SOLID-PHASE EXTRACTION OF POLAR ORGANIC POLLUTANTS

ES&T

M

ost organic c o n t a m i nants are present in environmental waters in trace a m o u n t s at t h e pg/L level and lower. As a result, very low detection limits are required for directly monitoring drinking water or studying the fate and transport of organics directly in the environment. In addition, many polar organic compounds cannot be analyzed at a trace level in water samples because of the lack of simple methods for their extraction. In general, water samples are too dilute and too complex to be analyzed by gas chromatography (GC) or liquid chromatography (LC) without some p r e l i m i n a r y sample preparation: extracting traces of the organic of interest from the aqueous media, concentrating these traces, and often removing other components from the matrix that have been coextracted and may interfere with the chromatographic analysis (cleanup). Recently interest has grown in trace enrichment techniques that use solid-phase extraction (SPE) as an alternative to the laborious and time-consuming liquid-liquid extractions (LLE). Moreover, recoveries of many polar analytes obtained using LLE are low because of their relatively high partial solubility in water. Automatic devices are now commercially available for the on-line coupling of SPE to GC or LC. Nevertheless, SPE does not appear as straightforward as LLE to many environmental chemists because of the many sorbent choices and because recoveries depend on the sample volume. We will discuss the similarity between the extraction process with apolar sorbents and classical elution LC. This similarity allows most of the parameters for SPE to be obtained from LC data. Some wellknown methods that predict retention in LC can also be applied to SPE and help in the choice of a sor-

SPE cartridge or it can be aspirated through the cartridge by vacuum. Various vacuum manifolds allow batches of up to 24 samples to be prepared simultaneously. Automation of the whole off-line SPE procedure is now possible using commercially available sample preparation workstations that perform each step sequentially. Compared to LLE, the off-line SPE technique is faster and saves a substantial amount of solvent. Samples can even be percolated through a sorbent in the field, avoiding the transporting and storing of voluminous samples. These adsorbed analytes generally store well. In addition, automated units reduce the risk of sample contamination. Neve r t h e l e s s , off-line procedures have the inherent disadvantage of losses during the evaporation step and loss of enDescription of the SPE procedure richment factor resulting from the SPE can be used off-line (i.e., the injection of an aliquot. sample preparation is completely On-line coupling of SPE to a GC or separated from t h e subsequent LC instrument avoids many of these chromatographic analysis) or onproblems. On-line coupling of SPE to line (i.e., it is directly connected to the chromatographic system) [1, 2). LC is particularly easy to perform in any laboratory (see box, p. 579A). In off-line methodologies, samples More accurate quantitative results are percolated through a sorbent can be expected because there is no packed in disposable cartridges or enmeshed in an inert matrix of a sample manipulation between the preconcentration and the analysis membrane-based extraction disk. A steps. The sample volume can be typical off-line SPE sequence is delower than off-line SPE because the scribed in the box (p. 578A). entire sample is transferred and anaDisposable cartridges contain a lyzed. The SPE sequence runs dursorbent bed that varies from 100 mg ing the analysis time of the previous to 1000 mg. The sample volume that sample. Automatic devices are now can be applied ranges from one to commercially available. several hundreds of milliliters; the The sorbents, chemistry, and large volumes are for cartridges deprinciples are identical in off-line signed with large reservoirs. The sample can be processed by injecting and on-line SPE. The only differences are in the packing size of sorthe contents of a syringe onto the bents—larger in off-line cartridges (40-60 pm) than in on-line precolu m n s ( 5 - 1 0 pm LC-grade sorM A R I E - C L A I R E H E N N I O N bents)—and the amount of sorbent, which may be 1000 mg or more in VALÉRIE PICHON off-line cartridges. The size of a preEcole Supérieure de Physique et de column cannot be increased beChimie Industrielles de Paris cause this will lead to band broadLaboratoire de Chimie Analytique ening during on-line elution. The 75231 Paris cedex 05, France amount of sorbent is less than 20 mg

bent. We will also present the limits of the common sorbent C18 silica for the preconcentration of polar organic (as related to t h e water— octanol partition coefficients of analytes). Finally, the potential of two other reversed-phase sorbents for extracting more polar organic is examined.

SOLI D-PHASE EXTRACTION OF POLAR ORGANIC

POLLUTANTS

FROM WATER

576 A

Environ. Sci. Technol., Vol. 28, No. 13, 1994

0013-936X/94/0927-576A$04.50/0 © 1994 American Chemical Society

in the classical precolumns (1 cm χ 0.2 cm i.d.) and, in general, less than 100 mg. Basic principles As the development of a SPE pro­ cedure is often an empirical process for those expert only in LLEs, we will give a simple approach to mod­ eling the SPE process, which can help select a convenient sorbent. To a first approximation, SPE can be considered as a simple liquid chromatographic process. The sor­ bent is the stationary phase, and the mobile phase is the water in the aqueous sample during the extrac­ tion step or the organic solvent dur­ ing the elution step. Organic com­ pounds that do not elute with the water are trapped on the sorbent. High enrichment factors are ob­ tained when analytes are strongly

retained by the sorbent in the pres­ ence of water and when there is a low retention with the eluting or­ ganic solvent. The choice of the sorbent is guided by the aqueous nature of the sample. Water should not easily elute the target organic compound. Thus bare silica and silica modified with polar groups are not usually good sorbents because water is the strongest eluting mobile phase. Bet­ ter sorbents are reversed-phase sta­ tionary phases for neutral organics and ion-exchangers for ionic com­ pounds. The main sorbents for re­ taining organic compounds in aque­ ous samples are reported in Table 1 with the corresponding separation mechanisms, the nature of the elu­ tion solvent, and the characteristics of the organic compounds that are preconcentrated.

These three reversed-phase sor­ bents will be compared in this study. The use of ion-exchangers requires some special attention because natu­ ral waters contain several milligrams of inorganic ions that very quickly overload the capacity of ion-ex­ changer sorbent. For this reason, we will not consider ion-exchange sor­ bents in the present study. Breakthrough. Breakthrough oc­ curs either when solutes are no longer retained by the sorbent or when the capacity of the sorbent has been overloaded. This latter reason is rather unlikely to occur in practi­ cal environmental analysis where concentrations are typically of the order of the pg/L level. Thus break­ through is mostly caused by insuffi­ cient retention. Breakthrough volume can be mea­ sured by monitoring the UV signal

Environ. Sci. Technol., Vol. 28, No. 13, 1994

577 A

Solid-Phase Extraction Steps

Λ

Typical steps in an SPE sequence are as follows: preparation of the sorbent, application of the sample, an eventual cleanup, and elution of the concen­ trated analytes. An example sequence is given for a cartridge packed with C18 silica: 1. The sorbent is activated with 3 - 5 mL of methanol and conditioned with the same volume of deionized water (it is important not to allow packing to dry out before adding the sample). 2. The aqueous sample is applied and trace organics of interest (+) are trapped by the sorbent while the water passes through. 3. Very often, other components (o) of the samples are either not retained by the sorbent or trapped with the analytes. A cleanup step can be added. Some of these interferences can be removed by applying 1-2 mL of a washing so­ lution made of water with a small amount of an organic solvent. 4. In the last step, the sorbent is dried for several minutes by vacuum suction. Then the concentrated analytes are eluted with 1 - 5 mL of organic solvent. It is recommended that the solvent wet the sorbent for 1 min and that the elution proceed at a dropwise rate.

of a water sample spiked with traces of a solute, S, which has an initial absorbance A0. The spiked sample is percolated through a precolumn. If the compound is retained by the sorbent, the effluent will not con­ tain it and the UV absorbance will be zero. A frontal or breakthrough curve is recorded beginning at a vol­ ume, Vb, usually defined as 1% of A0, up to a volume, Vm, defined as 99% of A0, where the effluent has the same composition as the spiked water sample (Figure la). Under ideal conditions, this curve has a bilogarithmic shape, the inflection point of which is the retention vol­ ume, Vr, of the analyte. When using the same precolumn in elution chromatography with water as the mobile phase and with the same flow rate, the injection of 10 or 20 pL of a concentrated solution of the same compound S will generate a peak detected at the same volume Vr as represented in Figure lb. The quantity Vb is a key parame­ ter for the preconcentration of the analyte and can be estimated to a 578 A

first approximation from Vr. The similarities between SPE and LC in­ dicate that data generated by LC for measuring or estimating VT are use­ ful. The knowledge of the retention behavior of analytes with hydro­ phobic sorbents should be applied. For example, ionizable compounds are only retained on C18 silica in their neutral form, so that the pH of the water sample is an important parameter, especially for the extrac­ tion of weak acids and bases. In practice, one first needs an ap­ proximate value of Vh in order to se­ lect an appropriate sorbent and the amount of the sorbent for off-line preconcentration with cartridges. More information on the relation between Vb and VT can be found in the box. Recoveries. Recovery is defined as the ratio between the amount ex­ tracted to the amount percolated and is theoretically 100% only for a sample equal to or lower than Vb. The maximum amount preconcentrated is reached at a sample vol­ ume of Vm (hatched area in Figure

Environ. Sci. Technol., Vol. 28, No. 13, 1994

la) and does not correspond to a 100% recovery. Therefore, the re­ covery in SPE depends on the sam­ ple volume and on the break­ through volume, related to the amount and the nature of the sor­ bent. Recoveries obtained with the same sorbent can be compared from one experiment with another only if the amount of sorbent and sample volume are known. It is always pos­ sible to obtain a 100% recovery by decreasing the sample volume be­ low the Vb value, and a simple cal­ culation indicates whether this vol­ ume will allow the r e q u i r e d detection or not. Limit of quantification. The abso­ lute detection limit of the chromato­ graphic detection system is usually expressed in pg or ng injected (e.g., 10 ng). Therefore, it is easy to calcu­ late the limit of quantification, C lim , which corresponds to the maximum amount that can be preconcentrated (hatched area in Figure la) or as equal to the product C lim χ Vr. If Vr is 100 mL, for example, then the concentration limit Clim is 100 ng/L. If a lower limit of concentration is required, the only remedy is to in­ crease Vb (or VT) by increasing the amount of sorbent or by selecting another sorbent that will provide a higher retention in the presence of water for the analyte of interest. Determination and prediction of breakthrough volumes. Recording breakthrough curves is time con­ suming, and reading Vh at 1% of A0 is difficult and not always accurate. Moreover, the sample should be spiked at a trace level in order to not overload the sorbent capacity, and the UV signal of the effluent should be monitored at very low absorbances (which may lead to prob­ lems with baseline stability or noise). Moreover, many compounds are difficult, and sometimes impos­ sible, to record because they have poor UV properties. A faster method that is easily per­ formed using the on-line setup con­ sists of preconcentrating water sam­ ples of increasing volumes, each containing the same amount of ana­ lytes, and then measuring the peak areas or heights eluted on-line from a precolumn (4). As the sample vol­ ume increases, the analyte concen­ tration decreases. Provided that breakthrough does not occur, the amount preconcentrated remains constant and the peak areas in the on-line chromatograms following elution are constant. When breakthrough occurs, the

On-Line Coupling of Solid-Phase Extraction HPLC pump Mobile phase

Precolumn Preconcentration pump

^

\ Analytical column

Waste Sample

Τ

A simple on-line setup consists of the following: The sorbent is packed in a small precolumn, which is placed in the sample loop position of a six-port switching valve. 1. The conditioning, sample application, and eventual cleanup occur via a simple pump in the injection valve in the "load position" (—). 2. The precolumn is coupled to an analytical column by switching the valve to the "injection position" ( — ) . The adsorbed compounds are then eluted di­ rectly from the precolumn onto the analytical column using a suitable mobile phase, which also enables the chromatographic separation of the trapped compounds.

TABLE 1

Sorbents that can be used for the solid-phase extraction of organic compounds present at a trace level in aqueous samples, LC separation mechanism involved, nature of the eluting solvent, and characteristics of the organic compounds that can be extracted from a sufficient water sample volume for trace-level determination Sorbent

Separation mechanism

Elution solvent

Alkyl bonded silicas (C8, C18)

Reversed-phase Organic solvents (methanol, acetonitrile . ..) Apolar styreneReversed-phase Organic solvents divinylbenzene (methanol, copolymers acetonitrile .. .) Graphitized carbons Reversed-phase Organic solvents (methanol, acetonitrile, THF . . .) Ion-exchangers Ion-exchange Water (pH adjusted in order that compounds are in their neutral form)

amount extracted decreases and the elution peak area decreases. Corre­ sponding recoveries can be calcu­ lated by dividing peak areas ob­ tained after breakthrough by those obtained before. This is shown in Figure 2. An advantage of this method is that the Vb values of sev­ eral compounds can be estimated simultaneously by preconcentra­ tion and on-line LC analysis under the real experimental conditions of

Extractable analytes Nonpolar to weakly moderately polar neutral compounds Nonpolar and moderately polar neutral compounds Nonpolar to polar neutral compounds Cationic and anionic organic compounds (pH of the water sample adjusted)

unknown sample analysis. Another method approximates the breakthrough volume using Vt, which is related to chromatographic data and precolumn characteristics by the equation Vr = V0(l + k'^), where V0 is the void volume of the precolumn or the cartridge and kw is the capacity factor of the solute eluted by water. V0 can be calcu­ lated from the porosity of the sor­ bent and the geometric volume of

the precolumn or sorbent bed in the cartridge. Values of k'w are esti­ mated from chromatographic mea­ surements using analytical columns packed with reversed-phase sor­ bents, such as C18 silicas, which are eluted with mobile phases com­ posed of water—methanol mixtures. The advantage of this method is that experimental data are obtained rap­ idly by measuring the capacity fac­ tor k" of the analyte in methanol— water p h a s e s . Over a range of methanol concentrations, often be­ tween 15% and 90%, there is a lin­ ear relationship between the loga­ rithm of the k1 and the percentage of methanol. This has been observed for alkyl silicas, apolar styrene— divinylbenzene copolymers PRP-1 (from Hamilton) or PLRP-S (from Polymer Laboratories), and porous graphitic carbon (PGC) (as shown in Figure 3 for phenol) (6). From rapid measurements with three or four mobile phases contain­ ing different methanol concentra­ tions, k'w can be estimated by graphically extrapolating to zero methanol content. In Figure 3, the experimental values of log k'm should be similar to predicted val­ ues because the relation is linear over the entire range. WerkhovenGoewie et al. [3) found good agree­ ment between Vr values for some chlorophenols derived from experi­ mental breakthrough curves and values calculated from extrapolated λ/νν values. Differences were on the order of 10-20%. However, according to Schoenmakers et al. (7), the relationship be­ tween log k1 and the methanol con­ tent is a better fit with a quadratic relationship for some compounds. Therefore, the extrapolated value is only an approximate value, but for trace enrichment studies this is ac­ ceptable. As an example, the log ^ w value of cyanuric acid was esti­ mated to 2.5 ± 0.2 using this method, which gave a Vr value be­ tween 30 and 70 mL for the precol­ umn used in Figure 2 (assuming a porosity of 0.75 for the porous carbon). The experimental Vb was measured as 50 ± 5 mL in Figure 2. Another aspect of the variations observed in Figure 3 is that the breakthrough volume decreases when some methanol is added. For phenol, using a cartridge packed with 100 mg of PRP-1, the break­ through volume should be 30 mL with water, 20 mL with water con­ taining 5% methanol, and 13 mL with 10% methanol. A small per­ centage of organic solvent can be

Environ. Sci. Technol., Vol. 28, No. 13, 1994

579 A

FIGURE 1

Similarity between SPE and elution chromatography (a) Breakthrough curve obtained by recording the UV signal of the effluent from a precolumn when percolating a water sample spiked with traces of a solute, S. The absorbance of the spiked solution is A^and was measured by direct connection of this solution to the UV detector.The breakthrough volume is usually defined as 1 % and Vm at 99% of the initial absorbance. V, is measured at the inflection point of the curve. The shaded area shows the maximum amount that can be preconcentrated. UV response Precolumn Detector Water spiked with traces (Ug/L) of S

v, Solid-phase extraction

vm Sample volume

(b) Elution peak obtained by injecting 10 \xL of a concentrated solution of S in the precolumn and eluted with a mobile phase of only water. The flow rate is the same as in (a), and the retention volume of the peak is Vr UV response Water

Precolumn Detector

Direct injection (K^Lofa concentrated solution of S) Elution chromatography

added to the raw sample without decreasing the recovery only if the retention is very high and if the sample volume percolated is less than the Vr value calculated from the corresponding log k" for this or­ ganic solvent concentration. Choice of the sorbent depending on the polarity of analytes The frontier between polar and nonpolar solutes is undefined. A parameter that characterizes the hydrophobicity and reflects the polar­ ity of a compound is its water—octanol partition coefficient, P oct . This parameter plays an important role in correlating phenomena of physicochemical, biological, and envi­ ronmental interest. Many log P o c t values are available in the litera­ ture, and calculation methods have been reported [8-10). We character­ ize solutes with log P o c t values above 3 as apolar, between 1 and 3 as moderately polar, and below 1 as polar. When log P o c t is less than zero, the compounds are more solu­ 580 A

Volume

ble in water than in organic sol­ vents. n-Alkyl silicas. Most of the off­ line environmental applications use SPE cartridges packed with η-alky 1 silicas (11). Because C18 silicas usually provide higher retention than C8 silica, C18 silicas should be used for extracting polar organics. Available cartridges packed with C18 silicas have different character­ istics and it is well known in LC that retention differs from one to an­ other. This is because retention de­ pends on the number of C18 chains bonded at the surface of the silica. Nevertheless, except for those with a low carbon content, kf^ is compa­ rable within a 20% variation for most C18 silicas, and that is accept­ able. B r a u m a n n (12) has gathered many log kf^ values that have been obtained with different C18 silicas u s i n g m e t h a n o l - w a t e r mobile phases. A linear relation was found between the average log A^w values and log P oct for closely related com­

Environ. Sci. Technol., Vol. 28, No. 13, 1994

pounds and even for compounds having different polarity and chem­ ical properties. For example, 60 compounds covering a wide range of structures from polar aniline (log P oct = 0.91) to the very hydrophobic ρ,ρ'-DDT (log P oct = 6.2) are related by log ^ w = 0.988 (±0.051) log P oct + 0.020 (±0.060). All the relations published in this work and other studies have similar coefficients, showing that values of log kf^ and of log P oct are very close. Thus, with C18 silicas, k'm can be approxi­ mated without any additional mea­ surements. Chlorophenol derivatives are used as an example. Depending on the number of substituents, this se­ ries ranges from apolar (log Pot:t = 5) to rather polar compounds (log P oct = 1.5). In Table 2, values of log Ayw, obtained by extrapolating from changes in kf with methanol con­ tent, are in good agreement with lit­ erature values (12). The values of VT in Table 2 for precolumns are calcu­ lated from V0 and V^ values. V0 is the product of the geometric vol­ ume of the precolumn and the po­ rosity of C18 silica. An average po­ rosity value is found between 0.65 and 0.7 for most available C18 sili­ cas. Vr values calculated for off-line cartridges are based on 100 mg of sorbent with an average density of 0.6 g/mL and a V„ calculated as 0.12 mL per 100 mg of sorbent. Val­ ues of Vh were also experimentally determined on precolumns for all the compounds according to the method described above in Figure 2. Good agreement is observed be­ tween the calculated VT values and experimental Vb values obtained with the precolumn. One can therefore rapidly obtain VT from log P 0(;1 and determine whether, depending on the concen­ tration limit required, C18 silica is suitable for the extraction. The data reveal a significant difference be­ tween apolar and moderately polar analytes: Vh values per 100 mg of sorbent are seven liters for pentac h l o r o p h e n o l , 350 mL for 3,5dichlorophenol (log P oct = 3.56), 16 mL for 2-chlorophenol, and < 5 mL for phenol. To extract moderately polar analytes from a sufficient vol­ ume requires an increase in the amount of C18 silica. But even with 1000 mg of sorbent, the sample vol­ ume is about 40 mL for phenol. The limitations of C18 silica are clear for some moderately polar organics, es­ pecially for on-line methods where the amount of sorbent cannot be in­ creased.

A Detailed Description of the Relationship between Vb and Vr The sorption of an analyte by SPE can be described by a frontal analysis model. Figure 1 shows that the derivative of a breakthrough curve is a Gaus­ sian peak (3). When defining the breakthrough volume, Vb, under the condi­ tions of Figure 1 ( 1 % of A0), the retention volume, Vr, is related to Vb by: Vb=Vr-2av

(1)

where σν is the standard deviation depending on the axial dispersion of the analyte along the bed of particles in the precolumn or cartridge. The break­ through volume is therefore controlled by retention and by kinetic parameters. The Vr term can be easily calculated from the capacity factor in water, /200 >200 175 + 25 60 ± 1 5 40+10 6 + 3 5±2