Steven 6. Hawthorne University oi NorIh Dakota I Energy and Environmental Research Center Grand Forks, NO 58202
Since the development of the transistor in 1948, analytical research has led to tremendous advances in instrumental techniques for organic analysis, particularly in the areas of chromatography and spectroscopy. Although some samples are inherently ready for analysis by these techniques, most require extraction of the analytes into a liquid solvent, especially when the analyst wishes to identify and quantitate minor or trace organic components from a bulk sample matrix. It is ironic that some modern chromatographic techniques have become so highly developed that they can separate, identify, and quantitate hundreds of sample components per hour, yet analysts often rely on sample extraction procedures-such as liquid solvent extraction in a Soxhlet apparatus-that were in use when Tswett first reported chromatography in 1906. An ideal extraction method should be rapid, simple, and inexpensive to perform; yield quantitative recovery of target analytes without loss or degra0003-2700/90/0382-633A/S02.50/0 @ 1990 American Chemical Society
dation; yield a sample that is immediately ready for analysis without additional concentration or class fractionation steps; and generate no additional laboratory wastes. Unfortunately, liquid solvent extractions frequently fail to meet these goals. They often require several hours or even days to perform, result in a dilute extract (which must be concentrated when trace analysis is desired), and may not result in quantitative recovery of target analytes. Recent concerns about the hazardous nature of many commonly used sol-
REPORT vents, the costs and environmental dangers of waste solvent disposal, and the emission of hazardous solvents into the atmosphere during sample concentration further support the development of alternative sample extraction methods. The limitations of conventional methods have fueled interest in the development of supercritical fluid extraction (SFE) as an alternative to extractions using liquid solvents. This REPORT describes current techniques and applications as well as the capabilities and limitations of using supercritical fluids for the analytical-scale ex-
traction and recovery of organic analytes from complex sample matrices such as environmental solids, polymeric resins, and biological tissues. Why use supercritical fluids for sample extraction?
A substance that is above its critical temperature and pressure is defined as a supercritical fluid. The combined gas-like mass transfer and liquid-like solvating characteristics of supercritical fluids have led to considerable excitement for their use as mobile phases for supercritical fluid chromatography (SFC) ( I ) , but their use for analyticalscale extraction has only recently received attention. For example, a computer search of Chemical Abstracts listings for ANALYTICAL CHEMISTRY and selected chromatography journals revealed only two papers on SFE prior to 1986, compared with 26 articles on this technique published from 1986 to mid-1989. Supercritical fluids are sometimes considered to be "super solvents," but this is not true when their solvating power is compared with that of liquids. The solvent strengths of supercritical fluids approach those of liquid solvents only as their density is increased, and the maximum solubility of an organic compound is frequently higher in liq-
ANALYTICAL CHEMISTRY, VOL. 62, NO. 11, JUNE 1. 1990
REPORT uid solvents than in supercritical fluids. Although supercritical fluids do not have any advantage over liquid sol-
vents in solvating power, comparison of several other characteristics of supercritical fluids with those of liquid sol-
vents demonstrates the potential for SFE to approach the idealized goals for analytical extraction. * SFE is fast. Mass transfer limitations ultimately determine the rate at which an extraction can be performed. Because supercritical fluids have solute diffusivities an order of magnitude higher (W4vs. cm2/s)and viscosities an order of magnitude lower (10-4 vs. N s/m2) than liquid solvents, they have much better mass transfer characteristics. Quantitative SFEs generally are complete in 10-60 min, whereas liquid solvent extraction times can range from several hours to days. The solvent strength of a supercritical fluid can easily be controlled. The solvent strength of a liquid is essentially constant regardless of extraction conditions, hut the solvent strength of a supercritical fluid depends on the pressure and temperature used for the extraction, as shown in Figure 1by the dependence of the Hildehrand soluhility parameter on temperature and pressure for supercritical COn (2).At a constant temperature, extraction at lower pressures will favor less polar analytes, and extraction at higher pressures will favor more polar and higher molecular weight analytes. This allows an extraction to be optimized for a particular compound class hy simply changing the pressure (and, to a lesser extent, the temperature) of the extraction. As discussed helow, this same characteristic allows class-selective extractions to be performed by extracting a single sample at different pressures. Many supercritical fluids are gases at ambient conditions. Liquid solvent extracts need to he concentrated prior to the determination of trace organic analytes, a step that requires additional time and can result in the loss of more volatile analytes. In contrast, hecause supermitical fluids are gases at amhient conditions, concentration steps after SFE are greatly simplified and direct coupling of the SFE step to chromatographic techniques is facilitated. Supercritical fluids have additional practical advantages. Most are relatively inert, pure, nontoxic, and inexpensive (on a per-extraction basis). Because fluids such as COSand N20 have relatively low critical temperatures (31 O C and 36 "C, respectively), SFE can he performed at low temperatures to extract thermally unstable compounds. The generation of liquid waste solvents and exposure of laboratory personnel to toxic solvents also can he reduced or eliminated.
Temperature CC) Figure 1. Effect of temperature and pressure on the Hlldebranc for supercritical C02. Torepresents uilicai temperature. (Adaptedlrom Rel~rence2.)
rrgure 2. Schematic of an SFE system. The 8uperunicai liuid (rW is pumped Io lhe exbaction vessel whsre lho analyies (purple) are exmcted lmm the sample m m x (brown). The anaiyies are then swept through me flow reseictor into the MIIBCtion device, and me depressurizedSweroritiCal liuid (now a gas. for most fluids) is vented.
ANALYTICAL CHEMISTRY, VOL. 62, NO. 11. JUNE 1, 1990
Experlmentai c o l l s i d e r a l i i A frequent misconception is that, hecause SFE and SFC both use supercri-
tical fluids, SFE must somehow rely on subsequent analysis using SFC techniques. Although analysis of supercritical fluid extracts using SFC is an option, SFE is an independent sample preparation method (analogous to Liquid solvent extraction), and supercritical fluid extracts can be analyzed by any technique that is appropriate for the extracted analytes. In addition to chromatographic methods, supercritical fluid extracts are frequently analyzed by spectroscopic, electrochemical, radiochemical, gravimetric, and other techniques. SFE is conceptually simple to perform (Figure 2). A pump is used to supply a known pressure of the extraction fluid to the extraction vessel, which is placed in a heater to maintain the vessel at a temperature above the critical temperature of the supercritical fluid. During the extraction, the soluble analytes are partitioned from the bulk sample matrix into the supercritical fluid, then swept through a flow restrictor into a collection device that is normally at ambient pressure. The fluids used for SFE are usually gases at ambient conditions and are vented from the collection device while the extraded analytes are retained. Several experimental variables must be considered and optimized for SFE to be successful,including the choice of supercritical fluid, pressure and temperature conditions, extraction time, sample size, the method used to collect the extracted analytes, and the equipment that is needed. Each of these aspects is discussed below. Choosing supercrnial fluids and extraction conditions
Predicting optimal extraction conditions. Perhaps because of the historical development of SFE in process engineering, the choice of analytical SFE conditions has most often been based on the pressure and temperature conditions where the target analytes have their highest solubility in the supercritical fluid. This condition can be approximated if the solubility parameter of the analyte is known and if certain correlations are used, such as one proposed by Giddings et al. (3):
(e.g., the extraction of fats from meat products), but because they are concerned with maximum solubilities, they become less useful when the concentrations of the target analytes are at minor and trace levels. For such samples, dissolving the maximum amount of the target analyte in the supercritical fluid is not of concern, and the analytes need only be soluble enough in the supercritical fluid to be transported out of the extraction vessel. King ( 4 ) suggests combining maximum solubility conditions with the threshold pressure (i.e., the pressure where the analyte becomes significantly soluble) to determine the range of pressures (or fluid densities) that are applicable to a particular extraction. Solubility data and correlations provide useful information for choosing initial SFE conditions. However, determining optimal extraction conditions for minor and trace components has been largely empirical, for two reasons. First, analytical SFE often involves the recovery of a complex variety of analytes (rather than a single target anaIyte). In such cases, extractions must be optimized for groups of compounds, which complicates the prediction of optimal extraction conditions. Second, and more importantly, solubility considerations address only part of the extraction problem. Because extraction of an analyte depends on its distribution between the supercritical fluid and the sorptive sites on the sample matrix, the ability of the supercritical fluid to compete with the analytes for the sorptive sites may be more important than solubility considerations for determining optimal extraction conditions. This is indicated by the comparative extractions using COZand NzO, discussed below. Unfortunately, the importance of the competition for active sites between the analytes and the supercritical fluid has received very little attention in the development of analytical SFE techniques (5).
Understanding the mechanisms and interactions among the matrix surface, the analytes, and the supercritical fluid, as well as the kinetics that control these interactions, will be important to the future design of analytical SFE methods. Suitable models of extraction mechanisms should be particularly useful for designing extractions using modified supercritical fluids, because it may be possible to select modifiers that compete best with the analytes for sorptive sites on the matrix surface as well as modifiers that increase analyte solubility in the supercritical fluid. Choosing a supercritical fluid. The characteristics of several fluids that have been used for SFE are listed in Table I. (Note that the Hildebrand solubility parameters listed in Table I are maximum values that are approached at very high pressures, as demonstrated in Figure 1 for CO2. Because of hardware considerations, SFE is normally performed at pressures that yield lower solvent strengths.) Although fluids with a range of polarities are available to optimize an extraction based on the polarity of the target analyte @),the choice of a fluid for analytical SFE frequently depends on practical considerations. Supercritical COZ has been the choice for most SFE studies, primarily because of attractive practical characteristics: relatively low critical temperature and pressure, low toxicity and reactivity, and high purity at low cost. Unfortunately, supercritical COz does not have sufficient solvent strength at typical working pressures (80-600 atm) to quantitatively extract analytes that are quite polar. In general terms, COzis an excellent extraction medium for nonpolar species such as alkanes and terpenes [6 - 6 8 (~al/cm~)’/~]. It is reasonably good for moderately polar species, including polycyclic aromatic hydrocarbons (PAHs), polychlorinated biphenyls (PCBs), aldehydes, esters,
6 = 1.25 P:”(p/pJ
where 6 is the Hildebrand solubility parameter, Pc is the critical pressure of the fluid, p is the density of supercritical fluid, and pt is the density of the fluid in its liquid state. Such correlations are quite useful when the target analytes represent a large percentage of the bulk sample ANALYTICAL
CHEMISTRY, VOL. 62. NO. 1 1 , JUNE I. 1990
REPORT alcohols, organochlorine pesticides, and fats (6 -&11), but is less useful for more polar compounds.
The results reported in the literature and our own experiences indicate that, as rule of thumb, compounds that can
be analyzed by conventional gas chromatographic techniques can be quantitatively extracted using supercritical Con. Some additional nonpolar organics, such as triglycerides (which have vapor pressures too low to be amenable to GC), are also easily extracted with
hvure 3. Comparison of the extraction efficiencies obtained using C02 and N20. me P A W chrywne and in&no[l.2.3-cd]pyrene. were enraCted for 10 rnin from marine sediment (6). The telrachlwodlbenzc-pdloxlns (TCWs) were enracled for 2 h horn municipal incineralor fly ash (7). TCOD eXlIaCtlonS Wldl CO2 We shown before and after acid besrment of the fly ash.
Flgure 4. Comparison of extraction efficiencies obtained using fied with methanol.
ANALYTICAL CHEMISTRY, VOL. 62. NO. 11, JUNE 1, 1990
A somewhat frustrating problem arises when fairly polar analytes need to be extracted. The first choice would be a fluid with higher solvent strength, hut the use of more polar fluids is severely limited by practical considerations. Supercritical ammonia would be very attractive from a solvent strength standpoint, but it is difficult to pump (it tends to dissolve pump seals), is chemically reactive, and is likely to be too dangerous for routine use. Supercritical methanol is also an excellent solvent but is less attractive because of its high critical temperature and because it is a liquid a t ambient conditions, which complicates sample concentration after extraction. For some analytelmatrix combinations, extraction efficiencies can be increased using fluids that would not be expected to yield better efficiencies based on their solubility parameters. For example, N20 gives more rapid extractions of some samples than can be achieved with COz (7,8 ) as demonstrated in Figure 3 for tetrachlorodibenzo-p-dioxins (TCDDs) and PAHs. This result is surprising given the similarity of NzO and CO2 in physical properties and solubility parameters (shown in Table I). (N20 does have a small permanent dipole moment; COS does not.) The results of Alexandrou and Pawliszyn (7) for the SFE of TCDDs from incinerator fly ash are particularly interesting. N20 yielded good recoveries from untreated fly ash, whereas COz yielded good recoveries only after the fly ash surface was m d i fied by acid treatment, demonstrating that displacement of the analytes from the sorptive sites on the fly ash-and not differences in their solubility-was responsible for the increased extract i o i efficiencies demonstrated by NzO. Because of the practical difficulties in using polar fluids such as ammonia for SFE, extractions of highly polar analytes have most often been done using CO2 containing a few percent added organic modifier. Modifiers can be introduced as mixed fluids in the pumping system, with the aid of a second pump, or by simply injecting the modifier as a liquid onto the sample before beginning the extraction. Figure 4 shows the increase in extraction efficiencies obtained in 30 min by adding a methanol modifier for extrac-
the $2,995 DL12.
The D112 Autotitrator. Because no two people see things exactly the same, You might see two profiles. Someone else, a crystal goblet. Which illustrates the fact that subjective judgments are hardly the stuff of consistent analysis. Unfortunately for many smaller labs, objective pH determination hasjust been too costly. Enter the DL12-a low-cost autotitrator that brings you METTLER accuracy at an unheard-of price. And exit the subjective determination of pH end points. With a crystal-cleardisplay, the DL12 brings any lab or quality control depallment expanded pH measurement wpabilities. Offen static measurementsto ~k.002units. And records results in user-selected increments. But accuracy is only half the picture. The METTLER DL12 is also foster and more cost effective than manual burettes. SO irs perfect for anyone involved in repetitive titrations for both in-process and end-product quality control. Just call 1-800-METTLERond we'll have you seeing eye-to-eye in no time.
CIRCLE 88 ON READER SERVICE CARD
Metller Instrument Corpomlion BOX 71 Hightstown, NJ 08520 1-800-METTLER (NJ 1-609-448-3000' 1
We understand Precisely
REPORT tions where the recovery using pure COz was not quantitative (9-12). In each case the extraction efficiencies increased dramatically, and even the ionic compound linear alkylbenzenesulfonate was quantitatively recovered. Although methanol has been the most widely used modifier, a variety of organic compounds have been used, ranging from alcohols, propylene carbonate, 2-methoxyethanol, and organic acids to methylene chloride and carbon disulfide. The selection of modifiers and their concentrations has been largely empirical because very little analyte solubility data exist for modified supercritical fluids. In addition, the competitive interactions between the modified supercritical fluid and the target analytes with the sorptive sites on the bulk matrix are poorly understood. Clues can be gleaned from SFC separations that have used modified COZ, and a reasonable starting point would be to select a modifier that is a good solvent in its liquid state for the target analyte. However, until the extraction mechanisms that control the distribution of the analytes between the matrix and the bulk supercritical fluid are better understood, optimization of SFE methods using modified fluids will frequently require testing modifiers with different polarities and concentrations as well as determining optimal temperature and pressure conditions. Class-selective extractions. As previously discussed, the solvent strength of a supercritical fluid can easily be controlled by changing the pressure (and, to a lesser extent, the temperature), which makes it possible to perform selective extractions with a single supercritical fluid by varying the extraction conditions. Although obtaining a pure analyte from complex matrices using SFE is unlikely, sequential extractions of different compound classes have been demonstrated. For example, alkanes can be extracted from urban air particulates with COZ at 7 5 atm (45 OC) whereas the PAHs remain unextracted until the pressure is raised to 300 atm. By sequentially extracting the air particulates at these two pressures, 85-90% selectivities can be achieved (13). In some cases, proper selection of extraction conditions can also be used to extract analytes from bulk matrix material that is itself soluble in the supercritical fluid under different conditions. For example, even though fat components are highly soluble in supercritical COz under most conditions, extracts containing organochlorine pesticides that are sufficiently fat-free to allow direct gas chromatographic 838A
analysis have been prepared by extraction at a pressure under the threshold solubility pressure (-120 atm at 40 OC) of the bulk fat matrix (4,141. Class-selective extractions have also been achieved by depositing the analytes onto a sorbent, then eluting them with supercritical fluids (although this may more properly be called SFC than SFE). This technique has been successfully coupled with GC (SFE/GC) for the class fractionation of alkanes, alkenes, and aromatics from gasoline using supercritical COz elution of each compound class from a silica column and transferring each compound class to a capillary gas chromatograph for separation of the individual components (15,16).Many other variations of class-selective SFE are possible, and a better understanding of extraction mechanisms will facilitate the development of class-selective SFE methods. SFE techniques and hardware Hardware requirements. Commercial extraction units are available for performing analytical SFE, and systems can easily be assembled in the laboratory. Analytical SFE is typically performed using syringe pumps that are similar or identical to those used for SFC, although less expensive alternatives are available because SFE is normally performed at constant pressures without the sophisticated pressure/ density ramp controllers required for SFC. Gas compressors are also useful (e.g., for COS), particularly for larger scale extractions (e.g., samples >50 g) for which a syringe pump may have insufficient capacity. Extraction cells can be purchased or constructed from common tubing fittings, but care must be taken to ensure that all of the extraction system components are rated for the working pressures. The temperature of the extraction cell is normally controlled by placing the cell in a chromatographic oven or in a simple tube heater. SFE can be performed in either a dynamic or a static mode. For dynamic SFE, the supercritical fluid is constantly flowing through the cell, and a flow restrictor is used to maintain pressure in the extraction vessel and allow the supercritical fluid to depressurize into the collection device. Both micrometering valves and short lengths of fused-silica tubing (-10-50-lm i.d.) have been widely used for flow restrictors. Static SFE is performed by pressurizing the cell and extracting the sample with no outflow of the supercritical fluid. After a set period of time, a valve (e.g., an HPLC switching valve) is opened to allow the analytes to be swept into the collection device. One
ANALYTICAL CHEMISTRY, VOL. 62, NO. 11, JUNE 1, 1990
advantage of dynamic SFE is that the supercritical fluid is constantly renewed during the extraction, but this technique requires more fluid than static SFE, particularly when extracting large samples. Specific criteria for selecting dynamic or static modes are not yet clear, and both methods have been used for quantitative SFE of a variety of samples. Sample size considerations. Although analytical SFE has been used with samples ranging from 1 mg to hundreds of grams, SFE of samples (10 g has been most common because larger samples require larger amounts of supercritical fluid for quantitative extraction, which can make quantitative trapping of the extracted analytes more difficult. Analytical SFE is most often conducted using fluids that are gases at ambient conditions; thus, the success of the trapping method depends on recovering the analytes from the expanded gas flow upon depressurization (e.g., a supercritical COz flow of 1mL/min results in a gas flow of 500 mL/min). In general, quantitative collection of the extracted analytes is simpler at lower extraction flow rates, particularly when the analytes are volatile. As discussed below, quantitative collection of relatively volatile analytes is convenient with supercritical fluid flow rates of up to at least 1 mL/min. At these rates, assuming that a typical sample has a void volume of -30%, passing 10 void volumes of extraction fluid through 1-,lo-, and 100-gsamples would require about 3,30, and 300 min, respectively. Such flow considerations indicate that to reduce extraction times, SFE of small samples is preferable unless larger samples are necessary to ensure sample homogeneity or to obtain sufficient sensitivity for trace analytes. (These same considerations indicate that extraction cells >1mL should be packed full of the sample to avoid significant void volumes in the extraction system, which would increase extraction times.) Fortunately, several approaches have been reported that are capable of quantitatively retaining extracted analytes under a wide variety of SFE conditions and sample sizes. Off-line vs. on-line collection techniques. Methods that have been used to collect extracted analytes upon depressurization of the supercritical fluid fall into two general categories: “off-line” SFE (analytes are collected for subsequent analysis) and “on-line” or “coupled” SFE (analytes are directly transferred to a chromatographic system). Off-line SFE is inherently simpler to perform because only the extraction step must be understood, and
the extract can be analyzed by any appropriate method. On-line SFE requires an understanding of both the SFE and the chromatographic conditions, and the sample extract is not available for analysis by a different method. The principal advantages of on-line SFE =&be elimination of sample handling between extraction and chromatographic analysis and the potential to achieve maximum sensitivity by quantitatively transferring the extracted analytes into the cbromatographic column. With off-line SFE, the analytes are most often collected in a few milliliters of a liquid solvent, and the analysis of the extract is conducted as it would be for any conventional liquid solvent extract. The cooling of the solvent caused by the expanding supercritical fluid prevents rapid evaporation of the collection solvent that would be expected because of the high gas flows, and even relatively volatile analytes are quantitatively recovered (e.g., decane is quantitatively retained in 3 mL of methylene chloride during a 15-min extraction using -1 mL/min of supercritical COn [ I Z ] ) .SFE effluents have also been collected on sorbents such as silica or bonded-phase packings and then eluted with liquid solvents or supercritical fluids for subsequent analysis. When large quantities of bulk matrix are extracted (e.g., fat from meats), direct depressurization into an empty receiving vessel has been successful (1 7),but with trace analytes, trapping in empty vessels (even with cooling) bas been
less successful because of losses from aerosol formation (9). Off-line SFE has been most extensively studied as a processing technique, but the use of SFE as a quantitative analytical procedure bas also been successful for a variety of analytes. All of the extractions listed in Table I1 yielded quantitative recovery of the target analytes, and the majority were completed in