Journal of Chromatography A, 1216 (2009) 334–345

Contents lists available at ScienceDirect

Journal of Chromatography A

journa l homepage: www.e lsev ier .com/ locate /chroma

Methods for the determination of phenolic brominated flame retardants,and by-products, formulation intermediates and decompositionproducts of brominated flame retardants in water

P. López ∗, S.A. Brandsma, P.E.G. Leonards, J. De BoerInstitute for Environmental Studies (IVM), VU University Amsterdam, De Boelelaan 1087, 1081 HV Amsterdam, The Netherlands

a r t i c l e i n f o

Article history:Available online 15 August 2008

Keywords:BromophenolsBromoanisolesBromoanilinesBromotoluenesWater analysisREAch

a b s t r a c t

Brominated flame retardants (BFRs) are the chemicals of high importance within the REAch frame-work. In addition to polybrominated diphenyl ethers (PBDEs), hexabromocyclododecane (HBCD) andtetrabromobisphenol A (TBBPA), other BFRs such as bromophenols, intermediates in FR formulation likebromoanilines, and their brominated and non-brominated by-products such as bromoanisoles, bromo-toluenes, bromoalkanes and 1,5,9-cyclododecatriene, respectively should be monitored and controlledbecause of their toxicity and their very low odour and taste thresholds, below sub-nanogram-per literlevels. In the present study several analytical methods for the simultaneous determination, i.e., com-bining one single sample treatment and one analysis step, of these compounds in water have beendeveloped, optimized and evaluated. The methods involve a (pre-concentration)-extraction technique,such as liquid–liquid (LLE), solid-phase (SPE), headspace (HS) extraction or solid-phase microextraction(SPME), followed by gas chromatography (GC)–mass spectrometry (MS) analysis with either electron cap-ture negative ionization (ECNI) or electron impact (EI) as ionization techniques. ECNI is more sensitive thanEI for analytes with more than one bromine atom. HS and SPME were previously optimized by means of amultifactorial experimental design. Extraction temperature and the liquid/headspace volume ratio werethe most significant factors in HS extraction. In SPME, the variables studied were the nature of the fiber, themode of extraction and the extraction temperature. Polydimethylsiloxane (PDMS) fibers appeared to bemore suitable than carboxen-polydimethylsiloxane (CAR-PDMS) for the analysis of the target compoundswith more than one bromine atom. The extraction of 2,4-dibromoaniline was only achieved in a directimmersion mode, in which the optimal extraction temperature was 60 ◦C. The methods LLE–GC–(ECNI)MS,LLE–GC–(EI)MS, SPE–GC–(ECNI)MS, SPE–GC–(EI)MS, HS–GC–(EI)MS and SPME–GC–(EI)MS were evalu-ated in terms of linearity, precision, detection limits and trueness. All methods, with the exception ofHS–GC–(EI)MS, were linear in a range of at least two orders of magnitude, giving recoveries above 75% anddetection limits at the low ng/L level for most of the target analytes. SPE–GC–(ECNI)MS is the most sensitiveand reliable method for the determination of most of the bromine compounds, whereas SPE–GC–(EI)MS

is the most suitable to quantify the three isomers of 1,5,9-cyclododecatriene. Both methods together withSPME–GC–(EI)MS (for qualitative confirmation) were applied to water samples from the Western Scheldt(The Netherlands), where 2,6-dibromophenol and 2,4,6-tribromoanisole could be detected at levels higher

r thr

1

pRC

at

0d

than their respective odou

. Introduction

On 29 October 2003, the European Commission adopted a pro-osal for a new EU regulatory framework for chemicals calledEAch (Registration, Evaluation, Authorization and Restriction ofhemicals), which entered into force on 1st June 2007 [1]. The

∗ Corresponding author. Tel.: +31 20 5989555; fax: +31 20 5989553.E-mail address: patricia.lopez@ivm.vu.nl (P. López).

tiotnuwT

021-9673/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.chroma.2008.08.043

esholds.© 2008 Elsevier B.V. All rights reserved.

im of REAch is to improve the protection of human health andhe environment through the better and earlier identification ofhe intrinsic properties of chemical substances, while maintain-ng the competitiveness and enhancing the innovative capabilityf the EU chemicals industry. A Chemicals Agency will act as

he central point in the REAch system: it will run the databasesecessary to operate the system, co-ordinate the in-depth eval-ation of suspicious chemicals and run a public database inhich consumers and professionals can find hazard information.

he Regulation also calls for the progressive substitution of the

togr. A

mi

dtcb(ispdbcpfb2imbpcabcbpc

tTcmocotti[ttOattrrbmia

m(sssrsmfcp

iwce[sbttb1ytmcol

cbtGenewFee

2

2

sB(4299(t12aS

sdGG

pamewu

P. López et al. / J. Chroma

ost dangerous chemicals when suitable alternatives have beendentified.

Flame retardants (FRs), which are a group of structurallyiverse chemicals which, added to materials, inhibit or suppresshe combustion process, are also a group of chemicals of highoncern within REAch. Thirty-nine percent of FRs is based onromine [2]. About one third of brominated flame retardantsBFRs) are polybrominated diphenyl ethers (PBDEs), another thirds 3,5,3′,5′-tetrabromobisphenol A (TBBPA) and the remainder con-ists of various other brominated compounds, within which otherhenolic compounds are included, such as 4-bromophenol, 2,4-ibromophenol and 2,6-dibromophenol. These compounds maye also formed as by-products of TBBPA, either in its photo-hemical degradation in water [3] or from the decomposition oflastics [4]. Like chlorophenols, bromophenols can be biotrans-ormed into the corresponding bromoanisoles via O-methylation byacterial microorganism [5]. 4-bromoaniline, 2,4-dibromoaniline,,6-dibromoaniline and 2,4,6-tribromoaniline are reported as

ntermediates in the formulation of FRs incorporated to plasticaterials [6]. Bromotoluenes and aliphatic bromides may also

e formed during the thermal decomposition of flame-retardedolystyrene [7] or in the water treatment processes. Hexabromo-yclododecane (HBCD) has been reported to biodegrade in fresh-nd wastewater via series of dihaloelimination steps, losing tworomines from vicinal carbons at each step and resulting in 1,5,9-yclododecatriene as the final degradation product [8]. Not allrominated compounds in the aquatic environment are of anthro-ogenic origin. Bromophenols, bromoanisoles and bromoalkanes,an also be produced by algae, polychaetes or coral [9–11].

Most of the studies in the presence of BFRs in environmen-al samples have focused on the occurrence of PBDEs [2,12–16],BBPA [17,18] and HBCD [19–21]. Only few studies have beenonducted on bromoanisoles, bromophenols, bromoanilines, bro-otoluenes and aliphatic bromides in water, hardly with detail

n their simultaneous determination. Bromophenols (BPhs) are ofonsiderable interest because of their extremely low taste thresh-lds, even at sub-nanogram-per liter levels in water. The tastehreshold of 2,6-dibromophenol is 0.5 ng/L in water [22]. Tastes dueo the presence of BPhs have been described as plastic or medic-nal [23], disinfectant, blench or plasticine [22] and iodoform-like24]. The threshold levels of brominated anisoles are even lowerhan those of the most organoleptically potent BPhs (e.g., the odourhreshold for 2,4,6-tribromoanisole (TBA) is 0.03 ng/L in water [25].n the other hand, the toxicity of BPhs, pentabromotoluene (PBT)nd aromatic anilines has already been demonstrated. BPhs dis-urb cellular Ca2+ signaling in neuro-endocrine cells [26], showhyroid-hormone like activity [27] and bind to the human estrogeneceptor [28]. PBT has been reported to stimulate aryl hydrocarboneceptor (AhR) DNA binding to levels comparable to that producedy TCDD (2,3,7,8-tetrachlorodibenzo-p-dioxin) [29], whereas bro-oanilines, in addition to their own carcinogenicity, may convert

nto N-nitroso compounds through reactions with nitrosylatinggents in the environment [30].

Most of the cited analytical methods are based on chro-atographic techniques. Gas chromatography–mass spectrometry

GC–MS) has been recognized as the method of choice in a wideeries of environmental analyses due to its higher selectivity andensitivity. Electron capture negative ionization (ECNI) is extremelyensitive for brominated compounds, with the mass-to-chargeatios (m/z) 79 and 81 as the dominant ones in the ECNI mass

pectrum. Electron impact (EI-MS) ionization mode would offerore selectivity and the ability to confirm a compound’s identity

rom its full mass spectrum [31]. Considering the expected con-entrations of these BFRs and derivates (ng/L) in water, applying are-concentration technique prior to the analytical determination

4sal

1216 (2009) 334–345 335

s advisable. Traditionally, liquid–liquid extraction (LLE) techniquesith organic solvents were employed [32], but those are time-

onsuming and hazardous. Currently, safer and more selectivextraction techniques are used, such as solid-phase extraction (SPE)33], solid-phase microextraction (SPME) [4,34–36], closed-looptripping analysis (CLSA) [37], purge and trap [24] and, recently, stirar sorptive extraction (SBSE) [38,39] and liquid–liquid microex-raction [40]. Nevertheless, and to the best of our knowledge,he simultaneous determination in a single run of bromophenols,romotoluenes, bromoanilines, bromoanisoles and the isomers of,5,9-cyclododecatriene in water samples has not been reportedet. Due to the broad spectrum of physical–chemical properties ofhe target compounds, several extraction techniques and ionization

odes should be investigated and optimized. The use of only oneonsecutive extraction and analysis steps in the overall analysis isf paramount relevance since it entails a significant reduction inaboratory resources.

The aim of the present study was therefore to develop andompare different methods for the simultaneous analysis of therominated compounds mentioned above. Four different extrac-ion techniques, LLE, SPE, headspace (HS) and SPME, combinedC–MS in two ionization modes (EI and ECNI) were tested andvaluated in terms of linearity, precision, detection limits and true-ess. Since SPME and HS extraction can be affected by severalxperimental variables, an optimization of each extraction processas previously conducted using multifactor experimental designs.

inally, the feasibility of the most reliable and robust methods wasvaluated by analyzing water samples from the Western Scheldtstuary in The Netherlands.

. Experimental

.1. Chemicals and materials

n-Decyl bromide (DeBr, 98%) and n-cetyl bromide (CeBr) wereupplied by Eurobrom B.V (Amsterdam, The Netherlands). 2-romotoluene (2BT, 99%), 3-bromotoluene (3BT), benzal bromideBBr), 4-bromotoluene (4BT, 98%), 2-bromoanisole (2BA, 97%),-bromoanisole (4BA, 99%), 3,5-dibromotoluene (3,5DBT, 97%),,5-dibromotoluene (2,5DBT, 98%), 2,4-dibromophenol (2,4DBPh,5%), 2,6-dibromoaniline (2,6DBAl), 2,6-dibromophenol (2,6DBPh,9%), 2,4-dibromoaniline (2,4DBAl, 98%), 2,4-dibromoanisole2,4DBA), trans-, trans-, trans-1,5,9-cyclododecatriene (EEE), trans-,rans-, cis-1,5,9-cyclododecatriene (EEZ, 98%), trans-, cis-, cis-,5,9-cyclododecatriene (EZZ), 2,4,6-tribromotoluene (TBT, 98%),,4,6-tribromoanisole (TBA, 99%), 2,4,6-tribromoaniline (TBAl, 98%)nd 2,3,4,5,6-pentabromotoluene (PBT, 99%) were supplied byigma–Aldrich Chemie B.V. (Zwijndrecht, The Netherlands).

Water (HPLC grade) and acetone for organic residue analy-is were purchased from J.T. Baker (Deventer, The Netherlands),ichloromethane for residue analysis from Promochem (Wesel,ermanay) and pure ethyl acetate from Riedel de Häen (Seelze,ermany).

Stock standard solutions of each analyte (1000 �g/mL) wererepared by weight in dichloromethane. Standard mixtures ofpproximately 2 �g/mL were weekly prepared in acetone. Standardixtures for calibration curves in dichloromethane for LLE tests, in

thyl acetate for SPE tests and in water for both SPME and HS testsere daily prepared. All solutions were stored at 4 ◦C in the darkntil their use.

The SPE extraction was carried out using a standard Millipore7 mm glass vacuum filtration apparatus (Bedford, MA, USA). Theintered piece of glass, acting as support for the glass fibre filtersnd SPE disks was removed and replaced by a removable stain-ess steel support with small holes. This construction facilitated

3 togr. A

avVt(pf

S1pt

2

sgd(2mpsnG

2

rcTtrt

Smsua9sewfadar

eidpssdea

scd

2

f(hptplTtfie

(a

a

2

otsP[wtulCh

paflppetfi

pUT

2

Giaawe

l

36 P. López et al. / J. Chroma

nd reduced the cleaning time of the extraction equipment. Theacuum source used was a diaphragm vacuum pump supplied byacuubrand GMBH + CO (Werheim, Germany). Glass microfiber fil-

ers GC/C from Whatmann (Maidstone, UK) on EmporeTM SDB-XC3M, St. Paul, MN, USA) high performance extraction disks, sup-lied by Varian Benelux (Middelburg, The Netherlands), were usedor the SPE experiments, as described by Westbom et al. [41].

The SPME manual holders and fibers were obtained fromupelco (Bellefonte, PA, USA). Two commercially available fibers,00 �m polydimethylsiloxane (PDMS) and 85 �m carboxen-olydimethylsiloxane (CAR-PDMS) were tested. They were condi-ioned as recommended by the manufacturer.

.2. Liquid–liquid extraction procedure

For the method validation experiments, 1 L of HPLC water waspiked with 1 mL of a mixture solution containing all the tar-et compounds in concentrations ranging from 2 to 2000 ng/L,epending on the type of validation test. Extraction solventdichloromethane) and extraction volumes (three times with 500,50 and 250 mL) were chosen based upon preliminary experi-ents conducted in the IVM laboratory. The organic phases were

ooled, concentrated to 20 mL and dried over anhydrous sodiumulphate. The extracts were further concentrated to 1 mL under aitrogen stream. One microliter of the extract was injected into theC–(EI)MSD and in the GC–(ECNI)MS systems.

.3. Solid-phase extraction procedure

SDB-XC is a poly(styrenedivinylbenzene) copolymer used as aeversed phase sorbent for solid-phase extraction. SDB-XC is a 100%opolymeric particle that is spherical, porous and cross-linked.his polymeric phase was selected because of its wider applica-ion range since it provides a unique selectivity, especially in theetention of moderately polar, water-soluble analytes, better thanhat provided by the C8 or C18 bonded silica.

Prior to the filtration and extraction, the system described inection 2.1 was conditioned, according to the procedure recom-ended by the SPE disks’ manufacturer (3M). In a first step, the

ystem was washed with 10 mL acetone until dryness under vac-um, followed by 10 mL iso-propanol. Then, 10 mL methanol wasdded, 1 mL was sucked out under vacuum and the remainingmL was left to soak the disk for 30 s. Subsequently, they were

ucked out with vacuum until a thin layer of solvent just cov-red the surface. On top of the methanol layer, 10 mL of wateras added and sucked through the disk until the water sur-

ace just covered the disk surface. The sample (500 mL) was thendded over the conditioned system and vacuum was applied untilryness. The flow rate, which depends on the vacuum sourcend the solids content of the sample, did not influence theecoveries.

In a first attempt, the target compounds from spiked water wereluted with dichloromethane, but low and inconsistent recover-es were obtained due to the inability of the dichloromethane toisplace and penetrate the water saturating the porous sorbentarticles to which the analytes were adsorbed. Thus, the elutiontep was modified incorporating ethyl acetate, a more water-oluble solvent, in the following order: 10 mL ethyl acetate, 10 mLichloromethane and 10 mL of a mixture 1:1 of both solvents. Theluate was passed through an anhydrous sodium sulfate column

nd was evaporated until 0.5 mL under a nitrogen stream.

For the method validation experiments, 0.5 L of HPLC water waspiked with 0.5 mL of a mixture solution containing all the targetompounds in concentrations ranging from 40 to 2000 ng/L, alsoepending on the type of validation test.

Aoahf

1216 (2009) 334–345

.4. Headspace experimental design

A central composite circumscribed (CCC) design has been per-ormed to investigate the effects of the extraction temperatureT), extraction time (tm), salt content (S), agitation (Ag) and theeadspace volume/sample volume ratio (%f). This CCC design com-rised a three level full-factorial design (+1, −1), superimposed byhe centre points (coded 0) and the star points (+˛, −˛). The staroints allow estimation of the curvature in the model and estab-

ish new extremes for the low and high settings for all the factors.he value of ˛ depends on certain properties of the design and onhe number of factors involved. In this study the value of ˛ wasxed at 1.5. The final regression model was used to find the optimalxtraction conditions.

Headspace (HS) sampling was conducted using a Combi PalCTC Analytics, Zwingen, Switzerland). CTC Pal is a multifunctionalutosampler for the headspace and liquid GC injection system.

All statistical analyses were carried out using the statistical pack-ge ModdeTM for Windows (Umetrics, Umeå, Sweden).

.5. Solid-phase microextraction design

Due to the number of variables and their intrinsic nature, theptimization was carried out in two steps. In a first step, the fac-or type of fiber, extraction mode and extraction temperature wereelected. Among all the commercially available fibers, fibers fromDMS and CAR-PDMS were selected based on previous studies4,34,35,42]. For each experiment, 20 mL amber headspace vialsere filled with 15 mL of HPLC spiked water with a final concen-

ration of 20,000 ng/L of the target compounds. Vials were sampledsing a CTC Pal injector. The desorption time was set at 2 min (split-

ess time) and the injection port at 250 ◦C for PDMS and at 300 ◦C forAR-PDMS. The fiber was additionally desorbed for 15 min in theeater unit available in the CTC Pal, thus reducing the carryover.

In a second step, a Box Behnken design [43] was selected toerform the optimization of the salt content, extraction time andgitation speed since it provided less uncertainty than other multi-actorial experimental designs. In this design, which requires threeevels of each factor, the treatment combinations are at the mid-oints of edges of the process space and at the center. The SPMErocedure was the same as the one explained above. The only differ-nces were that (i) some vials were also filled with NaCl, accordingo the design requirements and (ii) the desorption temperature wasxed at 250 ◦C.

All statistical analyses were carried out using the statisticalackage ModdeTM for Windows (Umetrics, Umeå, Sweden) andnscrambler software package, version 9.1 (CAMO Software AS,rondheim, Norway).

.6. Gas chromatography–mass spectrometry

LLE, SPE, HS and SPME analyses were performed in an AgilentC 6890N (Palo Alto, CA, USA) equipped with a CTC Pal sample

njector and an Agilent 5973N mass selective detector (MS) withelectron impact ion source. Extracts from SPE and LLE were also

nalyzed in another 6890N GC coupled in line with a 5975XL MSith a chemical ionization ion source. A microliter of extract from

ither SPE or LLE was injected in the GC–MS systems.The analyses using the 5973 MSD were performed on a 25 m

ength × 0.22 mm I.D., 0.25 �m thickness, SGE BPX5 column (SGE

nalytical Science, Bester, Amstelveen, The Netherlands). The GCven temperature program was: 40 ◦C for 2 min, heated to 120 ◦Ct 3 ◦C/min and hold for 5 min, heated to 180 ◦C at 6 ◦C/min andold for 2 min and finally heated to 290 ◦C at 15 ◦C/min and hold

or 5 min. Helium (purity 99.999%) was employed as carrier gas,

P. López et al. / J. Chromatogr. A 1216 (2009) 334–345 337

Table 1Molecular weights, selected ions for the quantification of the target compounds and comparison between the sensitivity of electron capture negative ionization (ECNI)detection and electron impact (EI) detection for a solution of 1 �g/mL of the target compounds

Compound MW Ions (m/z)a Areab

ECNI EI ECNI/EI

2-Bromotoluene 2BT 171.04 79 + 81 170 + 172 + 91 0.0484-Bromotoluene 4BT 171.04 79 + 81 170 + 172 + 91 0.2053-Bromotoluene 3BT 171.04 79 + 81 170 + 172 + 91 0.0124-Bromoanisole 4BA 187.04 79 + 81 171 + 173 + 186 + 188 0.0142-Bromoanisole 2BA 187.04 79 + 81 171 + 173 + 186 + 188 0.040trans-, trans-, trans-1,5,9-Cyclododecatrienec EEE 162.28 – 79 + 93 + 133 + 134 0trans-, trans-, cis-1,5,9-Cyclododecatrienec EEZ 162.28 – 79 + 93 + 133 + 134 03,5-Dibromotoluene 3,5DBT 249.94 79 + 81 169 + 171 + 250 1.232,5-Dibromotoluene 2,5DBT 249.94 79 + 81 169 + 171 + 250 1.01trans-, cis-, cis-1,5,9-Cyclododecatrienec EZZ 162.28 – 79 + 93 + 133 + 134 0Benzal bromide BBr 249.94 79 + 81 169 + 171 + 90 + 89 4.69Decyl bromide DeBr 221.22 79 + 81 135 + 137 0.0452,4-Dibromophenol 2,4DBPh 251.92 79 + 81 250 + 252 + 254 5.452,6-Dibromophenol 2,6DBPh 251.92 79 + 81 250 + 252 + 254 3.782,6-Dibromoaniline 2,6DBAl 250.93 79 + 81 249 + 251 + 253 2.302,4-Dibromoanisole 2,4DBA 265.93 79 + 81 264 + 266 + 268 + 251 2.562,4-Dibromoaniline 2,4DBAl 250.93 79 + 81 249 + 251 + 253 19.02,4,6-Tribromotoluene TBT 328.83 79 + 81 328 + 330 + 249 6.872,4,6-Tribromoanisole TBA 344.84 79 + 81 329 + 331 + 344 + 346 6.792,4,6-Tribromoaniline TBAl 329.83 79 + 81 329 + 331 5.00Cetyl bromide CeBr 303.35 79 + 81 135 + 137 0.0322,3,4,5,6-Pentabromotoluene PBT 486.65 79 + 81 486 + 488 + 407 3.10

wsai2

f(phfaa1a

2

Ha[

tl

dw[

pt[

wd

2

WCWsdCp

py2St

3

3n

ammbidMbw

a The sum of the selected ions was used for quantification purposes.b Compounds with better sensitivity in ECNI mode in bold.c Not detectable in ECNI mode.

ith a constant flow of 1 mL/min. The injector was operated in theplitless mode, with a splitless time of 2 min. The injector temper-ture was 200 ◦C for liquid injection and HS and 250 ◦C for SPMEnjection. Source, quadrupole and transfer line were 230, 150 and80 ◦C respectively.

The analyses using the chemical ionization ion source were per-ormed on a 60 m × 0.25 mm, 0.25 �m thickness, CP-Sil8 CB columnVarian, Middelburg, The Netherlands). The GC oven temperaturerogram was: 40 ◦C for 2 min, heated to 160 ◦C at 10 ◦C/min andold for 10 min and finally heated to 290 ◦C at 10 ◦C/min and hold

or 10 min. The injector was operated in the pulsed splitless, withpulse pressure of 4.30 bar kept for 1.5 min. The injector temper-

ture was 200 ◦C. Source, quadrupole and transfer line were 250,50 and 280 ◦C respectively. The ionization gas was methane withflow of 3.25 mL/min.

The ions chosen for quantification are shown in Table 1.

.7. Validation procedure

The validation of the different methods was performed withPLC water (no signal was detected at the retention times of thenalytes) according to EURACHEM guidelines [44] and ISO 572545].

The detection and quantification limits, LODs and LOQs respec-ively, were calculated as 3 (10) × [height of the noise]/[height ofowest standard] × [concentration lowest standard].

The linearity was established in the 0.05–100,000 ng/L range,epending on the methodology. Statistical analysis (ANOVA)as performed to check the goodness-of-fit and linearity

46].Precision, measured as intra-day repeatability and between-day

recision estimated over 3 days, was calculated in terms of RSD% at

wo concentration levels, performing three replicates at each level47].

Trueness was evaluated in terms of recovery by spiking HPLCater with two different concentrations of analytes, which alsoepended on the extraction methods.

t

tE8

.8. Sample collection

Four water samples were collected from two locations in theestern Scheldt, just in front of the mouth of the Ghent-Terneuzen

anal. The first location was close to the mouth of the canal and theestern Scheldt bank (sluice, sampling point 1 in Fig. 1) and the

econd one was situated approximately 3 km far north, in the mid-le of the Western Scheldt estuary, in front of the Ghent-Terneuzenanal (Middelplaat, sampling point 2 in Fig. 1). The sampling waserformed twice with one month in May 2007.

Water samples were collected in 1 L-amber glass bottles witholypropylene screw caps and stored at −18 ◦C. Prior to the anal-ses; the bottles were defrosted at room temperature for 24 ◦C.00 mL of each sample was used for SPE extraction and 15 mL forPME analyses. The samples were not filtered before performinghe extraction. Every sample was analyzed in triplicate

. Results and discussion

.1. Mode of ionization: electron impact (EI) or electron captureegative ionization (ECNI)

The most widely used detectors for the determination of BFRsre MS operated either in ECNI or EI mode [48]. However, the ECNIode has been scarcely used in the analysis of bromophenols, bro-otoluenes, bromoanisoles or bromoanilines. The sensitivity of

oth ionization techniques for the analytes of interest was checkedn a preliminary experiment. A solution of 1 �g/mL of each stan-ard was injected in a GC 6890N coupled in line with a 5975XLS under the same chromatographic conditions and in scan mode

ut while varying the ionization mode. The most abundant ions,hich were used for the optimization, the method validation and

he quantification are shown in Table 1.As it was expected the m/z ions 79 and 81, corresponding to

he bromine fragments [Br−], were the most abundant ones inCNI. Brominated compounds show a typical 79Br (50.5%) and1Br (49.5%) isotope distribution pattern [49]. In the EI mode, the

338 P. López et al. / J. Chromatogr. A 1216 (2009) 334–345

Bank

m[sscfidbo

oiS

3

aetlieac

aspw

TOa

V

EEASL

azpi

vmcutnttdntyPquadratic effects are concerned, the quadratic term of the extractiontemperature and the interaction between the temperature and theliquid volume ratio had a positive effect on the response, whereasthe interaction between temperature and agitation had a negative

Fig. 1. Sampling locations in Western Scheldt estuary ((1)

ajor ions formed were the molecular fragment [M+] and theM − Br]+. EI provides a much better sensitivity for mono-bromineubstitutes. The improvement of sensitivity with ECNI dependstrongly on the bromine content. Hence, the three isomers of 1,5,9-yclododecatriene gave no signal, whereas analytes with three andve bromine atoms exhibited greater areas in ECNI than in EI. Theifferences in sensitivity among the two dibromotoluenes and theenzal bromide, all with two bromide atoms, are due to the positionf the halogen in the chemical structure.

Unfortunately, the CTC Pal, used for HS and SPME extraction, wasnly available for the 5973 MS, the instrument without chemicalonization mode. Therefore, the ECNI mode could only be used forPE and LLE analyses.

.2. Optimization of the HS extraction

A CCC design was used to evaluate the significance of the mainnd interaction effects of the factors investigated, resulting in 29xperiments. The experimental domain (Table 2) of every fac-or was defined taking into account instrumental and operativeimits. Three replicate measurements at the centre of the exper-mental domain were performed in order to evaluate both thexperimental error and the existence of curvature. The sum of therea of all the target compound peaks was chosen as the responseriterion.

Fig. 2 shows the significant factors and their coefficients, as well

s the correlation (R2) and predictive coefficients (Q2). As can beeen both values are higher than 0.8, indicating a model with a highredictive power. The influence and significance of the variablesas evaluated by comparing the obtained coefficients for each vari-

able 2ptimal values for extraction temperature and time, agitation speed, salt contentnd ratio liquid–headspace volume in HS analysis

ariable Experimental domain Optimum

xtraction temperature (T, ◦C) 30–80 80xtraction time (tm, min) 15–60 15gitation speed (Ag, rpm) 250–600 250alt content (%S) 0–30 0iquid volume/Headspace volume (%f) 50–80 80

Fml

of the Ghent-Terneuzen Canal, (2) Western Scheldt Bank).

ble (y-axis value) with its error bar. When the error bar crosses theero-axis, the value of the corresponding coefficient could be eitherositive or negative, which implies that the factor has negligible

nfluence.Extraction temperature (T) and the liquid volume/headspace

olume ratio (%f) are positively correlated with the response, whicheans that the higher the temperature, the higher the amount of

ompounds released to the headspace and the more the vial is filledp, the lower the headspace volume, and thus the higher concen-ration of the analyte. Agitation speed (Ag), on the other hand, isegatively correlated with the response, indicating that the massransfer from the liquid sample is not favoured by an increase ofhe turbulence. The equilibrium was reached under the tested con-itions, as it was demonstrated by the fact that the response wasot influenced by the extraction time (tm). Although the overallrend showed no influence of the salt content on the extractionield, the factor positively affected the extraction of TBT, TBA andBT and negatively the removal of CeBr. As far as the interaction and

ig. 2. Scaled and centered coefficients for significant factors in the response surfaceodeling CCC design (R2 = 0.9549; Q2 = 0.8823; degrees of freedom = 18; confidence

evel = 0.95). For the identification of the factor see Table 2.

togr. A

er

u

3

ctt(twbiaav

tstoowv9ottmb

en

swaodntaTr

orgttao2ptacbp

FY

P. López et al. / J. Chroma

ffect. These three quadratic terms defined the curvature of theesponse surface.

The optimal HS extraction conditions, shown in Table 2, weresed for method validation.

.3. Optimization of the SPME procedure

In a first step, the parameters affecting the microextraction pro-ess were evaluated by a factorial design. These parameters werehe type of fiber coating (100 �m PDMS and 85 �m CAR-PDMS),he extraction temperature (40, 60, 80 ◦C) and the extraction modedirect immersion DISPME and headspace HSSPME). A multifac-or categorical 2*3*2 design involving 12 runs and three replicatesas selected. This standard design, consisting of all possible com-inations of levels and factors, has the advantage of providing

nformation of the main effects as well as the two-factor inter-ctions. Other variables such as extraction time, salt content andgitation speed were kept constant in this design experiment atalues of 30 min, 0% NaCl and 250 rpm respectively.

The analysis of variance (ANOVA) was the statistic test selectedo study the results and determine which factors had a statisticallyignificant effect on each compound, as well as which interac-ions between factors are significant (data not shown). The resultsf ANOVA are given as F-ratios, which measure the contributionf each factor on the variance of the response, and the p-values,hich test the statistical significance of each factor. When the p-

alue is less than 0.05, the factor has a significant effect at the5% confidence level. The most significant factor for the majority

f the compounds was the type of fiber, followed by the interac-ion between the type of fiber and temperature. The extractionemperature was found to be relevant for 17 analytes, being the

ost significant factor for the monobromotoluenes and the cetylromide. The mode of extraction was to be considered when

cm

ep

ig. 3. PLS plots showing the relationship between X = fiber (PDMS (p), CAR-PDMS (c)),= response area of the target compounds. See Table 1 for the identification of the compo

1216 (2009) 334–345 339

xtracting the more water-soluble target analytes, dibromophe-ols, bromoanisoles and bromoanilines.

The data were also analyzed statistically using partial leastquares regression (PLS) [50] (Fig. 3). PLS is a multivariate methodith the sample theoretical foundation as principal component

nalysis (PCA). In PLS, two matrices, X and Y, are related to eachther so that the information in Y has an influence on the bilinearecomposition of X into a low number of components (PLS compo-ents). In this work the x-variables (factors) consisted of the fiber,he mode of extraction (type in Fig. 3) and the extraction temper-ture, while the y-variables were the responses of each analyte.he data are presented as loading and score plots that show theelationships between the x- and y-variables and their importance.

Three principal components (PCs) were needed to explain 81%f the variance of the model (Fig. 3). PC1 (50% of variance) is cor-elated with the fiber used during the extraction process and, as aeneral rule, the PDMS fiber was the most effective for the extrac-ion of the heaviest compounds, whereas CAR-PDMS fiber providedhe best results for most volatiles, such as monobromotoluenesnd monobromanisoles, thus being placed on the negative x-axisf Fig. 3B. It is interesting to notice that such tendency is shifted for,4-DBPh and 2,4-DBA in immersion mode and at extraction tem-eratures higher than 60 ◦C. This outcome could be explained byhe differences in the sorption mechanism between the fibers. Thedsorption on particle surfaces is the preferred mass transfer pro-ess in case of CAR-PDMS, whereas absorption partitioning in theulk of the phase is dominant in case of PDMS [51]. Lighter com-ounds are retained into the pore of polymeric phase and heavier

ompounds might diffuse better into the pure phase. It is worthentioning that 2,4-DBAl was only extracted with PDMS.PC2 (17%) discriminated analytes according to their optimal

xtraction temperature. As can be seen, the differences with tem-erature are more noticeable with PDMS than with CAR-PDMS. For

type (headspace (hs), direct immersion (di)) and temperature (T 80, 60 40) andunds.

340 P. López et al. / J. Chromatogr. A 1216 (2009) 334–345

PME a

mdttt

fgbhiThnftww

pdltC

tdDCesCswiw

wstC6ar

Sa

AeIse

ifmstiNaatvesktertc

ascrtptFsw

we

3

Fig. 4. Response surface of the overall S

ost of the compounds, except for PBT, CeBr and monoBA, theifferences between fibers at 40 and 60 ◦C were more noticeablehan between 60 and 80 ◦C. The maximum overall response andhe minimum standard deviation was achieved at 60 ◦C, regardlesshe fiber.

PC3 (14%) described the behaviour of the target analytes as aunction of the type of extraction. As can be seen, most of the tar-et compounds are placed on the positive y-axis, which meant thatoth fibers were more effective in direct immersion (DI) than ineadspace (HS) mode and which could be also expected consider-

ng the low vapour pressures and the Henry’s law constant values.his is in agreement with other studies referring to haloanisoles andalophenols [25], but in contradiction with reports on bromophe-ols [4] and PBDEs [36], where the HS was the most successful mode

or sampling. However, a thorough look to these results revealedhat the HS response considerably increased with temperature,ith 100 ◦C as optimal value. In this work, temperatures over 80 ◦Cere avoided to prevent sample from boiling and reacting.

Confirming the ANOVA results, the interaction between tem-erature and type of fiber had also a negative influence on theescription of PC3, which could be observed for those compounds,

ike 2,5DBT, 2,4DBA, 2,4DBAl and 2,4DBPh, for which the extrac-ion yield continuously increased with temperature when usingAR-PDMS, but did not when PDMS.

Analyzing all information, a compromise should be reachedo achieve the optimal conditions for all the target compoundsue to their diverse behaviour and physical–chemical properties.espite the better response for the most volatile compounds withAR-PDMS, PDMS was the preferred fiber. In addition to its highfficiency for heavier compounds, it was the only one capable ofampling 2,4-DBAl and it did not show so much carryover as theAR-PDMS. When PDMS is used, low temperatures and immersioneemed to be the most effective conditions, except for PBT and CeBr,hose extraction yield strongly increased with temperature. Tak-

ng into account the overall tendency, immersion mode and 60 ◦Cere the values selected for the other two significant factors.

The effect of salt content, agitation speed and extraction timeas studied in a further multifactorial experiment design. The

alt content was evaluated over a range of 0–30%, the agita-ion speed between 250 rpm (minimum agitation speed of theTC Pal) and 600 rpm and the extraction time between 15 and0 min. The Box Benhken design (Fig. 4) was validated by ANOVA

nd their coefficients were estimated by partial least squaresegression.

Extraction time was a relevant factor for most of the analytes.PME is not an exhaustive extraction technique since analytesre partitioned between the bulk aqueous phase and the fiber.

dtLT

rea (y-axis = overall standardized area).

lthough maximum sensitivity is attained at equilibrium, completequilibrium is not necessary for an accurate and precise analysis.n this case, the negative value of the coefficient indicated that theteady equilibrium was already reached at the lowest value of thexperimental domain.

From these data, it can also be concluded that salt contents of paramount relevance for the extraction compounds with aunctional group, such as bromophenols, bromoanisoles and bro-

oanilines. Addition of salt increases the ionic strength of theolution. Depending on the solubility of the target analytes, extrac-ion is usually enhanced with increased salt concentration andncreased polarity of the target compounds (salting-out effect).evertheless, the increment of the viscosity and density of thequeous phase could negatively affect the kinetics of the processnd, consequently, the extraction efficiency. In summary, the addi-ion of salt might be favourable from a thermodynamic point ofiew but unfavourable from a kinetic point of view. This couldxplain why the extraction of CeBr, an apolar compound withlow kinetics, decreased and the extraction of TBA, with slowinetics but polar behaviour, increased. In addition, the salt con-ent also influenced the response through its interaction withxtraction time for all the analytes. Fig. 4A shows that the overallesponse is boosted with long extraction times at low salt concen-rations, whereas shorter extraction times are better at high saltoncentrations.

The interaction between extraction time and agitation speed issignificant factor for those analytes that were not affected by the

alt content, such as bromotoluenes and the three isomers of 1,5,9-yclododecatriene. Agitation of the sample enhances extraction andeduces the time to reach the thermodynamic equilibrium sincehe diffusion film in the aqueous phase, where the mass transferrocess occurs, is decreased as consequence of the increase in theurbulence flow [52]. This interaction could also be observed inig. 4B. Longer times were needed to reach the equilibrium withlow agitation speed, whereas this state was attained much fasterhen speeding up.

In view of all this data, the optimal SPME extraction conditionsere fixed at 30% salt content, 600 rpm agitation speed and 15 min

xtraction.

.4. Method validation

The methods were validated and compared in terms ofetection limits, linearity, precision and trueness by usinghe optimized conditions (SPME and HS procedures). SPE andLE samples were analyzed by GC–(EI)MS and GC–(ECNI)MS.ables 3–5 show the LODs, linearity, linear range, goodness-

P.Lópezet

al./J.Chromatogr.A

1216(2009)

334–345341

Table 3Method performance figures of the LLE–GC(EI)MS and SPE–GC–(EI)MS method (n.d.: non-detected)

Compound LLE–GC–(EI)MS SPE–GC–(EI)MS

LODs (ng/L) Linearrange (ng/L)

Truenessa

(RSD%, n = 3)R2 Repeatabilitya

(RSD% n = 3)LODs (ng/L) Linear

range (ng/L)Truenessb

(RSD%, n = 3)R2 Repeatabilityb

(RSD% n = 3)

2BT 3.6 3.58–294 n.d. 112% 0.9973 n.d. 12% 19 19–4579 63% 60% 0.9985 3% 11%4BT 3.7 3.7–146 n.d. 101% 0.9995 n.d. 13% 25 25–5889 64% 61% 0.999 4% 11%3BT 3.7 3.7–141 n.d. 107% 0.9995 n.d. 11% 22 22–5684 65% 62% 0.9987 4% 17%4BA 1.4 1.4–110 n.d. 117% 0.9976 n.d. 16% 42 42–4794 96% 91% 0.9972 n.d. 8%2BA 4.3 4.3–143 n.d. 119% 0.9995 n.d. 16% 17 17–5286 94% 89% 0.9974 14% 6%EEE 4.7 4.7–137 n.d. 94% 0.9998 n.d. 14% 38 38–6115 41% 38% 0.9991 18% 14%EEZ 11 11–144 n.d. 94% 0.9935 n.d. 14% 16 16–5117 41% 39% 0.9977 11% 21%3,5DBT 2.2 2.2–149 n.d. 106% 0.9945 n.d. 10% 23 23–6001 85% 81% 0.9993 16% 7%2,5DBT 1.9 1.9–143 n.d. 106% 0.9966 n.d. 11% 28 28–5756 85% 81% 0.9983 5% 7%EZZ 2.2 2.2–134 n.d. 75% 0.9992 n.d. 13% 24 24.1–5275 47% 45% 0.9956 n.d. 13%BBr 5.1 5.1–50 n.d. 78% 0.9998 n.d. 42% 171 171–5434 43% 41% 0.9952 17% 31%DeBrc >121 – n.d. 79% – n.d. 15% 27 27–4724 61% 58% 0.9981 22% 14%2,4DBPh >140 – n.d. 103% – n.d. 23% 235 235–5286 n.d. 93% 0.9933 n.d. 3%2,6DBPh >142 – n.d. 87% – n.d. 45% 32 32–5933 85% 80% 0.9951 26% 4%2,4DBA 6.0 6.0–148 n.d. 106% 0.9993 n.d. 8% 107 107–5912 87% 83% 0.9912 n.d. 3%2,4DBAl >140 – n.d. 112% – n.d. 36% 275 275–6323 75% 81% 0.983 n.d. 20%TBT 1.4 1.4–131 n.d. 106% 0.9989 n.d. 10% 27 27–6352 80% 86% 0.9977 14% 2%TBA 1.7 1.7–154 n.d. 110% 0.9988 n.d. 13% 33 33–5387 96% 91% 0.9979 18% 0%TBAl 3.0 3.0–144 n.d. 136% 0.9875 n.d. 29% 26 26–2548 109% 103% 0.992 27% 5%CeBr 2.6 2.6–133 n.d. 7% 0.9964 n.d. 39% 83 83–1047 109% 103% 0.9996 n.d. 7%PBT 0.3 0.3–45 87% 97% 0.9989 15% 15% 9.7 9.7–2925 107% 101% 0.9979 19% 3%

a Spiked at 2 and 2000 ng/L.b Spiked at 40 and 2000 ng/L.c Compounds in italics showed LODs higher than the highest point of the calibration curve.

342 P. López et al. / J. Chromatogr. A

Tab

le4

Met

hod

per

form

ance

figu

res

ofth

eLL

E–G

C–(

ECN

I)M

San

dSP

E–G

C–(

ECN

I)M

Sm

eth

od(n

.d.:

non

-det

ecte

d)

Com

pou

nd

LLE–

GC

–(EC

NI)

MS

SPE–

GC

–(EC

NI)

MS

LOD

s(n

g/L)

Lin

ear

ran

ge(n

g/L)

Tru

enes

sa

(RSD

%,n

=3)

R2

Rep

eata

bili

tya

(RSD

%n

=3)

LOD

s(n

g/L)

Lin

ear

ran

ge(n

g/L)

Tru

enes

sb

(RSD

%,n

=3)

R2

Rep

eata

bili

tyb

(RSD

%n

=3)

2BT

7.4

7.4–

290

n.d

.75

%0.

9956

n.d

.8%

1515

–240

876

%62

%0.

9954

11%

12%

4BTc

>144

–n.

d.77

%–

n.d.

6%50

50–3

097

78%

81%

0.98

44

n.d

.8%

3BT

>108

–n.

d.78

%–

n.d.

6%33

33–2

987

62%

62%

0.99

336%

10%

4BA

>148

–n.

d.78

%–

n.d.

5%18

19–2

522

109%

105%

0.99

667%

4%2B

A12

12–1

76n

.d.

90%

0.99

63n

.d.

3%16

16–2

779

88%

80%

0.99

49

1%2%

3,5D

BT

0.2

0.2–

1994

%95

%0.

9998

3%5%

0.2

0.2–

264

75%

69%

0.99

435%

9%2,

5DB

T0.

10.

1–18

94%

99%

0.99

943%

5%0.

50.

5–13

951

%67

%0.

9914

10%

13%

BB

r6.

76.

7–16

3n

.d.

66%

0.99

94n

.d.

47%

0.1

0.1–

239

20%

57%

0.99

5740

%24

%2,

4DB

Ph1.

51.

5–15

3n

.d.

76%

0.99

56n

.d.

13%

0.1

0.1–

233

71%

78%

0.99

5113

%6%

2,6D

BPh

1.4

1.4–

173

n.d

.72

%0.

9979

n.d

.36

%0.

10.

1–26

150

%76

%0.

994

817

%7%

2,6D

BAl

0.1

0.1–

1810

2%80

%0.

9993

9%5%

0.1

0.1–

269

67%

82%

0.99

5011

%3%

2,4D

BA0.

20.

2–18

102%

95%

0.99

937%

24%

0.1

0.1–

149

70%

81%

0.99

3311

%6%

2,4D

BAl

0.4

0.4–

1713

8%92

%0.

9992

1%6%

1.4

1.4–

278

46%

64%

0.99

394%

5%TB

T0.

30.

3–16

95%

96%

0.99

9213

%6%

0.1

0.1–

280

61%

71%

0.99

658%

9%TB

A0.

050.

05–1

996

%99

%0.

9991

2%5%

0.1

0.1–

237

65%

81%

0.99

5710

%6%

TBA

l0.

10.

1–18

107%

93%

0.99

972%

8%0.

10.

1–21

370

%82

%0.

9963

12%

4%C

eBr

2.0

2.0–

7.1

n.d

.76

%0.

9940

n.d

.15

%2.

22.

2–22

876

%85

%0.

9961

7%10

%PB

T0.

020.

02–1

597

%92

%0.

9993

5%2%

0.1

0.1–

219

64%

82%

0.99

6211

%6%

aSp

iked

at2

and

200

ng/

L.b

Spik

edat

40an

d20

00

ng/

L.c

Com

pou

nd

sin

ital

ics

show

edLO

Ds

hig

her

than

the

hig

hes

tp

oin

tof

the

cali

brat

ion

curv

e.

om

pawecddtiwwcptard

acsw(ncpwoeAPLwa

lelvbuapitt

lqotSf

scitiead

1216 (2009) 334–345

f-fit and repeatability, expressed as %RSD, of the six testedethods.In general, good linearity was demonstrated for most of the com-

ounds over two or more orders of magnitude. Only when HS ispplied, the linearity is reduced by one order of magnitude. It isorth mentioning that 2,6DBAl was also omitted from the SPME

xperiments since a peak from the fiber itself interfered with theompound peak, even in the SIM mode. To validate the regressionata, an analysis of variance (ANOVA) was performed 95% of confi-ence level. Since p-values for lack-of-fit test were greater or equalo 0.05, the linear-first-order models were suitable for the exper-mental data, with the exception of 2,4DBPh when SPME and 3BT

hen HS. Bromophenols are compounds with a high water affinity,hich makes their extraction with organic solvents more diffi-

ult. Derivatization to form the corresponding bromophenyl acetaterior to extraction would improve their isolation from water andheir analysis by GC techniques. The chromatographic resolutionnd peak shape of the derivatized bromophenol is vastly supe-ior to that of the underivatized compound, which results in loweretection limits [24].

Limits of detection (LODs) were calculated as the averagemount of analyte giving a response that is three times the heighthromatographic noise. If the methods with the same ionizationource were compared, the lowest overall LODs were obtainedith SPE (9.7–235 ng/L), followed by SPME (0.8–554 ng/L), LLE

0.03–(>140 ng/L)) and HS (77–(>33,600 ng/L)). The dibromophe-ols and 2,4DBAl were not detected at the highest level of thealibration curve when LLE or HS. If the ionization mode was com-ared, the LODs for compounds with two or more bromine atomsere lower when ECNI, which could be expected from the results

f Table 1. Monobromotoluenes and monobromoanisoles were notven detected at the highest level of the calibration curve when LLE.s for the individual target compounds and extraction technique,BT was the analyte with the lowest LODs, whereas the highestODs corresponded to DeBr and BBr when LLE, DeBr and 2,4DBPhhen SPME, diBPhs and 2,4-DBAl when HS and monoBT, monoBA

nd the isomers of 1,5,9-cyclododecatriene when SPE.Method precision was evaluated by testing at two concentration

evels. The intra-day repeatability is shown in Tables 3–5. The low-st RSD% values (1–40% for the low level and 2–24% for the highevel) were obtained when SPE–GC–(ECNI)MS, which is quite con-enient considering that no internal standard was used. It shoulde mentioned that there was a large difference in behaviour whensing LLE. Some compounds showed a good repeatability, but PBTnd BBr in particular showed a considerable variance. Between-dayrecision was evaluated verifying homoscedasticity and perform-

ng ANOVA on the data acquired over three days. ANOVA indicatedhat the mean values were not significantly different among thehree days (p > 0.05).

Trueness (extraction recoveries) was calculated at the same twoevels as the precision. Results showed that SPME–GC–(EI)MS wasuantitative for almost all the analytes. Good recoveries were alsobtained when LLE was applied for all the compounds with excep-ion of PBT and the bromides, regardless of the ionization mode.PE–GC–(ECNI)MS provided recoveries between 65 and 95%, exceptor BBr and 2,6DBPh.

In summary, none of the methods studied was found to beuitable for the simultaneous determination of all the targetompounds in real samples. The most convenient extraction andnstrumental technique for each target compound, together with

heir analytical features, are reviewed in Table 6. Consideringts high LODs and its inability to measure dibromophenols, HSxtraction should be excluded from the final selection. LLE waslso rejected since it entails the consumption of large volumes ofichloromethane and is rather time-consuming and ineffective for

P.Lópezet

al./J.Chromatogr.A

1216(2009)

334–345343

Table 5Method performance figures of the HS–GC(EI)MS and SPME–GC–(EI)MS method (n.d.: non-detected)

Compound HS–GC(EI)MS SPME–GC–(EI)MS

LODs (ng/L) Linear range (ng/L) Truenessa

(RSD%, n = 3)R2 Repeatabilitya

(RSD% n = 3)LODs (ng/L) Linear range (ng/L) Truenessb

(RSD%, n = 3)R2 Repeatabilityb

(RSD% n = 3)

2BT 353 353–7070 78% 78% 0.9949 4% 9% 29 29–87900 102% 101% 0.9993 19% 22%4BT 628 628–37000 78% 86% 0.9919 8% 9% 34 34–11300 103% 98% 0.9984 22% 22%3BT 364 364–35800c 91% 80% 0.9957 2% 9% 23 23–109000 100% 101% 0.9986 18% 22%4BA 1430 1430–27700 n.d. 75% 0.9902 n.d. 15% 40 40–101000 119% 100% 0.9968 20% 21%2BA 424 424–34800 70% 77% 0.9907 11% 7% 28 28–101000 120% 102% 0.9960 21% 17%EEE 460 460–34800 88% 110% 0.9911 11% 22% 8 8–117000 122% 99% 0.9990 9% 16%EEZ 1040 1040–37000 80% 78% 0.9902 5% 5% 14 14–98500 127% 101% 0.9992 11% 16%3,5DBT 2160 2160–38200 n.d. 96% 0.9814 n.d. 12% 9.0 9.0–115000 111% 102% 0.9995 25% 20%2,5DBT 1920 1920–36700 n.d. 90% 0.9821 n.d. 17% 11 11–110000 159% 108% 0.9989 19% 20%EZZ 2220 2220–33400 n.d. 104% 0.9903 n.d. 16% 11 11–101000 45% 132% 0.9995 11% 16%BBr >30500 – n.d. n.d. – n.d. n.d. 168 168–104000 129% 103% 0.9985 21% 16%DeBr 5500 5500–36000 n.d. 85% 0.9965 n.d. 7% 554 554–96700 n.d. 95% 0.9989 n.d. 12%2,4DBPh >27000 – n.d. n.d. – n.d. n.d. 4160 4160–52800c n.d. 146% 0.977 n.d. 18%2,6DBPh >31100 – n.d. n.d. – n.d. n.d. 41 41–71600 76% 67% 0.9986 21% 22%2,6DBAld 8060 8060–186000 n.d. n.d. 0.9919 n.d. 10% – – – – – – –2,4DBA 5900 5900–71900 n.d. n.d. 0.9965 n.d. 15% 0.6 0.6–113000 135% 120% 0.9992 27% 21%2,4DBAl >33100 – n.d. n.d. – n.d. n.d. 185 185–121000 108% 106% 0.9931 7% 19%TBT 65 65–14800 87% 100% 0.9959 2% 9% 2.0 2.0–121000 145% 105% 0.9999 16% 16%TBA 327 0.327–17900 106% 80% 0.9932 13% 11% 0.9 0.9–103000 139% 125% 0.9978 19% 18%TBAl 3430 3.43–16400 n.d. 45% 0.9923 n.d. 10% 7.0 7.0–93000 130% 123% 0.9936 24% 16%CeBr 518 518–15100 193% 90% 0.9904 4% 2% 6.0 6.0–8690 74% 86% 0.9991 16% 5%PBT 77 77–13400 121% 102% 0.9963 8% 7% 0.8 0.8–9340 70% 85% 0.9910 19% 16%

Compounds in italics showed LODs higher than the highest point of the calibration curve.a Spiked at 2000 and 20,000 ng/L.b Spiked at 200 and 20,000 ng/L.c p > 0.05.d It was omitted with SPME due to the presence of an interference from the fiber itself.

344 P. López et al. / J. Chromatogr. A 1216 (2009) 334–345

Table 6Recommended extraction and detection technique for each individual compound

Analyte Extraction Ion source LODs (ng/L) Trueness (RSD%, n = 3) Repeatability (RSD% n = 3)

2BT SPME EI 29 102% 101% 19% 22%4BT SPME EI 34 103% 98% 22% 22%3BT SPME EI 23 100% 101% 18% 22%4BA SPME EI 40 119% 100% 20% 21%2BA SPME EI 28 120% 102% 21% 17%EEE SPME EI 8 122% 99% 9% 16%EEZ SPME EI 14 127% 101% 11% 16%3,5DBT SPE ECNI 0.2 88% 80% 5% 9%2,5DBT SPE ECNI 0.5 75% 69% 10% 13%EZZ SPME EI 11 45% 132% 11% 16%BBr SPE ECNI 0.1 20% 57% 40% 24%DeBr SPE EI 27 61% 58% 22% 14%2,4DBPh SPE ECNI 0.1 71% 78% 13% 6%2,6DBPh SPE ECNI 0.1 50% 76% 17% 7%2,6DBAl SPE ECNI 0.1 67% 82% 11% 3%2,4DBA SPE ECNI 1.4 70% 81% 11% 6%2,4DBAl SPE ECNI 0.1 46% 64% 4% 5%TBT SPE ECNI 0.1 61% 71% 8% 9%TTCP

tsfstaaIosrtm1ab

3

oS

fcphtmro1(

D(gDs

F

BA SPE ECNI 0.1BAl SPE ECNI 2.2eBr SPE ECNI 0.1BT SPE ECNI 0.1

he quantitative extraction of the more water-soluble compounds,uch as bromphenols and bromoanilines. Based on its analyticaleatures SPE–GC–(ECNI)MS has been confirmed to be the mostuitable method for the extraction of the analytes with morehan one bromine atom, whereas SPME–GC–(EI)MS is the mostdvisable for those compounds that own one or none brominetoms in their chemical structure, with the exception of DeBr.n view of the main goal of this work, which is the validationf a unique method based on one extraction and one analysistep to determine simultaneously as many as brominated flameetardants and by-products as possible, and of its analytical fea-ures, SPE–GC–(ECNI)MS could also be convenient for monoBT and

onoBA determination. On the other hand, the three isomers of,5,9-cyclodocecatriene (EEE, EEZ and EZZ) could only be identifiednd quantified in EI ionization mode, thus SPME–GC–(EI)MS haseen selected for their determination.

.5. Analysis of real samples

The proposed methods were applied to study the occurrencef the target compounds in free water samples from the Westerncheldt.

wpi2o

ig. 5. Chromatogram of the water sample 2a (Peak identification: (1) 3,5DBT + 2,5DBT, (2

65% 81% 10% 6%70% 82% 12% 4%76% 85% 7% 10%64% 82% 11% 6%

The distribution of neutral organic contaminants among the dif-erence phases in surface water is predicted from the octanol–wateroefficients (Kow). Contaminants with Kow-values below 4 arereferably dissolved in the water matrix, whereas the moreydrophobic contaminants, with Kow-values higher than 4, tendo be adsorbed on the surface of the suspended particulate

atter (SPM) [53]. In this work, the SPM was not previouslyemoved by filtration, which implied that the determinationf the more hydrophobic analytes, such as the isomers of,5,9-cyclododecatriene (Kow = 5.52), CeBr (Kow = 8.54) and PBTKow = 6.99), was not altered.

The samples (Fig. 5) contained about 120 bromine compounds.espite the proximity of large cities, harbour, industries and users

textile) of bromine products are located upstream [54], the ori-in and source – biogenic and/or anthropogenic – is not clear.ebromination and bromination of compounds can occur under

pecific environmental conditions. Often, it is difficult to determine

hether an identified substance is of biogenic or of anthro-ogenic origin [55]. The SPE–GC–(ECNI)MS method allowed the

dentification and quantification of 2,5DBT, 3,5DBT, BBr, 2,4DBPh,,6DBPh, 2,6DBAl, 2,4DBA, TBT, TBAl, TBA and PBT. The presencef TBT, TBA and TBAl was confirmed with the SPME–GC–(EI)MS

) BBr, (3) 2,4DBPh, (4) 2,6DBPh, (5) 2,6DBAl, (6) 2,4.DBAl, (7) TBA, (8) TBAl, (9) PBT).

P. López et al. / J. Chromatogr. A

Table 7Concentration of target compounds in water samples from the Western Scheldtestuary (ng/L) using SPE–GC–(ECNI)MS as analytical method

Analyte 1aa 1bb 2aa 2bb

3,5DBT 6.5 1.1 2.7 <0.22,5DBT 12 1.9 21 15.5BBr 1.1 0.3 0.8 <0.12,4DBPh 8.5 12 25 8.32,6DBPh <0.1 0.2 1.6 <0.12,6DBAl <0.1 <0.1 0.7 1.72,4DBA 3.8 1.0 <0.1 <0.12,4DBAl 1.9 <1.4 2.2 <1.4TBT <0.1 <0.1 0.3 <0.1TBA 2.1 0.7 0.4 <0.1TBAl 0.7 0.5 0.3 <0.1P

mL

tosco

Crceb

4

ntsnrEap

bkc

A

DPt

R

[[[[

[[[[[[

[

[[[[[

[

[[

[

[[

[

[[[[

[[

[

[[

[

[[

[

[[

[

[

[

[

BT <0.1 <0.1 2.4 <0.1

a Sampling conducted at the beginning of the month.b Sampling conducted at the end of the month.

ethod (Table 7). The other target analytes were all below theODs.

The detected levels of 2,4DBPh (4 �g/L) are below its odourhreshold, whereas sample 2a exceeded the taste thresholdf 2,6DBPh (0.5 ng/L). As for TBA threshold, most of theamples values were greater than 0.03 ng/L [22,25], whichould be olfactory confirmed by the characteristic earthy-mustydour.

No conclusion about the contribution of the Ghent-Terneuzenanal to the overall contamination of Western Scheldt could beeached, as far as the target compounds were concerned. More dataoncerning to the flow of the two streams, the activity and the efflu-nts charges from the surrounding manufactories and so on, shoulde known to reach a conclusive statement.

. Conclusions

Due to its low detection limits, good linearity, precision, true-ess and selectivity SPE–GC–(ECNI)MS is the preferred method forhe simultaneous determination (one extraction and one analysistep) of bromoanisoles, bromanilines, bromotoluenes, bromophe-ols and bromoalkanes in water, whereas SPME–GC–(EI)MS isecommended to quantify the isomers of 1,5,9-cyclododecatriene.CNI is more sensitive for the analytes with more than two brominetoms, whereas EI is more suitable for the quantification of com-ounds with only one bromine atom.

The feasibility of the SPE–GC–(ECNI)MS has been also confirmedy analyzing some water samples from the Western Scheldt. To ournowledge, this is the first time that such wide range of brominatedompounds has been analyzed in a single run.

cknowledgements

We would like to acknowledge Mrs. M. van den Heuvel-Greve ofeltares, Delft, The Netherlands, for providing the water samples.. López acknowledges Caja Madrid (Spain) for personal fundinghrough its Fellowship Program.

eferences

[1] EC 1907/2007, OJL 396, 30.12.2006, p. 1.[2] J. de Boer, P.G. Wester, A. van der Horst, P.E.G. Leonards, Environ. Poll. 122 (2003)

63.[3] J. Eriksson, S. Rahm, N. Green, A. Bergman, E. Jakobsson, Chemosphere 54 (2004)

117.

[[[[

1216 (2009) 334–345 345

[4] M. Polo, M. Llompart, C. Garcia-Jares, G. Gomez-Noya, M.H. Bollain, R. Cela, J.Chromatogr. A 1124 (2006) 11.

[5] A.S. Allard, M. Remberger, A.H. Neilson, Appl. Environ. Microbiol. 53 (1987) 839.[6] Toa Gosei Chem. Ind. Ltd. (TOAG-C), Patent number: JP5105725-A, JP3021454-

B2.[7] E. Jakab, Md.A. Uddin, T. Bhaskar, Y. Sakata, J. Anal. Appl. Pyrolysis 68–69 (2003)

83.[8] J.W. Davis, S.J. Gonsior, D.A. Markham, U. Friederich, R.W. Hunziker, J.M. Ariano,

Environ. Sci. Technol. 40 (2006) 5395.[9] G.W. Gribble, Chem. Soc. Rev. 28 (1999) 335.10] W. Vetter, D. Janussen, Environ. Sci. Technol. 39 (2005) 3889.11] W. Vetter, G.W. Gribble, Environ. Toxicol. Chem. 26 (2007) 2249.12] J. de Boer, C. Allchin, R. Law, B. Zegers, J.P. Boon, Tr. Anal. Chem. 20 (2001) 591.13] J.P. Boon, W.E. Lewis, M.R. Tjoen-A-Choy, C.R. Allchin, R.J. Law, J. de Boer, C.C.

ten Hallers-Tjabbes, B.N. Zeggers, Environ. Sci. Technol. 36 (2002) 4025.14] A. Covaci, S. Voorspoels, J. de Boer, Environ. Int. 29 (2003) 735–756.15] R. Hites, Environ. Sci. Technol. 38 (2004) 945.16] L.Y. Zhu, R. Hites, Environ. Sci. Technol. 39 (2005) 3488.17] C.A. de Witt, Chemosphere 46 (2002) 583.18] E. Eljarrat, D. Barceló, Tr. Anal. Chem. 23 (2004) 727.19] M. Remberger, J. Sternbeck, A. Palm, L. Kaj, K. Stromberg, E. Brorstrom-Lunden,

Chemosphere 54 (2004) 9.20] H.J.C. Klamer, P.E.G. Leonards, M.H. Lamoree, L.A. Villerius, J.E. Akerman, J.F.

Bakker, Chemosphere 58 (2005) 1579.21] A. Covaci, S. Voorspoels, L. Ramos, H. Neels, J. Chromatogr. A 1153 (2007) 145.22] J.B. Adams, S.J. Lock, M.R. Toward, B.M. Williams, Food Chem. 64 (1999) 377.23] J.L. Acero, P. Piriou, U. von Gunten, Water Res. 39 (2005) 2979.24] J.W. Blythe, A. Heitz, C.A. Joll, R.I. Kagi, J. Chromatogr. A 1102 (2006) 73.25] L. Malleret, J. Dugay, A. Bruchet, M.-C. Hennion, J. Chromatogr. A 999 (2003)

135.26] T. Hassenklöver, S. Predehl, J. Pilli, J. Ledwolorz, M. Assman, U. Bickmeyer, Aq.

Toxicol. 76 (2006) 37.27] J. Legler, A. Brouwer, Environ. Int. 29 (2003) 879.28] C.M. Olsen, E.T.M. Meussen-Elholm, J.A. Holme, J.K. Hongslo, Toxicol. Lett. 129

(2002) 55.29] D.J. Brown, I. Van Overmeire, L.G. Goeyens, M.S. Denison, M.J. de Vito, G.C. Clark,

Chemosphere 55 (2004) 1509.30] L. Müller, E. Fattore, E. Benfenati, J. Chromatog. A 791 (1997) 221.31] A. Covaci, J. de Boer, J.J. Ryan, S. Voorspoels, P. Schepens, Anal. Chem. 74 (2002)

790.32] D. Khiari, A. Bruchet, T. Gittelman, L. Matia, S. Barrett, I.H. Suffett, R. Hund, Water

Sci. Technol. 40 (1999) 129.33] L.A. Reitzel, A. Ledin, J. Chromatogr. A 972 (2002) 175.34] A. Díaz, F. Ventura, M.T. Galcerán, J. Chromatogr. A 1064 (2005) 97.35] A. Weisz, D. Andrezejewski, I.R. Rasooly, J. Chromotogr. A 1057 (2004) 185.36] M. Polo, G. Gómez-Noya, J.B. Quintana, M. Llompart, C. García-Jares, R. Cela,

Anal. Chem. 76 (2004) 1054.37] K. Grob, F. Zürcher, J. Chromatogr. 117 (1976) 285.38] J.B. Quintana, R. Rodil, S. Muniategui-Lorenzo, P. López-Mahía, D. Prada-

Rodríguez, J. Chromatogr. A 1174 (2007) 27.39] N. Ochiai, K. Sasamoto, M. Takino, S. Yamashita, S. Daishima, A. Heiden, A.

Hoffman, Analyst 126 (2001) 1652.40] L.M. Zhao, L.Y. Zhu, H.K. Lee, J. Chromatogr. A 963 (2002) 239.41] R. Westbom, L. Thörnbey, S. Zorita, L. Mathiasson, E. Björklund, J. Chromatogr.

A 1033 (2004) 1.42] C. Pizarro, N. Pérez-del-Notario, J.M. González-Sáiz, J. Chromatogr. A 1143

(2007) 26.43] G. Box, D. Behnken, Technometrics 2 (1960) 455.44] The Fitness for Purpose of Analytical Methods: A Laboratory Guide to Method

Validation and Related Topics, EURACHEM Guide, first English edition 1.0-1998,LGC (Teddington) Ltd., http://www.eurachem.ul.pt/.

45] ISO 5725-1994, Accuracy (Trueness and Precision) of Measurement Methodsand Results, International Standardisation Organisation, Geneva, 1994.

46] N. Draper, H. Smith, Applied Regression Analysis, Wiley, New York, 1981.47] G.E.P. Box, W.G. Hunter, J.S. Hunter, Statistic for Experimental, Wiley, New York,

1978.48] J. de Boer, K. de Boer, J.P. Boon, in: J. Paasivirta (Ed.), The Handbook of Environ-

mental Chemistry, Springer, Berlin, 2000, p. 61.49] U. Sellström, Determination of some polybrominated flame retardants in biota,

sediment and sewage sludge, Ph.D. Thesis, University of Stockholm, Sweden,1999.

50] S. Wold, C. Albana, W.J. Dumm, U. Edlund, K. Esbensen, P. Geladi, S. Helberg, E.Johansson, W. Lindberg, M. Sjöström, in: B.R. Kowalski (Ed.), Mathematics andStatistics in Chemistry, D. Reidel, Dordrecht, 1984, p. 17.

51] J. Pawliszyn, Solid-Phase Microextraction: Therory and Practice, Wiley-VCH,New York, 1997.

52] M.A. Jeannot, F.F. Cantwell, Anal. Chem. 69 (1997) 235.53] F. Smedes, Int. J. Environ. Anal. Chem. 57 (1994) 215.54] S. Voorspoels, A. Covaci, P. Schepens, Environ. Sci. Technol. 37 (2003) 4348.55] N. Reineke, S. Biselli, S. Franke, W. Francke, N. Heinzel, H. Hühnerfuss, H.

Iznaguen, U. Kammann, N. Theobald, M. Vobach, W. Wosniok, Arch. Environ.Contam. Toxicol. 51 (2006) 186.