**2. Applications of PSE in the Development of New and Green Chemicals**

In this section, an introduction to ionic liquids and their potential applications is presented. A comprehensive review of the applications of ionic liquids in various industries is also provided. Furthermore, the latest developments and contributions in the field of ionic liquid design are reviewed. In addition, the challenges and limitations encountered in the design of optimal ionic liquids are discussed. Lastly, the recent applications of CAMD in the ionic liquid design are reviewed in detail.

## *2.1. Ionic Liquids*

Ionic liquids are organic salts that comprise organic cations with inorganic or organic anions and melt at temperatures below 100 ◦C. They were first introduced by Paul Walden in 1914 when he successfully synthesized ethyl ammonium nitrate which melts at room temperature [11]. The introduction of ionic liquids has attracted considerable attention from researchers, especially in replacing traditional solvents with ionic liquids due to their extraordinary characteristics. For example, ionic liquids are known to exhibit non-flammability, as well as high thermal and chemical stabilities owing to the strong ionic bonds [12]. Besides, ionic liquids also possess the attributes of "green" solvents since their vapor pressure is significantly lower than the conventional solvents. The high volatility of conventional organic solvents can cause air pollution and human health problems if they leak from process equipment.

Ionic liquids are widely recognized as "designer" solvents because of the flexibility in turning their properties by altering the cations. For instance, their melting point can be reduced by having large asymmetric organic cations [13]. By selecting various combinations of anions and cations, ionic liquids with an attractive set of physicochemical properties can be synthesized. In addition, ionic liquids' capabilities, which include a wide range of intermolecular interactions such as hydrogen bonding, ionic and covalent interactions as well as π-stacking, have also contributed to their tunable properties [14]. For these reasons, they have huge potential to reduce the environmental impact caused by organic solvents. Nonetheless, the major drawback of applying ionic liquids in various chemical processes is their high purification cost. As ionic liquids consist of highly charged ions, the ionic liquid recovery and purification process becomes difficult and cost-intensive [15]. In spite of the high recovery and purification cost, the unique properties of ionic liquids are still important and attractive from an industrial point of view. This can, at least in part, be seen by an increasing number of patent filings, especially related to the applications of ionic liquids [16]. Moreover, there is also an exponential increase in Science Citation Index (SCI) papers published on ionic liquids in the last few decades [17]. For these reasons, it is feasible that ionic liquids can become viable options for various applications and industrial processes.

#### *2.2. Potential Applications of Ionic Liquids*

In recent years, there is a surge in the applications of ionic liquids in various industries. Table 1 shows some potential applications of ionic liquids that will be discussed in detail in this section.


**Table 1.** Summary of potential applications of ionic liquids discussed in this review.

The first industrial application of ionic liquids, called the BASIL process, was developed by BASF [18]. This process is mainly used for producing generic photo initiator precursor alkoxy phenyl phosphines. It is the first commercial publicly announced process that uses an ionic liquid (named 1-methylimidazolium chloride) to scavenge acid from reaction mixtures. The extraordinary properties of ionic liquids are also advantageous for several reactions such as bio-catalyzed reaction, dimerization, isomerization, hydrogenation, hydroformylations, and alkylation [34]. Higher reaction rates and selectivity can be achieved when organic solvents are replaced with ionic liquid solvents. For instance, the hydrolysis rate of casein by lumbrokinase is enhanced when ionic liquids are added to the reaction [35]. The existence of ionic liquids at low concentrations improves the hydrophilicity of casein by forming an ionic liquid-casein complex.

Ionic liquids have also been reported as excellent candidates for heat transfer fluids owing to their low flammability, high heat stability, and low volatility characteristics. When ionic liquids were used as heat transfer fluids in shell and tube heat exchangers, a comparable or larger heat transfer area can be achieved compared to other heat transfer fluids [36]. The ionic liquid, 1-butyl pyridinium tetrafluoroborate (C9H14NBF4) was used as a geothermal fluid in an organic Rankine cycle was explored by Kazemi et al. [19]. Optimization and simulation work was carried out for a basic organic Rankine cycle using C9H14NBF4 and water as the geothermal fluids, respectively. In their study, economic, thermodynamic, and thermo-economic evaluations were chosen as objective functions for minimizing specific investment costs and maximizing exergy efficiency. The results indicated that the basic organic Rankine cycle performance was improved thermodynamically and economically when C9H14NBF4 was used as a geothermal fluid.

Moreover, ionic liquids have a high tendency to wet inorganic, metal, and polymeric surfaces [34]. With their wide liquidus range and high thermal stability, ionic liquids are ideal candidates to be used as lubricants in low pressure and/or high-temperature applications. Other than the aforementioned properties, ionic liquids are also applied as lubricating media in oxygen compressors due to their good lubricity and high chemical inertness [37]. Based on the long-term test conducted with pure oxygen, the result is both impressive and promising as the hydrocarbon concentrations in all tests were below 10 ppmv methane equivalents, which is beyond the threshold value of 1000 ppmv. In another recent study, it was proposed that the solid piston in reciprocating hydrogen compressors may be replaced by ionic liquids [20]. By replacing the solid piston in reciprocating compressors with a suitable liquid, efficiency can be improved along with significant cost reduction.

Apart from being suitable candidates for lubricants and thermal transfer fluids, ionic liquids have also been introduced in biomass conversion applications, such as lignocellulose pretreatment and fractionation. Most existing molecular solvents are not able to dissolve and hydrolyze cellulose into fermentable sugars as the cellulose chains are interconnected by strong hydrogen bonding in flat sheets [38]. Nonetheless, various ionic liquids have demonstrated their capabilities in dissolving and fractionating lignocellulose into its main constituent compounds. For example, cellulose isolated from sugarcane bagasse was pretreated with an ionic liquid called 1-Ethyl-3-methylimidazolium acetate ([Emim]Ac) at 90 ◦C prior to hydrolysis by cellulase [21]. After the pretreatment process, the glucose content is increased whereas the crystallinity index and degree of polymerization are reduced. The cellulose crystal structure has also transformed from cellulose I to cellulose II. These changes have enhanced enzymatic hydrolysis rate and glucose production yield. Furthermore, the effect of pretreating rubber woods with ionic liquids on thermodynamic properties and pyrolysis kinetics was explored by Khan et al. [22]. The pretreatment of rubber woods with both ionic liquids, 1-butyl-3-methylimidazolium chloride ([BMim][Cl]) and 1-butyl-3-methylimidazolium acetate ([BMIM][OAc]) successfully reduced the activation energy required for the pyrolysis reaction. Moreover, ionic liquids have also shown their potential in kraft lignin activation processes. After activating kraft lignin with ionic liquids, a better electrochemical performance was observed due to an increase in carbonyl groups, which are responsible for transporting protons and electrons in electrochemical applications [23].

Considering stringent environmental regulations as well as increased emphasis on green and clean manufacturing practices, ionic liquids have gained popularity in extraction processes. This is because most of the common organic solvents used in extraction processes are toxic and flammable, which leads to detrimental environmental issues when released into the atmosphere. The use of ionic liquids in extraction processes typically leads to shorter processing time and most importantly, the process can often be carried out at ambient temperature [39]. For this reason, the application of ionic liquid is widely found in the extraction and purification of bioactive compounds such as proteins, lipids, amino acids, and pharmaceuticals. Novel dual-chain ionic liquids were synthesized and applied to recover flavonoids from *Pinus massoniana* Lamb [24]. The results showed that ionic liquids are more efficient in extracting flavonoids than water and ethanol. In another study, a hydrophobic ionic liquid was shown to be capable of extracting keratin from chicken feathers [25]. The water extraction method

can be used to isolate keratin from the ionic liquid as hydrophobic ionic liquid is insoluble in water while keratin is soluble in water.

A classic challenge in the process has been to choose a suitable entrainer in separating azeotropic mixtures. Ionic liquids have been suggested as potential entrainers in an azeotropic system since their selectivity, capacity and thermal stability can be easily altered [40]. For example, Zhu et al. [26] performed a process simulation study of ionic liquids applied in an extractive distillation process to separate an ethyl acetate and ethanol mixture, which have close boiling points. The results showed that 1-Ethyl-3-methylimidazolium methyl sulfate was the best solvent for this separation. In another contribution, 1-butyl-4-methyl pyridinium tricyanomethanide ([4bmpy][TCM]) has been proposed as an entrainer in the extractive distillation process to separate cyclohexane and benzene [27] as overall solvent consumption can be reduced. Another contribution evaluated the capability of protic ionic liquids as an entrainer to separate ethanol-water azeotropic mixtures [41]. In this process, the relative volatility between ethanol and water is increased and the complete azeotrope elimination was achieved. Conductor-like screening model for real solvents (COSMO-RS) was used to study the intermolecular interactions between the binary ethanol-water and the ternary ethanol-water-protic ionic liquids mixtures. The results were in agreement with the experimental values.

The possibility of using ionic liquids to separate liquids was explored by Fadeev and Meagher [42], where an ionic liquid was used to recover butyl alcohol from a fermentation broth. Ionic liquids are also reported to be efficient in gas separation. Due to their hygroscopic nature, ionic liquids can remove water vapor efficiently from gas mixtures [43]. Likewise, the solubility of gas can be altered by the proper selection of cation, anion, and substituents. For example, when certain anions are used and the alkyl chain is shortened, the Henry's Law constant of carbon dioxide is reduced to half [34]. Based on the experiments performed by Anthony et al. [44], gases were found to have different solubilities in ionic liquids, suggesting that there is a great potential for designing ionic liquids to be used in specific gas separations.

Carbon dioxide (CO2) was first reported to be highly soluble in imidazolium-based ionic liquids like 1-butyl-3-methylimidazolium hexafluorophosphate by Blanchard et al. [30]. The finding reported that the dissolution of CO2 in imidazolium-based ionic liquids is totally reversible meaning that pure ionic liquids can be easily recovered after the desorption process. Since then, there has been an increasing number of scientific studies investigating the capability of ionic liquid compounds in capturing CO2. The efficiency of a new task-specific ionic liquid, which is an amino-functionalized ionic liquid named 1-butyl-3-propylamineimidazolium tetrafluoroborate ([NH2p-bim][BF4]), was comparable to conventional solvents in CO2 absorption processes [12]. To maximize the interaction sites for CO2 capture, a dual amino ionic liquid that contains a taurine anion and amino-functionalized imidazolium cation has been synthesized [45]. The result reported that 1 mole of the ionic liquid can absorb 0.9 mol of CO2 at atmospheric pressure. Besides, the application of both an ionic liquid and a zeolitic imidazolate framework (ZIF) as a separating agent in adsorptive absorption has demonstrated its capability in capturing CO2 [29]. The experimental and modeling results showed that CO2 is more soluble in the ionic liquid and ZIF mixture compared to that of pure ionic liquid.

Ionic liquids have also been introduced for absorbing other environmentally polluting gases such as hydrogen sulfide, ammonia, and sulfur dioxide. A mixed matrix membrane with imidazolium-based ionic liquids blended with poly ether-block-amide elastomer was synthesized for CO2 and H2S separation [28]. This study showed that the chosen ionic liquid has higher H2S selectivity than CO2. In another contribution, tetraglyme-sodium salt ionic liquids which possess high thermal stability appeared to be potential candidates for SO2 absorption [31]. The experimental results showed that SO2 absorption capacity improved by 30% when tetraglyme–sodium salt ionic liquids were used instead of tetraglyme. Three different types of ionic liquid including traditional ionic liquids, protic ionic liquids, and Brønsted acidic ionic liquids were synthesized and investigated for NH3 absorption [32]. Among the investigated ionic liquids, the protic ionic liquid (1-butyl imidazolium bis(trifluoromethylsulfonyl)imide ([Bim][NTf2])) achieved the highest NH3 absorption capacity. 1 mole

of [Bim][NTf2] was able to absorb 2.69 moles of NH3 at ambient pressure. Another study synthesized a series of metal ionic liquids and explored their potential for NH3 separation [33]. Bis(1-butyl-3-methyl imidazolium) copper tetrachloride salt ([Bmim]2[CuCl4]) and bis(1-butyl-3-methyl imidazolium) stannum tetrachloride salt ([Bmim]2[SnCl4]) showed great potential for NH3 separation due to their high selectivity and absorption capacities.

The abovementioned works illustrate that ionic liquids have been extensively explored for various applications. Nonetheless, considering that an enormous number of ionic liquids can be synthesized by combining various organic cations and anions, it is difficult to select the optimum candidate(s) from this huge search space. Experimentally testing random ionic liquid combinations is both time consuming and costly. Thus, in order to minimize the time and expense required in performing experiments, various theoretical/computational methods have been proposed to guide the ionic liquids screening process preceding an experimental campaign. The following section presents the latest developments and challenges in the design of optimal ionic liquids.

#### *2.3. Challenges in the Design of Optimal Ionic Liquids*

In order to design ionic liquids that suit a certain industrial application, it is first important to know how the anion, cation, and side chains on the cation affect physicochemical properties. This can be done through two pathways, either via experimental work or simulation approaches. Considering a very huge number of possible ionic liquids, it would be a very tedious task to select an optimum candidate through experimentation. Because of this, researchers have been working on developing systematic computer simulation methods including CAMD and/or process design simulations to design optimal ionic liquids. Computer-aided methods provide alternative routes in determining potential ionic liquids as they can explore a larger number of options in a shorter time frame. However, it does not mean that computer-aided methods can replace experiments entirely as identified candidates should be further tested via experiments to verify their performance.

CAMD techniques are shown to be effective in designing various types of chemical, biochemical, and material products. It is able to design molecular structures that meet a predefined property target. The success of employing the CAMD technique to design ionic liquids depends greatly on the availability and reliability of the associated/underlying predictive models. Only when thermophysical properties such as density and viscosity (which influence mass transfer rates) are satisfactorily characterized, can optimal ionic liquids and/or the process in which they are used be designed [46]. Despite the availability of significant thermophysical property data for an extensive range of ionic liquids in the free ILThermo database [47], predictive models for most thermophysical properties with adequate accuracy are still being developed [48]. The availability of adequate predictive models is important in the design/development of new processes, improvement of operating conditions in a process, and reduction of energy consumption [48]. Various approaches have been proposed to develop property prediction models for thermophysical and transport properties of ionic liquids. These approaches can be categorized into two main classes which include both theoretical and empirical methods [49]. The property prediction models classified under theoretical approaches include molecular dynamics (MD) simulations, equations of state (EOS), Monte Carlo (MC) simulations as well as rough hard-sphere theory (RHST). On the other hand, machine learning (ML) tools, GC methods, and quantitative structure-property relationships (QSPR) are considered empirical approaches. These prediction models are all widely applied in predicting physicochemical properties of ionic liquids, which include density, viscosity, thermal conductivity, surface tension, heat capacity, etc. A comprehensive review of the current developments in prediction models for pure ionic liquids was presented by Hosseini et al. [48]. Moreover, Hosseini et al. [48] also reviewed the most extensively applied semi-classical EOSs in estimating thermodynamic and physical properties of pure ionic liquids. EOSs are typically developed from empirical, theoretical, and semi-theoretical approaches. Semi-classical EOSs are developed from a combination of molecular thermodynamic theories of fluid (theoretical EOSs) and classical thermodynamic approaches (empirical EOSs). Nonetheless, most of

these existing prediction models for thermodynamic properties are limited to only a relatively small number of common families of ionic liquids.

To alleviate the limitations mentioned above, Chen et al. [3] presented newly developed GC models that are capable of estimating thermodynamic and physical properties for additional families of ionic liquids. The thermophysical properties include heat capacity, viscosity, surface tension, density as well as the melting point of the ionic liquids. New correlation equations with higher prediction accuracy that are easier have been proposed for the prediction of the aforementioned properties. The GC parameters were developed based on more than 13,300 experimental data points, including approximately 200 ionic liquids derived from 19 anions, 6 cations, and 4 substituents covering a large range of pressure and temperature conditions. These newly developed GC models offer the opportunity of applying them for CAMD to identify more potential ionic liquids solutions. Additionally, there is great potential for extending these methods to even more families of ionic liquids once experimental data become available.

With increasing awareness of environmental detriments, it is important to also consider environmental properties during the ionic liquid design stage. Since ionic liquids are normally highly soluble in an aqueous medium, they are readily discharged into the environment through wastewater [50]. Hence, the designed ionic liquids need to be environmentally friendly before they start accumulating in the environment. For this reason, various quantitative structure-activity response (QSAR) models were developed to assess the toxicity of ionic liquids. Recently, a comprehensive review of the expansion of QSAR models for toxicity prediction of ionic liquids was presented by Abramenko et al. [51]. The review concluded that the toxicity of ionic liquids is not solely reliant on the type of cation but increases with the length of the cation alkyl chain and the number of cation chain group branches. It was also found that the toxicity of ionic liquid is reduced with the presence of a polar side chain. Furthermore, the presence of polar side chains was found to increase the efficiency of biodegradation. However, most existing models for ionic liquid toxicity prediction only cover a small number of ionic liquids.

It is notable that pure ionic liquids may not always be the optimal candidates in satisfying the target properties, however, it is possible that ionic liquid mixtures could assist in addressing such problems. Unfortunately, property prediction models for ionic liquid mixtures are limited as this research area is still in its infancy. In a recent contribution, equations of state and artificial neural networks (ANN) were combined to predict volumetric properties of a limited number of pure and mixtures of amino acid ionic liquids [52]. Another contribution explored the behavior of some properties such as viscosity, diffusion coefficient, and electrical conductivity for binary mixtures of protic-protic ionic liquids as well as protic-aprotic ionic liquids [53]. The results presented that the electrical conductivity of ionic liquid mixtures is higher than that of the pure ionic liquids. Efforts have been devoted to developing property prediction models for ionic liquid mixtures, but the models are limited to only certain classes of ionic liquids. Future research should focus on developing reliable property prediction models for different families of ionic liquid mixtures.

#### *2.4. CAMD for Ionic Liquid Design*

The development of CAMD tools for the ionic liquid design has been limited due to the lack of accurate models in estimating ionic liquid properties. Most of the earlier work in the modeling of ionic liquids was based on molecular dynamics tools and quantum chemistry study. While these tools show reasonably accurate and reliable results, the computational time needed to complete the simulation of each candidate molecule is large. So, the applicability of these tools is more appropriate towards the final selection from a limited number of promising molecules. An early application of CAMD in ionic liquid design was the design of a suitable ionic liquid to be used in a refrigerant gas separation system. This work used molecular connectivity index based QSPR models to link the structure of the ionic liquids to target properties like diffusivity, Henry's law constant, and solubility [54]. A CAMD problem is formulated to determine the molecular structures is then formulated as a MILP. Another earlier application of CAMD tools for the design of ionic liquid was the extension of existing CAMD approaches based on GC models [4]. The original MINLP problem was decomposed into a set of sub-problems to reduce the complexity of the model. This approach solves one sub-problem each for structural constraints, physical property constraints, and mixture property constraints. Each of these sub-problems can eliminate several infeasible structures. The solutions obtained through the sub-problems are considered in a final sub-problem to solve for the optimum ionic liquid molecule based on the objective function and process models.

Figure 2 shows the general framework of ionic liquid design via CAMD. The procedure starts with an ionic liquid design problem definition where the product needs are identified. These requirements are then translated into targeted properties which will be validated using a model-based approach. Next, an extensive search of ionic liquids is conducted. The information collected from a vast number of literature sources and online databases is stored in the ionic liquid library. Various property models are also collected and stored in the property model library. In Step 4, UNIFAC and/or COSMO models are used to screen and predict ionic liquid with desirable properties. Lastly, the performance of the shortlisted ionic liquid candidates is validated either by experiments or simulations. This step is crucial to ensure that the ionic liquids identified are feasible and practical.

**Figure 2.** General framework of ionic liquid design via a computer-aided molecular design (CAMD) approach.

The application of GC-based models in CAMD has been well established and there are efficient algorithms for solving CAMD problems with different target properties. A comprehensive review of these works can be seen in the reviews of Austin et al. [55] and Chemmangattuvalappil [56]. In order to provide an accurate prediction of properties using GC models, the binary interaction parameters need to be obtained. QM calculations have been incorporated into several CAMD algorithms to address this issue in recent years [56]. One of the most popular advances in CAMD is the application of COSMO-based methods as the information of binary interaction parameters is not necessary [5]. COSMO-RS [57] and COSMO-SAC [58] are the two most popular methods applied in CAMD [5]. These methods only require the estimation of molecular volumes and sigma profiles for the thermodynamic calculations. This makes the COSMO based approaches extremely suitable for the design of ionic liquids.

A preliminary design strategy to screen a list of possible ionic liquids for CO2 capture by using COSMO-RS was proposed by Farahipour et al. [59]. COSMO-RS was applied for activity coefficients estimation whereas viscosity and melting point constraints are modeled by employing GC models. The proposed methodology holds the capability to consider different trade-offs and optimal ionic liquids were identified by enumerating all possible cation and anion combinations. As a result, 10 ionic liquids with promising characteristics were identified out of a few thousand ionic liquid candidates considered. This method was improved with an extended model library for ionic liquid properties

along with the consideration of several ionic liquid-based separation processes [60]. In their work, CAMD, property models based on GC, and process design simulations are combined and solved simultaneously to obtain the optimal ionic liquid separating agent. The developed method was then applied in case studies. For the first case study, CO2 capture from natural gas was considered while the focus of the second case study was ethanol-water azeotrope separation.

In recent years, systematic approaches have been developed for ionic liquid screening as extraction solvents. In one study, a sequential approach combining COSMO-RS-based LLE prediction, GC-based physical property prediction, and process simulation using Aspen Plus was applied [61]. Firstly, COSMO-RS is utilized to select the cation-anion solutions which possess desirable properties with common organic solvents as a benchmark. In the next step, GC tools have been applied to screen ionic liquid with desirable properties. Finally, ionic liquids that give the optimal process performance are selected by evaluation through process simulation. This method was then further improved by extending the UNIFAC-IL model to describe the extractive desulfurization of fuel oils system with extensive experimental data [62]. Optimal ionic liquid molecules for the extractive distillation can be identified by integrating both process and molecular design [63]. UNIFAC-IL model was selected for its database. The database fitting the group binary interaction parameters in the UNIFAC-IL model agreed well with the characteristics of the compounds in the system. An MINLP model was formulated to represent the design problem by combining UNIFAC-IL and GC models. The potential of the shortlisted ionic liquid candidates was then evaluated by process simulation. Song et al. [64] employed a similar approach for the design of ionic liquids for an alkane/cycloalkane extractive distillation process. In this design, the UNIFAC-IL model had been extended based on the consideration of the proximity effect in alkanes/cycloalkanes as distinct groups. Optimal ionic liquid as an entrainer had been modeled by formulating an MINLP problem and then through process simulation and economic analysis.

A CAMD approach was also applied to design optimal ionic liquids for cellulose dissolution. In this work, the solubility of cellulose in ionic liquids was estimated using a QSAR model which was developed using GC and ANN methods [65]. The design problem was formulated as a MINLP and solved by a genetic algorithm with an objective function maximizing the cellulose solubility in ionic liquids. Experimental analysis and characterization of the identified ionic liquid candidates were also performed. A simple method for CAMD in ionic liquids design as CO2 absorbents was proposed by Firaha et al. [66]. The strength of the interaction between CO2 and ionic liquid anions corresponded to chemical and physical absorption, respectively. The type of absorption was predicted according to geometry optimization accompanied by a solvation model. Solvated geometries provide Gibbs free energies which are analogous to the experimental values and correlate well with the experimentally observable absorption capacity.

An extensive GC model for infinite dilution activity coefficient (IDAC) of molecular solutes has been introduced to design ionic liquids for extractive desulfurization of gasoline [67]. In this work, ionic liquids are generated by applying purely empirical correlations of GC-based IDAC data and the least-squares support vector machine (LSSVM) method. The LSSVM-based model was capable of predicting IDACs based on the GCs of ions and Abraham's solvation descriptors of solutes. The reliability of the developed CAMD tool was verified by LLE simulation of ternary systems.

Another significant contribution is the ionic liquid design approach Karunanithi et al. [68] in which GC models and the COSMO-SAC thermodynamic model are combined to predict the surface charge density of ionic liquid structures predicted by density functional theory (DFT). The CAMD model can be solved via either deterministic or stochastic methods. This approach was tested on the design of ionic liquids for ibuprofen dissolution and ionic liquids with high electrical conductivity. The properties of the identified ionic liquid were verified with the data on the properties of the predicted structure. In another contribution, a multi-scale simulation approach was introduced to design ionic liquid solvents for separation processes [69]. An extended GC-COSMO approach was developed to estimate the σ-profiles and cavity volumes of ionic liquids. In this work, activity coefficients were

predicted by applying the COSMO-SAC model while the rest of the properties were predicted using semi-empirical models. A CAMD problem was formulated using an MINLP model. The MINLP was then solved using the branch and bound approach in order to obtain the optimal ionic liquid solvents. A hybrid process design approach has been developed by Chen et al. [70]. This approach combined ionic liquid design and process simulation to obtain the best ionic liquid for hybrid process schemes. Hybrid schemes that satisfy the desired demands were generated. The ionic liquid design problem was solved based on the hybrid schemes generated, structural constraints, along with the physical properties. Shortlisted ionic liquid candidates were evaluated by conducting the simulation of the process.

The advances in the ML tools have also been utilized in the ionic liquid design. In a recent contribution, a huge data library of ionic liquids has been used to train ML models [71]. ML models were developed using three learning methods simultaneously, including the random forests (RF), cubist, and gradient boosted regression (GBM). The ML models thus developed (trained) were applied to design ionic liquids by identifying the cation-anion pairs from a large database that covers varied chemical scaffolds. The performance of the proposed method was then validated theoretically and experimentally. Over 2600 promising ionic liquids were found to be potential ionic liquids for gas separation and cellulose dissolution applications.

#### **3. Applications of PSE in Integrated Biorefineries**

In this section, an introduction to the field of integrated biorefineries and the types of integrated biorefineries is presented. The different types of processes common in an integrated biorefinery are explained. Following this discussion, various design approaches and major contributions made in each of these approaches are reviewed in detail. Finally, the research gaps are identified and major current challenges in the design and current work on these challenging areas are presented.

#### *3.1. Introduction to Integrated Biorefineries*/*Types of Integrated Biorefineries*

Over the past few decades, in order to enhance the sustainability of chemical and energy production, there has been a shift from petroleum-based feedstock to biomass-based feedstock. Besides, the societal realization of limited non-renewable resources, concerning environmental issues, technological advancements, and the discovery of additional renewable energy resources have also contributed to this shift [72].

Early on, biomass was mainly used as renewable resources for energy production. Later, biomass has also been used as feedstock for the production of value-added products via different processing pathways (physical, biological, and thermal, etc.). Various independent processes (e.g., gasification, pyrolysis, fermentation, hydrolysis, palletization, etc.) for the production of bio-chemicals, biomaterials, and bio-specialty chemicals from biomass have been developed. The utilization of biomass as feedstock for the production of multiple products through a *biorefinery* has also gained attention from both industry and the scientific community in the past decades [6]. A biorefinery was first defined as *a complex system of sustainable, environmental, and resources-friendly technologies for the comprehensive utilization and the exploitation of biological raw materials (biomass)* by Kamm et al. [73]. Similar to a petroleum refinery, a biorefinery utilizes different biomass resources as feedstock, combines a diversified collection of conversion pathways to produce a wide range of value-added products for example bioenergy, bulk chemicals, and fine chemicals. According to Frost and Draths [74], the application of the biorefinery concept plays a crucial role in driving the change in the chemical industry to shift from using petroleum-based feedstock to biomass-based feedstock. As biomass consists of a wide range of organic constituents, it comes in a variety of forms with different properties and characteristics. Therefore, various processing technologies can be applied to convert biomass into higher-value market products. A range of pre-treatment systems (such as size reduction, drying system, acid, and base hydrolysis) are needed to standardize biomass into a form that can be further converted into products. In order to maximize the quality and performance of the biorefinery,

process integration and optimization shall be applied to synthesize a biorefinery comprised of multiple processing systems. This realization provides the foundation for the concept of integrated biorefinery which integrates various biomass conversion platforms [6].

According to Gravitis et al. [75], an integrated biorefinery represents a processing facility consisting of multiple technologies including feedstock handling, pretreatment processes, and different biomass conversion/upgrading processes. This allows the by-products and waste to be minimized while recovering the energy generated within the biorefinery. Therefore, the integrated biorefinery concept provides an opportunity to create a variety of value-added products while enhancing the sustainability performance in terms of the economic, environmental, and social impacts. A general representation of the integrated biorefinery concept is illustrated in Figure 1 as provided in the Introduction section.

As shown in Figure 1, an integrated biorefinery consists of different conversion pathways that convert different types of biomass feedstock into heat, power, and value-added products through depolymerizing and deoxygenating biomass components [76]. According to the U.S. Department of Energy/National Renewable Energy Laboratory (NREL), existing conversion pathways and technologies can be generally categorized into five different platforms based on the products produced [77]. These five platforms are the sugar platform, the thermochemical/syngas platform, the biogas platform, the carbon-rich chains platform, and the plant products platform. Table 2 summarizes the foci of these five biomass processing platforms as classified by NREL:


**Table 2.** Comparison of different biomass conversion platforms.

Other than classifying the biomass conversion processes into different conversion platforms, these conversion processes and technologies are more commonly categorized according to the nature of the processes. Based on the method and nature of the processes, these conversion processes and technologies can be divided into four main groups of physical/mechanical, thermochemical, chemical, and biochemical/biological processes [76]. A detailed discussion of each group of technologies is covered in the following sub-sections.
