Report
SPECTROSCOPY of SOLVENT CLUSTERING
S
upercriticalfluidsare receiving increasing attention from chemists as potentially useful media for chemical processing, and several labs have developed methods for extraction and separation of analytical-scale solid samples (1,2). Applications include pesticides, PCBs, and PAHs in soils; pollutants in plant, animal, and aqueous samples; fats from plants and animals; flavors, fragrances, and pharmaceuticals from plants; and additives and residues from polymers. Such methods are of particular importance for environmental monitoring in which extraction of the target analyte from a complex solid matrix can be the most difficult and time-consuming analysis stpn
Because the transport properties of supercriticalfluidsare 10-100 times greater than those of liquids, extraction efficiencies are often higher than those of liquid methods, resulting in a significant reduction in sample preparation times and an increase in sample throughput. SFE offers an economical alternative to standard liquid-extraction methods, which use simpler apparatus but can be quite slow. Another motive for studying supercritical fluids is their potential as solvents for industrial-scale solvent-intensive processes. Developing replacement processes based on supercritical C0 has
John F. Kauffman
lieved to influence solvation as well as chemA better ical reactivity in supercritical fluids. Fundastudies of the factors that influence understanding of how mental solvent clustering have been undertaken to improve predictive approaches to the desupercritical fluids velopment of supercritical fluid-based chemical processing. behave around solute In these investigations, it is important to distinguish between density augmentamolecules can help tion, caused by the presence of the solute molecule, and density fluctuations, which are characteristic of pure, superanalysts use them to criticalfluids.Typically, researchers use that are highly sensitive to their full advantage chromophores the solvent environment for these specstudies. Because the interacas they replace troscopic tions responsible for effects such as solvatochromism and fluorescence quenchconventional solvents ing are short-range interactions (10 11) studies of such phenomena provide inforbecome increasingly attractive, especially as mation on the solvent environment in the the cost increases for reprocessing and dis- cybotactic region The purpose of this Report is to provide an overview of fillTPTlt posing of noxious solvents. Supercritical spectroscopic approaches to studying denfluid replacement technologies have been sity augmentation in the cybotactic resuccessfully demonstrated at the commercial level and in many cases have resulled ii gion. product improvement (3-5). SFE methods often require the addiBecause of the potential of supercritical tion of polar modifiers such as ethanol. fluids, questions regarding the nature of sol- Presumably, these additives improve vation and the influence of the supercritiextraction efficiencies by binding to polar cal fluid environment on chemical reactivity matrix sites and inhibiting strong analyteare being investigated. During the past 15 matrix interactions. However, these years considerable evidence has indicated additives, as well as other matrix interferthe existence of localized regions of high ences, can influence analyte spectra. Thus, fluid density, or "solvent clusters," in the supercritical fluid clustering has imporimmediate vicinity of solute molecules, or tant implications for chemical analysis, the cybotactic region (6-9), which are beparticularly when quantitative determina-
University of Missouri 248 A
Analytical Chemistry News & Features, April 1, 1996
0003-2700/96/0368-248A/$12.00/0 © 1996 American Chemical Society
in SUPERCRITICAL FLUIDS
tions are based on spectroscopic data collected from analytes present in supercritical fluid environments. Background
Raised above its critical temperature, a pure substance cannot be liquefied, even under pressure. Figure 1 shows two pressure-density isotherms: one in the subcritical region and one in the supercritical region. The supercritical fluid is gaslike in that itfillsthe vessel in which it is contained. However, as the pressure is increased, the fluid density, unlike that of the liquid, can be continuously varied over the compressible region, which is the region over which the transition from gas-
like to liquidlike density occurs. As the liquid density is approached, the fluid begins to behave like a solvent; its solvating power depends on its density. Solute solvation generally begins to occur at pressures just above the critical pressure. The extent to which a solute is solvated is difficult to predict at the onset of solvation, because thefluidis highly compressible in this density regime. If the fluid is near gas density, its behavior can be modeled by the kinetic theory of gases in which only binary solvent-solvent interactions are likely to occur and are relatively infrequent. When thefluidis at its liquid density, the structure is determined largely by repulsive interactions. That is,
thefluiddensity at any given point within the container is essentially constant, because the molecules of thefluidare closely packed. The compressible region is intermediate between these extremes. The fluid molecules have a significant amount of free space; yet the environment is sufficiently dense, and multibody collisions are highly probable. Consider the effect of substituting a single solute molecule for one of the solvent molecules in a supercriticalfluid(Figure 2). If the interaction between the solute molecule and the nearby solvent molecules is attractive, the solute becomes a nucleation site for solvent clustering, particularly if solvent-solute interactions are
Analytical Chemistry News & Features, April 1, 1996 249 A
Report
Figure 1 . Pressure-density isotherms for subcritical and supercritical C 0 2 . Upon compression, the subcritical fluid liquefies, giving rise to a dense phase and a rarefied phase. Increasing the pressure does not significantly change the density of either phase, only the partitioning of molecules between the phases. In the supercritical fluid phase, the density can be continuously varied throughout the entire volume.
stronger than solvent-solvent interactions. Local solvent density augmentation, from attractive interactions, effectively results in solvation of the solute by the surrounding fluid, and the cluster of solvent around solute remains in solution. With this type of localized solvation, solubility in supercritical fluids can exceed estimations based on solute volatilization alone (i.e., ideal gas behavior) by orders of magnitude, particularly when modifiers are used (13). On the other hand, if the solvent-solute interactions repulsive the solute will reside in a rarefied region of the fluid A theoretical description of clustering can be developed using the radial distribution function formalism, a common statistical mechanical approach for describing density correlations in fluids (14-17). With this formalism, the local solvent density around a solute molecule p[2 can be related to the bulk solvent density p and 250 A
the radial distribution function g12(r) through (15,16) p12 = p(l + F[g12(r)])
(1)
The radial distribution function provides a measure of the density of molecules of type 1 within a shell around molecule 2, which is located at a distance r from its center. A direct connection can be made betweengl2(r) and the thermodynamic fugacity correction to the solubility of a pure solid in a pure liquid. This connection formally establishes the relationship between local density augmentation and enhanced solubility (18). Choice of the most pertinent form of F\g12(r)] is a topic of considerable importance currently under debate (17) and beyond the scope of this article. We can use physical reasoning, however, to establish anticipated properties of F[g12(r)], which we refer to as the local density aug-
Analytical Chemistry News & Features, April 1, 1996
mentation parameter (LDAP). The LDAP will be large in the compressible region of a supercritical fluid, with a positive value for attractive solutes and a negative value for repulsive ones. Also, the LDAP will diminish as the liquid density is approached because liquids are incompressible. That is, as the bulk fluid density approaches the liquid density, the local density also approaches the liquid density because the fluid is approaching a closely packed, incompressible state. Thus, we anticipate that the LDAP will vary with changes in the fluid bulk density. One of the goals of this work is to develop methods by which the sign, magnitude, and density dependence of the LDAP can be predicted. This objective will require consideration of both solvent-solvent and solute-solvent interactions on a molecular level The most common fluid under investigation is supercritical C02. Inexpensive and nontoxic, it is particularly attractive as an environmentally benign alternative to liquid solvents. However, C0 2 is rather nonpolar and is often unsuitable for chemistry involving polar reactants. Adding 1-10 mol% of a polar modifier (also called a co-solvent), such as ethanol or acetonitrile, to supercritical C02 has dramatically improved polar solute solubility (8 9) Spectroscopy is being used to develop a detailed understanding of cosolvent effects in mixed supercritical fluids The distribution of the comoonents of a mixed fluid is also affected bv the presence of solute molecules In these fluids local rtpnsity
augmentation fluid component and the co-solvent can ac cur around solute molecules and the ex
tent to which this occurs can be determined from spectroscopic studies. Instrumentation
The major limitation in these studies is sample container design, which must withstand high pressures while allowing the excitation and signal sources to enter and exit the cell. Design criteria depend strongly on the fluid to be studied and the temperature and pressure ranges of interest. Most of the phenomena discussed in this article occur at temperatures slightly higher than the critical temperature (e.g., at reduced temperatures between 1 and 1.1) and within the reduced pressure range of 1-3.
High-temperaturefluids,such as super- sorption and fluorescence features should and colleagues studied fluorescence from critical water, require extraordinary safety be related to solvent dielectric constant the charge transfer excited state of precautions in cell design. These fluids and refractive index. Numerous studies of DMABN in supercriticalfluoroform(25). tend to be highly corrosive to materials liquids show that spectral shifts of a variTheir results indicated that the local fluid such as quartz and stainless steel, which ety of solutes in polar and nonpolar sol- density was as much as a factor of 2 are normally considered robust. In addivents can be accurately predicted (10). greater than the bulk density in the comtion, the high temperatures required for Bulk solvent properties such as repressible region. the study of such fluids can alter the prop- fractive index and dielectric constant can Solvatochromic effects are generally erties of common materials in ways that be measured and accurately calculated as more pronounced in supercritical fluids make them unsuitable for such studies. a function of fluid density. Thus, remodified with a polar co-solvent (8,9,26). For example, stainless steel becomes brit- searchers interested in studying solvation Our studies of bis(4,4'-aminophenyl)sultle and loses its tensile strength when exhave tried to determine whether the relafone (APS) and its bis(dimethylamino) anposed to a high temperature for an ex- tionship between Av, refractive index, and alogue in mixed ethanol/C02 solvents tended period. dielectric constant holds for solutes disrevealed a nearly 10-fold enrichment of the solved in supercritical fluids. Yonker and cybotactic region in the polar modifier For low-temperature fluids, such as Smith demonstrated that the shift in the (8). The spectra in Figure 3 illustrate efC02, ethane, CF3H, and others with critifects typical offluorescencefrom solvatocal temperatures below 50 CC, stainless- absorbance maximum of 2-nitroanisole dissolved in supercritical C0 exhibits two chromic solutes dissolved in 5% (volume) steel cells can be designed to maintain rccfiincs of behavior: one at high density ethanol/C02. The broad feature centered pressures in excess of 10,000 psi. Failures and one at low density (7 20). They obaround 400 nm appears only in polar enviin windows and seals, on the other hand, served a steep linear dependence in the ronments and the feature at 340 nm apoften occur. For low-temperature studies, nonpolar environments The quartz is an acceptable window material. low-densitv regime and a much shallower linear deoendence in the high-density spectrum at 1422 psi (10 mol% ethanoll )i Also, sapphire is often used because its regime This variation was interpreted as strongly red shifted and is similar to the elastic limit is significantly higher than arising from clustering of the C0 solvent spectrum of APS in pure ethanol As the that of quartz, and it has a broad spectral around molecules in the transparency density regime leading to nearly liquidMany researchers use Teflon seals in like local densities at a bulk densitv of their cell designs. Teflon is not the best nl about 0 25 g/cm3 They rented sealing material because it tends to flow .. .. .jvi 1 1 J -4. and is not very elastic. However, most elasonce the liquidlike local density was tomeric materials that can withstand high UJ • • 4.1. U 11 J ' 4 . 1. pressures contain impurities within the achieved, increasing the bulk density has I J 4. • 41 4.1. I 1 1 elastomer matrix, which are extracted by only a moderate influence on the local solthe supercritical fluid. These impurities 4 J J 1 4. 1 1 vent density around solute molecules, can cause severe spectroscopic interfer1-1 L • 1 11 1 1 ences so Teflon is used to avoid such which results in a shallower dependence problems. It is important to remove these of the spectral shift on bulk density. sources of impurities because of the large Kim and Johnston demonstrated simitransport coefficients of these solvents lar effects with the solvatochromic dye and because diffusion-controlled prophenol blue dissolved in supercritical ethcesses such as fluorescence quenching ylene and fluoroform (21). They treated and excimer formation can be 1 or 2 orthe results using a model based on the raders of magnitude faster in snnercritical dial distribution function to calculate lofluids than i i lirmiHd cal solvent densities for the fluids. They demonstrated that the local density of fluoroform around the solute is signifiSolvatochromic studies cantly higher than that of ethylene, which Studies of solute solvatochromism are they attribute to hydrogen bonding perhaps the most widely applied spectroscopic approach to local density effects in interactions between phenol blue and flu- Figure 2. Excess solvent density in oroform. Morita and Kajimoto demon- the cybotactic region of a solution. supercritical fluids. The data are used to strated a similar local density enhancecorrelate spectral peak shifts in which lo(a) One solvent molecule is replaced by a solute (yellow) molecule to which the solvent cal composition and density of the fluid ment of supercritical ethane around dimis attracted, (b) Local density augmentation ethylaminobenzonitrile (DMABN) using surround a chromophoric solute moleresults from the presence of the solute cule. McRae theory (10,19) is used to cor- absorbance spectral shifts (22) This molecule in the compressible supercritical fluid. The shaded (blue) molecules have method has become quite common for relate solute spectral shifts Av to solvent diffused from outer regions into the voids characterizing density-dependent solvent properties. The results of McRae theory created by the attractive interaction around strengths of supercritical fluids (23,24)) predict that the frequency shift Av of abthe solute. Studies based onfluorescencespectral shifts have also been published. Kajimoto Analytical Chemistry News & Features, April 1, 1996 251 A
Report C02 pressure increases, the mole fraction of ethanol decreases from 10 mol% to 5 mol%, and a blue shift is observed in the spectrum. Yet analysis of the spectral shift indicates a local ethanol mole fraction of nearly 50%. These results are consistent with similar studies on phenol blue in C02/acetone mixed fluids (26). Pyrene
Solvatochromic studies are applicable only to solutes that demonstrate significant systematic spectral shifts with changing solvent environments. Such effects tend to be very strong in chromophores containing polar functional groups that have large charge asymmetry in the ground and/or excited states because of the interaction of the solute charge distribution with the solvent in the cybotactic region. The reaction field experienced by the solute results from the alignment of solvent molecules around the solute and is a direct COT!" of these interactions. Nonpolar solutes are often difficult to' probe by solvatochromic studies because their spectral shifts tend to be quite small. Yet a complete understanding of supercritical fluid clustering requires investigation of this phenomenon in a wide range of solvents and solutes. Several groups have examined the fluorescence spectrum of pyrene to probe solvent-solute interactions around a relatively nonpolar solute (6,27-34). Pyrene has a highly structured fluorescence spectrum containing five distinct vibronic bands. The intensity of the highest energy band (band 1) is known to be highly sensitive to the solvent environment, whereas the third band is not. The intensity ratio Py = / ] / / 3 (referred to as the Py solvent polarity scale) has been studied in numerous liquid solvents to establish an empirical correlation between Py and solvent dielectric properties. Investigations of Py in supercritical fluids indicate that significant clustering of the flui d s ftf*f*iit*Q firfiiind TIVFPTIP
Knutson and colleagues measured Py versus fluid bulk density of C02, and their results suggested a local density enrichment by as much as a factor of 2 over the bulk density in the compressible region (6). Molecular dynamics simulations agreed with these results when the local density was calculated only within thefirstsolva-
Figure 3. Fluorescence spectra of bis(4,4'-aminophenyl)sulfone dissolved in 5% (volume) ethanol/C0 2 .
tion shell. These results suggest that shortrange interactions are responsible for the sensitivity of Py to solvent polarity. Sun and co-workers made similar measurements and correlated the results with solvent bulk dielectric properties (28). The Py values calculated from bulk solvent properties agree with the measured Py values in the gaslike region and in the high-density region, but not in the compressible region. These results are similar to those of several solvatochromic studies, which suggest the possibility that clustering contributes to solvation of nonpolar solutes to a similar degree as polar solutes. Rice and colleagues observed similar behavior for pyrene in supercritical C0 2 and have extended these studies to examine differences in clustering in the ground and excited states of pyrene (27). They compared pressure-dependent absorption spectral shifts with Py measurements in supercritical C0 and conclude that the local density enhancement is significantly greater in the excited state of ovrene than in the ground state
252 A Analytical Chemistry News & Features, April 1, 1996
Rotational diffusion
Rotational diffusion experiments have also been used to study solvent clustering in supercritical fluids. Although these measurements are somewhat complex, they build on a conceptually simple phenomenon. When a solution containing a fluorescent chromophore is irradiated with polarized light, the light preferentially excites those molecules that happen to be oriented such that their absorption dipoles are aligned with the polarization vector of the light (11). Molecules that fluoresce immediately after absorption will preferentially emit light that is polarized parallel to the polarization vector of the excitation when their absorption and emission dipoles are parallel If the resulting fluorescence is collected through a polarizer oriented parallel to the pyritation the initial intensity will be high A low iniintensity however will be measured through a nolarizer oriented pernendiru lar to the e citation As time s the initiall
i t
i rT trihnf
fe
cited molecules will randomize because of diffusive rotational motion of the mole-
cules. The rate of this rotational diffusion is proportional to the local viscosity experienced by the solute molecules. Therefore, if one measures the polarization anisotropy of fluorescence from the solution, the local viscosity experienced by the solute can be determined (29,30). Betts and co-workers made rotational diffusion measurements of the polar chromophore (6-propionyl-2- (dimethylamine) naphthalene) in supercritical N 2 0 and reported a dramatic decrease in the rate of rotational diffusion in the compressible region, which they attributed to solvent clustering (29). They estimated that 25100 solvent molecules were required to form a cluster of sufficient size to cause this effect depending on the model used to interpret the results. Our lab has focused on developing rotational diffusion studies that can provide information regarding the influence of specific solvent-solute interactions on supercritical fluid clustering. We recently reported the results of a comparative densitydependent rotational diffusion study of diphenylbutadiene (DPB) and 4-hydroxymethylstilbene (HMS) (32). The DPB rotational diffusion time increases by 50% when the bulk density is increased from 0.3 g / cm3 to 0.8 g/cm 3 whereas the HMS rotational diffusion time increases by more than 300% over the same density range We developed a model based on Equation 1 that suggests that the local density around the HMS molecule is enriched by a factor of 1.4 over the bulk fluid density. Furthermore, because the average molecular features (e.g., solute dimensions, molecular volume, and number of JC electrons) of HMS and DPB are similar, these results indicate that interaction between the COo and the hydroxyl moiety of HMS is responsible for local density enrichment Future prospects Investigations such as these have already provided a wealth of evidence to support the existence of fluid clusters around solute molecules. Density fluctuations occur in pure supercritical fluids, so a particular region of space experiences density variations over time. Spectroscopic studies, however, indicate that significant density augmentation occurs around solutes, and these localized regions of high fluid density may not experience the
density variations that the bulk fluid experiences. Solvent clusters around solute molecules continuously exchange solvent molecules (31), but the time scale of the exchange and its relation to density variations in the pure fluid are issues that have not been fully explored. Rotational correlation time studies appear to have tremendous potential for contributing to our understanding of solvation in supercritical fluids. As theoretical and computational methods for the investigation of solvent-solute interactions advance, so will our ability to extract molecular-level information from studies of rotational diffusion and solvatochromism. Based on current information on local density effects in supercritical fluids, significant advances in our understandinc of
Supercritical fluids offer an economical alternative to standard liquid extraction methods. these systems can be anticipated over the next decade. Predictive models of local enrichment will be developed that can be applied not only to the study of solvation in supercritical fluids, but also to the examination of enrichment of one analyte around another in complex chemical systems. Such information will be important for improving supercritical fluid reaction kinetics and extractors. I would like to acknowledge the efforts of Robert Anderton and Richard Schulte in conducting spectroscopic investigations of supercritical fluid solvation in my lab. Our work has beer supported by the University of Missouri, the ACS Petroleum Research Fund, and the National Science Foundation (CHE-95-08744).
C. A; Debenedetti, P. G.; Chialvo, A A. ACS Symp. Ser. 1992, 488 (Supercrit. Fluid Technol.), 60-72. (7) Yonker, C. R; Frey, S. L; Kalkwarf, D. R; Smith, R. D.J. Phys. Chem. 1986, 90, 3022-26. (8) Schulte, R. D.; Kauffman, J. F.J. Phys. Chem. .994, 98,8793-8800. (9) Dobbs, J. M.; Wong, J. M.; Lahieree R J.; Johnston, K. P. Ind. Eng. Chem. Res. 1987,26, 56-65. (10) Mataga, N.; Kubota, T. Molecular Interactions and Electronic Spectra; Marcel DelD ker: New York, 1970. (11) Lakowicz, J. R Principles of Fluorescesce Spectroscopy; Plenum Press: NeN Yorko 1983. (12) Lee, L L. Molecular Thermodynamics of Nonideal Fluids; Butterworth: Boston, 1988. (13) Kurnik, R. T.; Reid, R. C. Fluid Phase Equil. 1982,8,93-105. (14) McQuarrie, D. A. Statistical Mechanics; Harper & Row: New York, ,9777 (15) Munoz, F.; Chimowitz, E. H. Fluid Phase Equil. 1992, 71,237-72. (16) Tom, J. W.; Debenedettt, P. G. .nd. Eng. Chem. Res. 1193,32,2118-28. (17) Chialvo, A A; Cummings, P. T. AIChEJ. 1994,40,1558-73. (18) Pfund, D. M.; Lee, L. L.; Cochran, H. DD Fluid Phass Equil. 1988,39,161-96. (19) McRae, E. G.J. Phys. Chem. 1957, 61, 562-69. (20) Yonker, C. R; Smith, R D.J. Phys. Chem. 1988,92,235-38. (21) Kim, S.; Johnston, K. P. Ind. Eng. Chem. Res. 1987,26,1206-13. (22) Morita, A.; Kajimoto, 0./. Phys. Chem. 1990,94,6420-27. (23) Phillips, D. J.; Brennecke, J. F. Ind. Eng. Chem. Res. 1193,32,943-51. (24) Bennett, G. E.; Johnston, K. P.J. Phys. Chem. 1994,98,441^47. (25) Kajimoto, O.; Futakami, M.; Kobayashi, T.; Yamasaki, K.J. Phys. Chem. 1988, 92, 1347-53. (26) Kim, S.; Johnston, K P. AIChEJ. 1987, 33,1603-15. (27) Rice, J. K; Niemeyer, E. D.; Dunbar, R. A; Bright, F. V.J. Am. Chem. Socc 1995, 117,5832-39. (28) Sun, Ya-P.; Bunker, C. E.; Hamilton, N. BB Chem. Phys. Lett. 1993,210,111-17. (29) Betts, T. A; Zagrobelny, J.; Bright, F. V.J. Am. Chem. Soc. 1192,114,8163-66. (30) Anderton, R. M.; Kauffman, J. F.J. Phys. Chem. .994,98,12117-24. (31) Eckert, C. A; Knutson, B. L. Fluid Phase Equil. 1993,83,93-100. (32) Anderton, R M.; Kauffman, J. F.J. Phys. Chem. .995, 99,13759-62.
References (1) Hawthorne, S.K. Anal Chem. .990,62, 633A-642 A John F. Kauffman is an assistant professor (2) Barnabas, I. J.; Dean, ,J R; Owen, S. P. An- at the University of Missouri. His research alyst 1994,119, 2381-94. interests include supercritical fluid solva(3) McHugh, M. A; Krukonis, V. J. Supercritical Fluid Extraction: Principles and Prac-tion, time-resolved fluorescence, and luminescent chemical sensors. Address corretice; Butterworth: Bosson, 1986. (4) Illman, D. L. Chem. Eng. News Sept. 5, spondence about this article to him at the 1994,22. University ofMissouri, Dept. of Chemistry, (5) Kirschner, E. M. Chem. Eng. News June 123 Chemistry Bldg.. Columbia, MO 65211 20,1994,13. (
[email protected]). (6) Knutson, B. L.; Tomasko, D. L.. Eckert, Analytical Chemistry News & Features, April 1, 1996 253 A