Talanta 80 (2009) 643–650 Contents lists available at ScienceDirect Talanta journal homepage: www.elsevier.com/locate/talanta Optimisation and validation of programmed temperature vaporization (PTV) injection in solvent vent mode for the analysis of the 15 + 1 EU-priority PAHs by GC–MS José Ángel Gómez-Ruiz 1 , Fernando Cordeiro, Patricia López, Thomas Wenzl ∗ European Commission, Joint Research Centre, Institute for Reference Materials and Measurements (IRMM), Retieseweg 111, B-2440 Geel, Belgium a r t i c l e i n f o Article history: Received 29 April 2009 Received in revised form 3 July 2009 Accepted 20 July 2009 Available online 28 July 2009 Keywords: 15 + 1 EU-priority polycyclic aromatic hydrocarbons (PAHs) Large volume injection (LVI) Programmed temperature vaporization (PTV) Gas chromatography mass spectrometry (GC–MS) Food a b s t r a c t This paper presents the optimisation of a programmed temperature vaporization solvent vent (PTV-SV) injection gas chromatographic mass spectrometric (GC–MS) method for the analysis of the 15 + 1 EUpriority PAHs in food extracts. Three operation parameters (vent time, vent flow and vent pressure) were optimised by applying a D-optimal experimental design. Among these variables, vent time showed the highest effect on the analytical response (signal intensity) of the target PAHs. The 15 + 1 EU-priority PAHs were analysed in solvent solutions and in extracts of fortified sausage. In addition, blank lamb meat extracts were prepared and spiked with the target PAHs prior to GC–MS analysis. The performance of the optimised PTV-SV injection GC–MS method was scrutinised for linearity, precision, matrix effects and robustness. All parameters were found satisfactorily. Compared to PTV injection in splitless mode, the PTV-SV injection method provided an enhancement of sensitivity for all target PAHs. Especially significant was the improvement of the S/N ratios of the compounds with the highest molecular mass. © 2009 Elsevier B.V. All rights reserved. 1. Introduction Polycyclic aromatic hydrocarbons (PAHs) are a large group of organic compounds containing two or more fused aromatic rings constituted of carbon and hydrogen atoms. PAHs constitute a large class of environmental contaminants. They are emitted from diesel engines, but also stationary sources, e.g. different kinds of industries [1]. The main source of exposure to PAHs for non-smoking humans is food. Food might be contaminated by PAHs through the environment or via formation during processing. The dietary intake of PAHs is of concern because some of them can cause cancer in humans [2]. The Scientific Committee on Food (SCF) assessed in 2002 the toxicity of 33 PAHs and concluded that 15 of them are potentially genotoxic and carcinogenic to humans [3]. In 2005, the Joint FAO/WHO Expert Committee on Food Additives (JECFA) identified an additional PAH (benzo[c]fluorene) as probably carcinogenic [4]. The joint set of PAHs is recognized as the 15 + 1 EU-priority PAHs (Table 1). This terminology serves to distinguish them from the 16 PAHs highlighted in the 1970s by the ∗ Corresponding author. Tel.: +32 14 571 320; fax: +32 14 571 783. E-mail address: Thomas.Wenzl@ec.europa.eu (T. Wenzl). 1 Current address: Instituto de Fermentaciones Industriales, Consejo Superior de Investigaciones Científicas (CSIC), C/ Juan de la Cierva 3, 28006 Madrid, Spain. 0039-9140/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2009.07.041 US Environmental Protection Agency (EPA) [5]. Though eight of the 15 + 1 EU-priority PAHs are identical with PAHs of the USEPA list they are challenging from the analytical point of view. Difficulties are related to both separation efficiency and sensitivity. High discrimination of the signal intensities of low and high molecular weight compounds is a common problem in the analysis of PAHs by gas chromatography [6,7]. The lower abundance of the heavier compounds causes also smaller signal-to-noise (S/N) ratios which result in the increase of the limits of detections (LOD). Among the 15 + 1 EU-priority PAHs are four dibenzopyrenes (Table 1), which are especially susceptible to discrimination of signal intensities due to their high molecular mass. Large-volume injection (LVI) techniques such as programmedtemperature-vaporization (PTV) and cool-on-column (COC) injection can be used to increase method sensitivity through the injection of larger volumes of the final extract. PTV injection has also been used to reduce discrimination of certain analytes compared to classical split–splitless injection [8]. In addition, PTV injection is preferred to on-column injection when the analyses involve complex samples [9]. Several studies have been recently reported using PTV methods for the determination of trace pollutants, such as phenols [10], pesticides [11] and polychlorinated biphenyls [12]. Likewise, PAHs have been also determined in environmental samples [12–17] and foodstuff [18,19] using PTV injection. However, it J.Á. Gómez-Ruiz et al. / Talanta 80 (2009) 643–650 644 Table 1 15 + 1 EU-priority PAHs under investigation (with acronyms). The right part of the table (in bold) shows those PAHs that are not included in the 16 EPA PAHs. 1 Benz[a]anthracene (BaA) (Mw 228) 2 Benzo[a]pyrene (BaP) (Mw 252) 10 Cyclopenta[cd]pyrene (CPP) (Mw 226) 3 Benzo[b]fluoranthene (BbF) (Mw 252) 11 Dibenzo[a,e]pyrene (DeP) (Mw 302) 4 Benzo[ghi]perylene (BgP) (Mw 276) 12 Dibenzo[a,h]pyrene (DhP) (Mw 302) 5 Benzo[k]fluoranthene (BkF) (Mw 252) 13 Dibenzo[a,i]pyrene (DiP) (Mw 302) 6 Chrysene (CHR) (Mw 228) 14 Dibenzo[a,l]pyrene (DlP) (Mw 302) 7 Dibenz[a,h]anthracene (DhA) (Mw 278) 15 5-Methylchrysene (5MC) (Mw 242) 8 Indeno[1,2,3-cd]pyrene (IcP) (Mw 276) +1 Benzo[c]fluorene (BcL) (Mw 216) must be stressed that none of the cited studies covered the whole set of 15 + 1 EU-priority PAHs. PTV injection offers more operating parameters than split/ splitless injection. Hence the optimum combination of parameter settings needs to be identified for each analysis task individually. Some authors [11,13] tackled the optimisation of the PTV injection by univariate procedures, whereas others applied multivariate statistical experimental designs for that purpose [12,14,16,20]. From the latter, only Yusà et al. [14] and León et al. [20] focused in their studies on the determination of PAHs, which however covered only a part of the analytes dealt with in this paper. The use of statistically based experimental designs (two-level full factorial designs, fractional factorials, central composite or Plackett-Burman designs) can simplify the optimisation of PTV injection through the use of a selected number of experiments that randomly combines the desired variables. This approach provides a method for the simultaneous investigation of multiple variables, estimating any interaction among them, and requires fewer experiments to complete the optimisation [14]. Only a few papers have been published on the optimisation of PTV injection for the determination of PAHs by GC–MS. They reported a significant increase in sensitivity by the use of PTV injection if compared to splitless injection [13,14,16]. These studies however have focused only on few selected PAHs [14] or on the 16 EPA PAHs [13,16]. A study has been published recently concerning 9 Benzo[j]fluoranthene (BjF) (Mw 252) the determination of high molecular mass PAHs by application of PTV injection GC–MS [21]. This paper reported on the measurement of 15 out of the 15 + 1 EU-priority PAHs in solvent solution and edible oil extracts by using PTV injection (1 ␮L) in splitless mode [21]. However PTV injection in splitless mode has the same limitations with regard to tolerated solvent volume as conventional splitless injection. If the solvent volume exceeds a certain level peak distortions have to be taken into account. PTV injection in solvent vent mode (PTV-SV) in contrary offers the possibility to perform large volume injection and gain thereby extra sensitivity. The limitation of this injection technique is provided by the larger amount of matrix that is injected into the instrument, which might interfere with the analyte signal. However, no studies are currently available that focus on the employment of PTV-SV injection for the GC–MS analysis of the 15 + 1 EU-priority PAHs in food samples. The aim of the present study was the optimisation of a PTV-SV injection method for the GC–MS analysis of the 15 + 1 EU-priority PAHs in foodstuff. The optimisation of different operating parameters of the PTV injection was carried out by application of a statistically based experimental design. Sausage samples prepared with oil fortified with the 15 + 1 EU-priority PAHs and noncontaminated lamb meat samples were used throughout the study. Different performance parameters such as linearity, precision, and matrix effects were evaluated for the optimised PTV-SV injection GC–MS analysis method. Sensitivity and discrimination of signal J.Á. Gómez-Ruiz et al. / Talanta 80 (2009) 643–650 intensities were compared to standard PTV injection in pulsed splitless injection mode (PTV-PSL). 2. Materials and methods 2.1. Reagents and reference materials All solvents were purchased in chromatographic grade from VWR International (Leuven, Belgium). The drying material poly(acrylic acid), partial sodium salt-graft-poly(ethylene oxide) was obtained from Aldrich (Milwaukee, USA) and Ottawa sand from Acros Organics N.V. (Geel, Belgium). The analytes benz[a]anthracene (BaA) CAS#56-55-3, chrysene (CHR) CAS#218-01-9, 5-methylchrysene (5-MC) CAS#3697-24-3, benzo[b]fluoranthene (BbF) CAS#205-99-2, benzo[j]fluoranthene (BjF) CAS#205-82-3, benzo[k]fluoranthene (BkF) CAS#207-089, benzo[a]pyrene (BaP) CAS#50-32-8, indeno[1,2,3-cd]pyrene (IcP) CAS#193-39-5, dibenz[a,h]anthracene (DhA) CAS#53-703, benzo[ghi]perylene (BgP) CAS#191-24-2, dibenzo[a,l]pyrene (DlP) CAS#191-30-0, dibenzo[a,e]pyrene (DeP) CAS#192-65-4, dibenzo[a,i]pyrene (DiP) CAS#189-55-9, dibenzo[a,h]pyrene (DhP) CAS#189-64-0, were commercially available BCR certified reference materials (IRMM, Geel, Belgium). Cyclopenta[cd]pyrene (CPP) CAS#27208-37-3, purity >99.0% by GC, was manufactured on request (Biochemisches Institut für Umweltkarzinogene, Großhansdorf, Germany). Benzo[c]fluorene (BcL) CAS#205-129, purity 99.99%, was purchased from Dr. Ehrenstorfer GmbH (Augsburg, Germany). The isotopically labelled compounds benz[a]anthracene (13 C6, 99%), benzo[a]pyrene (13 C6, 99%), benzo[b]fluoranthene (13 C6, 99%), benzo[ghi]perylene (13 C12, 99%), benzo[k]fluoranthene (13 C6, 99%), chrysene (13 C6, 99%), dibenz[a,h]anthracene (13 C6, 99%), dibenzo[a,e]pyrene (13 C6, 99%), dibenzo[a,i]pyrene (13 C12, 99%), indeno[1,2,3-cd]pyrene (13 C16, 99%) and pyrene (13 C3, 99%), which was used as internal standard for BcL, were obtained from LGC Promochem GmbH (Wesel, Germany). 9-Fluorobenzo[k]fluoranthene, used as injection standard, was purchased from Chiron AS (Trondheim, Norway). Stock solutions of the individual PAHs were gravimetrically prepared in toluene. The different working solutions containing the 15 + 1 EUpriority PAHs were prepared by mixing appropriate volumes from the individual stock solution and adding toluene to achieve the desired concentrations (6 different levels between 15 ng/mL and 80 ng/mL). 2.2. Sample preparation Contaminated sausage samples were prepared at the MaxRubner-Institut (MRI), Bundesforschungsinstitut für Ernährung und Lebensmittel, Institut für Sicherheit und Qualität bei Fleisch (Kulmbach, Germany) by applying spiked olive oil for the production of sausages. The sausage samples were split into portions of 50 g each and packed into aluminium cans. The final analyte contents ranged from 3.9 ␮g kg−1 to 9.9 ␮g kg−1 . Benzo[a]pyrene, the only PAH regulated by current EU legislation, was present at a level of 5.3 ␮g kg−1 , a value slightly above the maximum level specified in Commission Regulation (EC) No 1881/2006 (5 ␮g kg−1 ) [22]. Blank meat extracts were prepared from commercial minced lamb meat. They were fortified with the 15 + 1 EU-priority PAHs prior to the GC–MS analysis in order to evaluate matrix effects of the optimised PTV-SV method. The blank extracts were fortified to a concentration of 50 ng mL−1 for each PAH, a concentration similar to that expected in the final extract of the sausage sample considering sample intake and analyte enrichment. 2.2.1. Pressurised liquid extraction (PLE) Extraction was carried out in accordance with the method described by Jira et al. [23]. Briefly, 5 g of homogenised sausage 645 was mixed with the same amount of poly(acrylic acid), partial sodium salt-graft-poly(ethylene oxide) and 15 g of Ottawa sand. After adding 200 ␮L of the 13 C-labelled standard solution, the mix was placed into 33 mL extraction cells with cellulose filters at the bottom. The extraction was performed with n-hexane (at 100 ◦ C and 100 bars at a static time of 10 min) in an ASE 300 pressurised liquid extraction system (Dionex, Amsterdam, The Netherlands). The flush volume was 60% and the purge time 120 s. Two static extraction cycles were performed per sample. Afterwards, the extractant was evaporated in a Turbo Vap® workstation (Zymark, Hopkinton, USA) at 40 ◦ C using a stream of nitrogen. 2.2.2. Gel permeation chromatography (GPC) The evaporated PLE-extract was dissolved in approximately 5 mL of cyclohexane:ethyl acetate (1:1 v/v) and filtered through PTFE filters of 5 ␮m pore size (Millipore, Bedford, USA). GPC clean up was carried out on a GPC column (25 mm internal diameter and 320 mm length) filled with 50 g Bio-Beads S-X3 (Bio-Rad Laboratories, Nazareth-Eke, Belgium) applying a GPC Ultra instrument (LCTech, Dorfen, Germany) connected to a FW-20 detector (LCTech) that was operated at 254 nm. Chromatographic separation was achieved with cyclohexane:ethyl acetate (1:1 v/v) at a flow rate of 4 mL min−1 . The PAHs containing fraction eluted between 35 min and 80 min. The volume of this fraction was reduced to 4.5 mL by using an automated concentrator (CPC 2000-II Vacuubrand, Wertheim, Germany) that was integrated in the GPC system. The sensitivity of the UV detector was not sufficient to detect the elution of the PAHs. As an alternative the elution of toluene, which got into the extract as a residue of the spiking with internal standards dissolved in toluene, and the elution of the fat fraction were monitored for quality control purposes. Toluene eluted shortly after the fat fraction and directly in front of the first target analyte, benzo[c]fluorene. 2.2.3. Solid phase extraction (SPE) The GPC eluate was further reduced with a stream of nitrogen to 200 ␮L. Then 800 ␮L of cyclohexane were added and the mix was quantitatively loaded onto a preconditioned silica SPE cartridge (500 mg/4 mL, Alltech, Deinze, Belgium). The PAHs were eluted with 10 mL of cyclohexane. 2.2.4. Preparation for GC–MS analysis Toluene (200 ␮L) was added as a keeper to the SPE eluate, which was then concentrated to a volume of about 200 ␮L. Afterwards, 300 ␮L of toluene containing 9-fluorobenzo[k]fluoranthene was added, which served for monitoring the variability of injection. 2.3. Instrumentation and analytical conditions A gas chromatograph HP 6890N from Agilent Technologies (Waldbronn, Germany) with a programmable-temperature vaporization (PTV) injection port (septumless head) was used for the analysis of the target PAHs. The GC was coupled to an Agilent 5975B single quadrupole mass spectrometer (Agilent Technologies) operated in electron ionization (EI) mode at 70 eV. The analyses were performed in selected ion monitoring (SIM) mode recording the molecular ion of each compound (Table 1). Transfer line temperature and ion source temperature were maintained at 325 ◦ C and 300 ◦ C, respectively. The mass spectrometer was equipped with a 6 mm ultra large aperture draw out lens (Agilent Technologies). Injection was carried out using an automated GC PAL injection system (CTC Analytics, Zwingen, Switzerland). PTV-PSL injection was performed to set reference values on which basis the performance characteristics of the PTV-SV injection method were evaluated. The parameters of the PTV-PSL injection method were: injection volume 1 ␮L, injection pulse pressure 30 psi 646 J.Á. Gómez-Ruiz et al. / Talanta 80 (2009) 643–650 for 0.3 min, and injection port temperature 300 ◦ C. The purge valve was opened after 1 min (purge flow 30.5 mL min−1 ). The parameters of the PTV-SV injection were subject of the optimisation process and are discussed later in this paper. All analyses were carried out on a DB-17MS column 60 m × 0.25 mm internal diameter, 0.25 ␮m film thickness (Agilent Technologies, Diegem, Belgium). The oven temperature programme was 80 ◦ C (hold 1 min), to 250 ◦ C at 40 ◦ C min−1 , to 305 ◦ C at 25 ◦ C min−1 , to 315 ◦ C at 2 ◦ C min−1 , to 330 ◦ C at 40 ◦ C min−1 and hold for 35 min. Helium was used as carrier gas at 1.5 mL min−1 flow rate. 2.4. Statistical software Dedicated software for the design of experiments and optimisation, MODDE version 8.0 (Umetrics, Umeå, Sweden), was used for building up the parameter matrix and further evaluation of the results. 3. Results and discussion 3.1. Optimisation of PTV-SV injection All standards and sample extracts were prepared in toluene. Although toluene is not the preferred solvent for large volume injection due to its rather high boiling point, it is the most suitable solvent to dissolve PAHs with high molecular mass (i.e. dibenzopyrenes). Its application was also recommended to prevent the adsorption of micro-contaminants to glassware [24]. Large volume injection into a PTV injector can be done in different modes: at-once (i), or using multiple injection (ii) [25]. In mode (i), the sample is introduced at relatively high speed, whereas in mode (ii) the sample is introduced at a rate that is theoretically equal to that of evaporation. At-once injection was applied for this study. Different speeds were tested and the best results were obtained by injecting the sample at a speed of 5 ␮L s−1 . 3.1.1. Selection of the inlet liner The most appropriate PTV inlet liner has to be identified prior to the optimisation of operating parameters, as this device has a crucial role in the trapping and transfer of the analytes. The optimisation was carried out using a step-wise approach. Four different PTV inlet liners were tested in this study: the single-baffle, the single-baffle packed with glass wool, the multi-baffle, and the sintered glass inlet liner. Another type of liner, which is filled with Tenax was not considered in this study since it has been reported that their use can provoke problems during the transfer of high boiling compounds to the column, leading to peak distortion and losses among other negative effects [9]. Standard solutions containing 100 ng mL−1 of each target analyte were injected ten times into each liner applying PTV-SV injection, and repeatability of injection was evaluated. An injection volume of only 1 ␮L was selected to avoid chromatographic problems stemming from the, at that point of time, non-optimised PTV-SV injection. The sintered glass inlet liner showed the worst performance in terms of mass transfer and was therefore excluded from further studies. Fig. 1 shows the respective chromatograms for the baffle inlet liner (A) and the sintered glass inlet liner (B). As can be seen strong discrimination in terms of peak height occurred especially for the compounds with the highest molecular weight, the dibenzopyrenes. Table 2 shows the relative standard deviations (RSDs) obtained under repeatability conditions for the other three inlet liners. The Fig. 1. Chromatograms of the 15 + 1 EU-priority PAHs on a DB-17MS column 60 m × 0.25 mm internal diameter, 0.25 ␮m film thickness applying a PTV baffle liner (A) and a PTV sintered glass liner (B). multi-baffle inlet liner provided the best results in terms of precision for all compounds, although a similar level of precision was obtained with the single-baffle inlet liner for the compounds with low and medium molecular mass (from BcL to BaP). However the performance of the multi-baffle liner was superior for analytes with high molecular mass. The results obtained with the single-baffle inlet liner packed with glass wool were generally worse than those obtained with the other two baffle inserts. Noticeably high RSDs were found for BcL and for high molecular weight compound like DiP and DhP. This might be caused by losses of BcL during the solvent evaporation phase, and irreversible adsorption of the heavy PAHs on active sites of the glass wool. When increasing the injection volume from 1 ␮L up to 5 ␮L, peak symmetry was maintained only with the multi-baffle inlet liner. Therefore, the multi-baffle inlet liner was applied for further optimisation of the PTV-SV injection. 3.1.2. Injection volume The maximum tolerable injection volume of toluene was determined for the multi-baffle inlet liner by injecting different volumes of toluene into the PTV injector without connecting the column. The carrier gas flow was turned on in that experiment. According to Godula et al. [26] the injection volume can be increased until a solvent drop is observed at the exit of the inlet liner. A volume of 10 ␮L of toluene was injected into the multi-baffle inlet liner without visible overflow of the liquid and hence further-on used in this study. 3.1.3. Initial inlet temperature The initial injector temperature plays an important role in large volume injection in solvent vent mode. It has to be set to allow fast removal of the solvent without losses of the analytes. The temperature of the injection port was set in the described experiments to 55 ◦ C because higher temperatures caused losses of the most volatile target analyte, benzo[c]fluorene. This temperature reflects also the minimum temperature that can be reached in a J.Á. Gómez-Ruiz et al. / Talanta 80 (2009) 643–650 647 Table 2 Repeatability expressed as RSD % obtained for the injection of 1 ␮L of a standard solution containing the 15 + 1 EU-priority PAHs in toluene in a DB-17MS column 60 m × 0.25 mm internal diameter, 0.25 ␮m film thickness (100 ng/mL, n = 10). The RSD (%) of three different liners is shown. Benzo[c]fluorene Benz[a]anthracene Cyclopenta[c,d]pyrene Chrysene 5-Methylchrysene Benzo[b]fluoranthene Benzo[k]fluoranthene Benzo[j]fluoranthene Benzo[a]pyrene Indeno[1,2,3-cd]pyrene Dibenz[a,h]anthracene Benzo[ghi]perylene Dibenzo[a,l]pyrene Dibenzo[a,e]pyrene Dibenzo[a,i]pyrene Dibenzo[a,h]pyrene Single-baffle liner Single-baffle packed with glass wool liner Multi-baffle liner RSD (%) 1.32 1.73 1.37 2.38 1.54 1.10 1.43 1.63 1.27 3.01 5.05 2.98 3.95 6.01 6.61 6.20 37.84 2.43 5.07 8.45 1.62 4.57 1.58 5.01 0.88 3.09 9.38 3.98 2.76 2.40 24.19 26.18 3.13 1.81 1.51 2.10 1.46 0.61 0.85 1.13 0.39 0.90 0.99 0.77 1.65 1.19 1.96 2.33 reasonable period of time without applying cryogenic inlet cooling. Compressed air was used as replacement of cryogenic gases for facilitating the cooling process of the PTV injector. Consequently less than 5 min were necessary after each run to reach the initial inlet temperature. 3.1.4. Optimisation of PTV operating parameters The PTV-SV injection can be divided into four different phases: the injection, the solvent vaporization, the transfer of analytes and the cleaning phase. During the injection the split valve is open (solvent vent mode) and the sample is introduced into the injection port set at a temperature below the boiling point of the carrier solvent. The following step, the solvent vaporization, is critical with regard to the analyte transfer onto the column and its effect on the quality of separation. The aim of the solvent vaporization is to eliminate the excess of solvent without loosing the analytes. If the residual solvent volume is too high a long flooded zone is formed in the front part of the column leading to peak distortion (a phenomenon known as “peak band broadening in space”) [27]. Too excessive elimination of solvent will lead to losses of the most volatile compounds. Hence a small quantity of remaining solvent that is later-on transferred onto the column is desirable, not only to prevent analyte losses, but also to form a film at the front part of the column that traps and thereby focuses early eluting compounds [26]. The solvent elimination is influenced by different parameters that were considered in the optimisation process. They were in particular vent flow, vent time, vent pressure, and initial inlet temperature. The initial inlet temperature was set to 55 ◦ C, as detailed before. Hence the focus was put onto the other three variables. Previous reported studies [28] and own investigations revealed that a high final temperature of the injection port enhanced the transfer of the analytes with high molecular mass onto the column, while the slope of the temperature increase in the PTV injection port did not have any significant influence. Accordingly, the final temperature of the injection port was maintained throughout the study at 400 ◦ C and the temperature ramp at the maximum value of 600 ◦ C min−1 . Regardless the conditions selected during optimisation, the split valve was opened after 3 min (purge time) and kept in this position for 15 min at 400 ◦ C (cleaning step). Vent time, vent flow and vent pressure were the variables that had to be optimised in order to maximise the analytical response (peak area) of the different PAHs. The type of experimental design was selected taking into consideration the number of experiments to be performed and hence the time needed for measurements and the information level gained, which is expressed by the efficiency of the design. A full factorial experimental design including the three variables at two levels would consist of eight analyses, but would only allow the set up of linear models. Any non-linear relationships cannot be identified at two levels. A full factorial experimental design at three levels would consist without replication of 27 (33 ), and at four levels even of 64 (43 ) experiments. The advantage of full factorial designs is that all main effects and all multi-factor interactions can be described. However, if higher order factor interactions are not considered important fractional factorial experimental designs such as the Doptimal design may be applied. This is a computer generated design which consists of N runs out of all runs of the full factorial design, select in a way that maximum information can be gained with a minimum number of analyses [29]. The D-optimal design applied in the current work comprised 21 randomized chromatographic runs (10 corner experiments to which 11 axial points were added) out of 64 possible runs that span maximally the experimental region. Table 3 shows the parameter settings employed for each variable in the D-optimal experimental design. The correlation matrix obtained from the experimental data indicated an extremely high correlation between the responses obtained for all PAHs, which justifies the application of one single model for all the 15 + 1 EU-priority PAHs. The data was fitted by means of partial least square regression (PLS-R) providing an independent multivariate regression curve (model) for each PAH. No significant lack-of-fit was identified for any of the responses. Among the selected factors, vent time showed the highest influence on the analytical responses (negatively correlated with all the 15 + 1 EU-priority PAHs) followed by vent pressure and vent flow. Contour surface plots were used to visualise the modelled region and to assist in finding the optimal experimental conditions. Fig. 2 shows the contour plots of the response variable for the most volatile compound within the set of target analytes (BcL), Table 3 Experimental conditions employed for the optimisation of the programmed temperature vaporization (PTV) injection in solvent vent mode by applying a D-optimal experimental design. Parameter Vent time (min) Vent flow (mL min−1 ) Vent pressure (bar) Level Minimum Intermediate Maximum 0.5 25 0.25 1 and 1.5 50 and 75 0.5 and 1 2 100 1.5 J.Á. Gómez-Ruiz et al. / Talanta 80 (2009) 643–650 648 Fig. 2. Contour surface plots of BcL, BaP and DhP for two independent variables (vent time and vent flow) at a constant value of the other independent variable (vent pressure = 0.5 bar). Values inside the plots represent analytical responses in arbitrary units. for BaP, for which maximum levels in food are specified in EU legislation, and for the last eluting PAH (DhP) for the two independent variables vent time and vent flow at a constant value of the vent pressure (0.5 bar). The contour plots show that small variations in vent time provoke significant changes in the analytical response of the PAHs (the response decreases as the vent time increases). A minimum value of the vent time (0.5 min) combined with vent flow values ranging between 60 mL min−1 and 100 mL min−1 maximised the analytical response of all compounds. Long vent times showed a negative effect on the response even though the vent flow was kept low. The optimisation of the response for the individual analytes is not the way to success if the target is performance at the best for the whole set of analytes, because parameter settings to maximise performance for the volatile target analytes might be improper for the six-ring PAHs, which show low volatility. Therefore, the optimiser and the prediction list (two features provided by MODDE software) were used to predict the best experimental conditions considering that the response variables for all 15 + 1 EU-priority PAHs have the same importance. This is done by optimising an overall desirability function that is combining the individual desirability for each response. However, a finding that is not reflected in the models was that broad and fronting chromatographic peaks were obtained when short vent times were combined with low vent flows. Similar responses were gained for vent flows in the range of 60 mL min−1 to 100 mL min−1 when combined with low vent time (0.5 min) (Fig. 2). Therefore it was decided to set for the optimisation of the model the vent flow to the highest value to ensure adequate solvent elimination. Hence the following parameter settings were found most appropriate for the PTV-SV injection of PAH solutions: vent time 0.5 min, vent flow 100 mL min−1 and vent pressure 0.5 bar. 3.2. Method validation The optimised PTV-SV method was validated to assess its applicability for the analysis of the 15 + 1 EU-priority PAHs in different samples. Different parameters such as linearity, precision (repeatability and intermediate precision), influence of matrix effects, and robustness of the PTV-SV injection method were evaluated. 3.2.1. Linearity The linearity of the optimised PTV-SV injection GC–MS method was evaluated in the range of 15 ng mL−1 to 80 ng mL−1 from calibration standards applying Mandel’s fitting test. All calibration functions were found linear. Table 4 presents performance characteristics of the method that were determined from replicate injections of six PAH standard solutions, respectively sausage extracts. The coefficients of determination of the calibration functions were for all 15 + 1 EU-priority PAHs higher than 0.99. 3.2.2. Precision and matrix effects The precision of the PTV-SV method was evaluated both under repeatability and intermediate precision conditions. Standard solutions in solvent (toluene) and sausage extracts containing the 15 + 1 Table 4 Performance characteristics of the optimised PTV injection in solvent vent mode – GC–MS analysis method. Compound PTV injection in solvent vent mode Solvent solution Benzo[c]fluorine Benz[a]anthracene Cyclopenta[cd]pyrene Chrysene 5-Methylchrysene Benzo[b]fluoranthene Benzo[k]fluoranthene Benzo[j]fluoranthene Benzo[a]pyrene Indeno[1,2,3-cd]pyrene Dibenz[a,h]anthracene Benzo[ghi]perylene Dibenzo[a,l]pyrene Dibenzo[a,e]pyrene Dibenzo[a,i]pyrene Dibenzo[a,h]pyrene Meat extracts Calibration function R2 Repeatabilitya RSDr (%) Intermediate precisiona RSDi (%) Repeatabilityb RSDr (%) Intermediate precisionb RSDi (%) Recoveryc (%) 0.997 0.996 0.998 0.998 0.996 0.996 0.998 0.998 0.998 0.993 0.998 0.998 0.998 0.999 0.992 0.996 3.34 1.32 1.62 0.77 3.34 2.85 1.79 1.36 0.93 2.47 1.69 1.20 1.45 1.71 3.05 3.54 3.60 2.45 2.15 2.42 4.92 3.49 2.50 2.12 2.57 5.56 2.45 2.91 1.79 3.67 3.07 4.00 3.94 3.09 2.38 2.92 5.30 3.09 3.34 3.98 2.84 3.43 4.38 1.53 3.60 3.39 4.38 3.84 9.95 3.09 3.77 3.63 5.30 3.44 3.34 4.08 4.01 3.86 4.53 4.74 4.30 3.39 4.94 13.64 98.6 99.1 102.5 98.1 98.8 99.3 97.5 98.3 97.5 97.5 98.8 97.7 100.0 98.3 95.2 104.6 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 2.3 2.2 2.5 1.9 3.6 3.7 2.6 2.6 2.0 2.5 1.9 1.9 2.4 2.0 2.1 3.4 R2 : coefficient of determination from the analysis of 6 different concentrations in the range 15–80 ng mL−1 . a 50 ng mL−1 standard mix of the 15 + 1 EU-priority PAHs. b Sausage meat extract (3.9 ␮g kg−1 to 9.9 ␮g kg−1 for the 15 + 1 EU-priority PAHs). c Average recovery (%) plus standard deviation calculated from triplicate analysis of blank lamb meat extracts (n = 3) fortified with the 15 + 1 EU-priority PAHs (50 ng mL−1 ) prior to their analysis by GC–MS. J.Á. Gómez-Ruiz et al. / Talanta 80 (2009) 643–650 649 Fig. 3. SIM chromatograms of BcL (m/z 216), BaP (m/z 252) and DhP (m/z 302) obtained with PTV-SV and PTV-PSL injection of an extract of sausage fortified with the 15 + 1 EU-priority PAHs (concentration ranging from 3.9 ␮g kg−1 to 9.9 ␮g kg−1 ). Concentration of the presented compounds were 3.9 ␮g kg−1 , 5.3 ␮g kg−1 and 9.9 ␮g kg−1 for BcL, BaP and DhP, respectively. EU-priority PAHs were analysed. Repeatability was calculated from analysis results of three consecutive injections on the same day of either one PAH standard solution in toluene (50 ng mL−1 of each analyte), or one sausage extract sample (3.9 ␮g kg−1 to 9.9 ␮g kg−1 of the different PAHs), while intermediate precision was evaluated performing one-way ANOVA of three consecutive injections of both samples on three different days. Hence the calculated precision estimates reflect only the contribution of the GC–MS measurements on the overall precision of the analysis method. The relative repeatability standard deviations (RSDr ) of the PTVSV injection GC–MS measurements were satisfactory for the PAH solution in toluene, ranging between 0.77% for CHR and 3.54% for DhP. The respective relative intermediate precision standard deviation (RSDi ) ranged between 1.79% for DlP and 5.56% for IcP (Table 4). The influence of matrix effects on the GC–MS measurement can be deduced from the precision estimates of the sausage sample. For most analytes the intermediate precision was of the same order of magnitude as the repeatability, which indicates that the gradual accumulation of co-extractives in the injection liner did not negatively impact precision. They were significantly different only for BcL and DhP (Table 4). The precision estimates for the sausage samples were somewhat higher than the respective estimates for the solvent solution, which was to be expected. However the vaporization of the analytes from food extracts was not significantly different to that from solvent solutions. This was confirmed by the replicate injection of blank lamb meat extract samples that were spiked with the target analytes to a level of 50 ng mL−1 just before injection and comparison of the results to those for solvent solutions. The ratio of the results for the meat sample and of the solvent solution would indicate matrix effects, if it was significantly different from one. The last column of Table 4 contains this information, expressed as percent recovery. Most of the results were close to 100%, which reveals the absence of matrix interferences in the analysis of meat samples applying this particular combination of sample preparation and PTV-SV-injection-GC–MS analysis. 3.2.3. Robustness The robustness of the optimised PTV-SV method was evaluated using a screening fractional factorial design. Robustness measures the capacity of the method to remain unaffected by small but deliberate variations in the method parameters [30]. Assuming that the precision of the temperature control and of the control of pneumatics of the PTV injector is better than 5%, this value was selected for testing of the robustness of the PTV-SV injection. Therefore, values between 0.475 min and 0.525 min were selected for vent time, between 95 mL min−1 and 105 mL min−1 for vent flow and between 0.475 bar and 0.525 bar for vent pressure. A t-test was used to estimate the significance of each effect relative to the random variability (expressed as the observed standard deviation obtained under repeatability conditions at the centre of the experimental region). The PTV-SV method was considered robust since none of the three investigated factors had a statistically significant effect at the 95% confidence level on the analytical response of the different PAHs. 3.3. PTV in solvent vent mode and PTV splitless injection. Sensitivity of the optimised method The performance of the optimised PTV-SV injection method was compared to the PTV-PSL injection (injection volume 1 ␮L) in terms of the signal-to-noise (S/N) ratios of two different samples: a standard solution containing the 15 + 1 EU-priority PAHs (15 ng mL−1 ) and a sausage sample fortified with the 15 + 1 EU-priority PAHs (concentration ranging from 3.9 ␮g kg−1 to 9.9 ␮g kg−1 ). A PTVPSL injection method was already published by Bordajandi et al. [21]. The aim of the current experiments was to identify potential improvements of sensitivity of the PTV-SV injection method (10 ␮L) compared to the pulsed splitless injection method (1 ␮L). Fig. 3 shows selected SIM chromatograms of both PTV-SV injection and PTV-PSL injection of the meat extract for the ions with m/z 216, 252 and 302 corresponding to the base peak ions of BcL, BaP and DhP, respectively. The S/N ratios were for all analytes much higher in the chromatograms recorded after injection of the meat extract with the optimised PTV-SV method. Significantly increased was the sensitivity obtained for DhP, with an almost seven-fold higher S/N ratio after PTV-SV injection compared to PTV-PSL injection. The gain of sensitivity with the PTV-SV injection was noticeable also for the other six-ring PAHs (Table 5). An increase in S/N ratios J.Á. Gómez-Ruiz et al. / Talanta 80 (2009) 643–650 650 Table 5 Signal-to-noise (S/N) ratio of GC–MS analysis applying the optimised PTV injection in solvent vent mode (PTV-SV) and PTV injection in pulsed splitless mode (PTV-PSL) for the analysis of both a solution of the 15 + 1 EU-priority PAHs in solvent (15 ng mL−1 in toluene) and an extract of a sausage samples fortified with the15 + 1 EU-priority PAHs (content ranging from 3.9 ␮g kg−1 to 9.9 ␮g kg−1 ). Compound S/N ratio S/N ratios for the sausage extract normalised to BaP Solvent solution Benzo[c]fluorene Benz[a]anthracene Cyclopenta[c,d]pyrene Chrysene 5-Methylchrysene Benzo[b]fluoranthene Benzo[k]fluoranthene Benzo[j]fluoranthene Benzo[a]pyrene Indeno[1,2,3-cd]pyrene Dibenz[a,h]anthracene Benzo[ghi]perylene Dibenzo[a,l]pyrene Dibenzo[a,e]pyrene Dibenzo[a,i]pyrene Dibenzo[a,h]pyrene Sausage extract PTV-SV PTV-PSL PTV-SV PTV-PSL 538 7938 5918 2902 975 5072 541 3489 3551 926 4597 585 779 811 483 442 246 241 176 299 70 265 191 205 175 136 118 127 43 46 37 33 2484 1439 1111 2017 1231 2013 1722 2779 1420 1155 1794 1202 608 643 380 725 785 590 426 578 221 327 378 550 540 330 368 221 119 105 62 108 ranging between 5.1 and 6.7 times was observed. This is particularly important with regard to the typical discrimination of signal intensities that is observed in the analysis of PAHs by GC–MS [6,7]. However the effect of the injection technique on the discrimination of the signal intensities was visualised by determining relative signal-to-nose ratios in order to demonstrate that the enhancement of signal-to-noise ratios in PTV injection in solvent vent mode was not only the consequence of the higher amount of analyte introduced into the instrument. For that purpose the S/N ratio of the analytes were normalised to the S/N ratio of BaP (Table 5). For the last eluting PAHs (DiP and DhP) a 2.5-fold increase of the normalised S/N ratios was observed when applying PTV-SV injection compared to PTV-PSL injection. Likewise, the normalised S/N values of the other two dibenzopyrenes (DlP, DeP) as well as the rest of PAHs (except for BaA and CPP that showed similar values) were improved when the PTV-SV injection was used. Therefore, the PTV-SV injection does not only improve the sensitivity of the 15 + 1 EU-priority PAHs, it is moreover also able to reduce the discrimination of signal intensities between the low- and the high-molecular weight compounds. 4. Conclusions A PTV method in solvent mode (PTV-SV) for the analysis of the 15 + 1 EU-priority PAHs has been optimised and validated. The use of statistical experimental design simplified the optimisation of the most important parameters governing the PTV injection. As compared to PTV injection in splitless mode, the optimised PTV-SV method provided higher sensitivity for all studied analytes, irrespective of the type of injected sample being it a solvent solution of the target analytes or an extract of a fortified sausage sample. Especially significant was the improvement in the S/N ratio of the compounds with the highest molecular weight (more than six-fold increase for DeP, DiP and DhP). The optimised PTV-SV injection showed very good performance in terms of linearity, precision, and robustness. Matrix effects were not experienced in the analysis of meat extract samples. PTV-SV PTV-PSL 1.75 1.01 0.78 1.42 0.87 1.42 1.21 1.96 1.00 0.81 1.26 0.85 0.43 0.45 0.27 0.51 1.45 1.09 0.79 1.07 0.41 0.61 0.70 1.02 1.00 0.61 0.68 0.41 0.22 0.19 0.11 0.20 References [1] K. Ravindra, R. Sokhi, R. Van Grieken, Atmos. Environ. 42 (2008) 2895. [2] International Agency for Research on Cancer (IARC), IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, vol. 32, IARC, Lyon, France. [3] European Commission, Opinion of the Scientific Committee on Food, 2002. http://europa.eu.int/comm/food/fs/sc/scf/out153 en.pdf. [4] Joint FAO/WHO Expert Committee on Food Additives (JECFA), 2006. http:// www.who.int/ipcs/food/jecfa/summaries/summary report 64 final.pdf. [5] Environmental Protection Agency (EPA) of the United States of America (1999), Compendium Method TO-13A, EPA, Cincinnati, OH, USA. [6] E. Martínez, M. Gros, S. Lacorte, D. Barceló, J. Chromatogr. A 1047 (2004) 181. [7] A. Filipkowska, L. Lubecki, G. Kowalewska, Anal. Chim. Acta 547 (2005) 243. [8] F. Poy, S. Visani, F. Terrosi, J. Chromatogr. 217 (1981) 81. [9] E. Hoh, K. Mastovska, J. Chromatogr. A 1186 (2008) 2. [10] A. Vermeulen, K. Welvaert, J. Vercammen, J. Chromatogr. A 1071 (2005) 41. [11] D. Štajnbaher, L. Zupančič-Krajl, J. Chromatogr. A 1190 (2008) 316. [12] A. Esteve-Turrillas, E. Caupos, I. Llorca, A. Pastor, M. de la Guardia, J. Agric. Food Chem. 56 (2008) 1797. [13] F.M. Norlock, J.-K. Jang, Q. Zou, T.M. Schoonover, A. Li, J. Air Waste Manage. Assoc. 52 (2002) 19. [14] V. Yusà, G. Quintas, O. Pardo, A. Pastor, M. de la Guardia, Talanta 69 (2006) 807. [15] B.S. Crimmins, J.E. Baker, Atmos. Environ. 40 (2006) 6764. [16] V. Fernández-González, E. Concha-Graña, S. Muniategui-Lorenzo, P. LópezMahía, D. Prada-Rodríguez, Talanta 74 (2008) 1096. [17] J.L. Pérez Pavón, M. del Nogal Sánchez, M.E. Fernández Laespada, B. Moreneo Cordero, J. Chromatogr. A 1202 (2008) 196. [18] E. Ballesteros, A. García Sánchez, N. Ramos Martos, J. Chromatogr. A 1111 (2006) 89. [19] R. Rodil, M. Schellin, P. Popp, J. Chromatogr. A 1163 (2007) 288. [20] N. León, V. Yusá, O. Pardo, A. Pastor, Talanta 75 (2008) 824. [21] L.R. Bordajandi, M. Dabrio, F. Ulberth, H. Emons, J. Sep. Sci. 31 (2008) 1769. [22] Commission Regulation (EC) No 1881/2006 of 19 December 2006 setting maximum levels for certain contaminants in foodstuffs, Off. J. Eur. Union L 364 (2006) 5–24. [23] W. Jira, K. Ziegenhalsa, K. Speerb, Food Addit. Contam. 25 (2008) 704. [24] A. Covaci, S. Voorspoels, J. De Boer, Environ. Int. 29 (2003) 735. [25] A. Covaci, J. de Boer, J.J. Ryan, S. Voorspoels, P. Schepens, Anal. Chem. 74 (2002) 790. [26] M. Godula, J. Hajslova, K. Mastouska, J. Krivankova, J. Sep. Sci. 24 (2001) 355. [27] K. Grob, J. Chromatogr. 213 (1981) 3. [28] B. Veyrand, A. Brosseaud, L. Sarcher, V. Varlet, F. Monteau, P. Marchand, F. Andre, B. Le Bizec, J. Chromatogr. A 1149 (2007) 333. [29] L. Delgado-Moreno, A. Peña, M.D. Mingorance, J. Hazard. Mater. 162 (2009) 1121. [30] Citac/Eurochem, Guide to Quality in Analytical Chemistry—An Aid to Accreditation, Edition 2002, p. 31. http://www.eurachem.org/.