Environ. Sci. Technol. 2008, 42, 1724–1730
Life Cycle Assessment of an Ionic Liquid versus Molecular Solvents and Their Applications Y I Z H A N G , † B H A V I K R . B A K S H I , * ,† A N D E. SAHLE DEMESSIE‡ Department of Chemical & Biomolecular Engineering, The Ohio State University, Columbus, Ohio 43210, and Clean Processes Branch, United States Environmental Protection Agency, Office of Research and Development, National Risk Management Research Laboratory, Cincinnati, Ohio 45268
Received June 11, 2007. Revised manuscript received November 21, 2007. Accepted December 4, 2007.
Ionic liquids (ILs) have been claimed as “greener” replacements to molecular solvents. However, the environmental impacts of the life cycle phases and comparison with alternative methods have not been studied. Such a life cycle assessment (LCA) is essential before any legitimate claims of “greenness” can be made and is the subject of this paper. The model IL selected is 1-butyl-3-methyl-imidazolium tetrafluoroborate ([Bmim][BF4]) and its use as a solvent for the manufacture of cyclohexane and in a Diels–Alder reaction was assessed. These uses are compared with more conventional synthesis methods. The results indicate that processes that use IL are highly likely to have a larger life cycle environmental impact than more conventional methods. Sensitivity analysis shows that the result is robust to errors and variation in the data. For cyclohexane synthesis, the industrial gas phase process is the greenest, but the three solvents compared for the Diels–Alder reaction showed comparable life cycle impact. Although ILs are not the most attractive alternatives, the result may change if their separation efficiency, stability and recyclability are improved. Because there are many kinds of ILs, with many applications, two examples are not enough to reach any general conclusions about the greenness of all ILs. However, the life cycle data and approach of this study can be used for evaluating the greenness of more kinds of solvents, processes, and emerging technologies.
1. Introduction Ionic liquids (ILs) are organic salts that have melting points less than or approximately 100 °C and contain only ionic species without any neutral molecules (1). They possess many attractive properties, such as negligible vapor pressure, good thermal stability, high heat capacity and thermal conductivity, and nonflammability. They have therefore been explored for many applications in the chemical, petroleum, and allied fields for reaction media, electroplating, acid-gas scrubbing, and desulfurization of transportation fuels, often motivated by the perception of their being “green solvents” (2, 3). The main property supporting the claim of ILs as “green solvents” is negligible vapor pressures. Conventional industrial solvents, most of which are volatile organic compounds (VOCs), * Corresponding author e-mail:
[email protected]; tel: (614) 2924904; fax: (614) 292-3769. † The Ohio State University. ‡ United States Environmental Protection Agency. 1724
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 5, 2008
often cause fugitive emissions, which are a major source of environmental pollution. Therefore, the replacement of volatile solvents with ILs is expected to prevent the emission of VOCs. Although ILs are studied primarily in research laboratories, BASF has developed a commercial process that uses ILs to remove acids from a reaction mixture and the French Petroleum Institute (IFP) has commercialized the Difasol process using ILs as reaction solvents for dimerization of butane to octane (4). Despite their popularity, there is significant uncertainty regarding potential environmental impacts of ILs including but not limited to toxicity, thermal decomposition and hydrolysis during usage, difficulty of recovery, and resistance to environmental degradation (5). Certain ILs are as toxic as phenol and more toxic than benzene (6). Thermal and hydrolysis stability of certain ionic liquids has also been problematic. For example, 1-alkyl-3-methylimidazolium halides, such as [Bmim][AlCl4] can decompose to halomethane and alkylimidazole (7, 8). ILs containing [PF6]- can be hydrolyzed to HF, POF3, etc. when in contact with moisture (9). Thus, it is reasonable to assume that ILs are at least as hazardous as their decomposed or hydrolyzed products (10). Previous research on ILs has focused mainly on developing the fundamental understanding of ILs and developing their application and disposal phases. Actually, many upstream processes in the life cycle of ILs do involve volatile and hazardous organic chemicals. A comprehensive environmental analysis and evaluation of greenness of any product or technology must also account for the product life cycle because the environmental impact may simply shift to other stages of the life cycle. Existing work at the interface of environmental life cycle analysis and Green Chemistry includes identification of the characteristics of Green Chemistry from an LCA point of view (11), and a method for comparing the greenness of two reaction pathways for synthesizing a chemical (12). A previous study on the life cycle aspects of ILs has focused primarily on energy consumption without calculating the environmental impact due to other emissions (13). This article presents the first comprehensive life cycle assessment of 1-butyl-3-methyl-imidazolium tetrafluoroborate ([Bmim][BF4]) and a discussion of its use as a solvent for the manufacture of cyclohexane and in a Diels–Alder reaction such as cycloaddition of cyclopentadiene. Other processes for the synthesis of cyclohexane are also considered, including a gas phase industrial process and production in water. For the Diels–Alder reaction, three solvents [Bmim][BF4], lithium perchlorate–diethyl ether mixture, and water are compared. The selected reactions may represent bulk and fine chemical synthesis. The study is intended to convey the importance of a comprehensive life cycle study before labeling a process or solvent as green.
2. Cradle-to-Gate LCA of ILs versus Conventional Solvents LCA is a powerful tool to systematically study the broad environmental implications of ILs, but many challenges exist because of the emerging nature of these products. First, LCA usually relies on information about resource use, emissions, and impacts throughout the life cycle. For emerging technologies, such data are difficult to find in any LCA database. Second, inventory data about fine and specialty chemicals are also difficult to find, even for industrial products. In this work, the inventory data are estimated from laboratory experiments based on engineering heuristics and judgment. 10.1021/es0713983 CCC: $40.75
2008 American Chemical Society
Published on Web 01/25/2008
FIGURE 1. Life cycle tree of [Bmim][BF4]. The emissions and downstream treatment are hard to estimate or model, and often are ignored or considered via sensitivity analysis. For analyzing the inventory, a standard impact analysis method was used (14). The results are “midpoint” indicators, and can be compared only within each category because the measurement unit for each category is different. The IL selected for this study, [Bmim][BF4], was just one example of the hundreds of ILs described in the literature. [Bmim][BF4] is chosen because it has the advantage of being stable toward air, allows for facile recycling of solvents and catalysts, and is relatively well described in the literature. Many existing studies have used [Bmim][BF4] and consider it less toxic than many other ILs (15, 16). It is also commercially available. The traditional solvents evaluated in this section are benzene, acetone, deionized water, and lithium perchloratediethyl ether mixtures (LPDE). Among these, LPDE is not common, but it provides high yield in the Diels–Alder reaction studied in Section 3. The comparison in this section does not consider the use phase of each solvent, but the resulting inventory could be used as a data module in various applications, as done in the next section. 2.1. Life Cycle Tree of [Bmim][BF4]. The typical production process of [Bmim][BF4] requires two steps (10): synthesis of [Bmim]Cl from butyl chloride and 1-methylimidazole; and replacing the anion Cl– with [BF4]-, either through metathesis of [Bmim]Cl and a [BF4]- containing salt (e.g., NaBF4) or via an acid–base neutralization by HBF4. Organic solvents, like acetone, toluene or CH3CN, are usually used as the reaction medium in both steps (17). The reactions can leave the ILs contaminated with unreacted materials, coproducts, and solvents at the end of each synthesis step (18), requiring
purification steps. For the first step, a polar organic solvent, such as ethyl acetate, is usually used to remove unreacted material from [Bmim]Cl. The workup in the second step usually involves extraction of [Bmim][BF4] with methylene chloride from the coproduct NaCl, followed by washing the extract multiple times with water and vacuum drying or sparging with inert gas (19). These separation processes can potentially generate significant amounts of organic waste and polluted water and may consume a large amount of energy.
From a life cycle point of view, the reaction of butylchloride, 1-methylimidazole, and NaBF4 is the last step in preparing [Bmim][BF4]. A life cycle study also requires all the upstream information. Because life cycle information is only available for common chemicals, and not for specialized or new chemicals, such as [Bmim][BF4] and some of its precursors, it becomes necessary to generate a life cycle tree that connects materials for which the inventory is available to the final product. Such a tree for [Bmim][BF4] is displayed in Figure 1. The root of this tree is the selected IL, [Bmim][BF4]. The leaves, surrounded by a bigger box, are materials whose data are available in LCA databases. If more than one production procedure is feasible for a specific chemical in the tree, the most widely used technique is included for the LCA. Required solvents and energy for the synthesis step are on the left side of each process step, and byproduct generated are shown on the right side. Inorganic chemicals, organic compounds, and energy sources are also shown. Mining data VOL. 42, NO. 5, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1725
FIGURE 2. Comparing the life cycle environmental impacts of solvents per kilogram. Upper bars on IL represents the sensitivity analysis by including 10% of process loss as estimated emission. of Na2B4O7 is not available and data for mining NaCl is used instead because both compounds may be evaporites of ancient sea or lake water and are mined with similar equipment. The focus of this work is on obtaining information about processes between the leaves and the root. This information includes energy consumption, solvent use, unreacted materials, byproducts, emissions, labor, and equipment. Any data that are not available in the literature were either estimated on the basis of laboratory measurement, or ignored, or a range of values was considered via sensitivity analysis. If the data were ignored, it resulted in the “best case scenario” for the selected IL. Use of electrical or steam energy and their process emission data were incorporated in the analysis. Information about other solvents was obtained directly from LCA databases or their life cycle inventory was developed by the same strategy used for [Bmim][BF4]. Labor and equipment data were ignored due to lack of data, as is commonly done in most LCA studies. The unreacted materials, byproduct, coproducts, and lost solvents could potentially become pollutants if they are not collected, treated ,or disposed of carefully. Quantitative values for such pollutants from processes are usually missing in the literature. To be conservative, two scenarios are considered: base case and sensitivity analysis of estimated intermediate emissions. The base case includes life cycle emissions from the leaves, energy, and solvents, but excludes the estimated intermediate emissions, making the impact quantified by this case biased in favor of the IL. Sensitivity analysis is used to incorporate the impact of the unknown intermediate emissions. Unreacted materials and coproducts are calculated from knowledge of the conversion rate and product yields, and the byproduct are determined via kinetics of the side reactions. Use of volatile organic solvents usually results in losses. If the amount of volatile organic solvent loss is unknown, 5% of total solvent used is assumed to be lost in each cycle. The unreacted materials, useless coproducts and byproduct, and solvent loss are collectively referred to as “process loss”. 2.2. Comparison of ILs with Molecular Solvents. LCA data of water, benzene, and acetone are available in the SimaPro software (20), and the impacts can be calculated directly. However, LPDE is not found in any life cycle database, and its LCA calculation applies the same strategy as for [Bmim][BF4], as described in the Supporting Information. The impacts of the solvents per unit mass are displayed in Figure 2. Because of the wide variation of impacts in each category, all impacts are normalized by the impact of the [Bmim][BF4] base case. Although the original values are not 1726
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 5, 2008
shown in the figure, this does not affect the comparison because these midpoint indicators can only be compared within each category. The sensitivity bars for the IL represent sensitivity of the results if 10% of the process loss is emitted to the environment. This amounts to 0.5% of the total solvents and 10% of unreacted material and products, other than ionic liquid, being emitted. Such a sensitivity analysis is not deemed to be necessary for benzene and acetone because their data are from the commercial database in SimaPro, which is assumed to be more reliable. Figure 2 shows that [Bmim][BF4] has the most significant environmental impacts in most categories, whereas LPDE dominates POCP, and benzene dominates ODP. LPDE, benzene, and acetone have smaller impacts than [Bmim][BF4] in most categories. Deionized water has the smallest impact and is not visible in the figure. This result seems to be correlated with the length of the pathway from natural resources to the selected high grade solvent. Sensitivity analysis of [Bmim][BF4] worsens its impact, and especially changes the VOC emissions, POCP, and some ecotoxicity categories. This is because most of the emissions from intermediate processes are organic compounds. Making [Bmim][BF4] requires large quantities of organic materials and solvents, as shown in Figure 1, which is worth further investigation. The life cycle values of VOC emissions, use of organic solvents, and fossil fuel consumption per kilogram of the selected solvents are shown in Figure 3. Sensitivity analysis is only done for VOC emissions of IL because information about solvent use and material consumption is obtained from knowledge about their chemistry, which is more reliable than knowledge about VOC emissions. Figure 3a shows that even if VOC emissions from intermediate processes in the life cycle of the IL are ignored, this life cycle seems to generate more than twice the amount of VOCs as compared to VOCs from the life cycles of benzene or acetone. Figure 3b shows that IL uses significantly larger quantities of organic solvents than benzene or acetone. Washing processes of [Bmim]Cl and [Bmim][BF4] constitute the largest use of organic solvents, which may not be avoidable in present practice. These results indicate that claims about ILs such as [Bmim][BF4] being “greener” because of their causing less VOC emissions may be misleading since the use of organic solvents simply seems to shift to earlier steps in their life cycle. The fossil fuel consumption of [Bmim][BF4] is also higher than the requirement for benzene or acetone. Because most ILs are synthesized with a long supply chain involving organic compounds and require purification steps, it can be expected that other ILs may also have adverse environmental impact as demonstrated in this work for [Bmim][BF4]. However, this high environmental impact of
FIGURE 3. (a) VOC emission, (b) consumption of organic solvents, and (c) consumption of fossil fuels in the life cycle of [Bmim][BF4], benzene, and acetone (kg/kg).
FIGURE 4. Comparison of life cycle environmental impacts of cyclohexane production processes. Lower sensitivity bars represent 20 times of solvent reuse, upper bars represent additional 10% of process loss. the synthesis phase of ILs could be counter balanced in their use phase, due to their potential to increase yields of desired reaction products and the ease of recycling ILs.
3. LCA of Processes Based on Ionic Liquids versus Conventional Processes The use of [Bmim][BF4] for cyclohexane production and for a Diels–Alder cycloaddition reaction are selected for comparing the life cycle aspects of solvent use. Cyclohexane is a bulk chemical with a long production history, while the Diels–Alder cycloaddition reaction has great importance for synthetic chemistry since it is the primary tool in carboncarbon bond formation (21). The solvent treatment after usage is not considered because of the lack of data. 3.1. Cyclohexane Production. This section focuses on three processes for making cyclohexane. The industrial vapor phase process that is based on the hydrogenation of benzene, and processes that use H2O or [Bmim][BF4] as a reaction medium. The main reaction is shown in eq 1, but at high temperatures, side reactions such as isomerization or ring opening of cyclohexane are favored and can impact final product quality (22). C6H6 + 3H2 f C6H12 ∆Hr ) -205.5 kJ/mol
(1)
Cyclohexane is commonly produced in large liquid- or gas-phase reactors using nickel based or homogeneous catalysts (22, 23). Several new solvent systems have been studied recently, such as biphasic aqueous/supercritical ethane reaction medium, ILs, mixtures of ILs and water, and supercritical carbon dioxide. In the industrial process, benzene is hydrogenated at high temperature (200 °C) and pressure (20 bar) in a Gibbs reactor to form cyclohexane using heterogeneous vapor phase
catalysis or homogeneous liquid catalysis (22). This process can provide greater than 99% yield, where a polishing reactor completes benzene conversion making the separation step fairly straightforward. Emission data (24) show that the largest emissions are evident during storage and due to fugitive emission. Cyclohexane synthesis in water or ILs were selected for comparison on the basis of the processes that are still under laboratory development. The catalyst, Ru(η6 C10H14)(pta)Cl2, was selected because its stability has been studied, and the turnovers remain effectively constant after a series of five batches in both solvents. Therefore, solvents were assumed to be used 5 times and the life cycle impact of the catalyst was ignored because the quantity used is relatively small, and they are usually easy to separate. The turnover rate for this catalyst is 170 mol (mol cat)-1 h-1 in water, and 206 mol (mol cat)-1 h-1 in [Bmim][BF4] with a reaction time of 1 h (25). In such a situation, the yield is about 50% in both water and IL, and the effluent from the reactor is a mixture of benzene and cyclohexane. For a realistic comparison, all methods must yield the same product purity. Benzene and cyclohexane form an azeotrope, and hence the separation process using acetone and water as assisting solvents was modeled in AspenPlus. Sensitivity analysis is used to understand the effect of changing the emission rate from intermediate processes in the life cycle tree and frequency of solvent reuse. The emissions are increased from the base case to 10% of process loss as done for the synthesis of [Bmim][BF4] in Section 2.1. Sensitivity to the number of times of solvent reuse is evaluated by considering 5–20 times reuse. The latter case forms the lower end of the sensitivity bar in Figure 4, whereas the 10% process loss case forms the upper end. Two scenarios are considered for the industrial process. To keep the life cycle VOL. 42, NO. 5, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1727
boundary consistent with the “solvent base” case, the “industrial base” case does not consider emissions from intermediate processes although they are available (24), whereas the upper error bars for industrial case include all the emissions. Figure 4 compares results of the three processes after normalization by the base case of the IL-based process. Comparing base cases, their impact sequence in every impact category is I(IL base) . I(H2O base) > I(ind base) where, I denotes the life cycle impact. Because the environmental impact of deionized water is quite small, the life cycle results of the water-based process do not change much for more reuse and the lower sensitivity bars almost overlap with the base case. However, for the IL-based process the lower sensitivity bars are much less than those for five times reuse. This implies that most of the impact of the ILbased process is from the IL itself, and a long reuse time may reduce the overall impact significantly. However, even with 20 times reuse, processes that use water or IL still exhibit a larger life cycle impact than the industrial base case. It seems that even if the solvents can be reused for more cycles, it is still very difficult for the wateror the IL-based process to have a smaller life cycle impact than the industrial process. This is because these solventbased processes achieve a lower yield than the industrial process, making it necessary to separate benzene and cyclohexane with the help of acetone. Only if the yield of the solvent-based processes can match or exceed the yield of the industrial process is it likely that the former will be environmentally superior. Sensitivity to emissions from processes between the leaves and root of the life cycle tree are shown as upper sensitivity bars in Figure 4. The major pollutants in these intermediate processes for the IL- and water-based life cycles are benzene, cyclohexane, and acetone, which increase HTP, POCP, and VOC impacts. The emissions of the industrial process are benzene, cyclohexane and other VOCs, and so the affected categories are the same as in solvent-based processes. Comparing the three processes, the IL-based process is more sensitive to emissions than the other two processes. This is because intermediate processes in the life cycle of industrial and water-based processes are simply related to the reaction of benzene and hydrogen, and water as solvent. They are all common chemicals with emission data available in LCI databases. In contrast, intermediate emissions of the ILbased process also include the emissions from the life cycle of ILs (Figure 1). The intermediate emissions during IL preparation are large, as shown in the IL sensitivity analysis at Section 2. This also significantly affects the IL-based cyclohexane production. The sensitivity analysis of emissions from intermediate processesshowsthatforhumantoxicitypotential(HTP),theupper bound of industrial production is higher than the solventbased processes. This may indicate that under some circumstances, cyclohexane production in water or IL could have a smaller life cycle HTP than industrial production. However, this larger HTP of industrial production is due to the high loss of benzene during cyclohexane synthesis, primarily because of storage and fugitive emissions. These emissions are ignored in the solvent-based processes because of difficulty in accounting for them. Nevertheless, the results in Figure 4 still indicate that the industrial process is most likely to have a smaller life cycle environmental impact than the water and IL-based processes. Thus, replacing the mature and high-yielding industrial cyclohexane production with newer solvent-based techniques may not be promising from an environmental life cycle viewpoint. 3.2. Diels–Alder Reaction. Fine and specialty chemicals are expensive, produced at relatively small scales (
