Computers and Chemical Engineering 29 (2005) 16611676

Method for selection of solvents for promotion of organic reactions

Raqul Gani a, , Concepci n Jim nez-Gonz lez b , David J.C. Constable c o e a
a

CAPEC, Department of Chemical Engineering, Technical University of Denmark, Building 229, DK-2800 Lyngby, Denmark b GlaxoSmithKline, CEHS, 5 Moore Drive, Research Triangle Park, NC 27709, USA c GlaxoSmithKline, CEHS, 2200 Renaissance Blvd. Suite 105, King of Prussia, PA 19406, USA Received 7 July 2004; received in revised form 17 December 2004; accepted 7 February 2005 Available online 17 March 2005

Abstract A method to select appropriate green solvents for the promotion of a class of organic reactions has been developed. The method combines knowledge from industrial practice and physical insights with computer-aided property estimation tools for selection/design of solvents. In particular, it employs estimates of thermodynamic properties to generate a knowledge base of reaction, solvent and environment related properties that directly or indirectly inuence the rate and/or conversion of a given reaction. Solvents are selected using a rules-based procedure where the estimated reaction-solvent properties and the solvent-environmental properties guide the decision making process. The current method is applicable only to organic reactions occurring in the liquid phase. Another gas or solid phase, which may or may not be at equilibrium with the reacting liquid phase, may also be present. The objective of this method is to produce, for a given reaction, a short list of chemicals that could be considered as potential solvents, to evaluate their performance in the reacting system, and, based on this, to rank them according to a scoring system. Several examples of application are given to illustrate the main features and steps of the method. 2005 Elsevier Ltd. All rights reserved.
PACS: 07.05. Tp Keywords: Solvent selection; Organic reactions; CAMD; Green chemistry; Computer modeling and simulation

1. Introduction Solvents can inuence reactions in a number of ways. Solvents may be used as a reaction medium to bring reactants together, as a reactant to react with a solute when it cannot be dissolved, and as a carrier, to deliver chemical compounds in solutions to their point of use in the required amounts. This paper deals principally with the use of solvents as an organic reaction medium. Unless otherwise specied, the term reactions in this paper will be used to mean liquid-phase organic reactions. As a reaction medium, a solvent may be used in a number of ways. For example, in endothermic reactions, heat could be supplied through a heated inert solvent having a high heat capacity, while in exothermic reactions, the surplus heat can be

Corresponding author. Tel.: +45 45252882; fax: +45 45932906. E-mail address: rag@kt.dtu.dk (R. Gani).

removed by allowing the solvent to boil or absorb heat. If reactions involve solid reactants, solvents could be used to create a solution (reaction) medium through which the solid reactants can be brought into contact. Similarly, gas phase reactions, which are normally at high temperatures and/or pressures, could be performed in the liquid phase under signicantly lower temperatures and/or pressures. Also, reactants that are too reactive in one solvent could be safely studied in another solvent. Finally, solvents may be used to indirectly inuence the reaction by removing one or more products on-site. Despite the many benecial chemical uses for solvents, solvents present numerous environmental, health and safety challenges including human and eco-toxicity issues, process safety hazards and waste management issues. In general, solvents are subject to increasing and continuous regulatory scrutiny. The above issues, amongst others, highlight the need to avoid the use of organic solvents as much as possible. Following the Green Chemistry Principles (Anastas & Warner,

0098-1354/$ see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.compchemeng.2005.02.021

1662

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

Nomenclature Conc. EHS GWP KOW LC50 concentration environmental, health and safety global warming potential octanolwater partition coefcient lethal concentration 50, measure of toxicity (50% mortality of Fathead Minnow after 96 h) No. sec. number of secondary reactions Rxs ODP ozone depletion potential Rj reaction index j RME reaction mass efciency RSj reaction-separation property index j Sj solvent score for index j SP solubility parameter (goal value) SSG GSKs solvent selection guide TR reaction temperature Tb normal boiling point temperature Tm normal melting point temperature TPB toxic, persistent, bio-accumulative

1998), when organic solvents must be employed, their use must be minimised and optimised to enhance the reactions with the minimum environmental and operational concerns, as illustrated graphically in Fig. 1. Past efforts to develop various approaches and methodologies to facilitate solvent selection for reactive systems have been undertaken (Buxton, Livingston, & Pistikopoulos, 1997; Curzons & Constable, 1999; Folic, Adjiman, & Pistikopoulos, 2004; Jim nez-Gonz lez, Curzons, Constable, e a

Fig. 1. The goal is to use a solvent that promotes chemical reactivity and the associated downstream processing while minimising material and energy use as well as avoiding adverse EHS effects (for explanation of the abbreviations, see the nomenclature).

& Cunningham, 2005), but there is a clear need to develop an approach where solvents are ranked and compared in terms of their performance with respect to different types of organic reactions. In industry, solvent selection is routinely made by experts based on their experience and insights. There exists the need, therefore, to combine the calculation based approaches with the experience based approaches so that for any specied reaction system, the combined methodology would be able to formulate the solvent selection/design problem as well as solve it with the necessary computational tools. The objective of this paper is to describe a systematic computer-aided methodology that is able to guide a user through the selection/design of the best possible solvent for a specied reaction system, taking into consideration both chemical and environmental criteria. The methodology incorporates knowledge available from industrial practice as well as currently available computer-aided tools for solvent design and property estimation into a step-by-step algorithm for selection/design of solvents for promotion of liquid phase organic reactions. The implementation of the algorithm requires a number of computer-aided tools (such as, a knowledge base of solvents, reactions and industrial practice; a tool for estimating solvent properties in solution; a tool for estimating reaction equilibrium constants, a tool for generating solvent alternatives and a tool for screening their performance in specied reaction systems), which have been integrated into the different calculations steps of the methodology. Therefore, the new methodology is able to generate new molecules as well as evaluate known chemicals as potential solvents, producing thereby, a short list of candidate solvents ranked in terms of their performance so that expensive and time-consuming experiments is made for only the most promising ones. Depending on the specic solvent selection (or design) problem, different criteria for evaluation of the solvent performance may be used, for example, yield, reaction mass efciency, atom economy, exposure limits, purity prole, secondary reactions, energy requirements and environmental impact. The solvent selection occurs in two stages. In the rst stage, a solvent table, if available, is used to identify the solvents that match a sub-set of specied requirements (constraints) and based on this a score is allocated to each solvent. Alternatively, a CAMD technique is also used to generate a list of solvent candidates, which is then ranked with respect to their allocated scores. The needed reaction data is obtained from a database containing known reactions or provided by the user. The solvents data is retrieved from a list of 75 most common solvents while the hybrid computer-aided molecular design (CAMD) technique of Harper and Gani (2000) is used to generate and test chemically stable molecules that satisfy the solvent property requirements. In the second stage, the candidate solvents are used in further evaluations involving detailed calculations (such as properties of the reaction mixture) to identify a set of candidate solvents that can be recommended for nal selection based on experiments or other verication tests (for example, checking of properties databases).

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

1663

This paper highlights the main steps of the methodology together with its implementation as a computer-aided system for selection/design of solvents for promotion of liquid phase organic reactions. Four examples of application illustrating the main features of the methodology are also given.

2. Solvent selection/design methodology Given the details of an organic reacting system (reactants, products, reaction kinetics, etc.), the objective is to nd solvents that can promote the reaction (in terms of yield, reaction mass efciency, etc.) and rank them according to an established scoring/evaluation system. The solution method applied by the methodology consists of retrieving or generating reaction data (the minimum data needed to solve the problem) and based on these, allocates values to a set of reaction-indices (R). In the next step, using a combination of rules (based on industrial practice and physical insights) and estimated solvent properties, values are allocated to a set of reaction-solvent property indices (RS). In the next step, these generated RS values for each solvent are converted to their corresponding score-values (S). The solvents that have the highest scores and do not have more than one lowest score are listed as feasible and selected for further detailed study (for example, experimental verication). In the text below, rst the reaction-solvent properties that help to identify the suitable solvents are listed together with a brief description of their roles in the solvent selection methodology. This is followed by a brief discussion on the computeraided methods and tools that have been incorporated into the methodology. These methods and tools perform the necessary calculations to generate the data needed by the rules when allocating values to the R and RS indices. Some of these tools are also needed for calculations related to validation of solvent performance. The nal section provides details of the step-by-step solvent selection/design algorithm. 2.1. Reaction-solvent properties Properties have an important role in the selection and evaluation of solvents that may be added to a reaction system. The reaction-solvent properties considered in this work, are described below. 2.1.1. Reactivity of the solvent For those situations where the solvent has an indirect inuence on the reaction, the selected solvent must be neutral to the compounds (reactants, products and inerts) present in the reacting system. One option for determining solvent neutrality is to screen a database of reactions and very quickly determine if any of the known reactions in the database includes the candidate solvent as a reactant or product for any of the compounds present in the reacting system. This option, however, does not guarantee that all possible reactions have been checked. A better but calculation intensive option is to

set-up a stoichiometric reaction (based on atomic balance) and then calculate the chemical equilibrium constant for the reaction (Smith & Missen, 1982), using the pure component heats of formation at a specied temperature (or a default temperature of 298 K). If the chemical equilibrium constant is very small, no reaction is possible (that is, a catalyst becomes necessary to promote the reaction see also verication of reactivity). This calculation will ensure that the solvent itself will not react. For direct inuence of reactions, the solvent may act as a carrier of a catalyst or as a provider (of O, H, etc.). In this case, a detailed understanding of the reaction (kinetic) mechanism is necessary. In the rst version of this methodology, this option will be considered only if data for such reaction-solvent systems are available in the database. 2.1.2. Solvent must be liquid at the reaction conditions Boiling point, melting point and vapor pressure of the solvent and the reaction systems compounds will verify whether the solvent will be in the liquid phase at the reaction conditions. 2.1.3. Phase split The selected solvent may form another liquid phase or promote the precipitation of a product(s). This can be calculated very quickly by checking the condition of liquid phase stability (Michelsen, 1982) for all binary solvent-reactant pairs. If the condition for liquid phase stability is not satised, then the solvent in the presence of the reactant will cause a phase split, which could be a second liquid phase or a solid phase (causing a compound to precipitate). If the condition for liquid phase stability is satised, then the addition of the solvent will not cause a phase split. A similar analysis can also be carried out with the solvent-product pairs. 2.1.4. Solubility This needs to be checked for reactants (where the solvent is the carrier of one or more reactants), products (where the solvent is not miscible with the reactants but only with products) or catalyst (where the solvent is the carrier for the catalyst being used). A quick evaluation of the solubility can be made through the Hildebrand solubility parameters (Hildebrand, 1936), which will help to indicate which compounds are likely to dissolve a compound present in the reaction system and which compound is likely to cause a phase split. That is, the rst selection of the solvent could be made with this property and then veried through calculations of miscibility and solubility. A more accurate but time-consuming calculation of solubility can be made through models for estimation of liquid phase activities. For the purposes of this work, the Hildebrand solubility parameters are used in the generation of solvent candidates while the models for liquid phase activities are used for validation related calculations. 2.1.5. Heating/cooling properties Solvents may be used in the reaction as carriers of heat. For example, to supply heat to an endothermic reaction or to

1664

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

absorb heat to an exothermic reaction. In this case, it may be advantageous if the solvent is totally miscible in a carrier (for example, water) that may already be present in the reaction system. 2.1.6. EHS properties The solvents could be classied apriori in terms of EHS properties, and these properties may be used as preliminary screening criteria. Examples of these properties are LC50 or log KOW or specic EHS regulations that might restrict the use of a given solvent in the reaction system. 2.1.7. Association/dissociation Another property of the solvent to check under reactivity is that the solvent should not associate/dissociate (for example, ionic compounds), polymerise (compounds with double/triple bonds) and/or oligomerise (for example, isomers) under the conditions of the reaction system. For known chemicals, this data could be retrieved from a specially developed database containing a list of solvents that are likely to associate/dissociate, polymerise or oligomerise. For generated chemicals, this problem is avoided by not selecting building blocks (for molecular structure generation) that includes ionic compounds and/or double/triple bonds. 2.1.8. Selectivity If the reactant(s) as well as the product(s) are soluble in the selected solvent, the solvent should be more selective towards the product(s) than the reactant, as in the case of a non-isolation reaction that is to be followed by additional reactions that may be carried out in the same reaction vessel. In other situations, it may be better to make the product completely or mostly insoluble in the solvent system. For example, adding a solvent/anti-solvent mixture where the solvent dissolves the reactants and the anti-solvent promotes the precipitation of the product as is the case for the production of Ibuprofen (Karunanithi, Achenie, & Gani, 2004). Properties such as viscosity, polarity, H-bonding, heat of vaporization, etc., may also help to identify (or select from) potential solvent candidates. 2.1.9. Indicator for solvents The atom economy and reaction mass efciency dened by Constable, Curzons, and Cunningham (2002) could be used as an indicator for the need of solvents in the reaction. If the reaction mass efciency is less than 80%, addition of solvents may have a positive effect to the reacting system. 2.2. Computer-aided tools 2.2.1. Databases and reaction-solvent tables Examples of two databases developed for the solvent selection/design methodology are given here. The rst database provides a list of common list and their solvent-reaction related properties while the second database contains collected reaction data.

A database containing pure component properties for well-known solvents that are likely to play a role in reactionsolvent selection has been developed. The partial list of properties is given for 75 most common solvents in Table 1. This list of solvents has been obtained through discussion with several industrial companies and identifying the 75 most common single organic solvents. Note also that not all properties are given for all solvents. A database of reactions with reaction parameters likely to play a role in reaction-solvent selection has also been developed (Hostrup, 2002). This database contains data such as reaction type, reactive agent used, reaction phase, solvent impact, and many more. A sample of the data that can be found in this database is highlighted through Table 2. 2.2.2. Calculation tools (software) In addition to the databases listed above, a number of computer-aided methods and tools are needed. All the calculations discussed above (Section 2.1) can be performed through the Integrated Computer Aided System, ICAS (ICAS, 2001). Some of the important tools of ICAS that are used by the solvent selection/design methodology are briey described below (see also http://www.capec.kt.dtu.dk/ Software/ICAS-and-its-Tools/): Verication of the reactivity (neutrality) of the solvent with the reaction compounds (checked through the chemical equilibrium K-value calculation) needed data: heats of formation, which if not available, can be estimated (Marrero & Gani, 2001). Note that the log(pKa ) and the chemical equilibrium constant can be calculated from knowledge of reaction stoichiometry and the heats of formation. The pure component data are estimated through the Marrero and Gani method (ICAS, 2001), which is available as a toolbox in ICAS-ProPred (ICAS, 2001). Pure component properties (melting point, boiling point, Hildebrand solubility parameter, etc.) of solvent candidates as well as the reaction system compounds can either be retrieved from the CAPEC database or predicted through ICAS-ProPred (ICAS, 2001). Prediction of the equilibrium composition (based on ideal system assumption) or from known reaction (kinetic) rate model/data. Impact of solvent would be obvious if more reactants or fewer products remain in solution (assuming chemical equilibrium and the calculated pure component properties from above, the equilibrium composition can be calculated). ICAS (2001) provides a calculation option for this. Note that as this calculation needs to be repeated for each solvent candidate, it is performed only on the nal list of solvents. Creation of another phase (liquidliquid phase stability check veries this condition). Solvents that perform a phase split can be found through database search or through computer-aided molecular design techniques. The CAPEC database (Nielsen, Abildskov, Harper, Papaeconomou, &

Table 1 List of well-known solvents together with their related properties Solvent name CAS number Solvent classication Type 1,1,1-Trichloroethane 1,2-Dichloroethane 1,3-Dioxane 1,2-Dimethoxyethane (ethylene glycol dimethyl ether) 1-Butanol 1-Methyl-2-pyrrolidinone 1-Propanol 2-Butanol 2-Ethylhexanol 2-Methoxyethanol 2-Propanol Acetone Acetonitrile Benzene Bis(2-methoxyethyl) ether Butyl acetate Butyl cellosolve Carbon tetrachloride Chlorobenzene Chloroform Cyclohexane Cyclohexanol Cyclohexanone Dibasic esters (DBE) Dichloromethane Diethyl ether Diethylene glycol monobutyl ether Diisopropyl ether Dimethyl carbonate Dimethyl ether Dimethylpropyleneurea (DMPU) Dimethylsulphoxide (DMSO) Ethanol Ethyl acetate Ethyl benzene Ethylene glycol Fluorobenzene Formamide Furfural Heptane Hexane 000071-55-6 000107-06-2 000505-22-6 000110-71-4 000071-36-3 000872-50-4 000071-23-8 000078-92-2 000104-76-7 000109-86-4 000067-63-0 000067-64-1 000075-05-8 000071-43-2 000123-86-4 000056-23-5 000067-66-3 000110-82-7 000108-93-0 000108-94-1 000075-09-2 000060-29-7 000112-34-5 000108-20-3 000616-38-6 000115-10-6 000067-68-5 000064-17-5 000141-78-6 000100-41-4 000107-21-1 000075-12-7 000098-01-1 000142-82-5 000110-54-3 Halide Halide Ether Ether Alcohol Amide Alcohol Alcohol Alcohol Alcoholether Alcohol Ketone Nitrile Aromatic-HC Ester Chloride Chloride Cyclic-HC Alcohol Ketone Chloride Ether Etheralcohol Ether Ester Ether S-oxide Alcohol Ester Aromatic Alcohol Amide Etheraldehyde Alkane Alkane Polarity PNA PNA PNA PNA PA PA PA PA PA P-MF-A PA PNA PA NF PNA PNA PNA NF PA PNA PNA PNA P-MF PNA PNA PNA PNA PA PNA NF PNA PA P-MF NF NF H-bonding HBD HBD HBA HBA Charge NE/P NE/P NE/EPD NE/EPD E/N NE/EPD E/N E/N E/N E/N E/N E/NPG E/NPG NE/P NE/EPD E/PG NE/P NE/P NE/I E/N E/NPG NE/EPD NE/EPD NE/EPD NE/EPD E/PG NE/EPD E/NPG E/N E/PG NE/P E/N E/NPG NE/P NE/I NE/I 347.15 356.65 378.15 358.15 390.85 475.15 370.35 372.65 457.75 397.25 355.45 329.15 354.75 353.15 399.25 349.95 334.25 353.85 433.95 428.55 313.15 307.65 504.15 341.65 363.65 248.35 462.15 351.35 350.25 409.25 470.45 493.15 434.85 371.65 341.85 242.75 237.65 228.15 215.15 183.35 249.15 147.05 158.45 203.15 188.05 183.65 178.35 229.35 278.65 195.15 250.15 209.55 279.75 298.55 242.15 178.05 156.85 205.15 181.35 273.65 131.65 291.65 159.05 189.55 178.25 260.15 275.7 236.65 182.55 177.85 10.09 17.31 11.89 5.69 20.34 15.65 11.96 21.16 21.01 19.58 13.43 12.67 31.33 17.255 20.260 20.67 17.674 23.354 23.159 24.452 22.541 20.038 23.203 23.408 19.730 24.050 18.730 17.586 17.55 18.92 16.757 23.743 20.139 20.373 15.423 21.034 14.451 20.352 15.123 26.748 26.133 18.345 17.985 33.703 38.567 23.610 15.2 14.9 3.12 2.69 1.31 1.65 1.74 1.25 1.30 1.62 3.41 1.06 1.17 1.23 1.26 3.24 3.52 3.40 2.84 2.68 2.17 2.23 2.24 1.89 3.05 2.53 1.00 2.54 1.58 0.47 0.22 0.90 0.35 0.27 0.55 2.88 0.84 0.05 0.24 0.33 2.31 1.76 2.74 1.83 3.15 1.19 0.85 1.33 0.95 0.57 1.52 0.24 0.90 1.45 0.32 0.79 3.44 1.23 1.57 0.45 4.87 4.02 NBP (K) NMP (K)
0 Log Kp,f (at 300 K)

Sol. Par

LC50

Log P

Phase split in aqueous solution 4 4 0 1 3 0 0 2 0 0 0 0 0 5 4 5 4 5 1 1 2 2 0 3 5 1 1 0 1 5 0 R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

26.10 28.27 28.99

N HBA HBD HBD N HBA HBD HBA HBA HBA HBA HBA HBA N HBA

30.19 26.58 18.41 22.26

29.23 56.88 22.84 52.92

0.85 2.63 3.32

N N

1.60 0.13

3.06 3.80 3.47 3.02

1 5 5

1665

1666

Table 1 (Continued ) Solvent name CAS number Solvent classication Type Isoamyl alcohol Isobutanol Isopropanol Isopropyl acetate Mesitylene [1,3,5-trimethylbenzene] Methane sulphonic acid Methanol Methyl acetate Methyl cellosolve Methyl cyclohexane Methyl ethyl ketone Methyl isobutyl ketone Methyl tertbutyl ether m-Xylene n,n-Dimethyl formamide n,n-Dimethylacetamide N,N-Dimethylaniline n-Butanol n-Propyl acetate o-Xylene Pentane Petroleum ether Propionic acid p-Xylene Pyridine sec-Butyl alcohol t-Butyl acetate tert-Butyl alcohol Tetrachloroethylene Tetrahydrofuran Toluene Trichloroethylene Triethylamine Water Polarity H-bonding Charge NBP (K) NMP (K)
0 Log Kp,f (at 300 K)

Sol. Par

LC50

Log P

Phase split in aqueous solution

000067-63-0 000108-21-4 000108-67-8

Alcohol Ester Aromatic-HC

PA PNA NF

HBA N

E/N E/PG NE/P

355.41 361.65 437.89

185.28 199.75 228.42

-20.68

23.41 17.15 17.96

1.17 2.95 3.94

0.05 1.13 3.47

0 1 5

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

000067-56-1 000079-20-9 000108-87-2 000078-93-3 000108-10-1 001634-04-4 000108-38-3 000068-12-2 000127-19-5 000121-69-7 000109-60-4 000095-47-6 000109-66-0 000079-09-4 000106-42-3 000110-86-1 000540-88-5 000127-18-4 000109-99-9 000108-88-3 000079-44-8 000121-44-8 007732-18-5

Alcohol Ester Cyclic-HC Ketone Ketone Ether Aromatic-HC Amide Amide Amine Ester Aromatic, HC Alkane-HC Acid Aromatic, HC Amine Ester Chloride Ester Aromatic-HC Chloride Amine Aqueous

PA PNA NF PNA PNA PNA NF PA PA PA PNA NF NF PA NF PA PNA PNA PNA NF PNA PA

HBA N HBA HBA HBA N

E/N E/PG NE/I E/NPG E/NPG NE/EPD NE/P NE/NPG NE/NPG E/PA E/N E/PG NE/P NE/I E/PG NE/P E/PA E/PG NE/P NE/EPD NE/P NE/P NE/PA E/N

337.85 330.09 374.08 352.79 389.65 328.35 412.27 426.15 438.15 466.69 374.65 417.58 309.22 414.25 411.51 388.35 368.25 394.45 338.15 383.95 360.35 362.15 373.15

175.47 175.15 146.58 186.48 189.15 164.55 225.3 212.75 253.15 275.6 178.15 247.98 143.42 252.45 286.41 231.55 217.86 250.85 164.85 178.25 188.45 158.45 273.15

28.25

29.59 19.35 16.06 18.87 17.43 15.07 18.05 23.955 22.353 20.06 17.89 18.39 14.40 19.41 17.90 21.57 16.23 18.95 18.97 18.32 18.79 15.17 47.81

0.40 2.18 3.00 1.68 2.45 2.04 3.71 3.08 1.15 3.11 3.07 3.71 2.57 2.92 3.71 2.54 3.23 4.33 1.55 3.47 3.43 2.49

0.82 0.19 3.82 0.30 1.36 1.02 3.33 0.97 0.82 2.43 1.27 2.93 3.36 0.33 2.96 0.68 1.66 3.57 0.45 2.58 2.50 1.52

20.80

5 0 3 1 5 0 0 3

HBA N N

21.36 1.31

5 3 0 5 0

-21.19 33.17

HBA HBD HBA N HBA

21.32 19.43 39.79

5 1 5 2 1

Meaning of terms used in Table 1 polarity: PNA, polar and non-associating; PA, polar and associating; NF, normal uid; P-MF-A, polar (multifunctional) associating; P-MF, polar (multifunctional); Hbonding: HBD, hydrogen-bond donor; HBA, hydrogen-bond acceptor; N, neutral; charge: NE, non-electrolytic solvent; E, electrolytic solvent; P, polarizable; EPD, electron pair donor; I, inert; PG, protogenic 0 (proton donor); N, neutral (donor and acceptor); NPG, non-protogenic (proton acceptor); NBP, normal boiling point; NMP, normal melting point; log Kp,f is the chemical equilibrium constant of formation at 300 K = Gf /(2.303 RT); Sol Par, Hildebrand solubility parameter at 300 K (MPa2 ); LC50 , the aqueous concentration causing 50% mortality in fathead minnow after 96 h; log P, octanolwater partition coefcient.

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

Wide range

1667

Gani, 2001) can be used for known chemicals while the ICAS-Utility toolbox (ICAS, 2001) can be used for phase split and liquidliquid equilibrium calculations. Solubility of reactants/products can be determined through the calculation of liquid phase innite dilution activity coefcients. ICAS-Utility (ICAS, 2001) provides options to calculate these properties. A list of feasible solvent candidates satisfying all the desired property constraints can be generated through the search engine of the CAPEC database (Nielsen et al., 2001) and/or through ICAS-ProCAMD, which is a toolbox in ICAS for computer-aided molecular design (see ICAS, 2001). 2.3. Step-by-step solvent selection/design algorithm A systematic rule-based methodology has been developed to identify a suitable solvent for a specied reaction system. As highlighted in Fig. 2, the step-by-step algorithm consists of a sequence of steps through which the user is guided towards the solution. The rules employed in step 3 have been derived based on a combination of industrial practice, physical insights and generated property data. These rules are listed in Table 3. A scores table has also been developed so that scores can be assigned to the feasible solvent candidates. Table 4 lists the scores used. As illustrated in Fig. 2, for a given reaction system (supplied or retrieved from a database), a set of reaction indices (R) are generated. Using the reaction-solvent property table, a corresponding set of RS indices are generated. The solvents that match the RS indices are then listed as feasible solvents for further detailed study (for example, experimental verication). Step 1: Specify reactions and retrieve reaction data. Identify the reaction from the reaction database (Table 2) through the known reactants and or products. Generate also all the possible reaction paths (if necessary). If the desired reaction is not in the database, insert the corresponding data into the database. For each reaction, dene the reaction-solvent properties that will be considered for the reaction and retrieve the available reaction-solvent data. Step 2: Generate the values of R-indices. Based on the retrieved data from step 1, assign values to the reaction Rindices, using the data included in the solvent knowledge base (Table 1) and the reaction-solvent rules (Table 3). Step 3: Assign the reaction-solvent RS-indices. Following the reaction-solvent rules (Table 3), calculate the reactionsolvent indices (RS). Consult the known solvent database and identify the set of solvents that satisfy the reactionsolvent properties within 10% (score 10), 20% (score 8), 30% (score 6), 40% (score 4), 50% (score 2). The score for values outside the 50% range is one. Note that for index 9 (RS9j or S9j ), there are two subscripts i and j, where subscript i identies the index number (for example, 9) while subscript j indicates the identity of the EHS property.

PR (atm)

1300

300500

300600

300500

TR (K)

Low <373

Wide +ve ve HCL Liquid, gas Yes Alkylation Acylation Olens, alcohols, alkyl sulfates, alkyl halides Hydrocarbons Yes Liquid, gas ve
a

High

Wide +ve Carrier for catalyst, condensation capabilities Remove product

High

Dissolve reactant Remove product Dissolve reactant (agent)

Solvent

Impact

Remove product

+ve ?? +ve

+ve

+ve

ve (+ve for olens and hydroC) ve

HR

ve ve ve

+ve

H2 O HCL (also H2 O) H2 O, HCL, NH4 X (X is halide)

H2 O, (plus carbon-oxides)

ve

H2 O, (alcohol, HCL)

ve Alcohols

Producta

+ve

+ve

Create 2nd liquid phase, remove product Remove product

Remove product so it cannot be oxidised

Carrier for NH3

Function

150

Table 2 Reactions table

Amination (ammonolysis) Oxidation

Reaction type

Amination Halogenation Sulfonation

Hydrogenation

Esterication

Hydrolysis

Indicates only one of the reaction products that is usually found in the corresponding reaction type.

Catalyst

Liquid Liquid, gas Liquid, gas

Yes

Yes

Yes

Yes Yes Yes

Yes

Acid, alkaline solution CL, HCL SO3 (with H2 SO4 ), SO2 Cl2 sultes, acid sulfate NH3

O2 , alkaline, neutral and acid solutions, peroxides, ferric salts, copper salts H2

Agent

Reduction of ethylene-linkage, carbonyl to OH, carboxyl to OH, aromatic to cyclic Substitute a group (e.g. C2 H5 ) for the ionizable H of an acid Double decomposition of a compound (OH and O) Add alkyl radical into an organic compound Synthesis of ketones

Remove/reduce O atom Add halogen atoms Link SO3 H to C or N atom

Reduce O or add H (NH2 , NH3 )

Action

Add O (or remove H)

Acid, NH3 , acid chlorides, metal-salt, alkali-halides H2 O, acid, alkali, enzyme

Yes

Liquid, gas

Liquid, gas

Liquid, gas

Liquid, gas

Phase

Liquid, gas

1668

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

Fig. 2. Illustration of the main steps of the solvent selection methodology.

Table 3 Rules for calculating the reaction index (Ri ) and reaction-solvent indices (RSi ) Rule Rule A: Establish need for solvents Reaction index calculation (Ri ) If the stoichiometry of the reaction is known and the atom economy and the reaction mass efciency are acceptable, set R0 = 1 Otherwise, set R0 = 0 If R0 = 1, calculate the atom economy and reaction mass efciency If the reaction mass efciency is >80%, solvents might not be necessary, set R0 = 0 Otherwise, use of solvent is desirable, set R0 = 1 If R0 = 1, and if reactions must be occurring in the liquid phase, set R1 = 1 Otherwise, set R1 = 0 R2 = specied reaction temperature Reaction-solvent indices calculation (RSi ) NA

Rule B: Liquid phase reactions Rule C: The solvent must be liquid at the reaction temperature

NA

Rule D: Need for solvent as carrier

If one or more of the reactants are solids (or large molecules) set R3 = 1 Otherwise, set R3 = 0

Rule E: Need for solvents to remove reactants or products

If one or more of the products are solids (or large molecules), set R4 = 1

Remove solvents from the candidates-list that are not likely to be a liquid at R2 20 K, where R2 is the specied reaction temperature. That is to say, the melting point of the solvent must be lower, and the boiling point must be higher, than the reaction temperature. This rule is implemented as follows: retrieve the boiling point (Tb ) and melting point (Tm ) temperatures of each solvent and determine the average temperature (Tb + Tm )/2 = TS . Based on the calculated value of TS , assign the corresponding RS2 values according to the following rules: RS2 = 1 if TS R2 = 5 K; RS2 = 2 if TS R2 = 10 K; RS2 = 3 if TS R2 = 15 K; RS2 = 4 if TS R2 = 20 K; RS2 = 5 if TS R2 = >20 K Note that values for RS2 are needed only if R1 = 1 If R3 = 1, solvents are needed as carrier for the reactant in the liquid phase. Assign RS values for solvents according to the following rules: RS3 = 1 if solvents are totally miscible RS3 = 2 if solvents are highly soluble RS3 = 3 if solvents are soluble RS3 = 4 if solvents are slightly soluble RS3 = 5 if solvents are not soluble If R4 = 1, solvents are needed to remove the product from the reacting phase. Assign RS values for solvents according to the following rules:

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676 Table 3 (Continued ) Rule Reaction index calculation (Ri ) Otherwise, set R4 = 0 Reaction-solvent indices calculation (RSi )

1669

Rule F: Need for phase split

If phase split is necessary, set R5 = 1

Otherwise, set R5 = 0 Rule G: Matching of solubility parameters of solute and solvent The solvent must have a solubility parameter value, which is within 5% of the key reactant (if R3 = 1) or product (if R4 = 1) If R3 = 1 or R4 = 1, then R6 = 1 Otherwise, R6 = 0

Rule H: Neutrality of solvents

If R3 = 1 or R4 = 1, set R7 = 1 if the solvent must be neutral to all compounds present in the reacting system Otherwise, set R7 = 0

Rule I: Association/ dissociation properties of solvents

If the solvent must not associate or dissociate, set R8 = 1

Otherwise, set R8 = 0

Rule J: EHS property constraints

If EHS properties are to be used as constraints, set R9 = 1

Otherwise, set R9 = 0 Since there are a number of EHS properties, each R9 and its corresponding RS9 index has a second subscript to identify the specic EHS property. For example, R91 corresponds to log P and R92 corresponds to LC50 and so on. Currently, there are 10 EHS properties. Not all 10 properties are used in any specic solvent selection/design problem

RS4 = 1 if solvents are totally miscible RS4 = 2 if solvents are highly soluble RS4 = 3 if solvents are soluble RS4 = 4 if solvents are slightly soluble RS4 = 5 if solvents are not soluble If R5 = 1, check for solvent partial solubility with respect to one of the compounds from the reacting system (reactant, product or carrier) If solvent is partially soluble, RS5 = 1, 2, 3 or 4 Otherwise, RS5 = 5 If R6 = 1, retrieve the solubility parameter values for the feasible solvents and assign the RS6 values according to the following rules: use the solvents data (from the solvents table) to assign the RS values RS6 = 1, if SPS = SP 5% RS6 = 2, if SPS = SP 10% RS6 = 3, if SPS = SP 15% RS6 = 4, if SPS = SP 20% RS6 = 5, if SPS > SP 20% If R7 = 1, check for the solvent pKa value for the feasible solvents and assign RS7 values based on the following rules: RS7 = 1, if pKa > 3 RS7 = 2, if 2 < pKa < 3 RS7 = 3, if 1 < pKa < 2 RS7 = 4, if 0 < pKa < 1 RS7 = 5, if pKa < 0 If R8 = 1, check for the solvent molecule type data (solvents table) for the feasible solvents and assign RS8 values based on the following rules: RS8 = 1, if solvent is non-polar RS8 = 2 or 3, if solvent is polar non-associating RS8 = 4, if solvent is associating RS8 = 5, if solvent is ionic If R9 = 1, the set goal values for log P, LC50 , ODP, GWP, etc. and retrieve the solvent values for the corresponding properties (from the solvents table) and assign the RS9 values according to the following rules: RS9 = 1, if S = 5% RS9 = 2, if S = 10% RS9 = 3, if S = 15% RS9 = 4, if S = 20% RS9 = 5, if S > 20% Where, is the goal value of a specic EHS property S is the corresponding solvent property

Use the solvent data (from Table 1) to assign the value for RSi , and use the scores table (Table 4) to set the scores for the feasible solvents.

Step 4: Assign scores to solvent candidates. Create a list of feasible solvents that satisfy all the selection criteria. The scores are assigned from the calculated values of RS indices, using the scale shown in Table 4. Any solvent having one or more scores of 1 for any solvent index is rejected as infeasible or unsuitable. Step 5: Final selection and verication. Perform detailed analysis through rigorous (quantitative) calculations to
Table 4 Scores table Variable RS Score (Si ) Value 1 10 2 8 3 6 4 4 5 1

verify the performance of the solvents in the specied reactions.

3. Application examples The application of the solvent selection methodology is highlighted through four illustrative examples. For the rst two examples, each step of the solvent selection methodology has been highlighted. For the third example, the published results from a solvent selection problem are validated through the reaction-solvent index values, verifying thereby, the rules used to assign their values. For the fourth example, a replacement of a known solvent is proposed. In all

1670

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

Table 5a Partial list of solvents with their RS-index values for example 1 Solvent Solvent scores for reaction at TR = 307 K RS2 1,1,1-Trichloroethane 1,2 Dichloroethane Benzene Cyclohexane 2-Heptanone Diisopropyl ketone
a

RS3 2 2 1 1 2 2

RS5 3 3 1 3 1 1

RS6 2 3 5 1 1 1

RS7 2 2 2 1 1 1

RS8 2 2 2 2 2 2

RS91

RS92 5 5 5 5 2 2

Source Table 2 Table 2 Table 2 Table 2 ProCAMDDatabasea ProCAMDDatabasea

1 1 1 1 1 1

3 3

Obtained through ProCAMD and search engine of CAPEC database.

these examples, it is assumed that the reaction temperature is kept constant and therefore, heat effects are neglected (not considered). 3.1. Example 1: biotransformation of toluene to toluene cis-glycol in aqueous bio-catalyst toluene + O2 = toluene cis glycol (water + bio catalyst) 3.1.1. Problem denition The objective is to create an organic phase, which will remove the unreacted toluene and recycle it back with fresh toluene. The addition of solvent creates a second organic phase in addition to the water phase. The product is soluble in water and it is difcult to separate toluene from the product. The operation is in the fed-batch mode. That is, for a given charge, react, separate unreacted toluene from product and, recycle toluene with fresh charge. 3.1.2. Problem solution Step 1: Reaction data retrieved from the reaction database Step 2: Apply the rules and use the known reaction data to: Assign the R-indices: Addition of solvent needed to separate the unreacted toluene from the product (R0 = 1). Reaction in liquid phase (R1 = 1). Reaction temperature (R2 ) is 307 K. Solvent for reactant is necessary (R3 = 1). Solvent for products is not necessary (R4 = 0). Phase split is necessary (R5 = 1). Solubility parameter for reactant (toluene) is known. Therefore, set R6 = 1 and SP = 18.32 MPa0.5 . Solvent must be neutral (R7 = 1). Solvent must not associate/dissociate (R8 = 1).
Table 5b Partial list of solvents with their scores for example 1 Solvent Solvent scores for reaction at TR = 307 K S2 1,1,1-Trichloroethane 1,2 Dichloroethane Benzene Cyclohexane 2-Heptanone Diisopropyl ketone 10 10 10 10 10 10 S3 8 8 10 10 8 8 S5 6 6 10 6 10 10 S6 8 6 1 10 10 10

EHS properties are desirable, therefore, R9 = 1. Dene upper and lower bounds for solvent properties: Normal boiling point Tb > 400 K. Normal melting point Tm < 250 K. Hildebrand solubility parameter 17.0 MPa0.5 < SP < 19.0 MPa0.5 . Partial solubility in water. Octanolwater partition coefcient log P < 4. Toxicity log LC50 < 2. Step 3: Three sets of solvents were obtained. One set was obtained by consulting the list of known solvents from Table 2. Only four of the solvents that match most of the desired properties are listed (see Table 5a). A second set of potential solvents were identied through a search of the CAPEC database, which is available in ICAS. A third set was obtained through ProCAMD toolbox in ICAS. The 2 compounds found by the CAPEC database as well as ProCAMD are also listed in Table 5a. It can be noted that none of the solvents is perfect (that is, having 1 for the RS-indices). However, the two solvents identied from the CAPEC database appear to be quite appropriate if the condition of log KOW (also called log P) can be relaxed. Step 4: Using the scales shown in Table 4 and the assigned RS-indices, the solvents from Table 5a have been assigned scores, as shown in Table 5b. Step 5: Final selection considering that the reactant is toluene, it is important that none of the reactant is lost with the product and therefore, a solvent that is totally immiscible with water and to which toluene is soluble would be the most desirable solvent. Also, as some solvents may be lost with the product (or in this case aqueous solution), the solvent must also have a high S92 score. The two best solvents are 2heptanone and diisopropyl ketone. Note, however, that their log KOW value is less than 4 but greater than 2.5.

S7 8 8 8 10 10 10

S8 8 8 8 8 8 8

S91

S92 1 1 1 1 8 8

Action Reject Reject Reject Reject Study Study

6 6

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

1671

Fig. 3. Ternary liquidliquid equilibrium phase diagram.

Verication of the solvent has been performed in terms of ternary liquidliquid phase equilibrium calculations. It has been conrmed that toluene and 2-heptanone are totally miscible with each other while water and 2-heptanone are very slightly miscible. This means that most of the toluene from the aqueous (product) solution will move to the organic (solvent) phase. Fig. 3 shows the calculated ternary liquidliquid equilibrium phase diagram (using the UNIFAC-LLE model parameters from ICAS-Utility toolbox). 3.2. Example 2: esterication of alcohols alcohol + acid = ester + water Esterication of alcohols with fatty acids produce fatty esters and water. The reaction usually takes place in the presence of a catalyst. The reaction can be promoted by either removing the reactants or products from the reacting phase. One way to remove the products is to cause evaporation of water and/or the ester product by the addition of a solvent (this option has not been considered in this example). Another option could be to nd solvents that separate the reactants from the products. To assist the removal of the water, certain liquids that are insoluble in water and distill with it at temperatures below its regular boiling point are frequently added to the esterication mixture (Reid, 1952). Note that to complete the esterication, water needs to be removed and usually, either the alcohol or the acid is in excess. 3.2.1. Problem solution Step 1: Reaction data retrieved from the reaction database. Step 2: Apply the rules and use the known reaction data to: Assign the R-indices: Addition of solvent needed to separate the reactants from the products (R0 = 1). Reaction in liquid phase (R1 = 1). Reaction temperature (R2 ) is 373423 K (at pressures of 5 atm < P < 10 atm).

Reaction produces water and acetate and therefore, removal of product is favorable (R3 = 0) and (R4 = 1), or, R3 = 1 and R4 = 0. Phase split is necessary (R5 = 1). Solubility parameter is not necessary (R6 = 1). Solvent must be neutral (R7 = 1). Solvent must not associate/dissociate (R8 = 1). EHS properties are desirable, therefore, R9 = 1. Dene upper and lower bounds for solvent properties: Normal boiling point Tb > 450 K. Normal melting point Tm < 300 K. Miscibility with water and acetate; partial miscibility with alcohol and fatty acid, or, vice versa. Toxicity log LC50 < 3. Step 3: Two sets of solvents were obtained. One set was obtained by consulting the list of known solvents from Table 2. Only four of the solvents that match most of the desired properties are listed (see Table 6a). These solvents were reported to be particularly suitable for esterication reactions (Reid, 1952). Another set of potential solvents have been generated through ProCamd in ICAS. From this, the two most appropriate ones are also listed in Table 6a. It can be noted that none of the solvents is perfect (that is, having 1 for all the RS-indices). However, the two solvents identied through ProCAMD have at least, better EHS properties. The problem formulation and solution (generation of solvent candidates) through ProCAMD is highlighted in Appendix A. Step 4: Using the scales shown in Table 4 and the assigned RS-indices, the solvents from Table 6a have been assigned scores, as shown in Table 6b. Step 5: Final selection considering that the products are water and ester, the solvents found are totally immiscible with water and highly soluble with at least one of the reactants, if not both. They are also quite stable (high log(pKa) values). The main difculty with the known solvents is that they are not environmentally friendly. The two solvents gen-

1672

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

Table 6a Partial list of feasible solvents with their RS values for example 2 Solvent Solvent selection parameters for reaction at TR = 373423 K RS2 Carbon tetrachloride Toluene Benzene Chloroform 2-Nonanone 2-Decanone 2 1 1 2 1 1 RS3 1 1 1 1 1 1 RS5 1 1 1 1 1 1 RS7 2 2 2 2 1 1 RS8 2 2 2 2 1 1 RS92 5 5 5 5 2 2 Solvent source Table 2 Table 2 Table 2 Table 2 ProCamd ProCamd

Table 6b Partial list of feasible solvents with their scores for example 2 Solvent Solvent selection parameters for reaction at TR = 373423 K S2 Carbon tetrachloride Toluene Benzene Chloroform 2-Nonanone 2-Decanone 8 10 10 8 10 10 S4 10 10 10 10 10 10 S5 10 10 10 10 10 10 S7 8 8 8 8 10 10 S8 8 8 8 8 10 10 S92 1 1 1 1 8 8 Action Reject Reject Reject Reject Study Study

erated through ProCamd have much lower toxicity, are quite stable and have similar solvent properties as benzene and toluene. Thus, they would be able to separate the reactants from products. Verication of the solvent was performed in terms of a two-phase reactor (two liquid phases where reaction takes place only on one; note also that in practice, a vapour product is also removed; the objective here is only to highlight the removal of water into another liquid phase as has been done before) calculation and the yield evaluated with and without the addition of the solvent. A case study has been made for methanol and acetic acid reacting to give methyl acetate and water. First the verication of the creation of two liquid phases is conrmed in Fig. 4(a) and (b). Then, the change in the conversion of methanol to methyl acetate as a function of a decrease in the water content in the reacting phase has been calculated. For 50% removal of water to another phase, a 20% increase in the conversion (compared to the base case) is obtained.

3.3. Example 3: validation of industrial problem solution The basic chemistry is described in Fig. 5, where the reaction step is followed by extraction (Vinson, 2002). The rst reaction is a peptide coupling between compounds A and B with diisopropyl carbodiimide (DIC) as a coupling agent and N-hydroxybenzotriazole as a catalyst. This liquid phase reaction runs in a solvent mixture of 1:1 dimethylformamide (DMF) and methylene chloride (MeCl) and because of limited solubility of reactant A, the reaction runs over several hours. The second reaction is a saponication that hydrolyzes the ethyl group in compound C with 2.5N sodium hydroxide. The current process calls for no isolation between the rst and second reaction, which is common in the pharmaceutical industry if there are no purity concerns. The second reaction is followed by an extraction in methylene chloride, leaving the product in the aqueous phase. The nal operation (not shown in Fig. 5) involves an isoelectric precipitation that isolates the product as a zwitterion. The diisopropyl urea (DIU)

Fig. 4. (a and b) Calculated ternary liquidliquid equilibrium phase diagrams (conrmation that 2-nonanone will create a two-liquid phase system).

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

1673

Fig. 5. Reaction chemistry details for example 3 (from Vinson, 2002). Table 7 Reaction-solvent index values for the best solvents from Vinson (2002) Solvent Solvent selection parameters for reaction (Fig. 1) at TR = 300 K RS2 1-Butanol 2-Butanol Diethylene glycol monobutyl ether 1 1 1 RS4 2 2 2 RS5 1 1 1 RS6 1 1 1 RS7 1 1 1 RS91 1 1 1 RS92 1 1 1

by-product of reaction 1 is somewhat soluble in water, which leaves DIU with the product throughout the precipitation step and necessitates additional back-extractions after the second reaction to purify the aqueous phase. Vinson (2002) has proposed a solution to the above problem. In this paper, the solvent candidates proposed by Vinson are analyzed through the solvent selection methodology. The solvent selection problem, as proposed by Vinson, involved the selection of a solvent that preferentially removes the DIU impurity from the rst reaction mixture so that the need for additional operation after the second reaction would be reduced. The reaction temperature TR was assumed to be 300 K, the solubility parameter of DIU was found to be between 2123 MPa0.5 and the solvent must be inert to compounds A, B, C and HCl. The solvent must also create a phase split in the aqueous reaction mixture. Applying the reaction-solvent selection methodology, the following R-indices are dened: R0 = 1; R4 = 1; R8 = 0; R1 = 1; R5 = 1; R9 = 1 R2 = 1; R6 = 1; R3 = 0; R7 = 1;

3.4. Example 4: replacement of dichloromethane as a solvent in oxidation reactions 3.4.1. Problem denition

Add solvent to the above reaction to provide an inert reaction medium by dissolving the reactants and products and by keeping the reaction temperature low. A very well known solvent is dichloromethane, which has good dissolving power for many organic molecules, has favorable physical properties (related to reaction-separation) and is inert towards many types of reagents and reaction conditions (Ogawa & Curran, 1997). Since, however, dichloromethane does not satisfy the requirements of good chemistry (Curzon & Constable, 1999), the objective is to nd a replacement. The solvent needs to dissolve 3-octanol and 3-octanone, must be inert and must be liquid at the reacting condition (temperature = 298 K). Also, it must have favorable EHS properties and must have a density close to that of water. 3.4.2. Problem solution Based on the above problem denition, a list of candidate solvents are generated through ProCamd. Fig. 6 shows a screen shot from ProCamd providing a breakdown of number of feasible chemicals that have been found by ProCamd. ProCamd generated 2751 + 2185 molecules out of which 215 were found to satisfy the specied solvent property constraints.

Based on the above R-indices, the rules and the list of solvents selected by Vinson, the corresponding RS-index values are assigned (see Table 7). Vinson (2002) concluded and also veried that 1-butanol or 2-butanol were the best solvents. Also, according to the classication of solvents in terms of EHS properties (Jim nez-Gonz lez et al., 2005), these solvents have desire a able EHS characteristics.

1674

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

Fig. 6. Screen shots from ProCamd highlighting the generation of the list of feasible solvents for (step 2) for example 4.

One of the feasible solvent candidates is 2-pentanone, whose properties as calculated by ProCamd is also shown in the screen shot shown in Fig. 6. It can be seen from Fig. 6 that 215 chemicals (solvent candidates) have been generated. From this list, some of the most promising solvents are: 2-pentanone, sec-butyl acetate, 2oxepanone, gamma valerolactone, 2-ethoxy ethylacetate and many more. Among these solvents, 2-pentanone appears to be the most environmentally friendly.

4. Conclusions and future work This research provides a proof of concept for the methodology that provides a systematic approach for apriori solvent selection and design. This methodology should be considered as a rst step towards using chemical properties and environment, health and safety aspects to identify solvents that should be examined with greater rigor and through quantitative calculations and focussed experimentation. It also demonstrates that current industrial practice can be combined with computational tools for property estimation to develop methods that can help both the expert (by providing an integrated set of tools and data needed for problem solution and verication) and non-experts (by providing a systematic step-by-step method and their corresponding tools). While it is true that the solvent selection tables with scores need to be generated for each specic reaction considered, the methodology opens the door to the possibility of searching for solvents substitutes based on desired functions and characteristics. It therefore has the capacity to point

chemists to those solvents that can be most productively investigated. As part of future work in this area, a search engine connected to the solvents database and the reactions database is being developed to extract the necessary data for a specic reaction, nd the corresponding solvents and create the reaction-solvents property table with the corresponding scores. The examples presented here illustrate how the search engine would nd the data needed to create these tables, but the validation of the model and methodology with real world examples is undergoing. Also contemplated for future research is the automation of the quantitative calculations of the method, the investigation of other scenarios for use of solvents in organic synthesis (for example, to promote reactions); the inclusion of additional parameters that inuence the solvent-chemistryengineering-EHS relationships (mass and energy transfers); the investigation of potential effects of non-organic solvents (e.g. SCF, electrolytes); and the nal integration of different methods and tools. Finally, the methodology has been developed for liquid phase reactions and homogeneous reacting (catalyst) systems. Much work is needed to further extend the methodology to heterogeneous systems. Also, the methodology has been tested with mainly low molecular weight compounds in reacting systems. It needs to be veried with larger and more complex molecules. As most of the needed properties can be predicted through the Marrero and Gani method (2001), we are optimistic about extension to larger and more complex molecules. Here, mixtures of solvents and anti-solvents that form a solution, may be needed and will be investigated.

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676

1675

Appendix A. Use of ProCAMD for solvent design ProCAMD is based on the hybrid CAMD technique (Harper & Gani, 2000). It can be used for a wide range of molecular and mixture design problems and has been extensively tested for solvent design. For examples 14, it has been used to generate a list of feasible solvent candidates. It uses functional groups as building blocks to construct chemically feasible molecules and then tests them for the specied property constraints. It has a builtin collection of property models so that the generation and test steps can be done within ProCAMD. More details on the use of ProCAMD as well as tutorial documents highlighting the use of ProCAMD in solvent design can be found at http://www.capec.kt.dtu.dk/Software/ICASTutorials/ICAS-Tutorials-Workshops (the documents on computer aided product design related tools). In the screen shots below, the problem formulation and solution for example 2 are shown.

For the problem formulated in example 2, a total of 17502 solvent candidates have been generated and tested. Note that the testing of the generated solvents also required liquid phase stability calculations with respect to water and solvent. The response time of 107.73 s is very fast considering all the property calculations that were necessary.

One of the compounds out of the 17,502 feasible solvent candidates can be found in the CAPEC database. This is 2nonanone. The calculated properties for this compound as obtained by ProCamd are shown below. ProCamd is also integrated with ProPred (for pure component property estimation) and all other properties can be calculated through this option as well as verifying the results of ProCamd (see the screen on the right column).

1676

R. Gani et al. / Computers and Chemical Engineering 29 (2005) 16611676 Constable, D. J. C., Curzons, A. D., & Cunningham, V. L. (2002). Metrics to Green chemistry which are the best? Green Chemistry, 4(06), 521527. Curzons, A. D., & Constable, D. J. C. (1999). Solvent selection guide: A guide to the integration of environmental, health and safety criteria into the selection of solvents. Clean Products and Processes, 1, 8290. Folic, M., Adjiman, C. S., & Pistikopoulos, E. N. (2004). The design of solvents for optimal reaction rates, CACE-18. In A. Barbosa & H. Matos (Eds.), Proceedings of the ESCAPE-14, pp. 175180. Gani, R. (2001). ICAS Documentations, CAPEC Internal Report, Technical University of Denmark, Lyngby, Denmark. Harper, P. M., & Gani, R. (2000). A multi-step and multi-level approach for computer aided molecular design. Computers and Chemical Engineering, 24(2), 10911097. Hildebrand, J. H. (1936). The solubility of non-electrolytes. New York: Reinhold. Hostrup, H. (2202). Integrated approach to computer aided process synthesis. PhD Thesis. Technical University of Denmark, Lyngby, Denmark. Jim nez-Gonz lez, C., Curzons, A. D., Constable, D. J. C., & Cunninge a ham, V. L. (2005). Expanding GSKs solvent selection guide application of life cycle assessment to enhance solvent selections. Journal of Clean Technologies and Environmental Policy, 7, 4250. Karunanithi, A. T., Achenie, L. E. K., & Gani, R. (2004). Optimal (solvent) mixture design through a decomposition based CAMD methodology, CACE-18. In A. Barbosa & H. Matos (Eds.), Proceedings of the ESCAPE-14, pp. 217222. Marrero, J., & Gani, R. (2001). Group contribution based estimation of pure component properties. Fluid Phase Equilibria, 183184, 183203. Michelsen, M. L. (1982). The isothermal ash problem. Part II. Phase split calculation. Fluid Phase Equilibria, 9, 2140. Nielsen, T. L., Abildskov, J., Harper, P. M., Papaeconomou, I., & Gani, R. (2001). The CAPEC database. Journal of Chemical Engineering Data, 46, 10411044. Ogawa, A., & Curran, D. P. (1997). Benzotriuoride: A useful alternative solvent for organic reactions currently conducted in dichloromethane and related solvents. Journal of Organic Chemistry, 62, 450451. Reid, E. E. (1952). In P. H. Groggins (Ed.), Estericationin unit processes in organic synthesis (pp. 597650). New York: McGraw-Hill (Chapter X). Smith, W. R., & Missen, R. W. (1982). Chemical reaction equilibrium analysis: theory and algorithms. New York: Wiley. Vinson, J. (2002). CAMD for solvent selection in industry-II. In L. E. K. Achenie, R. Gani, & V. Venkatasubramanian (Eds.), Computer-aided molecular design: theory and practice. The Netherlands: Elsevier.

ProPred is able to estimate the properties of 2-nonanone as well as check with its database for known experimental values. The calculated log(LC50 ) is 3.9, which is less than 4 (this is not shown in the above screen shot but is also calculated by ProPred).

References
Anastas, P. T., & Warner, J. C. (1998). Green chemistry: theory and practice. Oxford: Oxford University Press. Buxton, A., Livingston, A. G., & Pistikopoulos, E. N. (1997). Reaction path synthesis for environmental impact minimization. Computers and Chemical Engineering (Supplement), 21, S959S964.