t on Supercritical and
      esis anjH Processing as
         rj^/j
|J~i;

-------
 Report on Supercritical and
Near-Critical CO2 in Chemical
Synthesis and Processing as
   Environmentally Benign
    Solvent Replacements
      EPA Order Number: 1W-0590-NASA
 Requisition/Reference Number: TM1150 QT-DC-01-001351
            Eric J. Beckman
      Bayer Professor of Chemical Engineering
         Associate Dean for Research
           School of Engineering
           University of Pittsburgh
             Pittsburgh, PA

-------
    Although the information in this document has been funded by the United
 States Environmental Protection Agency under contract number TM1150 QT-DC-
 01-001351 to University of Pittsburgh, it does not necessarily reflect the policy of
 the Agency and no official endorsement should be inferred. This document has
 been reviewed in draft by both internal EPA reviewers and external peer reviewers.
    Peer Reviews of the Report

    This document was reviewed in draft by both internal U.S. Environmental Pro-
tection Agency (EPA) reviewers and external peer reviewers who were chosen for
their diverse perspectives and technical expertise in environmental science and
engineering and in supercritical and near-critical CO, chemical topics. Dr. Richard
Engler and John Blouin from EPA's Office of Pollution Prevention and Toxics and
Dr. Endalkachew Sahle-Demessie from the Office of Research and Development
read the initial draft and provided written comments on the document.  Dr. Eric
Beckman addressed these comments and incorporated changes where necessary
into the draft in January 2002.

    Five external peer reviewers read the revised document and provided further
extensive written comments. The peer reviewers  included Dr. Martin Abraham,
University of Toledo; Dr. Joan Brennecke, University of Notre Dame; Dr. Joseph
DeSimone, University of North Carolina; Dr. Phillip Jessop, University of Califor-
nia, Davis; and Dr. Barbara Knutson, University of Kentucky. These comments
also were addressed and changes incorporated into the final document in May 2002.
Dr. Barbara Kara of EPA's Office of Research and Development, National Center for
Environmental Research, oversaw the project and reviews. Ted Just served as the
Contract Officer. Many thanks go to the efforts of the author and reviewers.

-------
                            Contents

Executive Summary	I
1    Introduction	7
     1.1  Supercritical Fluids	7
     1.2  Physical Properties of CO	9
     1.3  Environmental and Safety Advantages to Use of CO, in
         Chemical Processes	10
     1.4  Environmental and Safety Disadvantages Inherent to Use
         of CO, in a Process	11
     1.5  Chemical Advantages to Use of CO, as a Solvent	12
     1.6  Chemical Disadvantages to Use of CO, as a Solvent	15
     1.7  How We Will Approach Our Analysis	17
     1.8  Operating a Process Economically With CO,	18
     1.9  Scope of This Report	23
     1.10 A Note on Cleaning by Using CO,	24
     1.11  The Effect of Regulation on the Use of CO, in Green Chemistry
         and Chemical Processing	25
     Reactions Using Gases	29
     2.1  Hydrogenation	29
     2.2  Liquid-Phase Hydrogenations: Advantages to Use
         of Supercritical Sol vents 	29
     2.3  Heterogeneous Hydrogenation in CO,	31
     2.4  Homogeneous Hydrogenation in CO,	36
         2.4.1   CO,-Soluble Catalyst Design 	36
         2.42   Engineering Rationale for Homogeneous Versus
               Heterogeneous Catalysis	37
         2.4.3   Chemical Rationale for Homogeneous Catalysis	39
         2.4.4   Homogeneous Hvdrogenation and Material Synthesis	40
                                  111

-------
IV
          2.5  Industrial Activity: Hydrogenation in CO,	41
          2.6  Summary:  HydrogenationinCO,	41
          2.7  Hydroformylation in CO,	43
               2.7.1   Homogeneous Catalysis of Hydroformylation in CO,	44
               2.72   Heterogeneous Hydroformylation in CO,	46
               2.73   Industrial Activity: Hydroformylation in CO, 	46
               2.7.4   Summary: Hydroformylation inCO,	47
          2.8  Oxidation inCO,	47
               2.8.1   Oxidations in CO,: Experimental Results	48
               2.8.2   Industrial Activity: Oxidations in Supercritical Fluids	51
          2.9  Summary:  Gaseous Reactants in CO,	51
     3i   Polymerization and Polymer Processing	55
          3.1   Introduction	.•	55
          3.2  Polymerizations: General Background	55
          3.3  CO, as a Solvent for Polymer Systems	56
          3.4  Chain Polymerization and CO,	60
               3.4.1   Free Radical Solution Polymerization	60
               3.4.2   Heterogeneous Free Radical Polymerizations	61
                     3.4.2.1  Emulsion Polymerization in CO,	62
                     3.4.2.2  Dispersion Polymerization in CO,	63
                     3.4.2.3  Suspension Polymerization in CO,	64
                     3.4.2.4  CO, as Nonsolvent in Heterogeneous
                            Polymerizations	65
               3.4.3   Other Chain Polymerizations in CO,	65
               3.4.4   Industrial Activity: Chain Polymerizations in CO,	66
          3.5  Condensation Polymerizations	67
               3.5.1   Polyesters, Polyamides, Polycarbonates	67
               3.5.2   Polyurethanes	69

-------
3.6  Carbon Dioxide as a Monomer	70

3.7  Industrial Activity: Condensation Polymers and CO,
     as a Monomer	72

3.8  Postpolymerization Processing of Polymers Using CO,	72

3.9  Extrusion-Foaming Using CO,	74

3.10 Industrial Activity; Postpolymerization Processing	77

3.11  Use of CO, in Polymer Science Applied to the Microelectronics
     Industry  	77

3.12 Industrial Activity: CO, and Polymers in Microelectronics
     Manufacture	79

Other Reactions in CO,	81

4.1   Enzymatic Chemistry	82

4.2  Diels-Alder Chemistry	84

4.3  Lewis Acid Catalysis/Friedel-Crafts Chemistry 	85

4.4  CO, as Reactant and Solvent	86

4.5  Other Organic Reactions	87

4,6  Industrial Activity: Friedel-Crafts Chemistry and Other Name
     Reactions	88

4.7  Inorganic Chemistry: General 	88

4.8  Inorganic Chemistry: Metal Chelates	89

4.9  Inorganic Chemistry: Industrial Activity	92

Formation of Fine Particles Using CO,	95

5.1   Production of Particles Using CO,: RESS	95

52  Creating Fine Particles Using CO,: Nonsolvent Modes of
     Operation and PGSS  	.".	96

5.3  Production of Fine Pharmaceutical Powders:  Is This Green
     Processing?	97

5.4  Comparisons With Current Processes	98

5.5  Industrial Activity	100

-------
 6   Process Issues	 101

     6.1   Process Design Using Supercritical Fluids: Are CO,-Based
          Plants Inherently Uneconomical?	 101

     6.2   How Does One Economically Recover a Catalyst and/or
          Product From CO,?	 103

     6.3   Where Would Process Improvements Enhance Opportunities
          for Green Chemistry in CO,?	104

 7   Reactions at Interfaces and/or Multiphase Mixtures	 107

 8   Impact of the Technology for a Sustainable Environment (TSE)
     Program on Use of CO, as a "Green" Solvent	 Ill

     8.1   Description of Funded Projects in the TSE Program	 Ill

     8.2   Impact of the TSE Program	114

     8.3   Technology Transfer From TSE-Sponsored Programs	116

 9   Milestones in Green Chemistry Using CO,	119

10   Areas for Future Research on CO, Technology	123

11   References	125

-------
                  EXECUTIVE SUMMARY
    This review specifically examines the use of carbon dioxide (CO,) to create
 greener processes and products, with a focus on research and commercialization
 efforts performed since 1995. The literature reveals that use of CO, has permeated
 almost all facets of the chemical industry, and that careful application of CO, tech-
 nology can result in products (and processes) that are cleaner, less expensive, and
 of higher quality.

    Carbon dioxide is nonflammable and naturally abundant, with a Threshold Limit
 Value (TLV) lhat renders it less toxic than many organic solvents. Carbon dioxide is
 clearly a "greenhouse gas." but like water, if CO, can be withdrawn from the environ-
 ment, employed in a process and then returned to the environment "clean." no environ-
 mental detriment accrues. CO,'s combination of high TLV and high vapor pressure
 means that residual CO, left behind in substrates is not a concern with respect to human
 exposure—the same can certainly not be said to be true for many man-made and natu-
 rally occurring organic compounds. There is effectively no liability due to "residual"
 CO, in materials following processing—only water also enjoys this special situation.

    Chemical advantages to use of CO, as a solvent.  Carbon dioxide can pro-
 vide not only environmental advantages, but also chemical advantages when ap-
 plied strategically. First. CO, cannot be oxidized, and it is therefore particularly
 useful as a solvent in oxidation reactions. Next, cross-contamination of the other
phase during liquid-liquid extraction is not reallv contamination. Although CO,
 will "contaminate" an aqueous phase upon contact in a process, a mixture of CO,
 and water clearly does  not require remediation, unlike almost any other organic
 solvent. Further, CO, is generally immune to free radical chemistry. Carbon diox-
 ide does not support chain transfer to solvent during free-radically initiated poly-
merization, and is hence an ideal  solvent for use in such polymerizations. CO, is
miscible with gases in all proportions above 31°C. Gases  such  as hydrogen and
 oxygen are poorly soluble in organic  liquids and water, and  hence in  many two-
and three-phase reactors the rate is limited specifically by the rate at which the gas
diffuses across the gas-liquid interface. Liquid CO, can absorb much higher quan-
tities of H, or O, than typical organic solvents or water, and supercritical CO, is
completely miscible with such gases. Finally. CO, exhibits a  liquid viscosity only
 1/10 that of water. The surface tension in CO, also is much lower than that for
conventional organic solvents, and the diffusivity of solutes is considerably higher.
owing to carbon dioxide's low viscosity. Consequently. CO, would be expected to
wet and penetrate complex geometries better than simple liquids, and reactants
would be expected to diffuse faster within catalyst pores where CO, is the solvent.

    Chemical disadvantages to use of CO, as a solvent. Carbon dioxide exhibits
some inherent disadvantages where chemistry is concerned; some of these are unique

-------
 to CO,, and others are common to any number of solvents. First. CO, exhibits a
 relatively high critical pressure and vapor pressure. These characteristics guaran-
 tee higher capital costs for a CO,-based process relative to one using a conven-
 tional solvent, as well as the need for specialized equipment for laboratory work.
 These issues will not by themselves impede the commercialization of CO,-based
 processes, as many commercial processes operate at pressures above 10 MPa. Sec-
 ond, CO, is a Lewis acid.  When attempting to use amines as reactants. this can be
 a serious disadvantage, in that carbamate formation can slow the rate of the in-
 tended reaction and also  can alter the solubility characteristics of the substrate.
 CO, has been shown to react reversibly with a number of enzymes (lysine residues.
 specifically), leading to low activity in the presence of CO,. Unfortunately,  CO,
 can be hydrogenated in the presence of noble metal catalvsts to produce carbon
 monoxide (CO). If one is trying to hydrogenate a substrate in CO, over a heteroge-
 neous platinum catalyst, production of CO will poison the catalyst and produce
 toxic byproducts. There has been a certain degree of controversy recently as to
 what extent the same reaction occurs over palladium catalysts. Next, dense  CO,
 produces low pH (2.85) upon contact with water. This can render some biocatalytic
 and organic reactions problematic, in that many enzymes and catalysts are deacti-
 vated by low pH.  Finallv, CO, is a weak solvent. This is perhaps CO,'s greatest
 flaw, in that its inability to solvate compounds of interest at economical process
 pressures has inhibited its commercial use.

    Operating a process economically with CO,. Operation of any process at
 high pressure typically involves higher costs than the analogous process operated
 at one atmosphere. If such a process is considered "green," but cannot be created
 and operated economically, then the process will be of academic interest only and
 its potential green benefits unrealized. There are some simple "rules of thumb" that
 one can use to lower the cost of a CO2-based process. First, operate at high con-
 centration. One way in which to minimize  the cost of a  CO,-based process is to
 minimize the size of the equipment, hence minimize the amount of solvent (CO,,)
 flowing through the process. Next, operate at as low a pressure as possible. Use of
 CO,-philic functional groups  in the design  of substrates or catalysts can greatly
 lower the needed operating pressure, although it should be remembered that their
 use could easily raise raw material costs. Another classic technique for lowering
 operating pressure is to employ co-solvents—whether the use of co-solvent/CO,
 mixtures is green or not must be determined on a case-by-case basis. A final some-
 what obvious route to the lowering of the operating pressure is by operating at sub-
 ambient temperatures. Here, however, one must balance the advantage gained by
 reducing the operating pressure with the energy cost for cooling and any reduction
 in reaction rate. Although perhaps counterintuitive, recover products without high-
pressure drops. It has been noted that use of CO, as a solvent is advantageous
 because reduction of the pressure to one atmosphere results in the complete pre-
cipitation of any dissolved material, but use of such a route for product recovery
raises costs, as one then must compress the make-up of CO,. As gas compression is

-------
 energy-intensive and expensive, a greener route to product recovery is desirable.
 Fiiutllv, operate the process coniiinioitslv if possible.  The rationale lor operating
 in a continuous mode is that the equipment can be smaller. Although this is usually
 straightforward for liquid substrates, it can be much more difficult for the process-
 ing of solids at high pressure.

    This report  focuses on CO,-based processes where chemical  reactions are
 taking place (i.e.. green chemistry) or materials are being processed to create viable
 products. Hence, processes that contain only separations unit operations (for ex-
 ample, extractions and cleaning) were not specifically considered. Whereas this
 report does not explicitly address the state-of-the-art in cleaning using CO,, it does
 evaluate several technological issues  that are significant to the  advancement of
 CO,-based cleaning, including the design of high-pressure equipment and auxiliary
 agents for cleaning. In this report, a wide variety of process schemes were critiqued
 for their ability  to provide a more sustainable process as  compared  to existing
 technology,  using the 12 principles of green chemistry as a basis for judgments.
 Research and  development conducted over the previous 5 years (1997-present)
 has been emphasized.

    The extent to which  conventional solvents are regulated will  have a profound
 effect on the extent to which CO, is used as a solvent in the future. Carbon dioxide
 often has been described as a potential substitute for chlorofluorocarbons (CFCs).
 Because CFCs actually exhibit a number of advantageous properties, without the
 regulation restricting their use it is not likely that CO, would ever have been consid-
 ered as a viable competitor. Although  CFCs represent a somewhat extreme ease.
 regulation does exert more subtle effects on the use  of CO,. From an engineering
 perspective. CO, is nearly always more difficult to employ as a solvent because one
 needs  high-pressure equipment. Consequently, the  extent to which a particular
 solvent is regulated, and the subsequent obstacles to the use of such a solvent in
 a chemical process, can tip the scales either in favor or against use of CO,.

    Milestones in green chemistry using CO,. Perhaps the first true "green" appli-
cation  of CO,  was  the coffee decaffeination process scaled-up during the 1980s:
this is  a milestone  as it  showed that one  could successfully scale and operate a
CO,-based process economically given a good design. In the 1980s, conventional
 wisdom claimed that CO/s solvent power resembled that of n-alkanes, despite a
 large body of experimental  evidence to the contrary. The paper in Science by the
 DeSimone group on the CO,-philicity of  poly(perfluoroacrylates) in 1992 was a
milestone both from the scientific standpoint, and from a dissemination perspec-
tive, as this publication served to quash the "CO, is like hexane" heuristic. With
publications by Leitner's group and Tumas' group in the mid-1990s, showing the
use of fluorinated ligands in homogeneous  catalysis in CO,, green chemistry in CO,
began  to rapidly permeate the chemistry community and be applied broadly as a
solvent in organic transformations. In 1999. Brennecke published a study demon-
strating the potential for use of ionic liquid/CO, biphasic mixtures as media for

-------
 green chemistry—the first papers exploiting this biphasic system appeared in 2001.
 A number of researchers examined the strong potential for CO, to plastici/e poly-
 mers, with several important papers appearing between  1985 and 1994. Exploitation
 of this science appeared in 1996  through 2001. as both industry and acadcmia
 employed the plasticizing effect to enhance mixing in polymer systems.

     Regarding commercial milestones, the introduction of the CarDio process for
 continuous production of polyurethane foam using CO,  as the blowing agent has
 been extremely important, in that it is both green chemistry and commercially suc-
 cessful. Much more widely known is the construction (by DuPont) of a semi works
 facility to polymerize fluorinated monomers in  carbon dioxide. Another series of
 commercial milestones occurred in 2000 to early 2001.  when the pharmaceutical
 industry purchased (either in their entirety or substantial portions) Bradford Particle
 Design. Separex. and Phasex—three of the more  significant commercial enterprises
 relying primarily on supercritical fluids technology. It will be interesting to see whether
 this leads to more rapid commercialization of CO,-based processes or the reverse.

     Areas for future research on CO, technology. In each section of this report.
 mention has been made of potentially useful avenues for  future research: these are
 summarized below.

 •   The use of biphasic systems (including carbon dioxide as one component) for
    conducting reactions using gaseous components. Also, focus more on oxida-
    tions  and hydroformylations versus hydrogenation in CO,; the former reac-
    tions generate more waste and require more stringent  conditions than hydroge-
    nation, yet have received relatively less attention in the literature.

 •   Group contribution or, betteryet, first principles models  for the prediction of phase
    behavior in multiphase, multicomponent systems where carbon dioxide is one of
    the components; therefore, this should result in the design of "CO^-philes" that do
    not include fluorine. Prediction of basic transport properties is needed as well.

 •   The design of systems for the rapid high-pressure treatment of solid articles (as
    in the development of silicon wafers) or the continuous coating  of material
    using a CO,-based solution.

•   The use of CO, in  microelectronics processing, which will involve simulta-
    neous equipment and molecular design of great complexity.

•   An indepth understanding of the mechanism for generation of CO and subse-
    quent poisoning of noble metal catalysts in the presence of H, and CO,.

•   The design of catalysts for the generation of polyesters and commodity chemi-
    cals (aromatic acids) from CO,; activation of CO, at low pressures.

•   Explore the use of co-solvents for CO, in a more  systematic manner to find
    mixtures that are technically, environmentally, and economically successful.

-------
•   The design of additives that would allow greater use of CO, in the extrusion
    foaming of polymers.

•   The development of a set of fundamental design principles for the formation of
    particles via phase separation from mixtures that include CO, (under How in a
    known geometry).

•   Programs that focus on overcoming the various technical hurdles to the use of
    CO, in coating processes.

    The Technology for a Sustainable Environment Program (TSE) has been highly
successful in educating scientists in the use of CO, through their funding of sound
basic research in this area. A direct consequence of this is that CO, is no longer
considered an exotic solvent, but rather one more weapon in one's arsenal of sustain-
able technologies. In the past, the TSE program has solicited proposals from academia
in the general area of green chemistry, giving principal investigators (Pis) complete
freedom regarding the focus of the proposals. It might be useful for TSE staff also to
publish a list of current high-profile effluent problems where innovative green chem-
istry solutions might contribute to broad industry or societal benefit. However, it
should be stressed that moving from the current entirely curiosity-driven system to
one that is entirely "target-specific" is  not recommended, in that much of the truly
innovative discoveries in green chemistry and processing might be lost.

-------
                             Section  1

                          Introduction
1.1   Supercritical Fluids

    Supercritical fluids are materials at temperatures above their critical value (T.):
at the critical temperature, the meniscus that separates the gas from liquid phases
disappears, leaving a  single phase. As one also can see from the generic phase
diagram in Figure I. the compressibility (a function of "V/"P) diverges at the criti-
cal point. At temperatures above T. the density of the material can be varied smoothly
from "gas-like" to liquid-like values through variations to pressure [ I]. Because a
number of physical properties (viscosity, diffusivity. dielectric constant) are known
to depend on fluid density [2]. it is not surprising that physical properties of
supercritical fluids are said to lie intermediate to those of liquids and gases. "Near-
critical" is typically used to describe fluids at temperatures below the critical, but
within 50 K. It should be noted that no firm definition exists for "near-critical." and
                                                 IOC
FIGURE 1  Generic phase diagram: reduced density as a function of reduced
             pressure and reduced temperature [1].

-------
 hence, the term "liquid" will be employed to describe fluids at temperatures close
 to but  below T . Note that "liquid" can be  used to denote the situation  where a
 compressed liquid exists alone, or where a liquid exists in the presence of the asso-
 ciated vapor phase, although "saturated liquid" is normally used to denote the lat-
 ter situation.

    At constant pressure, the density of a fluid decreases as the temperature increases
 above T (see Figure 1). Further, to a first approximation, the solvent power of a fluid
 is proportional to its density [3]. Consequently, to successfully employ a supercritical
 fluid as a solvent, one should employ conditions such that the density is liquid-like.
 and hence, one should operate at temperatures relatively  close (either above  or be-
 low)  to T . Because many reactions of interest currently  carried out in liquids are
 conducted at temperatures between  273 and 423 K. those fluids that are  of most
 interest to the scientific community exhibit critical temperatures in this range.

    Critical temperature generally rises as the molecular weight of a material rises.
 and also if a material exhibits a significant self-interaction (hydrogen bonding, for
 example). Consequently, those fluids that exhibit T "s in the range of most interest
 tend to be nonpolar fluids of relatively low molecular weight, as shown in Table 1.
 Although they exhibit accessible critical temperatures, many of the fluids shown in
 Table 1 have not been extensively investigated (in a solvent role) owing to inherent
 disadvantages. For example, reports during the early 1990s of violent side reactions
 (oxidations)  have nearly extinguished the enthusiasm of the scientific community
 towards the use of N,O as a solvent, despite a number of characteristics that render it
 superior to CO, as a solvent. The alkanes are flammable, and hence, interest in these
 materials  as solvents has been limited to primarily academia over the past decade (a
 notable exception to this is the work by Magnus Harrod. described in a later section,
 on hydrogenation in propane). The fluoroalkanes and hydrofluoroalkanes exhibit a
 number of highly favorable physical characteristics, including nonflammability. com-
 mercial availability, and relatively low  vapor pressures. Unfortunately, these  fluids
 also are expensive and exhibit strong climate change potential, and interest in their
 use as solvents also has been relatively muted.

    The concerns listed in the previous paragraph have effectively reduced the list of
 viable supercritical solvents to carbon dioxide, given that CO, is inexpensive, non-
 flammable, and relatively inert (see Table 1). Consequently, this report will  focus
 entirely on the use of CO, (supercritical and liquid) as a solvent in the design  of
 greener reactions and processes. Note that some industrial processes currently em-
 ploy reactants (other than carbon dioxide) in the supercritical state (e.g., production
 of low density polyethylene from supercritical ethylene, and production of 2-butanol.
 and ultimately methyl ethyl ketone, from supercritical butene). These processes were
designed and constructed without "green" issues in mind, however. Indeed, advances
 in catalyst design (both Ziegler-Natta type and  metallocenes) have permitted the
polyolefins industry to avoid using high-pressure ethylene (reducing energy input)
while creating products with a broader array of physical properties.

-------
          TABLE  1   Critical Constants of Some Common Fluids [3].
Substance
CO,
SFh
Xenon
Ethylene
N,0
Ethane
Propane
Butane
CF4
CF;H
C:F,
C,F,H
C,F4H,
CH,OCH:
Propylene
Methanol
Water
Critical
Temperature ("K)
304. 1
3 1 8.7
289.7
282.4
309.6
305.4
369.8
425.2
227.6
299.3
293.0
339.2
474.2
400.0
364.9
512.6
647.3
Critical Pressure
(bar)
73.8
37.6
58.4
50.4
72.4
48.8
42.5
38.0
37.4
48.6
30.6
36.4
40.6
52.4
46.0
80.9
221.2
Critical Volume
(cc/mole)
160
198.8
118.4
130.4
97.4
148.3
203.0
255.0
139.6
132.7
222.0
209.7
235.3
178.0
181.0
118.0
57.1
    It should be noted that water exhibits all of the same favorable properties as CO,.
except for an accessible critical temperature. Water's T of 647 K (and Pt of 212 bar)
has for the most part  limited its use (in the supercritical state) to processes that
destroy organic compounds (complete oxidation) or to production of inorganic mate-
rials. Even this application is made difficult by the corrosive nature of supercritical
water solutions. Recently, however, a number of researchers have begun to explore
the use of superheated water (temperature between 373 K and 647 K) as a solvent for
use in "green" reactions [4]. Superheated water exhibits a relatively low dielectric
constant (-10). and thus can dissolve organic compounds. Superheated water also
exhibits a relatively high ionization constant, and hence, researchers have (for ex-
ample) employed superheated water to conduct Friedel-Crafts type acylations with-
out the use of the traditional aluminum halide [4]. The use of superheated water in
green chemistry, although a significant topic for inquiry, is outside the scope of this
report—interested readers can consult the reviews in reference 4.


1.2   Physical Properties of CO2
    The pVT properties of CO, have been known since the 1930s [5]; extensive data
sets are available in the literature and on the Web in the form of correlations of

-------
10
     density, viscosity, dielectric constant, etc., as functions of temperature and pres-
     sure [6J. Note that CO,'s critical pressure (and hence its vapor pressure in the "near-
     critical" or liquid regime) is significantly higher than analogous values for alkane,
     fluoroalkane. or hydrofluoroalkane fluids (Table 1). CO,'s anomalously high criti-
     cal pressure is but one result of the effect that  CO,'s strong quadrupole moment
     exerts on its physical properties. Although the high critical pressure is problematic,
     the most unfortunate outcome of the effect of quadrupole moment on physical prop-
     erties was the premise, first advanced during the late 1960s, that CO, might prove
     to be a solvent whose strength would rival or surpass that of alkanes and ketones
     [3]. Because early models employed to  calculate CO,'s solvent power relied on a
     direct relationship between the Hildebrandt solubility parameter (5) and the square
     root of  the critical pressure [(Po)";], the  solubility parameter of CO,  was
     overpredicted by 20 to 100 percent, leading to early inflated claims about the po-
     tential for using CO, to replace conventional organic solvents.


     1.3    Environmental and Safety Advantages to Use of CO2 in
           Chemical Processes

        Carbon dioxide is nonflammable, a  significant safety advantage in using it  as
     a solvent. Also, it is naturally abundant,  with a TLV (for airborne concentration at
     298 K to which it is believed that nearly all workers may be repeatedly exposed day
     after day without adverse effects) of 5,000 ppm [7], rendering it less toxic than many
     other organic solvents (acetone, by comparison, has a TLV of 750 ppm. pentane is
     600 ppm, chloroform is 10 ppm [7]). Carbon dioxide is relatively inert towards reac-
     tive compounds, which is another process/environmental advantage (byproducts
     owing to side reactions with CO, are relatively  rare), but CO,'s relative inertness
     should not be confused with complete inertness. For example, an attempt to con-
    duct a hydrogenation in CO, over a platinum catalyst at 303 K will undoubtedly lead
    to the production of CO, which could poison the catalyst [8], The same reaction run
    over a palladium catalyst under the same conditions will by contrast produce lesser
    amounts of CO as a byproduct [9], and knowledge of CO,'s reactivity is vital to its
    use in green chemistry.

       Carbon dioxide is clearly a "greenhouse gas," but also it is a naturally abundant
    material. Like water, if CO, can be withdrawn from the environment, employed in a
    process, then returned to the environment "clean,"  no environmental detriment
    accrues. However, although CO, could in theory  be extracted from the atmosphere
    (or the stack gas of a combustion-based power plant), most of the CO, employed in
    processes today is collected from  the effluent of ammonia plants or derived from
    naturally occurring deposits (for example, tertiary oil recovery as practiced in the
    United States f 10]). Because industrially available CO, is derived from manmade
    sources, if CO, can be isolated within a process  one could consider this a form of
    sequestration, although the sequestered  volumes would not be high. Ultimately.

-------
 one should consider the source of CO, used in a process to adequately judge the
 sustainability of the process.

    CO,'s combination of high TLV and high vapor pressure means that residual
 CO, left behind in substrates is not a concern with respect to human exposure—the
 same certainly cannot be said to be true for many manmade and naturally occurring
 organic compounds. Because there is effectively no liability due to "residual" CO,
 in materials following processing. CO, is not considered a solvent requiring pro-
 cess reevaluation by the U.S. Food and Drug Administration. Only water also en-
 joys this special situation. Indeed, most of the commercial operations employing
 CO, as a solvent were initiated to take advantage of CO,'s particular advantages in
 products designed for intimate human contact (such as food), or CO,'s non-VOC
 designation (such as the foaming of thermoplastics). The recent commercialization
 of fabric cleaning using CO, benefits both from CO,'s advantages in human-con-
 tact applications and situations where emissions appear unavoidable.

    The simultaneous use of both hydrogen and oxygen in a reaction is obviously
 problematic from a safety standpoint, given that H,/O, mixtures are explosive over a
 broad concentration range. Addition of CO, to mixtures of H, and O, expands the
 nonexplosi ve regime (in the gas phase), more so than if either N, or water vapor were
 added f 11 ]. At this point, it is not clear to what extent the nonexplosive regime will
 expand further as one raises the density of the mixture (and hence the heat capacity).

    In a final intriguing note regarding safety advantages inherent to the use of CO,
 as a solvent. DuPont scientists [12] discovered that addition of CO, to tetra-
 fluoroethylene enhances the stability of that notoriously difficult-to-handle mono-
 mer, although the exact mechanism for the enhanced stability has not been pub-
 lished. What has been revealed is that addition of CO, to TFE vapor inhibits run-
 away decomposition and explosion of the monomer. In addition, the CO/TFE mix-
 ture behaves like an azeotrope, in that boiling of a mixture of the two does not
 significantly change the concentration of either the liquid or the  vapor. According
 to the DuPont patent [12], this "azeotrope-like" behavior persists over a wide con-
 centration range, behavior that is quite unlike that of typical azeotropic mixtures.
 The enhanced safety of CO/1 hb mixtures relative to pure TFE is one of the rea-
 sons that DuPont constructed a semi-works polymerization plant employing CO,
 as solvent for the generation of fluoropolymers.


 1.4  Environmental and Safety Disadvantages Inherent to
      Use of CO2 in  a Process

    Because CO,'s vapor pressure at room temperature exceeds 60 bar. use of CO,
 in a process clearly requires high-pressure equipment, creating a potential safety
hazard relative to the same process operated at one atmosphere. In addition, un-
controlled release of large quantities of carbon dioxide can asphyxiate bystanders

-------
12
     owing to airdisplacemenl. These issues have not impeded the commercialization of
     CO,-based processes, nor is it  likely they will do so in the future. It should be
     remembered that the low density polyethylene polymerization process, first com-
     mercialized in the 1940s  and still  in operation today [13J. runs continuously at
     2.000 to 3.000 bar and 520 K with a highly flammable component, and hence, safe
     operation of a 100-200 bar CO,-based plant is readily  achievable using current
     technology. Operating an  exothermic reaction in a high-pressure environment is
     accompanied by additional safety concerns versus the analogous reaction run at
     one atmosphere.

        Whether to use liquid or supercritical CO, is a choice that actually involves
     safety as well as chemistry considerations. Although use of supercritical CO, al-
     most  always involves use of higher pressure (to achieve the same solubility of a
     given substrate as in the liquid phase), other factors also should be considered.
     First, supercritical CO, will exhibit a higher compressibility than liquid CO,, and
     hence the supercritical fluid will be better able to absorb excess heat evolved from
     an exothermic reaction whose rate suddenly exceeds typical operating conditions.
     On the other hand, use of saturated liquid CO, (in the presence of the vapor phase)
     would allow boiling to be used as a means to absorb excess heat. Use of supercritical
     CO, (versus liquid) could avoid complications owing to a phase separation occur-
     ring upon a departure from established temperature or pressure conditions within a
     given reactor. For example, if one is employing a mixture of oxygen, substrate, and
     liquid CO,  in a particular process, a sudden drop in pressure owing to a perturba-
     tion in the process could lead to formation of a flammable gaseous phase—use of
     a supercritical mixture could avoid this problem as no vapor-liquid separation will
     be encountered. Indeed, it also should be  remembered that the T of a mixture of
     CO, and other materials  will differ from that of pure CO, (see, for example, refer-
     ence 2 for useful correlations), and hence T-p conditions sufficient for supercritical
     operation with pure CO, may create a liquid in the case of the mixture.


     1.5   Chemical Advantages to  Use of CO2 as a Solvent

        Carbon dioxide can provide not only environmental advantages, but also chemi-
    cal advantages when applied strategically, as described below.

        CO, cannot be oxidized. In essence, carbon dioxide is the result of complete
    oxidation of organic compounds: therefore, it is particularly useful as a solvent in
    oxidation reactions. Use of almost any organic solvent in  a reaction using air or O,
    as the oxidant (the least expensive and most atom-efficient route) will lead to for-
    mation of byproducts owing to reaction of O, and the solvent. Indeed, the commer-
    cial anthraquinone process used to generate H,O, requires the removal and regen-
    eration (or incineration)  of substantial volumes of such solvent byproducts  [14].
    Oxidation reactions in CO, have consequently been investigated extensively over
    the past  decade (see Section 2.8).

-------
                                                                               13
    Because CO, is inert towards oxidation and also is nonflammable. CO, is one of
 the very few organic solvents that could be considered for the direct reaction of
 hydrogen and oxygen to form hydrogen peroxide [15]. This process has been under
 investigation for more than 2 decades, yet traditional  organic solvents are not
 sufficiently inert/safe, and water exhibits productivity disadvantages.

    CO, is benign, and hence cross-contamination of the other phase during liq-
 uid-liquid extraction is not really contamination. There are a number of large-scale
 chemical processes that employ biphasic (water-organic) mixtures—H,O, produc-
 tion and hydroformylation of low molecular weight alkenes are but two examples
 [14]. In any contact between aqueous and organic phases, some cross-contamina-
 tion is inevitable. The aqueous phase will require subsequent remediation  to elimi-
 nate the organic contamination, but the organic phase may require drying to allow
 further use in the process.

    Although CO, will "contaminate" an aqueous phase upon contact in a process.
 a mixture of CO, and water clearly does not require remediation (the CO, phase may.
 of course, require drying for further use). Consequently, CO, exhibits a particular
 advantage in processes where a biphasic reaction or liquid-liquid extraction against
 water is required. Eckert and colleagues [ 16] have, for example, investigated the use
 of phase transfer catalysts in CO,/water mixtures. Further, the coffee decaffeination
 process employs a liquid-liquid  extraction between CO, and water to recover the
 extracted caffeine [17].

    CO, is an aprotic solvent. Clearly, CO, can be employed without penalty in
 cases where labile protons could interfere with the reaction.

    CO, is generally immune to free radical chemistry. Because carbon dioxide
 does not support chain transfer to solvent during free radically initiated polymer-
 ization, it is an ideal solvent for use in such polymerizations, despite the fact that it
 is typically a poor solvent for high molecular weight polymers. In chain transfer, a
 growing chain (with a terminal radical) abstracts a hydrogen from a solvent  mol-
 ecule, terminating the first chain. The solvent-based radical may or may not support
 further initiation, and hence, chain transfer to solvent can lead to diminished mo-
 lecular weight and diminished polymerization rate. Research conducted during the
 1990s (primarily by J.M. DeSimone and coworkers) showed that CO,  does not
 support chain transfer, as it is inert towards polymer-based free radicals [18]. Other
researchers have examined small-molecule free radical chemistry in CO, to be viable
as well  [19]. Indeed, it is likely that most of the polymerizations currently conducted
by DuPont  in its semi-works facility are precipitation polymerizations, where the
 improved control over molecular weight and the enhanced safety inherent to use of
TFE/CO, mixtures (see Section 1.3) more than makes up for any difficulties caused
by polymer precipitation during  the reactions.

    CO, is  miscible with gases in all proportions above 31°C. The rate of  most
processes where  a gas reacts with a liquid is limited by the rate at which the gas

-------
14
     diffuses to the active site (either within a catalyst particle or simply to the liquid
     reactant). Gases such as hydrogen and oxygen are poorly soluble in organic liquids
     and water, and hence, in many two- and three-phase reactors the  rate is limited
     specifically by the rate at which the gas diffuses across the gas-liquid interface.

        Although phase separation envelopes exist with gases at lower temperatures, liq-
     uid CO, can absorb much higher quantities of H, orO, than typical organic sol vents or
     water [20]. Hence, one can eliminate the dependence of the rate on gas transport into
     the liquid phase by employing CO,. Although conventional wisdom might claim that
     this effect is achieved only through creation of a single phase (of CO,, gaseous reactant.
     and liquid substrate), recent work in the literature shows that one can achieve high gas
     solubility and hence high rate while remaining two-phase (see Section 2).

        It should be remembered that CO, will exhibit total miscibility with gases above
     304 K only if those gases also exhibit critical temperatures less than  or equal to
     304 K. This includes commonly used reactant gases such as H,. O,. and CO. for
     example. Further, addition of any third component (here, a gas such as H, or CO) to
     a mixture of CO, and substrate (and catalyst, perhaps) will alter the phase behavior
     of  the mixture. Because commonly  used  reactant gases such as H,. O,. and CO
     exhibit low critical temperatures [2],  at typical reaction temperatures (273-373 K).
     the density of these gases, even under relatively high pressures used to compress
     CO,, will be quite low (more gas-like than liquid-like). As such, we expect these
     gases to behave as nonsolvents towards the substrate and/or catalyst [21]. Thus.
     addition of large amounts of reactant gas  to the mixture may solve one problem
     (diffusion limitations) and create another (phase separation).

        CO, exhibits solvent properties that allow miscibility with both fluorous
     and organic materials. Carbon dioxide is miscible with a variety of low molecular
     weight organic liquids, as well as with many common fluorous (perfluorinated) sol-
     vents. The  literature has shown previously that one can create a homogeneous mix-
     ture of certain fluorous and organic liquids at one temperature, where phase separa-
     tion occurs on a temperature increase or decrease. Recently, Eckert's group has shown
     that one can employ  CO, as a phase separation "trigger" in much the same way—
     addition of CO, (at pressures as low as 20-30 bar) to a mixture of organic and fluorous
     liquids creates a homogeneous single phase, while removal (through depressuriza-
    tion) returns the system to a two-component, two-phase system [22].

        CO, exhibits a liquid viscosity only 1/10 that of water. At liquid-like densities.
    CO,'s viscosity is only 1/10 that of water,  and hence Reynolds numbers (pVD/u,
    where V is fluid velocity. D is  distance or length, p is fluid density,  and jj  is fluid
    viscosity) for flowing CO,  will be approximately 10 times those for conventional fluids
    at comparable fluid velocity. Because convective heat transfer is usually a strong
    function of Reynolds number, heat transfer in a CO, mixture can be expected to be
    excellent. On the other hand. CO,'s physical  properties also lead to significant natural
    convection causing problems in some  coatings processes. The extent to which natu-

-------
                                                                                15
 ral convection is an issue is directly related to the magnitude of the Grashof number
 [23]. which itself scales as p:/v:. Because CO, exhibits a liquid-like density and a gas-
 like viscosity. Grashof numbers for CO,-based processes can be significantly higher
 than for analogous liquid processes.

    The surface tension in carbon dioxide is much lower than that for conventional
 organic solvents, and the diffusivity of solutes is expected to be considerably higher.
 owing to CO,'s low viscosity. Consequently. CO, would be expected to wet and
 penetrate complex geometries better than simple liquids. Further, solutes would be
 expected to diffuse faster within catalyst pores where CO, is the solvent than in
 analogous systems using conventional liquids.


 1.6   Chemical Disadvantages to Use of CO2 as a Solvent

    Carbon dioxide exhibits some inherent disadvantages where chemistry is con-
 cerned: some ot these are unique to CO,, while others are common to any number of
 solvents.

    CO, exhibits a relatively high critical pressure and vapor pressure. As men-
 tioned above. CO, exhibits high critical and vapor pressures: these characteristics
 guarantee higher capital costs for a CO,-based  process relative to one  using a
 conventional solvent, as well as the need for specialized equipment for laboratory
 work. Exothermic reactions pose special problems for operation in CO,, given that
 high pressure is the baseline situation.

    CO, exhibits a low dielectric constant. Carbon dioxide exhibits  a dielectric
constant of approximately 1.5 in the liquid state; supercritical CO, will exhibit values
generally between 1.1 and 1.5, depending on density.  This low dielectric can be
both a process disadvantage and a chemistry disadvantage. Some reactions, for
example, require polar solvents for best results. Further,  low dielectric constant also
suggests poor solvent power, and hence solubility in CO, can require much higher
pressures for certain classes of solute than more polar compressible fluids (fluoroform.
for example, which exhibits a liquid dielectric of -10). On the other hand, the thermo-
dynamic interaction between CO,  and  nonpolar methylene groups is not  particularly
favorable, and hence, ethane often is a better solvent for hydrocarbons than CO,.

    CO, is a Lewis acid.  Carbon  dioxide will react with strong bases (amines.
phosphines, alkyl anions) [24]. When attempting to use amines as reactants, this
can be a serious disadvantage, in  that carbamate formation can slow the rate of the
intended reaction and also can alter the solubility characteristics of the substrate.
Although alkyl-functional primary and secondary amines react  readily with CO,.
tertiary amines are nonreactive. Further, the presence of electron-withdrawing groups
in close proximity to the nitrogen atom (as in anilines)  prevents formation of car-
bamates between CO, and such compounds. Carbon dioxide also will react (not
surprisingly) with metal alkoxides. metal alky Is, and  metal hydrides.

-------
16
        CO, has been shown to react reversibly with a number of enzymes (lysine
     residues, specifically), leading to low activity in the presence of CO, (although
     activity returns to normal following removal of the enzyme from the CO,-rich envi-
     ronment) [25]. Because carbamate formation is reversible, even at high pressure.
     researchers have employed CO, as a protecting group for amines [26]. and hence
     CO,'s reactivity with amines can be an advantage as well as a disadvantage. Fi-
     nally, because CO, reacts readily with carbanions to form relatively unreactive
     carboxylates. anionic polymerization cannot be conducted in carbon dioxide.

        CO, can be hydrogenated in the presence of noble metal catalysts to pro-
     duce CO. If one is trying to hydrogenate a substrate in CO, over a heterogeneous
     platinum catalyst, production  of CO will poison the catalyst  and produce toxic-
     byproducts. Unfortunately, this reaction takes place at relatively mild temperatures
     [8]. There has been a certain degree of controversy recently as to whether the same
     reaction  occurs over palladium catalysts. For example. Hancu and Beckman  [15]
     demonstrated that hydrogenations could be conducted successfully in CO, (over
     palladium), although it should be noted that the hydrogenation in question was
     very  fast and was conducted at 298 K.  Subramaniam's group [27]  was able to
     successfully conduct a hydrogenation reaction over palladium in a continuous
     reactor; no  loss in catalyst activity was observed over a period of 1-2 days. By
     contrast. Brennecke and Hutchensen [28] found that a palladium catalyst deacti-
     vated rapidly during batch hydrogenations in CO,. Subramaniam [29] recently in-
     vestigated these apparent contradictions and found that higher temperatures (>
     343 K) and greater residence times (such as would be found in batch reactions) lead
     to formation of CO that does ultimately poison the catalyst. This is an area where
     further research is certainly merited, given the potential importance of hydrogena-
     tion reactions.

        In addition to CO. it is likely that some formate could be created through hydro-
     genation  of CO, over noble metals; formate has been observed during homogeneous
     catalysis  [30] and theoretically could form under heterogeneous conditions as well.

        Dense CO2 produces low pH (2.85) upon contact with water. Carbon dioxide
    dissolves in  water at molar concentrations [31 ] at moderate pressures (< 100 bar).
    rapidly forming H,CO,. This can render some biocatalytic reactions problematic, in
    that many enzymes are denatured (unfolded and/or deactivated) by low pH. Johnston's
    group has shown that buffering is possible but that impractical!}'  high ionic strength
    (for enzymatic reactions) is needed [32]. On the  other hand, one could employ car-
    bonic acid as a reagent, in which case CO, could be treated  as a very low-cost, sus-
    tainable acid that does not require addition of a base for neutralization. Enick [33],
    for example, has employed carbonic acid, formed from CO,/water. to extract con-
    taminants from steel waste into water, where depressurization results in a rapid in-
    crease in pH and precipitation of the extracted materials. Carbonic acid formed from
    CO, and water reacts with hydrogen peroxide under basic conditions to produce a
    percarbonate species, which then can epoxidize alkenes [34].

-------
                                                                               17
    In summary, the low pH of water in contact with liquid CO, can he an advan-
 tage or disadvantage, depending on the circumstances. Hancu and Beckman [ 15).
 tor example, have investigated the generation of H,O, in CO,, where the product is
 stripped inlo water following synthesis in CO,. The optimum pH forH,O, stability
 is 2 to 4. so the low pH of water/CO, mixtures is an advantage for this process. The
 low pH of water in contact with CO, also enhances the back-extraction of caffeine
 in the decaffeination process for coffee. Clearly, however, the low pH of CO,-water
 systems is a detriment to the processing of biomolecules.

    CO, is a weak solvent (low polarizability per unit volume, low cohesive energy
 density). This is perhaps CO,"s greatest flaw, in that its  inability to solvate compounds
 of interest (hence requiring uneconomically high process pressures) has greatly inhib-
 ited its commercial use. This issue will be discussed in more detail in Section 3.3.

    CO, poisons Ziegler-type polymerization catalysts. CO, will terminate olefin
 polymerizations that employ classical Ziegler (titanium halide) catalysts, hence
 preventing such polymerizations from being conducted in carbon dioxide.


 1.7  How We Will Approach Our Analysis

    Reaction schemes will be critiqued on their ability to provide a more sustainable
 process as compared to existing technology, using the 12 principles of green chem-
 istry as a basis for judgments on sustainability. The basic principles of green chem-
 istry have  been  outlined by Anastas and Warner [35], and are listed below:

 I.   Prevention (alter process schemes and chemical  pathways to prevent the gen-
    eration of waste, rather than remediate waste once formed)

2.   Atom economy

3.   Less hazardous chemical synthesis

4.   Designing safer chemicals

5.   Safer solvents and auxiliaries (create and  employ solvents and process aids
    that, if emitted to the environment, exhibit a lower impact than currently used
    materials)

6.   Design for energy efficiency

7.   Use of renewable feedstocks

8.   Reduce derivatives

9.   Catalysis (create catalysts that are more selective than current analogues, and
    that therefore produce  lower volumes of byproducts during reactions)

 10.  Design for degradation

-------
18
     11. Real-time analysis for pollution prevention

     12. Inherently safer chemistry for accident prevention.

        If the properties of CO, and  its many proposed applications are examined.
     several common trends appear vis-a-vis the 12 principles shown above. CO, has
     been proposed as a benign alternative to common organic solvents, and hence prin-
     ciple 5 comes into play. If one assumes that some proportion of the organic solvent
     that is employed in  any chemical process will be emitted to the environment, then
     replacement of that solvent with CO, is a mode of prevention (principle I). as CO,
     emissions are less problematic. The toxicity of CO, is lower than for many organic-
     solvents (principle 4) and is naturally abundant (principle 7).

        It should be noted  that while use of CO, is  within the scope of several of the
     principles of green chemistry, improper or ill-considered process design could lead
     to egregious violation of some of the others. Indeed, if use of CO, as a solvent leads
     to higher energy consumption or an inherently unsafe process, then some of the 12
     principles will be followed while others are violated. Judgment of the net benefit
     must be done on a case-by-case basis.

        Finally, the source of CO, used in any process should be considered within the
     framework of the 12 principles of green chemistry. CO, is naturally abundant, yet
     CO, employed in an industrial process is typically not captured from the atmo-
     sphere. Carbon dioxide is a byproduct (of sizeable volume) of the commercial
     ammonia process [14], and much  of the commercially available CO, is derived
     from this source (after  purification). CO, also can be captured from fermentation
     processes, yet this is not generally practiced commercially  (owing to CO,'s low
     current value). Large deposits of CO, exist naturally in the United States: currently,
     these are tapped  for use in tertiary recovery of  petroleum in older fields in West
     Texas and Oklahoma [ 10]. If we examine the source of CO,, we can come to differ-
    ent conclusions of CO,'s worthiness as a benign  solvent. If, for example. CO, gen-
    erated by the ammonia process is employed, then one could consider this as pollu-
     tion prevention, because this CO, would otherwise be emitted to the atmosphere. If
     we employ CO, from natural deposits, this could be construed as "anti-sequestra-
    tion." as this CO, would ordinarily remain underground. If CO, could be  captured
    from the atmosphere (or power plant flue gas) in an energy efficient and economic
    manner, then used in a process, this would likely be the best source with respect to
    the 12 principles of green chemistry.


    1.8   Operating  a Process Economically With CO2

       Although use of CO, as a solvent is often considered to be "green." operation of
    any process at high pressure typically involves higher costs than the analogous process
    operated at one atmosphere. If such a process is  considered "green." but cannot be
    created and operated economically, then the process will be of academic interest only

-------
                                                                               19
 and its potential green benefits unrealized. There are some simple "rules of thumb" that
 one can use to render the cost of a CO,-based process as low as possible.

    Operate at high concentration. One way in which to minimize the cost of a
 CO,-based process is to minimize the size of the equipment. Given that CO, is
 typically proposed as a solvent (rather than a reactant). the most obvious means by
 which to minimize equipment size is to minimize the amount  of solvent (CO,)
 flowing through the process. Consequently,  one should try to choose or design
 substrates such that they exhibit high solubility in CO,. In addition, those processes
 where CO, is employed as the minor component (use of CO, as a plasticizer in
 polymer processing, for example) are likely to be favored economically.

    Another aspect of this issue is reflected in the typical phase behavior of com-
 pounds in CO, (see Figures 2 and 3). Note that in the typical phase diagram of a
 crystalline solid in CO,, an essentially pure solid phase exists in equilibrium with a
 solution. Given that the solid phase cannot be processed, one obviously  makes use
 of the solution where CO, is usually the major component. For the case of liquid-
 liquid phase behavior, a CO,-rich phase exists in equilibrium with a substrate-rich
 phase. However, because CO, has been shown to lower the viscosity of solutions
 substantially, one can actually pump and process the substrate rich phase. Further.
 one can operate  at lower pressure in addition to a higher  concentration. Conse-
 quently, it may be beneficial to employ systems where liquid-liquid phase behavior
 occurs rather than liquid solid. Efficient operation of a process is both economi-
 cally favorable and more environmentally friendly.

    Operate at as low a pressure as possible. Operation of a process at high
 pressure is more expensive than at one atmosphere, owing to equipment design and
 construction, as well as the additional safety  features that are necessary. Further,
                         0-8*         6.1*
                       NAPHTHALENE MOLE
                                                   9, If
FIGURE 2  Solid-fluid phase behavior [1]: CO,-naphthalene.

-------
20
     FIGURE 3  Liquid-liquid phase behavior [ 1J: CO,-hexane.


     the capital cost of a high-pressure process is not linear with pressure, because the
     pressure ratings of certain vital equipment (flanges, for example) are available in
     discrete steps (60 bar, 100 bar, for example). In addition, the number of companies
     with experience in high-pressure process design drops dramatically as the operat-
     ing pressure rises above 200 bar.

        Clearly, these caveats strongly recommend operating at the lowest pressure
     possible. One means by which to accomplish this is in the chemical design of reac-
     tants and/or substrates. It has been known for a number of years that certain func-
     tional groups are more "CO,-philic" (thermodynamically more CO,-friendly) than
     others. Use of CO,-philic functional groups in the design of substrates or catalysts
     can greatly lower the needed operating pressure, although it should be remembered
     that their use could easily raise raw material costs.

        Given that carbon dioxide is a relatively feeble solvent, a classic technique for
     lowering operating pressure (or raising operating concentration) is to employ co-
     solvents. Methanol and ethanol are most commonly used [ 1,36], but a wide range of
     organic solvents has been employed in this fashion, usually at concentrations below
     40 percent. Regarding whether the use of co-solvent/CO, mixtures is green, one must
     make a determination on a case-by-case basis. For example, in a conventional chemi-
     cal process, one must decide whether  it is more efficient to use a low-pressure pro-
     cess with  100 percent organic solvent or a high-pressure process using only 5-10
     percent organic solvent  (for example) with the balance CO,. To date, the typical
     answer has been to opt for the low-pressure, solvent-based process [37]. However, if
     the solvent (owing to the nature of the process) is to be emitted to the atmosphere.
     there are examples where the choice has been to opt for the CO,/cosolvent route. In
     the UniCarb coatings process developed by Union Carbide during the 1980s and
     1990s. CO, was  employed to replace  one component of a solvent mixture used in

-------
                                                                                21
spray coating, creating a CO,/co-sol vent-based process. The foaming of thermoplas-
tics such as polystyrene [38] often is conducted using a mixture of CO, and an al-
kane. a more efficient route than employing either  100 percent alkane or 100 percent
CO,. One also can employ relatively lower process pressures by operating in the two-
phase regime (gas-liquid) rather than employing pressures high enough to maintain a
single phase: more about this option will be described in a later section.

    Another somewhat obvious route to the lowering of the operating pressure is by
operating at subambient temperatures. Here, however, one must balance the advantage
gained by reducing the operating pressure with other impacts, such as the energy cost
for cooling and any reduction in reaction rate owing to reduced temperature. Whereas
dropping the temperature is an obvious mechanism to reduce the operating pressure,
there are others that have received far less attention. For example, the identification of
a minimum boiling azeotrope where CO, is  the majority component could provide a
solvent that is both green and exhibits a vapor pressure far lower than that of pure CO,.
Azeotropes are desirable in that process steps requiring flashing of the material  (or
small  leaks) will not change the composition of the solvent. Azeotropes can be at a
maximum boiling point (where the vapor pressure of the mixture is higher than either of
the pure component vapor pressures) or minimum boiling point (the opposite, and here
desired situation) [39]. Although addition of a second  component might lessen the
sustainability of the solvent, a solvent that is mostly CO, is typically better than one that
contains no CO,, and the reduction of the pressure through use of a minimum boiling
azeotrope might lower the operating pressure sufficiently to allow economical scale-up
of the process. Some  CO,-based azeotropes have been identified [40] as a result of
research by CFC-producing companies in a search for alternative refrigerants. Conse-
quently, most of the known CO, azeotropes are mixtures with fluorocarbons (also it is
known that ethane forms an azeotrope with CO,).

    Recover products without high-pressure drops. It has  been mentioned in
the literature that use of CO, as a solvent is advantageous because reduction of the
pressure to one atmosphere results in the complete precipitation of any dissolved
material, rendering easy product recovery. This may be true, but use of such a route
for product recovery raises costs, as one then must either recompress the CO, prior
to reuse or compress the make-up of CO,. As gas compression  is energy-intensive
and expensive, a greener route to product recovery is desirable.

    One example of product recovery without a high-pressure drop is liquid-liquid
extraction against water. A liquid-liquid extraction between an organic and aque-
ous phase inevitably cross-contaminates the phases, normally requiring remediation
of one. and probably  both phases. In the case of a water-CO, extraction, however.
the inevitable cross-contamination is benign (carbonated water). Indeed, the CO,-
based coffee decaffeination process employs a water-CO, extraction to recover the
caffeine, allowing the CO, to move in a loop at relatively constant  pressure (see
Figure 4). Further, the cross-contamination here is actually beneficial, as the low
pH in the "CO,-contaminated" water allows for a higher partition coefficient for

-------
22
     FIGURE 4  Process schematic for coffee decaffeination using CO, [1].
     caffeine, but the "water-contaminated" CO, is a better extractant for caffeine than
     pure CO,. Beckman and Hancu also employed a liquid-liquid extraction for the
     recover)' of H,O, synthesized in CO, [15].

         Operate the process continuously if possible. The rationale for operating in a
     continuous mode is that the equipment can be smaller while maintaining high pro-
     ductivity. Although this usually is straightforward for liquid substrates, it can be much
     more difficult for the processing of solids at high pressure. Indeed, there currently
     does not exist a viable means for introducing and removing solids continuously from
     a high-pressure (100 bar +)  process. Those commercial CO,-based processes that
     employ solids use  either batch or semi-batch mode. An example of the latter is the
     coffee decaffeination process, where dual extraction columns are employed, such
     that one  is in extraction mode while the other is being emptied and refilled [17].

        In the late 1980s, Chiang and colleagues at the University of Pittsburgh devel-
     oped a process (LICADO) for the cleaning of coal that employed a biphasic mixture
     of CO, and water [40]. Here, the coal was introduced to the process continuously as
     a slurry in  water. If the use of a  water slurry of solid substrate is tolerable, this is a
     useful means by which to introduce solids continuously into a high-pressure process.

        A clever example of the use  of phase behavior trends to accomplish continuous
     processing, as well as to recover products  without large pressure drops, is shown
     by Charpentier and colleagues [41] in the examination of the continuous polymer-
     ization of fluorinated monomers in carbon dioxide. The monomers are soluble in
     CO, (as  are many vinyl monomers), but the polymers are insoluble  (also a rela-
     tively general trend). Thus, monomers  can be recycled continuously through the
     continuously stirred tank reactor while the polymer precipitates and is collected.

-------
    Recover and reuse homogeneous catalysts and CO,-philes. The discover)'
 of CO,-philes in the early 1990s allowed for the exploration of a number of pro-
 cesses in CO, that had been heretofore untenable, owing to CO,'s feeble solvent
 power. Highly CO,-soluble surfactants and catalyst ligands became available, leading
 to a number of important discoveries regarding chemistry in carbon dioxide. How-
 ever, the new CO,-philes are significantly more expensive than their CO,-phobic
 counterparts, and it is important to the economics of a CO,-based process that any
 CO,-philes used in the process be recycled as extensively as possible. Note that the
 recycling of  CO,-philes not only makes good economic sense, but also is more
 sustainable than the case where the CO,-philes are simply disposed.

    Recovery and recycle of homogeneous catalysts are important whenever such
 catalysts are employed, because the metals employed in such catalysts are typically
 expensive.  In the case  of a CO,-based process, the ligands also are likely to be
 expensive (they must be designed to exhibit high CO, solubility), and the need for
 effective catalyst recycle is even more important.

    In summary, attention must always be paid to the economic viability of pro-
 cesses  employing CO, as reactant and/or solvent, and CO,-based  processes are
 generally thought to be "green": their benefits will never be realized if the cost of
 such processes dwarf that of conventional analogues.


 1.9   Scope of This Report

   This report will focus on CO,-based  processes where chemical reactions are
 taking place (i.e., green chemistry) or materials are being processed to create vi-
 able products. This will eliminate discussion of processes that contain only separa-
 tions unit operations (e.g., extractions and cleaning). Further, this report will focus
 on CO, as a  benign solvent (see Section 1.1 for rationale) as opposed  to other
 potential fluids. Research conducted over the previous 5 years (1997-present) will
 be emphasized.

   Clearly, a continuing challenge to the reader who is  interested or actively in-
 volved in research involving CO, as a solvent is "can the use of CO, create new
products, eliminate waste, save energy, and/or enhance safety to the point where
the costs of the product are reduced and a more sustainable process is created?"
The new DuPont fluoropolymer facility may be the first example of this, as the use
of CO, has eliminated the need for fluorinated solvents and has made working with
some of the monomers safer.

   In each of the following sections, recent research on various aspects of green
chemistry using CO, will be summarized. Whereas much of the published work in
this area emanates from academic groups, it should be noted that some industrial
concerns also  have been quite active. Industry quite naturally tends to patent before
they publish, and consequently a patent search was conducted for the period 1996-

-------
24
     2001  where finding the term "supercritical" in either the patent title or abstract was
     employed as the criteria defining a "hit." This search produced 450 hits for the time
     period in question. Well over one-half of these patents described inventions where
     CO, is used as the solvent in natural product extractions or cleaning. Of the remain-
     der, academic inventors filed nearly one-half. In addition, a search using "CO, or
     carbon dioxide" in title or abstract (without supercritical) produced 1,500 additional
     "hits." although the vast majority of these did not involve use of CO, as a solvent. For
     each of the sections on CO,-based research, a paragraph is appended that describes
     industrial activity (as described in patents) that is significant but not expressly men-
     tioned in the main body of the section. Without question, the most active industrial
     entities (in producing U.S. patents) on the use of supercritical fluids in green chemis-
     try/processing during 1996-2001 were DuPont. Micell Inc.. and Thomas Swan (UK).
     Not surprisingly, each of these companies also has supported major commercializa-
     tion efforts in CO,-based chemistry and processing (DuPont—polymerization of
     fluoropolymers in CO,; Micell—dry cleaning  in CO,; Thomas Swan—hydrogena-
     tions and alkylations in CO,). All three have strong research ties to universities.

        Applications for CO, that are not included in this report include extractions from
     natural products and food processing (such as the new orange juice pasteurization pro-
     cess from Praxair). Although these processes are important commercially, and are for
     the most part green, they also are mature and little innovation has emerged recently.


     1.10  A Note on Cleaning  by Using CO2

        There has been substantial effort made by both the academic and industrial com-
     munities to investigate the use of carbon dioxide in the cleaning of clothing, mechanical
     parts, and the surface of microelectronics components. Whereas this report will not
     explicitly address the state-of-the-art in cleaning by using CO,, it will evaluate several
     technological issues that are significant to the advancement of CO,-based cleaning.

        For example, although carbon dioxide is not a particularly strong solvent (see
     Section 3.3), it will readily solubilize low molecular weight, volatile, nonpolar com-
     pounds. If the "contamination" to be removed using CO, falls into this category.
     then no additional fundamental science is required, and the economics of the de-
     sign and construction  of the equipment will determine whether the technology is
     practiced. Breakthroughs in the design of high-pressure cleaning equipment that
     could rapidly process  individual  parts (see Section 5) would greatly help to pro-
     mote use of CO, as a cleaning solvent.

        CO, is a weak solvent,  and cleaning that requires the solubilization of polar.
     inorganic, or high molecular weight material will require the use of CO,-soluble auxil-
     iaries (surfactants, chelating agents). The discovery  that certain fluorinated com-
     pounds are "CO,-philic" during the early 1990s allows for rapid advancement in the
     design of such auxiliaries, and a discussion of the design of such auxiliaries is in-

-------
                                                                               25
 eluded in this report. For the future, the design of CO,-philic auxiliaries must likely
 include nonfluorinated building blocks, as fluorinated materials are very expensive
 (the fluoroalkyl sulfonate family) and are environmentally suspect [42].

    For the case of microelectronics processing, cleaning is accompanied by the
 need to perform chemistry (photolithography, etching). These topics are included
 in later sections (see Sections 3.11 and 3.12).

    Fabric cleaning recently has been commercialized by two groups in the United
 States (Micell Inc.. and Global Technologies/Dry Wash). Major issues confronting
 these groups in the future include design of inexpensive surfactants that clean effec-
 tively in CO,, the design of high-pressure cleaning equipment that renders the pro-
 cess cost competitive, and competition from other "benign" cleaning technologies
 (such as the use of high flash point alkanes. silicones. and water). The use of sili-
 cones (Green Earth [43]) seems to present significant competition, as these materials
 are promoted as being more benign than perchloroethylene or PERC (they are. if TLV
 is any indication), they are used at one atmosphere (equipment is relatively inexpen-
 sive), and their use is backed by some large, relatively wealthy corporations (GE for
 silicone production. Procter and Gamble for surfactant production [43]). Indeed, even
 the design of more efficient conventional  dry cleaning equipment (i.e.. that using
 PERC as the solvent) represents a commercial challenge [44]: the volume of PERC
 used by dry cleaners in the United States has dropped dramatically over the past
 decade primarily owing to the use of "tighter" equipment (lower fugitive losses dur-
 ing cleaning). Indeed, significant consolidation occurred in the CO,-based dry clean-
 ing industry during early  2002. Chart Industries, Inc.. a member of the DryWash
 consortium, decided to exit the CO,-based dry cleaning  business [45] after several
 years of disappointing growth ($126,000 net sales in 2001); the connection to the
 consortium was maintained by some of their employees as a spin-out company (Cool
 Clean). Cool Clean recently purchased the Hangers franchising operation from Micell.

    Finally, intellectual property (IP) issues also could inhibit the  use of carbon
 dioxide in fabric cleaning. Unilever, for example, has filed a number of patents (and
 continuations in part, etc.) on the use of surfactants in CO, for the purpose of fabric
 cleaning [46], yet it is not clear whether  this IP is meant to enhance Unilever's
 position in the field or merely to block the  use of CO, in fabric cleaning.

    In summary, this report will include several issues important to future cleaning
 applications for CO,, namely the design of effective, low-cost auxiliaries and the
design of lower cost equipment for use in  parts cleaning.


 1.11  The  Effect of Regulation on  the Use of CO2 in Green
      Chemistry and Chemical Processing

    The extent to which conventional solvents are regulated will have a profound
effect on the extent to which CO, is used as a solvent in the future. For example, we

-------
26
     can examine the recent history of CFCs (vis-a-vis CO,). CFCs were preferred as
     solvents for cleaning because they are non-flammable, relatively nontoxic (TLVof
     chlorodifluoromethane is 1 (XX) ppm [7]). and inexpensive. As a result of research
     performed during the 1970s and 1980s, it became apparent that CFCs contributed
     to the chemical erosion of the stratospheric ozone layer, leading to the Montreal
     Protocols that outlined a timetable for the withdrawal of CFCs from use as solvents
     (and refrigerants, etc.). Carbon dioxide often  is described as a potential substitute
     for CFCs in cleaning (and also refrigeration). Because CFCs exhibited a number of
     highly favorable properties, without international agreements restricting their use.
     it is not  likely that CO, would have been considered as a viable competitor.

        Although CFCs represent a somewhat extreme case, regulations and policies
     do exert more subtle effects on the use of CO,. This is seen most often when com-
     paring the advantages and disadvantages of using conventional solvents to use of
     carbon dioxide. From an engineering perspective. CO, is nearly always more diffi-
     cult to employ as a  solvent because one needs high-pressure equipment. Conse-
     quently, the extent to which a particular solvent is regulated, and the obstacles to
     the use of such a solvent in a chemical process, can tip the scales either in favor or
     against use of CO,. For example, acetone currently is not on the list of compounds
     that require reporting under Section 313 of the Emergency Planning and Commu-
     nity Right-to-Know Act (EPCRA. also known as the Toxics Release Inventory (TRI)
     [47]). Neither is it listed as a "Hazardous Air  Pollutant" [48] by the Office of Air
     Quality Planning and Standards at the U.S. EPA. Consequently, if a manufacturer
     currently was using carbon tetrachloride, for example, in a process where some of
     the solvent was emitted to the atmosphere, a  natural approach to "greening" the
     process might be to first determine whether acetone could be substituted for carbon
     tetrachloride (the latter is included on both the TRI and classified as a hazardous
     air pollutant). Use of acetone in  place of carbon tetrachloride would  likely not
     involve any changes to the equipment used in the process, but use of CO, would
     most certainly require equipment redesign. Choosing alternative solvents based on
     environmental considerations has been systematized as SAGE, the solvent alterna-
     tive guide, a Web-based interactive tool [49]; carbon dioxide is indeed one of the
     possible choices, depending on inputs, but no economic calculations are performed.

        As shown above, current regulations affect application of CO, by rendering some
     conventional solvents better or worse (from the cost of complying with current regu-
     lations) than carbon dioxide. In addition, it is possible to envision how future regula-
     tions or policies also might affect the use of CO, in green processing. Given that CO,
     has been  determined to play a role in global climate change, it is conceivable that the
     emission of CO, to the atmosphere will be regulated in the future. Consequently, a
     number of companies have begun instituting "trading credits" in CO, emissions, pri-
     marily on an internal basis. In these systems. CO, is assigned a "negative value," and
     thus use of CO, as a raw material allows one to theoretically reduce the cost of the
     process or product. If this practice becomes widespread (owing to future regulation

-------
                                                                                27
on CO, emissions), it likely will spur research and development on processes or
products that consume CO,.

    Another area where future regulation could greatly impact the use of CO, is if
restrictions are placed on the use of various fluorinated materials. Certain fluorinated
materials have been found to he highly CO,-soluble (see Sections 2.4.1. 3.3). and
these materials have been applied in the design of highly CO,-soluble auxiliaries
(surfactants and chelating agents). To date, the expense of fluorinated compounds
has greatly limited their use in commercial CO, technology, yet there are applications
areas (such as microelectronics) where the cost of fluorinated compounds might not
be an impediment to commercial use of CO, processing. However, it has been re-
ported recently that certain fluorinated surfactants persist in the environment, caus-
ing concern within the environmental and public health communities. The EPA has
proposed a Significant New Use Rule (SNUR) forperttuorooctanesultbnic acid and
closely related compounds [42] requiring manufacturers to notify EPA at least 90
days before commencing the manufacture or import of these materials for a signifi-
cant new use. This may be expanded to include perfluorinated carboxylic acids (and
their precursors) as  well.  If the use of fluorinated compounds is restricted in the
future, it could limit the use of CO, in certain areas of application. Needless to say.
design of nonfluorinated CO,-philic compounds would therefore become a priority
in advancing the state of the science.

-------
                              Section 2

                  Reactions Using Gases
2.1    Hydrogenation

    Hydrogenation is widely used in industry at scales ranging from grams per year to tons
per hour [50]. Hydrogenation is conducted at large scale in either the gas or liquid phase:
further, while gas-phase reactions are performed over a solid catalyst (heterogeneous cataly-
sis), liquid-phase reactions are conducted in either two (homogeneous catalyst, liquid and
gas each present) or three  (heterogeneous catalyst, liquid and gas each present) phase
modes. Finally, heterogeneous catalysis is conducted in batch, continuous slurry, and fixed
bed reactor configurations, although the latter is less common than the former two.

    Despite the broad range of potential reactor configurations and reactions, we can. by
examining the 12 principles of green chemistry described previously, make some general
comments as to how the use of supercritical fluids (CO, primarily) can enhance (and possibly
detract from) the sustainability and economic viability of a hydrogenation process. We will
restrict this discussion to those hydrogenations currently conducted in the liquid phase—
addition of a supercritical solvent to a gas-phase reaction will simply dilute the reactant
concentrations, reducing the rate significantly. With some exceptions (described below), it is
not likely that use of a supercritical solvent will enhance either the economic viability or the
sustainability of a gas-phase hydrogenation.

    Two areas where addition of CO, might benefit a gas-phase hydrogenation are flamma-
bility and catalyst defouling: addition of CO, to a mixture of hydrogen and a substrate will
enlarge the nonflammable region, and CO, could help to prevent catalyst fouling by dissolv-
ing compounds that contribute to coke formation [51].


2.2   Liquid-Phase Hydrogenations: Advantages to Use of
      Supercritical Solvents

    A number of hydrogenations (synthesis of unsaturated fatty acids, reduction of
fatty esters to alcohols) are conducted commercially in organic solvents, and re-

-------
30
     placement of these solvents with benign carbon dioxide will reduce both liability
     (reduced flammability. potential toxicity issues) and the potential for volatile or-
     ganic compound (VOC) emissions owing to fugitive losses. In addition, use of any
     supercritical fluid in a liquid-phase hydrogenation process can significantly alter
     the relative importance of fundamental processes governing the rate expression. In
     a three-phase hydrogenation. the rate can be governed purely by the kinetics of the
     reaction, but more likely will depend on the rate at which hydrogen diffuses from
     the gas phase to the active sites on the catalyst. The overall rate of transport is
     itself governed by three resistances in series:  (1) the resistance to transport of H,
     across the  gas-liquid interface. (2) the resistance to transport of H, through the
     liquid to the surface of the catalyst, and  finally  (3) resistance to transport of H,
     within the pores of the catalyst. Given that the overall rate is related to the sum of
     the resistances in  series [52J. one term can easily dominate the expression for the
     overall rate. Use of a supercritical fluid solvent (as opposed to a traditional liquid)
     eliminates  the gas-liquid interface, as low Tc gases such as H,, (X and CO are
     completely miscible  with fluids above their critical point. However, this does not
     necessarily mean  that the reaction will be kinetically controlled, as one must deal
     with the remaining two resistances to transport (i.e., bulk liquid to solid surface.
     interpore diffusion).  Because the diffusion constant is embedded in each of these
     resistances, the use of a supercritical fluid also can aid in their elimination, although
     simply switching from a conventional liquid  to  a supercritical fluid solvent for
     hydrogenation by no means guarantees that the reaction rate will depend solely on
     the underlying kinetics.

         It should be noted that significant effort is expended in hydrogenation reactor
     design to ensure that H, is well dispersed in the liquid phase—effective sparging
     greatly increases the contact surface area between the phases and hence the rate at
     which H, diffuses into the liquid. If use of a supercritical fluid allows for a reactor
     redesign (e.g., plug-flow  versus continuous-stirred tank given that gas sparging is
     unnecessary), then it may be possible to enhance the selectivity of the reaction
     through reactor design improvement, reducing waste.

         Indeed, selectivity is a major concern in any chemical process—hydrogenation
     is no exception. It is  well known that solvents affect the yield and selectivity of
     various hydrogenation reactions where "one very useful, although fallible, gener-
     ality is that in a series of solvents, the extremes in selectivity will be found at the
     extremes of the dielectric constant..." [50]. The supercritical fluids most often em-
     ployed as hydrogenation solvents, propane and CO,, exhibit dielectric constants at
     the lower end of the scale (1.5-1.7), and we might expect to see an effect on selec-
     tivity if a polar solvent is replaced by CO,. In addition, the physical properties of
     supercritical fluids are readily varied over a significant range through changes to
     pressure and temperature, and it may be possible to affect selectivity by altering
     these variables. Finally, addition of CO, or operation above the critical point of the
     reactant mixture could aid in coke removal from the catalyst, prolonging its life or

-------
                                                                                31
 maintaining favorable selectivity [51 ]. Clearly, enhancing selectivity of a reaction
 will ultimately reduce the volume of byproducts generated, and potentially the
 volume of waste emanating from a particular process.

    Hydrogenation is generally exothermic, and removing heat from the process is
 thus more of a problem than injecting heat [53]. In this case, the use of a supercritical
 fluid may or may not be advantageous. Liquids are useful as heat transfer fluids in
 that one can employ the heat of vaporization to absorb excess heat. Convecti ve heat
 transfer, which will depend on both fluid velocity and fluid physical properties, may
 or may not be more successful in a supercritical fluid, depending on the exact condi-
 tions. For example, the  magnitude of heat transfer is related both to the Prandtl
 number and Reynolds number [23]: Prandtl numbers for supercritical fluids are typi-
 cally lower than for liquids, while the Reynolds number for a supercritical fluid could
 be quite a bit higher (given that kinematic viscosity for supercritical fluids is high) at
 constant velocity. Heat removal is important, in that inability to effectively remove
 heat could lead to loss of selectivity. Liquid CO, could be useful in this regard, as
 boiling often is employed as a means by which to absorb excess heat.


 2.3  Heterogeneous Hydrogenation in CO2

    As mentioned above, the key "green" driving force behind the use of a super-
 critical solvent  rather than an organic solvent  in  a heterogeneous reaction  is the
 elimination of transport resistance (owing to diffusion of the gas across the liquid-
 vapor boundary)  and potentially a more efficient reaction. Ease of separation of
 products from reactants also is often mentioned, but not typically evaluated. In-
 deed, products  and reactants may be more easily separated in the conventional
 analog via a simple distillation. Baiker [54] has reviewed progress in heterogeneous
 reactions in supercritical fluids up to 1999; we will therefore cover only the most
 important discoveries made prior to 1999, and will focus on key strides made since
then.

    Harrod and colleagues [55] have successfully  performed the hydrogenation of
 fats and oils using supercritical propane; propane was employed to allow for solubil-
 ity of both the substrates (whose solubility in CO, is poor) and hydrogen, which is
completely miscible with any supercritical fluid. The homogeneous propane/Hysub-
strate mixture was fed into a  packed bed containing a commercial Pd catalyst—
extremely high reaction rates were indeed  achieved (gas-liquid transport resistance
being eliminated) and the concentration of trans fatty acids (an undesirable byproduct)
was reduced.  Hence, the green advantages to this reaction would include reduced
waste content and smaller and more efficient reactors. However, the use of propane is
problematic, and it is not clear whether the  process advantages due to faster reaction
rate balance the disadvantages deriving from use  of a flammable  solvent and the
problems inherent to  high-pressure process design/development. Further, the cata-
lyst deactivated quickly, an important problem for both economic  and sustainable

-------
32
     reasons [51.52]. Tacke and colleagues  [56] also investigated the hydrogenation of
     fats and oils (over a supported Pd catalyst), although they employed CO, as the
     supercritical solvent. Again, rates were shown to be significantly higher in the
     supercritical case (six-fold increase in space-time yields), and selectivity and catalyst
     lifetime also were improved. Each of these features contributes to enhancing the
     green potential of the process, while the need for high pressure operation detracts
     both from the cost and the sustainability (energy,  unit operation complexity). Macher
     and Holmquist [57] also examined the hydrogenation of an oil in supercritical pro-
     pane:  similar results to those found by Harrod were obtained. King and coworkers
     [58] examined the hydrogenation of vegetable oil and fatty acid esters over nickel
     catalysts using both CO, and propane as supercritical solvents, and  under condi-
     tions where either one or two fluid phases existed in the reactor. This approach is
     interesting, as it ultimately could prove to provide a useful engineering solution to
     the problem of solubilizing substrates in  CO, at moderate operating pressures.

        Indeed. Chouchi and coworkers [59] recently examined the hydrogenation of
     pinene (over Pd/C) in supercritical CO,. They  found that the rate of the reaction
     was significantly faster in the two-phase regime (i.e.. lower pressures) than when
     the pressure was raised to the point where only a single fluid phase existed. The
     reason for this seems clear: the Chouchi study was performed by charging a known
     amount of each of the ingredients to the reactor, then pressurizing with CO,. The
     partitioning of compounds between phases (in  the two-phase system) must have
     been such  that the concentration of reactants in the lower phase was higher than
     under single-phase conditions. In other words, raising the pressure to create a single
     phase simply diluted the reactants, lowering the rate. Note that the concentration of
     CO, in the lower phase (in the two-phase system) was likely to be substantial, as
     CO, should interact favorably with a volatile,  low molecular weight compound
     such as pinene. Further, the concentration of hydrogen in the lower phase also must
     have been substantial to support the high rate observed, and hence we see that CO,
     can swell an organic substrate significantly and carry substantial amounts of hydro-
     gen into a "swollen" liquid phase. Therefore, CO, could function as a "reversible
     diluent," much in the same way that it is employed as a "reversible plasticizer" in
     polymer science [60]. In this case, addition of CO, at relatively low pressures would
     enhance solubility of H, in the substrate, raising rates while not impacting process
     costs precipitously. Even safety could be improved, as previous work has shown
     that addition of CO, to a mixture of hydrogen and air expands the nonexplosive
     regime more so than addition of nitrogen [ 11 ]. As such, a sudden leak in the reac-
     tor, leading to a mixture of CO,, air, and hydrogen, would  still be safer than the
     same case where nitrogen was being used as the pressure-transmitting fluid. Use of
     CO, in such reactions could thus be green, safe, and practical.

        Bertucco [61 ]. and later Devetta [62], also showed the advantages of using a
     multiphase system in their work on the hydrogenation of an unsaturated ketone
     over a Pd/alumina catalyst. These researchers found that one could eliminate trans-

-------
 port resistance while operating in the three-phase (solid catalyst plus liquid plus
 gas) regime. Here again, the fact that CO,'s presence in the lower liquid phase
 greatly enhances the solubility  of  hydrogen in the  liquid (substrate  plus CO,)
 allows one  to eliminate transport resistance without the need to apply pressure
 high enough to create one phase. Consequently, one could conceivably render the
 reaction more efficient (and hence  less wasteful) and economically practical  by
 using moderate pressures.

    Arai and coworkers examined the hydrogenation of unsaturated aldehydes in
 both CO, and ethanol over a Pt/'Al,O, catalyst [63). The selectivity of the reaction
 towards  unsaturated alcohol in CO, was significantly belter than that in ethanol:
 while increasing the pressure in the CO, case improved selectivity, the opposite
 occurred when increasing the hydrogen pressure in the ethanol analog. Indeed, here
 is a case where the use of CO, appears to enhance selectivity, and thus reduce waste
 in a reaction versus the "liquid" analog.  It is not clear from the discussion by Arai
 whether this improvement in selectivity is enough to offset the difficulties involved in
 scaling up a high-pressure process, and whether the energy  input to the CO,-based
 analog is more  or less  than the liquid case.  Interestingly. Arai did not observe the
 rapid catalyst deactivation formerly observed by Minder and colleagues [64] during
 hydrogenation in CO, over a platinum catalyst. Minder's results were readily ex-
 plained by formation of CO and other poisoning species owing to the hydrogenation
 of CO, itself: it is not clear why Arai was able to avoid this problem.

    Poliakoff and colleagues [65] have evaluated the efficiency of hydrogenation
 of a wide variety of substrates in supercritical fluids (propane and CO,) over a Pd
 catalyst in a continuous flow reactor. Substrates included aromatic alcohols, alde-
 hydes, ketones,  unsaturated cyclic ethers, nitro compounds, oximes. and Schiff
 bases.  Reactions were conducted at temperatures ranging from 360 to 670 K at
 pressures between 80 and 120 bar. All of the substrates examined could be hydro-
 genated to some extent, with measured space-time yields exceeding 2 x 105 kg-h'-nr
 ' for the hydrogenation of cyclohexene. Given the high temperatures employed, the
 relatively low pressure, the presence of significant amounts of hydrogen, and the
 low volatility of some of the substrates employed, it is highly likely that two or more
 phases existed in the reactor during the initial phases of the process. CO,'s density
 will not be "liquid-like" at these pressures and temperatures, and hydrogen will act
 as a nonsolvent owing to its low critical temperature (and hence low reduced den-
 sity at the reaction conditions). Poliakoff  examined  the phase  behavior in the
cyclohexene-to-cyclohexane system and indeed found that multiple phases exist
 initially,  while a single phase  forms near the end  of the reaction. Single-phase
 behavior results because the temperature increases to a point above the critical
 temperatures of both cyclohexene and cyclohexane.  Whereas Poliakoff demon-
strated the breadth of continuous hydrogenation in CO,, lack of comparisons with
traditional hydrogenation reactions make it difficult to judge whether the technol-
ogy ultimately will be deemed "green." Catalyst lifetime, for example, is not men-

-------
34
     tioned—rapid loss in activity could render this technology less than adequate from
     both green and financial perspectives. If CO,-based hydrogenation allows for elimi-
     nation of significant volumes of solvent without greatly increasing energy or cata-
     lyst demand, then this technology ultimately could he both economically success-
     ful and green.

         Subramaniam and coworkers [27] also  examined the hydrogenation  of
     cyclohexene to cyclohexane (over Pd/C) in supercritical CO,, although under con-
     ditions where the system remained single phase throughout  the reaction and the
     temperature was held at a constant 343 K. The reaction remained stable over peri-
     ods exceeding 20 hours, and catalyst activity was maintained at a high level by
     pretreating  the cyclohexene feed to remove deleterious peroxides.  No CO or for-
     mate development was observed. Although this work does not suggest as to how
     or why such reactions could be considered "green," it does demonstrate that stable
     (with respect to temperature and pressure) catalytic hydrogenation in a continuous
     reactor using CO, as solvent is readily achievable. Again, the assumption here is
     that use of CO, will eliminate the gas-liquid interface, rendering the reaction more
     efficient and potentially less wasteful. Subramaniam has authored a comprehensive
     review on process design issues inherent to catalytic processes performed in car-
     bon dioxide [52].

        Hancu and Beckman [66] examined the hydrogenation of oxygen (production
     of H,O,) in CO, under both liquid and supercritical conditions. Hydrogen peroxide
     currently is  produced via hydrogenation  (over a Pd supported catalyst), then oxi-
     dation of a 2-alkyl anthraquinone (AQ) in an organic solvent (see Figure 5). Whereas
     H,O, is widely accepted as a green oxidant, the process by which it is manufactured
     exhibits  a number of less-than-green attributes. First, use of the organic solvent
     (coupled with the liquid-liquid extraction against water used to recover the prod-
     uct) creates  a significant contamination issue, one that currently is remedied using
     energy-intensive distillation. Further, because each of the reactions are transport
     controlled (again, by the rate of diffusion of H, orO, from the gas to liquid phase),
     CSTRs (continuous  stirred tank reactors) are used, allowing for a range of an-
     thraquinone residence times  and hence  over hydrogenation  of the AQ to form
     waste byproducts. Gelbein [67] has estimated that one-third of the cost of H,O, can
     be tied directly to anthraquinone and solvent makeup/regeneration; approximately
     150,000 tons of anthraquinone are produced each year simply to support consump-
     tion in the AQ process for producing hydrogen peroxide.

        Hancu first examined the use of CO2 as the organic solvent in the anthraquinone
     process by generating a highly CO,-soluble analog to conventional alky 1 anthraquino-
     nes (alkyl AQs exhibit solubilities in  CO, that are three orders of magnitude below
     what is employed in the commercial process). These fluoroether-functional AQs ex-
     hibited complete miscibility with CO,; maximum miscibility pressures were sensitive
     functions of anthraquinone composition and topology. Hancu  showed that kinetic
    control could be  obtained in both the hydrogenation and oxidation reactions using

-------
                                                                                  35
                                      : H.' I mtumn
FIGURE  5  Schematic of the Anthraquinone Route to Hydrogen Peroxide
              [15].
CO, as the solvent. Here, use of CO, eliminates the need for the distillation train, as
contamination of the aqueous phase by solvent and other byproducts is not an issue.
Further, although  the solvent in the conventional process is prone to both hydro-
genation and oxidation, this is not the case for the CO, analog.

    Despite  the promising laboratory results, Hancu's process in its original state ex-
hibited a critical economic flaw, yet one that could be corrected given recent results.
The fluoroether-functional AQ will be significantly more expensive than an alkyl AQ,
and pressures required to maintain a homogeneous mixture will be high, despite the use
of the CO,-philic AQ. If. however, we examine the results of Bertucco, Chouchi, and
Devetta [59, 61, 62], it is clear that an alternative route exists where one could take
advantage of the green aspects of CO, use while minimizing the AQ cost issues and
reducing the operating pressure. The works cited in the previous  sentence show that it
is quite possible that one does not need to achieve a single phase of hydrogen, CO,, and
substrate to eliminate gas-liquid diffusional limitations to reaction. In gas-liquid reac-
tion systems, often the primary resistance to transport is the low solubility of the reac-
tant gases in  the liquid phase and slow diffusion across the interface. The high degree of
swelling of a substrate by CO, can allow for significant increases in hydrogen solubility
in the liquid  phase, while the low viscosity of carbon dioxide enhances diffusion rates.
Thus, it is quite likely that one could derivatize an anthraquinone with an inexpensive
oligomer (such as  a short chain polypropylene oxide or silicone)  that would (a) not
raise cost significantly, (b) transform the crystalline, high melting alkyl AQ to a low
melting (or amorphous) derivatized AQ that would (c) swell significantly with CO, at
moderate pressures (less than 100 bar), allowing (d) a low viscosity liquid phase with
significant hydrogen solubility. This would render the oxidation process more tractable
as well, because one could employ air (instead of O,), where the nitrogen would by and
large remain in the upper gas phase. Hence, a CO,-based version  of the AQ process
could be rendered greener (through elimination of the solvent waste and energy load
reduction), while not detracting from the economics.

-------
     As noted in Section 1.6. a key future research issue that will impact heterogeneous
 hydrogenations in CO, is the lifetime of the catalysts, particularly the widely used pal-
 ladium catalysts. The literature contains examples of successful hydrogenations over
 Pd in CO,, and also examples where the rapid formation of CO led quickly to catalyst
 poisoning and deactivation. Subramaniam's group recently has presented a rationale
 [29] for the seemingly contradictory results in the recent literature. They showed (using
 high pressure FT-IR) that CO forms very quickly (within minutes) on Pd in a mixture of
 CO, and H,, and then over much longer times alters  its mode of binding to reduce
 catalyst activity. Temperature is a key parameter in this process, where temperatures
 above 343 K seem to greatly accelerate the process. Longer residence times (as would
 be experienced in batch reactors or CSTRs) also enhance the rate of poisoning.


 2.4   Homogeneous  Hydrogenation in CO2


 2.4.1   CO2-Soluble Catalyst Design

    Clearly, the most pressing issue one must deal with to conduct a homogeneous
 hydrogenation in a supercritical fluid is that of catalyst and substrate solubility. Carbon
 dioxide is without question the most popular solvent of those with a readily accessible
 (less than 370 K) critical temperature. However. CO, also is a feeble solvent [68, 69],
 whose inability to effectively solvate compounds of interest has greatly inhibited com-
 mercial development in the past. Although many metal-containing catalysts exhibit low
 solubility in carbon dioxide at moderate pressures, simple metal carbonyls are known
 to be miscible with CO, under relatively mild conditions [30. 70], and as such have
 been used successfully to catalyze reactions in carbon dioxide. In general, if the catalyst
 in question is a relatively volatile liquid, chances are good that it will exhibit accessible
 (less than 500 bar) miscibility pressures in carbon dioxide.

    For the case of those metal catalysts whose ligand design renders them poorly
 soluble in CO,, work performed since 1990 [71-73] has  identified  a number of
 functional groups that are decidedly "CO,-phiIic", such that derivatization of cata-
 lyst ligands with such groups enhances the solubility of catalysts in CO, to the
 point where homogeneous hydrogenation reactions are feasible. The most widely
 used of the CO,-philic groups for catalyst ligand preparation are—(CF:)-'s, used in
-(CH,)x(CF,)v-CF, "ponytails" where x ranges generally from 0 to 2  and y ranges
 from 0 to 6. The use of such groups creates a complex optimization problem for
 those wishing to scale up such processes:

•   The solubility  of the catalyst is sensitive to the  length (and number) of the
    fluorinated pony tails—longer (or more) tails tends to lower the pressure re-
    quired to solubilize a given concentration of catalyst [ 15,74.75]; lower operat-
    ing pressure means lower capital  investment. At the same time, increasing the
    percentage of fluorine in  the catalyst raises the cost owing both to  synthetic

-------
                                                                               37
    cost and increased catalyst molecular weight. The presence of the fluorines in
    the ligands can affect the electronic environment of the metal, either enhancing
    or detracting from the efficiency of catalysis.

 •   Recently, it has been shown that low molecular weight fluorinated sulfonate
    surfactants (perfluorooctanyl sulfonate (PFOS) and analogues) persist in the
    environment [42. 76]. If restrictions associated with PFOS type materials are
    extended to cover other low molecular weight fluorinated compounds, this
    would further raise the cost involved with the use of fluorinated catalysts.

    Whereas conducting homogeneous hydrogenation in an alkane lessens problems
 owing to the weak solvent power of CO,, the added liability due to the flammability of
 the mixture has dampened enthusiasm for such reactions. As mentioned previously.
 one must be aware that running a hydrogenation reaction in CO, can create byproducts
 owing to the reaction of hydrogen with CO, itself—such side reactions can be inhibited
 through proper catalyst design or choice of operating conditions.


 2.4.2  Engineering Rationale  for Homogeneous  Versus
        Heterogeneous  Catalysis

    In homogeneous hydrogenation. the catalyst has been designed such that it is
 soluble in the liquid phase; the  ligands of the catalyst usually are constructed to
 produce high selectivity to product. The rationale for conducting homogeneous
 hydrogenation reactions in CO, has three primary thrusts: (1) that operation in CO,
 eliminates the need for organic solvent, (2) operation in CO, eliminates the gas-liquid
 interface and hence allows for kinetic control over the reaction, and (3) use of CO, will
 alter the selectivity of the reaction (hopefully for the better). Much of the recent work
 on homogeneous hydrogenation has been directed at asymmetric synthesis, with the
 general hypothesis that use of CO, could possibly alter the enantioselectivity of the
 reactions concerned.

    The rate of a homogeneous hydrogenation  reaction conducted in an organic
 solvent or water is likely to be governed by the rate at which hydrogen diffuses
 across the vapor-liquid interface. As such, elimination of this interface (via opera-
 tion in CO,) eliminates this transport resistance. Indeed, because the catalyst in this
 case is soluble, elimination of the interface entirely eliminates transport resistance.
To allow direct replacement of the organic solvent in a homogeneous hydrogena-
 tion reaction with CO,, both the catalyst and the substrate must be soluble in CO,.
Consequently, the majority of the scientific effort in the literature on homogeneous
 hydrogenation in CO, is directed at synthesis of CO,-soluble analogues of conven-
 tional catalysts. Substrates must be chosen that are CO,-solubIe.  and hence one
observes predominantly "model" compounds employed rather than  necessarily
compounds of industrial interest.

-------
    One could pose the question. "If a liquid substrate is being employed, why not
 simply run the reaction using the homogeneous catalyst neat, in the ahsence of
 any solvent?" The solubility of hydrogen in organic liquids typically is quite low.
 and hence running the hydrogenation of a neat substrate will encounter signifi-
 cant transport resistance (of hydrogen across the interface) to reaction. If carbon
 dioxide readily dissolves or swells the liquid phase (catalyst and suhstra(e). the
 rate of reaction can increase owing to enhanced hydrogen concentration at the
 locus of reaction, despite the presence of CO,, a diluent.

    An example of the use of homogeneous catalysis to achieve an engineering
 goal was reported by Hancu and Beckman [ 15], who examined the generation of
 H,O, in CO, directly from H, and O, in a single step using a CO,-soluble palladium
 catalyst. This process has been examined in industry for more than 2 decades, as
 elimination of the anthraquinone from the process eliminates several unit opera-
 tions and greatly reduces raw material input. If one examines Gelbein's numbers for
 the economics of H,O, production [67], one would estimate that using the direct
 route would reduce the cost of production by more than 50 percent, a significant
 amount for a commodity process. Hancu proposed that one could generate H,O,
 in CO, (from H, and O,) using a soluble palladium catalyst, where the H,O, then is
 rapidly stripped into water. The green aspects of this process include elimination
 of solvent waste and anthraquinone input/byproducts, elimination of the distilla-
 tion train and the associated energy input, and elimination of several unit opera-
 tions and the associaied energy input. The process could be run  continuously
 and the product recovered  from CO, without a large pressure drop,  rendering the
 process economics more favorable. Previous work  on the direct route to H,O, has
 focused on the balance between safety  and productivity, where  most of the pat-
 ented processes employ water as the reaction  medium to maintain  safety. How-
 ever, because the solubility of H, and O, in water is so low, the productivity  of
 these processes is not sufficient to merit scale-up. In addition,  the Pd catalysts
 employed tend to catalyze degradation  of H,O, as well as formation, and hence
 running the reaction in water does not lead to the desired productivity. Hancu
 showed that one could employ a CO,-soluble catalyst, and run the reaction in CO,
 without transport limitations and in a nonexplosive concentration regime where
 rates are high. Future work is needed in this area with respect to optimizing cata-
 lyst performance and lifetime, yet this is a good example of the use of homoge-
 neous hydrogenation in carbon dioxide to  accomplish what are normally per-
 ceived  to be process goals.

    Unlike in the previous example, in cases where a separate aqueous phase is not
present, one may be able to take advantage of the favorable properties of CO, (with
 respect to hydrogenation) while avoiding some of the negative process issues by
employing a gas-liquid rather than one-phase system. For example, it is known thai H,
 is poorly  soluble in  most organic liquids, and it is expected that a hydrogenation  in
organic solvent would be transport limited.  If one knows the fundamental kinetic

-------
                                                                                39
 parameters of the reaction, one should be able to predict at what [H,] to [substrate]
 ratio the reaction could be controlled by the underlying kinetics, and calculate the target
 [HJ lor the reaction in the presence of CO,. If the substrate is a liquid, one should be
 able to find conditions where a two-phase system (H,-CO,-substrate) exists, yet where
 substantial amounts of hydrogen are dissolved in the lower phase. As described previ-
 ously, liquid-liquid phase diagrams of CO, and larger molecules are typically asymmet-
 ric, and hence operation at high concentrations of substrate is  possible at relatively
 lower pressures. Further, the catalyst would be required to dissolve in a mixture of
 (primarily) substrate and CO,, suggesting that one might not have to fluorinate the
 catalyst to achieve solubility in the proper phase. Thus, by operating in the two-phase
 region, one could operate at lower pressure with the original catalyst while also eliminat-
 ing the  need  for the organic solvent and the transport resistance  to reaction. Ideal
 substrates would be those that are relatively high in molecular weight, or are polar, yet
 also are liquids (or low melting solids, where CO, can depress the melting point [77]).

    Another interesting possibility would, in fact, involve functionalization of the
 catalyst (fluorination) to allow better solubility in CO, while also operating in the
 two-phase regime. Here, the presence of the CO, in the lower phase would serve to
 not only allow higher hydrogen concentrations but also would solubilize the cata-
 lyst. Upon removal of the CO,, the catalyst would precipitate, allowing recycle. This
 would support the CO,-based analogy to recent work by Gladysz and colleagues
 [78], where a fluorinated catalyst was developed that was insoluble in the reaction
 solvent, but dissolved upon heating. Temperature was used as the reversible trig-
 ger to allow catalyst use and recovery. Recently, it has been shown that CO, itself
 also could be employed as a reversible solvation trigger [79].


 2.4.3   Chemical Rationale for Homogeneous Catalysis

    The final reason for conducting a homogeneous hydrogenation in CO, is the
 premise that use of CO, would alter the selectivity of the reaction in a positive way.
 Xiao, for example [80], examined the asymmetric hydrogenation of tiglic acid (2-me-
 thyl-2-butenoic acid) in CO, using a ruthenium catalyst; enantiomeric excesses in CO,
 were essentially no better than those found for the same reaction in methanol. Tumas
 [81] examined the hydrogenation of dehydroamino acids in CO,  using a cationic
rhodium catalyst—the fluorinated counteranion (3,5-bis(trifluoromethyl phenyl) bo-
rate  [BARF]  or inflate) enhanced solubility of the catalyst in CO,. Tumas found
 somewhat better enantiomeric excesses for some substrates in CO, versus hexane or
 methanol, but overall the performance of CO, was comparable to that of the other
organic solvents. Leitner [82] has used chiral indium catalysts to perform the hydro-
genation of imines in CO,. The catalysts were modified (using fluoroalkyl ponytails)
to permit solubility in CO,. Enantiomeric excesses in CO, were comparable to those
 found for the same reaction in dichloromethane, while rates were found to be much
higher for some substrates in CO, versus CH,C1,.

-------
40
         Recently. Tumas (83) and Jessop [84] explored the use of biphasic mixtures of
     ionic liquids and carbon dioxide to perform hydrogenations. Ionic liquids are salts
     (typically ammonium or phosphonium) that exhibit melting temperatures near or be-
     low room temperature. Ionic liquids behave as polar solvents, yet exhibit vanishingly
     small vapor pressures. In both the Tumas and Jessop studies, a CO,-insoluble cata-
     lyst was dissolved in the ionic liquid, which  then is  brought into contact with a
     mixture of CO,, substrate, and hydrogen. As has been shown by Brennecke [85]. ionic
     liquids absorb large amounts of CO, (mole fractions above 0.5) at pressures below
     100 bar. Further, the ionic liquid does not measurably dissolve in CO,. Consequently.
     both Tumas and Jessop were able to conduct reactions in the ionic liquid at very high
     rates (the high CO, swelling allowed for high H, solubility), where the product could
     be stripped  from the ionic liquid into CO, and the catalyst retained in the ionic liquid
     for recycle. Note that this is an analogy of the two-phase CO,/H,/substrate mixture
     mentioned  above, where the high  swelling of the lower phase by CO, eliminates
     transport limitations, while two-phase operation permits use of moderate pressure.

        To date, the ionic liquids (ILs) being explored as solvents are primarily based on
     imidazolium or pyridinium cations (some work also has  been conducted on phospho-
     nium ILs). Whereas these ILs are proposed as benign solvents (owing to their near-
     zero vapor pressures), it must be remembered that the toxicity and fate (in the environ-
     ment) of such materials currently  is not known.  In addition, because large-scale
     manufacturing processes for these solvents have yet to be established, the impact of
     such processes on the environment also is not known. In summary, the current crop
     of ILs ultimately may or may not be judged to be benign solvents.


     2.4.4   Homogeneous Hydrogenation and Material Synthesis

        Watkins has explored a truly novel means by which to apply homogeneous hydro-
     genation in CO, to creation of metal nanoparticles and thin  metal films. Watkins has
     found that certain metal complexes exhibit millimolar solubility in CO, at pressures
     below 100 bar. Exposure of these complexes to hydrogen under mild conditions reduces
     the metal to the zero  valent state, inducing nucleation  of pure metal. Watkins first
     employed this reaction to create small metal particles within polymer monoliths [86]. The
     complex is added to CO,, and this solution is brought into contact with the polymer,
     which swells accordingly. Hydrogen then is introduced, which reduces  the complex
     within the polymer, forming the nanoparticles. Recently. Nazem, et al. [87], and Howdle's
     group [88] have examined the impregnation of polymers with silver particle precursors.
     performing the reduction in situ to form the nanoparticle-impregnated material. In Howdle's
     work, the polymers involved (polylactic acid and analogues) were found to resist attach-
     ment by bacteria owing to the antibacterial  properties of silver. Use of nanoparticles
     allowed for useful antibacterial properties, despite low loadings of silver.

        Watkins has further extended [89] this concept into the realm of green chemistry
     by adopting the process for use in creating thin  metal films. In the microelectronics

-------
                                                                               41
 industry, thin metal films can be generated on an inorganic substrate via vapor depo-
 sition, or via dip coating and reduction from an aqueous solution. The former can
 only be applied to volatile precursors, while the latter route produces very large vol-
 umes of  metal-contaminated aqueous waste. Watkins has found that homogeneous
 hydrogenation of metal complexes in CO, allows generation of con formal metal films
 on substrates with submicron  features, and that the only waste produced  is a low
 molecular weight alkane byproduct. Small trenches and pits can be coated easily
 because CO,'s low interracial tension permits wetting of even complex features.
 Watkins has demonstrated this concept with platinum, palladium, and nickel—a re-
 cent paper [89a] shows that the concept can be extended to copper as well.

    This  technology is undeniably green, and could be readily  applied to a variety
 of metal film applications, particularly if it can be demonstrated that metal deposi-
 tion can be targeted (patterned).


 2.5    Industrial Activity: Hydrogenation in CO2

    Of the relatively small number of patents (1996-2001)  that directly cover hy-
 drogenation in supercritical fluids, two are worthy of special consideration. First.
 Harrod and colleagues [90j describe the hydrogenation of fatty acids in supercritical
 fluids, technology that has formed the basis for a small startup company in  Europe.
 Likewise. Poliakoff and colleagues [91] have described the hydrogenation of a
 variety of substances in supercritical fluids, technology that has formed the basis/
 motivation for a pilot-scale plant constructed for Thomas Swan Company (Durham.
 UK) by Chematur (Karlskoga. Sweden). It should be noted that Chematur, a com-
 pany known for its supercritical water work (assets in both the United States and
 Europe),  has acquired the high pressure-related portion of Rauma (Finland), in-
creasing its capabilities in the design of processes capable of handling  supercritical
fluids. The Thomas Swan facility, which was scheduled to start up in September
 2001 (and did in early 2002). will be able to generate 1.000 tons per year of prod-
ucts, including the results of hydrogenations and Friedel-Crafts acylations and alky-
 lations conducted in supercritical fluids. At this time, it appears  that the Swan facil-
 ity will be used (at least in part) as a pilot-scale or semi-works facility to evaluate
the use of supercritical fluids as solvents in various chemical reactions.


2.6    Summary:  Hydrogenation in CO2

   In summary, hydrogenation in supercritical fluids has been extensively investi-
gated  over the past decade, and it is clear that hydrogenation reactions can be suc-
cessfully  conducted in CO, and other fluids. It is  not always  clear, however, what if
any green advantages are obtained via operation  in a supercritical solvent, as many
authors do not draw comparisons to conventional processes. Nevertheless, some gen-
eralizations can be made:

-------
42
      1.  The primary rationale for use of a supercritical solvent in hyclrogenalion reac-
         tions is the elimination of transport limitations to reaction through enhance-
         ment of the solubility of hydrogen at the reaction locus. Hydrogen is poorly
         soluble in conventional hydrocarbon liquids and water, and use of CO,  (and
         propane, to a lesser extent) as the solvent has been shown to enhance H,
         solubility and hence improve the efficiency of the reaction. Attaining kinetic
         control over the reaction can lead to reduced byproduct formation and lower
         energy input, although in the  case  of typically exothermic hydrogenations.
         energy removal is more important than energy addition.

     2.  A key point that arises if one examines the recent literature is that one does not
         need to create a single phase (of supercritical fluid, substrate, and hydrogen) to
         create a situation where transport limitations can be eliminated [59.61.  62]. For
         example, one can attain kinetic control over the reaction simply by ensuring that
         a significant amount of CO, is present in the liquid phase (maintaining a gas
         phase of CO,/H,). Here, the CO, functions as a diluent (and viscosity reducer)
         that enhances the solubility  of hydrogen in the lower  phase. The enhanced
         hydrogen solubility  more than makes up for the dilution effect from the CO,.
         Although elimination of the resistance owing to transport of H, into the liquid
         phase does not by definition create kinetic control over the reaction (resistances
         owing to diffusion to and within the catalyst also exist), the previous work has
         shown that the solubility of H, in the liquid is typically the limiting factor. The use
         of CO, as the "H, solubility enhancing diluent" could have broad  ramifications
         on the practicality for conducting hydrogenations in supercritical fluids, in that it
         could make the use of benign (and nonflammable) CO, more viable. For example.
         Harrod [55], as well  as others, has employed propane as a supercritical solvent
         solely to enable formation of a single phase with substrates whose solubility in
         CO, is poor. It may be possible to both employ CO, as the "diluent"  and eliminate
         transport limitations to reaction, rendering the reaction more efficient while avoiding
         the flammability problems inherent to propane. The use of CO, as "diluent" also
         could render the anthraquinone process described by Hancu [66] much more
         economically efficient as well as greener. This situation obviously best applies to
         liquids (or low melting solids) that are relatively nonvolatile. The  use of a two-
         phase (liquid-vapor) mixture also can help with heat transfer, as the  boiling of the
         liquid can be employed to absorb excess heat.

     3.   Regarding asymmetric hydrogenations. the key green advantages to this work
         seem to be the elimination of organic solvent and improved selectivity. However.
         the results in the literature have not established that significantly greater selectivi-
         ties are likely to be obtained solely through replacement of a conventional solvent
         with a supercritical fluid (primarily CO,). Solvent polarity does impact selectivity, so
         it is possible that reactions will be identified where use of CO, provides selectivity
        benefits. Most of the work on asymmetric hydrogenation has employed homoge-
        neous catalysts: catalyst lifetime and recovery are unresolved issues in this area.

-------
                                                                               43
 4.  The poisoning of noble metal catalysts via the formation of CO from CO, and H,
    could seriously impact the economic viability of hydrogenation processes con-
    ducted in carbon dioxide. Subramaniam [29] has begun to elucidate the effect of
    various process parameters on this process; more research in this area clearly is
    merited.


 2.7   Hydroformylation in CO2

    Hydroformylation, the reaction of hydrogen and CO with an alkene to form
 aldehydes (Scheme I), is practiced industrially (the "oxo" process) on an enormous
 scale using alkenes of various chain lengths [14].

 Scheme I:
    In one form of the process, cobalt is fed to a reactor containing the oxo gas (H,
and CO) and the alkene, where a reaction takes place to form the cobalt hydro-
carbonyl, the active catalyst species. Alkene then is converted to aldehyde in the
liquid phase (the liquid is either a mixture of alkene substrate and alkane solvent or
simply the alkene alone). The reaction takes place under rather severe conditions,
200 to 300 bar and temperatures between 410 and 450 K. The reaction produces
the needed aldehyde(s), as well as residual alcohols and alkane. The useful prod-
ucts are recovered, and the remainder are combusted. The selectivity of the process
is approximately 85 percent to the aldehyde products. The catalyst is recovered as
a cobalt "sludge" and regenerated/recycled. In a variation on the basic oxo process,
a water-soluble cobalt catalyst is employed that can be recovered via retention in
the aqueous phase at the end of the process. Hence, the reaction is biphasic  in
nature—poor solubility of higher alkenes limits this process to C, - C4 alkenes.

   The rationale for operating a hydroformylation reaction in a supercritical fluid is
similar to that for hydrogenation. Hydroformylation involves the use of two gaseous
reactants (CO and H,,), and hydroformylation of a nonvolatile or low volatility liquid
substrate will likely be limited by the solubility and transport of the gaseous reactants
from the  vapor to the liquid phase. As for the case of hydrogenation in supercritical
fluids, research on hydroformylation has been conducted using both homogeneous
and heterogeneous catalysts. Further, the majority of the studies reported employed
CO, as the solvent. The "green" rationale for exploring this class of reactions using
supercritical fluid solvents is that creation of a more efficient reaction (kinetically
controlled, more selective) will result in the  production of fewer byproducts and

-------
44
     perhaps require lower energy  input. Given the conditions under which the process
     currently is operated,  if one could produce the same space-time yield of product
     using lower pressure and/or temperature, the savings could he significant.

         In summary, the green premise behind conducting hydroformylation in CO, is
     not only to replace solvent (only a factor in some oxo processes), but also to create
     a more efficient reaction, and  hence reduce byproduct waste and energy input.


     2.7.1   Homogeneous Catalysis of Hydroformylation in CO2

         Rathke and colleagues [7()| reported the hydroformylation of an olefin in CO,
     in 1991. A cobalt carbonyl catalyst (soluble in CO, without modification) was used
     to promote the generation of butyraldehyde from propylene. CO,, and hydrogen.
     Rathke reported that operating the reaction in CO, produced a somewhat improved
     yield of linear to branched aldehyde. The rate of formation of both cobalt interme-
     diates and aldehydes was found to be similar to values found when the reaction was
     performed in conventional nonpolar solvents.

         Leitner's  group [92]. as well as Erkey and colleagues  [93]. reported hydro-
     formylation of an olefin in supercritical CO, using a homogeneous rhodium catalyst
     in 1998. where the now classic strategy of derivatizing the catalyst ligands with flu-
     orinated ponytails was used to enhance catalyst solubility. Leitner found that the
     reaction (hydroformylation  of l-decene) readily goes to completion in CO,, with
     catalyst activities similar to those reported in liquid systems. Erkey's results for l-
     octene are similar. As Leitner points out, the long-chain alkenes employed as sub-
     strates for the reactions in CO, would  likely not be soluble in water, and hence the
     well-known aqueous Rh/triphenyl phosphine trisulfonate catalyst system cannot be
     used to generate long-chain aldehydes. Potentially, this is a means by which to pro-
     duce valuable products while replacing an organic solvent with CO, (as long-chain
     aldehydes could only be produced in bulk or in organic solvent). Further, reaction in
     CO, will allow much higher CO and H, concentrations and potentially much faster
     rates. Indeed. Erkey and coworkers suspected that the high CO and H, concentra-
     tions were potentially the cause for differences in the rate expression between
     hydroformylation of l-octene conducted in CO, (using a fluorinated phosphine Rh
     catalyst) versus that in a conventional liquid. Interestingly. Leitner found that internal
     olefins,  which are "notoriously unreactive" in conventional solvents,  are
     hydroformylated with high rates and excellent yields. Erkey examined the effect of
     ligand structure (most notably, position and nature of the fluorinated ponytail) on the
     rate of hydroformylation, and found that the activity decreased as the basicity of the
     ligand decreased. Increasing the fluorine content of the ligand would tend to enhance
     the solubility of the catalyst in CO,, but decrease the activity. Indeed, increasing the
     fluorine content of the ligand also will increase the cost (both through an increase to
     molecular weight and the inherent cost of fluorinated compounds). Consequently, an
     optimization problem is created, where  increasing fluorine content to the ligand low-

-------
                                                                                45
 ers certain capital and operating costs owing to lower required operating pressure.
 while raising catalyst cost. A possible solution to this problem would be to decouple
 the effects that create the optimization problem (i.e.. find a way to enhance solubility
 of the catalyst without resorting to fluorination). Xiao's group at the University of
 Liverpool has examined this route [94], employing carbonyl groups attached to aryl
 phosphine ligands to enhance catalyst solubility in CO,.

    Akgerman's group has investigated homogeneous hydroformylation in super-
 critical CO, fora number of years [95]. In 1997, Guo and Akgerman reported the
 homogeneous hydroformylation of propylene in CO, using a soluble cobalt cata-
 lyst. Both the rate constant and the selectivity were found to be functions of pres-
 sure, each increasing significantly as pressure increased from 90 to 190 bar. The
 apparent effect of pressure on  the rate constant was attributed to potential limita-
 tions in catalyst solubility in the CO,/propylene mixture—as pressure increased.
 the catalyst solubility should increase, accounting for the observed effect. In a fol-
 low-on study published in 1999, Guo and Akgerman employed transition state theory.
 coupled with partial molar volumes calculated using the Peng-Robinson equation
 of state,  to attempt to explain the selectivity increase  with increasing pressure.'
 Calculations reproduced trends in both temperature and pressure-dependence of
 the rate and the selectivity. It is not clear whether this work has any "green" rami-
 fications, as the substrate employed (propylene) is a highly compressible fluid it-
 self, and might be expected to solubilize significant quantities  of hydrogen and
 CO. In this case, addition of CO, would tend to dilute the reactant concentrations.
 thereby slowing the rate. On the other hand, if it could  be shown that addition of
 CO, enhances the concentration of H, and CO significantly, then process advan-
 tages might be realized.

    Xiao and coworkers [96] also have examined homogeneous hydroformylation in
 CO,. They note, for example, that use of fluorinated aryl  phosphine ligands (as part
 of a rhodium catalyst) leads both to higher solubility in CO, and higher reaction rates
 (the latter owing to both electronic affects and solubility limitations of alkyl-ated
 phosphine catalysts). Comparison of the rates of hydroformylation of acryl-ates in
 CO, and toluene showed the expected enhancement (in CO,), owing to the consider-
 able increase in solubility of the reactants (CO and H,) in CO, versus toluene at the
 same pressure. Selectivities remained the same. As in other research on hydrogena-
 tion and hydroformylation in CO,, the "green" advantages of the process are sug-
 gested to be the increased rates owing to the higher solubility of H, and CO in CO,
 versus typical organic solvents,  plus the inherently benign nature of CO, versus other
 solvents.  However, these attributes may be offset by the high  pressure required to
operate in CO, (energy and capital requirements likely will be higher) and the in-
creased cost and potential environmental problems owing to the use of fluorinated
catalyst ligands needed to provide reasonable  solubility in CO,.

    It would be quite useful to explore the use of CO, as  a swelling agent  for a
liquid hydroformylation system, where the dilution effect is offset by the enhanced

-------
46
     solubility of gaseous reactants in the liquid phase owing to the presence of CO,.
     Catalysts could still be homogeneous yet not require tluorinated ligands. given
     that the continuous phase would be primarily alkyl-functional substrate (and prod-
     uct). Consequently, one could eliminate gas-liquid transport resistance while oper-
     ating at substantially lower pressures than those required for single-phase opera-
     tion. This indeed might be the process compromise that would provide the "green-
     est" operation. Note that this is  the opposite to what many authors recommend
     [97]—whereas a single phase is the best option for some processes, in cases where
     CO,/liquid substrate/gas reactive mixtures are being considered, two-phase opera-
     tion has significant advantages. Indeed, if one could operate a hydroformylation at
     high space-time yield at lower pressures and temperatures than the current process
     owing to the presence of CO,, the process would be both green and economically
     viable. As in the case of hydrogenation, the use of a two-phase (liquid-vapor)
     system would allow easy heat removal through boiling (and later condensation) of
     the liquid.

     2.7.2  Heterogeneous Hydroformylation in CO2

        Several research groups have evaluated heterogeneous catalysis of hydro-
     formylation in CO,; generally, yields were good and selectivities to linear aldehyde
     were excellent. For example, Poliakoff [98] used a rhodium complex (aryl phosphine
     ligands) immobilized on silica—selectivity to linear aldehyde was more than 90
     percent at 10 percent alkene (1 -octene) conversion. Clearly, use of an immobilized
     catalyst eases catalyst recovery and reuse issues. Poliakoff found no drop in cata-
     lyst activity after 30 hours of continuous use. Abraham [99] also has examined
     heterogeneous hydroformylation of propylene, focusing on the design of the cata-
     lyst to optimize performance. At first, Abraham's group focused on support design
     to try to minimize product sorption, while more recent work has targeted the design
     of "tethered" rhodium catalysts to try to achieve the advantages of both homoge-
     neous and heterogeneous catalysts. It is again interesting that  researchers have
     neglected to examine the question "Under what conditions will the use of CO,
     provide better results than when  using neat substrate?" Given that gases such as
     CO and hydrogen are poorly soluble in organic liquids, if CO, will swell the sub-
     strate substantially, then conditions may  exist where the concentration of hydro-
     gen in the liquid phase (of a two-phase mixture) may be such that the rate in such a
     situation is higher than  in the neat substrate case, despite the presence of a diluent
     (CO,). Such comparisons  would be useful  for the purposes of determining the
     viability of such CO,-based processes.


     2.7.3   Industrial Activity:  Hydroformylation in CO2

        Only one industrial patent of note [100], assigned to Mitsubishi Chemical Co.,
     was identified during our patent search. No scale-up work seems to have followed.

-------
                                                                               47

2.7.4   Summary: Hydroformylation in CO2

    In summary, one could report many of the same  conclusions regarding
hydroformylation in CO, as for hydrogenation in CO,. In hydroformylation, how-
ever, process conditions for the industrial route are rather severe, and hence if one
could obtain the high yields and selectivities of the industrial process but at mod-
erate conditions (p. T) via use of CO, as a solvent, the  process would be  both
greener and less expensive. A rich area for further work is in hydroformylation in
two-phase systems where CO, acts  as the "reversible diluent."


2.8  Oxidation in CO2

    At first glance, CO, appears to be an ideal solvent for use in oxidations. Unlike
most organic solvents, CO, will not oxidize further in the presence of oxygen and
catalysts, and use of CO, as the  solvent eliminates the solvent byproduct waste
stream that usually is expected in oxidations.

    Many of the conclusions found from recent research on  hydrogenation and
hydroformylation in CO, also can be applied to oxidations conducted in CO,. How-
ever, although hydrogenation and hydroformylation focused exclusively on H7 (and
H/CO) as reagents, oxidations conducted in CO, have been pursued using a variety
of oxidants. Clearly, however, the use of O, as a benign oxidant has received the most
attention, as it is ultimately the least expensive and most atom-efficient route. Re-
search on oxidation of substrates using O, in CO, has targeted the elimination of
transport resistance (as for hydrogenation and hydroformylation) through the elimi-
nation of the gas-liquid interface.  This is proposed to enhance the efficiency of the
reaction, leading to fewer byproducts. As in the preceding cases, it would be ex-
tremely interesting to examine oxidation in a single-phase system where CO., is the
minor component (a diluent for the  substrate or swelling agent) or in a two-phase
system where the substrate resides primarily in the lower phase.  The role of the CO,
is simply to enhance the solubility of oxygen in the substrate-rich phase, where we
assume  that the dilution effect owing to CO,'s presence is more than offset by the
enhanced oxygen concentration.  This would allow lower pressure operation and
might eliminate the need for fluorinated catalyst ligands (for homogeneous processes)
in that the catalyst must be soluble in a concentrated substrate-CO, mixture, rather
than a mixture that is primarily CO,. Indeed. Wu and colleagues [101] examined pre-
cisely this type of system, although it is not clear from the paper whether they
recognized the ramifications of their  work. Wu studied the oxidation of cyclohexane
with oxygen in the presence of an iron porphyrin catalyst and acetaldehyde where
CO, was the  solvent. The  yield (of cyclohexanol/cyclohexanone) increased with
pressure up to approximately 100 bar, then decreased sharply  at higher pressures.
Phase behavior measurements were not made, but qualitative observations (via sap-
phire windows in the  reactor) suggested that the drop in yield coincided with a

-------
48
     transformation from two phase to one phase. In this system, the presence of signifi-
     cant quantities of CO, in the lower phase of a two-phase mixture allows for solubili/a-
     tion of substantial quantities of oxygen, providing fora high rate of reaction. Trans-
     formation to a one-phase mixture merely produced a dilution effect, lowering the rate.

         An additional consideration that recommends the use of CO, as "diluent" rather
     than major component ("solvent") is  that oxidations using O, typically are con-
     ducted using air (O,/N,). Air is superior from an economic standpoint, as use of O,
     mandates somewhat energy-intensive O,-N, separation (and hence inadvisable
     from a green perspective). However,  if one were to use O,/N, in a single-phase
     system where CO,  is the primary solvent, nitrogen would build up in the system
     unless a concerted effort (pressure reduction) was made to continuously remove it.
     In a two-phase mixture where CO, is the minor component, the nitrogen concentra-
     tion in the lower phase would quickly saturate (equilibrium would be established
     with the upper phase), and this additional pressure drop and separation step is  not
     needed (a green advantage).


     2.8.1   Oxidations in CO2: Experimental Results

        Clearly, the oxidation of cyclohexane (first to cyclohexanone/cyclohexanol, sub-
     sequently to adipic acid) is one of the more commercially important oxidations
     performed industrially (Scheme II) [14].

     Scheme El:
        Cyclohexane is oxidized in the liquid phase using air (at temperatures of 395 to 435
     K and pressures in the  10-20 bar range)  to a mixture of cyclohexanone and
     cyclohexanol. Magnesium or cobalt salts are employed to catalyze the reaction. Srini vas
     and Mukhopadhy ay [ 102] examined the oxidation of cyclohexane in CO, with oxygen
     at temperatures between 430 and 470 K and pressures up to approximately 200 bar.
     Interestingly, a catalyst is not mentioned by the authors, despite the fact that one is
     employed industrially. The authors found that the condition of the feed (one phase.
     two phase, proximity to a phase boundary) exhibited a strong effect on the product
     profile and the rate of product  formation. Not  surprisingly,  given the discussion
     above, the highest rates (for both cyclohexane and cyclohexanol formation) were
     observed in the single-phase system where CO, was the minor component: that is.
     CO, was employed to homogenize the mixture of cyclohexane and oxygen, leading to
     high concentrations of each reactant and high rates.

-------
                                                                              49
    Another oxidation process of great import industrially is the formation of ep-
 oxides from alkenes. Most important is probably the generation of propylene oxide
 from propylene. Currently, propylene oxide is produced via one of three processes
 (primarily). First, chlorohydrin (from chlorine and propylene) can be reacted with a
 base to generate propylene oxide and salt (Scheme III); a very large volume of wash
 water (40 times the volume of product, which is then treated and emitted) is required
 to work up the product.

 Scheme III:
                     Ca(OH),  	»      /\   +CaCl,+
    One also can produce propylene oxide (PO) via a coproduct process where an
intermediate is peroxidized with oxygen, and the oxygen transferred to propylene,
creating propylene oxide and a byproduct alcohol (which then is transformed to a
coproduct) [14]. The most widely used coproduct processes for PO production
also create styrene or methyl tertiary butyl ether (Scheme IV).
Scheme IV:
                 C)    	*  ^V-OOH   +   J>\—OH
          OOH   +
    There is significant interest in designing a process that only produces PO from
propylene and oxygen, as methyl tertiary butyl ether is now environmentally sus-
pect and the demand for styrene tends to fluctuate and that for PO remains consis-
tently strong. As such, propylene oxide production is more energy intensive and
wasteful than desired because a coproduct must be produced along with PO. Con-
sequently, Baiker and colleagues [103] investigated the oxidation of propylene with
an oxygen/hydrogen mixture using a Pt/Pd on TS-1 (titanium silicate) catalyst in a
two-phase system (methanol was employed as the primary solvent). The reaction
proceeds via formation of hydrogen peroxide from H, and O, over the Pd, followed
by oxidation of propylene to PO. Both nitrogen and CO, were employed as solvents
for the H,/O, mixture. Baiker found that the yield of PO increased markedly  on
switching from nitrogen to CO, in the upper phase of the mixture, and that increas-
ing pressure enhanced the yield still further. As in previous cases, these results

-------
50
     may derive from the simple fact that use of CO, as the solvent for the reactant gases
     allows for greatly enhanced concentrations of these gases in the lower (or liquid)
     phase, thereby enhancing the rates.

         Eckert and colleagues, as well as Beckman. et al.. have investigated an interest-
     ing route to alkylene oxides [104]. As shown originally by Richardson and col-
     leagues [105], hydrogen peroxide will react with a bicarbonate salt under basic
     conditions to form the percarbonate ion. which then will react with alkenes to form
     the epoxide. This reaction is an  analogy to epoxidation using a hydroperoxide
     (such as t-BuOOH). Liquid CO, will dissolve in molar quantities in water, forming
     carbonic acid. Beckman and Eckert each showed that a biphasic CO,/H,O,/water
     mixture also will form percarbonate (upon the addition of appropriate amounts of
     base), and will epoxidize olefins such as cyclohexene oxide (Scheme V).

     Scheme V:
        The addition of a base is critical for achieving high activity. In general, sodium
     hydroxide is more effective than bicarbonate (likely as it raises the pH more effec-
     tively). Given Beckman's results, it would appear that percarbonate is formed both
     via reaction of H,O, and bicarbonate and via direct reaction between CO, and H,O,.
     Further, because the reaction is biphasic, addition of a CO,-philic surfactant en-
     hanced the rate dramatically, as would be expected. Likewise, addition of a phase
     transfer catalyst (a tetraalkyl ammonium halide) also enhanced  the rate. These
     epoxidations are intriguing as they employ only water, CO,, and H,O, as reactants
     and a catalytic amount of base. The primary drawback to this route is that hydrogen
     peroxide, although usually considered a commodity chemical, currently is too ex-
     pensive to use as an oxidant to produce PO.

        A number of other researchers have examined the oxidation of alkenes to ep-
     oxides using a variety of chemical strategies in carbon dioxide. Birnbaum [ 106], for
     example, employed a fluorinated (and hence CO,-soluble) porphyrin catalyst to
     oxidize cyclohexene to cyclohexene oxide. Not surprisingly, Birnbaum found that
     the selectivity  was significantly higher in CO, than in organic solvent, as operation
     in CO, does not produce solvent oxidation products. Loeker [107] examined the
     oxidation of olefins in CO, using oxygen and aldehydes as sacrificial co-oxidants.
     The reaction was heterogeneous, although it was the steel walls of the high-pres-
     sure reaction vessel that were employed as the catalyst. Finally, Haas and Kolis
     [108] found that one could readily oxidize olefins in CO, using t-butyl hydroperox-
     ide and a soluble Mo(CO)6 catalyst as an oxygen  transfer medium. Regarding
    epoxidations, the direct generation of propylene oxide from propylene would be the

-------
                                                                               51
 most significant "green" advance to be made in this area, yet use ot anything hut
 oxygen (or air) as the oxidant currently is too expensive.

    Wacker chemistry (the oxidation of an alkene to a ketone using a PdCi,/CuCl,
 catalyst) also has been examined using CO, as the sole solvent. Li and colleagues
 [109] examined the  oxidation of  1-octene in CO, and found that operation in a
 mixture of CO, and  methane!  led to higher selectivity to the methyl ketone than
 operation in either CO,  or methanol  alone. Because the phase behavior of the
 system  was not measured, the effects reported by Li cannot be explained com-
 pletely. For example, although it is known that the PdCl, and CuCI, catalysts are
 soluble in methanol and poorly soluble in CO,, it is not clear as to their solubility in
 the mixture of MeOH and CO,. Li's group also examined the oxidation of acrylic acid
 to the analogous 3.3-diaIkoxy propionate using  a similar catalyst system.

    In early 2002. Subramaniam's group [110] published the results of an interesting
 study on homogeneous oxidation performed in mixtures of carbon dioxide and conven-
 tional organic solvents (primarily acetonitrile). This study showed vividly that one can
 use judicious mixtures of solvent and CO, to truly  optimize the performance of a reac-
 tion. Use of CO, alone necessitated high pressures (hundreds of bar to dissolve both
 substrate and catalyst), and the low  polarity of pure CO, provided a nonideal medium
 for the catalyst. On the other hand, use of pure acetonitrile allowed operation at one
 atmosphere and provided the catalyst with a suitably polar environment, but the solu-
 bility of oxygen in the liquid phase was poor. When the right mixture of acetonitrile/CO,
 was employed, the catalyst activity  was high, and all components (oxygen, substrate.
 and catalyst) dissolved at pressures of only tens of bar. Study of more examples of this
 type of system may yield processes that are both greener than current methods and
 economically practical, particularly if one can ultimately eliminate the need for the or-
 ganic solvent and work with neat liquid substrates.


 2.8.2   Industrial Activity:  Oxidations in Supercritical Fluids

    In a 1997 patent [111], Pitchai and colleagues (ARCO Chemical Co.. now Lyondell
 Chemical Co., a leading producer of propylene  oxide via the coproduct process)
describe a process where propylene is converted to propylene oxide directly using
 a silver catalyst, where addition of CO, enhances the efficiency of conversion. It is
 known that Lyondell is actively working on a coproduct free route to propylene
oxide [ 112]. although it is not yet  clear whether supercritical fluids are being em-
ployed in the current  work.


2.9   Summary:  Gaseous  Reactants in CO2

   Clearly, carbon dioxide exhibits some significant advantages as a solvent in
systems where one or more of the reactants is a gas under typical operating condi-
tions. In such cases, operation in a liquid solvent almost always sets up a situation

-------
52
     where the reaction is controlled by diffusion of the gas through the gas-liquid
     interface. Consequently, use of CO, as the solvent can produce (at suitable pres-
     sure and temperature conditions) a single-phase substrate-gaseous reactant-CO,
     mixture, and eliminate transport resistance owing to the presence of the gas-liquid
     interface. This, in turn, can render the reaction more efficient and potentially lead to
     lower energy usage, smaller processes, and less waste. In addition, it is clear that
     use of CO, as the solvent exhibits special advantages in certain reactions where
     oxygen is employed as a reactant—because CO, will not oxidize, no solvent-based
     oxidation waste products will be produced in CO,-based systems. Further, when
     hydrogen and oxygen are used together in a process (as in Baiker's [103] and
     Beckman's [15] work), use of CO, as the solvent can greatly enhance the safety of
     the process. Despite the successes noted in the literature, there are some interest-
     ing avenues of research in the  general area of "use of gaseous reactants in CO,"
     that have not been, but should be pursued.

        First, a minority of the papers published on use of H,. O,. and/or CO in CO,-
     based reaction systems employ a two-phase mixture in which to conduct the reac-
     tion: researchers opt instead to raise the pressure to a point  where a single phase
     forms. Because  CO, usually swells organic liquids extensively,  conducting the
     reaction in a two-phase mixture could eliminate the transport resistance owing to
     gas diffusion into the liquid phase while permitting use of relatively low operating
     pressures. In many cases, if one simply knew the phase behavior  of the gas/CO,/
     substrate mixture, one could predict those conditions where high (enough) con-
     centrations of gaseous reactant  would exist in the lower, substrate-rich phase. Use
     of lower pressures facilitates equipment design and results in  less stringent utilities
     requirements, and is thus a "green" advantage. In addition, operation in a two-
     phase mixture would  allow use of air as an oxidant without a slow buildup of
     nitrogen in the mixture. Finally, as in the case for hydrogenations. use of a two-
     phase mixture would allow for heat transfer via liquid boiling and condensation.

        Another significant point to be made regarding heterogeneous catalysis in
     CO,-based systems is that elimination of the transport resistance owing to gas-
     liquid diffusion may not render the reaction kinetically controlled, as one also must
     account for liquid-solid transport and pore diffusion within the catalyst. Typically,
     the effect of pore diffusion on the control of the reaction is mitigated by employing
     smaller catalyst particles, but this solution is not always practical at  larger scales. In
     addition, it often is easier to operate using a fixed bed of catalyst rather than a slurry
     of particles. Because CO, is a low viscosity fluid, it may be possible in some situa-
     tions to move from a slurry of particles to a fixed bed without sacrificing rate.

        Finally, a number of researchers have shown that one can design catalysts that
     are soluble in CO,, and can be operated without any transport constraints despite
     employing gaseous reactants and catalysts. However, recovery of a homogeneous
     (and typically valuable) catalyst from CO, is not a trivial problem, and its solution is
     required to allow homogeneous reactions in CO, to be both green and economically

-------
viable. Naturally, one solution is to design catalysts that are relatively nontoxic and
whose activity is high enough such that recovery is not necessary (as is the case
currently with ethylene polymerization catalysts). In the case of all catalysts (ho-
mogeneous  and heterogeneous), the  effect of the presence of CO, on catalyst
deactivation (perhaps through the formation of CO during hydrogenalion) is an
area that merits further scrutin\.

-------
                            Section 3

   Polymerization and Polymer Processing
3.1   Introduction

    Polymerization and polymer processing in/with CO, exhibit some interesting yet
seemingly contradictory trends. Some of the most successful commercial processes-
that employ CO, as a solvent involve polymeric substrates, yet the vast majority of
polymers produced worldwide are produced in the complete absence of solvent.
Indeed, polyolefins (polyethylene), vinyl polymers (styrenics, acrylontrile, butadi-
ene), polyamides (nylons), and polyesters are generated principally in bulk polymer-
ization processes [ 113]. Further, for the most part, commercial polymers are poorly
soluble (many, in effect, are insoluble) in CO,. However, owing to the asymmetry of
polymer-CO, phase envelopes, even polymers that are poorly soluble in CO, will
swell extensively under moderate CO: pressure, allowing for a number of applica-
tions using CO, as reversible diluent/plasticizer. CO, is used extensively in the foam-
ing of polymers (both styrenics and polyurethanes), has been used as the solvent in
coating processes (Union Carbide's UniCarb process), and currently is being ex-
plored at the developmental level in fluoropolymer synthesis (DuPont) and powder
coating processing (Ferro Industries).


3.2   Polymerizations:  General  Background

   Polymerizations typically are classified by the mode of polymerization (ring-
opening, free-radical, etc.), by the type of monomer used (styrenics, acrylates) or
by the type of linkage formed during polymerization (polyamides, polyesters). In
addition, polymerizations can be conducted in the bulk state, in solution, or in one
of many so-called "heterogeneous modes"—namely precipitation, suspension, dis-
persion, or emulsion.

   Because CO, typically is proposed/employed as a benign solvent, the follow-
ing discussion of polymer formation and processing in CO, will focus on those
                                  55

-------
56
     applications where solvents ordinarily are used. However, where examples can he
     found in which use of CO, in a formerly solventless process can provide sustain-
     able and other benefits, such applications also will be discussed.


     3.3   CO2 as a Solvent for Polymer Systems

         Polymers present special problems regarding dissolution in any solvent—the
     very low entropy of mixing in polymer/solvent binaries (owing to the long chains
     of the polymer) requires  a very favorable enthalpic interaction between polymer
     segments and solvent to ensure dissolution of substantial polymer concentrations
     [114]. This problem is magnified in the case of CO,, given that CO,'s solvent power
     is admittedly weak.

        Although a significant portion of academic polymer-supercritical fluid phase
     behavior work has considered solutions where the polymer is the minor compo-
     nent, it is important to remember that the full-phase diagram offers several interest-
     ing regimes with regard to possible green applications. In Figure 6. we see a ge-
     neric phase diagram of a polymer and a supercritical fluid [115]. showing the vari-
     ous phase separation envelopes and the behavior both above and below the solvent
     critical temperature. As can be seen in Figure 6. the liquid-liquid phase envelope is
     asymmetric (owing to the  large disparity in size between polymer and solvent) with
     the liquid-liquid critical point shifted towards the 100 percent solvent axis. This is
     important—it means that solubilization of low concentrations  of polymer in sol-
     vent will require the highest pressures. Swelling of the  polymer by the solvent
     (moving to the right along the x-axis in Figure 6) requires significantly lower pres-
     sures. Thus, in certain polymer-supercritical fluid mixtures, one can observe very
     high degrees of swelling (>25 percent in polyacrylate-CO, mixtures, for example)
     at pressures of 100 bar and below [116]. The relatively low pressures required to
     elicit high degrees of swelling may be one reason why applications where CO, is
     the minor component have been successfully commercialized, while those employ-
     ing dilute polymer solutions have not.

        High-pressure phase behavior studies of polymers and supercritical fluids have
     been conducted since the  late 1940s; the early work was performed to support the
     high-pressure polyethylene process.  Ehrlich's group performed some of the best
     early work on the phase behavior of polyolefins in supercritical alkanes and alk-
     enes [117]; these studies have been followed by  numerous others on
     polyethylene:alkane orpolyethylene:alkene mixtures [118].

        In the late 1960s. Giddings  suggested a simple correlation  between solubility
     parameter and critical pressure that indicated that CO,'s solvent power should be
     similar to that of pyridine  [3J. However, the strong quadrupole moment of CO, af-
     fects carbon dioxide's pVT properties (including the critical pressure) without influ-
     encing its solvent strength. Consequently, early calculations of the solubility param-

-------
                                                                              57
FIGURE 6  Qualitative P-x diagram of a poiymer-CO, binary mixture, both
             above and below the critical temperature of the solvent [81].
             Figure includes liquid-liquid (LL), vapor-liquid (VL), and three
             phase vapor-liquid-liquid (VLL) types of phase envelopes.

eter were invariably inflated. This actually was confirmed by the very study that
proposed that CO,'s solubility parameter should approach that of pyridine; polymers
that would dissolve in pyridine were not soluble in carbon dioxide. Subsequent cal-
culations performed during the early 1980s [see, for example, 119] using CO,'s equa-
tion of state, strongly suggested that CO,'s solubility parameter should approach that
of normal alkanes. However, experimental work by Heller's group on the phase be-
havior of polymers performed during that time [ 120] clearly demonstrated that CO, 's
solvent power is inferior to that of n-alkanes—very few polymers tested by Heller
showed any significant solubility in carbon dioxide at moderate (less than 200 bar)
pressures. Experimental work by Johnston's group [121] suggested that solubility
parameter was not the best means by which  to characterize the solvent power of
compressible fluids such as carbon dioxide. Johnston suggested instead that polariz-
ability/volume is a better measure of solvent power; by this standard, CO, is judged
to be a feeble solvent, in line with experimental evidence.

   During this same time period, a number of researchers found that silicones
[122] and fluorinated materials [1,69,123] exhibited miscibility with CO, at pres-
sures well below those of alkanes of comparable chain length. Indeed, a calcula-
tion of the solubility parameter of CO, using the heat of vaporization and molar
volume (of the liquid) would suggest values similar to those of fluoroalkanes or
silicones [124]. In 1992, DeSimone and colleagues published the first reports that

-------
58
     describe a truly "CO,-philic" polymer, a fluorinated poly aery late [73]. Further work
     [125] showed that block copolymers of fluorinated acrylates and "CO,-phobic"
     polymers were both soluble and able to form micelles in carbon dioxide.

         It is interesting that the role of fluorine in the design of CO,-philic materials has not
     been completely established. For example, although the poly(perfluoroacrylates) are
     the most CO,-philic polymers known, it also is true that more poorly soluble
     fluoropolymers have been identified than highly soluble variants [118,126]. Samulski
     and colleagues [ 127] have found experimentally that fluorine interacts specifically with
     the electron-poor carbon  on CO,, which would explain  why addition of one or two
     fluorine atoms to aryl phosphine ligands  or chelating agents tends  to enhance CO,-
     solubility significantly. Calculations using various levels of theory tend to predict no
     specific interactions with fluorine [ 128], suggesting that fluorine's role in the design of
     CO,-philic materials is simply to lower the cohesive energy density. McHugh recently
     has suggested that fluorination can significantly enhance the "CO,-philicity" of poly-
     mers if the fluorination creates a dipole in the material, providing a locus for quadrupole-
     dipole interactions with CO, [ 126a]. This appears to be an area where more fundamental
     research would help to create a clearer picture of the underlying phenomena.

        As interest in applications for CO,-philic polymers exploded in the 1990s [ 129],
     a small group of researchers continued to probe the fundamentals of CO, behavior
     with special regards to polymer solubility. Johnston's and Eckert's groups, using IR
     spectroscopy and computer calculations, proposed that Lewis acid-base interac-
     tions between CO, and carbonyl groups could explain the  high swelling of
     polyacrylates by  carbon  dioxide [130, 131]. Calculations using various levels of
     theory tend to support the experimental evidence, at least where carbonyl groups
     are concerned [132]. Further, the specific interactions between Lewis base groups
     and CO, exhibit a much more significant effect on poIymer-CO, phase behavior than
     small molecule-CO, phase behavior. McHugh's group published  several seminal
     papers [ 118,133] on the phase behavior of CO, and various homo-  and copolymers
     in the mid-1990s. Conventional wisdom of the time would suggest that, because
     CO, is a low dielectric, low cohesive energy density solvent, it should only solvate
     polymers of similar characteristics. However, for the case of ethylene-acrylate co-
     polymers, McHugh found that increasing the acrylate content lowered miscibility
     pressures, despite the fact that the acrylate is the polar comonomer. McHugh pos-
     tulated quadrupole-dipole interactions as the cause; clearly, Lewis acid-base inter-
     actions could have played a role as well. For the case of n-alkyl acrylates, McHugh
     found that  increasing the side  chain length of the polymer initially would lower
     miscibility pressures, ostensibly due to the increased polymer free volume (and
     hence entropy of mixing). However, because enthalpic interactions between CO,
     and methylene groups are not favorable, increasing the side chain length beyond a
     certain point led to decreased miscibility. Johnston recently reported that polymers
     that exhibit low interfacial tensions (and hence low cohesive energy densities) also
     tended to exhibit low miscibility pressures in carbon dioxide [68].

-------
                                                                               59
    Clearly, the phase behavior of polymers in CO, is tied to carbon dioxide's low
 cohesive energy density, but its Lewis acid character also will play a significant role
 if the polymer contains Lewis  base groups. For example, Beckman found that
 polybutadiene, a very low cohesive energy density polymer, is more "CO,-philic"
 than other vinyl polymers of higher cohesive energy density [134]. However, both
 polypropylene oxide and polyvinyl acetate exhibit lower miscibility pressures than
 polybutadiene. likely owing to the presence of Lewis base groups in each of the latter
 polymers despite exhibiting higher cohesive energy densities than polybutadiene.

    Topology also plays a role in determining phase behavior. Beckman and Lepilleur
 [ 135] found that increases to polymer chain branching generally lowers miscibility
 pressure in  CO,. This result confirms earlier results on branched polyolefins in
 alkanes [ 136]. Finally, McHugh found that topology can play an extraordinary role
 in determining the phase behavior of polymers in CO,. The miscibility pressures of
 polyvinyl acetate, for example, lie at pressures hundreds to thousands of bar lower
 than those for polymethyl aery late (an isomer of PVAc) [ 133]. The underlying mecha-
 nism for this behavior is entirely unknown.

    In the late 1990s, Beckman's group [ 137] proposed a hypothesis for design of CO,- •
 philic polymers that incorporated the earlier conclusions reached by both McHugh and
 Johnston. Beckman and colleagues proposed that CO,-philic polymers should incorpo-
 rate monomers (or functional groups) that contain several features: high flexibility (and
 thus low Tg), low cohesive energy density, and Lewis base groups to provide loci for
 specific interactions between the polymer and CO,. They demonstrated the effective-
 ness of the hypothesis by designing highly CO,-soluble ether-carbonate copolymers.
 Modified polydimethyl siloxane (PDMS) also was examined [ 138]—experimental work
 by Kiran [139] had shown that PDMS exhibits upper critical solution temperature (UCST)
 type phase behavior at room temperature, suggesting that  the enthalpic interaction
 between PDMS and CO, is nonoptimal. Fink, et al.. then showed that addition of Lewis
 base groups (in side chains) to PDMS lowered miscibility pressures in CO, by hundreds
 of bar. Finally, Wallen [ 140] has proposed that CO, can exhibit specific interactions other
 than simple Lewis acid-base type. Wallen has found, via both simulation work and
 experiment, that an aldehyde will exhibit interactions between the carbonyl oxygen and
 the carbon atom in CO, as well as a weak  hydrogen bonding interaction between the
 aldehyde H and the oxygen in CO,.

    In summary, we have made great strides in our understanding of CO,-polymer
phase behavior since the days when "CO, is like hexane" was conventional wis-
dom. However, as shown by recent work from McHugh, Beckman, and Johnston, a
fundamental understanding of CO,-polymer thermodynamic behavior is still lack-
 ing. Poly(fluoroacrylates) are the most CO,-philic polymers known, but their high
cost renders their application problematic. If one could, from first principles, design
a nonfluorinated, truly CO,-philic polymer, this would greatly enhance the potential
 for industrial application of CO,, both in polymer science and general chemical
processing.

-------
60

     3.4   Chain Polymerization and CO2

         In chain polymerizations, an initiating species is formed that contacts a mono-
     mer, creating the beginning of an active chain. This chain then grows rapidly to
     form the polymer molecule. Finally, a chain-terminating event may take place (or
     monomer may be depleted), ending growth of the chain in question. The various
     chain polymerization types then are further subdivided based on the type of initiat-
     ing species, and also the relative rates of initiation and growth [141].


     3.4.1   Free Radical Solution Polymerization

         In free radical chain polymerization, an initiator (through thermal, chemical, or
     photochemical stimulation) forms an active radical that contacts a vinyl monomer.
     forming the growing chain. Termination takes place either through chain coupling
     or disproportionation. Molecular weight distributions can be broad (> 2.0). and
     average molecular weight rises rapidly with conversion, leveling off as long chains
     are continuously formed. Low-density polyethylene, polyacrylates, polystyrene.
     poly vinyl chloride, and other materials are formed using free radical initiation. Much
     of the total commercial volume of such polymers is synthesized in the absence of
     solvent in continuous processes containing only monomer, polymer, and initiator
     at temperatures sufficient to create a pumpable polymer melt.

         As described above, the solubility of most polymers in carbon dioxide is relatively
     poor, and it is not surprising that early work on polymerization in CO, was relegated to
     precipitation polymerizations [ 142]. Although it could be claimed that the plasticizing
     effect of CO2 on the precipitated polymer might enhance transport of monomer to the
     growing chain end, no significant advantages (versus the added complication of work-
     ing at elevated pressure), green or otherwise, were realized from such processes, possi-
     bly because the presence of the monomer itself tended toplasticize the polymer. Conse-
     quently, one would only expect to observe a significant effect of added CO, during the
     later stages of polymerization, when the presence of CO, might inhibit the well-known
     Trommsdorf, or autoacceleration effect (the latter occurs when the increased viscosity
     of a polymer melt inhibits chain termination, leading to rapid increases in rate). Because
     CO, is a diluent, its presence also would lower the rate in general, which is a disadvan-
     tage [143]. Finally, vinyl polymerizations are exothermic, and great care would need to be
     taken to prevent uncontrolled pressure increases. In summary, the disadvantages in-
     herent to operating a vinyl polymerization in CO, have greatly outweighed any advan-
     tages to date.  In general, it is very hard to justify (from a "green" perspective) adding
     solvent to a solventless process.

        One exception to this rule is in the surfactant-free precipitation polymerization
     of fluoromonomers [144], recently scaled up  by DuPont  to  a semi-works size in
     North Carolina. Typically, fluoropolymers are generated via suspension polymer-

-------
                                                                               61
 ization in water; the use of carbon dioxide as the solvent provides for a chain-
 transfer free solvent and eliminates the need for the surfactant (as noted previously.
 the EPA recently has filed a SNUR regarding fluorinated surfactants of the
 fluorosulfonate variety, possibly restricting their use in the future |42]). Interest-
 ingly, most fluoronionomer polymerizations are precipitation polymerizations (as
 shown by McHugh [ 126]. many fluoropolymers are insoluble in CO,).  However,
 addition of CO,  stabilizes tetrafluoroethylene, eliminates the need for fluorinated
 solvents and surfactants, and eliminates chain transfer to solvent. Indeed, a recent
 conversation with a DuPont customer [145] revealed that the fluorinated copoly-
 mers produced in CO, exhibit superior performance during extrusion, owing to
 fewer gels and a tighter composition distribution. In fluoropolymer polymeriza-
 tion. CO, provides green advantages, safety advantages, and product advantages.

    Another possible application for precipitation polymerization in carbon dioxide
 involves acrylic acid [146]. Currently, poly(acrylic acid) is generated in an emulsion
 or suspension polymerization  in a hydrocarbon continuous phase; removal of the
 alkane from the product is both energy intensive and waste forming. Use of CO, as
 the continuous phase allows the generation of dry. free-flowing, granular material.

    Carbon dioxide also has been proposed as a diluent (reversible plasticizer) for
 reactions on preformed polymers, reactions that often take place within extruders
 during polymer  processing. In theory, the plasticizing effect of CO, will reduce
 transport limitations of the reactants (in the otherwise highly viscous melt), leading
 to enhanced rate and thus more complete reaction in the same residence time. How-
 ever, O'Neill and Beckman [ 143] found that in the case of the poly vinyl acetate-to-
 butyrate transition (a highly successful industrial process), the presence of the low
 molecular weight reactants was sufficient  to plasticize the melt. CO, acted merely
 as a diluent, lowering the rate  by reducing the concentration of the active species.


 3.4.2   Heterogeneous  Free Radical Polymerizations

    Heterogeneous polymerizations are those where the polymer is not soluble in
 the continuous phase, or solvent [141]. These polymerizations can be subdivided
further based on the thermodynamic affinity of the monomer for the solvent and the
 nature of the polymer stabilization:

•   Emulsion

•   Dispersion

•   Suspension.

    Although simple precipitation can be  considered as a form of heterogeneous
polymerization, it has been considered separately in the previous section.

-------
62

     3.4.2.1   Emulsion Polymerization in CO2

         In emulsion polymerization, neither the monomer nor the polymer is soluble
     (to any appreciable extent, there is always some measurable monomer solubility)
     in the continuous phase, and sufficient surfactant is present to form micelles (the
     locus of the polymerization) and to stabilize the large droplets of monomer that
     also are present (the latter form monomer reservoirs). The kinetics of the emul-
     sion polymerization are such that (unlike in bulk or solution free radical poly-
     merization) both high rate and high molecular weight are possible. Carbon diox-
     ide, although not a powerful solvent, is miscible with a large variety of volatile,
     low molecular weight vinyl monomers [147]. As such, identifying a suitable can-
     didate for emulsion polymerization is problematic, as one must find a monomer
     that exhibits a sizeable phase envelope under the conditions of interest, yet under
     conditions where the surfactant to be employed is miscible (in CO,, the converse
     is much simpler to identify—a mixture where the monomer is miscible and the
     surfactant is not). This has proved to be difficult, and to date only acrylamide,
     acrylic acid, and N-vinyl formamide have been investigated in any detail [148].
     The case for acrylamide is complicated further by the  fact that it is  a solid at
     temperatures below 353 K and has been employed as an aqueous solution—the
     presence of the water renders subsequent polymer particle size analysis difficult.
     Emulsion polymerization of water soluble monomers in  CO, is a viable target in
     the context of green chemistry, in that the commercial route employs a kerosene-
     based continuous  phase and also requires significant energy input to separate
     product from emulsion following polymerization.

        The key issue  in emulsion polymerization is the design of the surfactant—it
     must be soluble in CO, at moderate pressures, effective,  and relatively low cost.
     Early work employed fluorinated surfactants (nonionic and anionic), as these were
     known to be CO,-philic [148]. Results showed that one could indeed generate high
     polymer at high rates, but the surfactants employed were more valuable (even at 1
     percent loading and below) than the polymers being generated, and recycle is dif-
     ficult to achieve economically. Although silicone-functional surfactants also have
     been evaluated [ 149] in emulsion polymerization, their performance is not as good
     as their fluorinated cousins, and their cost can be quite high  (for siloxane-based
     materials generated from the cyclic tetramer (D4), cost is about 5 to 10 times as
     high as traditional hydrocarbon surfactants. For monofunctional materials created
     from the D3 cyclic  trimer, the cost approaches that of fluorinated materials). The
     practicality of the process would be greatly enhanced by discovery of an effective
     yet low-cost surfactant. In work to date, AIBN (azo bis(isobutyrnitrile)) usually
     was employed as the initiator, and process temperatures were set at 330-340 K to
     achieve reasonable polymerization rates (AIBN half-life at 343 K is - 4 hours). As
     such, process pressures were relatively high (> 200 bar). Clearly, use of an initiat-
     ing system that operates at lower temperatures (photochemical or redox [141])
     would lower the required process pressure and also render emulsion polymeriza-

-------
                                                                              63
 tion in CO, more practical (see. for example. [150]). It should be noted that such an
 initiator system would be more expensive than that currently employed, an added
 cost that must be factored into the total.


 3.4.2.2   Dispersion Polymerization in CO2

    Dispersion polymerization [151]. where  the monomer is soluble in the con-
 tinuous phase (here CO,) while the polymer is not. has seen extensive research
 activity over the past decade. Because most, if not all. vinyl monomers are miscible
 with CO, at relatively modest pressures (complete miscihility below 100 bar at 313
 K in many cases), although high polymers are  notoriously insoluble, dispersion
 polymerization seems well suited to adaptation  to carbon dioxide. If a dispersion
 polymerization was conducted in a conventional liquid, a low  molecular weight
 alcohol or alkane would be the preferred continuous phase, and thus CO, could
 replace a significant volume of organic solvent. Separation of the product polymer
 from the continuous phase in a CO, system would not require dry ing/devolatilization,
 a potentially significant energy savings. Because many vinyl monomers lend them-
 selves to dispersion polymerization in CO,, the key requirement to successful dem-
 onstration was finding a suitable stabilizer. Finally, because a successful dispersion
 polymerization produces a stable latex that then can form the basis for a coating
 formulation, it was hoped that the analogous process in CO, would produce a coat-
 ing formulation that could be sprayed without VOC release.

    Stabilizers for dispersion polymerization in conventional systems require a
 soluble component and an anchoring component—DeSimone's group prepared the
 first successful stabilization system from homo- and copolymers of fluoro-acrylate
 monomers [152]. Small amounts of these copolymers permitted the rapid polymer-
 ization of methyl methacrylate (MMA) in CO, in the form of monodisperse par-
 ticles approximately 1  micron in size. Johnston's group later showed that stabiliza-
 tion of the particles was due in large part to effective solvation of the CO,-philic,
 fluorinated blocks of the copolymer [153]. If conditions (temperature  and pres-
 sure) were such that the fluorinated chains would collapse, flocculation of the par-
 ticles would take place. Beckman and Lepilleur [154] also examined the disper-
 sion of MMA in CO,;  comb-type copolymers (acrylate backbone and fluoroether
 side chains) were employed. Once the backbone was above a certain chain length,
 monodisperse, micron-size particles could be formed rapidly. Finally. Howdle and
colleagues [155] found that one could create a very simple but effective stabilizer
 for MMA polymerization, a fluoroether  carboxylic acid. Hydrogen bonding be-
 tween the acid and MMA's carbonyl provided anchoring sufficient to stabilize the
dispersion and form small PMMA particles.

    As in the case tor  emulsion polymerization, practical dispersion polymeriza-
tion in CO, ultimately will require a stabilizer that is both sustainable and inexpen-
sive, and the  fluorinated materials investigated  heavily during  the 1990s are not

-------
64
     likely to be applied industrially. A reactive silicone (polydimethyl siloxane. aery late
     terminated) has been applied as a stabilizer in MMA polymerization [ 156). but its
     performance was far less satisfying than the various fluorinated stabilizers that
     have been evaluated. As in the case of emulsion polymerization, use of an initiating
     system that operates at low temperature (versus the typical thermally triggered azo-
     and peroxide compounds) would lower process temperature (and hence pressure)
     substantially. Finally, although micron-size particles of MMA (and other mono-
     mers) were readily formed. latex stability was relatively poor, with material set-
     tling out  in a matter of hours (versus the desired days and weeks). This is not
     entirely surprising, as the low viscosity of CO, (1/10 that of water) produces a
     relatively high terminal  settling velocity. If the cost of the stabilizer could be low-
     ered and the stability of the latex  improved, a CO,-based dispersion could form the
     basis of a low-VOC coating system.

        A potentially sustainable CO,-based (and hence solvent free) coating formula-
     tion might be developed even if the rapid settling of the latex cannot be  corrected.
     If polymer particles, produced either in water or in CO, then recovered  and dried.
     could subsequently be redispersed in CO,, then the dry particles could be shipped
     from manufacturer to remote customer and still employ a non-VOC (CO,-based)
     spray-coating system. Use of such a system would save the large amount of energy
     needed to transport essentially solvent (CO, or water) for long distances. Johnston
     and colleagues have investigated the mechanics of particle redispersal and also the
     design of surfactants that would allow such polymerization and redispersal [157].
     Their initial results are promising. Although not entirely similar, the commercial
     UniCarb process [37] was an early attempt to address the stability versus sustain-
     ability balance in spray coatings.  The conventional  coatings process  employed
     polymer beads dispersed in  a mixture of a good solvent and a poor, yet volatile
     solvent. The UniCarb process replaced the poor solvent with CO, (also a poor, yet
     volatile solvent) while retaining  the good solvent  to maintain the stability of the
     dispersion. Replacement of the poor solvent with CO, reduced VOC emissions by
     60 percent.

        One area where CO, would exhibit advantages over both water and organic
     solvents would be dispersion polymerization of hydrolytically sensitive monomers.
     In such a case, water would be green but technically infeasible, but apolar organics
     would be technically feasible yet not sustainable. DeSimone and Shiho have illus-
     trated this using a glycidyl methacrylate monomer [ 158]. Again, if an effective, yet
     inexpensive surfactant could be identified, use of CO, in such an application would
     be both green and technically efficient.


     3.4.2.3   Suspension Polymerization in CO2

        In suspension polymerization, neither the monomer nor the polymer are soluble
     in the continuous phase, but the stabilizer structure and concentration are such that

-------
                                                                             65
only droplets are formed (no micelles), and the kinetics of the polymerization re-
semble that of hulk polymerization. Suspension polymerization typically is ap-
plied to hydrophohic vinyl monomers in water, a process that is itself relatively
green (although water remediation and energy use for drying represent targets for
improvement). CO, has been used in the suspension polymerization of acrylic acid
in CO, in the hopes of replacing the conventional hydrocarbon continuous phase.
Polyacrylic acid is a very low-cost commodity material, and such a process must
produce dry. free-flowing powder at relatively low pressure  and with an inexpen-
sive stabilizer (159).
3.4.2.4  CO2 as Nonsolvent in Heterogeneous
          Polymerizations

    Cooper and colleagues [160] have explored a novel application of CO, in
heterogeneous polymerization. CO, is used as the porogen in the suspension poly-
merization of styrene/divinyl benzene, where the resulting porous beads form the
basis for ion exchange resins. Typically, a hydrocarbon porogen is employed, must
be separated from the product,  and disposed after use. A good porogen must be
miscible with the monomer (as  is the case with CO, and styrene) yet immiscible
with the polymer (as in CO,/polystyrene). Generally, one alters the pore size and
total surface area of the beads through alterations to porogen composition: Coo-
per showed that the same tunability  could  be achieved through pressure alter-
ations to CO,.


3.4.3   Other Chain Polymerizations in CO2

    Carbon dioxide has been employed as a solvent for cationic and metal-catalyzed
ring-opening polymerization of various monomers in CO,. Biddulph and Plesch first
examined cationic chain polymerization of isobutylene in CO, in 1960 [161]; Kennedy
later examined this reaction [ 162]. This work demonstrated that cationic polymeriza-
tion is indeed viable, but the precipitation of the polymer lessens any advantages one
might have derived from use of a green solvent. DeSimone later applied knowledge
of CO,-philic compounds to greater advantage by examining the homogeneous cat-
ionic polymerization of fluorinated monomers (both vinyl and functional oxetane) in
CO, [163]. As the  DeSimone group demonstrated earlier, polymerization of fluori-
nated monomers in CO, is a very effective technique for polymer production without
the use of hydro fluorocarbon solvents.

    Metathesis polymerization also is viable in CO,, yet the hydrocarbon mono-
mers employed produce polymers that rapidly precipitate on attaining even modest
chain length [164]. The same is true for oxidative polymerizations of either pyrrole
or dimethyl phenol. It has been  shown that one can prevent the seemingly inevi-
table precipitation  through use of fluorinated stabilizers (and hence formation of a

-------
66
     dispersion), but the high cost of the stabilizers has inhibited further consideration
     of such routes.

         Not surprisingly, anionic polymerization in CO, produces, at best, carboxy-
     terminated oligomers. as the terminal anion reacts quickly with CO, to produce the
     less reactive carboxylate. Carbon dioxide also is an efficient chain terminator in
     Ziegler-Natta and metallocene-type catalyst systems—as such. CO, currently can-
     not be used as a solvent in controlled olefin polymerizations, the largest volume of
     polymerizations. Because these polymerizations tend to be low pressure gas-phase
     reactions of ethylene and propylene. it is not clear what role carbon dioxide could
     play even if the catalysts could tolerate its presence.


     3.4.4   Industrial Activity: Chain Polymerizations in CO2

        DuPont has filed a number of patents [165] describing the use  of CO, as a
     solvent for chain polymerization of fluorinated monomers. This technology, plus
     patents filed by coworkers at the University of North Carolina [144], formed the
     basis for the construction of a semi-works facility in North Carolina with an annual
     capacity exceeding 1.000 tons of fluoropolymer (there are plans to expand this
     capacity significantly by 2006). 3M and Xerox also have obtained recent patents in
     this area [ 166], although their supercritical CO, research efforts were discontinued
     several years ago.

        The European Union (EU) funded (1.5 million Euros. 12/97-12/00) a multi-year
     study (Superpol project)  linking four universities with polymer manufacturers—
     Solvay, Goldschmidt. and DSM—to explore the use of supercritical fluids in poly-
     mer production. Although the consortium includes both prestigious universities and
     well-known companies, the results to date [167] have not significantly added to the
     information described above. Indeed, nothing shown in the most recent presentation
     (November 2001) by the  Superpol group to the EU would be considered new by
     experienced researchers in the field, and strangely, no literature or patent references
     were included with this report. There is discussion of heterogeneous polymerization
     using fluorinated or silicone stabilizers, with results shown for the dispersion poly-
     merization of methyl methacrylate. There is a discussion of additives  for use with
     polyolefins, but the relationship of this work to CO,-based processing is not clear.
     The synthesis of fluoropolymers in CO, is described, noting the advantages of using
     CO,  as a solvent versus water or hydrofluoroalkanes. Mention is made that carbon
     dioxide could be used effectively in cleaning of a variety of substrates. Needless to
     say, much of this information was known (through publications and patents) prior to
     1997; thus, it is not clear what future impact the Superpol project will have on the use
    of carbon dioxide in polymer synthesis and processing in  Europe. Solvay recently
     has acquired the fluoropolymers business of Ausimont. and may invest in CO,-based
    fluoropolymer polymerization technology in the future.

-------
                                                                             67

3.5   Condensation Polymerizations


3.5.1    Polyester, Polyamides, Polycarbonates

    Condensation polymerization [141] occurs through the step-wise addition of
difunctional monomers to each other, usually in a reaction that produces a small
molecule byproduct (water or an alcohol, for example). Polyesterification (reaction of
diol with diester or diacid) and polyamidation (reaction of diamine with diacid or
diester) are two classic examples of great industrial importance. Because of the nature
of these polymerizations, there are key differences with respect to chain polymeriza-
tions. Condensation polymerizations usually are endothermic, and heat must be ap-
plied to achieve a high rate of reaction. Unlike chain polymerization, molecular weight
builds slowly in condensation reactions. Indeed, the statistics of condensation poly-
merization show that the extent of reaction of the active end groups must reach at
least 95 percent to create polymer chains of reasonable length. Because each conden-
sation (chain building) reaction is governed by equilibrium, removal of the small
molecule byproduct is crucial in achieving high extent of reaction and high chain
length.

    Continuous industrial condensation polymerization processes all exhibit the
same general elements [113]. The two monomers are added to the system in the
correct proportions, and then heated and pumped into a U-shaped tubular  reactor
with the appropriate catalyst. Steam (or alcohol) is flashed from the reactor at its
exit, and the resulting oligomer is pumped to a "finishing stage." Vacuum or flowing
N, is applied to remove the small molecule, and slow mixing creates surface area to
enhance the reaction rate. The oligomers are transformed to polymers. Tempera-
tures in the process must be high enough to melt the polymer, and temperatures of
520-570 K are not uncommon.

    Given the nature of condensation polymerizations, CO., has been applied as a
diluent/plasticizer to enhance the removal of the small molecule, hence increasing
molecular weight [168]. By dissolving in the polymer melt, CO, should reduce the
viscosity and increase the rate of removal of the condensation byproduct. Clearly, for
the process to be most successful, the small molecule should partition preferentially
to the CO, phase. The green aspect of such a scheme is that use of CO, could allow
better removal of the condensation byproduct at lower temperature, saving energy.
The best example of this use of CO, is probably the work of Kiserow and DeSimone
[ 169] on the CO,-enhanced solid-state polymerization of polycarbonate. In bisphenol
A polycarbonate production, diphenyl carbonate is reacted with bisphenol A to pro-
duce the polymer plus phenol. Many end users of polycarbonate (as well as nylon
6,6) practice "solid-state polymerization," where the purchased polymer is charged to
a vacuum oven to increase molecular weight through additional reaction and byproduct
removal.  DeSimone  showed that CO, could be employed to remove phenol from

-------
68
     polycarbonate oligomers at temperatures well below the T of the polymer (420 K).
     raising molecular weight substantially [ 169]. Later work [ 170] by Shi. et al.. showed
     that limitations to the increase in molecular weight are due primarily to an imbalance in
     the concentration of the two types of endgroup on the polymer (hydroxyl and termi-
     nal carbonate)—this is a common problem in the solid-state polymerization of con-
     densation polymers.

         A general problem with using CO, to enhance condensation byproduct removal
     is the low solubility of some common byproducts in carbon dioxide. Water, the most
     common byproduct in polyamide generation, is poorly soluble in CO,. In the forma-
     tion of polyethylene terephthalate (the highest volume polyester), the polymer is
     formed via the self-condensation of the adduct of two moles of ethylene glycol and
     dimethyl terephthalate (see Scheme VI): the byproduct is ethylene glycol. also poorly
     soluble in CO,. Indeed, the use of CO, to plasticize polymer melts and remove conden-
     sation byproducts is sound, sustainable processing, but this technique will only be
     truly effective if the byproduct is designed to partition strongly to CO,.

     Scheme VI:
     HOOC
      2   HQ^x^
                                                                      OH
        Although energy reduction is an admirable part of green chemistry, the most
    significant targets for green chemistry in condensation polymers are probably not
    the polymerizations themselves, but rather the  synthesis of the monomers. For
    example, diphenyl carbonate (monomer for polycarbonate) is synthesized from phos-
    gene and phenol, and a sizeable effort has been made by industry  to optimize the
    catalytic production of diphenyl carbonate from phenol and CO [171]. Bisphenol A
    (also a precursor to polycarbonate) is under scrutiny  for possible deleterious ef-
    fects on humans. Terephthalic acid (precursor for polyesters) is generated via an
    oxidation of p-xylene that produces some problematic  waste streams [ 14]. DuPont
    has expended considerable effort in a joint venture with Genencor to create a bio-

-------
                                                                               69
chemical route to propane diol. another precursor to aromatic polyesters. Pilot-
scale biological production of propane diol has been achieved, and a full-scale
production is planned for the future [ 172]. Non-phosgene routes to di-isocyanates
(precursors to polyurethanes) using CO, as a raw material have been investigated
by both industry and academia [173]. Finally, the oxidation route  to adipic acid
(precursor to nylon 6,6), and the synthesis of caprolactam (precursor to nylon 6)
are frequent targets of scientists involved in green chemistry, given the significant
waste streams emitted by current processes [174]. Consequently, it would appear
that real  breakthroughs in green chemistry applied to condensation  polymers will
and should come in the  area of more sustainable monomer synthesis. In some of
these cases, CO, could play a significant role, but the primary research need ap-
pears to be more atom-efficient synthetic routes.


3.5.2   Polyurethanes

    Polyurethanes are indeed condensation polymers but represent a special case,
in that a small molecule is not produced during the primary polymerization reac-
tion (where  a hydroxyl  group and isocyanate react to form a urethane linkage).
Whereas polyurethanes  are applied as fibers, coatings, and thermoplastics, their
primary relevance to this report owes to their extensive use in foamed articles.

   Polyurethane flexible slabstock foam has been produced via the "one-shot"
process since the late 1950s [175]. A stream of polyol (a multifunctional hydroxy-
terminated oligomer, typically a polyether) is blended with water, catalysts, surfac-
tants, and "blowing agents," then injected into a high-intensity mixing chamber
with a multifunctional isocyanate. The resulting liquid blend is pumped evenly
onto a moving belt, where polymerization occurs as hydroxyl groups react with
isocyanates  to form urethane linkages. Further, water reacts with isocyanate to
form an amine group plus CO,, where the amine subsequently reacts with another
isocyanate to form a urea linkage. The heat of reaction boils the "blowing agent";
this plus the CO, released during the polymerization creates the foam, which  is
stabilized until cure by the added surfactant.

   For decades, the preferred blowing agent was either a chlorofluorocarbon or
methylene chloride; note that these blowing agents were simply emitted to the atmo-
sphere during foam formation. Following adaptation of the Montreal Protocols in
1986, foam producers  searched for alternatives. Compounds such as pentane and
hydrofluoropropane have been evaluated and applied, yet these do not fully amelio-
rate the emissions problem (and, of course, hydrocarbons are flammable). In the late
1980s and early 1990s, Crain Industries created a CO,-based process (CarDio [ 176])
where liquid CO, (3-5 percent by weight) is injected into the polyol stream at pres-
sures above the vapor pressure of CO,. The pressure then is gradually reduced, such
that the pressure in the high intensity mixer is only 10-20 bar. The pressure then is
reduced further via the use of a "gate-bar" assembly that expands the mixture to one

-------
70
     atmosphere and spreads it evenly onto the moving belt. The liquid mixture remains
     single phase through the mixing chamber because polyols absorb significant amounts
     of C(X even at low pressures. Plants operate the CarDio process in both Europe and
     the  United States. The Bayer Corporation also has commercialized a CO,-based.
     continuous polyurethane process [177]. In both the CarDio and Bayer processes,
     CO, directly replaces a large volume of organic solvent that would have been emitted
     to the atmosphere with little additional energy input (cooling the liquid CO,). Conse-
     quently, polyurethane foam production using CO, as the blowing agent is an excel-
     lent example of green chemistry using carbon dioxide. It is interesting to note that the
     first patent proposing the use of CO, as the blowing agent for polyurethane foam was
     filed in 1959 [ 178]—it was only after perfection of the gate bar assembly in 1991,
     that Grain was able to successfully scale up a CO,-based polyurethane foam line.
     Thus, the success of a green, CO,-based chemical process can depend as much on
     mechanical design as on chemical design.


     3.6   Carbon Dioxide as a Monomer

        Since 1969, it has been known that carbon dioxide can be copolymerized with
     oxiranes to form poly(ether-carbonates) [179]. Production of a polycarbonate us-
     ing CO, instead of phosgene (the usual route) is indeed a green process, in that not
     only is a harmful chemical replaced with a benign alternative, but also the produc-
     tion of substantial quantities of salt (the usual byproduct in polycarbonate produc-
     tion) is avoided. Poly(ether-carbonates) formed from oxiranes and CO, could be
     applied as degradable  surfactants (using ethylene oxide) or low energy alternatives
     to polyester polyols in polyurethane manufacture (using propylene oxide). They
     also have been found to be the most CO,-philic, nonfluorinated materials yet iden-
     tified [137], and they could enhance the  wider use of CO, as a  benign  solvent.
     There are, however, some key technical hurdles that have substantially prevented
     the commercialization of a CO,-based route to a polycarbonate to date:

     •   Most of the catalysts developed to date have  not demonstrated particularly
        high activity when used with either ethylene oxide or propylene oxide, the
        comonomers most likely needed to produce economically viable copolymers
        [ 180]. On the other hand, a number of catalyst systems have been shown to be
        highly effective in the copolymerization of CO, with cyclohexene oxide [181],
        although this copolymer has not attracted any significant industrial interest owing
        to monomer cost versus polymer properties.

    •  Those catalysts that have shown high activity in CO,/propylene oxide copoly-
       merizations have not  permitted significant incorporation of CO, into the co-
       polymer (typically less than 10 percent carbonate) [182].

    •  Catalysts developed to date tend to produce substantial amounts of low mo-
       lecular weight, cyclic carbonate when  used with either ethylene oxide or pro-

-------
                                                                              71
    pylene oxide. In many cases, more than 80 percent cyclic material is produced.
    The low molecular weight cyclic cannot be polymerized, and current catalysts
    could not be employed economically.

    Early work (1970s-1980s) focused on the assessment of zinc catalysts for the
copolymerization of oxiranes and CO, [ 180]. These catalysts typically employed a
reaction between a dialkyl zinc and a multihydroxyl-containing compound to cre-
ate the active catalyst. Polymerization times were relatively long and significant
amounts of cyclic carbonate were produced, yet alternating copolymer (100 per-
cent carbonate) could be generated. Molecular weight distributions in these poly-
merizations could be very broad, often above 5.0. Nevertheless, a zinc system even-
tually was used to synthesize an ethylene oxide-CO, alternating copolymer that
was applied commercially (PC Corp., Wilmington, DE) as a ceramic binder (this
copolymer degrades cleanly to gaseous byproducts at temperatures above 470 K).

    Recent work in this area has focused on the development of "single-site" style
catalysts to allow better control over molecular weight [181]. However, although
these new catalysts have proved to be very effective in  the copolymerization  of
cyclohexene oxide and CO,, none have been able to solve the problems observed
during copolymerizations of CO, and either ethylene oxide or propylene oxide.  In
general, in copolymerizations of CO, and propylene  oxide, catalysts derived from
aluminum exhibit high activity and produce predominantly copolymer with a nar-
row molecular weight distribution, yet allow little CO, incorporation into the co-
polymer  [182]. Zinc catalysts allow for high levels of CO, in the copolymer, yet
produce predominantly low molecular weight alkylene carbonate.

    Indeed, the generation of copolymers of CO, and either propylene or ethylene
oxide would represent green chemistry, as these materials would have ready mar-
kets and alternative routes to their production (via phosgene) are highly problem-
atic from a sustainable viewpoint. Until the technical hurdles to efficient copoly-
merization (see above) can be overcome, a CO,-based route to aliphatic polycar-
bonates, and indeed, aliphatic polycarbonates in general, will not enjoy widespread
use. Whereas a variety of other polymers also have been generated from CO, [183],
either the properties of these new materials (vis-a-vis their cost) have not been
promising, or the efficiency of the polymerization has been low, and hence they are
technical curiosities rather than potential avenues  for green chemistry. Indeed,  to
achieve the highest impact (with respect  to green  chemistry), research should be
directed at creating catalysts that target the efficient copolymerization of propy-
lene oxide (or perhaps ethylene oxide) and CO,.

    Generation of an aliphatic polyester from CO, and an  olefin would be a superb
example of green chemistry with a ready market for the material. Aliphatic polyes-
ters, although "green" materials in their own right (they degrade cleanly to nontoxic
fragments in the environment), require multiple steps to prepare the monomers and
then the polymer, and also significant energy input along the way. A chain polymer-

-------
72
     ization route to aliphatic polyesters starting from olefins and CO, would be both
     greener and less expensive than the current method. With the exception of one or two
     references in the late 1970s [ 184] and a 1949 patent [ 185]. there has been no published
     scientific activity on this problem, despite the technical and commercial importance.
     Calculations performed at the University of Pittsburgh suggest that formation of a
     lactone (the immediate precursor to a polyester) from CO, and several olefins should
     be thermoneutral. and the reaction is at least theoretically tractable.


     3.7   Industrial Activity: Condensation Polymers and CO2 as a
           Monomer

        As mentioned earlier, both Grain and Bayer have commercialized the use of CO,
     as the blowing agent in continuous polyurethane foam production—20+ plants
     currently operate using this technology. Further, PC Corp. (DE, USA) sells aliphatic
     polycarbonate (used as a ceramic binder) generated via the copolymerization of
     CO, and ethylene oxide.

        Xerox has patented [ 186] a process where bisphenol A polycarbonate is gener-
     ated from bisphenol A and diphenyl carbonate using CO, to extract the residual
     phenol. Further, Akzo-Nobel patented [ 187] the formation of a degradable surfac-
     tant via the copolymerization of ethylene oxide and CO,, where the polymerization
     is terminated by a fatty acid. However, it is known that Xerox has ceased its re-
     search efforts on polymerization in CO,, while Akzo-Nobel shut down its research
     efforts on CO,/alkylene oxide copolymerizations in early 1998, during elimination of
     its corporate research department.


     3.8    Postpolymerization Processing of Polymers Using CO2

       Polymers require far more postsynthesis processing than do small molecules,
    and it is not surprising that CO, plays a role in green postpolymerization processing
    of polymers. First, as mentioned previously, CO, will swell many polymers exten-
    sively, even those normally considered "CO,-phobic." As shown in the generic
    phase diagram (Figure 6), this is  because of the asymmetry of the liquid-liquid
    phase envelope, itself arising from the disparity in size (and hence vapor pressure)
    of the  solvent and  solute. Swelling a polymer with CO, will drop its viscosity
    significantly (depending on temperature, by orders of magnitude). This large drop
    in viscosity allows for a number of CO,-enhanced processes. For example. Berens
    and Huvard [ 188a] demonstrated that the swelling of a polymer by carbon dioxide
    enhances the rate of infusion of model compounds. Kazarian and Eckert [ 188b] later
    exploited this effect in a novel way; they have shown that one can greatly  enhance
    the kinetics of mixing of a CO,-incompatibIe dye with a polymer. In this work, the
    dye and polymer are thermodynamically compatible, but the rate of infusion of the
    polymer by the dye is glacially slow. CO, plasticizes the polymer (while not actually

-------
                                                                              73
dissolving very much, if any, of the dye), lowering the viscosity and allowing fast
blending. The dying of fabric and fibers using CO, has been extensively examined
in Europe and the United States [ 189, 190]; the dye and polymer are thermodynam-
ically compatible, and the dye is sparingly soluble in CO,. Consequently, the dye
partitions preferentially into the swollen polymer, where the CO, diluent enhances
the kinetics of the thermodynamically favorable process. It is interesting to note
that Johnston [191] outlined the fundamentals for such a process several years ago
using a silicone polymer. CO,, and  toluene as the model "infusant." The green
aspect to this work is a reduction in energy required for mixing, as well as elimina-
tion of the aqueous waste stream commonly  associated with dying operations.
Further, use of CO, in place of water reduces air emissions and the need for drying
of the fibers after dying [ 192]. Major challenges remaining in this process are in
many ways "mechanical." How does one design a treatment chamber that allows
fast charging, fast sample changeover, and rapid dying? Is there sufficient thermo-
dynamic and transport information available to model and hence scale up the pro-
cess? Note that this situation is analogous to that described for continuous poly-
urethane production using carbon dioxide—the chemical challenges were  over-
come long before the mechanical issues were  settled. A further challenge would
include redesigning conventional dyes to allow for higher CO, solubility, which
would provide for more even coating.

    Applying the concept of carbon dioxide as a "reversible plasticizer," Shine and
Gelb [ 193] showed that one could mix a thermally labile bioacti ve compound (here a
vaccine)  into polycaprolactone. Howdle and colleagues  [194] recently expanded
this work into the tissue engineering field. CO, was used  to swell an  aliphatic
polyester, depressing its T, to well below room temperature. A temperature and
shear-sensitive enzyme then was mixed with the swollen polymer; upon depressur-
ization. the enzyme  was found to be dispersed throughout the now foamed poly-
mer, and to have retained its activity. Such a process allows the blending of tem-
perature sensitive compounds with polymers without the need for additional sol-
vent-based processing.

    Powder coating processing provides another potential application for CO, as a
sustainable and reversible plasticizer. Powder  coatings (blends of low molecular
weight functional polymer, crosslinking agent, pigments, and stabilizers) are con-
sidered green materials, as they can be applied directly to automobile and appliance
bodies without any solvent. However, the means for production of powder coat-
ings is itself wasteful and expensive. The raw materials are charged to an extruder
for high shear mixing; the resulting pellets then are ground and sieved to create the
proper size distribution. Waste from the grinding process cannot be re-extruded, as
the polymers are quite naturally thermally sensitive. Ferro Corporation first pat-
ented [195] a process where CO, is used to swell the polymer, depressing its T
(normally 310-320 K) to well below 270 K. The additives (pigments, etc.) then are
mixed with the swollen polymer. Finally, the material is rapidly depressurized through

-------
74
     a nozzle to form a granular mixture. Note that material processed in this way can
     actually be recycled if necessary, as temperatures employed are low (313 K). PPG
     Corporation [ 196] also supported work in this area using hydrofluorocarbon fluids;
     this work was targeted at small colored batches. Other patents also have appeared
     recently [ 197]. The remaining challenges include elimination of a significant degas-
     sing problem upon film  formation and the need to lower the operating pressure as
     much as possible to remain economical. Regarding the degassing problem, conven-
     tional powder coating formulations  use benzoin as the degassing agent (to help
     eliminate air during film formation). However, it is not currently known why benzoin
     is effective as a degassing aid in conventional formulations, and therefore, the
     design of analogues for use in material processed in CO, currently is not possible.
     Indeed, both Ferro Corporation and  PPG have ceased (at least  for now) their re-
     search and development efforts in this area, owing  to an inability to rapidly over-
     come these technical hurdles.


     3.9   Extrusion-Foaming Using CO2

        The extrusion-based foaming of polymers [38] is inherently  sustainable in that
     small amounts of raw material (the polymer) are used to create valuable, lightweight
     parts. The  low weight and/or low thermal conductivity of these parts ultimately
     saves energy in applications ranging  from home and appliance insulation to trans-
     portation components. Although the  parts themselves can be considered sustain-
     able, the conventional method of fabrication releases a large volume of solvent to
     the atmosphere. Prior to the late 1980s, CFCs often were employed as blowing
     agents (pore-forming agents), as  these solvents are low boiling,  nontoxic, and
     nonflammable. Subsequent to the acceptance of the  Montreal  Protocols (1986),
     most foam producers switched from  CFCs to hydrofluorocarbons, hydrocarbons,
     or mixtures of hydrocarbons and CO,. Generally, there is a desire within the foam-
     producing industry to move to 100 percent CO, as the blowing agent in extrusion
     foaming, although some serious technical hurdles remain. A variety of polymers are
     extrusion-foamed, including polyolefins, polystyrene, and polyesters. It should be
     noted that although injection of a volatile blowing agent to the extruder is probably
     the most common means to induce foaming, use of "chemical" blowing agents (i.e.,
     compounds that thermally decompose to form gases) also is employed.

        The extrusion-based  foaming of polymers is conceptually simple, yet requires
    complex analysis to fully understand the system. In the case of polystyrene, a fluid
    is injected  into the extruder, where  the pressure and temperature are sufficient
    (ostensibly) to create a single-phase mixture of blowing agent and polymer. Mixing
    is enhanced through strategic screw design. Following mixing, the melt is cooled (in
    some cases in a second, tandem extruder) to build melt strength, as the addition of
    the fluid greatly lowers the melt viscosity. The die is cooler still. Upon exiting the
    die, the rapid pressure drop creates a supersaturated solution, where small pores
    containing CO, nucleate and grow (nucleating agents often are added to stimulate

-------
                                                                               75
this process). The pores grow until the rapidly rising viscosity of the polymer
(owing to cooling and loss of blowing agent) restricts further expansion. In con-
ventional extruded foam, the cells are on the order of 100 to 1.000 microns in diam-
eter. Microcellular foam [198]. formed in much the same way albeit with higher
concentration of CO, in  the polymer melt, exhibits cells 50 microns and below in
size.

    The generation of foamed thermoplastics using CO, as the sole blowing agent is
most definitely  "green"  processing, as the CO, replaces either organic or hydro-
fluorocarbon agents that would otherwise directly enter the atmosphere. A number of
researchers have investigated the fundamentals of foam formation using high pres-
sure CO,, and several important conclusions have arisen [ 199]:

•   The number of cells nucleated during a pressure quench in a CO,-swollen
    polymer depends directly on the degree of swelling of the polymer. Swelling, in
    turn, rises as pressure rises and as temperature falls. To create more cells, con-
    ditions must be adjusted to ensure higher degrees of swelling.

•   The growth of cells  is dependent upon the degree to which CO, diffuses into
    the nuclei, and also the degree to which CO, expands as pressure drops. At the
    same time,  growth is inhibited by the retractive force of the polymer melt.
    which increases as the temperature drops and CO, diffuses from the melt. To
    make smaller cells, growth must be restricted soon after nucleation, by vitrify-
    ing the system before the pressure drops to the point where CO, begins to
    expand significantly. If one desires to make a large number of very small cells.
    then in theory one should start with a high degree of swelling of the polymer by
    CO,, and vitrify  the material as soon as possible after nucleation of pores. Un-
    fortunately, very high degrees of swelling lower the melt  strength (related to
    viscosity) significantly, and pores tend to coalesce during growth [200].

•   Our understanding of the fundamental processes that control foam morphology
    derives, in large part, to fundamental studies performed in academia and industry
    during the late 1980s and early 1990s. For example, early studies of the effect of
    pressure on the swelling of polymers by CO, by Berens and Huvard [201], Liao
    and McHugh [202], and Wissinger and Paulaitis [116] paved the way for future
    work on polymer foaming. Wang and Kramer [203] first explored the behavior
    of the glass  transition of a polymer versus CO, pressure in 1983; this was fol-
    lowed by a seminal study by Condo and Johnston [60]. Fundamental studies of
    the viscosity of polymer-CO, melts, for example, were performed by Manke and
    also by Khan [204]. These studies provided the data that made later studies of
    foam formation more tractable. Although it is likely that similar work was per-
    formed in industry, little of it can be found in the open literature, and the aca-
    demic work has been vital in providing a basis for recent foam research.

    Foam formed using CO, as the sole blowing agent has been commercialized in
a number of cases, yet the process is nonoptimal. as foam properties using CO, still

-------
76
     do not approach those when CFCs are employed as blowing agents. Although the
     foam-forming process is understood from an academic sense, a number of scien-
     tific/technical challenges remain before optimization can occur. These include:

     •   Shear effects on phase behavior. The phase behavior of CO,-polymer mixtures
         generally is measured (in academia) under static conditions: there have been
         reports that the phase behavior of CO,-polystyrene. for example, depends sig-
         nificantly on shear [205]. Measurement of high-pressure phase behavior under
         shear presents a significant experimental challenge, yet one that may have to
         ultimately be conquered if a full understanding of extrusion foaming is to be
         found.

     •    Pressure limitations in conventional extruders. Although extruders theoreti-
         cally can be operated at very high pressures (300+ bar), the typical operating
         pressure for a polystyrene foam extruder is approximately 100 bar at temperatures
         in excess of 470 K. At the same time, the swelling of polymers such as polystyrene
         is not sufficient under these conditions to produce foam of the same quality as
        can be produced with liquid blowing agents. Although raising the pressure is the
        usual remedy for insufficient swelling,  it is not a viable one in this case, and
        additives must be developed that will allow enhanced swelling of "CO,-phobic"
        polymers by CO, [206]. Further, these additives must be designed so as  to be
        effective at low loadings (or else foam physical properties and cost will be ad-
        versely impacted).

     •   Rapid diffusion of CO,. Compared to conventional blowing agents. CO, diffuses
        rapidly from foam pores—this rapid diffusion in practice contributes to foam col-
        lapse [207], Consequently, there is a need to develop additives that will partition to
        the CO,-polymer interface, then set up a barrier against CO, diffusion.

     •   High thermal conductivity of CO,. Insulation is a prime application for foamed
        polymeric materials. Further, the effective thermal conductivity of a polymer
        foam, at low-foam density, is a strong function of the thermal conductivity of
        the gas inside the pores. Because CO, exhibits a significantly  higher thermal
        conductivity than CFCs [208], larger quantities of foam may have to be em-
        ployed to provide the same  level of insulation if CO, is employed as the blowing
        agent. The blowing agent, although originally entrapped within the foamed
        polymer, will eventually diffuse out and be replaced by air diffusing in—the
        high diffusion coefficient of CO, renders this exchange faster with CO, than
        with chlorofluorocarbons. Thus, an additional challenge  is to achieve  high
        insulating value while employing CO,.

        Finally, a general conclusion that can be drawn from the extensive previous work on
    foaming is that, using the "swell-quench" method, one can generate a foam with either
    small pores (less than 10 microns) or low bulk density (less than 0.05 g/cc), but not both.
    Low bulk density requires the generation of very large numbers of small pores, and high
    swelling  (and hence high nucleation  density) but limited growth. Unfortunately, as

-------
                                                                                77
 mentioned previously, high swelling also leads to low melt strength and pore coales-
 cence. The lower limit for cell size in extruded foam with low hulk density « 0.1  g/cc)
 appears to he approximately 50 microns. Consequently, researchers have explored new
 strategies for forming low hulk density, fine-celled foams. For example, Enick and col-
 leagues [209] have generated molecules that will dissolve in CO,, then self assemhle to
 form gels. Removal of the CO, (viadepressurization) leaves behind a porous structure
 with submicron cell size and hulk density below 0.05 g/cc.

    In summary, the foaming of thermoplastics using CO, as the sole blowing agent
 is undeniably green  polymer processing, in that use of CO, directly replaces or-
 ganic solvent that ultimately would enter the atmosphere. The challenges to effi-
 cient use of CO, in foam production are given earlier—it should be noted that these
 are entirely technical and would provide excellent targets for future research.


 3.10 Industrial Activity:  Postpolymerization Processing

    As mentioned earlier, a large number of patents have been issued for both the
 foaming of polymers with CO, and the use of CO, to dye textiles. For the case of
 polymer foaming, the technology has achieved commercial status, both macrocellular
 foam formation (Dow, for example) and microcellular foam formation (Trexel has
 licensed technology developed at MIT by Nam Suh and colleagues [210]). The tex-
 tile work has been advanced to the pilot stage in Germany and in the United States.


 3.11 Use of CO2 in Polymer Science Applied to the
      Microelectronics Industry

    The preparation of an 8-inch silicone wafer requires hundreds of individual pro-
cess steps, of which approximately one-half involve washing [211]. It has been esti-
 mated that a single fabrication line will use more than 1 million gallons of solvent each
 year. In photolithography, the technique used to create patterned microelectronic com-
ponents, a polymer layer is applied to an inorganic substrate by spin coating from
 solvent, then selectively imaged and developed (washed off) to create a pattern. To
create the pattern, a mask is applied to the polymer layer, after which radiation is em-
 ployed to either crosslink the accessible areas (leaving the hidden areas uncrosslinked).
or degrade the accessible areas (leaving the hidden areas intact). The mask then is
removed and the soluble material (in either case) is washed away. Photolithography
currently employs significant volumes of either solvent or water to accomplish the
developing (washing) step, and generates a substantial liquid effluent stream. The key
 to successful developing is to be able to efficiently change the solubility characteristics
of the exposed portion of the resin. Carbon dioxide is a particularly intriguing solvent
 for use in microelectronics applications, not only because it is environmentally benign,
 but also because its vanishing low interfacial tension allows it to successfully wet and
penetrate very small features on a component.

-------
78
         Initial work to apply carbon dioxide to the coating and photolithography processes
     dates to the mid-1990s: researchers at IBM and Phasex Corporation examined the
     design of resins  specifically for use in CO,-based developing [212]—the work by
     DeSimone's group on the miscibility of perfluoropolyacrylates showed the IBM re-
     searchers that such  a process was feasible. A number of fluorine and silicon-
     containing polymers were examined, and a photoacid generator was employed to de-
     velop the patterns: the most viable system seemed to be one where a random copolymer
     of a fluorinated aery late and t-butyl methacrylate was used. Ober and colleagues [213]
     also have designed a photolithography system that could be developed using  carbon
     dioxide. A block  copolymer of a fluoroacrylate (CO,-soluble) and tetrahydropyrano
     methacrylate was synthesized. The polymer was spun-cast onto a substrate from a con-
     ventional solvent, and a photoacid generator was added. The system was masked, pat-
     terned (using 193 nm radiation) and developed with CO,, demonstrating that 0.2-mi-
     cron  features could be produced. DeSimone aJso has postulated the design of fluori-
     nated copolymers for use in photolithography [214]; both negative and positive resist
     systems are described. Interestingly, fluorinated materials are both highly CO,-soluble
     and known to be relatively transparent  to radiation in the 130 to 190 nm range [215]
     (the wavelengths to be employed in next generation systems).

        DeSimone and colleagues have described a free-meniscus coating methodol-
     ogy using CO, to apply polymers to inorganic substrates, potentially eliminating
     the significant volume of solvent  currently used for that purpose [214. 216].
     DeSimone has demonstrated the concept using fluorinated polyethers. polymers
     whose high solubility in CO, is well known.

        As suggested in recent articles in  Chemical and Engineering News [217] and
     Technology Review [218], interest in the use of CO, in microelectronics processing is
     growing. To date, most of the industrial ventures involve partnerships between large,
     well-known chemical suppliers to the electronics industry (Praxair, Air Products) or
     microelectronics companies (IBM) and small firms with expertise in the design of
     high-pressure equipment (Supercritical Systems [219]  (Fremont. CA; purchased by
     Tokyo Electron) and SC Fluids (Nashua, NH)). The efforts to date have focused on
     the use of mixtures of CO,  and cosolvents as a means to overcome the feeble solvent
     power of CO, without having to resort to the design of CO,-philic materials. Clearly.
     technical challenges for the future include the ability to design CO,-phiIic materials
     for use in microelectronics processing that also are acceptable (from both technical
     and environmental perspectives) to the industry. Indeed, do we possess a firm  under-
     standing as to the underlying molecular foundation for high CO, solubility as well as
     transparency to radiation of a particular wavelength? Today, the answer is "no." Will
    these underlying mechanisms ultimately conflict with one another? Further, given the
    rapid throughput in the industry, can high-pressure systems be developed that will
    allow use of CO, at the throughputs required? Finally, the work  to date on polymers
    for use in lithography has created materials where the exposed portion of the polymer
    is rendered insoluble in carbon  dioxide  (through action of a photochemically  gener-

-------
                                                                             79
ated acid on a protected carboxylic acid). It is somewhat surprising that we have yet
to see a system created where the exposed portion of the material is rendered soluble
in CO, instead.

    It is clear that if CO, can make significant inroads into the microelectronics
processing industry, then potentially large volumes of organic solvents, and just as
importantly water, could he replaced with CO,—once again, there are clear techni-
cal challenges to he overcome.


3.12 Industrial Activity:  CO2 and Polymers in
      Microelectronics Manufacture

    It was recently announced that Air Products and Chemicals had agreed to pur-
chase equipment from SC Fluids for use in photoresist development using carbon
dioxide  [219]. SC  Fluids also is  working with ATMI  (chemical supplier to the
microelectronics industry) and IBM on photoresist development using CO, [219].
Ashland Specialty Chemicals has formed an alliance with Dainippon Screen and
Kobe Steel to develop technology for microelectronics processing using CO, [220].

    In addition to using CO, to strip material from wafer surfaces, industry has
applied carbon dioxide processing to create porous materials that will function as a
low dielectric substrate or film [221].

-------
                            Section 4
                 Other Reactions in  CO,
    Researchers in both academia and industry (although most of the publications
come from academic laboratories) have conducted a large number of reactions in
carbon dioxide, demonstrating the feasibility for use of CO, in a broad range of
applications. Again, we must pose the question "Is this green chemistry?" Further,
what is the impact of this work on the greater chemical industry?

    If we examine the "E-Factors," or mass of waste per mass of product for vari-
ous industries, we find that chemicals and Pharmaceuticals produce waste at a rate
several orders of magnitude higher than that for bulk chemicals or petrochemicals
(see Table 2). However, if we examine the impact of each industry (related to the
E-Factor times the production rate), we see that the commodity segments still exer-
cise the greater impact.

   TABLE 2  Production and E-factors for various industry segments [222].
Industry Segment
Oil Refining
Bulk Chemicals
Fine Chemicals
Pharmaceuticals
Production (tons/year)
10" -10*
lO'-lO6
KF-IO4
10- 103
E-Factor (mass waste/mass
product)
-0.1
100
   Hence, if one had to choose the industry segments on which to focus research
efforts in use of CO, in green chemistry, it would seem that the obvious choice
would be bulk chemicals and petrochemicals. On the other hand, because fine chemi-
cals typically are produced in batch mode in small volumes, the cost of high-pres-
sure equipment for these industries may not be as much of an impediment as it
would be for their commodity cousins.

   Finally, as noted in a later section, the education of chemists in the use of CO, as
a solvent has a value of its own, and as such, the publication of papers on reactions
                                  81

-------
82
     that fall into this chapter has done much to "demystify" CO,. Therefore, these papers
     have significant educational value.


     4.1    Enzymatic Chemistry

        At first glance. en/yme/CO, mixtures appear as ideal reaction systems for the
     performance of green chemistry. Enzymes are naturally derived catalysts that are
     highly selective, but CO, is a naturally abundant, benign solvent. However, research
     into enzymatic reactions in CO, has dropped precipitously since the mid-1990s. and
     no commercialization of such processes currently is anticipated. The reasons for this
     are straightforward and scientifically based, deriving from the substantial research
     performed in this area during the 1990s.

        Enzymes are naturally derived catalysts, proteins whose primary, secondary, and
     tertiary structures have evolved to create a catalyst that is highly selective and very
     active under a set of narrowly defined conditions. Enzymes are green catalysts, and
     their means of production (fermentation) also is typically a green process. In nature.
     enzymes perform their catalytic function in water, yet Klibanov (and others) showed
     that enzymes would function adequately (not as well as in water) in organic media
     provided that a small amount of water remains bound  to the enzyme [223]. Further.
     although lipases (and other analogous enzymes) naturally perform hydrolysis reac-
     tions in an aqueous environment, these same enzymes were shown to perform esterifi-
     cation in an organic environment. Because enzymes do not dissolve in the organic
     solvents under consideration, enzymatic chemistry in organic solvents is governed by
     heterogeneous reaction kinetics. This, however, is not a drawback, as catalyst recovery
     is easier than for a homogeneous system. Given this background, enzymatic reactions
     in CO, seemed an ideal combination of green solvent with green catalyst.

        During the early 1990s. a number of enzymes were evaluated in carbon dioxide.
     primarily in support of esterification reactions [224]. For the most part, activities
     were very low, much lower than for the same reaction conducted in a conventional
     organic solvent. In addition, rates in CO, were substantially lower than rates in other
     compressible fluids (ethane, propane, fluoroform). In some key publications, Russell
     and colleagues outlined the reason for CO,'s low activity—apparently carbon diox-
     ide reacts with primary amine residues (primarily from lysine) to form carbamic acid
     and/or ammonium carbamates [225]. This derivatization was observed experimen-
     tally, and is apparently responsible for the reduced activity of many enzymes in CO,
     (note that not all enzymes suffer from this reduced activity, consistent with the fact
     that enzymes exhibit a range of protein sequences and macrostructure). Carbamate
     formation is reversible, as removal of the enzyme from CO, followed  by examination
     of the rate in either water or another organic solvent reveals no change in inherent
     activity. Even bubbling of gaseous carbon dioxide through a suspension of enzyme in
    organic solvent can produce the reversible drop in activity. Consequently, interest in
    enzymatic chemistry using enzyme powder in CO, diminished greatly.

-------
                                                                              83
    At this same time, advancements in the design of CO,-philic surfactants al-
lowed for the possibility of performing enzymatic chemistry in the aqueous core of
micelles formed in carbon dioxide, a situation that would eliminate the problems
due to carbamate formation (polar solvents destabilize the carbamates). Indeed,
work by Randolph and Johnston [226], as well as Beckman and colleagues [227],
showed that one could solubilize an  enzyme in the core of a  micelle,  and then
recover the protein via depressurization. However, CO, dissolves in water and forms
carbonic acid, and not surprisingly the pH within the micelles was shown to be less
than 3.0. Although Johnston showed that one could buffer such a system to pHs
ranging from 5.0 to 6.0 [32], the ionic  strength required was far higher than would
normally be recommended for use with an active enzyme. Thus, realization of the
full "green" potential of enzyme-CO,  systems was again blocked by technical re-
alities. Because CO, was the most widely preferred supercritical solvent, interest in
enzymatic chemistry in CO, declined rapidly after the publication of the key papers
described earlier.

    Other issues to note regarding the use of enzymes in CO, include the need by
the enzyme  for a certain  amount of bound water and the equilibrium nature of
many of the reactions. Although CO,  usually is considered a nonpolar solvent, it
will solubilize approximately 2,500 ppm water at moderate pressures (100 bar,
room temperature). Because enzymes will not function in organic media if stripped
of all of their water, care must be taken to prevent CO, from dehydrating the en-
zyme. In addition, many of the enzymatic reactions that one might wish to perform
in CO, are governed by equilibrium,  and one must examine means by which to
remove the byproduct or product  from the neighborhood of the enzyme.

    A final obstacle to use of enzymes in supercritical fluids lies in the poor solu-
bility of many of the polar substrates that one might want to transform. For ex-
ample, although many of the literature studies performed during the early 1990s
examined esterifications,  the starting  material (carboxylic acid) usually was not
particularly soluble in CO2 (hardly surprising given what is known about CO,).

    The previous paragraphs make plain the technical hurdles that would need to be
overcome to render enzymatic chemistry in CO, generally practical and useful. Either
enzymes must be identified (or developed through a directed evolution-like process)
that do not form carbamates with CO, (or where carbamate formation does not impede
activity) or a way must be found to buffer a CO,/water mixture without resorting to an
ionic strength that will harm the enzyme. Conversely, identification of enzymes that
thrive at low pH or high ionic strength also would be worthwhile in this  regard.

    If the problems described above could  be overcome, a number of issues re-
garding the use of enzymes in compressible fluids could be evaluated. For ex-
ample, work by Russell [228] using fluoroform showed that pressure (through its
effect on fluid properties) could  be used to tune enzyme activity and also, to a
certain extent, selectivity for a given reaction path. However, given the preference

-------
84
     for CO, versus other compressible fluids, until the problems regarding CO, and
     enzymes are dealt with, enzymatic chemistry in compressible fluids will likely con-
     tinue at only a very low level of research activity.


     4.2   Diels-Alder Chemistry

         The Diels-Alder reaction is employed on a large scale industrially to purify
     cyclopentadiene, and to a lesser extent, to manufacture anthraquinone [ 14]: it should be
     noted that these reactions proceed without solvent. A substantial body of literature ex-
     ists concerning Diels-Alder chemistry in supercritical fluids. CO, in particular. For the
     most part, research on this particular reaction has been used (via analysis of the rate
     constants), to confirm the influence of concentration fluctuations (present near the criti-
     cal point) on the rate of the reaction. In general, the rate reaches a maximum near T\
     dropping at both higher and lower pressures. However, this work currently is of scien-
     tific interest only, as control of a reaction in the neighborhood of a critical point  is
     problematic at large scale. Tester and colleagues [229] report that most  Diels-Alder
     rate constants in CO, can be correlated using a simple Arrhenius expression provided
     that the preexponential term varies linearly with fluid density, similar to what Roberts
     [230] observed using propane as the solvent. Lewis acid catalysts are effective (if soluble).
     as shown by Matsuo and coworkers using scandium triflate in CO, [231 ].

        Although the literature on Diels-Alder chemistry in CO, at first glance appears
     uninteresting (from a green chemistry viewpoint), there are some publications that merit
     closer scrutiny. For example, Ikushima, et al.  [232], published the results of a study of
     the cycloaddition of isoprene and methyl acrylate (Scheme VII), reporting that while
     one atmosphere conditions produced primarily the para isomer of the methyl acetoxy
     cyclohexene product, operation in CO, produced significant amounts (at some pres-
     sures the major component) of the meta isomer. If true, such a result suggests that use of
     CO, can alter product selectivities and would significantly impact the field of green

     Scheme VII:
             X

-------
                                                                               85
 chemistry in critical fluids. However, subsequent work by Danheiser and Tester [233]
 revealed that the Ikushima group failed to note that multiple phases were present in the
 reactor, and that adequate sampling of the phases revealed that all conditions produced
 a 67-31  split of para and meta isomers. This again shows the importance of understand-
 ing the phase behavior of any reaction mixture under evaluation.  Indeed, subsequent
 work by Danheiser and Tester on a wide range of Diels-Alder substrates revealed no
 effect of CO, pressure on regioselectivity.

    Some additional observations on Diels-Alder chemistry in CO, include reports
 by Clifford and colleagues [234J that the endo:exo ratio of products in the reaction
 between methyl acrylate and cyclopentadiene exhibits a maximum  versus pressure
 in CO,. Totoe and colleagues [235] also observed differences in  product selectivity
 between toluene and CO, in a 1.3-dipolar cycloaddition.

    In summary, although there have been some intriguing reports on variations in
 selectivity in CO, versus conventional solvents, most of the research on  Diels-
 Alder chemistry in CO, has been directed at deriving fundamental parameters rather
 than creating opportunities for green chemistry perse. The work by Danheiser and
 Tester should stand as a warning to those involved in chemistry  in supercritical
 fluids—ignoring phase behavior effects is at one's own peril.


 4.3  Lewis Acid Catalysis/Friedel-Crafts Chemistry

    Friedel-Crafts chemistry is used extensively to perform liquid-phase alkyla-
 tions and  acylations. although  it  should be noted that the largest scale industrial
 processes do not employ solvent, and some have switched from the typical alumi-
 num halide "catalyst" to support acidic catalysts  [14]. However, fine chemical syn-
 theses often employ relatively toxic solvents during Friedel-Crafts reactions, and
 this reaction presents a viable target for use of CO,. Because Friedel-Crafts chem-
 istry is usually performed in polar media, an obvious question is  whether CO, (with
 its low dielectric  constant) can actually support such reactions. Further, the pri-
 mary environmental drawback to Friedel-Crafts chemistry is  the need for large
amounts of aluminum halide. and much recent research has focused on finding true
catalysts for the various alkylations and acylations. Interestingly, many of the newer
Friedel-Crafts "catalysts" are fluorinated and highly CO,-soluble.

    Chateauneuf and Nie [236]  examined the alkylation reaction between methoxy
benzene and triphenyl methanol using trifluoroacetic acid as catalyst. Kobayashi
and coworkers  [237] found that  rhenium triflate promoted the acylation of aro-
matic compounds (as in Chateauneuf's work, if electron donating substituents were
present on the aryl compound) with an anhydride. The reaction proceeded smoothly
 in either organic solvents or CO,. Finally, Poliakoff's group examined the Friedel-
Crafts alkylation of various activated aryl compounds using a supported (Deloxan)
acid catalyst in CO, [238]. Although not large, the literature on Friedel-Crafts chem-

-------
86
     istry in CO, demonstrates that this reaction is indeed feasible, and that many of the
     Lewis acids proposed as catalysts are readily CO,-soluhle.

         Olah and colleagues [239] examined the acid catalyzed isobutene-isobutylene
     reaction in carbon dioxide: they found that CO, acted as a weak base, and use of CO,
     as a solvent lowered the acidity of the system and hence the alkylate quality. How-
     ever, in cases where the acidity was increased to counteract this effect, the use of CO,
     decreased the amount of acid needed to perform the alkylation. Further, use of CO,
     increased the octane number of the product.

         In a final intriguing note, Pernecker and Kennedy [240], during an investiga-
     tion into the Lewis acid catalyzed polymerization of isobutylene in CO,, found that
     addition of only the Lewis acid to carbon dioxide formed a product, either a solid
     precipitate or a second liquid. Removal of the CO, regenerated the original Lewis
     acid. On the other hand, incubation of a Lewis acid with the polymerization initia-
     tor, followed by addition to CO,, resulted in  no "CO,-product" formation.
     Pernecker's results suggest that one might activate CO, itself for further reaction
     using a Lewis acid, but if the Lewis acid is presented with a more  reactive sub-
     strate, it will preferentially bind to this substrate.

         In summary, Friedel-Crafts chemistry is (in fine chemical synthesis) performed
     in solvent, and CO, represents a potentially useful and green substitute. Catalysts
     that would be used ordinarily to perform such reactions are soluble in CO, without
     further modification. The effective use of CO, then depends on substrate solubility.


     4.4   CO2 as  Reactant and Solvent

        This section will discuss those reactions where  CO, is employed as reactant and
     solvent, yet where small molecules  (rather than polymers, see Section 3) are formed as
     products. A large number of reactions using CO, as a raw material have been demon-
     strated in the laboratory, but very few such reactions are practiced commercially. For
     example, it has been shown in the  literature that one can generate formic acid [241],
     dimethyl formamide [242], carboxylic acids [243], and methanol [244] using CO, as a
     reactant (and in many cases the solvent as well). To date, however, the economics of
     such processes have not been sufficiently favorable to warrant significant industrial
     attention. Part of the problem is that use of CO, to create commodities such as those
     listed above competes directly with use of highly reactive CO to create the same mol-
    ecules. For example, methanol is produced  from CO and hydrogen (synthesis gas, or
    syngas) in an atom-efficient process [14]. Further, the needed synthesis gas can be
    readily generated from coal, natural  gas, or petroleum. To form methanol from CO,, an
    additional clean and inexpensive source of hydrogen would be needed. Further, the
    thermodynamics of the two routes are such that one can obtain twice the yield of metha-
    nol from the syngas route (at 470 K, for example) than the CO, route [245]. At present,
    CO, is used only to supplement syngas during methanol production if the ratio of hy-

-------
                                                                                87
drogen to CO is significantly higher than 2.0 (which can occur when natural gas is used
as the syngas source). Other small molecules such as formic acid, formates, and
formamides then are generated from methanol (plus CO. ammonia, alkyl amines)—
this chemistry also is  atom-efficient and alternative routes using CO, as a starting
material have been unable to compete. In general, it is presumed that CO,-based routes
for basic commodity chemicals would be competitive if a relatively inexpensive, non
CO,-producing source of hydrogen can be developed [245]. Granted. CO is a much
more toxic material  than CO,, yet syngas has been used successfully for decades in
chemical processes, so this factor currently carries little weight.

    The generation  of dialkyl carbonates presents a similar example to those de-
scribed above—a number of researchers have investigated the synthesis of dialkyl
carbonates from CO, and alcohols using alkoxy tin catalysts [246]—in this process
the equilibrium must be pushed towards product  via the removal of alcohol. Mean-
while, the commercial process operates very effectively from CO and alcohol over
relatively inexpensive copper catalysts [247].

    Despite the negative results described above, it is important to note that approxi-
mately 110 megatons of CO, are consumed each year to produce low molecular
weight products [245]. Most of this is consumed  to generate urea: in addition, sali-
cylic acid is synthesized (Kolbe-Schmitt reaction) from CO, and a phenolic salt, but
alkylene carbonates  are generated from the analogous alkylene oxides and CO,. The
alkylene carbonates  are considered relatively benign solvents (they exhibit low tox-
icity and low vapor pressure), and their synthesis from CO, is an example of green
chemistry. Monsanto, as well as academic researchers,  have studied the synthesis of
isocyanates from CO, [ 173]. Although the traditional route reacts amines with phos-
gene, creating the isocyanate plus salt, the CO,-based routes react the amine with
CO, in the presence of strong dehydrating agent. The yields of such CO,-based reac-
tions are excellent, yet the cost of the dehydrating agent (or rather, its regeneration)
has inhibited commercialization of such chemistry. Behr, among others, has reviewed
a range of small molecule reactions that employ CO, as a reactant [248].

   In summary. CO, has the potential to be a useful C: synthon but recent work,
while scientifically  interesting, has not led to processes that can effectively com-
pete with existing routes/plants. Further, when considering CO, as a green reactant.
one always must be cognizant of any energy differences required to employ CO, in
a synthetic scheme versus a conventional reactant (such as CO). If use of CO, is
more energy intensive, then one might create a situation where more CO, is created
than chemically "sequestered."


4.5   Other Organic Reactions

   As was mentioned previously, volatile metal carbonyls (for example) exhibit
sufficient solubility  (or sufficiently low miscibility pressures) to support catalysis

-------
88
     in CO, without catalyst modification. As such, there are a number of examples in the
     literature where CO, has been used as a "drop-in" replacement for catalytic reac-
     tions ordinarily conducted in organic solvents. Nevertheless, once Leitner and
     Tumas demonstrated  in 1997 that one could perform homogeneous catalysis in
     CO, if the catalyst ligands were properly designed, a number of researchers have
     extended this work, examining a wide range of name reactions in CO,. The impor-
     tance of the Leitner and Tumas papers was perhaps to demonstrate that any catalyst
     could be rendered CO,-soluble.  if the fluorination of the ligands could be accom-
     plished synthetically. Consequently, carbonylation [249], Heck and Stille couplings
     [250], vinylic substitution [251 ]. hydrosilation [252]. isocyanate trimerization [253],
     dechlorination [254]. Pauson-Khand cyclization [255], and others have been per-
     formed successfully in carbon dioxide. The use of fluorinated catalyst ligands is
     common, providing the solubility needed for the reaction to proceed smoothly.

        Although these papers demonstrate  the scope of "chemistry  in CO,." the
     impact of such work on the overall aims of green chemistry is not clear. Granted.
     such reactions would ordinarily be performed in an organic solvent, and use of CO,
     replaces such solvent use. On the other hand, the reactions described above typi-
     cally are used for small  volume, batch reactions, and the overall impact of this
     work on the greening of industrial chemistry will be small. Perhaps the most sig-
     nificant impact of this work on green chemistry is in its ability to show chemists
     that CO, is a viable solvent for a  variety of reactions, and  the greatest value of the
     work may be to educate the next generation of chemists.


     4.6  Industrial Activity: Friedel-Crafts Chemistry and Other
          Name Reactions

        Both  Poliakoff [256] and Subramaniam [257]  have  patented alkylations in
    supercritical fluids, albeit using different types of catalysts. Each of these academic
    groups is/was working with an industrial partner (Thomas Swan and Engelhard.
    respectively), and the work ultimately may be transferred to industry.

        Schiraldi and colleagues, as well as Harris and coworkers [259] have patented
    the esterification of specific substrates in carbon dioxide. Finally, a group at BASF
    has patented the generation of alpha-tocopherol (and derivatives) in carbon diox-
    ide [260]. It is not clear at this time if these inventions are being pursued further by
    the companies involved.


    4.7   Inorganic Chemistry:  General

       Obviously, most inorganic compounds are not soluble in carbon dioxide, and
    inorganic chemistry performed in or with CO, has been accomplished by finding
    ways around this  seemingly intractable thermodynamic hurdle. The first inorganic

-------
                                                                                89
chemistry performed in a supercritical organic solvent was probably the work by
Matson [261 [ at Battelle Pacific Northwest Laboratories in the late 1980s—an
emulsion was formed in a supercritical alkane. and inorganic particles were gener-
ated via a reaction at the micellar interface between an inorganic and an organic
precursor (note that when Matson performed his study, it was not possible to form
micelles in CO,). Recently, several research groups have adopted the same strategy
to create metal nanoparticles within micelles formed in carbon dioxide. Naturally.
the great strides  made during the 1990s in the identification and application of
CO,-philes paved the way for this research. Both Fulton [262] and Roberts [263]
have reported the formation of metals particles with diameters less than 20 nm by
(a) creating an emulsion in CO, where the aqueous cores of the micelles contain
metal ions as well as water, and (b) adding a reducing agent to the CO, such that a
reaction occurs at the micellar interface between ion and reducing agent to nucle-
ate the particles. Particle growth then occurs through micelle-micelle collisions—
Roberts has shown that one can control the particle growth rate via control over the
degree to which the micelles can collide and exchange contents. Further, changing
the physical properties of the compressible continuous phase can alter the micellar
collision rate.

    An obvious question is "Is this green chemistry?" Because there currently is no
sizeable industrial market for metal nanoparticles. this question is difficult to answer.
Production of metal  nanoparticles in a CO,-continuous emulsion likely will be more
environmentally friendly than the analogous reaction in an organic solvent. However,  if
such metal nanoparticles are ultimately applied commercially, there also may be other
means by which to synthesize them, means that require no solvent at all. As can be seen
by this and other such situations, it can be difficult to judge whether a process is green
unless taken in context with competing processes—green seems not to be an absolute
but rather a relative concept.


4.8  Inorganic  Chemistry:  Metal Chelates

    Although separations will not expressly be covered in this report, the use of
chelating agents for metal extraction should be noted. Many conventional chelat-
ing agents and their associated metal complexes are poorly soluble in carbon diox-
ide, and concepts on the design of CO,-philic materials were applied very early to
the design of CO,-soluble chelating agents [264].  showing that fluorination im-
proved solubility. On the other hand, tri-alkyl phosphates and tri-alkyl amines, known
to bind several types of metals, have been shown to be miscible with CO, at mod-
erate pressures despite containing no fluorine. Various research groups [265] have
demonstrated that one can extract metals (using the appropriate agent) from both
solid and liquid  matrices at high yields. Also, it has been shown that the phase
behavior of the metal chelate  can be substantially different from that of the agent
(not surprising because  at the very least the molecular weight of the chelate is much

-------
90
     greater than that of the agent). Finally, one of the first advances in the design of
     nonfluorous CO,-philes came about as a result of work by Siever's group on chelating
     agent structure-solubility relationships [266]. It was shown that, in the case of cop-
     per-beta-diketone complexes, the solubility of analogues containing branched alkyl
     groups was superior to fluorinated analogues.

         Again, we must pose the question. Is the use of chelating agents in carbon dioxide
     green chemistry/processing? The two most important cases for examination, where
     metals are processed/purified for sale and where metals must be removed from solid or
     liquid matrices to remedy an environmental problem, will be examined.

         Regarding the  first case, both copper and precious  metals  (platinum group
     metals, or PGMs) are purified using solvent extraction. In the case of copper, sol-
     vent extraction and electrowinning (SX-EW) have captured approximately  15 to 20
     percent of the total  amount of copper produced worldwide [267]. replacing  the
     significantly less green (owing to energy use and air emissions) conventional smelt-
     ing process. In SX-EW. the metal first is extracted from the ore using sulfuric acid
     (along with substantial amounts of silver, lead, iron. zinc, and arsenic, plus a wide
     variety of minor components) via heap leaching, where the acid simply is allowed to
     flow by gravity  through an ore pile. This acidic solution then is contacted with an
     organic solvent containing an extractant (one of a variety of amines, phosphates, or
     oximes) to draw the copper selectively into the organic phase (usually a high flash
     point alkane mixture). The copper is back-extracted into water,  from where it is
     electrochemically reduced (electrowinning) to pure (99.99+ percent) copper. The
     solvent extraction step is, from a process perspective, somewhat simple, consisting
     of a series of mixer-settler tanks that are open to the environment.

        Previous work has shown that copper can be extracted into carbon dioxide:
     further, it is likely that one could synthesize a highly CO,-soluble analogue to one of
     the currently used commercial extractants for copper. Hence, one could construct a
     CO,-based analogue to the current solvent extraction process.  However, it is  not
     likely that the cost of such a step would justify the move away from the currently used
     organic solvents. At present, the solvent extraction/back extraction steps contribute
     approximately 10 to 20 percent of the $0.2/lb processing cost of copper using SX-EW.
     assuming that more than 90 percent of the extractant is recovered after each use [268].
     Indeed, perhaps  a far better target for green processing applied to copper refining
     would involve either conversion of the remaining traditional smelters over to SX-EW
     [269] or finding  ways in which to lower the energy demand of the ore excavating/
    crushing/grinding process or the electrowinning step [270]. A further complication is
    that most copper refining is performed in either South America or Africa, where  the
    regulatory and/or societal driving force for adopting green chemical processing is
    substantially less than in either Europe or the United States.

        Platinum group metals, either those derived from ore or during the recycling of
    catalytic converters or electronics components, also are refined using solvent extrac-

-------
                                                                                91
 tion [271]. The metal is extracted using strong acid (usually HC1). then purified by
 extraction into organic  solvent using an auxiliary, where selectivity is achieved via
 both the design of the auxiliary and subsequent aqueous washing steps to remove
 unwanted trace metals.  The extraction is multi-step, so as to sequentially remove the
 gold, platinum, palladium, and other PGMs. The metals then are reduced either chemi-
 cally or electrochemical!} and recovered. The  opportunities for the use of carbon
 dioxide to replace organic solvents in such processes mirror those in copper refining:
 however, the value of the metal is five orders of magnitude greater. Further, it has been
 shown that CCX-soluble analogues can be designed to those compounds used  to
 extract PGMs into organic solvents [272]. However, just as the value of PGMs makes
 the use of CO, more viable, so too does it promote the development of competing
 technologies. For example. IBC (Utah) has developed solid metal absorbents com-
 prised of macrocycles tethered to polymeric resins [273]. These resins have been
 shown to selectively hind PGMs of various types, where the metals are recovered by
 back extraction following processing. If CO, is to  be competitive in this arena, the
 ligands must be selective, should be as inexpensive as possible, and/or one must be
 able to recover them following binding and release of the metal. Both the ligands and
 their metal complexes must be  highly soluble at low pressures (preferably CO,'s vapor
 pressure), as throughputs in this application will be very high. As in the case of coffee
 decaffeination.  it would be highly preferable to reduce and/or capture the  metals
 without depressuri/ation of the CO,. Given Watkin's research, it may be possible, for
 example, to reduce the metals using added hydrogen. Unlike in the case of conven-
 tional organic solvents, adding hydrogen to CO, produces neither safety nor mass
 transport problems. There are two features of this process that weigh in favor of CO,:
 (a) the metal concentration is relatively low. meaning that employing a high ligand:metal
 ratio still allows for dilute ligand concentrations: and (b) aqueous flow rates  can be
 higher than the point that causes breakthrough problems for solid sorbents. There
 may be opportunities for use of CO, in this industry.

    Another application of potential interest is in the upgrading of so-called vacuum
 resid (or vacuum residual) in petroleum refineries [274]. Vacuum resid refers to low
 vapor pressure (hence relatively high molecular weight) fractions of the initial petro-
 leum stream. In addition to hydrocarbons, this fraction contains a substantial quan-
tity (more than 1.000 ppm) of a wide spectrum of metals (owing to the concentration
effects of numerous upstream unit operations). Included in this mix of metal contami-
nants are considerable amounts of vanadium and nickel, metals that can deactivate
the catalysts employed  to crack petroleum into useable (salable) materials. Further.
both the nickel and vanadium  are complexed by porphyrin type materials present in
the vacuum resid. If these metals could be easily and economically extracted, more of
the initial petroleum stream could be employed to create saleable products, meaning
 less is simply burned.

    Aqueous waste from electroplating operations generally contains substantial
amounts of dissolved metals in a low pH (2.0 and below) medium. Chelating agents

-------
92
     dissolved in carbon dioxide can he used to extract many of the relevant metals From
     such low pH media |275J. provided that the agents are designed to operate under
     such conditions. Generally, the strategy by which dictating agents are rendered
     CO,-soluble involves the attachment of "CO.-philic" functional groups to a moiety
     known to bind certain metals, and as such there are in theory no restrictions as to
     the type of chclating agent employed, so long as the functionalization chemistry
     can be performed. The competing technologies for CO, extraction include the use
     ot precipitants. compounds that react with dissolved metals to form insoluble spe-
     cies, as  well as chelating agent-functional ion exchange resins (solid sorbents).
     Precipitants are inexpensive, yet they produce a sludge that must be collected and
     disposed. Ion exchange resins (following back extraction) produce instead a con-
     centrated (ideally)  solution of the metals, which must be subsequently treated to
     recover the metal.

         The most problematic application to analyze is where CO,, plus a chelating agent.
     is being used to remove metals from a matrix to accomplish remediation. Indeed, the
     primary focus of green chemistry is the elimination of waste production, rather than
     the clean up of existing problems. Yet, the use of CO, to remediate metal contamina-
     tion may be  considered green processing in some circumstances. First,  it has been
     shown by various research groups that a variety of metals can be extracted from solid
     matrices (including  soil [276]) using chelating agents dissolved in carbon dioxide. If
     CO, was to be used to replace either an organic solvent or water in the  washing of
     contaminated soil, this could be considered green processing, provided that the en-
     ergy required for the process was equal to or less than that employed for the conven-
     tional route.  A large amount of sludge (as much  as  15 percent of soil throughput.
     created from suspended fine particles) is produced, for example, when soil is washed
     with water. Because carbon dioxide is a low density, low  viscosity, low interfacial
     tension fluid, it is likely that sludge production would be greatly reduced if CO, were
     used to wash soil. On the other hand, because soil washing typically involves excava-
     tion of the contaminated material, remediation strategies that eliminate the problem
     without excavation (in  situ remediation) should be preferred. Such strategies range
     from the use of  green plants to absorb and concentrate metals, to the addition of
     agents to the oil that stabilizes the metals, thereby preventing their transport.


    4.9  Inorganic  Chemistry:  Industrial Activity

        Materials Technology Limited has obtained several patents [277] describing the
    use of high-pressure CO, to enhance the rate  of curing of concrete, where the CO,
    actually dissolves in the concrete mixture and reacts with the matrix. Although one
    might consider this as sequestration of CO,, and hence green chemistry, it should be
    remembered that the preparation of the concrete precursor involves the calcining of
    the raw material, where CO, is driven off while injecting significant energy. Thus, more
    CO, is probably produced during this sequence than is sequestered.

-------
                                                                               93
    Both Texas Instruments [278] and Micron Technology [279] have patented
inventions where inorganic chemistry is performed in CO, to support cleaning/
processing of silicone wafers. The Micron patent describes the use of mixtures of
CO, and etching chemicals to pattern inorganic substrates, while the Texas Instru-
ments patent describes a process where inorganic contamination on wafers is first
derivatized. then dissolved in CO, and removed. Note that in these patents, the use
of CO, is designed to replace the use of water. In many parts of the world, signifi-
cant water usage by industry is not sustainable, and there is a need to find replace-
ment technologies for large-scale water usage.

-------
                            Section 5
        Formation of Fine Particles  Using
                                 C02
    The controlled formation of particles (or powders) is important to several dis-
parate industries, including those that manufacture pigments, Pharmaceuticals, and
catalysts. Needless to say. these diverse applications mandate a diverse set of speci-
fications for the production of such particles. Not surprisingly, supercritical fluids
(and carbon dioxide in particular) have  made inroads into particle production to
varying degrees, with penetration more significant in some industries versus oth-
ers. In particular, the benign properties of carbon dioxide (vis-a-vis intimate con-
tact with humans) have created substantial interest within the pharmaceutical pro-
duction community for use of CO, in the generation of therapeutic paniculate prod-
ucts. In  some cases, the use of CO., is proposed to supplant the use of organic
solvents, and such a process could  rightly be termed green  processing. In other
cases, the use of CO, (plus auxiliaries, as will be described below) might actuajly
be less "green" than a current process, but the characteristics of the product are
superior, providing a performance rather than an environmental advantage. Fur-
ther, because regulatory approval on new products or processes (in the pharmaceu-
tical industry) can require years to obtain, the industrial impact of CO, processing
of pharmaceutical powders may not occur for some time (if at all, naturally). How-
ever, recent industrial investment (by entities in the pharmaceutical industry) in
supercritical fluid technology suggests that the level of interest remains high.


5.1   Production of Particles Using CO2: RESS

   The earliest particle formation process using CO, as the solvent is probably the
often cited paper by Hannay and Hogarth in the 19th century, where depressuriza-
tion of a CO,-based solution created a precipitate "like snow" (see [1] for descrip-
tion). During the 1980s, researchers at Battelle's Pacific Northwest Laboratories
created the RESS (Rapid Expansion of  Supercritical Solution) process, where a
solution (of solid in supercritical alkane) was sprayed through a nozzle (where the
                                  95

-------
96
     outlet was at atmospheric pressure), creating fine particles [280]. Other research-
     ers have explored the use of RESS to form particles since then, both from an ex-
     perimental and theoretical standpoint [281 ]. As mentioned previously, CO, is not a
     particularly  powerful solvent, and many of the solutes one might like to process
     using RESS  require very high pressures (500 bar and above) to dissolve even small
     quantities of material—high CO, throughput will be needed to produce relatively
     small amounts of particles. The high CO, throughput (with  its associated  costs.
     capital and operating) has effectively inhibited the  use of RESS on a commercial
     basis. This has rendered RESS generally less interesting than some competing CO,-
     based particle  formation technologies; these will be described below.

         The most successful (from a developmental, if not yet truly commercial point of
     view) particle-forming processes are those that have taken what is known about CO,'s
     thermophysical properties and applied these characteristics strategically. For example.
     as has been mentioned previously, it is well known that CO, is a rather feeble solvent—
     although problematic when attempting to use CO, in a RESS process, this characteris-
     tic is quite useful when CO, is employed as a nonsolvent to induce precipitation of a
     solute from organic solvent. Further, whereas high pressure is required to create dilute
     solutions of large molecules in CO,, low pressures are sufficient to create  solutions of
     CO, in large molecules (or solutions of compounds in organic solvent), as suggested in
     Figure 6. Saturated solutions of CO, (in either polymers or solute/solvent mixtures),
     sprayed through nozzles, have been used to generate fine particles.


     5.2   Creating Fine Particles Using CO2: Nonsolvent Modes
           of Operation  and PGSS

        Jung and Perrut have written an excellent review of the use of supercritical fluids to
     generate fine particles [281 ]; other reviews have appeared recently as well [282]. These
     reviews describe the wide variety of materials that have been micronized via CO,-
     based processing, and the various modes in which such particle processes operate.

        During the 1980s, Krukonis and colleagues [283] found that CO, could be em-
     ployed as a nonsolvent to induce  controlled precipitation of various solutes from
     organic solvent solution. The success of this approach derives from CO,'s generally
     feeble solvent power yet its miscibility with a variety of volatile organic solvents.
     The use of CO, as a nonsolvent to produce particles has expanded significantly since
     then, where the typical "process" employs one of several nozzle designs to create an
     aerosol simultaneous with the induced-phase separation. As shown in the review by
     Jung and Perrut [281], an extraordinary variety of materials (many bioactive com-
     pounds) have been processed via one of the many nonsolvent routes, typically gener-
     ating micron-size particles and smaller.

        As noted in the section on polymer processing, the pressure required to create
     a concentrated mixture  of polymer and CO, is significantly  lower than that re-

-------
                                                                               97
quired to create a dilute solution of polymer in CO, (see Figure 6). As such, a
number of researchers have explored the use of gas-saturated solutions (of either
CO, in a polymer, or CO, in an organic solvent/solute mixture) to produce fine
particles. The CO,-saturated mixture is sprayed through a nozzle, and the rapid
vaporization of CO, creates an aerosol and removes any organic solvent. The work
by Ferro Corporation on the generation of powder coating formulations using CO,
is an example of this type of processing, sometimes referred to as PGSS (particles
from gas-saturated solutions).

    Although a variety of materials have been micronized using carbon dioxide, it
is clear that most of the industrial interest in such processes arises from pharma-
ceutical manufacturers. As such, we will focus on bioacti ve particle manufacture in
discussing the green potential of these processes.


5.3  Production of Fine Pharmaceutical  Powders: Is This
      Green Processing?

    To determine whether CO,-based particle formation processes are "green," one
must first examine the ways in which particles currently are generated. First, it seems
clear that the pharmaceutical industry is truly interested in the production of fine
powders (panicles) of controlled size and known purity. The design and testing of
inhalable drugs is an ongoing area of significant research and business activity.

    The CO,-based particles processes described in the literature are green (and eco-
nomical) to varying degrees. For example, although RESS employs CO, as the only
solvent, the need for high CO, throughputs (owing to  low solubility of target com-
pounds) means that the energy budget for such a process will be high (energy' needed
for compression and purification of large volumes of CO,). On the other hand, pro-
cesses such as PGSS or the various nonsolvent modes of operation employ carbon
dioxide at relatively low pressure and flow rates. Many of the antisolvent processes
employ organic solvents (dimethyl sulfoxide (DMSO) most frequently), and care must
be taken to "close the loop" on these solvents to avoid lowering the sustainability of the
process. Because CO,-based particle production processes are, at most, at the pilot
scale, it is not clear to what extent the organic solvent can actually be recycled. Further,
if the particle process requires regulatory approval (for use in manufacture of pharaia-
ceuticals), it is not clear to what extent solvent recycle will be permitted.

    Many pharmaceutical compounds are readily  soluble in water, but they  are
poorly soluble in even  polar  solvents such as  DMSO. Researchers at Bradford
Particle Design dealt with this situation in a CO^-based nonsolvent process by in-
corporating a cosolvent (an alcohol) that is miscible with both water and CO, [284],
Use of a coaxial nozzle and this cosolvent  allowed  Bradford Particle Design to
produce  fine particles from a variety of water-soluble compounds. Sievers and
colleagues [285] have dealt with this problem via use of colliding streams of aque-

-------
98
     ous solution and CO, (prior to exiting the high pressure environment at a nozzle).
     where the CO, helps form (and dry) an aerosol of the aqueous solution. These two
     processes are noted because they each accomplish the formation of small particles
     of valuable compound using entirely sustainable solvent systems—CO,/water/etha-
     nol by Bradford Particle Design and CO,/water by Sievers and colleagues. This
     mode of operation would seem to exhibit the highest green potential of the various
     CO,-based powder  processes.


     5.4   Comparisons With Current Processes

         The literature suggests that milling, crystallization, and spray drying currently are
     the most common means by which to generate powders (particles) from pharmaceu-
     ticals [282.286]. Milling [287] is a relatively energy-intensive process, but requires no
     solvent and is readily scalable. Milling (including jet pulverizing) has been demon-
     strated to  be able to create particles in the 1 to 5 micron range. The design and
     performance characteristics of various types of mills are known, and the process is
     readily scalable and can  be rendered continuous [288]. However, temperature in-
     creases during milling can damage labile compounds, and strict control over particle
     size and particle morphology may either be lacking or inconsistent. Milling can cre-
     ate substantial waste if the distribution of particle sizes exhibits a substantial tail at
     the lower end of the scale. Replacement of milling with a CO,-based process would
     seem to owe more to product concerns than to "green" concerns, if one of the various
     CO, processes can generate product consistently with the correct characteristics (size.
     distribution, shape, morphology).

        Spray drying [288,289] involves the atomization of a solution (product in sol-
     vent), the mixing of the droplets with a hot gas (usually air), followed by the drying
     of the droplets to form the particles. Particles can be produced whose sizes range
     from 2 microns up to 500 microns; theory on design and operation of spray dryers
     has been well studied. If water is being  employed as the solvent, then  the only
     significant "green" complaint that one might have with spray drying is that water's
     high heat of vaporization requires a significant energy input to the process. On the
     other hand, as in the case of milling, if the CO,-based process generates particles of
     higher quality (closer adherence to size and morphology constraints) at a competi-
     tive  price, then  the  CO, process could dominate despite potentially being  less
     green. Obviously, if one is spray drying from organic solution, then recycle of the
     solvent is an additional consideration.

        As for the cases of both milling and spray drying, crystallization is an often-
     used industrial process where numerous variations are possible [288, 290). Design
     principles for crystallizers have been investigated in depth in the past, and proce-
     dures for the design of crystallizers are readily available. If water is being used as
     the solvent, crystallization is already a relatively green process where perhaps
     high-energy input owing to the use of water as a solvent (recall the need to dry the

-------
                                                                               99
 product) or the need to treat the wastewater from the process could he seen as
 negatives. Again, however, crystallization may not be able to produce the particle
 characteristics desired by the end-users.

    In summary, the use of carbon dioxide as a nonsolvent for the production of
 particles (primarily pharmaceutical particles) is not substantially more "green" than
 competing technologies (in some cases it could be less green). However, the use of
 CO, could  provide better product, and its relatively green status provides no com-
 plications from a sustainability perspective. What seems to differentiate CO,-based
 processes from their conventional competitors (crystallization, spray drying, mill-
 ing) is a general lack of basic design equations that would allow ready creation of
 a design schematic, given product specific inputs (the usual situation in computer-
 aided design of a unit operation of process). Research by DeBenedetti's group
 during the  1990s  [291] suggested that the process by which panicles are created
 during spraying of a solution into CO, could be modeled by considering the forma-
 tion of fluid droplets and the transport of both CO, and solvent between the con-
 tinuous phase and the droplet phase. However, recent work by Randolph and col-
 leagues [292] suggests that true droplets never form in the spray process, and that
 particle formation can be described by gas-phase nucleation and growth within the
 expanding  plume. Whereas this may seem (to an outsider) as merely an academic
 debate, accurate models of the particle formation process inevitably  result jn the
 identification of the correct dimensionless groups associated with the phenomena
 and the underlying mathematical relationships that will ultimately permit process
design from first principles. Although there is general agreement that phase behav-
 ior (thermodynamics) and transport play roles in the effects of process conditions
on particle characteristics, it is not clear that a universal set of design guidelines
currently exists.

    In summary, what appears to be needed in this CO,-based sub-field is research on
building a true engineering model for such processes, where the input of fundamental
thermophysical parameters allows for the design and operation of equipment that can
deliver product with the desired characteristics.  Indeed,  the proliferation of acro-
nyms associated with CO,-based particle production (see [281]) lends the impression
that the various processes are in some way fundamentally different from one another,
and thus one  must experimentally evaluate each option (for a particular solute) to
determine the proper operating mode to produce a given particle size and distribu-
tion. The lack of a defined "unit operation" with acknowledged theoretical underpin-
ning makes it difficult to perform an engineering design and scale-up of such pro-
cesses, hindering their wider use. Equipment for CO,-based particle production is
rather treated as "custom."

    Another  avenue of research (in this area)  that has  received  relatively scant
attention in recent years is the use of CO, to process/produce well-defined particles
from pigments. It is known that pigment particle size (and extent of particle ag-
glomeration) exhibits a strong effect on the ultimate color of the article receiving

-------
100
     the pigment. Pigments usually are milled mechanically: the use of a CO,-based anti-
     solvent process could allow for the production of pigments with good control over
     the size and size distribution. Texter [293] has reviewed a number of solution-based
     methods (homogeneous and multiphase systems) for generating fine panicles from
     pigments—most seem to rely upon controlled precipitation of pigment from a pre-
     cursor solution (or emulsion) to form the particles. Naturally. CO, presents some
     advantages, as it can be readily separated from the organic solvent and it is itself
     benign. Whether such  advantages allow CO,-based processes to supplant tradi-
     tional milling (which obviously  uses no solvent) remains an open question, al-
     though preliminary results are promising [294].


     5.5  Industrial Activity

        There has been an interesting spate of industrial activity on particle formation
     using carbon dioxide over the past 3 years, much of it not expressly technology
     based. Bradford Particle Design (UK) helped  pioneer the development of the
     "SEDS" process (solution-enhanced dispersion by supercritical fluids), where etha-
     nol is added to an aqueous solution while it is sprayed into CO, to form particles. In
     early 2001. Inhale Therapeutics acquired Bradford Particle Design, demonstrating
     the interest  by the pharmaceutical community  in this technology. Interestingly,
     Bradford previously announced that Bristol-Myers-Squibb had licensed their tech-
     nology for use in pharmaceutical manufacture; it is  not clear as to the state of that
     alliance at this time. At nearly the same time (late 2000) as the Bradford acquisi-
     tion, Lavipharm (Greece) announced the acquisition of Separex (France) and the
     purchase of a 30 percent stake in Phasex (US). Both Separex  and Phasex are well
     known to the supercritical fluid community, having each worked on the fundamen-
     tals and design of numerous supercritical fluid processes.

        The review by Jung and Perrut lists many of the patents awarded on CO,-based
    processing for the generation of  fine particles.  In  addition to Bradford Particle
    Design [284], a number of academics  have patented aspects of the nonsol vent route
    to particle production, including Randolph [295] and Sievers [285]. at the Univer-
    sity of Colorado, and Subramaniam at the University of Kansas [296].

       Regarding the PGSS-type processes, many of the patents that have appeared
    are related to applications in the coatings industry, including the Unicarb Process
    (mentioned previously), and powder coatings applications from Ferro (mentioned
    previously) and Morton [297].

-------
                           Section  6

                      Process Issues
    One of the foremost reasons why more supercritical CO,-based processes have
not appeared over the past decade is that they are thought to be (and in many cases
are) "too expensive," owing primarily to the added cost required to design, con-
struct, and operate a high-pressure process. Although the discovery of "CO,-philes"
in the early 1990s was rightly hailed as a scientific/chemistry breakthrough, its
impact would most likely have been felt during process design and scale-up, as the
use of CCX-philic substrates and catalysts would permit lower operating pressures.
and would lower the cost of the process. Unfortunately, the high cost of fluorinated
CO,-phiIes more than negated their effect on lowering operating pressure. Conse-
quently, the recent work by Beckman on design of nonfluorous CO,-philes [137],
while nominally a "polymer science" issue, is actually an example of the use of
molecular design to lower process costs. Indeed, the work by Wallen  on specific
interactions between CO,  and carbonyls [140], by DeSimone and Ober on design
of polymers for photolithography with CO, developing [213, 214],  and by
Subramaniam on the use of expanded liquids for oxidations [110] are all examples of
the use of molecular design to try to lower process pressure. As such, chemistry
and engineering are inextricably intertwined when trying to optimize a CO,-based
process.


6.1   Process Design Using Supercritical Fluids:  Are  CO2-
      Based Plants Inherently Uneconomical?

    The number of processing plants operating worldwide that employ supercritical
CO, is slightly above 100 and growing steadily [298]. Most of the current plants
use CO, to process food in some way (extraction of fractionation), yet  other types
of plants obviously are being brought on stream (for example, fluoropolymer syn-
thesis by DuPont. hydrogenation by Thomas Swan, coatings by Union Carbide,
polyurethane processing by Grain Industries). Despite this steady growth, there is
a general sense (or unease) within both the academic and industrial communities

-------
102
     that there are elements connected to the design and construction of CO,-based
     plants that effectively block greater use of the technology.

        Several authors have reviewed various aspects of process design and costing of
     •'supercritical" plants [299]: these reviews typically focus on a specific industry.
     For example, Perrut  reports that for the case of extraction, the relative cost of a
     supercritical plant scales as (V*Q)I/J. where V is the column volume and Q the flow
     rate. This is consistent with what we report in Section 1.8. where minimizing equip-
     ment size and flow rate will minimize process cost.

        Each of the authors who has reviewed process design using supercritical CO, em-
     phasizes that one needs access to the relevant fundamental parameters to complete and
     optimize the design. Such parameters include both the relevant thermodynamic model
     for the mixture(s) in question with the appropriate binary interaction parameters, reac-
     tion data (rate constants, heats of reaction, Ahrrenius constants), and transport con-
     stants (densities, diffusivities. and viscosities). Note that these parameters are exactly
     the same as would be required to design a one atmosphere process, and there is nothing
     inherently  "foreign" about a CO,-based process that inhibits design and costing.  In-
     deed, high pressure alone is not sufficient to explain the perceived inhibition of CO,-
     based process scale-up, given that hydroformylation operates at 200-300 bar at large
     scale, but low density polyethylene is produced at more  than 2.000 bar. If one has
     access to the necessary basic information, one can employ software such as ASPEN to
     accomplish the process design, and ICARUS to handle the costing (the author has done
     so successfully with colleagues as part of consulting contracts).

        We must conclude that, if the inhibition in the scale up of CO,-based processes is
     real rather than perceived, then it must be due to a lack of the fundamental parameters
     needed for process design, plus other factors that would inhibit the commercializa-
     tion of any "new" technology. For example, as mentioned previously, it is difficult at
     present to predict the effect of molecular structure on phase behavior in CO, of mol-
     ecules that exhibit any substantial degree  of complexity. Carbon dioxide exhibits
     both nonpolar tendencies (low dielectric constant) and "polar" properties (Lewis
     acidity, strong quadrupole moment), and predictions of phase behavior are not straight-
     forward (as in the case of alkanes or alkenes). Recent work [300] has shown that the
     Statistical Associating Fluid Theory (SAFT) can provide good descriptions of the
     phase behavior of complex mixtures including CO,, yet the complexity of this model
     and/or lack of suitable parameters currently may limit its  use  industrially. Group
     contribution models have been applied to CO, solutions somewhat narrowly, gener-
     ally targeting a single class of solutes [301]. What appears to be needed is a means to
     easily predict the properties of mixtures involving CO,, such that confident predic-
     tions of process requirements  and costs can be made using conventional process
     software such as ASPEN. It also may be necessary for academics to conduct their
    own preliminary estimates of process cost for CO,-based alternatives to conventional
     processes to move such technologies beyond the discussion stage.

-------
                                                                              103
6.2   How Does One Economically Recover a Catalyst and/or
       a Product From CO2?

    The problems associated with product recovery through use ot a large pressure
drop have been discussed in a previous section: although depressurization precipi-
tates product efficiently, it is expensive (and energy-intensive) to continually com-
press and recompress CO, (assuming the process is continuous). As discussed pre-
viously, the coffee decaffeination process strips the extracted "product" (the caf-
feine) from CO, into water with only a small pressure drop across the liquid-liquid
extraction column. Beckman and Hancu employed the same process to recover
hydrogen peroxide generated in CO, from O, and H,.

    The fabric cleaning process developed by Micell [302] illustrates another means
by which to recycle CO, without employing a large pressure drop. The CO, is
employed as a pressurized liquid: following the cleaning cycle, the CO, is trans-
ferred to a vessel where a small pressure drop is used to create a vapor-liquid two-
phase system. CO, then is drawn from the top of the vessel and recondensed. leav-
ing behind a residue for disposal and allowing reuse of the CO,. Energy is captured
and reused as much as possible during the cycling of the CO,.

    Catalyst recycle is a more pressing need for supercritical fluid processes (ow-
ing to the custom design of CO,-philic ligands) than conventional analogues, but
also presents a more difficult problem.  Homogeneous catalysts are  designed to
provide enhanced selectivity and kinetic control of reactions, yet without effective
recycle their added cost prevents economical scale-up. Consequently, any green
advantages gained through use of CO, as a solvent are more than counteracted by
the green and economic disadvantages incurred by use of a homogeneous catalyst.
As such, investigations into means by which to recover homogeneous catalysts
from CO, play a vital role in enhancing the viability of green chemistry in CO,.

    For example, a collaboration between Tumas and the DeSimone group has in-
vestigated the design of metal catalysts that are tethered  to crosslinked.
polyfluoroacrylate polymer beads [303],  As noted earlier, fluoroacrylate polymers
are the most CO,-philic materials yet identified; because the crosslinked versions
employed by Tumas cannot dissolve  (they are, after all, crosslinked), they will swell
in the presence of CO, to 300 percent of the their initial volume. Because the metal-
ligand construct is tethered to the beads, the catalysts can be readily recovered after
the reaction and potentially reused. Crooks [304] also has tried to address the catalyst
recycle issue through design of dendrimer-supported metal catalysts; they have cre-
ated Pd nanoparticles within dendrimers and employed these to support hydrogena-
tion and other  reactions. The outer shell of the dendrimers can be decorated with
fluoroalkyl groups, and these macrocatalysts can be employed  in CO,. Finally,
Keurentjes and coworkers [305] recently have published a method where catalysts
are tethered to microporous inorganic supports for use in catalysis  in CO,.

-------
104
        The strategies employed by these three groups are extremely important, in that
     each has attempted to preserve the benefits of a homogeneous catalyst while co-
     opting the primary benefit of a heterogeneous catalyst—the ability to easily re-
     cover the valuable metal. For each case, some key issues remain to be discussed:
     Docs each "supported" catalyst preserve the activity and selectivity of the soluble
     parent? Are the reactions kinetically controlled or diffusionally limited? How fast
     does the metal "leach" from the supported catalysts?

        Eckert [ 16J. Tumas [306]. and others have examined the use of phase transitions
     to allow recycle of catalysts and other valuable components in a CO, process. Eckert
     has found that addition of CO, to a mixture of organic and fluorocarbon solvents
     induces mixing, but removal of the CO, (by depressurization) rapidly leads to com-
     plete phase separation. Consequently. CO, can be employed as a reversible and be-
     nign "trigger" to allow a catalytic reaction while ultimately allowing segregation of
     the catalyst following reaction. Tumas has examined the use of a "pressure trigger" to
     attempt to recover the  catalyst from a CO,-continuous emulsion. At elevated pres-
     sure, a water-in-CO, emulsion forms where the catalyst is localized in the aqueous
     micellar cores. Reduction of the pressure breaks the emulsion, leading to a distinct
     aqueous phase housing the catalyst (which then could be reused).


     6.3   Where Would Process Improvements Enhance
          Opportunities for Green Chemistry in CO2?

        As in the previous section, examples described here are not directly related to
     green chemistry, but solution of such problems would greatly enhance the viability
     of CO,-based processes, and are intimately tied to green chemistry in carbon diox-
     ide. For example,  there remains no efficient means by which to inject and remove
     granular solids from a high-pressure system (screw feeders have been tried with
     limited success). There clearly are a number of areas (food processing) where con-
     tinuous injection and removal of solids would greatly enhance the economic vi-
     ability of a CO,-based process, yet lack of the mechanical means by which to ac-
     complish this relegates the process to batch or semi-batch operation. Note that the
     chemical basis for continuous polyurethane foam production using liquid CO, as
     the blowing agent was established in the early 1960s, whereas  commercialization
     only occurred after development of the proper equipment in the early 1990s.

       Over the past decade, there has been significant academic and industrial interest
     in cleaning processes using CO,-cleaning of metal pans, electronics components,
     and fabrics. CO, is ideally suited to such applications owing to its low viscosity and
    environmentally benign nature, yet mechanical issues complicate application of CO,
    to these processes. For each of these applications, individual "pieces" must be rap-
    idly inserted into a high-pressure chamber, the chamber then is sealed and pressur-
    ized, the "piece" is cleaned, and the chamber is depressurized and emptied. In an
    atmospheric operation, the operation of such a situation is trivially simple and easy to

-------
                                                                                 105
scale (cost per part drops as chamber volume rises). The opposite currently is true for
high-pressure operation; scale-up is nontrivial and the cost of the system rises rapidly
as the size of the chamber rises. More efficient "piecework" operations at high pres-
sure will not only render cleaning operations less expensive, but also coating and
fabric dying operations. Finally, many proposed CCX-based processes (including spin
coating, lithography and developing, free meniscus coating) that are under examina-
tion in academic/industrial laboratories would benefit greatly from breakthroughs in
the  design of equipment designed to efficiently transfer parts in and out  of high-
pressure environments.

-------
                             Section  7

  Reactions  at Interfaces and/or Multiphase
                              Mixtures
    Reactions at interfaces (or transport across interfaces to facilitate reaction) in CO,-
based systems have been proposed as a useful means by which to support green chem-
istry in carbon dioxide while easing separation problems post-reaction. Indeed, if cata-
lyst, reactants, and products can be segregated effectively in various phases in the reac-
tor, downstream separation is certainly easier. However, one is now also faced with
thermodynamic (phase behavior) and transport limitations to reaction. Akey proviso in
attempting to use a biphasic system (with CO,) to perform green chemistry is that the
continuous component of each phase (CO, and the second component) should either
both be environmentally benign (and hence cross-contamination is irrelevant) or should
be immiscible over essentially the entire concentration regime. Thus, only the compo-
nents of interest (reactants, products) are moving across the phase boundary.

    Reactions making use of the CO,/water biphasic mixture have long been pro-
posed as green alternatives to conventional reactions. Each of these solvents is inher-
ently benign, they are immiscible over a broad range of concentrations, and the inevi-
table crosscontamination that occurs on phase contact does not require remediation.
Eckert and colleagues [307] first examined the use of a conventional phase-transfer
catalyst in a CO,/water mixture and found that despite the lack of"CO,-philic" ligands,
the tetra-alkyl ammonium bromide was effective at catalyzing the reaction across the
interface. Although Eckert employed a phase-transfer catalyst, Johnston and colleagues
(and later Tumas) enlarged the interfacial surface area through creation of an emul-
sion [308]. The enhanced surface area in the emulsion greatly enhanced the rate of
the model reactions  performed by these two groups (see Figure 7). Beckman  and
Hancu [34b] also examined the use of added surfactant to enhance reactivity  in a
CO,/water biphasic system. CO, dissolves in aqueous hydrogen peroxide, forming
percarbonate (through two distinct mechanisms). The percarbonate ion (basic condi-
tions are employed) then reacts with an alkene at the interface, forming the epoxide.
The addition of surfactant to this system substantially enhanced the reaction rate, as
did the addition of a phase-transfer catalyst. The usual caveat in CO,/water biphasic
                                   107

-------
108

             100
Q                   *"•    i
  •                   ,0.1
                                                                  i
                 jc
         *
                 i*
                 i    *
                0     100200300400500600700
     FIGURE  7  Time profile of formation of ethyl benzene from hydrogenation
                  of styrene performed in biphasic water/toluene (*), biphasic wa-
                  ter/CO, (•) and in emulsions using PFPE MW = 2500 (A), PFPE
                  MW ="740 (n), Lodyne 106A (o), or PBO-PEO (D) as surfac-
                  tants. Reaction conditions: 50/50 wt °/c water/CO.,, 1 .5% surfac-
                  tant, 80 mM styrene, 1 mol % catalyst (to substrate), Rh/L = 1/6,
                  40 °C,  4000 psi. TOP values at 50% conversion are given as a
                  comparison for biphasic H^O/toluene. H,O/CO,,  and H,O/CO,
                  emulsion systems [308].

    mixtures is that the low pH can cause problems for some reactions [309]. Quadir and
    colleagues [3 10] used the CO,/water biphasic system in an intriguing way; CO, was
    employed to alter the particle size distribution emanating from an emulsion polymer-
    ization in water.

        The recent intense scientific interest in ionic liquids has created another possible
    biphasic system for use with carbon dioxide. Ionic liquids are salts (to date, ammo-
    nium and phosphonium salts) that exhibit melting temperatures close to or below
    room temperature. These materials exhibit manageable viscosities and essentially
    negligible vapor pressures, and are considered potentially benign solvent media. In
    1 999, Brennecke  [311] observed that ionic liquids would absorb large quantities of
    CO, at relatively low pressure (mole fractions of- 0.6 at pressures below 100 bar),
    yet the amount of ionic liquid  dissolved in CO, was below the detection limit of the
    instrument employed (and thus below 10"5 mole fraction). As such, the phase behav-
    ior of an ionic liquid in equilibrium with CO, resembles that of a crosslinked polymer
    in equilibrium with CO, (see Figure 8). Further, like polymer-CO, mixtures, the ap-

-------
                                                                              109
            «•   ^
            a.
                 20-
            3
            OS
                            1 Phase
                     0      0.25    0.5     0,75      1
                            C02 mole fraotion

FIGURE 8  Phase behavior of carbon dioxide with the ionic liquid 1-butyl-
             3-methyl imidazolium hexafluorophosphate [311 ].
parent volume change resulting when mixing for an ionic liquid-CO, mixture is large
and negative, such that the volume change upon swelling of the ionic liquid is rather
small, despite the amount of CO, absorbed. Further, because CO, dissolves readily in
the ionic liquid, transport across the interface is rapid.

    A number of researchers have since exploited ionic liquid/CO, biphasic mixtures
as media for green chemistry. Tumas [83] employed CO, as a reactant in the forma-
tion of dimethyl formamide from amines, postulating that the ionic liquid would
stabilize the polar intermediate in the reaction. Both Cole-Hamilton [312] and Leitner
[313] conducted catalytic reactions in an ionic liquid, employing CO, to both extract
products (leaving the catalysts behind), and enhanced the solubility of gaseous reac-
tants in the ionic liquid phase. Jessop and Eckert [84] examined asymmetric hydro-
genation in an ionic liquid, again where the product is stripped into CO,, leaving the
catalyst behind. It would not be surprising to see other such efforts in the future. The
previously stated (see Section 2) caveats regarding ionic liquids naturally still apply.

    In theory, one also could conduct reactions across a CO,-soIid interface (other
than heterogeneous catalysis) and a CO,-organic liquid  interface, although little
work has been reported to date. The one notable example is the work by Eckert's
group [16], where a phase-transfer catalyst (PTC) is used to promote the displace-
ment reaction of benzyl chloride with solid potassium bromide (no reaction occurs

-------
110
     in the absence of the PTC). Brennecke  [314] found that a simple esterification
     reaction conducted in a biphasic CCX/organic mixture proceeded to a greater degree
     of conversion, possibly because the product partitioned preferentially to the upper.
     CO,-rich phase. To render any of these interfacial reactions practical, the thermody-
     namics of the system must be well understood. Clearly, the extent to which reac-
     tants. products, byproducts, and solvents partition between the phases will deter-
     mine the rate of reaction and the ability to recover both products and catalysts. In
     the case of ionic liquids, data and or models on the pVT and mixture behavior are
     entirely lacking, and partitioning behavior must still be determined experimentally.

-------
                          Section 8

 Impact of the Technology for a Sustainable
 Environment (TSE) Program on Use of CO2
                 as a "Green" Solvent
8.1   Description of Funded Projects in the TSE Program

•  Are there concentrations or foci among the various projects funded?

•  Are certain programmatic areas underrepresented?

   The titles of projects funded under the TSE program (those that involve super-
critical fluid technology) are shown in Table 3. We can further subdivide these
projects by their primary focus, as has been done in the body of this report. For
example, Grant Nos. R826115 (DeSimone) and R825338 (Russell) involve inves-
tigations of polymerizations in carbon dioxide and can be grouped together. In the
DeSimone work, the primary emphasis is on the design of highly CGvsoluble
amphiphiles for support of dispersion polymerization in carbon dioxide. This work
draws on previous successful demonstration by the DeSimone group that fluori-
nated polyacrylates are probably the most CO,-philic polymers yet discovered, and
that can be used to create COysoluble stabilizers [73, 125. 152]. The Russell pro-
gram focuses on  the use of enzymes to generate polyesters at low temperature
using activated substrates (divinyl esters). Supercritical fluids are employed in an
attempt to control the molecular weight and molecular weight distribution.

   The second group (Grant Nos. R826734 and R824731, Brennecke; 961355,
Cummings; 9985598, John) contains programs that focus on the measurement of
basic thermodynamic data/properties, information that would be needed to support
design of green materials or processes where CO, is involved. For example, the
program by Cummings attempts to use a combination of theoretical and experi-
mental techniques to design and evaluate surfactants to be employed in carbon
dioxide. The two programs directed by Brennecke each examine ways by which
                               111

-------
112
                 TABLE 3  Projects Funded by the TSE Program.
Principal
Investigate
Abraham
Akgerman
Beckman
Brennecke
Brennecke
Busch
Cummings
DeSimone
John
y
Russell
Sievers
Tanko
Grant No
R828206
R828135
R824730
R826734
R824731
9815321
9613555
R826115
9985598
R828129
R825338
R824728
9524986
Title
Development of a Heterogeneous
Catalyst forHydroformylation in
Supercritical CO,
Homogeneous Catalysis in
Supercritical Carbon Dioxide with
Fluoroacrylate Copolymer
Supported Catalysts
Design and Synthesis of CO,-
Soluble Affinity Ligands for Use in
CO, Extraction of Proteins
Multiphase Reactive Equilibria in
CO,-Based Systems
Phase Equilibria of CO,-Based
Reactions Systems
Catalytic Oxidations in Supercritical
Carbon Dioxide
Molecular-Based Study of
Reversed Micelles in Supercritical
CO, for Solvent Substitution in
the U.S. Chemical Industry
Nonionic Surfactants for Dispersion
Polymerization in CO,
Clathrate Hydrates in Water-in-CO,
Microemulsions
Water as a Solvent for Metal-
Mediated Carbon-Carbon Bond
Formations
Biocatalytic Polymer Synthesis in
and from Carbon Dioxide for
Pollution Prevention
Replacement of Organic Solvents
by Carbon Dioxide for Forming
Aerosols in Coatings Processes
Supercritical CO, and CHF, as
Alternative Solvents for Pollution
Prevention
Institution
University of
Toledo
Texas A&M
University
University of
Pittsburgh
University of
Notre Dame
University of
Notre Dame
University of
Kansas
University of
Tennessee
University of
North Carolina
Tulane
University
Tulane
University
University of
Pittsburgh
University of
Colorado
Virginia
Polytechnic
University

-------
                                                                                13
the often complex phase behavior involved in supercritical fluid reaction systems
might be better modeled (predicted). Finally, the program directed by John exam-
ines the thermodynamics behind CO, clathratc hydrate formation, specifically in
micellar systems.

    The programs by Tanko (9524986) and Li (R828129) examine chemical reactions
(or classes of reactions), where use of CO, as the primary solvent might render the
process inherently greener. Tanko's work focuses on free radical chemistry in CO,,
and that of Li examines C-C bond formation in a broad sense and examines reactions
in water. CO,, and ionic liquids.

     Another group of  grants proposes to examine the creation of homogeneous
catalysts for use in CO,—here fluorinated ligands are employed to render traditional
organometallic catalysts CO,-soluble. allowing the examination of a broad range of
reactions. Busch (9815321) proposes to examine oxidation chemistry, but Akgerman
(R828135) proposes to study hydrogenations and hydroformylations. The Akgerman
work adopts a slightly different tactic to create the CO,-soluble catalysts, where a
fluoroacrylate polymer is employed as the support for a number of metal species.

    Finally,  the remaining projects examine other facets of green chemistry and
engineering in CO,. The Abraham work (R828206) examines hydroformylation in
CO, using a heterogeneous catalyst. Sievers proposes to create small inorganic
particles of controlled size and composition by using CO, to create an aerosol from
an aqueous  solution. The Beckman  program (R824730) studied the  design and
application of highly CO,-soIubIe amphiphiles for use in extraction.

    Just as interesting as the nature of the projects supported by the TSE program are
the research areas that are, in a sense, missing. It is important to note that the TSE
program is small relative to the total  number of  U.S. research grants from various
agencies that support CO,-based green chemistry, and that it cannot be  expected to
provide support in all of the potentially interesting areas relevant to green chemistry
in CO,.  Nevertheless, there are  some important subcategories that are not repre-
sented in Table 2. For example, it is likely that the use of CO, in microelectronics
processing is going to receive increased research  scrutiny in the next 5 years by
industry, yet there are no TSE projects that address this application area. As shown in
Section 3, the groundbreaking papers have, in fact, already been published by re-
searchers such as Ober (Cornell), Watkins (University of Massachusetts), and
Wetmore-Gallagher (Phasex  Corporation).  Other  research areas that  are
underrepresented include the use of CO, in polymeric foams, CO, as a raw material,
and any number of topics involving process design and optimization. Regarding
fundamentals, it is somewhat strange  that there are no programs in Table 2 that ask
the question, "Why are CO,-philes CO,-philic?" or "Can we predict, using first prin-
ciples, the chemical structure of CO,-philic materials?" Or finally, "Can one adequately
predict the phase compositions in multiple-phase systems where CO, is present?"
These would seem to be useful areas for fundamental academic research.

-------
14

    8.2   Impact of the TSE Program

    •   Is the TSE program leading the field or following?

    •   Can results (publications and patents) from TSE-sponsored programs be con-
       sidered milestones in the field (green chemistry in supercritical fluids)?

       Because the nature of the work sponsored by the TSE program (the CO,-based
    programs) is spread over a number of different subcategories, the impact differs as
    one moves from one group of projects to the next. For example, the key publications
    regarding the design of "CO,-philic" materials were published before 1995: these
    included  the papers by the DeSimone group in Science [73, 125, 152], those by the
    Beckman group [71.123], and those by the Johnston group [72]. These papers served
    to establish that certain fluorinated materials were highly CO,-soluble, and that one
    should be able to design functional CO,-soluble materials from them. In this respect,
    the programs directed by DeSimone and Beckman can be said to follow up on earlier
   discoveries. Further, it could even be proposed that the groundbreaking papers men-
   tioned above were made possible by work on CO,-fluorocarbon  interactions pub-
   lished during the mid to late 1980s [ 123a, 315]. However, during the 1980s, conven-
   tional wisdom claimed that  CO, behaved like an alkane insofar as solvent strength
   was concerned, and changing that wisdom required a significant body of published
   work. The "CO,-is-like-hexane" heuristic likely survived well into the 1990s;  the
   DeSimone Science paper in  1992 [73] did as much as anything else to eliminate it. In
   summary, the DeSimone and Beckman programs could rightly be considered "follow-
   on" work to successful earlier discoveries, and one might propose that industry alone
   could have supported them. However, the chemical industry generally is conserva-
   tive, owing to a sizeable investment in currently operating facilities, and it is not likely
   that the DeSimone or Beckman programs could have been entirely industry sup-
   ported during the time period that the TSE program provided support.

      The same issues apply to the  programs focusing on the design of CO,-soluble
   homogeneous catalysts (Busch, Ackerman, Li). The groundbreaking work in this
   field was  performed by Leitner's group at Max-Planck (Muelheim) and the Tumas
   group at Los Alamos National  Laboratory from 1994 to 1996. These two groups
   showed that homogeneous catalysis in CO, could be conducted via redesign of the
   catalyst ligands to produce high CO,-solubility. As in the case of the DeSimone and
   Beckman work, the Busch, Li, and Akgerman programs then could be considered as
   follow-on work rather than  truly pathbreaking. However, one also could  question
   why, given that DeSimone's paper appeared in Science in  1992, did the design of
   fluorinated ligands to support  homogeneous catalysis in  CO, not appear in  the
   literature  until 1995-96? The answer may be that transfer of the basic knowledge
   surrounding design of CO,-philic compounds did not pass easily across disciplin-
   ary boundaries. On the other hand, it may owe as much to the fact that the chemis-
   try community as a whole had yet to embrace the use of homogeneous catalysts.

-------
                                                                               115
    The Tanko and Abraham programs focus on a particular chemistry (hydro-
formylation. tree radical reactions), and their relatively narrow focus lends to limit
their impact. Indeed, the early work by DeSimone on free radical polymerization in
CO, (pre-1995) showed clearly that free radical reactions were quite feasible in
CO,, provided that the reactants were themselves soluble. Whereas the description
of the Li program mentions CO, as a green solvent. Li's work has focused almost
entirely on use of water as a green solvent. Consequently. Li's impact in the litera-
ture has been significant, but not where use of CO, as a solvent is concerned.

    The two programs directed by Brennecke and that of Cummings are designed
to support the eventual scale up  and analysis of processes employing CO, as a
green solvent. If one is eventually to commercialize a  CO,-based process employ-
ing several reactants and products, knowledge of the phase behavior as a function
of conversion/time will be essential in optimizing the process design. Generally.
the modeling of the phase behavior occurs later in the design process than analysis
of the chemistry (i.e.. rates, byproducts), and we would not be expected to see the
impact of  such modeling work until a number of CO,-based processes are brought
forward to the pilot stage and beyond. In general, good models for phase behavior
are extremely important to process design, but fail to  attract the glamour inherent
to the design of a new catalyst system.

    It should be noted that one outcome of the Brennecke work is that one can
readily predict the existence of azeotropes, given some fundamental knowledge of
the phase  behavior of a binary mixture. As mentioned earlier, minimum boiling
azeotropes where CO, is one  of the components could be "next generation" green
solvents, depending on the identity  of the  second component. Such azeotropes
would maintain their composition during various unit operations while exhibiting a
lower vapor pressure than pure CO,, rendering their use somewhat more practical.
If the Brennecke work could be used to predict the existence of such azeotropes. it
could speed the discovery  of such next generation solvents.

    The other projects in Table 3  are somewhat outside the mainstream of green
chemistry  in CO, and have failed to attract widespread interest. The work by Russell.
for example, is noteworthy in the fact that a rapid polyesterification reaction was
demonstrated at low temperature (producing high molecular weight polymer), rather
than demonstrating a potential new application for use of CO, as a solvent.

    In summary, the most significant observation of the projects in Table 2 is that
most could rightly be considered follow-on work to (in many cases) the investiga-
tors' own earlier groundbreaking results. Given that most new science is incremental
in nature, this  is not entirely surprising. Further, one of the more important aspects of
the TSE program is the education of the next generation of scientists and engineers to
appreciate  the use of CO, as a green solvent—use of CO, as a solvent is fast becom-
ing "mature." an achievement due  in large part to sustained funding in the area from
EPA and National Science Foundation.

-------
16

    8.3   Technology Transfer From TSE-Sponsored Programs

    •   Are results from  TSE-sponsored projects being transferred to the industrial
       sector.

    •   Which chemistries/processes currently are being investigated at the pilot scale?
       Full scale?

    •   Have new companies been created to commercialize results derived from TSE-
       sponsored programs?

       The largest industrial applications for liquid/supercritical CO, are presently ter-
    tian,' oil recovery, foaming of polymers (polyurethanes and thermoplastics), and food
    processing (extraction of coffee, tea. hops, and other natural products). In addition.
    several  new applications are in  various stages of development: pasteurization  of
    orange juice, fabric dry cleaning, and polymerization of fluoromonomers. The latter
    two applications were transferred from academia to industry, yet these pre-date the
    TSE program.

       There are a number of reasons why more CO,-based processes are not transferred
    from academia to industry, with or without TSE funding. For example, although techni-
   cally highly successful, fluorinated CO,-philes have not found industrial application
   simply because they are very expensive. Of the 13 projects shown in Table 2.6 (Beckman.
   DeSimone. Busch. Akgerman. Li. and Cummings) rely  heavily on the use of fluori-
   nated CO,-philes. Although each of these projects has been technically successful and
   added to the overall knowledge base on CO, technology, they have not been transferred
   to industry, and are not likely to be transferred, owing to the cost of the CO,-phile (and/
   or lack of a viable means by which to recycle the CO,-phile).

      Indeed, in cases where homogeneous catalysis in CO, is  under consideration,
   effective recycle/recovery of the catalyst is a primary hurdle preventing transfer  to
   industry, even if the ligands are nonfluorinated. The projects by John (clathrate for-
   mation in micelles) and Akgerman (use of polymeric supports for homogeneous ca-
   talysis) could be considered to incorporate aspects of  catalyst recovery, but these
   techniques currently are not employed by industry.

      Regarding the other projects, some (Brennecke's two projects) involve prima-
   rily basic science that would be  used to support scale-up,  while others (Abraham.
   Tanko) have focused on chemistries that are relatively narrow in  scope and scale-up
   is dependent on economic comparisons between a CO,-based  process and conven-
   tional processes. For  the  case  of the program by Russell,  although the enzymatic
   route to  formation of an aliphatic polyester was very successful, the use of the acti-
   vated diester renders the process too expensive for further consideration by industry.

      The  Sievers  project  focused on the creation of small inorganic particles via
   formation  of an aerosol from  water using CO, as the  "propellant." Although the

-------
                                                                                117
work described in the TSE abstract has not been transferred to industry, Sievers
has commercialized a related spin-off—the formation of small drug particles (inhalable
Pharmaceuticals) via formation of an aerosol from water using CO, as propellant.

    In summary, if perhaps little of what TSE has funded in the CO, arena has been
directly transferred to industry at this point, it has kept CO, technology on industry's
radar screen—if some of the technical hurdles mentioned in this report can be over-
come, it is likely that industry will embrace the technology. Thanks to funding from
TSE and other such programs, CO, is no longer considered "exotic" technology by
industry. This is perhaps the most significant accomplishment of the program. Fi-
nally, the TSE program has supported the training of a number of scientists and engi-
neers in CO,-based technology; the movement of these younger scientists into indus-
try represents a very effective means of technology transfer.

    A final interesting aspect of the use of CO, as a process  solvent is that a large
number of large chemical companies in the United States and Europe have experi-
mented with CO, technology, often attempting to target CO, at narrow problems within
their respective businesses. Historically, this has been a somewhat "hit-or-miss" ap-
proach, and companies will form task forces to explore CO, technologies, eliminate
them after short periods if a promising application does not present itself, then re-form
them  (often years later)  if another opportunity seems to present itself. A company
such as Air Products and  Chemicals has had several distinct flirtations with CO, over
the years, some resulting in spin-off companies. There have been several occasions
where a company will explore a CO,-based technology, then cancel the project while
patenting and/or presenting the work publicly. The topic is subsequently picked up
by an academic group and broadened, then reintroduced to industry through the usual
channels of publications and scientific meetings. Consequently, the technology trans-
fer proceeds via an industry-to-academia-to-industry route. For example, some of the
earliest work  on the solubility of fluorinated polymers in CO, was conducted by the
Phasex Corporation (itself working for other industrial clients). This work ultimately
filtered into academia, which greatly broadened the scope and advanced the science.
Subsequently, the use of fluorinated CO,-philes has completely  permeated both
academia and industry.

-------
                            Section 9

  Milestones in Green Chemistry Using CO,
    Designating particular achievements as milestones is, of course, subjective.
There are several types of milestones that one can consider with regard to green
chemistry in carbon dioxide—purely scientific milestones, milestones in the dis-
semination of information on use of CO,, and milestones in commercialization.
Perhaps the first true "green" application of CO, was the coffee decaffeination
process scaled up during the 1980s; this is a milestone as it showed that one could
successfully scale a CO,-based process and operate such a process economically
when given a good design.

    For example, in the 1980s, conventional wisdom claimed that CO,"s solvent
power resembled that of n-alkanes, despite a large body of experimental evidence to
the contrary. During the period 1988-1992, a number of research groups(Smith [261],
Johnston [72], Enick [123a] & Brady, Beckman [ 123b]) reported that fluorinated ma-
terials, as well as silicones, exhibited significantly better thermodynamic compatibil-
ity with CO, than alkanes. The paper in Science by the DeSimone group [73] on the
CO,-phiIicity of poly (per-fluoroacrylates) in 1992 was a milestone both from the
scientific standpoint and from a dissemination perspective, as this publication served
to quash  the "CO,-is-Iike hexane" heuristic. Interestingly, it was another 3 years
before the information of the CO,-philicity of fluorinated materials found its way into
the synthetic organic chemistry community. With publications by Leitner's group
[241 ] and Tumas' group [81] showing the use of fluorinated ligands in homogeneous
catalysis in  CO,, green chemistry in CO, began to rapidly permeate the chemistry
community. Once it  was demonstrated that effectively any catalyst could ultimately
be rendered CO,-soluble, CO, was applied broadly as a solvent in organic transfor-
mations by both the  academic and industrial communities. In 1999, Brennecke [85]
published a study demonstrating the potential for use of ionic liquid/CO, biphasic
mixtures as  media for green chemistry—the first papers exploiting this biphasic sys-
tem appeared in 2001.

    Beckman published the first reports claiming that a nonfluorinated CO,-philic
polymer could be designed in 2000. This work was based on earlier fundamental
                                  119

-------
120
     studies on CO,-polymer phase behavior conducted by McHugh [133]. Johnston
     [68]. and Eckert [ 131 ]. It is too early to say whether such technology will ultimately
     permeate the community and render design of CO,-philic materials more economi-
     cally viable.

        A  number of researchers  examined the strong potential for CO, to plasticize
     polymers, with several important papers appearing between 1985 and 1994. Ex-
     ploitation of this science appeared in 1996 through 2001, as both industry (Ferro
     [ 195]. PPG [ 196]) and academia (Howdle [88], Eckert [ 188b]) employed the plasticiz-
     ing effect to enhance mixing in polymer systems.

        Regarding commercial successes, the introduction of the CarDio process for
     continuous production of polyurethane foam using CO, as the blowing agent has
     been extremely important, in that it is both green chemistry and commercially suc-
     cessful. However, because CarDio was conducted entirely by industry with no R&D
     support from academia, it is little known within academic circles. Much more widely
     known is the construction (by DuPont) of a semi-works facility to polymerize fluori-
     nated monomers in carbon dioxide, as this technology was transferred (in part) from
     academia (work by DeSimone's group at North Carolina). The same probably is true
     for the cleaning of fabrics (dry cleaning) using CO,. The recent development of a
    large-scale process to pasteurize orange juice using CO, also is an important step
    forward in green processing using CO,, but like CarDio, process development is
    occurring entirely within industry and the project has not received wide notice.

       The introduction of CO, to microelectronics processing began with prelimi-
    nary work  by  the Phasex Corporation and IBM in 1995 and 1996, given the
    DeSimone Science paper showing that perfluoroacrylate polymers are readily mis-
    cible with CO,. Again, because the preliminary work was conducted primarily by
    industry and was disseminated to a relatively narrow audience (the microelectron-
    ics industry), extensive interest in this topic did not begin until several years later,
    when both Obers group (Cornell) and the DeSimone group (University of North
    Carolina) began to play active  roles. Now, the use of CO, in microelectronics pro-
    cessing is considered sufficiently noteworthy to merit an article in Chemical &
    Engineering News. The work by Watkins on creation of thin metal films via chem-
    istry in  CO, [89] will likely enhance interest further.

       Another series of commercial milestones occurred in late  2000/early 2001.
    when the pharmaceutical industry purchased (either in their entirety or substantial
    portions) Bradford Particle Design, Separex, and Phasex—three of the more sig-
    nificant commercial  enterprises relying primarily on supercritical fluids technol-
    ogy. It will be interesting to see whether this leads to more rapid commercialization
    of CO,-based processes or the  reverse.

       In summary, milestones in green chemistry using CO, have occurred on scientific
    achievement, as was the case with the discovery of CO,-philic polymers by DeSimone
    in 1992. and also the dissemination of fundamental science to industries or commu-

-------
                                                                                121
nities for whom CO, had previously been considered an exotic technology. In this
report, a number of technical hurdles to increased use of CO, in green chemistry have
been outlined. It is hoped that future milestones will occur by overcoming these
hurdles. Finally, it should be noted that some scientific milestones that have occurred
in this field might be considered the result of a particular researcher recognizing the
broader implications of a narrowly focused study published previously. Although
some of the projects supported by the TSE program could be said to be somewhat
narrowly focused, it may be too early to judge whether they could lead to break-
throughs. Unfortunately, breakthroughs cannot be readily designed a priori.

-------
                           Section 10

        Areas for Future  Research on CO2
                          Technology
    In each of the previous sections, mention has been made of potentially useful
avenues for future research; these are summarized below (in no particular order).

•   The use of biphasic systems (including carbon dioxide as one component) for
    conducting reactions using gaseous components.)

•   A greater focus on oxidations and hydroformylations versus hydrogenation in
    CO,; the former reactions generate more waste and require more stringent con-
    ditions than hydrogenation, yet have received relatively less attention in the
    literature (with respect to the sub-field of reactions in CO,).

•   Group contribution or. better yet. first principles models for the prediction of
    phase behavior in multiphase, multicomponent systems where carbon dioxide
    is one of the components. Prediction of basic transport properties is needed as
    well.

•   A fundamental understanding of the effects of chemical structure and topology
    on the phase behavior of molecules in carbon  dioxide; this should, therefore,
    result in the design of "CO,-philes" that do not include fluorine. This is per-
    haps a subset of the previous bullet, but no less important.

•   An understanding of the fundamentals behind solvation of hydrophilic com-
    pounds (including water) in CO,-based emulsions; also, thermodynamics and
    transport properties of the CO,-water interface are important. This would ad-
    dress the frustrating observation that not all CO,-soluble amphiphiles can solu-
    bilize water.

•   The design of equipment that would allow rapid injection and removal of solids
    from high pressure. CO,-rich environments. Also, the design of systems for the
    rapid high-pressure treatment of solid articles (as in the development of silicon
    wafers) or the continuous coating of material using a CO,-based solution.
                                  123

-------
124

     •   The use of CO, in microelectronics processing.

     •   An indepth understanding of the mechanism for generation of CO and subse-
        quent poisoning of noble metal catalysts in the presence of hydrogen and CO,.

     •   The design of catalysts for the generation of polyesters and commodity chemi-
        cals (aromatic acids) from CO,; activation of CO, at low pressures.

     •   The identification of azeotropes that include CO,. Also, it would be useful to
        explore the use of co-solvents for CO, in a more systematic manner and to find
        mixtures that are technically, environmentally, and economically successful.

    •   The design of additives that would allow greater use of CO, in the extrusion
        foaming of polymers. Also, the generation of low density, fine-celled foams
        using CO, as the blowing agent.

    •   The development of a set of fundamental design principles for the formation of
        particles via phase separation from mixtures that include CO, (under flow in a
        known geometry).

    •   Programs that focus on overcoming the various technical hurdles to the use of
        CO, in coating processes. For example, although problems in using CO, to
        process powder coating formulations differ greatly from problems encountered
        in preparing emulsion coating formulations using CO,, the problems are inher-
        ently technical in nature.

        With respect to the TSE program, in the past, the program has solicited propos-
    als from academia in  the general area of green chemistry, giving investigators com-
    plete freedom regarding the focus of the proposals. It also might be worthwhile for
    the  program  directors to (not in place of) set some green chemistry targets for
    investigators to attack. For example, in the recent  past the Monsanto Company
    identified environmental problems  of paramount importance (to Monsanto,  of
    course), and then asked for proposals from  the scientific community at-large re-
    garding potential solutions. The winning investigators received $500,000 to  $1
    million  to generate a prototype of their solution. Innovention has created a Web
    site  that lists synthetic problems of primary importance to this  company and  its
    clients, with "rewards" listed for successful solutions contributed from the outside.
    Given that the staff at the U.S. EPA is likely  more familiar with pressing environ-
    mental problems than typical academic chemists and engineers, it also might  be
    useful for TSE staff to publish a list of current high-profile effluent problems where
    innovative green chemistry solutions might contribute to broad industry or societal
    benefit.  TSE then could fund some proposals that attack the priority problems, and
   some that are more curiosity-driven or more fundamental. However, it should  be
   stressed that moving from the current curiosity-driven system to one that is entirely
   "target-specific" is not recommended, in that much of the truly innovative discov-
   eries in green chemistry and processing might be lost.

-------
                          Section  11

                          References
I.     McHugh. M.A.. Krukonis. V.J. 1994. Supercritical Fluid Extraction. 2nd
      ed. Butterworth-Heinemann  Boston. MA; (b) Brunner. G Gas extraction.
      Darmstadt (FRG): Steinkopff Verlag.

2.     Reid. R.C..  Prausnitz, J.M..  Poling. B.E. 1987. The Properties of Liquids
      and Gases. 4th ed. New York: McGraw-Hill.

3.     Giddings, J.C..  Myers, M.N.. McLaren, L..  Keller. R.A. 1969. Science
      162:67: (b) Giddings, J.C.. Myers, M.N.. King. J.W. J. Chromatogr. Sci.
      7:276.

4.     Siskin, M.. Katrizky, A.R. 2001.  Chem. Rev.  101:825: (b) Katritky. A.R..
      Nichols, D.A., Siskin, M., Murugan, R., Subramaniam, B.M. 2001.  Chem.
      Rev. 101:837; (c) Savage, P.  1999. Chem Rev.  99:651.

5.     Michels, A.. Michels, C. 1936. Proc. Roy. Soc. (London) A. 153:201; (b)
      Michels, A.. Michels, C.. Wouters. H. 1936. Proc. Roy. Soc. (London) A.
      153:214.

6.     Angus, S., Armstrong, B., de Reuck  K.M. 1976. (on basis of surveys and
      equations produced by V.V. Altunin) Carbon Dioxide. Oxford, New York.

7.     Praxair Material Data Safety Sheet, P-4574-H. May 1999: (b) CRC Hand-
      book of Chemistry and Physics, 76th ed. D.R. Lide (ed). 1996. CRC - Boca
      Raton, FL. Chapter 16.

8.     Minder. B., Mallat. T., Baiker, A. 1996. 3rd International Symposium on
      High-Pressure Chemical Engineering. Zurich.  139.

9.     Solymosi, F., Erdohelyi. E.. Lancz, M. 1985. J. Catal 95:567. Erdohelyi,
      A., Pasztor. M.. Solymosi, F.  1986. J. Catal. 98:106.
                                 125

-------
126
     10.    Lake. L.W. 1989. Enhanced Oil Recovery: Englewood Cliffs. NJ:  Prentice
           Hall; (b) Chakma. A.. Islam. M.R.. Berruti. F. (eds.). Enhanced Oil Recov-
           ery.  1991. Am. Inst. Chem. Eng. Symp. Set: 280. New York.

     11.    Pande. J.O.. Tonheim. J. 2001.  Proc. Safety Pmgr. 20:37.

     12.    Van  Bramer. D.J.. Shiflett, M.B.. Yokozeki. A.  1994. US Patent No.
           5.345.013.

     13.    Rodriguez. F.  1996. Principles of Polymer Systems.  4th ed. New York:
           McGraw-Hill.

     14.    Weissermel. K.. Arpe. H-J. 1997. Industrial Organic Chemistry. 3rd ed.
           Weinheim (FRG): VCH-Wiley.

     15.    Hancu. D.. Beckman. EJ.  2001. Green Chem. 3:80.

     16.   Eckert. C.A.. Liotta. C.L.. Gulp. C.W., Lamb D.R. In: Jessop. P.G. Leitner.
          W. (eds.). 1999. Chemical Synthesis Using Supercritical Fluids. Weinheim.
          GER: Wiley VCH. p 446.

     17.   Zosel. K. US Patent No. 3.806.619. April 23,1974; (b) Prasad. R.. Gottesman.
          M.. Scarella, R.A.  US Patent No. 4.246.291. January 20, 1981.

     18.   Kendall. J.L.. Canelas. D.A.. Young. J.L.. DeSimone. J.M.  1999. Chem.
          Rev. 99:543.

     19.   Tanko. J.M.. Blackert. J.F. 1994. Science 263:203: Hadida. S.. Super. M.S..
          Beckman, E.J., Curran, D.R 1997. J. Am. Chem.  Soc.  119:7406.

    20.    Tsang, C.Y., Street. W.B. 1981.  Chem. Eng. Sci. 36:993.

    21.    Hsiao. Y-L., DeSimone. J.M. \997.J.Polym.Sci. PartA: Polym.Chem. 35:
          2009.

    22.    Eckert.  C.A. Presented at the  International Symposium on Supercritical
          Fluid Technology. Myrtle Beach, SC, August 18-23,2001.

    23.    Welty, J.R., Wicks. C.E.. Wilson, R.E.  1984. Fundamentals of Momentum,
          Heat, and Mass Transfer. 3rd ed. New York:  John Wiley & Sons.

    24.    Inoue. S., Yamazaki. N. 1982. Organic and Bioorganic Chemistry of Car-
          bon Dioxide. New York: John Wiley & Sons; (b) West, K.N.. Wheeler. C.,
          McCarney, J.P.. Griffith. K.N., Bush. D., Liotta, C.L.. Eckert, C.A. 2001. J.
          Phys. Chem. A 105:3947.

    25.    Mesiano. A.J.. Beckman, E.J., Russell, A.J. 1999. Chem Rev 623.

    26.    Wu. Y.  US Patent No 4,483.888. November 20.1984.

-------
                                                                              127
27.   Arunajatesan. V.. Subramaniam. B.. Hutchenson. K.W.. Herkes. F.E. 2001.
      Chem. Ent>. Sci. 56:1363.

28.   Hutchensen. K.W.. Herkes. F.E.. Walls. D.J., Das. T.K.. Brennecke. J.R 2001.
      Presentation at  the ACS Annual Meeting.

29.   Arunajatesan. V.. Subramaniam. B.. Hutchensen. K.W.. Herkes. F.E. 2001.
      Presentation 176d at the AIChE Annual Meeting. Reno. NV.

30.   Jessop. P.G. Hsiao. Y.. Ikariya, T.. Noyori. R.  1996. J. Am.  Chem. Soc.
      118:344.

31.   Wiebe. R.. Caddy. V.L.  1940. J. Am. Chem. Soc. 815.

32.   Holmes. J.D.. Ziegler. K.J.. Audriani, M., Lee. C.T.. Bhargava. P.A.. Steytler.
      D.C.. Johnston, K.P.. 1999. J. Phys. Chem. B 103:5703.

33.   Enick,R.M.,Beckman.E.J.,Shi.CM.,Xu,J.H. Chordia.L.  2001. Energy &
      Fuels 15:256.

34.   Nolen. S.A.. Liu, J.. Brown. J.S.. Pollet, P., Eason. B.C.. Griffith. K.N.. Glaser.
      R., B.h. D.. Lamb, D.R.. Liotta, C.L., Eckert. C.A.. Thiele. GF, Bartels. K.A.
      2002.  Ind.Eng. Chem. 41:316: (b) Hancu. D., Beckman. E.J.  2002. Ace.
      Chem. Res. 35:757.

35.   Anastas. P.T.. Warner, J.C.  1998. Green Chemistry. Theory and Practice.
      New York: Oxford University Press, p 30.

36.   Eckert. C.A.. Bush. D.. Brown, J.S., Liotta, C.L. 2000.  Ind. Eng. Chem. Res.
      39:4615; (b) Lucien. P.P.. Foster, N.R. 2000.  / Supercrit. Fl. 17:111.

37.    Chinsoo. L. Hoy. K.L., Donohue, M.D. US Patent No. 4,923.720. May 8,1990;
      Chinsoo. L.. Hoy, K.L., Donohue, M.D. US Patent No. 5,027,742. July 2,1991; (c)
      Hoy. K.L.. Nielsen. K.A.. Chinsoo, L. US Patent No. 5.108,799. April 28,1992.

38.   Suh, K.W. 1991. Polvstyrene and Structural Foam in Pol\meric Foams. D.
      Klempner. K.C. Frisch (eds.). New York: Oxford University Press, p 151.

39.    Khoury, P.M. 1995. Predicting the Performance of Multi-Stage Separation
      Processes. 2nd ed. Boca Raton, FL: CRC Press.

40.    Chiang. S-H..Klinzing,GE.  US Patent No. 4.613,429. September 23,1986.

41.    Charpentier. P.A.. DeSimone.  J.M., Roberts. GW. 2000. Chem. Eng. Sci.
      55:5341; (b) Charpentier. P.A.. Kennedy, K.A.. DeSimone, J.M., Roberts. GW.
      1999. Macromolecules  32:5973.

42.    Perfluorooctyl sulfonates; proposed Significant New Use Rule.  2000. Fed.
      tfeg.65(202):62319.

-------
128
     43.    www.greenearthcleaning.com: (h) Berndt. D.R.. Griffiss. J.M. US Patent No.
           5.942.007. August 24. 1999.

     44.    Feder, B.J. New York Times February 15.2000. Section C. p 1.

     45.    www.chart-ind.com/news.html.

     46.    Jurelier. S.H.. Kerschner. J.L.. Murphy. D.S. 2(XX). US Patent No. 6.148.644: (h)
           Jureller. S.H.. Kerschner. J.L.. Harris. R. 1997.  US Patent No. 5.683.473: Jurelier.
           S.H.. Kerschner. J.L.. Bae-Lee. M.. Del Pizzo. L.. Harris. R.. Resch. C.. Waja. C.
           1996. PCTInt. Appl. WO 9627704.

     47.    US EPA. Office of Environmental Information. EPA 260-B-01-001. March 2001.

     48.    Air Toxics Web Site-CAA: original list of hazardous air pollutants: http//
           www.epa.gov/ttnatw01/origl89.html.

     49.    http://www.clean.rti.org.

     50.    Rylander, P.N.  1985. Hydmgenation Methods. Orlando. FL: Academic Press.

     51.    Subramaniam. B. 2001. Appl. Catal. A: Gen. 212. 199; (b) Subramaniam. B..
           Arunajatesan. V.  Lyon, C.J.  1999.  Stud. Surf.  Sci. Catal.  126:63: (c)
           Subramaniam. B., Ashraf. J. In: Innovations in Supercritical Fluids. 1995.
          ACS Symposium Ser. 608:246; (d) Ginosar. D.M.. Subramaniam. B. 1994. Stud.
          Surf. Sci. Catal. 88:327; Ginosar, D.M., Subramaniam. B. 1995. J. Catal.  152:31:
          Baptiste-Nguyen. S.. Subramaniam. B. 1992. AlChEJ. 38:1027.

    52.    Subramaniam, B.. Lyon, C.J.. Arunajatesan. V.  2002. Appl. Catal. B: Envir.
          37:279.

    53.    Satterfield, C.N. 1980. Heterogeneous Catalysis in Practice.   New York:
          McGraw-Hill.

    54.    Baiker, A. 1999. Chem. Rev. 99:453.

    55.    van den Hark. S.. Harrod. M.. Moller. P. 1999. / Am. Oil Chem. Soc. 76:1363:
          (b) van den Hark, S.Harrod,M. 2001. Appl. Cat. A-Gen 210:207.

    56.    Tacke. T., Wieland. S., Panster. P. 1996. 3rd International Conference on High
          Pressure Chemical Engineering, Zurich, pp 17-21.

    57.    Macher, M.B.. Holmquist, A. 2001. Ear. J. Lip. Sci. Tech.  103:81.

    58.    King, J.W.. Holliday. R.L.. List. GR.. Snyder. J.M. 2001. J. Am. Oil Chem. Soc.
          78:107; (b) Andersson. M.B.O., King, J.W., Blomberg. L.G 2000.  Green
          Chem. 2:230.

-------
                                                                              129
59.   Chouchi. D.. Gourgouillon. D.. Courel. M. Vital. J.. da Pome. M.N. 2001. Ind.
      Eng. Client. Res. 40: 2551.

60.   Condo. P.O.. Paul. D.R.. Johnston. K.P. 1994. Macmnwleciilex 27:365: (b)
      Condo. P.O.. Johnston. K.P. 1994. J. Polym. ScL. Pail B. Polym. Pliys. 32:523.

61.   Bertucco.A..Canu. P.. Devetta. L..Zwahlen. A.G  1997. Ind Eng. Client.
      Res. 36:2626.

62.   Devetta. L.. Giovanzana. A.. Canu. P., Bertucco. A.. Minder. B.J. 1999.  Catal.
      Today 48:337.

63.   Bhanage.B.M..Ikushima.Y..Shirai.M..Arai.M. 1999. Catal. Lett. 62:175.

64.   Minder. B.. Mallat. T. Pickel. K.H.. Steiner. K., Baiker. A.  1995. Catal. Lett.
      34:1.

65.   Hitzler. M.G. Smail. F.R.. Ross. S.K., Poliakoff. M. 1998. Org. Proc. Res. De\.
      2:137.

66.   Hancu. D.. Beckman. E.J. 1999.  Industr. Eng. Chem. Res. 38:2824; (b) Hancu.
      D.. Beckman. E.J. 1999.  Industr.  Eng. Chem. Res. 38:2833: (c) Hancu. D..
      Beckman. E.J. 2000. Ind.& Eng. Chem. Res. 39:2843.

67.   Gelbein. A.P. 1998. CHEMTECH 28:1.

68.   O'Neill. M.L.. Cao. Q.. Fang. M.. Johnston. K.P. Wilkinson. S.P.. Smith.
      CD., Kerschner. J.L.. Jureller. S.H. 1998.  Ind. Eng. Chem. Res. 37:3067.

69.   Consani. K.A.. Smith. R.D. 1990. J. Supercrit. Fl 3:51.

70.   Klingler. R.J.. Rathke.. J.W.  1994. J. Am.  Chem. Soc. 116:4772: (b) Rathke.
      J.W.. Klingler. R.J.. 1993. US Patent No. 5,198,589; (c) Rathke. J.W., Klingler,
      R.J.. Krause.T.R. \99\.Organometallics  10:1350; (d) Warzinski,R.P.,Lee.
      C-H., Holder. GD.  1992. J. Supercrit.  Fluids 5:60.

71.   Fink. R.. Hancu. D.. Valentine. R.. Beckman. E.J. 1999. J. Phys.  Chem B
      103:6441.

72.   Harrison. K.. Goveas. J.. Johnston. K.P., O'Rear, E.A.  1994.  Langmuir
      10:3536.

73.   DeSimone. J.M.. Guan. Z.. Elsbernd. C.S.  1992. Science 257:945.

74.   Palo. D.R.. Erkey. C. 2000. Organomet. 19:81.

75.   Smith, D.C.. Stevens.  E.D.. Nolan. S.P. 1999. Inorg. Chem. 5277.

76.   2000. Chem. Eng. News 78(21 ):9:2001. Chem. Eng. News 79(5):7.

-------
130
     77.    Jessop. P.G.. DcHaai. S.. Wynne. D.C. 2(XX). Client. Commun. 693.

     78.    Wendc.M..Mcier.R..Gladysz.J.A. 2(X)I. /,4m. Client. Soc. 123:1 1490.

     79.    Jessop. P.G. Olmstead. M.M.. Ahlan. C. Grahcnaucr, M. Sheppard. D.. Eckert.
           C.A.. Liotta.C.L.. 2(X)2. Inoi^. Client. 41:3463: (h) Jessop. P.G. Brown. R.A..
           Pol let. P.McKoon. E.. Ngo.T..Eckert.C.A.. Liotta.C.L. In: Jessop. P.G:
           Brown. R. A.: Pellet.  P.: McKoon. E.: Ngo. T:  Eckert. C. A.: Liotta. C. L.
           (eds.). 2(X)I. The Use of Neoteric Solvents for Hydrogenation anil Other
           Asymmetric Reactions. Boulder. CO.

     80.    Xiao. J.. NefTvins. S.C. A.. Jessop. P.G. Ikariya. T.. Noyori. R.  1996. Tetrahed.
           Len. 37:2813.

     81.    Burk. M.J.. Feng. S.. Gross. M.F.. Tumas. W. 1995.  J. Am. Chem. Soc. 117:8277.

     82.    Kainz. S..  Brinkman. A.. Leitner. W.. Pfaltz. A.  1999. J. Am. Chem. Soc.
           121:6421.

     83.    Liu. F.. Abrams. M.B.. Baker. R. T.. Tumas. W. 2001. Chem. Commun. 433.

     84.    Brown. R.A., Pollet. P.. McKoon, E.. Eckert. C.A.. Liotta. C.L.. Jessop. P.G.
           2001. J.Am. Chem. Soc. 123:1254.

     85.    Blanchard. L.A.. Hancu. D.. Beckman. E.J., Brennecke. J.F. 1999. Nature
           399:28.

     86.    Watkins. J.J.. McCarthy. T.J. 1995.  Chem. Mat. 7:1991.

     87.    Nazem. N.. Taylor. L.T.. Rubira. A.F. 2002. J. Supercrit. Fl. 23:43.

    88.    Morley. K.S.. Marr. PC.. Webb. P.B., Berry. A.R., Allison. F.J.. Moldovan. G.
           Brown. P.O.. Howdle. S.M.  2002. J. Mat. Chem. 12:1898.

    89.    Blackburn. J.M.,Long.D.P..Cabanas.A.. Watkins. J.J. 2001. Science294:]4\:
          (b) Fernandas. N.E.. Fisher. S.M., Poshusta. J.C., Vlachos. D.G. Tsapatsis,
          M.. Watkins. J.J. 2001. Chem. Mat.  13:2023; (c) Blackburn. J.M.. Long. D.P..
          Watkins, J.J. 2000.  Chem.  Mat. 12:2625; (d) Long. D.P.. Blcakburn.  J.M..
          Watkins. J.J.  2000. Adv. Mat. 12:913: (e) Watkins. J.J.. Blackburn.  J.M..
          McCarthy. T.J.  1999. Chem. Mat. 11:213.

    90.    Harrod. M.. Moller. P. July 24.2001. US Patent No. 6.265.596.

    91.    Poliakoff. M.. Swan. T.M.. Tacke. T. Hitzler. M.G. Ross, S.K.. Wieland. S.
          December 5.2000. US Patent No. 6.156.933.

    92.    Koch. D.. Leitner. W.  1998. /. Am. Chem. Soc. 120:13398: (b) Kainz. S., Leitner.
          W. 1998. Catal. Len. 55:223:  (c) Francio, G, Leitner. W. 1999. Chem. Commun.
          1663.

-------
                                                                           131
93.    Palo, D.R., Erkey, C.  1998. Ind. Eng. Chem. Res. 37:4203; (b) Palo. D.R., Erkey.
      C. 1999. Ind. Eng. Chem. Res. 38:3786: (c) Davis. T. Erkey. C. 2000. Ind Eng.
      Chem. Res. 39:3671.

94.    Hu. Y., Chen, W.. Xu. L.J., Xiao. J.L. 2001. Organometallics 20:3206.

95.    Guo, Y. Akgerman, A. 1997.  Ind. Eng. Chem. Res. 36:4581; (b) Guo. Y.
      Akgerman.A. 1999. J. Supercrit. Fl. 15:63; (c) Lin, B.. Akgerman. A. 2001.
      Ind. Eng.  Chem. Res. 40:1113.

96.    Banet Osuna, A.M.. Chen. W., Hope, E.G. Kemmitt, R.D.W., Paige, D.R..
      Stuart, A.M., Xiao, J.. Xu, L. 2000. J.  Chem. Soc. Dalt. Trans. 22:4052.

97.    Kie, J., Han, B., George, M.W., Yan, H., Poliakoff, M. 2001. J. Am. Chem.
      Soc.  123:3661.

98.    Meehan. N.J.. Sandee, A.J., Reek. J.N.H..  Kamer, P.C.J., van Leeuwen.
      P.W.N.M., Poliakoff. M. 2000.  Chem. Commun. 1497.

99.    Tadd, A.R., Marteel, A.. Mason, M.R., Davies, J.A.. Abraham, M.A. 2002.
      Ind. Eng. Chem. Res. ASAP; (b) Snyder, G, Tadd. A., Abraham, M.A. 2001.
      Ind. Eng. Chem. Res. 40:5317; (c) Dharmidhikari, S., Abraham. M.A. 2000.
      J. Supercrit. Fl 18:1.

100.  Ojima, I., Urata, H.  US Patent No. 5,962,744. Oct. 5. 1999.

101.  Wu, X-W., Oshima. Y., Koda, S. 1997. Chem. Lett. (Japan) 1045.

102.  Srinivas, P., Mukhopadhyay, M. 1997.  Ind. Eng. Chem Res. 36:2066.

103.  Jenzer, G, Mallat, T., Maciejewski, M., Eigenmann, E, Baiker, A.  2001.
      Appl. Catal.A:Gen.  125.

104.  Nolen, S.A., Lu, J., Brown, J.S., Pollet, P., Eason,  B.C., Griffith, K.N.,
      Glaeser, R., Bush, D., Lamb, D.R., Liotta, C.L., Eckert, C.A., Thiele, G.F.,
      Bartels, K.A. 2002. Ind. Eng. Chem. Res. 41:316; (b) Hancu, D., Green, J.,
      Beckman,E.J.  2002. Industr.Eng. Chem. 41:4466.

105.   Yao, H.. Richardson, D.E. 2000. J. Am. Chem. Soc. 122:3220; (b) Richardson,
      D.E., Yao, H., Frank, K.M., Bennett, D.A. 2000. J. Am. Chem. Soc.  122:1729.

106.   Bimbaum, E.R., Le Lacheur, R.M., Horton, A., Tumas, W. 1999. J. Molec.
      Cat. A: Chem. 139:11.

107.   Loeker, F, Leitner, W. 2000.  Chem. Eur. J. 6:2011.

108.   Hans,GR.,Kolis,J.W. 1998. Organomet. 17:4454.

109.   Jiang, H.,Jia,L., Li. J. 2000.  Green Chem. 2:161.

-------
132
     110.   Busch, D.H.. Wei. M., Music. G.T.. Subramaniam. B. 2002. 7. Am. Chem. Soc.
           124:2513.

     111.   Pitchai.R.,Kahn,A.P.,Gaffney,A.M. US Patent No. 5.625,084. April 27.1997.

     112.   McCoy, M. 2001. Chem. Eng. News 79(43): 19.

     113.   Kumar, A., Gupta. R.K.  1998. Fundamentals of Polymers. New York: McGraw-Hill.

     114.   Flory, P.J.  1953. Principles of Polymer Chemistry.  Ithaca, NY: Cornell
           University Press.

     115.   Wiesmet, V.,  Weidner, E.,  Behme, S., Sadowski, G, Arlt, W.  2000.  J.
           Supercrit. Fl.  17:1.

     116.   Wissinger. R.G, Paulaitis, M.E. 1987. J. Polym. ScL, Part B, Polym. Phys.
           25:2497; (b) Wissinger, R.G, Paulaitis, M.E.  1991. J. Polym. ScL, Part B.
           Polym. Phys.  29:63J.

     117.  Ehrlich, P.  1965. J. Polym. Sci. A 3:31; (b) Ehrlich, P., Mortimer, GA.
           1970. Adv.  Polym. Sci. 7:386.

     118.  Kirby, C, McHugh, M.A. 1999. Chem Rev.  99:565.

     119.  Allada, S.R.  1984. hid. Eng. Chem. Proc. Des. Dev.  23:344.

     120.  Orr, P.M., Heller, J.P., Taber, J.T., Card, R.J.  1983. CHEMTECH 482.

     121.  McFann, GJ., Howdle, S.M., Johnston, K.P.  1994. AIChE J. 40:543.

    122.  Sikorski, M.E., Lundberg, J.L.  1983. Abstract HS 22, American Physical
          Society Meeting, Detroit, MI; (b) Harris, T.V..  Irani, C.A., Pretzer. W.R.
          September 3,  1991. US Patent No. 5,045,220.

    123.  lezzi. A., Bendale, P., Enick, R.M., Turberg, M.. Brady. J. 1989. Fl. Phase
          Equil. 52:307; (b) Hoefling, T.A., Enick, R.M., Beckman, E.J.  1991.  J.
          Phys. Chem. 95:7127 (c) Hoefling, T.A., Newman, D.A., Enick, R.M.,
          Beckman, E.J. 1993. / Supercrit. Fl. 6:165: (d) Hoefling, T.A., Newman,
          D.A.. Enick, R.M., Beckman, E.J. 1993.  J. Supercrit. Fl. 6:205.

    124.  Ghenciu, GG  1997. Extraction of proteins in liquid carbon dioxide. Ph.D.
          thesis. University of Pittsburgh.

    125.  McClain, J.B., Belts, D.E., Canelas. D.A., Samulski, E.T.. DeSimone, J.M.,
          Londono. J.D., Cochran, H.D., Wignall, GD., Chillura Martino, D., Triolo, R.
          \996.Science 274:2049.

    126.   McHugh, M.A., Park, I-H., Reisinger, J.J., Ren, Y. Lodge, T.P.. Hillmyer, M.A.
          2002. Macromolecules 35:4653; (b) DiNoia, T.P., Conway, S.E., Lim, J.S.,

-------
                                                                             133
      McHugh, M.A. 2000. /. Polym. ScL Pan B: Polym. Phys. 38:2832; (c) Lora, M.,
      Lim, J.S., McHugh, M.A. 1999. / Phys. Chem. B 103:28 18; Mertogan, C.A.,
      McHugh, M.A., Tuminello, W.H. 1 999. J. Appl. Polym. Sci. 74:2039.

127.   Dardin,A.,DeSimone,J.M.,SamuIski,E.T. \998.J.Phys.Chem. 102:1775;(b)
      Dardin, A., Cain, J.B., DeSimone, J.M.. Johnson, Jr., J.C.S., Samulski, E.T. 1 997.
      Macromolecules 30:3593.

128.   Diep, P., Jordan, K.D., Johnson, J.K., Beckman, E.J.  1998. J. Phys. Chem. A
      1022231.

129.   Supercritical fluids. 1999. Chem. Rev. 99; many of the articles in this issue
      describe the use of CO^-philic materials.

130.   Meredith, J.C., Johnston, K.P.,Seminario,J.M.,Kazarian,S.G,Eckert,C A 19%.
                  100: 10837.
131.   Kazarian, S.G, Vincent, M.F., Bright, F.V., Liotta, C., Eckert, C.A. 19%. J. Am.
      ChenLSoc.llS:n29.

131   Nelson. M.R., Borkman, R.F. 1998. J. Phys. Chem. A 102:7860.

133.   Rindfleisch,F.,DiNoia,XP., McHugh. M.A. 19%. J. Phys. Chem. 100:15581.

134.   Beckman, E.J. (unpublished results)

135.   Lepilleur,C, Beckman, EJ. 1997. Fluid Phase Equil. 134:285.

136.   Kleintjens, L.A., Koningsveld, R., Gordon, M. 1980. Macromolecules 13:303;
      (b)Chen,S-J.,Banaszak,M.,Radosz,M. 1995. Macromolecules 28:1812.

137.   Sarbu. X, Styranec, X, Beckman, E J. 2000. Nature 405: 165.

138.   Fink, R.,Hancu,D., Valentine, R., Beckman, EJ. 1999. J. Phys. Chem. B 103:6441.

139.   Xiang,Y.,Kiran,E. 1995. Polymer 36:48 17.

140.   Wallen, S. International Symposium on Supercritical Fluid Xechnology. 2001 .
      Myrtle Beach, SC.

141.   Odian,G. Principles of Polymerization. 3rded. 1991. New York: John Wiley &
      Sons.

14Z   Fukui, K., Kagiya, X, Yokota, K, Xoriuchi, Y, Kuniyoshi, F.  1 970. US Patent No.
      3322228.

143.   O'Neill, M.L., Newman, D., Beckman, EJ. 1999. Polym. Eng. ScL 39:862.

144.   Romack,XJ.. DeSimone, J.M. 1995. Macromolecules 28:8429; (b)Romack,XJ.,
      DeSimone, J.M. 1997. US Patent No. 5,674,957; (c)Romack,XJ., DeSimone,

-------
134
           J.M.  1999. US Patent No. 5.939.501 ;(d)Romack.T.J..DeSimone. J.M. 1999.
           US Patent No. 5,939,502; (e) Romack, T.J.. DeSimone. J.M. 1999. US Patent
           No. 5.981.673.

     145.   George, E.R. (Zeus. Inc.) 2002. (personal communication)

     146.   Romack,T.J., Maury, E.E., DeSimone J.M. 1995. Macmmolecule.s28:9\2.

     147.   Hyatt, J A 1984. J. Org. Chem. 49:5097.

     148.   Adamsky, F., Beckman, E.J. 1996. The Polymeric Materials Encyclopedia.
           Salamone, J.C. (ed.).  CRC Press. Boca Raton, FL Volume 1: A-B.p54:(b)
           Singley, E.J. 1997. Development of fluoroether amphiphiles and their applica-
           tions in heterogeneous polymerizations in supercritical carbon dioxide. Ph.D.
           thesis. University of Pittsburgh.

     149.   Fink, R., Beckman, E.J., Hildebrandt, V. 2000. PCTInt.Appl.WO0053639:(b)
           Fink, R., Beckman, E.J. 2002. hid. Eng. Chem. Res. (submitted)

     150.   Bunyard, W.C., Kadla, J.F., DeSimone, J. M. 2001. J. Am. Chem. Soc. 123:7199.

     151.   Barrett, K.E.J. Dispersion Polymerization in Organic Media.  1975.  New
           York:  John Wiley  and Sons.

     152.  DeSimone, J.M., Maury, E.E.. Mencelogu, Y.Z., McClain, J.B., Romack. T.R..
          Combes, J.R. 1994. Science 265:356; (b) Canelas, D.A., Belts, D.E.. DeSimone.
          J.M. 1996. Macromolecules 29:2818; Canelas, D.A., DeSimone. J.M. 1997.
          Macromolecules 30:5673.

    153.  O'Neill, M.L., Yates, M.Z.. Harrison, K.L., Johnston. K.P., Canelas, D.A..
          Belts, D.E., DeSimone, J.M., Wilkinson, S.P. 1997.  Macromolecules 30:5050;
          (b) Yates, M.Z., O'Neill, M.L.. Johnston, K.P., Webber, S., Canelas, D.A..
          Belts, D.E., DeSimone, J.M. 1997. Macromolecules 30:5060.

    154.  Lepilleur,C, Beckman, E.J.  1997. Macromolecules 30:745.

    155.  Christian, P., Giles, M.R., Griffilhs, R.M.T., Irvine, D.J., Major, R.C.. Howdle.
          S.M. 2000. Macromolecules 33:9222.

    156.   Shaffer, K.A., Jones, T.A., Canelas,  D.A., DeSimone, J.M., Wilkinson, S.P.
          1996. Macromolecules 29:2704.

    157.   Yates, M.Z., Li, G, Shim. J.J., Maniar. S.. Johnston. K.P., Lim. K.T.. Webber. S.
          1999. Macromolecules 32:1018; (b)  Li, G, Yates, M.Z., Johnston. K.P.. Lim.
          K.T., Webber, S. 2000. Macromolecules 33:1606.

    158.   Shiho, H., DeSimone, J.M. 2001. Macromolecules 34:1198.

    159.   Romack, T.J., Maury, E.E.. DeSimone, J.M.  1995. Macromolecules 28:912.

-------
                                                                            135

160.   Wood, C.D.. Cooper, A.I. 2001. Macromolecules 34:5.

161.   Biddulph. R.H., Plesch, P.H. 1960. J. Chem. Soc. 3913.

162.   Pernecker, T., Kennedy, J.P. 1994. Polym. Bull. (Berlin) 33:13.

163.   Clark. M.R., DeSimone, J.M.  1994. Polym. Prep. (Am Chem Soc. Di\. Polym.
      Chem.) 35:482: (b) Clark. MR.. DeSimone. J.M.  1995. Macromolecules
      28:3002.

164.   Mistele, C.D.. Thorp, H.H., DeSimone, J.M.  1996. 7. Macromol. SciA 33:953:
      (b) 1998. US Patent No. 5,840.820.

165.   Debrabander. E.F.. Brothers, P.O. April 18,2000.  US Patent No. 6.051.682: (b)
      Wheland, R.C., Brothers. P.O. August 22,2000. US Patent No. 6.107,423; (c)
      Wheland. R.C., Brothers, P.O., Anolick, C., Stewart, C.W. May 8,2001. US
      Patent No. 6.228,963; (d) Brothers, P.D. August 15, 2000.  US Patent No.
      6,103,844.

166.   Cunningham, M.F.. Mahabadi, H.K. May 2,2000. US Patent No. 6,057,409.;
      (b)Eian,GL..Elsbernd,C.L.S. September 24,1996. US Patent No. 5,559,198.

167.   http://europa.eu.int/comm./research/growth/gcc/pressroom.html

168.   Burke, A.L.C., Givens, R.D., Jikei, M., DeSimone, J.M.  1997. Polym. Prep.
      38:387; (b)Givens.R.D.,Jikei,M., DeSimone, J.M. 1997. Polym. Prep. 38:468.

169.   Gross, S.M.. Roberts, GW., Kiserow, D.J., DeSimone, J.M. 2000. Macromol-
      ecules 33:40.

170.   Shi, C., DeSimone, J.M., Roberts, G, Kiserow, D.J. 2001. Macromolecules
      34:7744.

171.   Chaudhari, R.V., Gupte,  S.P., Kelkar, A.A.,  Kanagasabapathy, S.,
      Radhakrishnan. S.  US Patent No. 5,917,077. June 29,1999.

172.   2001.  Chem. Eng. News 79(21 ):27.

173.   McGhee, W.D., Stern, M.K.. Waldman, T.E. August 3, 1993. US Patent No.
      5,233,010.

174.   Hoelderich, W.F.. Dahlhoff.G 2001. Chem. Innov. 29.

175.   Klempner, D.. Frisch, K.C. Handbook of Polymeric Foams.  1991. New York:
      Oxford University Press.

176.   Fiorentini. C., Griffiths, A.C.M. May 13,1997. US Patent No. 5,629,027; (b)
      Fiorentini.C, Griffiths, A.C.M. Septem,ber9.1997. US Patent No. 5.665,287;
      (c) Fiorentini. C., Griffiths, A.C.M. June 17,1997. US Patent No. 5,639.483.

-------
136
     177.   Sulzbach, H-M., Steilen, H., Raffel, R., Eiben. R.. Ebeling. W. October 3,2000.
           US Patent No. 6,127,442.

     178.   Merriman, P.  May 18,1965. US Patent No. 3,184,419.

     179.   Inoue, S., Koinuma, H., Tsuruta, T. 1969.  Makromol. Client. 130:210; (b)
           Inoue, S., Koinuma, H.,Tsuruta,T.  1969. J. Polym. Sc/. £7:287.

     180.   Inoue, S. 1976. CHEMTECH588; (b)Rokicki, A., Kuran, W. 1981. Macmmol.
           Sci. -Rev. Macmmol. Chem. C21:135; (c ) Kuran, W. 1998. Prog. Polym. Sci.
           23:919.

     181.   Aida. T., Inoue, S. 1982. Macromolecules 15:682; (b)Aida.T..Ishikawa,M,
           Inoue, S. 1986. Macromolecules 19:8; (c) Darensbourg. D.. Holtcamp, M.
           1995.  Macromolecules. 28:7577; (d) Super. M.. Berluche, E., Costello, C.,
           Beckman, E.J. 1997. Macromolecules 30:368; (e) Cheng, M.. Moore, D.R.,
           Reczek.JJ., Chamberlain, B.M.,Lobkovsky,E.B.,Coates,GW. 2001.  J.Am.
           Chem.Soc. 123:8738.

     182.   Buzdugan, E., Beckman, E.J.  Presented at the 6th International Conference
           on CO, Utilization. Breckenridge, CO, September 10-14.2001.

     183.   Super, M., Beckman, E.J. 1997. Trends. Polym. Sci. 5:236.

     184.   Soga, K., Hosoda, S., Tazuke, Y., Dceda. S. 1975. J. Polym. Sci. - Polym. Lett.
           13:265; (b) K. Soga, M. Sato, S. Hosoda. S. Ikeda. 1975. J. Polym. Sci. -
          Polym. Lett. 13:543.

     185.  Sargent, D.E.  1949. US Patent No. 2,462,680.

     186.  Odell, P.G December 16,1997. US Patent No. 5,698.665.

     187.  Kim, K-S., Danishevsky, S., Peterson, C.B.  August 8.2000. US Patent No.
          6,100,372.

     188.  Berens, A.R.,  Huvard, GS., Korsmeyer, R.W.. Kunig. F.W.  1992. /. Appl.
          Polym. Sci. 46:231; (b) Kazarian, S.G. Brantley. N.H.. West. B.L., Vincent,
          M.F., Eckert, C.A. 1997. Appl. Spectr. 51:491.

    189.   Montero, GA., Smith, C.B., Hendrix. W.A.. Butcher. D.L. 2000. Ind. Eng
          Chem. Res. 39:4806.

    190.   Bach, E., Cleve, E., Schollmeyer, E., Bork, M.. Korner, P. 1998. Melliand Int.
          3:192; (b) Bach, E., Cleve, E..  SchoJlmeyer, E., Vardag, T., Korner, P.  1999.
          2:165; (c) Argyle, M.D., Propp, W.A.  January  20. 1999. US Patent No.
          5,709,910; (d) Eggers, R., von Schnitzler, J., Truckenmuller. K. August 17,
          1999. US Patent No. 5,938,794; (e) Schollmeyer. E., Bach. E.. Cleve, E., Bork,
          M., Steinhauer, M., Korner, J-P. September 21,1999. US Patent No. 5,953,780;

-------
                                                                             137
       
-------
138
           37:2771; (b) Royer. J.R., Gay. Y.J.. DeSimone. J.M.. Khan, S.A. 2000. J. Polym.
           Sci. - Polym. Pins. 38:3168.

     205.   Suh.  K. Extruded thermoplastic foams produced with supercritical fluids.
           Paper presented at the Conference on Supercritical Fluids in Materials Pro-
           cessing and Synthesis. Switzerland: Davos. September 26-October 1. 1999.

     206.   Wilkes. GR.. Ely. K.A.. Dunbar. H.A.. Uhl. E.R.. Zwynenburg, J.L. November
           30. ] 999. US Patent No. 5.993.706.

     207.   Cebien.J. Bayer Corporation (personal communication).

     208.   Perry. R.H.. Chilton. C.H. 1973. Chemical Engineering Handbook. 5th ed.
           New York: McGraw-Hill.

     209.   Shi. C., Huang. Z., Kilic, S.. Xu, J., Enick. R.M.. Beckman. E.J.. Carr, A.J.,
           Melendez.R.E.. Hamilton. A.D.  1999. Science 286:1540.

     210.   Park. C.B., Suh. N.P.. Baldwin. D.F. April 18,2000. US Patent No. 6.051,174.

     211.   International Roadmap for Semiconductors.  1999. Semiconductor Industry
           Association. San Jose. CA.

     212.   Allen. R.D., Wallraff. G September9.1997.  US Patent No. 5.665.527.

     213.   Pryce Lewis, H.G. Weibel. GL., Ober, C.K., Gleason, K.K. 2001.  Chem. Vap.
          Dep. 7:195; (b) Sundararajan, N., Yang, S., Ogino, K.. Valiyaveettil, S., Wang,
          J., Zhou. X.. Ober, C.K., Obendorf. S.. Allen, R.D. 2000. Chem. Mat. 12:41;(c)
          Yang, S.. Wang, J., Ogino, K., Valiyaveettil, S., Ober. C.K.  2000. Chem. Mat.
           12:33.

    214.  Taylor, O.K., Carbonell, R.. DeSimone, J.M. 2000. Ann. Rev. Energy Envir.
          25:115; (b) Kendall, J., DeSimone, J.M., Carboneil, R.G, McAdams, C.L. 2002.
          PCTlnt. Appl. WO 0231596.

    215.  Brodsky, C., Byers, J., Conley, W., Hung, R., Yamada, S.. Patterson,  K.,
          SomerveU, M., Trinique, B., Iran, H. V, Cho, S.. Chiba, T, Lin. S-H., Jamieson,
          A., Johnson, H., Vander Heyden, T., Willson, C.G 2000. / Vac. Sci. Tech. B
          18:33%.

    216.   Carbonell. R.C., DeSimone, J.M., Novick, B.J. US Patent No. 6.083.565; (b)
          DeSimone, J.M.. Carbonell. R.C. US Patent No. 6.001.418.

    217.   McCoy, M.  2001. Chem. Eng. News 79(38): 10.

    218.   Voss, D. 2002.  Tech. Rev. 105:27.

    219.   2001.  Chem. Eng. News 79(38): 10.

-------
                                                                             139
220.  2(X) 1. Chem. Eng. News 79(50): 14.

221.  Agarwal. V.K. US Patent No. 6.306.754. October 23.2001: (b) Lee. H-J.. Jeng.
      D.G-K. November 20.2(X)1. US Patent No. 6.319.858.

222.  Sheldon. R. A. 2001. Coll. Led. of Summer Schools on Green Chemistry.  In:
      Tundo. P.. Clemen/a. L.. Perosa. A. (eds.). 41.

223.  Klibanov.A.M.  1997. Trends Biotechnol. 15:97.

224.  Mesiano. A.. Beckman. E.J.. Russell, A.J. 1999. Chem. Rev. 99L:623.

225.  Kamat. S.. Barrera. J.. Beckman. E.J.. Russell. A.J.  1992. Biotech. Bioeng.
      40:158.  (b) Kamat. S.. Critchley, G. Beckman.  E.J.. Russell. A.J.  1995.
      Biotech. Bioeng. 46:610.

226.  Johnston. K.P..  Harrison, K.L., Clarke. M.J.. Howdle. S.M., Heitz. M.P..
      Bright. F.V.. Carlier. C.. Randolph, T.W.  1996.  Science 271:624.

227.  Ghenciu. E.. Russell. A.J.. Beckman, E.J.  1998.  Biotech. Bioeng. 58:572.
      (b) Ghenciu. E.G.. Beckman, E.J.  1997.  Industr. Eng. Chem. 36:5366.

228.  Chaudhary. A.K..  Beckman. E.J.. Russell, A.J.  1995. J. Am. Chem. Soc.
      117:3728.

229.  Weinstein. R.D.. Winslo. A.R.. Danheiser. R.L., Harris, J.G, Tester, J.W.
      1996. J. Phys. Chem. 100:2337.

230.  Reaves. J.T.. Roberts. C.B.  1999. hid. Eng. Chem. Res.  38:855.

231.  Matsuo. J., Tsuchiya. T., Odashima. K., Kobayashi, S. 2000. Chem. Lett. 178.

232.  Ikushima. Y.. Saito. N.. Arai, M.  1992. J. Phys. Chem. 96:2293.

233.  Renslo. A.R.. Weinstein. R.D.. Tester. J.W.. Danheiser. R.L. 1997.  J.Org.
      Chem. 62:4530.

234.  Oakes. R.S.. Heppenstall. T.J.. Shezad, N., Clifford,A.A.. Rayner, C.M. 1999.
      Chem. Commuii. 1459.

235.  Totoe. H.. McGowin. A.E.. Turnbull. K. 2000. J. Supercrit. Fl. 18:131.

236.  Chateauneuf. J.E.. Nie. K. 2000. Adv. Envir. Res. 4:307.

237.  Kawada, A.. Mitamura. S., Matsuo, J., Tsuchiya. T. Kobayashi, S.  2000.
      Bull. Chem.  Soc. Japan 73:2325. (b) Kobayashi. S.. Manabe, K.  2000.
      PureAppl. Chem.  72:1373.

238.  Hitzler. M.G. Smail. F.R.. Ross, S.K., Poliakoff. M.  1998. Chem. Commun.
      359.

-------
140
     239.   Olah. GA.. Marinez. E., Torok. B.. Surya Prakash, GK. 1999. Catal. Lett.
           61:105.

     240.   Pernecker. T.. Kennedy. J.P.  1994. Polym. Bull. Berlin 32:537.

     241.   Fornika.R..Gorls.H..Seemann,B..Leitner.W. 1995. Client. Conmnui. 1479:
           Leitner. W. 1995. Angew. Chem. Int. Ed. Eng. 34:2207.

     242.   Jessop, P.E.. Hsiao. Y. Ikariya, T., Noyori, R. 1996. J. Am. Chem. Sot: 118:344.

     243.   Shi, M.. Nicholas. K.M. 1997. J. Am. Chem. Sac. 119:5057.

     244.   Ushikoshi. K.. Mori. K.. Watanabe, T.. Takeuchi. M.. Saito. M.  1998. Ad-
           vances in chemical conversions for mitigating carbon dioxide. In: T. Inui. M.
           Anpo. K. Izui. S. Yanagida.T. Yamaguchi (eds.). Studies in Surface Science
           and Catalysis.  New York: Elsevier Science. Volume 114. p 357.

    245.   Arakawa. H.. et al. 2001. Chem. Rev. 101:953.

    246.   Yamazaki, N.. Nakahama. S.. Higashi, F. 1979. hid. Eng. Chem. Prod. Res.
           Dev. 18:249.

    247.   Shaikh.A-A.G. Sivaram,S.  19%. Chem. Rev. 951.

    248.   Behr, A.  1988. Carbon dioxide activation by metal complexes. New York:
           VCH; (b) Halmann, M.M. 1993. Chemical fixation of carbon dioxide: meth-
           ods for recycling CO2 into useful products. London: CRC Press; (c) Gibson.
          D.H.  1996. Chem. Rev. 96:2063. Also, see reference 124 from our reference
          245.

    249.  Jia,L., Jiang. H.. Li, J. 1999.  Green Chem. 1:91;(b)Kishimoto.Y. Ikariya. T.
          2000.  J. Org. Chem. 65:7656.

    250.   Morita, O.K., Pesiri, D.R., David, S.A.. Glaze. W.H.. Tumas. W.  1998. Chem.
          Commun. 1397; (b) Chen, W., Xu. L.. Xiao. J. 2000. Org. Lett. 2:2675; (c)
          Bhanage. B.M.. Ikushima, Y, Shirai, M.. Arai. M. 1999. Tetrahed. Lett. 6427;
          (d) Osswald. T., Schneider, S.. Wang, S., Bannwarth. W. 2001. Tetrahed. Lett.
          42:2965.

    251.   Cacchi,S.,Fabrizi.G.Gasparrini.F..Villani.C. 1999. Synlen. 3:345.

    252.   He. L-N., Choi. J-C, Sakakura, T.  2001.  Tetrahed. Lett. 42:2169.

    253.   Montilla, F.. Clara. E., Aviles. T., Casimiro. T.. Aguicar Ricardo, A.. Nunes de
          Ponte,M. 2001. J. Organomet. Chem. 626:227.

    254.   Kabir, A., Marshall. W.D. 2001. Green Chem. 3:47.

    255.   Jeong, N., Hwang, S.H. 2000. Angew. Chem. Int. Ed. Eng. 39:636.

-------
                                                                              141
256.   PoliakofT. M.. Swan. T.M.. Tackc. T.. Hitzlcr. M.G.. Ross. S.K.. Wieland. S..
       Small. F.R. October 16.2(K)I. US Patent No. 6.303.840.

257.   Sutmimamam. B.. Clark. M.C. May 25.1999. US Patent No. 5.907.075.

258.   Ritter. S.K. 2001.  C/iem. Eng. News 79(40):63.

259.   Harris. R.. Jureller. S.H.. Kerschner. J.L.. Trzasko. P.T.. Humphreys. R.W.R.
       November 2. 1999. US Patent No. 5.977,348.; (b) Schiraldi. D.A.. Brown. J.S..
       Chandler. K.. Ecken. C.A.. Hurley. J.S.. Lamb. D.R.. Lesutis. P.H.. Liotta. C.L.
       March 20.2001. US Patent No. 6.204,386.

260.   Lowack. R.. Meyer. J.. Eggersdorfer. M., Grafen. P. June 4. 1996.  US Patent
       No. 5,523.420.

261.   Matson. D.W.. Fulton. J.L.. Smith. R.D.. Consani. K.A. August 24.1993.  U.S.
       Patent No. 5.238.671.

262.   Ji, M.. Chen. X.Y.. Wai. C.M.. Fulton. J.L. 1999. J. Am. Chem. Soc. 121:2631.

263.   Cason. J.P.. Khamwasadkar. K.. Roberts. C.B.  2000.  Ind. Eng. Chem.  Res.
       39:4749.

264.   Laintz. K.E.. Wai. C.M.. Yonker, C.R.. Smith. R.D. 1991. / Supercrit. Fl. 4:194.

265.   Erkery, C. 2000. J. Supercrit. Fl. 17:259. (b) Wai. C.M.. Wang. S. 1997. J.
       Chronmt. A 785:369. (c) Smart. N.G, Carleson. T.. Kast. T.. Clifford. A.A..
       Burford.M.D..Wai.C.M. 1997. Talanta 44:137.

266.   Lagalante.A.F.Hansen.B.N.. Bruno.T.J.. Sievers, R.E.  1995. Inorg.Chem.
       34:5781. (b) Anderson. W.C., Sievers, R.E.. Lagalante,A.F., Bruno,T.J. 2001.
      J. Chem.  Eng. Data 45:1045.

267.   Dicinoski. W. 2000. Min. Mag.  182:258.260.262; (b) White. R.H. 1999. In:
      Copper Leaching. Solvent Extraction, and Electrowinning. Soc. Min. Metall.
      Explor. Littleton. CO: p 229: (c)Townsend. B., Sievers. K.J. 1990. Min.
      Mag. 162:26,29.32.35: (d) Flett. D.S. 1974. Inst.  Mining Met. Trans. Sect. C
      83C30.

268.   Olper. M., Maccagni. M.. Buisman. C.J.N., Schultz. C.E. 1999. Proceedings
      of the Copper  '99-Cobre 99th International Conference (4th). Young.  S.K.
      (ed.). Min.. Met., & Mat. Soc. 4:597; (b) Valic. D.,Tombalakian. A.S., Alfantazi.
      A., Moskalyk. R.R.  1999. Proceedings of the Copper'99-Cobre 99th Interna-
       tional Conference (4th). Young. S.K. (ed.). Min., Met.. & Mat. Soc. 3:711.

269.   Davenport. W.G. 1999. Proceedings of the Copper '99-Cobre 99th Interna-
      tional Conference (4th). Young. S.K. (ed.). Min., Met.. & Mat. Soc. 1:55.

270.  Braun, T.B. 1981.  J. Met. 33:59.

-------
142
     271.   Benner.L.S.. Su/uki.T..Meguro. K..Tanaka.S.(eds.K 1991. Precious Met-
           als: Science and Technology. Int. Prec. Met. Inst.. Allentown. PA. (h) Harris.
           G.B.A. 1993.  In: Mishra. R.K. (ed.). Toronto. Canada:  Precious Metals
           Pergamon Press, p 351.

     272.   Powell. C.J.. Beckman. E.J. 2001. hid. Eng. Chein. Res. 40:2897.

     273.   Izatt. R.M.. Bradshaw. J.S.. Bruening. R.L. 1996.  PureAppl. Cliem. 68:1237:
           (b) list of publications on IBC Web site, http://www.ibcmrt.com/publica-
           tions

     274.   Sherman. J.H.. Hershberger. J.W.. Taylor. R.T.. Garrett. GM. 2000.  PCT
           Int. Appl.  WO 0056842: (b) Bertus. B.J., Walker. D.W.  1988.  US Patent
           No. 4.727.053: (c) Macedo. J.C.D.. Duarte. M.A. I., 1994. Appl. Cat. A
           110:87: (d) Roussel. J-C. Boulet. R. 1995. Pet. Refin. /(Crude Oil. Petro-
           leum Products. Process Flowsheets) 1 -15:453-60; (e) Pearson. C.D.. Green.
           J.B. 1993. Energy & Fuels 7:338; (0 Mitchell, P.C.H.  1990. Catal. Today
           7:439: Reynolds. J.G  1988. US Patent No. 4,789.463.

     275.   Li. J-T. Beckman, E.J.  1998.  Ind. Eng. Ghent. Res.  37:4768: (b) Laintz.
           K.E.. Hale. CD.. Stark. P.. Rouquette. C.L.  1998. Anal. Chen,. 70:400.

    276.   Yu. J.-J.. Chiu. K-H., Wang, S. 2001. J. Chin. Inst. Chem. Eng. 32:263; (b)
           Lin. Y., Wu. H.. Smart. N.G, Wai. C.M. 2001. Sep. Sci. Tech. 36:1149; (c)
           Gervais. F.. Perre, C.. Sarrade. S.. Moszkowicz, P.. Barna, L. 1999.  Recents
           Progres en Genie des Precedes. 13:197; (d) Shamsipur, M., Ghiasvand. A.R..
           Yamini. Y. 2001.  J. Supercrit. Fl. 20:163; (e) Wai. C.M., Waller, B. 2000.
          Ind. Eng. Chem. Res. 39:4837; (f) Kersch, C., Woerlee. GF. Witkamp. GJ.
          2002. In:  Liddell,K.C(ed.). Metal Separation Technologies Beyond 2000:
          Integrating Novel Chemistry With Processing.  Min..  Met, and Mat. Soc.,
          Warrendale. PA.

    277.   Baglin. F.G US Patent No. 5,897,704. April 27,1999; (b) Jones, R.H.  October
           12,1999. US Patent No. 5.965,201.

    278.   Douglas. M. A.. Templeton.A.C. February 9.1999. US Patent No. 5,868.856.

    279.   Vaartstra. B.A. November 12,2000.  US Patent No. 6.149,828.

    280.   Matson, D.W., Peterson. R.C.. Smith, R.D.  1987. J. Mat. Sci. 22:1919.

    281.   Jung, J., Perrut, M. 2001. J. Supercrit. Fl. 20:179.

    282.   Bungert, B.. Sadowski. G, Arlt, W. 1998. Ind. Eng. Chem. Res. 37:3208; (b)
          Kompella.  U.B., Koushik. K. 2001. Crit. Rev. Ther. Dwg Can: Sys. 18:173: (c)
          Reverchon, E.. Delia Porta. G 2001.  PureAppl. Chem. 73:1293; (d)Marr.R..
          Gamse.T. 2000. O!e«L£H£.Proc39:19;(e)Palakodaty,S.,York.P. 1999. Pharm.
          Res. 16:976; (0 Reverchon. E.  1999. J. Supercrit. Fl. 15:21.

-------
                                                                              143
283.  Gallagher, P.M., Coffey. M.P., Krukonis. V.J., Klasutis. N. 1989. In: Johnston.
      K.P., Penninger, J.M.L. (eds.). Supercritical Science and Technology. ACS
      Symp. Ser. 406, American Chemical Society, Washington. DC.

284.  Hanna, M., York, P. 1994. World Patent WO 95/01221: (b) Hanna. M.. York.
      P. 1995. World Patent WO 96/00610; (c) Hanna. M.. York. P. 2000. US Patent
      No. 6,063,138.

285.  Sievers, R.E., Karst, U.  1995.  Eur. Pat. 0 677 332. 1997. US Patent No.
      5,639,441; (b) Sievers, R.E., Karst, U., Milewski. P.O., Sellers. S.P., Miles.
      B.A., Schaefer, J.D., Stoldt, C.R., Xu, C.Y. 1999. Aer. Sci. Tech. 30:3; (c)
      Sievers, R.E., Karst, U. 2000. US Patent No. 6.095,134.

286.  Maa, Y-E, Prestrelski, S.J. 2000. Curr. Phamia.Biotech.  1:283.

287.  Thibert. R., Tawashi, R. Microspheres, microcapsules. and liposomes. 1999.
      Prep.Appl. 1:327.

288.  Perry, R.H., Chilton, C.H.  1973. Chemical Engineer's Handbook. 5th ed.
      New York: McGraw-Hill.

289.  Kileen,M. 1994. Powder Bulk Eng. 8:39; (b)Mortensen, S.,Lohmann. D.
      Chem-Tech. 1992.21:72; (c) Weers, J. 2000. Innov. Pharma. Tech. 1:111,114;
      Jono, K., Ichikawa, H., Fukumori, Y. 2000. Powd. Tech. 113:269.

290.  Shekunov, B., York, Y.P. 2000. J. Cryst. Growth 211:122.

291.   Weber, M., Russell, L.M., DeBenedetti, P.G 2002. J. Supercrit. Fl. 23:65; (b)
      Werling, J.O., Debenedetti, P.G. 2000. / Supercrit. Fl. \8:11; (c) Werling. J.O.,
      Debenedetti, P.G 1999. J.Supercrit. Fl. 16:167;(d)Kwauk,X.,Debenedetti.P.G
      1993. J.Aerosol.Sci.24:445;(e)Tom,J.W.,Debenedetti,P.G 1991. J.Aerosol
      Sci. 22:555; (f) Debenedetti, P.G 1990. AIChEJ. 36:1289.

292.  Lengsfeld, C.S., Delplanque, J.P., Barocas, V.H.. Randolph, T.W. 2000. J.
      Phys. Chem. B 104:2725.

293.  Texter, J. 2001. Reactions and synthesis in surfactant systems. Surfact. Sci.
      Ser. 100:577.

294.  Hong. L.. Guo, J., Gao, Y, Yuan, W-K. 2000. Ind. Eng. Chem. Res. 39:4882.

295.  Manning, M.C., Randolph, T.W., Shefter, E., Falk. R.F. 1999.  US Patent No.
      5,981,474; (b) Manning, M.C., Randolph, T.W.. Shefter, E., Falk. R.F. 1998.
      US Patent No. 5,770,559.

296.  Subramaniam. B., Saim, S., Rajewski, R.A.. Stella. V. 1999.  US Patent No.
      5.874.029: (b)Subramaniam,B.,Saim,S.,Rajewski.R.A.. Stella. V. 1998. US
      Patent No. 5,833,891.

-------
144
     297.   Daly. A.T.. Decker. O.H.. Wursthorn.R.  1998. US Patent No. 5.766.522: (b)
           Daly. A.T.. Shah. N.B.. Cornell. GD.. Wursthorn. K.R. US Patent No. 5.708.039.

     298.   Chordia. L. Presented ai the 5th Internaional Symposium on Supercritical
           Fluids. Atlanta. GA. April 8-12.2000.

     299.   Perrut. M. 2000. Ind. Ent>. Client. Res. 39:4531: (b) Hauthal. W.H. 2001.
           Chemosphere 43:123: (c) Smith. C.W.. Huse. G.  1998.  In:  Me Hardy. J..
           Sawan. S.P. (eds.).  Supercrital Fluid Cleaning.  Westwood. NJ:  Noyes
           Publishing.p245.: (d)Cygnarowicz-Provost.M. 1996. In: King, J.W.. List.
           GR. (eds.). Supercritical Fluid Technology for Oil and Lipid  Chemistry.
           Champaign. IL:  AOCS Press, p 155: (e)Gani.R..Hytoft.G.Jaksland.C. 1997.
           Appl. Therm. Eng. 17:889.

     300.   Colina. CM. Hall. C.K.. Gubbins. K.E. 2002. Fl. Phase Equil.  194-197:553:
           (b)Blas.F.J..Galindo.A. 2002. Fl. Phase Eauil. l94-197:50l:(c)Byun.H-S..
           Kim. K.. McHugh. M.A. 2000. bid. Eng. Chem. Res. 39:4580: (d)Zhang. Z-Y..
           Yang.J-C..Li.Y-G. 2000. Fl. Phase Equil. 169:1.

     301.   Diaz. S.. Espinosa. S.. Brignole. E.A.  2000. Computer-Aided Chem. Eng.
           8:319; (b)Artal. M.. Munoz Embid. J.. Velasco. I.. Berro. C.. Rauzy. E. 2001.
           Fl. Phase Equil. 178:119: (c) Espinosa. S.. Diaz. S.. Brignole, E.A.  2000.
           Comp. Chem. Eng. 24:1301: (d) Jaubert, J-N.. Coniglio. L. 1999. Ind. Eng.
           Chem. Res. 38:5011: Keshtkar. A.. Jatali. F.. Moshfeghian. M.  1997.  Fl.
           Phase Equil. 140:107.

    302.   http//www.micell.com

    303.  Crette, S.A.. DeSimone. J.M.. Carbonell. R.G. Tumas. W.. Brady. J.T. Ab-
          stracts of papers for ACS Meeting, August 20,2000,220:413, part 2.

    304.  Crooks, R.M., Lemon. B.I.. Sun. L.. Yeung. L.K.. Zhao, M.Q. 2001. Top. Curr.
          Chem.2\2SL

    305.  van der Broek. L.J.P.. Goetheer, E.L.V., Verkerk. A.W.. de Wolf, E., Deelman.
          B.J., van Koten, G. Keurentjes, J.T.F. 2001. Aug. Chemie Int. Ed. 40:473.

    306.  Jacobsen. GB.. Lee. C.T.. Johnston. K.P.. Tumas. W.  1999. J. Am. Chem. Sot:
          121:11902.

    307.   Chandler. K.. Culp. C.W.. Lamb. D.R.. Liotta. C.L., Eckert. C.A.  1998. Ind.
          Eng. Chem. Res. 37:3252.

    308.   Jacobsen. GB., Lee. C.T.. Johnston. K.P.  1999. J. Org. Chem. 64:1201. 1207.

    309.   Bonilla, R.J.. James, B.R.. Jessop. P.G 2000.  Chem. Commun. 941.

-------
                                                                            145
310.   Quadir. M.A.. Snook. R., Gilbert, R.G, DeSimone. J.M.  1997. Macromol-
      ecules 30:6015.

311.   Blanchard, L.A.. Hancu. D.. Beckman. E.J., Brennecke, J.F. 1999. Nature
      399:28.

312.  Sellin. M.F., Webb, P.B., Cole-Hamilton. D.J. 2001. Client. Commun. 781.

313.  Bosmann, A., Francio, G, Janssen, E., Solinas, M., Leitner, W.. Wasserscheid.
      P.  2001. Angew.  Chemie Int. Ed. Eng. 40:2697.

314.  Blanchard. L.A., Brennecke, J.F. 2001.  Green Chem. 3:17.

315.  Sigman,M.E.,Lindley,S.M.,Leffler,J.E.  1985. J.Am.Chem.Soc. 107:1471.

-------