-
0n line Monitoring of Aquatic Samples Automated procedures that increase the speed of analysis and improve analyte detectability U.A.Th. B R I N K M A N
T
race-level determination of a wide variety In the past five years, work has been carried out of pesticides, industrial chemicals, and their in the Department OfAnalytical Chemistry of the Free (bi0)degradation products in surface, University (The Netherlands) to design analyzers ground, and tap waters is an important based on automated on-line solid-phase extractiontopic in environmental analysis. These de- liquid chromatography (SPE-LC)and solid-phase exterminations often involve identification traction-gas chromatography (SPE-GC).The LC sysand confirmationbefore quantification,requhing that tems are equipped with diode-my W; thermospray, skilled operators use sophisticated instrumentaor particle beam mass spectrometric detectors, tion and adequate analytical procedures. Moreover, whereas the GC systems rely on selective GC or masswith the numbers of samples being analyzed rap- selective detectors. Work on what has become hown idly increasing and legislated threshold values for in- as the SAMOS (System for Automated Monitoring of dividual pollutants and groups of environmental con- Organic micropollutants in Surface water) aptaminants becoming stricter, it is not surprising that proach has been performed within the framework of the keywords in modern trace-level environmental the international Rhine Basin Program, initiated by analysis are speed, selectivity, sensitivity and, be- Hewlett Packard on their 50th anniversary ( I ) . The cause of environmental concerns, solvent conservation. These factors are driving monitoring and screening in the direction of fully on-line automatable systems,which combine sample preparation and separation-cum-detection in one analytical setup. ana isis of water samples. Sample preparation in on-linemethods generally inws volves solid-phased o n (SPE).Compared with traditional methods of sample ueatment such as liquidliquid or W e t extraction, or even off-line SPE, online SPE is much less laborious and time consuming. In addition, because of the "closed" nature of an online system, contamination of the sample and sample extract during handling is greatly reduced, analyte losses from evaporation do not occur, and the complete sample rather than a small aliquot (typically 1-570) of the h a l extract is analyzed. Therefore, analyte detectability (in terms of concentration units) increases. Some detectability can be sacrificed to duce sample sizes fmm the conventional 0.5-1.0 L to about 100 mL (columnliquid chromatography)or as little as Y n 0SPEKl 10 mL (capillary gas chromatography). Finally, beWaste cause no liquid-liquid extraction of aqueous samples 1.2. ana 3 hign.pressrre vafves.4. mace-ennchmenl can# age, 5 solenoid valve. is involved, very little organic solvent (c 1 mL) is re6. pdse damper. 7 p ~ r g epump. 8' LC solvent a e *Very system. 9 analp c a co m n 10. dsade a r m detector. 11 wOlknat on 12: or mer. 13. OleDarallVR quired in the on-line procedurescompared with the f ~ quently used 10&300 mL of dichloromethane per 1u sa" it. analysis. i
0013-936Xn51W29-79A$09.00/00 1995 American Chemical Society
VOL. 29, NO. 2, 1995 I ENVIRONMENTAL SCIENCE &TECHNOLOGY. 7 8 A
rmm Selected pesticides found in European rivers using the SAMOS" LC system Garonne
Meuse Meuse Uster (Switzerland) Rhine (Germany) Rhine (Netherlands) Rhine (Netherlands) Rhine (Netherlands) Thames
Nov. 1991 2.3-Dichlorobenzidine, dimethoate June 1992 Atrazine. diuron, simazine Oct. 1992 Benzothiazole Oct. 1991 Alachlor April 1992 Atrazine, barban, fluometuron Oct. 1991 Chloridazone April 1992 Atrazine Nov. 1993 Dimethoat Feb. 1992 3,dDichloroaniline
'System lor Automated Monitoring of Organic micropollutants m Surface water
potential and limitations of the SAMOS appro,,:. fur on-line environmental monitoring of aquatic samples is explored in some detail in this article.
On-line SPE-LC systems The general setup of a fully automated LC-based analyzer for water monitoring is shown in Figure 1. Next to the pumps, key parts of the on-line SPE-LC system are the disposable cartridge containing a small precolumn (No. 4;typically 5-10 nun length x 3-2 nun i.d.1, the alkyl-bonded silica analytical column (No. 9). and the diode-array W (DAD W) detector (No. 10). Dunng an analysis, a suitablevolume of the aqueous sample, often approximately 100 mL, is loaded
onto the precolumn at a speed of 5-10 mL/min. The analytes of interest are retained on the precolumn, which is usually packed with a C18bonded silica or a more hydrophobic styrenedivinylbenzene copolymer such as PLRP-S to trap the analytes (trace enrichment). The solution that runs through the precolumn passes to a waste stream. After a rapid cleanup with a few milliliters of HPLC-grade water, desorption is performed hy coupling the precolumn on-line with the analytical column and starting an acetonitrile-aqueous phosphate buffer mobile-phase gradient. Reversedphase LC-DAD UV proceeds according to established rules. For polar pesticides, which are common target compounds, experience has shown that nearoptimum W detection conditions for a large numher of analytes are found by monitoring at 210/220/ 245/280 or 220/230/245/270/300 tun. Data handling includes searching for target analytes within the proper LC retention time windows, matching the experimentally observed and library DAD W spectra, quantification, and report printing (2.3).With the whole procedure automated, it is possible to identify and quantify some 100 organic micropollutants at concentrations in water samples of 0.5-1.0 pg/L. This is a good result because in the European Union the alert and alarm levels for individual pesticides in surface water are set at 1 pg/L and 3 pg/L, respectively. Several examples that illustrate on-line SPELC-DAD W performance are presented below. In the past two years, samples from the rivers mine (Germany/The Netherlands), Thames (U.K.), Garonne (France), Meuse (Belgium/The Netherlands). and Ehro (Spain) have been analyzed. Pesticides and the- '--akdov- --?ducts were detected
Ietection of a test set of 27 polar pesticides king SPE-LC DAD UV t o analyze 100 mL of Amsterdam tap water.
180
-
I
160 140
120 3
: 100
1 v)
80 60 40
20
,
2 Timelmin)
80 A
VOL. 29. NO. 2,1995 I ENVIRONMENTAL SCIENCE & TECHNOLOGY
I
in many instances. Some examples are presented in Table 1, which is by no means exhaustive. In all but one case (diuron; Meuse, June 1992), concentrations were below the alert level [Le., in the 0.05-0.7 pg/L range). In several cases, the results were confirmed by on-line LC-MS or SPE-GC procedures. Preliminary studies have shown that the present system can easily be modified and adapted to ion pair-based separations. This mode has been used to detect the well-known industrial chemicals anthracene-1,8-disulfonic acid, naphthalene-1,5-disuIfonic acid (NDS),and 4-nitrotoluene-2-s~olfonic acid in mine waters, typically at 1-10 pg/L concentrations. Because drinking-water utilities probably do not use ion-pairing agents to remove strongly acidic sulfonic acids, these compounds also can be expected to be present in tap water. Analysis of Amsterdam tap water, which is produced from Rbine River water, indeed showed the presence of NDS at a level of about 1 pg/L (4). Validation of the performance of an analytical system is of primary importance. Extensive studies were carried out using a test set of 2530 pesticides.Tbe results produced RSD values of 0.2-1.5'70 (n = 20) for the retention times, and 1-9% (n = 8) for peak area measurements in the low microgram-per-literrange. (21ibration plots were h e a r over the concentsationrange of 0.1-10 pg/L in all but two cases, and the detection limitsfor actual samples were at or below the alert level of 1 pg/L for all hut two of the test analytes. Tbis is illustrated for Amsterdam tap water in Figure 2. Although the large hump attributable to humic and fulvic acids creates a rather disturbing looking baseline in the early part of the chromatogram, the actual loss in analyte detectability &e., signalto-noise ratio) in this section compared with the later flat-baseline part is not more than twofold. There were virtually no maintenance problems during the 6-month validation period.
phenylurea isoproturon (< 0.1 pg/Ll was detected. Further work has revealed that a large number of triazines, phenylureas, N-methylcarbamates, organophosphorus pesticides, and chlorophenols can be detected down to 0.01-1 pg/L in surface water (5). Figure 3 s h m the analysis of a water sample from the Rhine River. Time-scheduled monitoring allowed the detection of 30-100 ng/L concentrations of the phenylureas monuron and isoproturon. The peak eluting in front of monuron is not attributable to a phenyhma and has not been identified.The mass spectrum of monuron, which is included, reveals a weak point ofTSP/MS in general: Next to the molecular ion ( d z1991,only few fragment ions appear. Consequently, unambiguous identification often is not possible. One way to improve the analysis is to use both TSP/MS and PBlMS for detection (67). The advantage of the latter alternative is that electron impact mass spectra are generated; these provide a wealth of structural information that can be compared with GC-MS libraries. The main drawback is the somewhat poor sensitivity of PB/MS, mainly because of the low transport efficiency through the interface. In one study, the final breakdown product of diuron, 3,4-dichloroaniline, was identified in a surface water sample by combining positive chemical ionization (molecular mass) and electron impact PB/MS information. Next, on the basis of similarly combined mass spectral information and a retention time (12.8 min) in between those of the aniline (11.1 min) and the parent compound, diuron (13.4
-
1
-scheduled monitoring of Rhine River wate i-LC-TSP 01 50 mL of Rhine River water
SPE-LC-MS Whenever pesticide concentrations are suspected to be close to or higher than the alert level and when relativelylargeunidentified peaks show up in a chromatogram, confirmation and identification make it mandatory to use MS-based procedures. LC-MS has certainly proved to be a useful technique in environmental analysis. However, the sensitivity often leaves much to be desired, both with the thermospray (TSP) and particle beam (PB) interfaces. To meet the current demands of environmental analysis, we use on-line (though not fully automated) SPE-LC-MS,with sample sizes of 50-250 mL, similar to those used with the LC system discussed above. The LC eluent conditions are roughly similar to those used in LC-DAD W analysis, although with the MS Engine mass spectrometer frequently used by us, methanol is preferred to acetonitrile as a modifier with ammonium acetate added as an ionization promoter. In one study, 50 mL of a Meuse River sample was analyzed using SPE-LC-TSP/MSwith time-scheduled detection to improve analyte detectability. The presence of three target analytes, diuron (1.4 pg/L), atrazine (0.4 pg/L), and simazine (0.6pg/L), was confirmed. In addition, a small peak attributable to the
2
4
6
8
10
12
14
16
18
20
280
3
Time Imin)
c
1W
180
2W
220
240
260
Masdcharpe m/z Monitoring at m/zlLWandZOI 10-10 min). dz207 Il0-14minl. andmh275IlC20mi to detect monuron. isoprmron. nnd neburon Inat present in sample).respenivslv.
VOL. 29. NO. 2.1995 i ENVIRONMENTAL SCIENCE d TECHNOLOGY
81 A
min), another peak could be assigned to the intermediate, N-(3,4-dichlorophenyI)-N'-methylurea. In other surface water samples the bisphenols, bis(4-hydroxypheny1)methaneand bis(2-hydroxyphenyllmethane (Ebro River),and N-butylbenzenesulfonamide (ObRiver, Siberia)were identified. Clark et al. (8) reported the presence of the latter compound in various water samples, in addition to other plasticizers and plastic additives, by using LCPB/MS and GC-MS with off-line sample pretreatment.
On-line SPE-GC systems In most studies, capillary GC should be one's first choice as a separation technique because of its high separation efficiency,high speed of analysis, and wide range of sensitive and often selective detectors. However, analysis by GC suffers from one weak point: sample introduction. In many procedures injection volumes are only a few microliters. Because final extracts typically have volumes of between 50 pL and 1mL, 9599%of all analyte(s) collected is discarded in the last step before analysis. Fortunately, the state of the art of GC injection techniques is such that retention gaps-which essentially are uncoated deactivated fused silica capillaries several meters in length, used to refocus analyte handsare graduallyhemming standard equipment (9).Consequently, injection of up to 100 pL of solution is not a problem anymore, provided a suitable organic solvent is selected. If aqueous samples have to be handled, SPE procedures similar to those discussed for the various LCbased systems, combined with efficient desorption using a relatively small volume of organic solvent, open up the possibility to design on-line E-GCbased or SPE-GC-based water analyzers. A typical setup is shown i: Yigure i. 2: :he context of the
present review, only the main aspects are discussed. Because of the better performance of GC detectors, the aqueous sample volumes are generally 1-10 mL rather than the 100 mL used with the LC techniques. In most instances, it is mandatory to remove even the last traces of water before desorption of the analytes into the GC part of the system. To this end, 15 min of N, drying at ambient temperature is required. Experience shows that this procedure does not lead to significant losses for medium-volatile analytes. Drying is facilitated if the conventional precolumn is replaced with a cartridge holder containing 3-5 small (3-4-mm diameter) extraction disks. These disks allow a high flow rate of gas to pass through their pores. Desorption requires 50-150 pL of an organic solvent such as ethyl acetate at a flow rate of about 50 pL/min. During the 1-3-min procedure, the analytes are quantitatively desorbed and transferred to the retention gap under so-called partially concurrent solvent evaporation conditions. After transfer, separation and detection are performed according to conventional GC procedures. We have used automated setups equipped with flame ionization (FlD), thermionic (NPD), flame photometric (FPD), and mass-selective (MSDI detectors. Studies on raw and spiked surface water and drinking water included the analysis of more than 100 samples hy means of SPE-GC-FlDwithout exchanging any part of the system. When analyzing 1mL of a surface water spiked with 10-40 pg/L of nine chlorinated phenols, even with the nonselective FID, detection limits were about l pg/L. Much work has been done in the area of organophosphorus pesticides, triazines, and organosulfur compounds. For organophosphorus compounds, detection limits of 0.1-0.2 pg/L were
lutomated SPE-GC-MSD system eneral setup of on-line and automated system
SPE camidge
2 4 4 pinch
solenoid valve
8 2 A . VOL. 29.
NO. 2,1995 I ENVIRONMENTAL SCIENCE & TECHNOLOGY
On-column injector
obtained in surface water from several rivers using SPE-GC-NPD and 2.5-mL samples (10).The target compoundsmevinphos, diazinon, feniimthion, fenthion, triazophos, and coumaphos-bracket the polarity range of a large part of the GC-amenable organophosphorus pesticides. The system was also used to detect and quantify atrazine and simazine in surface and drinking waters. The results were compared with SPE-GC-MSusing either multiple ion detection or full-scan acquisition (11). As an example, the data of Table 2 show good agreement among the various analytical procedures for atrazine, even at the low concentration of 10-100 ng/L. Figure 5 shows the total ion current recorded for 10 mL of Meuse River water with fullscan acquisition as well as the reconstructed ion traces for both triazines and their mass spectra. Fullscan acquisition offers the considerable advantage over multiple ion detection that other target compounds or unknowns can be studied. Recent work has focused on an alternative to drying with N2. To that end, a short drying cartridge containing anhydrous sodium sulfate or silica was inserted between the outlet of the trace-enrichment precolumn and the retention gap (Figure 4). Such a column can be regenerated between runs by means of heating and can be reused as many as 100 times. There are virtually no losses because of adsorption for the many classes of organic micropollutants that have been tested (12).The system has been used for chlorobenzenes; chlorophenols; and phosphorus-, sulfur-, and nitrogen-containing pesticides. In one series of applications the detection limits in Amsterdam tap water (IO-mL sample) invariably were lower than 0.1 pg/L, even with FID detection. However, with NPD and FPD detectors, seleaivitywas markedly better. Therefore, these detectors were primarily used for surface water analysis. Because selectivity is a major concern with all trace-level studies, it is useful to examine the recently introduced atomic emission detector (AED). For this detector, an important problem is that, according to many analysts, more than 1-2 pL of an organic solvent cannot be injected. Our experience finds that a minor modification of the sample introduction system-inserting an early solvent vapor exit between the retention gap and the retaining precolumn (Figure4)-enables the introduction of up to 100 pL of an extract. To date, most of our work has been limited to the off-line combination of SPE and GC-AED. That is, 50-mL samples were preconcentrated by SPE and, after drying with N,, the analytes were desorbed with about 500 pL of ethyl acetate. 'Ikenty percent of this extract was then injected into the GC-AED instrument. The detection limits of 11 organophosphorus pesticides were in the 0.1-0.5 pg/L range, and the method was successfully used to confirm the presence of henzothiazole (S and N channels, 1 pg/L) in surface water and of triphenylphosphine oxide (P channel, 0.1 pg/L) in tap water (13).Recently, a fully on-line SPE-GC-AED system has been set up. Preliminary work indicates that the expected fivefold increase in analyte detectability is indeed observed and there are no undue maintenance problems.
-
-
Atrazine determination using on-line methods Concentrationof atrazine (nglL)using
1 mL Samole
MID'
Tap water Meuse, Feb. 1992 Meuse,June1992 Rhine Detection limit
25
45 45
50 5
10 mL
FSb NPDC MID FS NPI
25 50 50 55 30
45 40
15 60 40 70 0.5
10 50 40
20
60
65
3
4
50
'SPE-GC-MS wilh multiple ion detection (MID), M S quamilication Bf m/ZZW. SPE-GC-MS wlh full-scan (FSI acquisition. SPE-GGNPD.
'
ConclusiullJ In the past few years, the development and use of analytical systems based on the on-line combination of analyte trace enrichment and chromatographic separation-plus-detection have taken great strides forward. The use of integrated and often semi- or fully automated procedures has increasedthe speed of analysis and improved analyte detectahility (i.e., sensitivityand selectivity).Moreover, compared with older setups, online systems markedly reduce consumption of organic solvents. The state of the art,how-, is clearly not the same in SPE-LC-W-,SPE-LC-MS-,and SPE-GC-based procedures. Today, LC systems with diode-my W detection and similar SPE-LC-W systems (14 are being used routinely in a number of laboratories. Satisfactoryresults are obtained for all but the highly polar analytes (earlyelution, rapid breakthrough) and the small number of priority pollutants displaying insufficient W absorbance. If full automation is not a main concern, no special problems will be encountered by experienced analytical chemists who want to construct their own integrated analyzers from commercially available instrumentation. For LC-MS,the situation is different. The monitoring of large numbers of organic microcontaminants has not been given the attention it deserves. Fortunately, E - M S studies on ever increasing numbers of such analytes are being published (5.6, 15). Another problem is that traditionally interfaces have been emphasized, mainly TSP/MS and PB/ MS, which display deficiencies with regard to analyte detectahility and compatibility with LC eluent composition. Somewhat surprisingly,the manyproblems observed with interfacing have not really stimulated the utilization of SPE-LC-MS techniques. Recent studies (5, 7Jshow that on-line SPE-LC-MS is a highly desirable option that enables trace-level analysis for a large number of compounds. The introduction of other interfaces such as electrospray and atmospheric pressure chemical ionization will no doubt further stimulate research and development. Finally, running an on-line LC-GCseparation was truly exceptional 10 years ago, and such heartcut operations were restricted to the use of normal-phase LC eluents. Currently many papers have been pubVOL. 29, NO. 2,1995
I ENVIRONMENTAL SCIENCE & TECHNOLOGY. 83 I
-"
I 5u
L.
.I.
11,
I, Simazine
Atrazine
I
150
mJr
dz
)Total ton current(TlCltncs,lbl re~onmuctedion trecesfor atreme ldz2MI end simezine (m/1201) Mass spectra rscordsd forlcl atrazin8 id Id) sirname peaks
lished in this area. Even on-line reversed-phase LC-GChas been used successfully by several workers (for an overview, see Reference 16). In addition, several approaches have been reported that enable the on-line analysis of water samples by GC. lbm interesting alternatives are the on-line combination of SPE (16)and GC or liquid-liquid extraction and GC (17). The former option is better suited to effect trace enrichment and to encompass a wide range of analytes of varying polarity in one run.This is the SPE-GC approach that, with its NF'D, FPD, or MS selective detection, opens up new horizons in analyte identification and quantificationin water samples at the trace level. With mass-selective detection steadily becoming less expensive and more user-friendly, online SPE-GC-MSD water analyzers may have a bright future. However, currently such SPE-GC-based analyzers certainly are not as rugged as the SPE-LC procedures and, therefore, are not suited for the routine laboratory. Attempts to reach that stage are presently in p r o p s and already amact attentionbecause of the interesting high selectivity and sensitivity of all procedures involving a GC separation.
Acknowledgment This study was financially supportedby the Eumpean Union, project EVSV-(392-0105. 84 A .
VOL. 29. NO. 2.1995 i ENVIRONMENTAL SCIENCE &TECHNOLOGY
References (11 van Hout. E1.M.; Brinkman. UAT. Eur WaterPoNut. Confro1 1993,3,29. (21 Slabodnik. J. et al. Anal. Ckim. Acta 1992,268.55. (31 Slobodnik. J. et al. I. Ckmmafogr: 1993,642,359. (4) Brouwer, E. R. et al. Quim Anal. 1993,12.88. (51 Bagheri, H. et al. I. Ckromatogr 1993,647,121. (6)Miles, C. I.; Doerge, D. R.; Bajic, S. Arch. Environ. Conram. Toxicof. 1992.22,247. (71 Bagheri, H. et al. Ckromarographin 1993,37,159. (8) Clark, L.B. et al. Intern.1. Environ.Anal. Ckem 1991.45, 169. (9)Gmb, K On-linecoupled ZGGC; Huthig Heidelberg, 1991. (10)Kwakman. ELM. et al. Ckromatograpkia 1992,34,41. ill1 Bulterman, A-I. et al. High-Res.Ckmmamgr: 1993.16,397. (12)Pic6,Y et al. Analysr, 1994,119. 2025. (13)Rinkema, E D.;Louter, A.I.H.; Brinkman, U.A.T. I. Ckmmnmm: 1994.678.289. (141 R e u p k R.; Zobe, 1.; Ploeger, E. LC-GC, Intern. 1992,5, 6. (15)Volmer, D.;Levsen,K.;Wunsch, G.I. CkmmatogrA 1994, 660,231. (16)Vreuls, 1. J. et al. J.Assoc.On Anal. Ckem. 1994,77,306. (17)ValcBrcel. M.; Ballesteros, E.; Gallego. M. 7hndsAnol. Ckem. 1994,13,68.
U.A. Th. Brinkman is professor of analytical chemistry at theFree University (Amsterdam). His m a i n research is in hyphenated and coupled-column separation techniques, with environmental analysis as the area of a p plication.