient and environmentally impactful. 197 In order to deal with this issue, use of alternative solvents such as ionic liquids, has attracted a considera ble amount of attention in the past few years. 198 Ionic liquids, have the potential to be tailored to possess the desired properties of organic solvents (e.g. having high solvency power towards CO 2 ) without possessing their undesired properties. 199

PAGE 130

116 Studies indicate that based on the type of anion selected, ionic liquids (ILs) can be utilized to promote physical or chemical absorption of CO 2 The physical absorption, which is merely based on the dissolution of CO 2 in the liquid phase (IL rich phase), is only feasible when a great abundance (i.e. a high partial pressure) of CO 2 is available in the gaseous mixture. The pre combustion capture is the process which deals with the removal of CO 2 from the fuel streams (mainly containing CO 2 and H 2 ) before they under go the chemical combustion. The higher concentration of CO 2 in the feed stream (compared to flue gas), make the physical absorption using an ionic liquid feasible In chapter 3, we showed that a large inventory of anions, cations, and functional group s either exist or can be synthetized. 200 202 A large number of ILs can potentially be formed through the combination of these distinct building blocks (estimated to be as many as 10 14 feasible ILs). 203,204 6.2 Forward problem A ctivity coefficients of comp ounds in a multicomponent mixture can be used to predict the equilibrium condition of a system. Similar to the approach used in the previous chapter on the design of an IL for aromatic/aliphatic separation, UNIFAC model can be utilized to predict the activity coefficients of the components based on the concept of group contribution. Due to limited availability of CO2 solubility data in ionic liquids we followed a predictive approach similar to the one explained in chapter 5.

PAGE 131

117 COSMO RS, a promising tool for solvent screening for different applications, was used to predict the activity coefficients and solubility of CO2 in selected ILs. H is a partitioning coefficient between vapor (gas) and liquid phases which measures the tendency of a compound to remain in the gas phase versus being dissolved in the liquid given temperature) in the gas phase divided by its concentration in the liquid phase w hen the constant implies that the compound of interest has higher volatility thereby less likely to stay in the liquid phase (i.e. lower concentration in the liquid ph good index to evaluate whether a chemical (a molecular compound or an ionic liquid) is a good solvent for CO 2 capture. In order for an ionic liquid to be a good solvent for a carbon capture CO 2 system should be as low as possible. Ionic liquids generally have very low vapor pressures and hence t 2 in ILs can be defined as following ( 6 1) is the activity coefficient of CO2 in the liquid (IL rich) phase and is inversely proportional to its solubility in the IL and is the vapor pressure of CO 2 at the given temperature. In this study, the experimental values of the vapor pressur e of CO 2 were used. The activity coefficients of CO 2 in different ionic liquids were calculated using COSMO RS model 213 implemented in COSMOtherm software. 214 In order to verify the ability of the COSMO RS model to predict the activity coefficients (hence the solubility) of CO 2 in

PAGE 132

118 different ILs, a comparison between experimental and COSMO constants for different IL CO 2 mixtures is presented All calculations in COSMOtherm were performed using BP level of density functional theory (DFT) and TZVP basis set (the recommended settings for predicting the thermophysical properties of chemical compounds in chemical engineering applications). 215 In order to find f CO 2 in different ILs, we tested all different BP TZVP basis sets available in COSMOtherm. We identified that for IL CO 2 systems, BP_TZVP_C21_0108 parameterization set predicts the atic error between the predicted and experimental values was observed, therefore a linear relationship shown in eqn. (6 2 ) between experimental and COSMO predicted values of H (based on BP_TZVP_C21_0108 basis set) was developed to reduce the average errors of COSMO based predictions. H mod = 0.7403H pred + 7.9697 ( 6 2) Table 6 1 lists the experimental and COSMO predicted values (both before and after modification ) of Henr CO 2 systems which were not part of the original data set used to develop the linear relationship

PAGE 133

119 Table 6 1 : A comparison between experimental and COSMO based values of constant H exp [bar] H Cosmo pred [bar] H Cosmo modified [bar] 57.8 76.09 64.30 67.9 93.12 76.91 77.0 112.60 91.33 50.58 61.17 53.25 57.49 68.98 59.04 64.17 76.91 64.90 69.35 85.45 71.23 80.02 104.22 85.13 103.38 136.98 109.37 139.94 205.09 159.80 47.46 60.83 53.00 52.77 67.73 58.11 57.28 75.30 63.71 62.72 83.54 69.82 71.26 101.29 82.95 80.46 121.24 97.73 94.18 155.87 123.36 97.94 168.26 132.54 Average error 36.6% 12.7% E q n. (6 2 ) software. The modified values of H along with the experimental values of vapor pressure of CO 2 were used to reverse calculate its activity coefficients in different ILs.

PAGE 134

120 Based on t he approach used in the previous chapter, the COSMO predicted activity coefficients (after correcting for systematic errors ) were used to develop group parameters ( R and Q ) and binary interaction parameters (a ij and a ji ) of the UNIFAC model for systems con sisting of CO 2 and ILs. A list of cations, anions, and side chain groups used in this study are listed in Appendix D The UNIFAC parameters developed in this study can be used in any computer aided ionic liquid design (CAILD) framework to explore a wide variety of ionic liquids (estimated to be as many as 21,000) with an aim to find the most optimal ionic liquid for a CO 2 capture process. The accuracy of the CAILD approach was tested using available experimental data on the solubility of CO 2 in several different ionic liquids and the results along with the relative errors are tabulated in Table 6 2. Table 6 2 : Experimental and UNIFAC predicted values of CO 2 solubility in different ILs Ionic Liq uid (IL) Solute Ref 1 butyl 3 methylimidazolium Tf 2 N CO 2 283.1 100 0.0373 0.008 0.0498 216 1 butyl 3 methylimidazolium PF 6 CO 2 313.3 103 0.0162 0.0016 0.0135 217 1 (3 hydroxypropyl)pyridinium Tf 2 N CO 2 303.58 64.3 0.01175 0.00071 0.0102 218 1 (3 hydroxypropyl)pyridinium Tf 2 N CO 2 323.47 69.9 0.00578 0.00035 0.00635 218 1 (3 hydroxypropyl)pyridinium Tf 2 N CO 2 333.46 66.5 0.0038 0.00023 0.00451 218 1 ethyl 3 methylimidazolium Tf 2 N CO 2 283.43 100 0.03996 2e 05 0.0412 219 1 ethyl 3 methylimidazolium Tf 2 N CO 2 293.39 100 0.03203 1e 05 0.03152 219 1 ethyl 3 methylimidazolium Tf 2 N CO 2 303.39 100 0.02626 1e 05 0.02446 219

PAGE 135

121 Ionic Liq uid (IL) Solute Ref 1 butyl 3 methylimidazolium BF 4 CO 2 283.1 99.9 0.0197 0.0028 0.032715 220 1 butyl 3 methylimidazolium PF 6 CO 2 283.15 49.9 0.0139 0.006 0.017019 221 1 butyl 3 methylimidazolium PF 6 CO 2 283.15 99.9 0.0267 0.006 0.031079 221 1 ethyl 2,3 dimethylimidazolium Tf 2 N CO 2 283.15 49.9 0.016 0.004 0.019515 221 1 ethyl 2,3 dimethylimidazolium Tf 2 N CO 2 283.15 99.8 0.0323 0.004 0.039781 221 6.3 Reverse p roblem In this section we propose a computer aided IL design model which integrates a process model within the solvent design framework and can design optimal ionic liquid for CO 2 capture process is presented 6.3.1 Case s tudy A feed stream consisting of pure CO 2 at 1 bar pressure enters a single stage absorption column where it comes into contact with pure IL. The absorption process occurs at 288 K and 1 bar. The IL enriched phase which now has absorbed CO 2 will enter s the desorption column in which the CO 2 will be separated from the liquid phase by increas ing the temperature. The desorption process occurs at 318 K and 1 bar. The goal of the CAILD framework is to design an IL with high CO 2 solubility (i.e. high solvency power towards CO 2 ) and the lowest performance index ( PI ) possible. This means that the designed ionic liquid will strongly absorb CO 2 at lower temperatures and releases it effectively at higher

PAGE 136

122 temperatures. A schematic of the coupled absorption desorption process of interest is depicted in Figure 6 1. Figure 6 1 : A schematic of single stages CO 2 absorption desorption processes Detailed Formulation of the Problem In this optimization framework the objective is to maximize the solubility of CO 2 in the IL phase during the absorption step while at the same time minimize the solubility of CO 2 in the IL at a higher temperature during the desorption step The objective function was defined as the performance index ( PI ) as shown below : Objective function = min ( ) ( 6 3) Feas ibility constraints (6 4)

PAGE 137

123 (6 5) (6 6) (6 7) (6 8) (6 9) Solution property constraints (6 10) (6 11) (6 12) (6 13) As described in previous chapter, the optimal IL designed using the CAILD framework must satisfy specific rules to ensure the chemical feasibi lity of the designed compound. These rules, also named as structural constraints, include feasibility rules (e.g. octet rule), bonding rule, and complexity rules (which deals with the size of cation as well as the side chains attached to the cation base). Eqn (4 9 ) deals with the size of each side chain attached to the cation base that is an upper bound of 15 for the number of CH 2 groups in this study. Eqn (4

PAGE 138

124 10 ) ensures that an ionic liquid with the solvent power better than that of methanol (a conventio nal solvent used for physical absorption of CO 2 ) is selected as the optimal solvent. Eqn (4 11 ) guaranties that the system has reached equilibrium condition where the mole fraction of CO 2 in the liquid phase ( reaches its maximum value at that temperature. In other words, when a system consisting of CO 2 and an IL has reached the equilibrium condition, the composition of the two phases (IL rich phase and CO 2 rich phase) will not change with time; meaning that the IL cannot dissolve more CO 2 at that temperature. Since ILs mostly have negligible vapor pressures, we do not expect to detect any trace of them in the gas phase (neither in absorption nor in the desorption stages). Therefore it is safe to say that in both the stages only liqu id phases consist of more than one compound (i.e. CO 2 and IL). Eqn s. (4 12 ) and (4 13 ) ensure that the summation of the mole fraction of the two components in the liquid phase is equal to 1 It is also worth mentioning that due to the systematic and model errors in predicting the solubility and activity coefficients of the compound in different phases, it is impossible to satisfy the equi valence s of E qns. (6 11 ) through (6 13 ) with 100% accuracy. Therefore a ( tolerance was applied to the equations to enable the optimization program converge. 6.4 Results The CAILD optimization program was run several times to ensure that it converged to a global optima. The optimization model (CAILD) was used to identify the top 5 optimal ionic liquids (ILs) from different categories (different cation head groups) with best solvent

PAGE 139

125 properties. The structure of the optimal solvents along with their structures are listed in Table 6 3 Table 6 3 : Name, symbol and structure of the optimal ILs Ionic Liquid (IL) Symbol Chemical Structure 1 (12 phenyloxydodecyl) 3 (5 phenylpentyl)imidazolium Tf 2 N IL 1 1 (3 methoxypropyl) 4 ethylpyridinium ethyl sulfate IL 2 1 (15 benzyloxypentadecyl) 1 hexylpyrrolidinium Tf 2 N IL 3 Decyl (4 pentadecylbenzyl) methylphosphonium triflate IL 4

PAGE 140

126 Ionic Liquid (IL) Symbol Chemical Structure 4 (8 phenoxyoctyl) 4 (phenylnonyl)morpholinium Tf 2 N IL 5 As discussed before it is critical for the optimal ILs to have reasonably low viscosity and melting point to enable them to be used as the process solvent in industry The initial selection of anions used in this study was based on this fact that they contribute to the lowe r values of melting points of the ILs they make. The viscosity and melting point values of the 5 optimal ILs, listed in Table 6 4, were calculated from correlations developed in chapter 2 and presented in our previous study. 222 These values, along with the predicted solubility of CO 2 in these ILs and the solvency power of the optimal ILs, are listed in Table 6 4 Table 6 4 : Pure (physical) and mixture properties of the optimal ILs IL SP H abs @ 288 K [bar] H des @ 318 K [bar] PI T m [K] Viscosity @ 288 K [cP] IL1 4.532 0.0706 9.7498 20.649 4.6034 201.95 587.91 IL2 1.438 0.0309 10.9397 22.934 5.2183 312.04 66.10 IL3 4.717 0.0798 9.1024 19.888 4.1660 228.08 724.92 IL4 4.848 0.0843 7.5657 16.518 3.4653 230.94 845.34 IL5 3.963 0.0677 11.2954 23.779 5.3654 224.24 1521.50

PAGE 141

127 IL SP H abs @ 288 K [bar] H des @ 318 K [bar] PI T m [K] Viscosity @ 288 K [cP] Bmim Tf 2 N 1.755 0.0340 48.7733 90.8902 26.1726 267.1 81.60 Bpy BF 4 0.911 0.0211 74.4796 136.7710 40.5583 272.1 313.1 BmPyr triflate 1.5522 0.0301 54.0972 103.0201 28.4071 277.56 299 Methanol 0.429 0.00853 225.0339 419.1725 155.168 175.55 0.6220 Bmim: 1 butyl 3 methylimidazolium Bpy: 1 butylpyridinium BmPyr: 1 butyl 1 methylpyrrolidinium triflate: Trifluoromethanesulfonate

PAGE 142

128 Life Cycle Environmental Implications of Ionic Liquids Chapter 7: 7.1 Life Cycle Perspectives on Aquatic Ecotoxicity of Common Ionic Liquids As it was mentioned earlier, field of ILs is developing rapidly, thus it is important to consider their environmental, ecological, and hu man health impacts at the design stage for their long term acceptance. Despite the large amount of attention, ILs are still a fairly new class of materials and several of their environmental characteristics have only been studied recently. The non volatile nature of ILs can greatly limit atmospheric pollution and for this reason, they are often promoted as green chemicals with the potential to replace volatile organic solvents 223 However, in recent years, many studies have reported that several ILs possess relatively high level of toxicity towards freshwater organisms. While both points are valid, a realistic picture can only emerge through analysis of the life cycle ecological impacts that considers the upstream ecological impacts associated with producing them as well as the downstream ecological impacts due to their use. The current study addresses this research gap by focusing on understanding the production side and end of life fresh water ecotoxicity impacts related to ILs. To date there have been ver y few Life Cycle Assessment (LCA) studies performed on processes that involve ILs. 224,225,260, 261 These studies indicate that, from a life cycle perspective, IL based processes do not necessarily improve the environmental performance in comparison to molec ular solvents based processes. One common limitation in all of these studies is that they did not consider any possible impacts associated with the direct release of ILs into the environment. Therefore, it is clear that the environmental fate as well as to xicity

PAGE 143

129 information related to ILs has not been incorporated in any LCA study to date. This is because toxicity based characterization factors for ILs are not yet available in impact assessment tools such as TRACI 2.1. One of the main objective of this pape r is to address this research gap by developing, freshwater ecotoxicity characterization factors for the following five common ILs: 1 butyl 3 methylimidazolium bromide ([Bmim] + [Br] ), 1 butyl 3 methylimidazolium chloride ([Bmim] + [Cl] ), 1 butyl 3 methylim idazolium tetrafluoroborate ([Bmim] + [BF 4 ] ), 1 butyl 3 methylimidazolium hexafluorophosphate ([Bmim] + [PF 6 ] ), and 1 butylpyridinium chloride ([BPy] + [Cl] ). In the recent past, several studies have provided new insights into IL chemistry, environmental fate and toxicity. Many ILs exhibit significant solubility in water 227 and even water immiscible ILs show limited water solubility and stability. 228 This, combined with the fact that they are non volatile, makes wastewater discharge as the most likely rout e through which ILs will eventually be released into the environment Once discharged, ILs will interact with aquatic ecosystems through a variety of mechanisms and can cause damage to aquatic species. 229 To evaluate the ecotoxicity of ILs, a broad range of testing models (bacteria, fungi, algae, plants, and animals) have been used. 230 Standard ecotoxicological tests show that many ILs have high toxicity towards freshwater organisms ( e.g. algae and D. magna). 231 Wells et al. 231 examined four different classe s of ILs, which all had relatively high ecotoxicity impacts (EC50 < 100 mg L 1 ). Certain ILs, such as those with shorter side chain 1 alkyl 3 methyl imidazolium and pyridinium, show only moderate ecotoxicity to bacteria, algae, and invertebrates, but when the side chains are longer than C 8 their ecotoxicity profiles become significantly worse. This is also true for phosphonium and ammonium based

PAGE 144

130 ILs, which have higher molecular weights. The environmental fate of ILs depend on several biotic and abiotic f actors. Only in the past few years have the major abiotic mechanisms of ILs ( e.g. their sorption in different type of soils) been understood. 232, 233 The extent of the final toxicity impact of ILs will depend heavily on their physicochemical properties, int eractions with surrounding environmental media, and chemical and biological transformations. This information is essential to understand how these compounds are transformed and how long they will persist in the environment. On the other end of the spectru m, there is still very little understanding of the environmental and ecological impacts associated with the production of ILs. This is because ILs are emerging materials and are not yet produced in commercial scales. Consequently, no primary data is availa ble on material/energy consumption. Further, very little data is available about the precursors needed to produce these ILs. This study integrates the existing 224, 225 and newly developed life cycle inventories for the production of the above mentioned five ILs with an aim to understand their production side upstream freshwater ecotoxicity impacts. Through an in depth analysis, we compare the relative contributions of the different IL life cycle stages (production phase and use phase loss) towards ecotoxici ty impacts and discuss their significance. We would like to point out that similar studies on other materials, such as carbon nanotubes (Eckelman et al. 248 ) have revealed important information on comparing production and use phase impacts.

PAGE 145

131 7. 2 Methods 7. 2.1 Goal, Scope, System Boundary The main goal of this study was to understand and compare the relative freshwater ecotoxicity impacts of ILs related to their production phase and direct release to the environment during use phase. The scope of this study include s: 1) building cradle to gate life cycle inventory for production of five common ILs; 2) developing characterization factors for their freshwater ecotoxicity impacts; and 3) comparing the potential ecotoxicity impacts of the production and use phas e release. A functional unit of 1 kg of the five ILs was considered.. The system boundary (shown in Figure E 1 in Appendix E ) includes all upstream steps necessary for the production of ILs, treatment, and recovery as well as the direct release of IL to th e environment during use. 7. 2.2 Life Cycle Inventory of Ionic Liquid Production and Data Sources Inventory of upstream steps include raw material inputs and emissions related to the production of reactants, precursors, reagents and ancillary materials as well as the extraction, conversion and delivery of energy inputs. The emissions from the construction and maintenance of chemical plants are assumed to be relatively low 225 and hence neglected. Data related to electricity, thermal energy (steam), transport ation systems, and chemical production were derived either from Ecoinvent, 258 or USLCI database integrated in SimaPro software. 259 Data availability permitting, for consistency purposes we assumed that all life

PAGE 146

132 cycle steps involved in the production of IL s to be within the United States. Whenever non U.S. data had to be used, every effort was made to adjust the data to corresponding US energy mix. The inventory of all materials (e.g. precursors, reagents, reactant etc.) and energy used for the production o f the ILs along with the data sources are listed in Table E 2 of Appendix E Please note that data for chemicals that were not available in standard LCI databases (ILs and some of their precursors) were assembled using mass and energy balances derived from chemical process simulations (CPS) 225,234 supplemented with theoretical calculations based on sound judgment and good engineering practice. A detailed description of the procedure adopted to generate the mass and energy balance is presented in Appendix E 7. 2.3 Fresh Water Ecotoxicity Impacts of Ionic Liquid Production The life cycle impact assessment methodology, Tool for the Reduction and Assessment of Chemical and other environmental Impacts (TRACI 2.1.) developed by U.S. Environmental Protection Agen cy utilizes characterization factors derived from the USEtox model 238 Using these characterization factors, emissions of organic and inorganic materials released during the upstream life cycle steps of IL production was translated into total freshwater ecotoxicity impacts.

PAGE 147

133 7. 2.4 Development of Ecotoxicity Characterization Factors for Ionic Liquids Characterization factors are used to quantify the extent to which a given pollutant contributes to environmental impacts. In this work we utilized the USEtox m odel 238 which is a state of the art modeling framework based on scientific consensus for characterizing ecotoxicity impacts of chemicals, to develop freshwater ecotoxicity characterization factors for the five ILs. In this approach, the aquatic ecotoxicit y characterization factors are estimated as a product of fate factor (FF), exposure factor (XF) and effect factor (EF) as shown in e qn (7 1 ) (7 1) EF relates to the inherent toxicity of the substance of interest, FF relates to the fate and transport of the substance, and XF relates to potential routes of exposure and intake of the substance. In the following section, we describe the procedure for the development of each of these factors for ILs. Effect Factor: EF ref lects the relationship between the concentration of an IL and potentially affected fraction (PAF) of aquatic organisms. It is defined as the slope of the concentration response relationship up to the point when the PAF reaches 50% ( e qn (7 2) ). (7 2) HC50 corresponds to the geometric mean of species specific EC50 data and EC50 represents the effective concentration of the IL of interest at which 50% of the population of a particular species experiences a response. Table 7 1 summarizes the aquatic toxicity data (EC50) of the

PAGE 148

134 five ILs collected from the literature. It can be observed that, there is a wide distribution of EC50 values associated with ecotoxicty of ILs towards different species. In the next step, as suggested by USEtox, the geometric mean of EC50 values for each IL was calculated. Based on the guidelines provided for the use of USEtox model acute toxicity data were converted to chronic data using acute to chronic ratio of two. Similarly, when no observed effect c oncentration (NOEC) was reported we extrapolated NOEC to EC50 by a factor of nine. Next, the geometric mean of individual EC50 data for each IL, HC50, was used to estimate the effect factor. Fate Factor : Mul timedia fate and transport models are used to derive the environmental fate factors. In these models the environment is represented as a number of homogeneous compartments (e.g. air, water, and soil). The intermedia transfer of chemical substances between different compartments is modeled as a set of mass balance equations.

PAGE 149

135 Table 7 1 : Toxicity values of selected Ionic Liquids Type Species Test type Reported toxicity values (mg/L) Geometric mean of EC50s [Chronic] (mg/L) Ref [Bmim] [Br] [Bmim] [Cl] [Bmim] [BF4] [Bmim] [PF6] [BPy] [Cl] [Bmim] [Br] [Bmim] [Cl] [Bmim] [BF4] [Bmim] [PF6] [BPy] [Cl] Bacteria Photobacterium phosphoreum growth inhibition (Acute) 408.020 --------204 --------239 Bacteria Aliivibrio fischeri Luminescence (Acute) 256 2250 429 897 799 802 334 929 257 440 491.6 216.5 399.5 278.5 168.1 240 241 228 Algae Pseudokirchneriella subcapitata growth of algal biomass or photosynthesis O2 evolution (Acute) 229 5260 504 568 375 63.9 393.1 252 284 187.5 32 242 243 Daphnia (Crustacean) Daphnia magna first brood number of neonates & immobilization (Acute) NOEC = >3.2 EC50 = 8.03 NOEC= >3.2 EC50= 6.3 10.7 NOEC= >3.2 EC50= 19.9 20 10.7 9.5 5.4 16.9 10 229 Animalia Physella acuta egestion rate & death (Acute) NOEC = 4 LC50 = 229 -----LC50 = 123 --64.2 ----61.5 --244 fish Danio rerio death & frond number (Acute) LC50>=100 LC50>=100 EC50=59.5 LC50>=100 LC50>=100 LC50>=100 50 38.6 50 50 50 257 HC50 105.2 66.9 74.2 77.0 40.5

PAGE 150

136 The fate factor, which represents the persistence of a substance (residence time in days) in a particular environmental compartment, depends on physicochemical properties, partition coefficients, and biodegradation rates of the sub stance of interest. Most of the IL fate and transport parameters needed as input to the model were gathered from literature, and are listed in Table 7 2. A small value of 1 x 10 20 for biodegradation rate was assigned to all ILs that were categorized as non biodegradable but had no quantitative values 255 Exposure Factor : The environmental exposure factor for freshwater ecotoxicity is defined as the fraction of IL dissolved in freshwater and is determined using e qn (7 3 ). (7 3 ) Where, represent the partition coefficient between water and suspended solids and represents the partition coefficient between water and dissolved organic carbon respectively. BAF represents the bio concentration factor in fish. SS (15 mg L 1 ), DOC (5 mg L 1 ), and BIOmass (1 mg L 1 ) are default values for suspended matter, dissolved organic carbon and biomass concentrations, respectively. 248 The exposure factor (XF) for each IL was also calculated bas ed on parameters listed in Table 7 2. We would like to note that there are certain limitations in using the proposed approach to the case of ILs. The USEtox model is suited for relatively small organic and inorganic materials while ILs typically have high molecular weights. However, other studies have reported results using the USEtox model for complex materials such as carbon nanotubes. 248 In addition, even though we assume that ILs can be treated as neutral entities, the ionic nature of these compounds may influence the results. Despite these limitations, we believe that, as per current state of knowledge, the

PAGE 151

137 approach of using the USEtox model represents the best possible way towards ecotoxicity impact assessment of ILs.

PAGE 152

138 Table 7 2 : Environmental Properties of the studied Ionic Liquids Parameter Unit Value [Bmim] + [Br] [Bmim] + [Cl] [Bmim] + [BF 4 ] [Bmim] + [PF 6 ] [BPy] + [Cl] Molecular weight g mol 1 219.12 174.67 226.03 284.18 171.67 Octanol water partition coeff. K OW 0.0033 0.0029 0.0030 0.0218 0.0020 Organic carbon water partition coeff. K OC L k g 1 398.11 398.11 398.11 398.11 141.61 (K H ) a Pa kg mol 1 1 10 20 1 10 20 1 10 20 1 10 20 1 10 20 Water solubility (25 ) mg L 1 1 10 6 5.5 10 5 1 10 6 1.9 10 4 4.32 10 4 Dissolved carbon water partition coeff. (K DOC ) b L k g 1 18.27 18.27 18.27 18.27 6.50 Suspended solids partition coeff. (Kp SS ) b L k g 1 18.27 18.27 18.27 18.27 6.50 Sediment water partition coeff. (Kp sd ) b L k g 1 18.27 18.27 18.27 18.27 6.50 Soil water partition coeff. (Kp sl ) b L k g 1 18.27 18.27 18.27 18.27 6.50 Biodegradation rate in air c s 1 1 10 20 1 10 20 1 10 20 1 10 20 1 10 20 Biodegradation rate in water c s 1 1 10 20 3.47 10 8 1.36 10 8 1 10 20 1 10 20 Biodegradation rate in sediment c s 1 1 10 20 1 10 20 1 10 20 1 10 20 1 10 20 Biodegradation rate in soil c s 1 1 10 20 1 10 20 1 10 20 1 10 20 1 10 20 Bioaccumulation factor in fish (BAF fish ) d L k g 1 1.58 10 4 8.55 10 4 8.94 10 4 5.73 10 3 6.04 10 4 a) The value 1 10 20 for K H was assigned to the ionic liquids, as all ILs have very small (negligible) vapor pressures. b) The values for soil water partition coefficient of all ionic liquids were calculated using the formula of K d = K OC *f OC in which f oc represents the fraction of organic carbon in the soil 0.0459 and K OC is the organic carbon water partition coefficient. This value of K d was assigned to all four parameters K DOC Kp SS Kp sd and Kp sl c) The small value of 1 10 20 was used for all non biodegradable ionic liquids which had a biodegradability rate < %1 within 28 days. d) Bioaccumulation factor in fish (BAF fish ) of the ionic liquids were calculated using this formula: l og BAF= 0.68 + 0.94 log K OW

PAGE 153

139 7. 2.5 Fresh Water Ecotoxicity Impact of Direct Release of Ionic Liquids Large scale us e of ILs will inevitably lead to their release into the aquatic environment through wastewater disposal or accidental leakage. 249 The ecotoxicity impact of this release needs to be considered in any type of IL assessment. At a large commercial scale it wou ld be reasonable to expect that industrial plants using ILs will have an appropriate treatment unit to remove and recover ILs from wastewater streams. This is a likely scenario since ILs are expensive and they can also be recovered easily thereby providing significant economic incentive to recover and reuse them. Studies show that in a typical chemical plant, loss of volatile organic solvents is usually estimated to be close to 10% with the majority of the loss due to air emissions 250 For example, monoetha nolamine, a solvent used for CO 2 capture requires a makeup of 10% solvent with <1% emissions to water. 250 For the case of ILs, their non volatile nature would eliminate the possibility of any air emissions while water emissions would represent the most sig nificant source of exposure. As typical water emission of solvents is < 1% we conservatively assume a maximum of 2% loss of ILs (during the use phase) to the wastewater stream which can be further significantly reduced by management of industrial wastewate r. The assumed 2% loss of ILs to wastewater is in fact a very conservative estimate as nearly complete recycle and reuse of these compounds is feasible and since ILs are lot more expensive than organic solvents an economically viable process/technology th at uses ILs would involve recovery and recycle of as much IL as possible. Further, environmental regulations would limit levels of IL discharge to the environment through wastewater effluents. Several treatment methods such as oxidative, 251

PAGE 154

140 thermal, 252 and photocatalytic degradation 253 have been proposed for IL removal. Other approaches such as adsorption in activated carbon 254 and use of salts 249 have also been suggested. For this study, we selected the treatment option based on salting out using alumin um based salts as proposed by Neves et al 249 due to the fact that this approach makes use of inorganic salts commonly used in current waste treatment plants and also has high recovery. Neves et al 249 reported a recovery efficiency of 96% to 100% for a wi de variety of ILs. We assume a range of 96 to 98 percent recovery of the ILS lost to the waste water stream which results in a final release fraction of 0.04% to 0.08% of ILs into freshwater streams. The ecotoxicty impact associated with the direct release of the IL to freshwater was estimated as a product of the IL characterization factor the mass of IL used (i.e. 1 kg) and the final release fraction to the freshwater (unrecovered IL). 7. 2.6 Uncertainty This section deals with the treatment of uncertainty related to the ecotoxicity estimates of the two life cycle stages, production phase and release during use. Monte Carlo simulation was employed and critical uncertain parameters were varied to generate multiple samples. Upon reviewing the LCI related to t he production of ILs we identified that in comparison to other emissions, metals had very high ecotoxicty characterization factors. In addition, the USEtox model developers report caution while using the characterization factors of inorganic materials (met als). Therefore, the characterization factors (CFs) for all metals were varied by order of magnitude using a uniform distribution to generate Monte Carlo inventory samples with an aim to quantify the model uncertainty related to the production of ILs at

PAGE 155

141 95% confidence To capture the uncertainties related to ecotoxicity impacts associated with release during use phase a similar Monte Carlo simulation was performed within the USEtox model. A uniform distribution with one order of magnitude variation in ea ch direction was assigned to each model input parameter tabulated in Table 7 2 as well as to HC50 data shown in Table 7 1. A range of 0.4% 2% (0.4% representing 80% removal through water treatment and 2% representing no water treatment) of the produced ILs was assumed to be released to the freshwater compartment. The Monte Carlo simulation randomly generated 1000 input parameter sets resulting in 1000 model outputs for the freshwater ecotoxicty characterization factors. The sample statistics were used to ge nerate the mean ecotoxicty impacts as well as uncertainty levels at 95% confidence. 7. 3 Results and Discussion The effect factor (EF), fate factor (FF), and exposure factor (XF) of the five ILs studied are shown in Table 7 3. Table 7 3 : USEtox based effect factors, fate factors, exposure factors, and characterization factors for different ionic liquids IL EF [PAF m 3 /kg] FF [days] XF CF (CTUe/kg) [Bmim] + [Br] 4.75 131.5 0.9996 624.375 [Bmim] + [Cl] 7.47 100.1 0.9996 747.448 [Bmim] + [BF 4 ] 6.73 122.4 0.9996 823.422 [Bmim] + [PF 6 ] 6.49 142.9 0.9996 927.05 [BPy] + [Cl] 12.34 143.3 0.9998 1767.97

PAGE 156

142 The fact that XFs are very close to 1 or 100% shows that the aquatic organisms will be exposed to nearly all of the ILs once released to the freshwater bodies. The overall characterization factors for freshwater ecotoxicity of the five ILs are also shown i n Table 7 3. These characterization factors along with release fraction was used to estimate ecotoxicity impacts of release during use phase. For comparison purposes, we present these IL characterization factors along with CFs of few other conventional che mical compounds (gathered from TRACI 2.1 database) in Table E 3 of Appendix E Cradle to gate LCI of the production of the five ILs, each consisting of 102 air emissions (97 for pyridinium), 121 water emissions, and 35 soil emissions was translated into freshwater ecotoxicity impacts using ecotoxicity characterization factors from TR ACI 2.1 database. The error bars based on 95% confidence interval of the Monte Carlo results are reported in Figure 7 1 along with the mean value of ecotoxicty impacts for both production and release of ILs. Figure 7 1 : Ecotoxicity impacts related to production and use phase release of ILs

PAGE 157

143 Results from Figure 7 1 reveal that for all studied ILs the mean freshwater ecotoxicity impacts attributable to the production phase of the ILs is nearly one order of magnitude higher than mean impacts due to their estimated release to the freshwater during use phase. This find ing suggests that future research should not only focus on designing and identifying ILs with minimal toxicity but, more importantly, focus on improving the environmental profile of upstream production steps. Next, for the IL production phase, we examined relative contribution of ecotoxicity impacts due to upstream cradle to gate energy use versus cradle to gate direct release of materials and chemicals (i.e. chemical release during production of reagents, precursors, reactants etc.). We find that majority of ecotoxicity impacts of production of ILs arise from chemical releases associated with the upstream production steps rather than energy use. Table 7 4 breaks down the contributions) of these two categories. For the five ILs considered in this study the average energy related ecotoxicity impacts was about 17% while the average chemical or material related ecotoxicity impacts was about 83%. Table 7 4 : Breakdown of energy and material related ecotox icity impacts of IL IL Energy Related Ecotoxicity Impacts (%) Material Related Ecotoxicty Impacts (%) [Bmim] + [Br] 7.5 92.5 [Bmim] + [Cl] 20 80 [Bmim] + [BF 4 ] 27 73 [Bmim] + [PF 6 ] 21 79 [BPy] + [Cl] 8 92

PAGE 158

144 Since ILs are a combination of cations and anions, we can separate their production related ecotoxicity impacts into cation related impacts and anion related impacts. Table 7 5 lists these numbers for the IL [Bmim] + [Br] About 12.7% of the total ecotoxic ty impacts can be attributed to cation production steps and the remaining can be attributed to anion production steps. In addition, we find that almost 86% (out of the 87.3%) of the anion production ecotoxicty impacts are due to the direct emission of the three chemicals Bromobenzene, Benzene, 1,2 dibromobenzene. As for cation production steps, metal emissions are fully responsible for the ecotoxicity impacts (12.7%). Further analysis shows that for production of [Bmim] + [Br] approximately 5% of the total c radle to gate life cycle ecotoxicity impacts are due to upstream emissions related to natural gas/electricity production for energy purposes. The remaining 95% are associated with the direct release of materials during upstream processes (including the p rocess of natural gas extraction/processing for non energy purposes). The above results show that we cannot decrease the production side ecotoxicity impacts of [Bmim] + [Br] by simply switching to alternate energy sources (such as renewables). Therefore, it is clear that researchers need to identify less impactful synthesis routes or control emissions associated with certain key precursors to further minimize freshwater ecotoxicity impacts of this IL. Note that the three chemical (Bromobenzene, Benzene, 1,2 dibromobenzene) releases that contribute the most (86%) towards [Bmim] + [Br] production ecotoxicity impacts are all by products in the synthesis steps related to 1 Bromobutane. Therefore, alternate methods or routes to produce 1 Bromobutane or tight cont rol of chemical releases during this step, has the potential to significantly decrease

PAGE 159

145 the total ecotoxicty impacts. Note that for [Bmim] + [Cl] ecotoxicity impacts due to cation and anion production are roughly equal while for [BPy] + [Cl] the cation relate d impacts are significantly higher (76%) than that of anion. As for the IL [BPy] + [Cl] the major portion (76.1%) of ecotoxicity impacts can be attributed to the release of just one chemical chloramine during the upstream step of production of pyridine whi ch is a key precursor for the cation. Another interesting fact emerges when we compare the ecotoxicity impacts of [Bmim] + [Br] and [Bmim] + [Cl] Both ILs have the same cation, similar (halogen) anion, similar precursor materials and very identical synthesi s routes. However, their production side ecotoxicty impacts vary a lot (48.3 CTUe and 16.4 CTUe respectively). This variation can entirely be attributed to the higher ecotoxicity impact associated with bromobenzene emission during the production of reactan t 1 bromobutane in comparison to lower ecotoxicity impacts of chlorobenzene emission during the production of reactant 1 chlorobutane. These results point to the fact that overall production side ecotoxicity impacts are very sensitive to certain key chemic al releases during the upstream synthesis steps. Next, we compare the sensitivity of the results to the production of one of the precursors, hydrogen chloride (HCl) or hydrochloric acid (HCl.H 2 O), through four different routes/processes.

PAGE 160

146 Table 7 5 : Breakdown of freshwater ecotoxicity impacts of ILs associated with use phase release IL Impact (CTU e ) Reactants Emissions Impact (CTU e ) % [Bmim] + [Br] 48.3 1 Methylimidazole 1 Bromobutane (BuBr) Metals 6.12 12.7 Bromobenzene Benzene 1,2 dibromobenzene 38.6 2.03 0.809 80.0 4.2 1.7 [Bmim] + [Cl] 16.4 1 Methylimidazole 1 Chlorobutane (BuCl) Metals 7.68 46.8 Chlorobenzene Benzene 1,2 dichlorobenzene 6.4 1.16 0.46 39 7.1 2.8 [BPy] + [Cl] 53.2 Pyridine 1 Chlorobutane (Bu Cl) Chloramine 40.5 76.1 Chlorobenzene Benzene 1,2 dichlorobenzene 6.65 1.2 0.48 12.5 2.2 0.9 HCL is used in the production of 1 chlorobutane which is needed for production of both [Bmim] + [Cl] and [BPy] + [Cl] The most common industrial process for the production of HCl is the chlorination of benzene which has very high ecotoxicty impacts as seen in Table E 4 ( Appendix E ). In comparison, we find that three other less common al ternate processes that can be used to produce HCl have very nominal impacts (See Table E 4). This analysis shows that identification and scale up of alternate efficient processes has the potential to significantly decrease the production side ecotoxicity i mpacts of ILs. It is also clear that ecotoxicity impacts due to the direct release of ILs during use phase is very sensitive to the release fraction. Therefore, through efficient manufacturing processes, careful control of IL

PAGE 161

147 release into water, and effic ient recovery and reuse we can further reduce use phase ecotoxicity impacts. In summary, the findings of this paper show that, for the studied ILs, ecotoxicity impacts related to the emissions associated with upstream IL production steps are significantly greater than impacts due to their estimated direct release during use phase. Further analysis revealed that ecotoxicity impacts due to chemical releases during upstream production steps outweigh upstream emissions related to energy use. It is also evident that one or two key precursor chemicals involved in the upstream steps of IL production contribute disproportionately high towards overall ecotoxicity impacts. We also see that different approaches (processes) to produce key precursor chemicals can resul t in widely varying ecotoxicty impacts. We propose that future research should focus on developing synthesis and purification steps that reflect green chemistry and green engineering principles 256 with an emphasis on lowering the life cycle impacts of the IL production phase. Few areas of importance would include tight control of chemical release, improve reaction yields and process efficiencies, identify alternate production processes for key precursors, and chemical recovery and reuse during the synthesis steps of ILs and their precursors. The results from this paper encourage further investigation into life cycle ecotoxicity impacts of other types of ILs. Future technologies based on ILs should consider the full life cycle ecotoxicity impacts in order to assess their risks and benefits. Due to the emerging nature of IL applications, life cycle assessment (LCA) studies are crucial now, as they are most beneficial during the early stages of technology development and can help avoid unintentional shift of env ironmental burdens

PAGE 162

148 from one stage to another. Such studies will enable integration of life cycle impacts as part of 7. 4 Life c ycle a ssessment of e nergetic i onic l iquids Energetic materials are used as explosives or as fuels. They release large amount of energy when they decompose. In the case of explosives all energy is released rapidly while in the case of fuels energy is released in a controlled manner. These material s derive their energy content from oxidation of the carbon backbone or from their high positive heats of formation. The general requirements for energetic materials are high energy density, thermal stability, low sensitivity to impact and low toxicity. 2 62 Traditional energetic materials that are commonly used in explosive formulations are HMX (1,3,5,7 tetranitro 1,3,5,7 tetraazacyclooctane), RDX (1,3,5 trinitro 1,3,5 triazacyclohexane) and TNT (2,4,6, trinitrotoluene). 263 Hydrazine derivatives are widely u sed as energetic fuel in rocket propulsion systems. 2 6 4,265 When discharged to the environment energetic materials will interact with biological systems. Use of energetic materials such as TNT, and RDX can leave residues which can potentially impact environ mental and human receptors. 2 66 Monitoring studies reveal that some of these munition compounds persist at the sites where they were produced or processed. 267 Unexploded and low order detonation residues containing TNT, RDX and HMX have been pointed out as main sources of groundwater contamination in military training ranges. 2 68 Indeed, munition compounds, such as RDX, have been detected in sole source drinking water aquifers in military ranges such as Camp Edwards. 2 69 These

PAGE 163

149 chemicals have been found to be m oderately to highly toxic to freshwater organisms. 266 In addition low concentrations of explosive compounds have been measured in marine sediments. 2 70 With an aim to address some of the above environmental concerns researchers are exploring other green ene rgetic material formulations. There is growing interest in the development of new energetic ionic salts and liquids for use as aerospace propellants and explosives. 2 64 As energetic materials, ionic salts offer several advantages over conventional energetic molecular compounds that include negligible volatility (ease of handling) and high density. 26 3 Energetic ionic salts can be prepared by combining energetic cations such as 1,2,3 triazolium with energetic anions such as nitrates, perchlorate and dinitramid e. The high heats of formation of these salts are primarily due to the presence of nitrogen containing cations and anions. 263 Nitrogen rich heterocyclic energetic salts are of particular interest. 271,275,276,277,278 A large number of ionic salts that are b ased on a triazole derivative have been proposed as energetic materials. 273,274,280 Triazole has a molecular formula of C 2 H 3 N 3 with a five membered ring that contain three nitrogen atoms located at 1,2,3 or 1,2,4 positions. 1,2,4 triazole and 1,2,3 triazole have heats of formation values of 109 KJ/mol and 272 KJ/mol respectively. As discussed previously, one of the main driving forces for the discovery and development of new energetic materials, such as ionic salts, is the mitigation of environmental and toxicological hazards associated with currently used materials. Manufacture of chemicals through environmentally friendly appro aches represents a fundamental industrial challenge.

PAGE 164

150 The energetic ionic salts possess lower vapor pressures and higher densities compared to non ionic molecules. 26 3 Due to their negligible vapor pressure they are usually considered as to volatile molecular compounds. In addition, ionic salts have tunable physical and chemical properties that enable us to tailor their structures for task specific applications such as energetic materials. Since ionic liquids and salts are of inherently l ess risk to human health and the environment they are considered as green chemicals. However, a more fundamental definition of green chemistry involves reducing or eliminating the use or generation of hazardous substances in the design, manufacture and app lication of chemical products. 281 In order to legitimately evaluate the greenness of ionic salts as energetic materials, it is not only enough to consider the inherently benign nature of the chemical, but also need to take a holistic view that considers en vironment and health impacts associated with the entire life cycle of their production including direct environmental emissions during the production phase and indirect emissions associated with energy use in their production. This study presents the first comprehensive cradle to gate life cycle assessment that considers all stages involved in production of 1,2,3 triazolium nitrate, a triazole based energetic ionic salt, and compares it with the environmental impact associated with production of TNT on a fu nctional unit basis. This approach will allow us to systematically investigate whether ionic salt based energetic materials provide any environmental benefits in comparison to traditional energetic materials. The chemical structure of 1,2,3 triazolium nitr ate and TNT are shown in Figure 7 2

PAGE 165

151 Figure 7 2 : a) 1,2,3 triaolzium nitrate; b) TNT There are several challenges involved in performing an LCA of ionic salts. Most of these challenges are due to the fact that ionic salts are a new class of compounds that are emerging. Ionic salts are not yet produced in large scales in commercial plants and there is no primary data available on material/energy consumption and direct environmental discharges. Process desig n and simulation software cannot be used to model production processes of ionic salts due to lack of comprehensive physical and thermodynamic property models for these salts and their precursors. Therefore, simulation of material and energy balances of ion ic salt production processes becomes very difficult. The other important challenge in modeling environmental impacts is that, there are no emissions factors available in LCA databases such as Ecoinvent and Gabi for several precursors (reactants) that are r equired for IL production. Due to these limitations no LCA study has been done on ionic salts. To our knowledge, even for ionic liquids, there have been only few LCA studies that have considered them in their analysis. 282,283,284 In order to overcome the a bove mentioned challenges we use a theoretical approach to estimate theoretical energy requirements for reaction and separation steps involved in ionic salt production. Then we adjust the theoretical energy requirements to actual energy consumption by acco unting for energy losses through the use of data from a

PAGE 166

152 comparable industrial process. Direct discharges of the ionic salt and its precursors to the environment during the production phase are assumed to be negligible. This energy and associated environmen tal loads constitute the inventory for life cycle assessment (LCA) method. 7. 4 .1 Process and E nergetic requirements for triazolium nitrate and TNT synthesis Synthesis of 1,2,3 triazolium nitrate : The main reaction for synthesis of triazolium nitrate propos ed by Drake et al. 26 2 and shown in e qn ( 7 10) is adopted for this study. However, the emission factors (life cycle emissions) for the reactants involved are not available in standard LCI databases such as Ecoinvent. Therefore, we consider a series of upst ream reactions e qn s. (7 4 a ) to (7 9), that constitute the life cycle tree for the production of ionic slat 1,2,3 triazolium nitrate. We calculate theoretical energy requirements for each of these steps that are part of the reaction tree. Major energy consumption in these batch processes would relate to the reaction and separati on stages. Wherever appropriate, we make further assumptions of minimal separation energy requirements (and therefore ignore them) when the products are in two different phases (easy to separate) or the product is of high yield (no need to separate small q uantities of by product). Actual industrial scenarios involving potential future scale up are expected to be more energy intensive, as in an industrial plant the actual energy consumption is few times greater than theoretical energy requirements due to hea t and energy losses. To capture this effect, we did a comprehensive review of several studies and found this factor to vary between 3 to 5 times that of theoretical energy requirement. In order to make an adjustment we selected a comparable process (synthe tic production of sodium

PAGE 167

153 carbonate) for which industrial energy consumption data was available. 285 We calculated the theoretical energy requirement for this process and compared it with actual energy consumed and found that actual electricity consumption i s 3.2 times higher than theoretical electricity requirement while actual natural gas consumption is 4.2 times higher than theoretical natural gas requirements. We also assume that for exothermic reactions electricity is used for cooling and for endothermic reactions natural gas is used for heating. We use the two correction factors in all our calculations to transform theoretical energy requirement to actual energy consumption. ( 7 4 a) ( 7 4 b) ( 7 5 ) ( 7 6 ) ( 7 7 ) ( 7 8 ) ( 7 9 ) (7 10 )

PAGE 168

154 The production of sodium is based on well known electrolysis cell process e qn s. (7 4 a ) and (7 4b) The potential required to oxidize Cl ions to Cl 2 is 1.36 volts and the potential needed to reduce Na + ions to sodium met al is 2.71 volts. Therefore, a potential of at least 4.07 volts is required to drive this reaction. 286 In the second step, sodium (solid) and ammonia (gas) are reacted at 375 to produce sodium amide and hydrogen e qn (7 5) 287 Sodium amide is in liquid phase and hydrogen is in the gas phase at this temperature. Thus, we assume that the energy requirement for separation of the two phase products in a small scale batch plant is equal to the energy required for cooling sodium amide from reaction temperatur e to room temperature. We calculated the theoretical heat of reaction as 2.12 MJ/Kg and theoretical heat of separation as 0.634MJ/Kg. Accounting for correction factors this translates into a total cooling load requirement of 2.0448 KWh per kg ionic salt a nd total heating load of 2.66 MJ per kg ionic salt (equivalent to 0.0616 m 3 of natural gas/kg).In the next step ammonia (gas phase) and oxygen are reacted at room temperature, to produce nitrous oxide and water (Eq. 3). Nitrous oxide is in gas phase and wa ter is in liquid phase at this temperature. Thus, we assume no significant energy requirement for separation in a small scale batch plant. We calculated the theoretical heat of reaction to be 15.47MJ/Kg. Accounting for correction factor this translates to 11.162 KWh per kg ionic salt. In the next step, sodium amide (solid phase) and nitrous oxide are reacted at 200 o C, to produce sodium azide (solid phase), sodium hydroxide (solid phase) and ammonia (gas phase). The energy requirement for separation stage is based on solid solid separation of the two solid products and the theoretical heat of reaction was 3.295 MJ/Kg. Accounting for correction factors, the total cooling load requirement translates to 2.87 KWh/kg and total

PAGE 169

155 heating load requirement translat es to 2.60 MJ/Kg (0.0602 m 3 of natural gas per kg of sodium azide). In the next step sodium azide (solid phase) is reacted with hydrochloric acid (liquid phase) at 65 o C, to produce hydrazoic acid (HN 3 ) and sodium chloride salt. 288 Since one of the product s is a gas and the other a solid, the two phases can be separated through a one step flash drum. Therefore we assume that for a batch plant the energy required for separation is minimal. We calculated the theoretical heat of reaction to be 2.9 MJ/kg. Acco unting for correction factor, this translates to 2.096 KWh/kg ionic salt. In the next step hydrazoic acid (HN 3 ) and acetylene gas are reacted at 25 o C, to produce 1,2,3 triazole ( 289 The yield for this reaction is 99 %. In view of very high yield ( low un reacted materials) and presence of no important byproducts, for all practical purposes 1,2,3 triazole ( can be considered pure. Thus, we assume no significant energy is required for separation. We calculated the theoretical heat of reaction as 2.275 MJ/kg. Accounting for correction factor, this translates to 3.68 KWh per kg ionic salt. In the next step 1,2 ,3 triazole ( and nitric acid (HNO 3 ) are reacted at 25 o C, to produce 1,2,3 triazolium nitrate (the energetic salt) with a yield, 98.9 %. Due to high yield of reaction (very low non reacted materials) and no important byproducts the product can be c onsidered as a pure component. Thus, we assume no significant energy is required for the separation part and the main energy consumption is the reaction phase. We calculated the theoretical heat of reaction as +1.73 MJ/Kg. Accounting for correction factor, this translates to 7.27 MJ/Kg of Ionic salt (0.168 m 3 Natural gas per kg of Ionic salt). The entire energy and material balance of the life cycle tree to produce 1 Kg of the ionic salt is shown in Figure 7 3

PAGE 170

156 Figure 7 3 : Material and energy flows associated with the life cycle tree for producing the ionic salt 1,2,3 triazolium nitrate Synthesis of TNT: TNT production is based on the synthesis procedure reported by Tadeusz 290 In the considered process toluene and nitric acid are reacted at 80 (both in liquid phase) to give trinitrotoluene (TNT). The industrial results show that 95 % of the product is and the rest is distributed between and We as sume that the major energy requirements for TNT production relates to the reaction step and separation step (separation of from byproducts). The theoretical energy

PAGE 171

157 requirement for the reaction step equals to heat of reaction calculated as 1.768 MJ/ kg. The actual energy consumed by accounting for energy losses was estimated using comparable plant data as described previously. The estimated actual electricity requirement for cooling in reaction and separation steps is 2.638 KWh per kg TNT. Energy requ irement for separation purposes has been predicted to be 0.902 MJ/Kg. 7. 4 .2 Life cycle assessment (LCA) of energetic ionic salts Functional Unit : Energy content is the most appropriate functional unit for this comparative study. 1,2,3 triazolium nitrate an d TNT have different energy content and their energy release mechanism also differs. While energy release from TNT is based on oxidation, the ionic salt relies on heat for formation. Therefore the heat of combustion for TNT and heat of formation for ionic salt were used as measures of energy content. A reference of 1 MJ energy content was used as the basis of comparison. On a mass equi valence basis this translates to following reference flow: 1 Kg of TNT equivalent to 1.62 Kg of ionic salt. System Boundary: The system boundary includes the final step of ionic salt production (reaction and separation), upstream reaction/separation steps for the precursors (as defined by the reaction tree), electricity and natural gas production, upstream processes involved in electricity and natural gas production including raw material extraction and transportation. Life Cycle Inventory: Life cycle inventory (LCI) represents the collection of data on the material and energy inputs and emissions associated with the production of the energetic ionic salt. Material and energy flows constructed in the previous section were used as inputs

PAGE 172

158 to the life cycle inventory (LCI). Since majority of processes are either new for which industrial scale up has not been developed or physical, c hemical and thermodynamic properties of the precursors are not available and hence chemical process simulation was not possible, we used data from the approach outlined in the previous section as inputs to the inventory. Due to the challenges associated wi th performing an LCA of new chemicals that were outlined earlier we consider this simplified approach as adequate for the scope of this study. Emission factors for production of electricity, natural gas and other starting materials of the life cycle tree were obtained from the U.S. life cycle inventory database. 291 The emission factor for electricity from grid was assumed as 70% generation from bituminous coal and 30% generation from natural gas, based on 2008 U.S. grid electricity data. 292 Since contribution of other renewable energy and nuclear sources to the grid were either very small or vary significantly depending on the location we assumed all grid electricity is from coal and natural gas. Emission factors for electricity production from bit uminous coal included emissions from coal mining & transport and emissions from power plant. Emission factors for electricity production from natural gas (NG) included emissions from NG extraction from ground and transport, emissions from NG processing and emissions from power plant. Emission factors for natural gas combustion include emission from NG extraction from ground, emissions from NG processing and emissions from NG combustion in an industrial boiler. Life cycle emission factors for some of the mat erials in the life cycle tree (sodium chloride [NaCl], ammonia [NH 3 ], oxygen [O 2 ], hydrochloric acid [HCl], ethylene [C 2 H 4 ] and nitric acid [HNO 3 ]) that were available in the US LCI database were used. Life cycle emission factors of the remaining materials (sodium [Na], sodium amide [NaNH 2 ], nitrous

PAGE 173

159 oxide [N 2 O], sodium azide [NaN 3 ], hydrazoic acid [HN 3 ] ethylene [C 2 H 2 ], 1,2,3 triazole [C 2 N 3 H 3 ]) were calculated using the theoretical approach explained in section 2. These emission factors have been applied t o the inputs to calculate the life cycle emissions for ionic salt and TNT production thereby completing the output side of the inventory (LCI). Life Cycle Impact Assessment: The life cycle impact assessment methods describe environmental impacts based on c haracterization factors. These characterization factors are developed by consideration of inherent characteristics of chemicals (for example toxicity) as well as information on fate and transport and possible mode of exposure. The life cycle impact assessm ent (LCIA) methodology based on Tools for the Reduction and Assessment of Chemical and other Environmental Impacts (TRACI) developed by U.S. Environmental Protection Agency was used in this study. This method was considered the most appropriate since it is based on United States data and models. This study considers only mid point impacts as end point impact modeling brings in additional uncertainty to the results and TRACI is primarily a mid point impact assessment method. Midpoint impact categories quanti fy the relevant emissions and resources from the life cycle inventory in terms of common reference substances (e.g. Kg CO 2 eq). The impact categories considered are: 1) Global warming; 2) Acidification; 3) Eutrophication; 4) Smog formation; 5) Human heath criteria; 6) Human health cancer; 7) Human health non cancer; and 8) Ecotoxicity. Classification and Characterization steps of LCA were applied to relate individual elementary flows in the inventory to the impact categories and to identify relevant charact erization factors based on the media to which the emissions occur. Normalization was not considered in this study as normalization factors based on U.S. data were not available.

PAGE 174

160 Sensitivity Analysis : The main source of uncertainty in this study relates to the conversion factors used for translating theoretical electricity and thermal energy requirement to actual energy consumed in an industrial plant. A sensitivity analysis is performed to study the effect of varying the conversion factors 30%. We examine in detail how sensitive the results are to changes in these conversion factors. 7. 4 .3 Results and d iscussion This section summarizes the main findings from comparing the ionic salt with TNT. The total scores of each environmental impact category for 1,2,3 triazolium nitrate and TNT are shown in Table 7 6 The impact profiles resulting from production of 1.62 kg of ionic salt and 1 kg of TNT are shown in Figure 7 4 with ionic salt impact set at 100% and TNT displayed as a level relative to the former. Rela tive comparisons between the ionic salt and TNT (Figure 7 4 ) show that in all of the analyzed categories ionic salt had significantly higher environmental/health impact than TNT. With respect to climate change, ionic salt production has roughly 3 times hig her environmental burden than TNT production. In the category of human health, ionic salt is roughly 3, 4 and 4 times more impactful than TNT for criteria, cancer and non cancer cases respectively. The environmental burden of IL is higher by approximately 2.5, 2, 2 and 4 times that of TNT for acidification, eutrophication, smog formation and ecotoxicity respectively. The climate change indicator, global warming air (GW), is dominated by CO 2 emissions during the life cycle of both IL and TNT production. CO 2 emissions account for 96% of total

PAGE 175

161 GW for both IL and TNT with methane accounting for the remaining 4%. Sulfur dioxide (70% for IL and 58% for TNT) and nitrous oxide (27% for IL and 33% for TNT) emissions dominate acidification indicator while photochemic al smog is dominated by nitrogen oxide emissions (90% for both IL and TNT). Eutrophication potential is dominated by nitrogen oxide (98.98% for IL and 96.22% for TNT) emissions. The human health criteria indicator is dominated by sulfur dioxide emissions ( 93% for IL and 90% TNT) while human health cancer and non cancer indicators are entirely due to benzene emissions. The ecotoxicity indicator is also ent irely due to benzene emissions. Table 7 6 : Impact of ion ic salt and TNT (functional unit: 1 MJ energy content) Category Units Ionic salt TNT Global Warming Air Kg CO 2 eq. 29.4738851 9.07309769 Acidification Air Kg H + mole eq. 9.489402937 3.752113828 HH Criteria Air Kg PM 10 eq. 0.023712006 0.007939867 Eutrophication Air Kg N eq. 0.00281911 0.001383166 Eutrophication water Kg N eq. 2.89911E 05 5.42903E 05 Smog Air Kg O 3 eq. 1.588845 0.777182 Ecotoxicity (Fresh air) CTU eco 0.111861804 3.3806E 06 Human health (Cancer) CTU cancer 2.47276E 11 6.34994E 12 Human Health (Non cancer) CTU noncancer 6.26146E 12 1.60791E 12 The results of the cradle to gate life cycle comparison unequivocally shows that energetic ionic salts, such as 1,2,3 triazolium nitrate, have larger environmental burden than traditional energetic materials such as TNT. This disproves the commonly accepte d notion that ionic their negligible vapor pressure, this fact alone does not make them green. A holistic analysis that includes the inherent properties of the ionic s alt, emissions associated with their

PAGE 176

162 production, exposure and end of life impacts need to be considered. This study provides greater insights into the greenness of energetic ionic materials through a more holistic approach. Closer examination of the result s reveal that majority of the life cycle environmental burden can be attributed to energy consumption (electricity and natural gas). This is due to the fact the emissions during energetic material production phase (reaction and separation) dominates other phases such as raw material extraction and transportation. Moreover the environmental footprint of the ionic salts is much larger than TNT due to the fact that steps involved in producing ionic salts and their precursors are much more energy intensive than the steps to produce TNT. This translates into significant increases in environment and health impacts across all categories. Since most of the emissions in this study can be attributed to electricity and natural gas consumption, using more efficient indu strial plants, and/or finding alternate synthesis pathways can result in net energy savings and help offset some of the emissions. Utilization of renewable resources such as solar, wind and waste biomass to produce electricity can significantly reduce the overall environmental footprint of the production process.

PAGE 177

163 Figure 7 4 : Comparison of scaled impacts of ionic salt and TNT (functional unit of 1MJ energy content): GWP (Global Warmin g Potential), AP (Acidification Potential), EP (Eutrophication Potential), HH (Human Health). Uncertainty Uncertainties related to conversion factors used for extrapolating theoretical energy calculations to actual energy consumption were addressed via sensitivity analysis. Additional sources of uncertainties are identified as follows: a) all calculations were based on the life cycle tree e qn s. (7 4a) to (7 10), and it is p ossible that alternate methods (reactions) exist for producing one or more of the precursors; b) the reaction yields are based on lab scale experiments from literature; c) we assume majority of the environmental impact for ionic salt and precursor producti on processes (i.e. life cycle tree) can be attributed to the energy intensive reaction and separation steps which could be a source of uncertainty; d) the calculations in this study are based on the assumption of small scale batch processes for producing t he ionic salt and precursors which could be another source of uncertainty if future manufacture plants are continuous. 0 20 40 60 80 100 Ionic Salt TNT

PAGE 178

164 Sensitivity theoretical energy to conversion factor is shown in Table 7 7 and Figure 7 5 The conversion factor was varied by 30% to study how sensitive the results are to this parameter. The error margins indicate that the results are highly robust to changes in this parameter and all conclusions that were previously arrived at are fully valid.

PAGE 179

165 Figure 7 5 : Sensitivity analysis of the scaled impacts of ionic salt and TNT (functional unit of 1 MJ): GWP (Global Warming Potential), AP (Acidification Potential), EP (Eutrophication Potential), HH (Human Health). The proportional size of the error bars indicate that the impacts a ssociated with the ionic salt are more sensitive to this parameter (conversion factor) than for TNT. This is due to the fact that there are more upstream reactions steps (shown in the life cycle tree) for ionic salt and hence the conversion factor is appli ed multiple times (in comparison to TNT). For bigger conversion factors the difference between ionic salt and TNT impacts will be even more than predicted. A comprehensive search of industrial data and a detailed analysis indicates that the considered conv ersion factors (3.2 and 4.2) fall in the lower end which implies that the LCA results presented here are conservative estimates. Table 7 7 : Sensitivity analysis Substance GWP AP EP Smog HH (cancer) HH (non cancer) Ecotox. Ionic Salt (IS) 29.4739 9.4894 0.00285 1.5888 2.473E 11 6.261E 12 1.32E 05 IS (+30%) 34.3751 11.029 0.00328 1.833 2.93E 11 7.420E 12 1.56E 05 IS ( 30%) 24.5725 7.9499 0.00241 1.3447 2.015E 11 5.103E 12 1.07E 05 TNT 9.0731 3.7521 0.00144 0.7772 6.35E 12 1.608E 12 3.38E 06

PAGE 180

166 Substance GWP AP EP Smog HH (cancer) HH (non cancer) Ecotox. TNT (+30%) 9.48553 3.8905 0.00148 0.7988 6.616E 12 1.675E 12 3.52E 06 TNT ( 30%) 8.65707 3.6126 0.00140 0.7554 6.081E 12 1.540E 12 3.24E 06 Scenario analysis In order to investigate the influence of the use of renewable energy on ionic salt production processes, we developed two hypothetical scenarios as follows: 1) In the first scenario cooling energy (electricity) required for all materials needed during ionic salt production (final reaction/separation unit a s well as all upstream reactions/separation units) comes from wind (renewable source); 2) In the second scenario wind energy is used only in the last two meaningful du e to the fact that an ionic liquid production plant is likely to purchase the primary raw materials (acetylene, hydrazoic acid and nitric acid upstream processes) from other industries that would likely use electricity from grid. For both scenarios, natu ral gas is assumed to deliver the required heating energy. In both cases of comparison it is assumed that fossil energy is completely used for TNT production. The results for the two scenarios are summarized in Table 7 8 and Table 7 9 respectively. It can be concluded that under scenario one the environmental impact of ionic salt, in most categories, is lower than TNT. However, under scenario two (more realistic) the environmental impact of ionic salt is still significantly higher than TNT. It will be reaso nable to assume that under a scenario where only renewable energy is used (in all stages) for both ionic salt and TNT production the magnitude of environmental impact for both cases will be lowered proportionally, however

PAGE 181

167 the final analysis and conclusions presented in relation to the comparison between them will still hold true (since ionic salt production processes consume significantly more life cycle energy than TNT production processes). Table 7 8 : Enviro nmental Impact for scenario 1 Category Units Ionic salt TNT Global Warming Air Kg CO 2 eq. 2.778853 9.073098 Acidification Air Kg H + mole eq. 0.541855 3.752114 HH Criteria Air Kg PM 10 eq. 0.000932 0.007940 Eutrophication Air Kg N eq. 0.000347 0.001383 Eutrophication water Kg N eq. 2.7504E 05 5.42903E 05 Smog Air Kg O 3 eq. 0.197913 0.777181 Ecotoxicity (Fresh air) CTU eco 3.94306E 06 3.3806E 06 Human health (Cancer) CTU cancer 7.40643E 12 6.34994E 12 Human Health (Non cancer) CTU noncancer 1.87544E 12 1.60791E 12 Table 7 9 : Environmental Impact for scenario 2 Category Units Ionic salt TNT Global Warming Air Kg CO 2 eq. 27.29388 9.0730977 Acidification Air Kg H + mole eq. 8.758716 3.752114 HH Criteria Air Kg PM 10 eq. 0.021851 0.007939 Eutrophication Air Kg N eq. 0.002617 0.001383 Eutrophication water Kg N eq. 2.88697E 05 5.42903E 05 Smog Air Kg O 3 eq. 1.475256 0.777181 Ecotoxicity (Fresh air) CTU eco 1.24115E 05 3.3806E 06 Human health (Cancer) CTU cancer 2.33131E 11 6.34994E 12 Human Health (Non cancer) CTU noncancer 5.90329E 12 1.60791E 12

PAGE 182

168 The analysis presented in this study considers only the life cycle energy consumption. Environmental impacts associated with emissions were not assessed. In order to get a more complete picture it is necessary to perform a cradle to grave life cycle assessment that will include indirect emissions, direct material emissions (of ionic salt, precursors etc.) during production phase, as well as end of life (after use) impacts of the energetic materials. However, this is not currently possible as the state of the art impact assessment methods, such as TRACI 293 and Eco Indicator 294 do not contain characterizati on factors for ionic salts and many of their precursor materials. Therefore, there is a great need for future research to focus on fate, transport, and mechanism of damage to human and ecosystem species by ionic salts, ionic liquids and their precursor mat erials. This will allow us to develop characterization factors for these compounds and help us investigate the exposure and end of life impacts of the emitted material. The results presented in this study are subject to several uncertainties.

PAGE 183

169 Conclusion s and Future Work Chapter 8: This study focused on the development of a novel computer aided molecular design framework and environmental impact assessment methods of ionic liquids, a newer generation of materials which have attracted much attention during the past few years Ionic liquids are shown to have the capability of replacing organic compounds used in industri al plants for a vast number of applications; from liquid liquid extraction to thermal energy storage. One of the important components o f green chemistry is attributed to the solvent medium in which a chemical reaction or an extraction process is carried out 29 5 Traditional molecular solvents are proven to cause adverse environmental and human health impacts and therefore they should be re placed with greener alternatives whenever a better candidate, which can do similar tasks, is available. Jessop 296 states that one of the major challenges in the search for environmentally benign or green solvents is to ensure their availability. Jessop us ed the Kamlet Taft plots to show that current list of green solvents populate only a small region of the entire spectrum of solvents needed for different applications and argues that large unpopulated areas of this diagram mean that future process chemists and engineers need solvents having specific desirable properties which are also green. Many studies have concluded that ionic liquids offer a great potential to satisfy this need by serving as the potential candidate to be used for several different appli cations. However, the true greenness of any chemical can be ascertained only when a holistic view of its environmental impacts is considered. Towards, this end this dissertation has made contribution in the realm of understanding and characterizing life cy cle environmental performance of ionic liquids.

PAGE 184

170 This dissertation also presented an overarching framework, termed as CAILD, that can be utilized to design or tune ionic liquids through a modeling perspective The details of the CAILD optimization approach along with several important case studies were presented The success of ionic liquid design requires that the physical properties of these new compounds be predicted with an acceptable level of accuracy. Chapter two of this dissertation focused on contributions related to the prediction of two pure/physical properties of ionic liquids namely melting point and viscosity, as it is a necessary step for successful application of ionic liquids This was accomplished through a Quantitative Structure Prop erty Relationship (QSPR) approach were quantum chemistry (QC) based descriptors were used to develop physical property correlations. The QC models which normally do not need any structural group s related parameters can be combined with QSAR/QSPR approach to predict the physical properties of chemical compounds with an acceptable precision. The used QC models rely on the energy profile and the charge density of the chemical compound of interest and can be utilized when there is no reliable group contributio n approach available. Chapters five and six of this dissertation were focused mainly on the prediction and use of solution propert ies in this case activity coefficients, of ILs and solutes in a multicomponent mixture Due to lack of data we proposed a new approach for fitting the interaction parameters based on QC and COSMO predictions. The activity coefficients estimated through the fitted UNIFAC model were used to predict the solubility of several solutes in ILs. Th e approach was used to find optimal sol vents with desired properties (e.g. high solvency power towards the solute of interest low melting point etc. ) for different

PAGE 185

171 applications. Chapter five was focused on designing an optimal IL for the extraction of aromatic compounds from an aliphatic arom atic mixture In c hapter six a case study was presented where an optimal IL was designed to act as the solvent for a CO 2 capture process Chapter 7 focused on studying the environmental impacts associated with the production and release of common ILs through development of their characterization factors. The se characterization factors (CF) were used to calculate the freshwater ecotoxicty impacts associated with the direct release of the ionic liquids into the environment In the case of freshwater ecot oxicty impacts, the impacts associated with the production of ILs far outweighed the impacts linked to the potential release in to freshwater resources To conclude, the l ast section of chapter 7 described the results of comparative cradle to gate life cyc le assessment s (LCA) between two energetic materials one being an energetic ionic liquid and the other being a convectional energetic compound. The results showed that for this specific case, when it comes to the life cycle impacts associated with the pro duction, ILs might not be more environmentally benign than their molecular counterparts. This finding is interesting in the sense that it contradicts with the commonly accepted belief of ILs as green materials. This fact along with the high toxicity of som e ILs towards aquatic organisms This also shows that ionic liquid s need to be tuned/designed specifically for a given application by considering not only their pe rformance but also their environmental impacts before they are produced in larger commercial scales.

PAGE 186

172 8.1 Limitations and Recommendations F or a computer aided design framework to successfu lly search for optimal candidate s in different cases it needs to cover a broad range of structural groups acting as building blocks of the compound being designed through the model That being said, the first challenge in computer aided design of ionic liquids is that the available group contribution models of the phys ical properties of ILs do not span the entire spectrum of cations, anion and functional groups Although, i t is not impossible to overcome this challenge since there are only a limited number of structural group s mak ing an enormous number of ILs this has not yet happened. W e strongly propose that future research should focus on experimental measurement s and data collection of the physical properties of ionic liquids to expand the available group contribution methods to cover a more diverse set of cations, anions and /or functional groups. The second challenge is that for certain physical properties such as surface tension, there are no group contributions available. In addition, for certain physical properties such as viscosity, the available models either fails to predict the property of newer compounds with an acceptable range of accuracy or they are very complicated and cannot be integrated directly within a CAILD framework. In this study, we made use of quantum chemistry (QC) models combined with specif ic QSAR/QSPR correlations to predict viscosity and melting point of ILs. The limiting part was that even though, through QC models based correlations, we were able to predict the two physical properties of ILs with a relatively better accuracy (compared to existing group contribution methods), but it was not feasible to integrate them

PAGE 187

173 directly within a CAILD framework. Therefore, in certain case studies we first had to find a set of optimal and feasible IL candidates based on the objective function and othe r constraints and then evaluate the list of ionic liquids for their melting point and viscosity requirements using the developed correlations in chapter 2. The third challenge relate s to the accuracy of the group contribution models that we used as part o f the CAILD model and the intrinsi c uncertainty they carried. The group contribution (GC) models are meant to predict the physical properties of newer compounds, based on the type and number of the structural groups present in these compounds. In order to develop a good GC model for a physical property, we need to have diverse set of chemicals with overlapping structural groups. In the next step for all of the selected chemicals, the experimental data on the physical property of interest should be either m easured or collected from literature so that the data can be used to fit group contributions for different structural groups. These contributions along with the information on the type and number of groups can be utilized to predict the physical properties of newer compounds. Although this approach seems to be straightforward, the limitation resides in the uncertainty of the experimental data used to develop group contribution models. This problem can be addressed if the uncertainty of the experimental data is reported with the data, but for most ionic liquids that was not the case. In addition to that, experimental data collected from different sources normally do not have the same level of uncertainty. This means that the intrinsic uncertainty embedded in the group contribution models which are used to predict the physical properties of ILs in a CAILD model is not accurately measureable. These uncertainties will also be carried on to the design results i.e. the structure of the optimal IL. This study, was n ot able to capture these

PAGE 188

174 uncertainties as, for the most part, we were restricted to using group contribution models available in the literature for which uncertainty values were not reported. Another challenge relates to the limitation of USEtox model whic h was used to develop the ecological characterization factors of common ionic liquids. The model itself is a state of the art and most comprehensive model available on the fate and transport modeling of chemical compounds in the environment. Although, the USEtox model is appropriate for our calculations the liming part is that it was primarily developed for molecular compounds and the model has not been previously used for the case of ionic liquids. The large and asymmetric shape of the cations and anions i n ionic liquids along with their electrical charges make the accuracy of USEtox for the case of ionic liquids questionable. Despite this limitation, other studies have used USEtox to develop characterization factors for materials such as Carbon Nanotubes ( CNT) which are made of extensively large unit cells. Another limitation of using the USEtox model for the case of ionic liquids can be attributed to their potential dissociation in the water media since that would change the measured physical properties. T he studies on the behavior of ionic liquids in the aquatic environments and the data on their dissociation factors are still very limited. The promising part is that some studies have suggested that ionic liquids mostly act as Lewis acids in water rather t han as an ionic material such as sodium chloride. The aforementioned fact implies that ILs can be assumed to be mostly dissolved in water because of their hydrogen bonding interaction with water molecules rather than being completely disassociated. The fac t that some ionic liquids are not soluble in water, despite their ionic nature, can add further credence to this hypothesis.

PAGE 189

175 The results of the life cycle studies which were performed in this dissertation are limited to a few numbers of ionic liquids and should not be used to make any general conclusion or comments about ionic liquids in general. Conclusions such as ionic liquids are more impactful compared to the case of organic compounds or similar to that should be only made case by case and after care ful consideration of the results obtained from the comprehensive life cycle assessments. The fact that there is no actual data on the production of ionic liquids make results of their life cycle assessment studies more limited and uncertain. 8.2 Contr ibutions Even though there is still dispute over the time when the first ionic liquid was discovered, Ethanolammonium nitrate was first reported by Gabriel and Weiner 297 in 1888 and it was as early as 1943 that the term ionic liquid was first used. Since t hen, thousands of scientific papers and patents have been published in the areas of synthesis and use of ionic liquids. Ionic liquids are one of the most widely studied categories of materials during the past few years. Not a single day goes by without peo ple thinking about a new application for ionic liquids. That being said, the process of choosing the best and most optimal ionic liquid is quite challenging. Ionic liquids have a very wide range of properties so any change in their structures can change th eir physical and solution properties significantly. An ionic liquid which is well suited for a given application might be a bad choice for another application. The refore the challenge of optimal IL selection can be addressed accordingly through their intel ligent design.

PAGE 190

176 Despite the unique characteristics and promising properties of ionic liquids, only a few hundred (compared to 10 14 theoretically possible structures) ionic liquids have been synthetized and tested to this date. Therefore, people have only tried a very small subset of all feasible ionic liquids for any application of interest presenting a great platform to discover new ionic liquids On the other hand, property prediction through structure property models is limited because such models have not extensively been developed. The above limitations, need to be addressed before any attempt of large scale commercialization of ILs is made. The ultimate aim of the CAILD methodology, would be to narrow down to a smaller set of ionic liquid candidates from the millions of available alternatives The final IL, which is most optimal for a given task can then be selected by ab initio computational chemistry calculations or by experimental verification of the designed compounds. The progress tow ards designing newer ionic liquids through the proposed CAILD model, will not only contribute towards our understanding of the relationship between cation and anion structures and the physical properties of ionic liquids, but will also provide a mechanism to engineer new ionic liquids which can also be environmentally benign for a given application. Another important contribution of this study was to perform life cycle assessment on well known ionic liquids. The environmental impacts associated with the pr oduction of an energetic ionic salt was compared with a conventional explosive material. The approach we used for this study was innovative and unique in the sense that certain conversion factors were exclusively developed for this study which were used to convert theoretical energy

PAGE 191

177 requirements of the production of the ionic salt to actual industrial data. Since there was no actual data available on the production of ionic salts this approach made the life cycle assessment and the calculation of energy re lated impacts during the production stages possible. Finally, t he freshwater ecotoxicity characterization factors related to five common ionic liquids were calculated for the first time through the use of a comprehensive fate and transport model USEtox. T he characterization factors developed were later used to preform cradle to grave life cycle assessments of these ionic liquids. T he life cycle assessments enabled us to see the difference s between life cycle impacts of production steps and the hypothetic al release of ionic liquids into the environment. 8.3 Future work on ionic liquid applications A general and comprehensive CAILD model has now been developed and several case studies were solved to show the effectiveness of this approach. Below several suggestions that can advance the approach and expand the model furthermore are proposed. Once more, w e strongly emphasize that future research should focus on data collection as well as experimental measurement s of physical and thermodynamic properties of ionic liquids This will enable us to expand the available group contribution models to cover more diverse set of cations and anions including those which are less common or have not been studied yet. Further we also suggest development of new group contr ibution models of physical properties for which no model currently exists.

PAGE 192

178 We also propose that process simulation models should be developed to measure the actual energy required to produce ionic liquids at a larger scale. This data would allow more acc urate LCA analysis of ionic liquids enabling us to make more informed decisions about their environmental impacts before investing on a particular ionic liquid. In addition, the studies on toxicity and corrosivity of ionic liquids are still limited. There fore, we strongly propose that future research should focus on measuring and predicting these two properties of ionic liquids since it might be very costly to use an ionic liquid with a high degree of corrosivity or toxicity. This information will help in making an informed selection of an optimal ionic liquid for a particular application and it is necessary that before any investment on the production of ionic liquids in large scales can be made. The use of the CAILD model, is not limited to the case stud ies we introduced in this dissertation. In this section of dissertation we propose several more cases for which a CAILD model can be developed accordingly to help in finding ionic liquids. Absorption chillers One of the potential where ionic liquids appli cations can be utilized is in absorption chiller, where the absorbent acts as a compressor. The cooling agent which is normally water would be absorbed into the absorbent agent, normally a molten salt such Li + Br and will get condensed. In the next step, water in the vapor phase will be separated from the absorbent in the generator by heating the solution. Furthermore, the water vapor gets cooled using the cooling water and then the liquefied vapor will be again evaporated to produce the chilling

PAGE 193

179 effect. I onic liquids have shown the capability of replacing molten salts in closed cycle absorption chiller, since they usually have an acceptable range of melting point and high decomposition temperature. The ionic liquid for this process needs to be optimized to have an acceptable crystallization temperature and high affinity towards water. An absorption chiller which performs using the optimal IL, designed through the CAILD model, would have a Coefficient of Performance (COP) comparable to that of a conventional absorption chiller without having their drawbacks such as easy crystallization. 298 Electrolytes One of the other applications of ionic liquids is as electrolyte in an electrolytic capacitor or in a battery. The high electrical conductivity of certain ionic liquids along with their tunable viscosity make them interesting for this application. An optimal ionic liquid with high ionic conductivity and electrochemical window which also has accept able values of melting point and viscosity and is not corrosive and/or explosive can be designed through CAILD for electrochemical applications. 299 Lubricants One of the applications of ionic liquids which has attracted a lot of attention is their use as a lubricant agent. A low viscosity ionic liquid with low corrosivity can be used to reduce the friction between the metallic compounds in a particular system. A CAILD model can be developed to design an ionic liquid with good thermal properties, low viscos ity and low melting point, and low corrosivity to be exploited as a lubricant of a system. Such anionic

PAGE 194

180 liquid would flow throughout the system to cool down the moving objects by absorbing the heat being produced. 300 Surfactants A s urfactant is a compound that has the capability of lower ing the surface tension between two liquid phases or between a liquid and a solid. If further research in the future allows us to have precise group contribution models to estimate surface tension of ionic liquids as well a s the accurate correlative models to predict the interfacial tension between an ionic liquid and other chemical compounds, a CAILD framework can be developed to design surfactants. The model would be able to design an optimal ionic liquid with an objective of lowering the surface tension between the two compounds of interest while having acceptable values for melting point, viscosity, toxicity and other necessary thermal properties. 301 Cellulose dissolution Several studies have shown that certain type of ionic liquids can act as a good solvent for cellulose extraction process. Cellulose, the most abundant natural polymer on earth, can be extracted from plant residues wood pulp or cotton by means of a solvent which can dissolve the cellulose out of the primary cell walls of green plants. The extracted cellulose can be mainly used to produce paperboard and paper. The advances on the field of ionic liquids have led to room temperature ionic liquids with relatively low viscosity that can dissolve cellulose. A CAILD model can be used to design an optimal ionic liquid with highest solvency power towards cellulose while also having a relatively low value of melting point and viscosity. In order to do that, group contributio n models, such as UNIFAC, capable of predicting the

PAGE 195

181 solubility of a polymer (cellulose) in a solvent, should be advanced to include. The developed model would allow us to predict the activity coefficients of the cellulose in any ionic liquid of interest wh ich can be further translated to the solubility of cellulose. This would help us find an ionic liquid, as a solvent for this process, with highest solvency power possible. 302

PAGE 196

182 R EFERENCES (1) Rogers, R. D., & Seddon, K. R. (2003). Ionic liquids -solvents of the future?. Science 302 (5646), 792 793. (2) Armand, M., Endres, F., MacFarlane, D. R., Ohno, H., & Scrosati, B. (2009). Ionic liquid materials for the electrochemical challenges of the future. Nature materials 8 (8), 621 629. (3) Walden, P. (1914). Molecular weights and electrical conductivity of several fused salts. Bull. Acad. Imper. Sci.(St. Petersburg) 8 405 422. (4) Welton, T. Room temperature ionic liquids. So lvents for synthesis and catalysis. Chemical reviews 1999, 99, 2071 2084. (5) Plechkova, N. V.; Seddon, K. R. Applications of ionic liquids in the chemical industry. Chemical Society Reviews 2008, 37, 123 150. (6) Gani, R., Nielsen, B., & Fredenslund, A. (1991). A group contribution approach to computer aided molecular design. AIChE Journal 37 (9), 1318 1332. (7) Achenie, L., Venkatasubramanian, V., & Gani, R. (Eds.). (2002). Computer aided molecular design: theory and practice (Vol. 12). Elsevier. (8) Harper, P. M., Gani, R., Kolar, P., & Ishikawa, T. (1999). Computer aided molecular design with combined molecular modeling and group contribution. Fluid Phase Equilibria 158 337 347. (9) Maranas, C. D. (1996). Optimal computer aided molecular design: A polymer design case study. Industrial & engineering chemistry research 35 (10), 3403 3414. (10) Karunanithi, A. T., Achenie, L. E., & Gani, R. (2005). A new decomposition based computer aided molecular/mixture design methodology for the design of optimal solvents and solvent mixtures. Industrial & engineering chemistry research 44 (13), 4785 4797. (11) Karunanithi, A. T., Achenie, L. E., & Gani, R. (2006). A computer aided molecular design framework for crystallization solvent design. Chemical Engineering Science 61 (4), 1247 1260. (12) Pretel, E. J., Lpez, P. A., Bottini, S. B., & Brignole, E. A. (1994). Computer aided molecular design of solvents for separation processes. AIChE Journal 40 (8), 1349 1360. (13) Karunanithi, A. T., & Mehrkesh, A. (2013). Computer aided design of tailor made ionic liquids. AIChE Journal 59 (12), 4627 4640.

PAGE 197

183 (14) Chong, F. K., Foo, D. C., Eljack, F. T., Atilhan, M., & Chemmangattuvalappil, N. G. (2014). Ionic liquid design for enhanced carbon dioxide capture by computer aided molecular design approach. Clean Technologies and Environmental Policy 1 12. (15) Hada, S., Herring, R. H., Davis, S. E., & Eden, M. R. (2015). Multivariate characterization, modeling, and design of ionic liquid molecules. Computers & Chem ical Engineering (16) Harini, M., Jain, S., Adhikari, J., Noronha, S. B., & Rani, K. Y. (2015). Design of an ionic liquid as a solvent for the extraction of a pharmaceutical intermediate. Separation and Purification Technology (17) Gardas, R. L., & Coutinho, J. A. (2008). Extension of the Ye and Shreeve group contribution method for density estimation of ionic liquids in a wide range of temperatures and pressures. Fluid Phase Equilibria 263 (1), 26 32. (18) Gardas, R. L., & Coutinho, J. A. (2008). A group contribution method for viscosity estimation of ionic liquids. Fluid Phase Equilibria 266 (1), 195 201. (19) Luis, P., Ortiz, I., Aldaco, R., & Irabien, A. (2007). A novel group contribution method in the development of a QSAR for predicting the to xicity (Vibrio fischeri EC 50) of ionic liquids. Ecotoxicology and environmental safety 67 (3), 423 429. (20) Carlisle, T. K., Bara, J. E., Gabriel, C. J., Noble, R. D., & Gin, D. L. (2008). Interpretation of CO2 solubility and selectivity in nitrile func tionalized room temperature ionic liquids using a group contribution approach. Industrial & Engineering Chemistry Research 47 (18), 7005 7012. (21) Gardas, R. L., & Coutinho, J. A. (2008). A group contribution method for heat capacity estimation of ionic liquids. Industrial & Engineering Chemistry Research 47 (15), 5751 5757. (22) Lazzs, J. A. (2012). A group contribution method to predict the melting point of ionic liquids. Fluid Phase Equilibria 313 1 6. (23) Lazzs, J. A. (2012). A group contributi on method to predict the thermal decomposition temperature of ionic liquids. Journal of Molecular Liquids 168 87 93. (24) Gharagheizi, F., Ilani Kashkouli, P., Mohammadi, A. H., Ramjugernath, D., & Richon, D. (2012). Development of a group contribution method for determination of viscosity of ionic liquids at atmospheric pressure. Chemical Engineering Science 80 326 333.

PAGE 198

184 (25) Stoimenovski, J., MacFarlane, D. R., Bica, K., & Rogers, R. D. (2010). Crystalline vs. ionic liquid salt forms of active pharma ceutical ingredients: a position paper. Pharmaceutical research 27 (4), 521 526. (26) Lapkin, A. A., Plucinski, P. K., & Cutler, M. (2006). Comparative assessment of technologies for extraction of artemisinin. Journal of natural products 69 (11), 1653 166 4. (27) Stoimenovski, J., MacFarlane, D. R., Bica, K., & Rogers, R. D. (2010). Crystalline vs. ionic liquid salt forms of active pharmaceutical ingredients: a position paper. Pharmaceutical research 27 (4), 521 526. (28) Karkamkar, A., Aardahl, C., & Autrey, T. (2007). Recent developments on hydrogen release from ammonia borane. Material Matters 2 (2), 6 9. (29) Rao, C. H., Venkatesan, K. A., Nagarajan, K., & Srinivasan, T. G. (2008). Dissolution of uranium oxides and electrochemical behavior of U (VI ) in task specific ionic liquid. Radiochimica acta 96 (7/2008), 403 409. (30) Wu, B., Reddy, R., & Rogers, R. (2001). Novel ionic liquid thermal storage for solar thermal electric power systems. Solar Engineering 445 452. (31) Pham, T. P. T., Cho, C. W., & Yun, Y. S. (2010). Environmental fate and toxicity of ionic liquids: a review. Water research 44 (2), 352 372. (32) Pretti, C., Chiappe, C., Pieraccini, D., Gregori, M., Abramo, F., Monni, G., & Intorre, L. (2006). Acute toxicity of ionic liquids to the zebrafish (Danio rerio). Green Chemistry 8 (3), 238 240. (33) Zhao, D., Liao, Y., & Zhang, Z. (2007). Toxicity of ionic liquids. Clean soil, air, water 35 (1), 42 48. (34) Ranke, J., Stolte, S., Strmann, R., Arning, J., & Jastorff, B. (2007). Desig n of sustainable chemical products the example of ionic liquids. Chemical Reviews 107 (6), 2183 2206. (35) Mehrkesh, A., & Karunanithi, A. T. (2014, July). New Perspective on Computer Aided Molecular Design: A Life Cycle Assessment Approach. In Proceedings of the 8th International Conference on Foundations of Computer Aided Process Design (Vol. 34, p. 369). Elsevier. (36) Li, X., Zhao, J., Li, Q., Wang, L., & Tsang, S. C. (2007). Ultrasonic chemical oxidative degradations of 1, 3 dialkylimidazolium ionic l iquids and their mechanistic elucidations. Dalton transactions (19), 1875 1880.

PAGE 199

185 (37) Smiglak, M., Reichert, W. M., Holbrey, J. D., Wilkes, J. S., Sun, L., Thrasher, J. S., ... & Rogers, R. D. (2006). Combustible ionic liquids by design: is laboratory saf ety another ionic liquid myth?. Chemical Communications (24), 2554 2556. ( 38 ) Karunanithi, A. T., & Mehrkesh, A. (2013). Computer aided design of tailor made ionic liquids. AIChE Journal 59 (12), 4627 4640. ( 39 ) Cole, A. C., Jensen, J. L., Ntai, I., Tran, K. L. T., Weaver, K. J., Forbes, D. C., & Davis, J. H. (2002). Novel Brnsted acidic ionic liquids and their use as dual solvent catalysts. Journal of the American Chemical Society 124 (21), 5962 5963. ( 40 ) Wilkes, J. S. (2004). Properties of ionic liquid solvents for catalysis. Journal of Molecular Catalysis A: Chemical 214 (1), 11 17. ( 4 1) Preiss, U., Bulut, S., & Krossing, I. (2010). In silico prediction of the melting points of ionic liquids from thermodynamic considerations: a case study on 67 salts with a melting point range of 337 C. The Journal of Physical Chemistry B 114 (34), 11133 11140. ( 42 ) Fredlake, C. P., Crosthwaite, J. M., Hert, D. G., Aki, S. N., & Brennecke, J. F. (2004). Thermophysical properties of imidazolium based ionic liqui ds. Journal of Chemical & Engineering Data 49 (4), 954 964. ( 43 ) Hu, H. C., Soriano, A. N., Leron, R. B., & Li, M. H. (2011). Molar heat capacity of four aqueous ionic liquid mixtures. Thermochimica Acta 519 (1), 44 49. ( 44 ) Vila, J., Fernndez Castro, B., Rilo, E., Carrete, J., Domnguez Prez, M., Rodrguez, J. R.,... & Cabeza, O. (2012). Liquid solid liquid phase transition hysteresis loops in the ionic conductivity of ten imidazolium based ionic liquids. Fluid Phase Equilibria 320 1 10. ( 45 ) Bradle y, J C., (2011), Alfa Aesar melting point data now openly available ( 46 ) Hughes, T. J., Syed, T., Graham, B. F., Marsh, K. N., & May, E. F. (2011). Heat capacit ies and low temperature thermal transitions of 1 hexyl and 1 octyl 3 methylimidazolium bis (trifluoromethylsulfonyl) amide. Journal of Chemical & Engineering Data 56 (5), 2153 2159. ( 47 ) 2012). Solubility advantage of pyrazine 2 carboxamide: Application of alternative solvents on the way to the future pharmaceutical development. Journal of Chemical & Engineering Data 57 (5), 1525 1533. ( 48 ) Bradaric, C. J., Downard, A., Kennedy, C., Rober tson, A. J., & Zhou, Y. (2003). Industrial preparation of phosphonium ionic liquids. Green Chemistry 5 (2), 143 152.

PAGE 200

186 ( 49 ) Coker, T. G., Ambrose, J., & Janz, G. J. (1970). Fusion properties of some ionic quaternary ammonium compounds. Journal of the American Chemical Society 92 (18), 5293 5297. ( 50 ) ChemFiles, Enabling Technologies Ionic Liquids, Vol.5 No.6. www.sigmaaldrich.com/content /dam/sigma aldrich/docs/Aldrich/Brochure/al chemfile_v5_n6.pdf ( 51 ) Li, H., Zhao, G., Liu, F., & Zhang, S. (2013). Physicochemical Characterization of MF m Based Ammonium Ionic Liquids. Journal of Chemical & Engineering Data 58 (6), 1505 1515. ( 52 ) Vere & Schick, C. (2013). Does alkyl chain length really matter? Structure property relationships in thermochemistry of ionic liquids. Thermochimica Acta 562 84 95. ( 53 ) Nishida, T., Tashiro, Y., & Yamamoto, M. (2003). Physical and electrochemical properties of 1 alkyl 3 methylimidazolium tetrafluoroborate for electrolyte. Journal of Fluorine Chemistry 120 (2), 135 141. (5 4 ) Fletcher, S. I., Sillars, F. B., Hudson, N. E., & Hall, P. J. (2009). Physical properties of selected ionic liquids for use as electrolytes and other industrial applications. Journal of Chemical & Engineering Data 55 (2), 778 782. ( 55 ) Wachter, P., Schweiger, H. G., Wudy, F., & Gores, H. J. (2008). Ef ficient determination of crystallisation and melting points at low cooling and heating rates with novel computer controlled equipment. The Journal of Chemical Thermodynamics 40 (10), 1542 1547. ( 56 ) Zhang, Z. H., Tan, Z. C., Li, Y. S., & Sun, L. X. (2006) Thermodynamic investigation of room temperature ionic liquid. Journal of thermal analysis and calorimetry 85 (3), 551 557. ( 57 ) MacFarlane, D. R., Meakin, P., Amini, N., & Forsyth, M. (2001). Structural studies of ambient temperature plastic crystal ion conductors. Journal of physics: condensed matter 13 (36), 8257. ( 58 ) advantage of pyrazine 2 carboxamide: Application of alternative solvents on the way to the future p harmaceutical development. Journal of Chemical & Engineering Data 57 (5), 1525 1533. ( 59 ) Liu, Q. S., Yang, M., Li, P. P., Sun, S. S., Welz Biermann, U., Tan, Z. C., & Zhang, Q. G. (2011). Physicochemical properties of ionic liquids [C3py][NTf2] and [C6py ][NTf2]. Journal of Chemical & Engineering Data 56 (11), 4094 4101.

PAGE 201

187 ( 60 ) Carvalho, P. J., Regueira, T., Santos, L. M., Fernandez, J., & Coutinho, J. A. (2009). Effect of Water on the Viscosities and Densities of 1 Butyl 3 methylimidazolium Dicyanamide and 1 Butyl 3 methylimidazolium Tricyanomethane at Atmospheric Journal of Chemical & Engineering Data 55 (2), 645 652. ( 61 ) Harris, K. R., Kanakubo, M., & Woolf, L. A. (2007). Temperature and pressure dependence of the viscosity of the ionic liqui d 1 butyl 3 methylimidazolium tetrafluoroborate: viscosity and density relationships in ionic liquids. Journal of Chemical & Engineering Data 52 (6), 2425 2430. ( 62 ) Seddon, K. R., Stark, A., & Torres, M. J. (2000). Influence of chloride, water, and organ ic solvents on the physical properties of ionic liquids. Pure and Applied Chemistry 72 (12), 2275 2287. ( 63 ) Baker, S. N., Baker, G. A., Kane, M. A., & Bright, F. V. (2001). The cybotactic region surrounding fluorescent probes dissolved in 1 butyl 3 methy limidazolium hexafluorophosphate: effects of temperature and added carbon dioxide. The Journal of Physical Chemistry B 105 (39), 9663 9668. ( 64 ) Tokuda, H., Tsuzuki, S., Susan, M. A. B. H., Hayamizu, K., & Watanabe, M. (2006). How ionic are room temperatu re ionic liquids? An indicator of the physicochemical properties. The Journal of Physical Chemistry B 110 (39), 19593 19600. ( 65 ) Jacquemin, J., Husson, P., Padua, A. A., & Majer, V. (2006). Density and viscosity of several pure and water saturated ionic liquids. Green Chemistry 8 (2), 172 180. ( 66 ) Almeida, H. F., Passos, H., Lopes da Silva, J. A., Fernandes, A. M., Freire, M. G., & Coutinho, J. A. (2012). Thermophysical properties of five acetate based ionic liquids. Journal of Chemical & Engineering Data 57 (11), 3005 3013. ( 67 ) Ghatee, M. H., Zare, M., Moosavi, F., & Zolghadr, A. R. (2010). Temperature dependent density and viscosity of the ionic liquids 1 alkyl 3 methylimidazolium iodides: experiment and molecular dynamics simulation. Journal of Chemical & Engineering Data 55 (9), 3084 3088. ( 68 ) ... & Coutinho, J. A. (2011). Thermophysical characterization of ionic liquids able to dissolve biomass. Journa l of Chemical & Engineering Data 56 (12), 4813 4822. ( 69 ) Schreiner, C., Zugmann, S., Hartl, R., & Gores, H. J. (2009). Fractional walden rule for ionic liquids: examples from recent measurements and a critique of the so called ideal KCl line for the Walde Journal of Chemical & Engineering Data 55 (5), 1784 1788.

PAGE 202

188 ( 70 ) Deive, F. J., Rivas, M. A., & Rodrguez, A. (2011). Thermophysical properties of two ionic liquids based on benzyl imidazolium cation. The Journal of Chemical Thermodynamics 43 (3), 4 87 491. ( 71 ) Restolho, J., Serro, A. P., Mata, J. L., & Saramago, B. (2009). Viscosity and surface tension of 1 ethanol 3 methylimidazolium tetrafluoroborate and 1 methyl 3 octylimidazolium tetrafluoroborate over a wide temperature range. Journal of Chemic al & Engineering Data 54 (3), 950 955. ( 72 ) Neves, C. M., Carvalho, P. J., Freire, M. G., & Coutinho, J. A. (2011). Thermophysical properties of pure and water saturated tetradecyltrihexylphosphonium based ionic liquids. The Journal of Chemical Thermodynamics 43 (6), 948 957. ( 73 ) Crosthwaite, J. M., Muldoon, M. J., Dixon, J. K., Anderson, J. L., & Brennecke, J. F. (2005). Phase transition and decomposition temperatures, heat capacities and viscosities of pyridinium ionic liquids. The Journal of Chemical Thermodynamics 37 (6), 559 568. ( 7 4) Sanchez, L. G., Espel, J. R., Onink, F., Meindersma, G. W., & Haan, A. B. D. (2009). Density, viscosity, and surface tension of synthesis grade imidazolium, pyridinium, and pyrrolidinium based room temperatu re ionic liquids. Journal of Chemical & Engineering Data 54 (10), 2803 2812. ( 75 ) Influence of molecular structure on densities and viscosities of several ionic liquids. Jour nal of Chemical & Engineering Data 56 (12), 4984 4999. ( 76 ) Zhang, S., Sun, N., He, X., Lu, X., & Zhang, X. (2006). Physical properties of ionic liquids: database and evaluation. Journal of physical and chemical reference data 35 (4), 1475 1517. ( 77 ) Ionic Liquid Database (IL Thermo). NIST Standard reference database # 147 ( http://ILThermo.boulder.nist.gov/ILThermo/mainmenu.uix ). Last accessed data of ( 78 ) Holbrey, J. D., & Seddon, K. R. (1999). Ionic liquids. Clean Products and Processes 1 (4), 223 236. ( 79 ) Rogers, R. D., & Seddon, K. R. (2003). Ionic liquids -solvents of the future?. Science 302 (5646), 792 793. (8 0 ) Ohno H. Electrochemical Aspects of Ionic Liquids. (2005) Hoboke n: John Wiley & Sons Inc.

PAGE 203

189 ( 81 ) Wood, N., & Stephens, G. (2010). Accelerating the discovery of biocompatible ionic liquids. Phys. Chem. Chem. Phys. 12 (8), 1670 1674. ( 82 ) Yan, T., Burnham, C. J., Del Ppolo, M. G., & Voth, G. A. (2004). Molecular dynamics simulation of ionic liquids: The effect of electronic polarizability. The Journal of Physical Chemistry B 108 (32), 11877 11881. ( 83 ) Cadena, C., & Maginn, E. J. (2006). Molecular simulation study of some thermophysical and transport properties of triazolium based ionic liquids. The Journal of Physical Chemistry B 110 (36), 18026 18039. ( 84 ) Kroon, M. C., Buijs, W., Peters, C. J., & Witkamp, G. J. (2007). Quantum chemical aided pre diction of the thermal decomposition mechanisms and temperatures of ionic liquids. Thermochimica Acta 465 (1), 40 47. ( 85 ) Vatamanu, J., Borodin, O., & Smith, G. D. (2010). Molecular insights into the potential and temperature dependences of the different ial capacitance of a room temperature ionic liquid at graphite electrodes. Journal of the American Chemical Society 132 (42), 14825 14833. ( 86 ) Achenie, L., Venkatasubramanian, V., & Gani, R. (Eds.). (2002). Computer aided molecular design: theory and practice (Vol. 12). Elsevier. ( 87 ) Gani, R., & Brignole, E. A. (1983). Molecular design of solvents for liquid extraction based on UNIFAC. Fluid Phase Equilibria 13 331 340. ( 88 ) Buxton, A., Livingston, A. G., & Pistikopoulos, E. N. (1999). Optimal des ign of solvent blends for environmental impact minimization. AIChE Journal 45 (4), 817 843. ( 89 ) Sinha, M., Achenie, L. E., & Gani, R. (2003). Blanket wash solvent blend design using interval analysis. Industrial & engineering chemistry research 42 (3), 516 527. ( 90 ) Chavali, S., Lin, B., Miller, D. C., & Camarda, K. V. (2004). Environmentally benign transition metal catalyst design using optimization techniques. Computers & chemical engineering 28 (5), 605 611. ( 9 1) Karunanithi, A. T., Achenie, L. E., & Gani, R. (2006). A computer aided molecular design framework for crystallization solvent design. Chemical Engineering Science 61 (4), 1247 1260. ( 92 ) Chemmangattuvalappil, N. G., Eljack, F. T., Solvason, C. C., & Eden, M. R. (2009). A novel algorithm fo r molecular synthesis using enhanced property operators. Computers & Chemical Engineering 33 (3), 636 643.

PAGE 204

190 ( 93 ) Harper, P. M., Gani, R., Kolar, P., & Ishikawa, T. (1999). Computer aided molecular design with combined molecular modeling and group contribution. Fluid Phase Equilibria 158 337 347. ( 94 ) Samudra, A. P., & Sahinidis, N. V. (2013). Optimization based framework for computer aided molecular design. AIChE Journal 59 (10), 3686 3701. ( 95 ) Sheldon, T. J., Folic, M., & Adjiman, C. S. (2006). Solvent design using a quantum mechanical continuum solvation model. Industrial & engineering chemistry research 45 (3), 1128 1140. ( 96 ) Odele, O., & Macchietto, S. (1993). Computer aided molecular design: a novel method for optimal solvent selection. Fluid Phase Equilibria 82 47 54. ( 97 ) Gardas, R. L., & Coutinho, J. A. (2008). A group contribution method for viscosity estimation of ionic liquids. F luid Phase Equilibria 266 (1), 195 201. ( 98 ) group contribution and artificial neural networks. Industrial & Engineering Chemistry Research 48 (6), 3254 3259. ( 99 ) Qiao, Y., Ma, Y., Huo, Y., Ma, P., & Xia, S. (2010). A group contribution method to estimate the densities of ionic liquids. The Journal of Chemical Thermodynamics 42 (7), 852 855. ( 100 ) Ye, C., & Shreeve, J. N. M. (2007). Rapid and accurate estimation of densities of room temperature ionic liquids and salts. The Journal of Physical Chemistry A 111 (8), 1456 1461. ( 101 ) Gardas, R. L., & Coutinho, J. A. (2008). Extension of the Ye and Shreeve group contribution method for density estimation of ionic liquid s in a wide range of temperatures and pressures. Fluid Phase Equilibria 263 (1), 26 32. ( 102 ) Huo, Y., Xia, S., Zhang, Y., & Ma, P. (2009). Group contribution method for predicting melting points of imidazolium and benzimidazolium ionic liquids. Industrial & engineering chemistry research 48 (4), 2212 2217. ( 10 3) Gardas, R. L., & Coutinho, J. A. (2009). Group contribution methods for the prediction of thermophysical and transport properties of ionic liquids. AIChE Journal 55 (5), 1274 1290. ( 104 ) Gardas, R. L., & Coutinho, J. A. (2008). A group contribution method for heat capacity estimation of ionic liquids. Industrial & Engineering Chemistry Research 47 (15), 5751 5757.

PAGE 205

191 ( 105 ) Roughton, B. C., Christian, B., White, J., Camarda, K. V., & Gani, R. (2012). Simultaneous design of ionic liquid entrainers and energy efficient azeotropic separation processes. Computers & Chemical Engineering 42 248 262. ( 106 ) Luis, P., Ortiz, I., Aldaco, R., & Irabien, A. (2007). A novel group contribution method in the development of a QSAR for predicting the toxicity (Vibrio fischeri EC 50) of ionic liquids. Ecotoxicology and environmental safety 67 (3), 423 429. ( 107 ) Fredenslund, A., Jones, R. L., & Prausnitz, J. M. (1975). Group contribution estimation of activity coefficients in nonideal liquid mixtures. AIChE Journal 21 (6), 1086 1099. ( 108 ) Wang, J., Sun, W., Li, C., & Wang, Z. (2008). Correlation of infinite dilution activity coefficient of solute in ionic liquid using UNIFAC model. Fluid Phase Equilibria 264 (1), 235 241. ( 109 ) Lei, Z., Zhang, J., Li, Q., & Chen, B. (2009). UNIFAC model for ionic liquids. Industrial & Engineering Chemistry Research 48 (5), 2697 2704. ( 110 ) Nebig, S., & Gmehling, J. (2011). Prediction of phase equilibria and excess properties for systems with ionic liquids using modified UNIFAC: Typical results and present status of the modified UNIFAC matrix for ionic liquids. Fluid Phase Equilibria 302 (1), 220 225. ( 11 1 ) Sinha, M., Achenie, L. E., & Ostrovsky, G. M. (1999). Environmentally benign solvent design by global optimization. Computers & Chemical Engineering 23 (10), 1381 1394. ( 112 ) Sahinidis, N. V., & Tawarmalani, M. (2000). Applications of global optimization to process and molecular design. Computers & Chemical Engineering 24 (9 ), 2157 2169. ( 113 ) Marcoulaki, E. C., & Kokossis, A. C. (2000). On the development of novel chemicals using a systematic synthesis approach. Part I. Optimisation framework. Chemical Engineering Science 55 (13), 2529 2546. ( 114 ) Xu, W., & Diwekar, U. M. (2005). Improved genetic algorithms for deterministic optimization and optimization under uncertainty. Part II. Solvent selection under uncertainty. Industrial & engineering chemistry research 44 (18), 7138 7146. ( 115 ) Karunanithi, A. T., Achenie, L. E., & Gani, R. (2005). A new decomposition based computer aided molecular/mixture design methodology for the design of optimal solvents and solvent mixtures. Industrial & engineering chemistry research 44 (13), 4785 4797. ( 116 ) Achenie, L. E. K., Gani, R., & Venkatasubramanian, V. (2002). Genetic algorithms based CAMD. Computer Aided Molecular Design: Theory and Practice 12 95.

PAGE 206

192 ( 117 ) Poling, B. E., Prausnitz, J. M., & O'connell, J. P. (2001). The properties of gases and liquids (Vol. 5). New York: McGraw Hill. ( 118 ) Aguirre, C. L., Cisternas, L. A., & Valderrama, J. O. (2012). Melting point estimation of ionic liquids by a group contribution method. International Journal of Thermophysics 33 (1), 34 46. ( 119 ) Widegren, J. A., Saurer, E. M., Marsh, K. N., & Magee, J. W. (2005). Electrolytic conductivity of four imidazolium based room temperature ionic liquids and the effect of a water impurity. The Journal of Chemical Thermodynamics 37 (6), 569 575. ( 120 ) Nishida, T., Tashiro, Y., & Yamamoto, M. (2003). Physical and electrochemical properties of 1 alkyl 3 methylimidazolium tetrafluoroborate for electrolyte. Journal of Fluorine Chemistry 120 (2), 135 141. ( 121 ) Tokuda, H., Tsuzuki, S., Susan, M. A. B. H., Hayamizu, K., & Watanabe, M. (2006). How ionic are room temperature ionic liquids? An indicator of the physicochemical properties. The Journal of Physical Chemistry B 110 (39), 19593 19600. ( 122 ) Tomida, D., Kenmochi, S., Tsukada, T., Qiao, K., & Yokoyama, C. (2007). Thermal conductivities of [b mim][PF6],[hmim][PF6], and [omim][PF6] from 294 to 335 K at pressures up to 20 MPa. International Journal of Thermophysics 28 (4), 1147 1160. ( 123 ) Van Valkenburg, M. E., Vaughn, R. L., Williams, M., & Wilkes, J. S. (2005). Thermochemistry of ionic liquid heat transfer fluids. Thermochimica Acta 425 (1), 181 188. ( 124 ) Ge, R., Hardacre, C., Nancarrow, P., & Rooney, D. W. (2007). Thermal conductivities of ionic liquids over the temperature range from 293 K to 353 K. Journal of Chemical & Engineering Data 52 (5), 1819 1823. ( 125 ) Meindersma, G. W., Podt, A. J., & de Haan, A. B. (2005). Selection of ionic liquids for the extraction of aromatic hydrocarbons from aromatic/aliphatic mixtures. Fuel Processing Technology 87 (1), 59 70. ( 126 ) Gmehling, J. G., Anderson, T. F., & Prausnitz, J. M. (1978). Solid liquid equilibria using UNIFAC. Industrial & Engineering Chemistry Fundamentals 17 (4), 269 273. ( 127 ) Anastas, P. T., & Warner, J. C. (2000). Green chemistry: theory and practice Oxford university press. ( 128 ) Jessop, P. G. (2011). Searching for green solvents. Green Chemistry 13 (6), 1391 1398.

PAGE 207

193 (129) MacFarlane, D. R., Tachikawa, N., Forsyth, M., Pringle, J. M., Howlett, P. C., Elliott, C. D., Davis, J. H., Watanabe, M., Simon, P., & Angell, C. A. (2014). Energy applications of ionic liquids. Energy & Environmental Science 7 232 250. (130) Moens, L., Blake, D. M., Rudnicki, D. L., & Hale, M. J. (2003). Advanced thermal storage fluids for solar parabolic trough systems. Journal of solar energy e ngineering 125 (1), 112 116. (131) Armand, M., Endres, F., MacFarlane, D. R., Ohno, H., & Scrosati, B. (2009). Ionic liquid materials for the electrochemical challenges of the future. Nature materials 8 (8), 621 629. (132) Paulechka, Y. U. (2010). Heat capacity of room temperature ionic liquids: a critical review. Journal of Physical and Chemical Reference Data 39 (3), 033108 1 033108 23. (133) Gani, R., & Brignole, E. A. (1983). Molecular design of solvents for liquid extraction based on UNIFAC. Fluid Phase Equilibria 13 331 340. (134) Camarda, K. V., & Maranas, C. D. (1999). Optimization in polymer design using connectivity indices. Industrial & Engin eering Chemistry Research 38 (5), 1884 1892. (135) Achenie, L., Venkatasubramanian, V., & Gani, R. (Eds.). (2003). Computer aided molecular design: theory and practice 12 Elsevier. (136) Karunanithi, A. T., Achenie, L. E., & Gani, R. (2006). A comput er aided molecular design framework for crystallization solvent design. Chemical Engineering Science 61 (4), 1247 1260. (137) for optimal reaction rate constants. AIChE journal 53 (5), 1240 1256. (138) Samudra, A. P., & Sahinidis, N. V. (2013). Optimization based framework for computer aided molecular design. AIChE Journal 59 (10), 3686 3701. (139) Stavrou, M., Lampe, M., Bardow, A., & Gross, J. (2014). Continuous molecular t argeting computer aided molecular design (CoMT CAMD) for simultaneous process and solvent design for CO2 capture. Industrial & Engineering Chemistry Research 53 (46), 18029 18041. (140) McLeese, S. E., Eslick, J. C., Hoffmann, N. J., Scurto, A. M., & Cam arda, K. V. (2010). Design of ionic liquids via computational molecular design. Computers & chemical engineering 34 (9), 1476 1480.

PAGE 208

194 (141) Roughton, B. C., Christian, B., White, J., Camarda, K. V., & Gani, R. (2012). Simultaneous design of ionic liquid entrainers and energy efficient azeotropic separation processes. Computers & Chemical Engineering 42 248 262. (142) Karunanithi, A. T., & Mehrkesh, A. (2013). Computer aided design of tailor made io nic liquids. AIChE Journal 59 (12), 4627 4640. (143) Chong, F. K., Foo, D. C., Eljack, F. T., Atilhan, M., & Chemmangattuvalappil, N. G. (2014). Ionic liquid design for enhanced carbon dioxide capture by computer aided molecular design approach. Clean Technologies and Environmental Policy 1 12. (144) Hada, S., Herring, R. H., Davis, S. E., & Eden, M. R. (2015). Multivariate characterization, modeling, and design of ionic liquid molecules. Computers & Chemical Engineering (145) Valderrama, J. O., & Zarricueta, K. (2009). A simple and generalized model for predicting the density of ionic liquids. Fluid Phase Equilibria 275 (2), 145 151. (146) Valderrama, J. O., & Rojas, R. E. (2009). Critical properties of ionic liquids. Revisited. Industrial & Engineering Chemistry Research 48 (14), 6890 6900. (147) Valderrama, J. O., Martinez, G., & Rojas, R. E. (2011). Predictive model for the heat capacity of ionic liquids using the mass connectivity index. Thermochimica Acta 513 (1), 83 87. (148) a)Lazz s, J. A. (2012). A group contribution method to predict the thermal decomposition temperature of ionic liquids. Journal of Molecular Liquids 168 87 93. b)Lazzs, J. A. (2012). A group contribution method to predict the melting point of ionic liquids. Flu id Phase Equilibria 313 1 6. (149) Harper, P. M., & Gani, R. (2000). A multi step and multi level approach for computer aided molecular design. Computers & Chemical Engineering 24 (2), 677 683. (150) Karunanithi, A. T., Achenie, L. E., & Gani, R. (2 005). A new decomposition based computer aided molecular/mixture design methodology for the design of optimal solvents and solvent mixtures. Industrial & engineering chemistry research 44 (13), 4785 4797. (151) Pistikopoulos, E. N., & Stefanis, S. K. (19 98). Optimal solvent design for environmental impact minimization. Computers & Chemical Engineering 22 (6), 717 733.

PAGE 209

195 (152) Venkatasubramanian, V., Chan, K., & Caruthers, J. M. (1994). Computer aided molecular design using genetic algorithms. Computers & Chemical Engineering 18 (9), 833 844. (153) Marcoulaki, E. C., & Kokossis, A. C. (1999). Stochastic screening and scoping of separation sequences using detailed simulation models. Computers & Chemical Engineering 23 S97 S100. (154) Gebreslassie, B. H., & Diwekar, U. M. (2015). Efficient ant colony optimization for computer aided molecular design: Case study solvent selection problem. Computers & Chemical Engineering 78 1 9. (155) Eljack, F. T., Abdelhady, A. F., Eden, M. R., Gabriel, F. B., Qin, X., & El Halwagi, M. M. (2005). Targeting optimum resource allocation using reverse problem formulations and property clustering techniques. Computers & chemical engineering 29 (11), 2304 2317. (156) Chemmangattuvalappil, N. G., Solvason, C. C., Bommareddy, S., & Eden, M. R. (2010). Reverse problem formulation approach to molecular design using property operators based on signature descriptors. Computers & Chemical Engineering 34 (1 2), 2062 2071. (157) Patkar PR, Venkatasubramanian V. Genetic Algorithms Based CAMD. In: Achenie LEK, Gani R, Venkatasubramanian V, (2003). Editor. Computer Aided Molecular Design: Theory and Practice. Amsterdam: Elsevier Science. ( 158 ) Anastas, P., & Eghbali, N. (2010). Green chemistry: principles and practice. Chemical Society Reviews 39 (1), 301 312. (1 59 ) Gupta, G. D., Misra, A., & Agarwal, D. K. (1991). Inhalation toxicity of furfural vapours: an assessment of biochemical response in rat lungs. Jou rnal of Applied Toxicology 11 (5), 343 347. (1 60 ) Arts, J. H., Muijser, H., Appel, M. J., Kuper, C. F., Bessems, J. G., & Woutersen, R. A. (2004). Subacute (28 day) toxicity of furfural in Fischer 344 rats: a comparison of the oral and inhalation route. F ood and chemical toxicology 42 (9), 1389 1399. (1 61 ) Gordon, C. J., Dyer, R. S., Long, M. D., & Fehlner, K. S. (1985). Effect of sulfolane on behavioral and autonomic thermoregulation in the rat. Journal of Toxicology and Environmental Health, Part A Curr ent Issues 16 (3 4), 461 468. (16 2 ) Tilstam, U. (2012). Sulfolane: A versatile dipolar aprotic solvent. Organic Process Research & Development 16 (7), 1273 1278.

PAGE 210

196 (1 63 ) Garca, S., Larriba, M., Garca, J., Torrecilla, J. S., & Rodrguez, F. (2012). Separation of toluene from n heptane by liquid liquid extraction using binary mixtures of [bpy][BF 4] and [4bmpy][Tf 2 N] ionic liquids as solvent. The Journal of Chemical Thermodynamics 53 119 124. (1 64 ) Meindersma, G. W., & de Haan, A. B. (2008). Conc eptual process design for aromatic/aliphatic separation with ionic liquids. chemical engineering research and design 86 (7), 745 752. (1 65 ) Ravilla, U. K., & Banerjee, T. (2012). Liquid liquid equilibria of imidazolium based ionic liquid+ pyridine+ hydroc arbon at 298.15 K: Experiments and correlations. Fluid Phase Equilibria 324 17 27. ( 166 ) Karunanithi, A. T., & Mehrkesh, A. (2013). Computer aided design of tailor made ionic liquids. AIChE Journal 59 (12), 4627 4640. ( 167 ) Zhang, S., Sun, N., He, X., Lu, X., & Zhang, X. (2006). Physical properties of ionic liquids: database and evaluation. Journal of physical and chemical reference data 35 (4), 1475 1517. ( 168 ) Ionic Liquid Database (IL Thermo). NIST standard reference database #147. Available at: http://ILThermo.boulder.nist.gov/ILThermo/mainmenu.uix Last accessed data: 05/10/2013. ( 169 ) Holbrey, J. D., & Seddon, K. R. (1999). Ionic liquids. Clean Products and Processes 1 (4), 223 236. ( 170 ) Rogers, R. D., & Seddon, K. R. (2003). Ionic liquids -solvents of the future?. Science 302 (5646), 792 793. ( 171 ) Wood, N., & Stephens, G. (2010). Accelerating the discovery of biocompatible ionic liquids. Phys. Chem. Chem. Phys. 12 (8), 1670 1674. ( 172 ) Achenie LEK, Gani R, Venkatasubramanian V. (2003) Computer Aided Molecular Design: Theory and Practice. Amsterdam: Elsevier Science ( 173 ) Gani, R., & Brignole, E. A. (1983). Molecular design of solvents for liquid extraction based on UNIFAC. Fluid Phase Equilibria 13 331 340. ( 174 ) Buxton, A., Livingston, A. G., & Pistikopoulos, E. N. (1999). Optimal design of solvent blends for environmental impact minimization. AIChE Journal 45 (4), 817 843.

PAGE 211

197 ( 175 ) Sinha, M., Achenie, L. E., & Gani, R. (2003). Blanket wash solvent blend design using interval analysis. Industrial & engineering chemistry research 42 (3), 516 527. ( 176 ) Chavali, S., Lin, B., Miller, D. C., & Camarda, K. V. (2004). Environmentally benign transition metal catalyst design using optimization techniques. Computers & chemical engineering 28 (5), 605 611. ( 177 ) Karunanithi, A. T., Achenie, L. E., & Gani, R. (2006). A com puter aided molecular design framework for crystallization solvent design. Chemical Engineering Science 61 (4), 1247 1260. ( 178 ) Chemmangattuvalappil, N. G., Eljack, F. T., Solvason, C. C., & Eden, M. R. (2009). A novel algorithm for molecular synthesis using enhanced property operators. Computers & Chemical Engineering 33 (3), 636 643. ( 179 ) Patkar PR, Venkatasubramanian V. (2003) Genetic Algorithms Based CAMD. In: Achenie LEK, Gani R, Venkatasubramanian V, editor. Computer Aided Molecular Design: Theory and Practice. Amsterdam: Elsevier Science ( 180 ) Fredenslund, A., Jones, R. L., & Prausnitz, J. M. (1975). Group contribution estimation of activity coefficients in nonideal liquid mixtures. AIChE Journal 21 (6), 1086 1099. ( 181 ) Wang, J., Sun, W., Li, C., & Wang, Z. (2008). Correlation of infinite dilution activity coefficient of solute in ionic liquid using UNIFAC model. Fluid Phase Equilibria 264 (1), 235 241. ( 182 ) Lei, Z., Zhang, J., Li, Q., & Chen, B. (2009). UNIFAC model for ionic liquids. In dustrial & Engineering Chemistry Research 48 (5), 2697 2704. ( 183 ) Nebig, S., & Gmehling, J. (2011). Prediction of phase equilibria and excess properties for systems with ionic liquids using modified UNIFAC: Typical results and present status of the modif ied UNIFAC matrix for ionic liquids. Fluid Phase Equilibria 302 (1), 220 225. ( 184 ) Roughton, B. C., Christian, B., White, J., Camarda, K. V., & Gani, R. (2012). Simultaneous design of ionic liquid entrainers and energy efficient azeotropic separation processes. Computers & Chemical Engineering 42 248 262. ( 185 ) based tetrafluoroborates and toluene. Journal of Chemical & Engineering Data 55 (9), 3051 3054. ( 186 ) Co rder, S., Gonzlez, E. J., Calvar, N., & Domnguez, (2012). Application of [HMim][NTf 2],[HMim][TfO] and [BMim][TfO] ionic liquids on the extraction of toluene

PAGE 212

198 from alkanes: Effect of the anion and the alkyl chain length of the cation on the LLE. The Journal of Chemical Thermodynamics 53 60 66. ( 187 ) Gonzlez, E. J., Requejo, P. F., Domnguez, ., & Macedo, E. A. (2013). Phase equilibria of binary mixtures (ionic liquid+ aromatic hydrocarbon): Effect of the structure of the components on the solubil ity. Fluid Phase Equilibria 360 416 422. ( 188 ) liquid extraction of toluene from heptane using 1 alkyl 3 methylimidazolium bis (trifluoromethylsulfonyl) imide ioni c liquids. Journal of Chemical & Engineering Data 56 (1), 113 118. ( 189 ) pentyl 3 methylimidazolium bis (trifluoromethylsulfonyl) imide for desulfurization, denitrification and dearomatizati on of FCC gasoline. The Journal of Chemical Thermodynamics 76 1 15. ( 190 ) Meindersma, G. W., Podt, A. J., & de Haan, A. B. (2006). Ternary liquid liquid equilibria for mixtures of toluene+n heptane+ an ionic liquid. Fluid phase equilibria 247 (1), 158 168. ( 19 1 ) Gonzlez, B., Corder, S., & Santamara, A. G. (2013). Application of 1 alkyl 3 methylpyridinium bis (trifluoromethylsulfonyl) imide ionic liquids for the ethanol removal from its mixtures with alkanes. The Journal of Chemical Thermodynamic s 60 9 14. ( 19 2 ) Naser, S. F., & Fournier, R. L. (1991). A system for the design of an optimum liquid liquid extractant molecule. Computers & chemical engineering 15 (6), 397 414. ( 19 3 ) Mehrkesh, A.; Karunaithi, A.T. (2015) A Quantum Chemistry based Correlative Approach for Prediction of the Melting point and Viscosity of Ionic Liquids, ready for submission ( 194 ) Metz, B., Davidson, O., de Coninck, H., Loos, M., & Meyer, L. (2005). Carbon dioxide capture and storage. ( 195 ) Yu, C. H., Huang, C. H., & Tan, C. S. (2012). A review of CO2 capture by absorption and adsorption. Aerosol and Air Quality Research 12 (5), 745 769. ( 196 ) Rochelle, G. T. (2009). Amine scrubbing for CO2 capture. Science 325 (5948), 1652 1654. ( 197) Rao, A. B., & Rubin, E. S. (2002). A technical, economic, and environmental assessment of amine based CO2 capture technology for power plant greenhouse gas control. Environmental science & technology 36 (20), 4467 4475.

PAGE 213

199 ( 198 ) Zhang, X., Zhang, X., Dong, H., Zhao, Z., Zhang, S., & H uang, Y. (2012). Carbon capture with ionic liquids: overview and progress. Energy & Environmental Science 5 (5), 6668 6681. ( 199 ) Anastas, P., & Eghbali, N. (2010). Green chemistry: principles and practice. Chemical Society Reviews 39 (1), 301 312. ( 200 ) Karunanithi, A. T., & Mehrkesh, A. (2013). Computer aided design of tailor made ionic liquids. AIChE Journal 59 (12), 4627 4640. ( 201 ) Zhang, S., Sun, N., He, X., Lu, X., & Zhang, X. (2006). Physical properties of ionic liquids: database and evaluation. J ournal of physical and chemical reference data 35 (4), 1475 1517. ( 202 ) Ionic Liquid Database (IL Thermo). NIST standard reference database #147. Available at: http://ILThermo.boulder.nist.gov/ILThermo/mainmenu.uix Last accessed data: 05/10/2013. ( 203 ) Holbrey, J. D., & Seddon, K. R. (1999). Ionic liquids. Clean Products and Processes 1 (4), 223 236. ( 204 ) Rogers, R. D., & Seddon, K. R. (2003). Ionic liquids -solvents of the future?. Science 302 (5646), 792 793. ( 205 ) Achenie LEK, Gani R, Venkatasubramanian V. (2003) Computer Aided Molecular Design: Theory and Practice. Amsterdam: Elsevier Science. ( 206 ) Gani, R., & Brignole, E. A. (1983). Molecular design of solvents for liquid extraction based on UNIFAC. Fluid Phase Equilibria 13 331 340. ( 207 ) Buxton, A., Livingston, A. G., & Pistikopoulos, E. N. (1999). Optimal design of solvent blends for environmental impact minimization. AIChE Journal 45 (4), 817 843. ( 208 ) Sinha, M., Achenie, L. E., & Gani, R. (2003). Blanket wash solvent b lend design using interval analysis. Industrial & engineering chemistry research 42 (3), 516 527. ( 209 ) Chavali, S., Lin, B., Miller, D. C., & Camarda, K. V. (2004). Environmentally benign transition metal catalyst design using optimization techniques. Com puters & chemical engineering 28 (5), 605 611. ( 210 ) Karunanithi, A. T., Achenie, L. E., & Gani, R. (2006). A computer aided molecular design framework for crystallization solvent design. Chemical Engineering Science 61 (4), 1247 1260.

PAGE 214

200 ( 211 ) Chemmangattuvalappil, N. G., Eljack, F. T., Solvason, C. C., & Eden, M. R. (2009). A novel algorithm for molecular synthesis using enhanced property operators. Computers & Chemical Engineering 33 (3), 636 643. ( 212 ) Patkar PR, Venkatasubramanian V. (2003) Genetic Algorithms Based CAMD. In: Achenie LEK, Gani R, Venkatasubramanian V, editor. Computer Aided Molecular Design: Theory and Practice. Amsterdam: Elsevier Science. ( 213 ) Klamt, A., Eckert, F., & Arlt, W. (2010). COSMO RS: an alternative to simulatio n for calculating thermodynamic properties of liquid mixtures. Annual review of chemical and biomolecular engineering 1 101 122. ( 214 ) Eckert F, Klamt A. (2013) "COSMOtherm Version C3.0, Release 13.01. COSMOlogic GmbH & Co. KG, Leverkusen, Germany" Available at www.cosmologic.de. ( 215 ) Palomar, J., Gonzalez Miquel, M., Polo, A., & Rodriguez, F. (2011). Understanding the physical absorption of CO2 in ionic liquids using the COSMO RS method. Industrial & Engineering Chemistry Research 50 (6), 3452 34 63. (216) Anthony, J. L., Anderson, J. L., Maginn, E. J., & Brennecke, J. F. (2005). Anion effects on gas solubility in ionic liquids. The Journal of Physical Chemistry B 109 (13), 6366 6374. (217) Afzal, W., Liu, X., & Prausnitz, J. M. (2013). Solubilities of some gases in four immidazolium based ionic liquids. The Journal of Chemical Thermodynamics 63 88 94. (218) Deng, Y., Husson, P., Delort, A. M., Besse Hoggan, P., Sancelme, M., & Costa Gomes, M. F. (2011). Influence of an oxygen functiona lization on the physicochemical properties of ionic liquids: density, viscosity, and carbon dioxide solubility as a function of temperature. Journal of Chemical & Engineering Data 56 (11), 4194 4202. (219) Jacquemin, J., Husson, P., Majer, V., & Gomes, M. F. C. (2007). Influence of the cation on the solubility of CO2 and H2 in ionic liquids based on the bis (trifluoromethylsulfonyl) imide anion. Journal of solution chemistry 36 (8), 967 979. (220) Anthony, J. L., Anderson, J. L., Maginn, E. J., & Brenneck e, J. F. (2005). Anion effects on gas solubility in ionic liquids. The Journal of Physical Chemistry B 109 (13), 6366 6374. (221) Cadena, C., Anthony, J. L., Shah, J. K., Morrow, T. I., Brennecke, J. F., & Maginn, E. J. (2004). Why is CO2 so soluble in im idazolium based ionic liquids?. Journal of the American Chemical Society 126 (16), 5300 5308. ( 222 ) Mehrkesh A & Karunanithi A T. (2015) A Quantum Chemistry based Correlative Approach for Prediction of the Melting point and Viscosity of Ionic Liquids, ready for submission.

PAGE 215

201 (223) Rogers R D ., & Seddon K R. (2003). Ionic liquids -solvents of the future? Science 302 (5646), 792 793. (22 4 ) Zhang, Y., Bakshi, B. R., & Demessie, E. S. (2008). Life cycle assessment of an ionic liquid versus molecular solvents and their applications. Environmental science & technology 42 (5), 1724 1730. (22 5 ) Righi, S., Morfino, A., Galletti, P., Samor, C., Tugnoli, A., & Stramigioli, C. (2011). Comparative cradle to gate life cycle assessments of cellulose dissolution with 1 butyl 3 methylimidazolium chloride and N methyl morpholine N oxide. Green Chemistry 13 (2), 367 375. (22 6 ) Imperato, G., Koenig, B., & Chiappe, C. (2007). Ionic green solvents from renewable resources. European journal of organic chemistry 2007 (7), 1049 1058. (22 7 ) Pham, T. P. T., Cho, C. W., & Yun, Y. S. (2010). Environmental fate and toxicity of ionic liquids: a review. Water research 44 (2), 352 372. (22 8 ) Ventura, S. P., Marques, C. S., Rosatella, A. A., Afonso, C. A., Gona lves, F., & Coutinho, J. A. (2012). Toxicity assessment of various ionic liquid families towards Vibrio fischeri marine bacteria. Ecotoxicology and environmental safety 76 162 168. (22 9 ) Bernot, R. J., Brueseke, M. A., Evans White, M. A., & Lamberti, G. A. (2005). Acute and chronic toxicity of imidazolium based ionic liquids on Daphnia magna. Environmental Toxicology and Chemistry 24 (1), 87 92. (2 30 ) Petkovic, M., Seddon, K. R., Rebelo, L. P. N., & Pereira, C. S. (2011). Ionic liquids: a pathway to env ironmental acceptability. Chemical Society Reviews 40 (3), 1383 1403. (23 1 ) Wells, A. S., & Coombe, V. T. (2006). On the freshwater ecotoxicity and biodegradation properties of some common ionic liquids. Organic Process Research & Development 10 (4), 794 7 98. (23 2 ) Matzke, M., Thiele, K., Mller, A., & Filser, J. (2009). Sorption and desorption of imidazolium based ionic liquids in different soil types. Chemosphere 74 (4), 568 574. (23 3 ) Mrozik, W., Jungnickel, C., Ciborowski, T., Pitner, W. R., Kumirska, & Stepnowski, P. (2009). Predicting mobility of alkylimidazolium ionic liquids in soils. Journal of Soils and Sediments 9 (3), 237 245.

PAGE 216

202 (23 4 ) Hu, S., Wang, A., Lwe, H., Li, X., Wang, Y., Li, C., & Yang, D. (2010). Kinetic study of ionic liquid synthesis in a microchannel reactor. Chemical Engineering Journal 162 (1), 350 354. (23 5 ) Wu, M., Wang, M., Liu, J., & Huo, H. (2007). Life cycle assessment of corn based butanol as a potential transportation fuel (No. ANL/ESD/07 10). Argonne National Laboratory (ANL). (23 6 ) National Renewable Energy Laboratory (NREL), US LCI Database Project. (2010). U.S. Department of Energy: Golden, CO; www.nrel.gov/lci/about.html (237) Chris Goemans, Athena Institute 2010 NREL US LCI Database N.A. Electricity Generation by Fuel Type Update & Template Methodology. (23 8 ) Rosenbaum, R. K., Bachmann, T. M., Gold, L. S., Huijbregts, M. A., Jolliet, O., Juraske, R., ... & Hauschild, M. Z. (2008). USEtox t he UNEP SETAC toxicity model: recommended characterisation factors for human toxicity and freshwater ecotoxicity in life cycle impact assessment. The International Journal of Life Cycle Assessment 13 (7), 532 546. (23 9 ) Garcia, M. T., Gathergood, N., & Sc ammells, P. J. (2005). Biodegradable ionic liquids Part II. Effect of the anion and toxicology. Green Chemistry 7 (1), 9 14. (2 40 ) Ranke, J., Mlter, K., Stock, F., Bottin Weber, U., Poczobutt, J., Hoffmann, J., ... & Jastorff, B. (2004). Biological effec ts of imidazolium ionic liquids with varying chain lengths in acute Vibrio fischeri and WST 1 cell viability assays. Ecotoxicology and environmental safety 58 (3), 396 404. (24 1 ) Docherty, K. M., & Kulpa Jr, C. F. (2005). Toxicity and antimicrobial activit y of imidazolium and pyridinium ionic liquids. Green Chemistry 7 (4), 185 189. (24 2 ) Cho, C. W., Pham, T. P. T., Jeon, Y. C., Vijayaraghavan, K., Choe, W. S., & Yun, Y. S. (2007). Toxicity of imidazolium salt with anion bromide to a phytoplankton Selenastrum capricornutum: Effect of alkyl chain length. Chemosphere 69 (6), 1003 1007. (24 3 ) Pham, T. P. T., Cho, C. W., Min, J., & Yun, Y. S. (2008). Alkyl chain length effects of imidazolium and pyridinium ionic liquids on photosynthetic response of Pseudokirchneriella subcapitata. Journal of bioscience and bioengineering 105 (4), 425 428. (24 4 ) Bernot, R. J., Kennedy, E. E., & Lamberti, G. A. (2005). Effects of ionic liquids on the survival, movement, and feeding behavior of the freshwater snail, Phy sa acuta. Environmental Toxicology and Chemistry 24 (7), 1759 1765.

PAGE 217

203 (24 5 ) Ropel, L., Belvze, L. S., Aki, S. N., Stadtherr, M. A., & Brennecke, J. F. (2005). Octanol water partition coefficients of imidazolium based ionic liquids. Green Chemistry 7 (2), 83 90. (24 6 ) Mrozik, W., Nichthauser, J., & Stepnowski, P. (2008). Prediction of the adsorption coefficients for imidazolium ionic liquids in soils using cyanopropyl stationary phase. Pol. J. Environ. Stud 17 383 388. (24 7 ) Centre for Environmental Resea rch and Sustainable Technology (UFT), http://www.il eco.uft.uni bremen.de (24 8 ) Eckelman, M. J., Mauter, M. S., Isaacs, J. A., & Elimelech, M. (2012). New perspectives on nanomaterial aquatic ecotoxicit y: production impacts exceed direct exposure impacts for carbon nanotoubes. Environmental science & technology 46 (5), 2902 2910. (24 9 ) Neves, C. M., Freire, M. G., & Coutinho, J. A. (2012). Improved recovery of ionic liquids from contaminated aqueous streams using aluminium based salts. RSC Advances 2 (29), 10882 10890. (2 50 ) Moser, P., Schmidt, S., & Stahl, K. (2011). Investigation of trace elements in the inlet and outlet streams of a MEA based post combustion capture process results from the test pr ogramme at the Niederaussem pilot plant. Energy Procedia 4 473 479. (25 1 ) Siedlecka, E. M., & Stepnowski, P. (2009). The effect of alkyl chain length on the degradation of alkylimidazolium and pyridinium type ionic liquids in a Fenton like system. Envir onmental Science and Pollution Research 16 (4), 453 458. (25 2 ) Awad, W. H., Gilman, J. W., Nyden, M., Harris, R. H., Sutto, T. E., Callahan, J., ... & Fox, D. M. (2004). Thermal degradation studies of alkyl imidazolium salts and their application in nanoc omposites. Thermochimica Acta 409 (1), 3 11. (25 3 ) Itakura, T., Hirata, K., Aoki, M., Sasai, R., Yoshida, H., & Itoh, H. (2009). Decomposition and removal of ionic liquid in aqueous solution by hydrothermal and photocatalytic treatment. Environmental Chem istry Letters 7 (4), 343 345. (25 4 ) Anthony, J. L., Maginn, E. J., & Brennecke, J. F. (2001). Solution thermodynamics of imidazolium based ionic liquids and water. The Journal of Physical Chemistry B 105 (44), 10942 10949. (25 5 ) Gathergood, N., Garcia, M T., & Scammells, P. J. (2004). Biodegradable ionic liquids: Part I. Concept, preliminary targets and evaluation. Green Chemistry 6 (3), 166 175. (25 6 ) Anastas, P. T., & Warner, J. C. (2000). Green chemistry: theory and practice Oxford university press.

PAGE 218

204 (25 7 ) Pretti, C., Chiappe, C., Pieraccini, D., Gregori, M., Abramo, F., Monni, G., & Intorre, L. (2006). Acute toxicity of ionic liquids to the zebrafish (Danio rerio). Green Chemistry 8 (3), 238 240. (25 8 (25 9 ) Consultant, PR. "SimaPro 7.3.3 LCA software", (2012). Amersfoort, Netherlands. (2 60 ) Mehrkesh, A., & Karunanithi, A. T. (2013). Energetic ionic materials: How green are they? A comparative life cycle assessment study. ACS Sustainable Chemi stry & Engineering 1 (4), 448 455. (26 1 ) Farahipour, R., & Karunanithi, A. T. (2014). Life Cycle Environmental Implications of CO2 Capture and Sequestration with Ionic Liquid 1 Butyl 3 methylimidazolium Acetate. ACS Sustainable Chemistry & Engineering 2 (11), 2495 2500. ( 2 6 2 ) Darke, G., Hawkins, T., Brand, A., Mckay, M., & Ismail, I. (2003). Energetic, Low Melting Salts of Simple Heterocycles (No. AFRL PR ED TP 2003 007). AIR FORCE RESEARCH LAB EDWARDS AFB CA SPACE AND MISSILE PROPULSION DIV. (2 6 3 ) Singh, R. P., Verma, R. D., Meshri, D. T., & Shreeve, J. N. M. (2006). Energetic Nitrogen Rich Salts and Ionic Liquids. Angewandte Chemie International Edition 45 (22), 3584 3601. ( 2 6 4 ) Gao, H., Joo, Y. H., Twamley, B., Zhou, Z., & Shreeve, J. N. M. (2009). Hypergolic Ionic Liquids with the 2, 2 Dialkyltriazanium Cation. Angewandte Chemie 121 (15), 2830 2833. ( 2 6 5 ) Schneider, S., Hawkins, T., Rosander, M., Vaghjiani, G., Chambreau, S., & Drake, G. (2008). Ionic liquids as hypergolic fuels. Energy & F uels 22 (4), 2871 2872. ( 2 6 6 ) (2006). Dissolution and transport of TNT, RDX, and Composition B in saturated soil columns. Journal of Environmental Quality 35 (6), 2043 2054. ( 2 6 7 ) Talmage, S. S., Opresko, D. M., Maxwell, C. J., Welsh, C. J., Cretella, F. M., Reno, P. H., & Daniel, F. B. (1999). Nitroaromatic munition compounds: environmental effects and screening values. In Reviews of environmental contamination and toxicology (pp. 1 156). Springer New York. ( 2 6 8 ) transport of RDX, HMX, TNT and DNT in the volcanic soils of Hawaii: A laboratory and modeling study. Journal of hazardous mate rials 185 (2), 1600 1604.

PAGE 219

205 ( 2 6 9 ) Clausen, J., Robb, J., Curry, D., & Korte, N. (2004). A case study of contaminants on military ranges: Camp Edwards, Massachusetts, USA. Environmental Pollution 129 (1), 13 21. ( 2 7 0 ) Nipper, M., Carr, R. S., Biedenbach, J. M., Hooten, R. L., & Miller, K. (2002). Toxicological and chemical assessment of ordnance compounds in marine sediments and porewaters. Marine pollution bulletin 44 (8), 789 806. ( 2 7 1 ) Pimienta, I. S. O.; Elzey, S.; Boatz, J. A.; Gordon, M. S. Pentazole Based Energetic J. Phys. Chem. A 2007, 111 691 703. ( 2 7 2 ) Schmidt, M. W., Gordon, M. S., & Boatz, J. A. (2005). Triazolium based energetic ionic liquids. The Journal of Physical Chemistry A 109 (32), 7285 7295. ( 2 7 3 ) Xue, H. O. N. G., & Shreeve, J. M. (2005). Energetic Ionic Liquids from Azido Derivatives of 1, 2, 4 Triazole. Advanced Materials 17 (17), 2142 2146. ( 2 7 4 ) Drake, G., Kaplan, G., Hall, L., Hawkins, T., & Larue, J. (2007). A new family of energetic ionic liquids 1 amino 3 alkyl 1, 2, 3 triazolium nitrates. Journal of Chemical Crystallography 37 (1), 15 23. ( 2 7 5 ) Drake, G. W., Hawkins, T. W., Boatz, J., Hall, L., & Vij, A. (2005). Experimental and Theoretical Study of 1, 5 Diamino 4 H 1, 2, 3, 4 Tetrazolium Perchlorate. Propellants, Explosives, Pyrotechnics 30 (2), 156 163. ( 2 7 6 ) Zorn, D. D., Boatz, J. A., & Gordon, M. S. (2006). Electronic structure studies of tetrazolium based ionic liquids. The Journal of Physical Chemistry B 110 (23), 11110 11119. ( 2 7 7 ) Gutowski, K. E., Holbrey, J. D., Rogers, R. D., & Dixon, D. A. (2005). Prediction of the formation and stabilities of energetic salts and ionic liquids based on ab initio electronic structure calculations. The Journal of Physical Chemistry B 109 (49), 23196 23208. ( 2 7 8 ) Si mple, nitrogen rich, energetic salts of 5 nitrotetrazole. Inorganic chemistry 47 (13), 6014 6027. ( 2 7 9 ) Darwich, C., Klaptke, T. M., Welch, J. M., & Suceska, M. (2007). Synthesis and Characterization of 3, 4, 5 Triamino 1, 2, 4 Triazolium 5 Nitrotetrazolate. Propellants, Explosives, Pyrotechnics 32 (3), 235 243. ( 2 8 0 ) Darwich, C., Klaptke, T. M., & Sabat, C. M. (2008). 1, 2, 4 Triazolium Cation Based Energetic Salts. Chemistry A European Journal 14 (19), 5756 5771.

PAGE 220

206 ( 2 8 1 ) Anastas, P. T., & Warner, J. C. (2000). Green chemistry: theory and practice Oxford university press. ( 2 8 2 ) Zhang, Y., Bakshi, B. R., & Demessie, E. S. (2008). Life cycle assessment of an ionic liquid versus molecular solvents and their applications. Environmental scienc e & technology 42 (5), 1724 1730. ( 2 8 3 ) Righi, S., Morfino, A., Galletti, P., Samor, C., Tugnoli, A., & Stramigioli, C. (2011). Comparative cradle to gate life cycle assessments of cellulose dissolution with 1 butyl 3 methylimidazolium chloride and N methyl morpholine N oxide. Green Chemistry 13 (2), 367 375. ( 2 8 4 ) Ott, D., Kralisch, D., & Stark, A. (2010) Perspectives of Ionic Liquids as Environmentally Benign Substitutes for Molecular Solvents. Handbook of Green Chemistry ( 2 8 5 ) IPPC BAT Reference Document Large Volume Solid Inorganic Chemicals Family Process BREF for Soda Ash; (2004) European Soda Ash Producers Association; EUROPEAN CHEMICAL INDUSTRY COUNCIL: Brussels, Belgium. ( 2 8 6 ) Atkins, P. (1997). Physical Chemistry 6 th edition; W.H. Freeman and Company: New York, U.S. ( 2 8 7 ) Greenlee, K. W., Henne, A. L., & Fernelius, W. C. (1946). Sodium amide. Inorganic Syntheses, Volume 2 128 135. ( 2 8 8 ) Jobelius, H. H., & Scharff, H. D. (2000). Hydrazoic acid and azides. Ullmann's Encyclopedia of Industrial Chemistry ( 2 8 9 ) Author, A. (1961). Proceedings of the Chemical Society. October 1961. Proceedings of the Chemical Society 1961 (October), 357 396. ( 290 ) Hudak, C. E., & Parsons, J. B. (1977). Industrial process profiles for environmental use: chapter 12: the explosives industry. In Industrial process profiles for environmental use: chapter 12: the explos ives industry NTIS. ( 291 ) National Renewable Energy Laboratory (NREL), (2010). US LCI Database Project; US Department of Energy: Golden; www.nrel.gov/lci/about.html (accessed 02/19/2013) ( 29 2 ) IEA Statistics and Balances ; (2008). International Energy Agency; www.iea.org/stats/index.asp (accessed 02/19/2013) ( 29 3 ) Bare, J. (2002). Tool for the Reduction and Assessment of Chemical and Other Environmental Impacts (TRACI); U.S. EPA: Cincinnati; www.epa.gov/nrmrl/std/traci/traci.html (accessed 02/19/2013 )

PAGE 221

207 ( 29 4 ) Eco indicator 99 Manual for Designers: A damage oriented method for Life Cycle Assessment, Ministry of Housing, Spatial Planning and the environment; PR Consultants B.V. : Amersfoort, Netherlands, 2000; www.pre sustainability.com/download/manuals/EI99_Manual .pdf (accessed 02/19/2013) (295) Anastas, P. T., & Warner, J. C. (2000). Green chemistry: theory and practice. Oxford university press. (296) Jessop, P. G. (2011). Searching for green solvents. Green Chemistry, 13(6), 1391 1398. (297) Gabriel, S., & Weiner, J. (1888). Ueber einige abkmmlinge des propylamins.Berichte der deutschen chemischen Gesellschaft, 21(2), 2669 2679. (298) Zhang, X., & Hu, D. (2011). Performance simulation of the absorption chiller using water and ionic liquid 1 ethyl 3 methylimidazolium dimethylphosphate as the working pair. Applied Thermal Engineering, 31(16), 3316 3321. (299) electrolytes. E lectrochimica Acta, 51(26), 5567 5580. (300) Ye, C., Liu, W., Chen, Y., & Yu, L. (2001). Room temperature ionic liquids: a novel versatile lubricant. Chemical Communications, (21), 2244 2245. (301) Merrigan, T. L., Bates, E. D., Dorman, S. C., & Davis Jr J. H. (2000). New fluorous ionic liquids function as surfactants in conventional room temperature ionic liquids. Chemical Communications, (20), 2051 2052. (302) Kosan, B., Michels, C., & Meister, F. (2008). Dissolution and forming of cellulose with ioni c liquids. Cellulose, 15(1), 59 66.

PAGE 222

208 Appendix A : A comprehensive list of IL structural groups used in CAILD Cation Structure Valence Anion Structure Groups Valence Imidazolium 1 2 3 4 5 CH 3 1 Pyridinium Values 1,2,3,4,5,6 CH 2 2 Pyrrolidinium Values 1,2,3,4,5,6 CH 3 Ammonium Values 1,2,3,4 Halogen C= 2

PAGE 223

209 Cation Structure Valence Anion Structure Groups Valence Phosphonium Values 1,2,3,4 Mesylate C 1 Piperidinium Values 1,2,3,4,5,6,7 Propionate OH 1 Triazolium Values 1,2,3,4,5 Benzoate ACH 2 Thiazolium Values 1,2,3,4 Acetate AC 3 Pyrazinium Values 1,2,3,4,5 Trifluoroacetate CH 3 C=O 1 Sulfonium Values 1,2,3 Phosphate CH 2 C=O 2 Toluenesulfonate HCO 1 Hydrosulfate CH 3 COO 1 Azide CH 2 COO 2

PAGE 224

210 Cation Structure Valence Anion Structure Groups Valence Perchlorate HCOO 1 CH 3 O 1 CH 2 O 2 CHO 3 CH 2 NH 2 1 CHNH 2 2 COOH 1 COO 1 CH 3 S 1 CH 2 S 2

PAGE 225

211 Appendix B : IL structural Groups used in CAILD for Aro/Ali separation Cation Structure Valence s Anion Structure Groups Valence Imidazolium 1,2 CH 2 N 2 Pyridinium 1,2 CH 3 N 1 Pyrrolidinium 1,2 CH 2 C 2 Ammonium 1,2,3,4 Acetate CH 3 C 1 Phosphonium 1,2,3,4 Dicyanamide ( CH 2 OH) 1 Piperidinium 1,2 Methyl sulfate ( OCH 2 ) 2 Sulfonium 1,2,3 Ethyl sulfate ( OCH 3 ) 1 Benzyl 1 Trifluoromethanesulfonate

PAGE 226

212 Appendix C : UNIFAC parameters Table C 1 : UNIFAC Group parameters Groups R k Q K CH 2 _N 13.8418 11.1728 CH 3 _N 0.6141 0.1000 CH 2 _C 3.2961 2.5364 CH 3 _C 9.9824 9.4787 Benzyl 40.5232 34.8127 Benzene 15.5701 14.0744 Toluene 22.8394 19.8815 ( OCH 2 ) 14.3225 12.0009 ( OCH 3 ) 21.9956 18.8325 CH 2 OH 7.3858 3.9473 Imidazolium (Im) 37.6794 25.3520 Pyridinium (Py) 27.7218 15.7289 Acetate 48.2941 58.7731 BF 4 13.9150 15.1420 dicyanamide 14.7336 17.4006 Ethyl sulfate 15.8121 27.1192 Methyl sulfate 20.7785 28.2590 PF 6 13.1277 9.7641 Tetrachloroaluminate 11.1443 17.7546 Tf 2 N 7.0149 20.7695 Trifluoromethane sulfonate 10.6059 17.9915

PAGE 227

213 Table C 2: UNIFAC binary interaction parameters CH 2 _N CH 3 _N CH 2 _C CH 3 _C Benzyl Benzene Toluene ( OCH 2 ) ( OCH 3 ) CH 2 OH Im Py acetate BF 4 Dicyanamide Ethyl sulfate Methyl sulfate PF 6 Tetrachloroal uminate Tf 2 n Trifluoromethane sulfonate CH 2 _N 0.00000 1.16579 25.7773 6.45722 15.1489 1.48005 18.8038 5.88507 1.10437 2.02013 16.9582 5.79161 17.7573 15.99783 4.96941 0.05499 10.93569 0.80336 13.15483 15.05248 11.37369 CH 3 _N 1.68084 0.00000 4.85109 5.06299 0.59709 2.01512 6.62040 5.40423 0.82860 1.20805 6.06809 0.37559 3.80043 1.34011 0.36836 0.18676 1.47925 0.05868 1.19747 1.23485 2.19861 CH 2 _C 25.28551 5.99498 0.00000 47.09916 10.5870 15.0526 11.7414 19.6074 10.7467 7.95056 62.0555 75.0840 48.2112 15.77255 4.94174 21.71303 21.17017 48.3189 3.46539 16.69048 12.74337 CH 3 _C 0.70579 5.06994 52.6398 0.00000 4.03457 4.29872 12.8932 40.9540 6.30428 2.09730 38.6775 49.7486 15.9366 29.41371 18.41411 11.70081 14.01750 10.8254 28.51723 21.49728 0.41066 Benzyl 14.00093 0.17840 11.2833 2.66468 0.00000 4.15252 2.02645 0.91927 0.35096 0.82274 4.84100 12.3605 17.1604 10.00296 1.72898 5.80848 13.08336 8.87472 5.45786 9.95203 8.21795 Benzene 2.30388 2.90971 18.3311 7.72913 1.14559 0.00000 0.89930 32.7881 0.29892 1.14945 2.54493 0.95936 0.39166 43.52084 30.26628 8.45436 10.15082 34.8469 17.31842 12.61899 17.02574 Toluene 15.64443 7.70630 18.0393 14.48392 2.79985 0.96391 0.00000 22.5853 0.84185 32.5409 37.5683 29.3082 3.13926 44.78354 33.04389 11.60019 14.54582 45.5756 7.33549 2.29163 21.66514 ( OCH 2 ) 5.79003 4.96635 24.8178 44.89187 0.13243 33.7713 26.2788 0.00000 0.09726 0.96165 12.7951 21.0561 16.3011 8.05758 2.03792 1.09650 8.06744 0.36425 12.14903 13.34696 12.60302 ( OCH 3 ) 1.56356 0.38124 3.14236 6.94162 0.50097 0.83967 3.03527 1.00263 0.00000 0.73816 1.73568 3.08941 14.7912 10.15447 8.17613 13.91870 14.50267 22.7935 25.38931 3.77102 1.47467 CH 2 OH 1.95315 1.57426 7.94395 2.33672 0.13845 0.40771 34.9896 0.91483 0.36147 0.00000 3.45168 22.8505 5.23318 1.11358 2.11460 5.48680 3.60385 0.10905 2.39802 4.74375 1.98723 Imidazolium 17.16286 5.09030 57.6604 32.85272 5.33940 0.98912 25.7026 10.3675 0.24214 3.61899 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 Pyridinium 5.53185 0.05547 84.3573 56.34316 16.6069 0.42367 31.6206 13.7473 1.17319 22.0478 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 acetate 17.91113 2.94794 44.0661 11.43326 16.0744 1.77493 2.49335 16.8226 14.2412 3.60559 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 BF 4 17.19043 0.84555 11.9109 36.46595 9.14039 46.2113 45.1081 8.24678 7.28194 0.64549 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 Dicyanamide 6.80295 0.42248 0.04199 21.01523 3.65490 30.1238 34.4092 3.34343 8.44273 2.60012 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 Ethyl sulfate 1.69681 0.93907 27.5389 12.53314 1.11396 7.76601 15.5473 0.95280 13.2171 5.30778 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 Methyl sulfate 11.52280 1.69572 22.9545 16.14358 12.3407 9.29195 15.0417 6.17558 14.0437 3.33744 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 PF 6 0.92455 0.49328 59.8144 9.51712 10.4494 35.0327 47.0096 0.17842 24.6162 0.43686 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 Tetrachloroal uminate 5.26286 1.53392 1.44656 29.38851 6.91276 17.5325 6.62032 11.7053 14.6398 2.21476 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 Tf 2 n 16.02941 0.32853 11.5435 23.38245 8.33547 12.7067 4.49287 10.7186 2.74782 5.19038 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 Trifluorometh ane sulfonate 11.36804 2.21452 14.3549 3.59546 6.18483 14.2896 24.2871 10.9620 1.77800 2.26958 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000

PAGE 228

214 Appendix D : IL structural Groups used in CAILD for CO 2 capture Table D 1: Structural groups used in CAILD to design an optimal IL for CO2 capture process Cation Structure Valence s Anion Structure Groups Valence Imidazolium 1,2 CH 2 2 Pyridinium 1,2 CH 3 1 Pyrrolidinium 1,2 ( OH) 1 Ammonium 1,2,3,4 Acetate ( O ) 2 Phosphonium 1,2,3,4 Dicyanamide Phenyl 1 Piperidinium 1,2 Methyl sulfate Benzyl 1 Sulfonium 1,2,3 Ethyl sulfate Trifluoromethanesulfonate

PAGE 229

215 Appendix E : Life Cycle Inventory of Ionic Liquids Production The life cycle inventory data of the production of Ionic Liquids (ILs) were derived from a combination of mass and energy balances from literature, theoretical calculations, and chemical process simulation. The methodology adopted to derive this inventory is presented below. Detailed material and energy flows for [Bmim] + [Br] is presented in Fig ure E 3, while a consolidated inventory of materials and energy requirements for the production of the other four ILs are presented in Table S1. The data sources for material and energy inputs are presented in Table S2. [Bmim] + [Br] : Chemical reaction an d separation steps involved in the final synthesis of [Bmim] + [Br] using the reactants bromobutane and 1 methyl imidazolium was modeled utilizing Aspen Plus. A schematic diagram of the process is shown in Figure E 2. Kinetics data needed to model the chem ical reaction was derived from Shaozheng et al. 234 This experimental study used a micro channel reaction system consisting of a micro mixer and a tubular reactor to investigate the kinetics of the butylation of 1 methyl imidazolium (MIM) towards the synthesis of the ionic liquid, [Bmim] + [Br] as shown in e qn (E 1) BrBu+MIM [Bmim] + [Br] ( E 1 ) Our simulation included a micro channel reactor with three tubes with an inner diameter of 1.8 mm and length of 1.13 m as specified in the experimental work by Shaozheng et al 234 The residence time was 2 5 minutes. The molar flow rates of the reactants 1 bromobutane and

PAGE 230

216 1 methylimidazole were set at 0.1 mol/hr and 0.08 mol/hr respectively. The reactants entered the reactor at and 1 bar. The butylation process is an endothermic reaction and the reactor was simulated as an adiabatic reactor. The simulation results showed that the output flows were at about 33 and a conversion of 56% was achieved. The products were separated in a flash drum operating at 262 and 0.5 bar, where the unreacted 1 bromobutane and 1 methylimidazole were removed from the main product The bottom product of the flash drum was enriched with ionic liquid having a purity of about 98%. The thermal energy needed for the separation step (flash drum) and for pre heating the reactants was assumed to be provided by natural gas combusted in an industrial boiler. Electricity from grid was used for compression of gases, pumping the liquids, and cooling. The energy and materials required to produce 1 kg of [Bmim] + [Br] (based on the simulation results) are listed in the last block of Figure E 3. Inventory for the production of the reactant methyl imidazolium (MIM) was not available in any standard LCI database and therefore mass and energy needed was obtained from CPS results presented in Righi et al. 225 Inventory for the reactant bromobutane was also not available in any LCI database. Therefore we utilized the CPS mass and energy balances of chlorobutane provided in Righi et al. 225 to adjust the data for bromobutane (BuBr) as follows. Bromobutan e is produced from hydrogen bromide and butanol through the reaction shown in e qn (E 2 ) HBr + C 4 H 9 OH C 4 H 9 Br + H 2 O ( E 2 )

PAGE 231

217 This reaction has the same stoichiometry and is similar to chlorobutane production except that hydrogen bromide is used as the reactant instead of hydrogen chloride. Inventory data for HBr was also not available in LCI databases. Hence, life cycle inventory of HCl (hydrogen chloride) was used since both reactions have the same stoichiometry and uses same process of production. The thermal and electrical energy needed to produce 1 chlorobutane reported in Righi et al. 225 was adjusted to 1 bromobutane based on its molecular weight and physical properties: heat of reaction and heat of vaporization. Similarly, the byproduct chlorob enzene in the LCI of HCl was replaced with equivalent amount of bromobenzene (same molar amount). Note that >95% of ecotoxicity impacts of the HCl production process is due to chlorobenzene release and substitution of it with equivalent amount of bromobenz ene instead is a critical and crucial adjustment. Inventory data related to n butanol was derived from Ecoinvent while data for other precursor materials (formaldehyde, Hydrobromic acid, ethylene glycol, ammonia, methanol, N & P fertilizers, and lime) were gathered from Ecoinvent or USLCI database (See Table S2). Whenever we had to use Ecoinvent data for chemicals derived from European databases we made a critical change of adjusting the energy mix to U.S. data (we refer to this as Ecoinvent adjusted to US) The energy and material data for the production of ionic liquid [Bmim] + [Br] are listed in Figure E 3. [ Bmim] + [Cl] : Life cycle inventory for [Bmim] + [Cl] production was derived from CPS mass and energy balances collated from Righi et al. 225 The suggested industrial process for the production of [Bmim] + [Cl] was a three step batch process 225 Note that their inventory data

PAGE 232

218 was based on European energy mix while we applied US energy mix to their energy data to generate our inventory. [Bmim] + [BF 4 ] : The ionic liquid [Bmim] + [BF 4 ] can be produced through an anion exchange reaction as shown below [Bmim] + [Cl] + NaBF 4 [Bmim] + [BF 4 ] + NaCl ( E 3 ) To build the LCI for [Bmim] + [BF 4 ] production, inventory data of NaBF 4 from Ecoinvent database was combined with LCI data of [Bmim] + [Cl] discussed before. A stochometric calcualtion based on molar mass of reactants and products was used to estimate the amount of material needed to produce 1 kg of [Bmim] + [BF 4 ] As for the io n exchange reaction step e qn (E 3), the amount of thermal and electrical energy required was assumed to be similar to that of the final reaction step of [Bmim] + [Cl] production. [Bmim] + [PF 6 ] : The same approach was utilized to derive the LCI for production of [Bmim] + [PF 6 ] This ionic liquid was assumed to be produced through the below anion exchange reaction: [Bmim] + [Cl] + HPF 6 [Bmim] + [ PF 6 ] + HCl ( E 4)

PAGE 233

219 The inventory data for HPF 6 was not available in standard LCI databases and hence were derived based on theoretical estimations (using heat of reactions and heat of vaporization for reaction and separation steps) of mass and energy balances required to produce HPF 6 through the react ion shown below: 1 H 3 PO 4 (aq) + 6 HF (aq) 1 HPF 6 (aq) + 4 H 2 O(l) ( E 5 ) For the upstream steps, the theoretical values of energy requirements were translated into industrisal scale data using conversion factors which were specifically d eveloped for this purpose. These factors were derived by comparing the indutrial scale energy consumption of producing several common chemicals with their corresponding theoritcal energy requirements. In the final step of reaction tree (ion exchange step ), e qn E 4 the thermal and electrical energy required, were assumed to be similar to the final step of [Bmim] + [Cl] production. [BPy] + [Cl] : The ionic liquid [BPy] + [Cl] was assumed to be produced thorugh reaction shown in e qn E 6. BrBu + C 5 H 5 N (Pyridine) + C 4 H 8 O 2 (Ethylacetate) [BPy] + [Br] ( E 6) Life cycle inventory of the precursor materials of the ionic liquid, pyridine and ethylacetate, were collected from Ecoinvent while the inventory for 1 bromobutane was assembled as

PAGE 234

220 discu ssed in the section of [Bmim] + [Br] production. The theoretical energy needed for the final step e qn E 6 was calculated based on heat of reaction and heat of vaporization as explained in the prevouis sections. The input material required to produduce 1 kg of the ionic liquid was calculated from the stoichiometry values of the reactants and products. The actual energy needed to calculate the life cycle inventory of this ionic liquid was estimated by applying the aformentioned conversion factors.

PAGE 235

221 Figure E-1: System boundaries of the Cradleto -Grave Life Cycle Assessment IL Recovery Recycling Energy Production (Heat and Power) Production of Precursors (Chemicals) Fuel Production IL P roduction Ionic Liquid Use IL Production Phase Natural Resources Transport IL Use Phase IL Emission (Release to Water) Emissions to Environment IL End of Life IL release

PAGE 236

222 Figure E 2: A schematic of chemical process simulation for Production of [Bmim] + [Br]

PAGE 237

223 Figure E 3: E nergy and material inventory for the production of [Bmim] + [Br] All values are adjusted to 1 kg of final product ([Bmim] + [Br] )

PAGE 238

224 Table E-1: Consolidated energy and material requirements for production of ILs Material (1 kg) Inputs Precursors Thermal Energy (MJ) Natural gas () Electricity (Kwh) (US grid) [Bmim] + [Cl] 1 methylimidazole (0.49 kg) [Righi e t al. ] 224 1.5 0.058 1 chlorobutane (0.61 kg) [Righi et al. ] 224 Ethyl acetate ( 0.04 kg) [USLCI] [Bmim] + [BF 4 ] Sodium tetrafluoroborate [NaBF 4 ] (0.48 kg) [Ecoinvent adjusted to the US] 1.49 0.062 [Bmim] + [Cl] (0.77 kg) (Righi et al. ) 224 [Bmim] + [PF 6 ] HP F 6 (0.51 kg) [This study] 1.48 0.061 [Bmim] + [Cl] (0.61 kg) [Righi et al. ] 224 [BPy] + [Cl] 1 chlorobutane (0.63 kg) [Righi et al. ] 224 2.1 0.06 Pyridine (0.36 kg) [Ecoinvent adjusted to the US] HPF 6 Phosphoric acid (0.67 kg) [Ecoinvent, RER] 1.17 0.316 H ydrogen fluoride (0.82 kg) [Ecoinvent, RER]

PAGE 239

225 Table E 2: D ata sources Inventory Reference Glyoxal Righi et al 224 1 methylimidazole Righi et al 224 Methylamine Ecoinvent adjusted to US Butanol Ecoinvent adjusted. to US Hydrobromic acid (HBr) HCL from USLCI adjusted to HBr Electric Energy US grid, 2010 Heat US Industrial Boiler 1 bromobutane BuCl data adjusted to BuBr Propylene Ecoinvent, RER adjusted to the US Hydrogen, liquid Ecoinvent, RER adjusted to the US Nitrogen, liquid Ecoinvent, RER adjusted to the US Ammonia USLCI Methanol USLCI Formaldehyde Ecoinvent adjusted to the US Ethylene glycol USLCI

PAGE 240

226 Inventory Reference HPF 6 Calculated and adjusted to the US NaBF 4 Ecoinvent adjusted to the US pyridine Ecoinvent adjusted to the US ethylacetate Ecoinvent adjusted to the US

PAGE 241

227 Table E 3: F reshwater ecotoxicty Characterization Factors (CFs) of ionic liquids and conventional chemicals IL CF (CTUe/kg) [Bmim] + [Br] 624.375 [Bmim] + [Cl] 747.448 [Bmim] + [BF4] 823.422 [Bmim] + [PF6] 927.05 [BPy] + [Cl] 1767.97 Formaldehyde 297.42 Toluene 55.91 2,3,4,6 tetrachlorophenol 24927.8 Furfural 386.82 Benzyl chloride 818.11 Benzylamine 182.04 Imidazole 185.92 2 aminopyridine 684.75 Pyrene 885597.45 2,6 diphenylpyridine 71589.60 2,4 D Butyl ester 16378.85 Butanone 12515.26 2,4,6 Trinitrotoluene 9399.86

PAGE 242

228 Table E 4: Ecotoxicity impacts related to different HCL production processes Process Ecotoxicity Impacts (CTU e ) Benzene Chlorination / EU 8.22 Reaction of propylene with chlorine (36% in H 2 O) /EU 0.027 Manheim process / EU 0.353 Reaction of H 2 with Cl 2 (chlorine) /EU 0.553