Rapid automatic identification and quantification of ...

11 downloads 42298 Views 2MB Size Report
Mar 26, 2013 - incomplete recovery. Nonane was added so the final .... be moved from “program and settings” to the C drive. 3. Results and discussion. 3.1.
Analytica Chimica Acta 778 (2013) 54–62

Contents lists available at SciVerse ScienceDirect

Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca

Rapid automatic identification and quantification of compounds in complex matrices using comprehensive two-dimensional gas chromatography coupled to high resolution time-of-flight mass spectrometry with a peak sentinel tool Yasuyuki Zushi ∗ , Shunji Hashimoto, Akihiro Fushimi, Yoshikatsu Takazawa, Kiyoshi Tanabe, Yasuyuki Shibata Center for Environmental Measurement and Analysis, National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba, Ibaraki 305-8506, Japan

h i g h l i g h t s

g r a p h i c a l

a b s t r a c t

• Analytical

method by GC × GC–HRTOFMS with automatic peak sentinel tool was developed. • PCDD/Fs and PCBs in crude lake sediment were automatically and accurately analyzed. • False positives/negatives were not observed except for when co-elution occurred. • GC × GC–HRTOFMS with the tool “TSEN” is useful for rapid and accurate screening.

a r t i c l e

i n f o

Article history: Received 18 October 2012 Received in revised form 9 March 2013 Accepted 17 March 2013 Available online 26 March 2013 Keywords: Two-dimensional peak sentinel Automatic screening GC × GC–HRTOFMS Comprehensive analysis Sediment analysis Organic micropollutants

a b s t r a c t Comprehensive two-dimensional gas chromatography coupled to mass spectrometry (GC × GC–MS) is a powerful tool for comprehensive analysis of organic pollutants. In this study, we developed a powerful analytical method using GC × GC for rapid and accurate identification and quantification of compounds in environmental samples with complex matrices. Specifically, we have developed an automatic peak sentinel tool, T-SEN, with free programming software, R. The tool, which consists of a simple algorithm for on peak finding and peak shape identification, allows rapid screening of target compounds, even for large data sets from GC × GC coupled to high resolution time of flight mass spectrometry (HRTOFMS). The software tool automatically assigns and quantifies compounds that are listed in user databases. T-SEN works on a typical 64 bit workstation, and the reference calculation speed is 10–20 min for approximately 170 compounds for peak finding (five ion count setting) and integration from 1–2 GB of sample data acquired by GC × GC–HRTOFMS. We analyzed and quantified 17 PCDD/F congeners and 24 PCB congeners in a crude lake sediment extract by both GC × GC coupled to quadrupole mass spectrometry (qMS) and GC × GC–HRTOFMS with T-SEN. While GC × GC–qMS with T-SEN resulted in false identification and inaccurate quantification, GC × GC–HRTOFMS with T-SEN provided correct identification and accurate quantification of compounds without sample pre-treatment. The differences between the

Abbreviations: GC×GC–MS, comprehensive two-dimensional gas chromatography coupled to mass spectrometry; T-SEN, two-dimensional peak sentinel tool; HRTOFMS, high resolution time of flight mass spectrometry; AMDIS, automated mass spectral deconvolution and identification system; AIQS-DB, automatic identification and quantification system with database; CRM, certified reference material. ∗ Corresponding author. Tel.: +81 29 850 2914; fax: +81 29 850 2575. E-mail address: [email protected] (Y. Zushi). 0003-2670/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.aca.2013.03.049

Y. Zushi et al. / Analytica Chimica Acta 778 (2013) 54–62

55

values measured by GC × GC–HRTOFMS with T-SEN and the certified values for the certified reference material ranged from 7.3 to 36.9% for compounds with concentrations above the limit of quantification. False positives/negatives were not observed, except for when co-elution occurred. The technique of GC × GC–HRTOFMS in combination with T-SEN provides rapid and accurate screening and represents a powerful new approach for comprehensive analysis. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Comprehensive two-dimensional gas chromatography coupled to mass spectrometry (GC × GC–MS) is one of the most important techniques for comprehensive analysis of organic micropollutants. Development of two-dimensional chromatographic techniques has been reviewed elsewhere [1]. The GC × GC technique can effectively separate chemicals using a combination of capillary columns of different polarities. However, coupling these techniques to MS generates a much larger volume of data compared to traditional GC–MS, hence the need for enhanced data processing/alignment techniques. Automatic peak detection algorithms for effective data analysis of two-dimensional (2D) chromatograms have been developed in previous research and two effective algorithms have been discussed in detail [2,3]. One of these is the watershed algorithm, which was originally used in image analysis for finding segmented regions where water collects. This algorithm was developed and applied to peak detection in 2D chromatograms by Reichenbach and implemented in GC Image software [4]. The other algorithm is a two-step procedure in which two step data processing are conducted for peak detection and integration. In the first step, general peak finding is conducted on a one-dimensional chromatogram. In the second step, the previously detected peaks, which have been split by modulation, are merged and their total intensity values are calculated [5]. Those algorithms have mainly been applied to total ion chromatograms (TIC) and extract ion chromatograms (EIC), with the aim of determining how effectively and accurately the peaks in a chromatogram can be isolated and their total signal intensity values calculated. Further developments are needed to implement these algorithms in automatic comprehensive analysis of the numerous compounds present in real environmental samples. The DotMap algorithm has also been developed for finding target compounds in mass spectra from 2D chromatography. Compounds are identified, based on similarities between the measured mass spectrum and that in a database of reference spectra. This program is useful for finding peaks in 2D chromatography with clean samples or samples containing the target compounds with high intensity signals. Parallel factor analysis has also been used to deconvolute overlapping peaks in a 2D chromatogram (GC × GC) for pure components [6]. An Automated Mass Spectral Deconvolution and Identification System (AMDIS) has been used for traditional gas chromatography coupled to quadrupole mass spectrometry (GC–qMS) for automatic peak assignment and quantification [7]. In the AMDIS procedure more than 100,000 spectra are searched for all TIC chromatogram peaks. The automatic identification and quantification system with database (AIQS-DB) is used in traditional GC–qMS to assign and quantify compounds in databases more accurately than AMDIS [8]. The AIQS-DB uses the selected ion monitoring (SIM) mode for sensitive quantification, in addition to scan mode and permits measurement of up to approximately 1000 compounds within one analysis run. Given that more than 10,000 peaks can be detected in an environmental sample, especially in GC × GC analysis, there has been great interest in developing automatic peak identification and quantification procedures for comprehensive analysis. However, several problems can occur with automatic peak identification and quantification in complex matrices, such as false positive/negatives and inaccurate quantification. These problems can arise with

co-elution of compounds and/or for MS measurements acquired at low-resolution. Recently, there have been some studies of GC × GC coupled to high resolution time-of-flight mass spectrometry (GC × GC–HRTOFMS) [9–12]. Full scan analysis by HRTOFMS provides accurate mass spectra of molecular/fragment ions by electron impact ionization at high frequency (25 Hz). The scan data collected at high frequency can be used to obtain sharp peaks in GC × GC by pulse modulation. If environmental samples that contain very complex matrices are analyzed by GC × GC–HRTOFMS, the measured mass spectral data size will be 15–20 GB in profile format. Even if the data are converted to centroid format, the data size is still approximately 2 GB. Although GC × GC–HRTOFMS provides accurate mass spectra for well separated compounds and could be used for qualitative and quantitative comprehensive analysis of real samples, further development is needed in data handling techniques. There has been one previous study which handled the GC × GC–HRTOFMS data. In the study, automatic peak assignment has been conducted by a tool called smart template which included in GC Image software [13] and peak-region approach. This approach allowed comparative analysis of 2D chromatograms for a series of samples such as breast-cancer tumors. The approach was able to classify the sample by grade of cancer with a 78% success rate. From the aspect of quantification in GC × GC–HRTOFMS, automatic tools for data processing are underdeveloped and their possibilities have yet to be properly evaluated and exploited. In this study, we have developed an analytical method based on GC × GC–HRTOFMS and a two-dimensional peak sentinel tool (T-SEN) for automatic, rapid and accurate identification and quantification of compounds listed in user databases. This tool may be used to extract and process data for trace-level components that are present in the mass spectra of compounds. The tool was written in the free programming software R and uses a twostep procedure for peak quantification with several modifications. Recently, R has begun to be used for data processing in comprehensive analysis. Smart et al. suggested the use of R (R package; R-XCMS) for modification of peak integration by AMDIS [14]. Peaks from LC-Orbitrap-MS (mass resolution: 60,000) can be detected and integrated by enviMass, which can be used by RExcel, for target screening and non-target analysis [15]. OrgMassSpecR is an R package that can be used for library searches or manual searches of non-target compounds, based on mass spectra obtained from GC × GC–LRTOFMS. [16]. We applied T-SEN, written in R, to GC × GC–HRTOFMS data from a lake sediment certified reference material (CRM) for peak finding and quantification of the concentrations of 17 congeners of polychlorinated dibenzo-p-dioxins (PCDD) and dibenzofurans (PCDFs) (PCDD/Fs) and 24 congeners of polychlorinated biphenyls (PCBs). Within the 24 PCB congeners, the concentrations for 12 PCB congeners were in agreement with the certified values. We also compared the results obtained for 17 PCDD/F and 12 PCB congeners with the certified values. For evaluation of the method, T-SEN was used to analyze data obtained from the same sample by GC × GC coupled to quadrupole mass spectrometry (qMS). No cleanup procedures were required for sediment samples before analysis by GC × GC–HRTOFMS in combination with T-SEN. The calculation was complete within a suitable time frame for routine analysis.

56

Y. Zushi et al. / Analytica Chimica Acta 778 (2013) 54–62

2. Materials and methods 2.1. Standards The following standards (Table S-1) were purchased from Wellington Laboratories (Guelph, Canada): NK-ST-A (contains seven PCDDs and 10 PCDFs); NK-LCS-A (contains carbon-13 (13 C) labeled PCDD/Fs that served as internal standards (ISs)); 68B-PAR (contain 27 PCBs); 68B-LCS (contains 13 C labeled PCBs that served as ISs); BDE-MXE (contains 27 polybrominated diphenyl ethers (PBDEs)); and MBDE-MXG (contains 13 13 C labeled PBDEs that served as ISs). The following standards (Table S-1) were purchased from CIL (Andover, MA); ES-5465 (contains 27 persistent organic pollutants (POPs)); and ES-5465-5X (contains 25 13 C labeled POPs that served as ISs). 2.2. Samples and sample preparation Air-dried lake sediment CRM from Lake Ontario (WMS-01, Wellington Laboratories), which is certified for 17 PCDD/F congeners and 12 PCB congeners that correspond to the species in the afore-mentioned standards for all of the PCDD/Fs and 12 PCBs (Section 2.1) was selected for study. Two grams (dry weight) of the lake sediment sample (n = 3) were placed in a solvent extraction cell (33 mL), with the sample being sandwiched between glass beads. The sample was then extracted using an Accelerated Solvent Extraction unit (Thermo Fisher Scientific, Waltham, MA). Hexane was used for the first extraction at the following settings: 100 ◦ C, 1.03 × 107 Pa (1500 psi), 50% flush thorough three cycles, and 60 s purge time. Next toluene was used under the same conditions, except that the temperature was increased to 150 ◦ C. The extracts were concentrated in a rotary evaporator (BUCHI Labortechnik AG, Flawil, Switzerland) and then combined. The combined extracts were then purged with N2 and mixed with the ISs (amounts of octachlorodibenzo-p-dioxin (OCDD) and octachlorodibenzofuran (OCDF) were 10 ng; all other ISs were 5 ng). The concentrations reported in this work were not corrected for incomplete recovery. Nonane was added so the final volume of each sample was 100 ␮L. Recoveries were checked by adding 20 ng of OCDD and OCDF and 10 ng of each of the other standards to the lake sediment samples (n = 4) and performing the extraction using the procedure, as described above. The average recoveries of the compounds were between 50 and 130% (Table S-2). 2.3. GC × GC–qMS and GC × GC–HRTOFMS The prepared samples were analyzed using a 6890GC (Agilent Technologies, Palo Alto, CA) with a KT2004 GC × GC system (Zoex, Houston, TX) coupled to a JMS-T100GC (JEOL, Tokyo, Japan) HRTOFMS. Also, a 7890GC (Agilent Technologies) with a GC × GC system coupled to a 7000A (Agilent Technologies) tandem mass spectrometer was used for GC × GC–qMS in single-scan mode. Details of this configuration are described elsewhere [11]. The first GC column (GC1) was an InertCap 5MS/Sil (60 m × 0.25 mm I.D., 0.1 ␮m film thickness, GL Sciences, Tokyo, Japan) and the second GC column (GC2) was an SLB-IL59 (1.5 m × 0.1 mm I.D., 0.08 ␮m film thickness, Sigma–Aldrich, St. Louis, MO). The SLB-IL59 is an ionic liquid column whose properties are similar to a polar column, although it has a slightly different polarity and lower column bleed even at 300 ◦ C [17]. The injection volume was 2 ␮L, in splitless mode at 410 kPa. The oven temperature was held at 70 ◦ C for 1 min, then increased to 180 ◦ C for 50 ◦ C min−1 , 230 ◦ C at 3 ◦ C min−1 , 300 ◦ C at 5 ◦ C min−1 and then held at 300 ◦ C until 50 min after the start time. The modulation period during the analysis was 4 s. The same settings for GC were employed for GC × GC–qMS and GC × GC–HRTOFMS. The HRTOFMS could acquire 25 spectra per

second with a mass resolution of 5000 (full width at half maximum). The ionization voltage for electron impact (EI) ionization was set at 45 V and the mass range was m/z 50–700 for HRTOFMS. The qMS could acquire 16 spectra per second at nominal mass. The ionization voltage for EI ionization was set at 70 V and the mass range was m/z 150–530 for the qMS. 2.4. User database preparation The standards (Section 2.1) were analyzed by GC × GC–qMS and GC × GC–HRTOFMS to establish a database. GC Image software ver. 2.2b2 (Zoex, Houston, TX) was used for database preparation and visualization. The user database contained information on the retention times (RTs) of the compounds on GC1 (RT1) and GC2 (RT2) and pairs of fragment ions (with accurate mass spectra) and the relative ratios of the fragment ions for the target compounds, in descending order of the ratio. The user needed to input those data to identify and quantify the compounds of interest and to prepare the database. Approximately 30 pairs of fragment ions and their relative ratios were recorded for each compound in this study. A demonstration version of the database is available as supplementary data. Because different ionization voltage settings were used for qMS (70 V) and HRTOFMS (45 V) measurements, we prepared individual user databases for qMS and HRTOFMS to avoid the possibility of inappropriate calculations through use of inappropriate ion fragment patterns and ratios. 2.5. T-SEN The T-SEN program was developed using the free and open source programming software R 2.14.2 (64 bit version). T-SEN consists of a simple algorithm for on-peak finding and peak shape identification and it permits rapid screening of target compounds even for large data sets from GC × GC–HRTOFMS. The program automatically detects peaks for compounds using information stored in a spectral database and, in addition, calculates the total intensity of each peak in a 2D chromatogram (or 1D chromatogram, optionally) [18]. Peak quantification with this program is similar to a previously reported two-step procedure [5], but with several modifications. One of the major modifications was to conduct the peak finding and quantification process on combined EICs, not on a TIC. The combined EIC is a combination of several ion chromatograms for a compound listed in the user database. The combined EIC for a compound in a specified domain, which is defined by the RTs of the compound and a search-range setting (described later), is newly calculated in each search process. The user can choose the number of ions for the peak finding. For example, if the user sets the ion number to five, the five ions with the highest relative intensity will be selected from the database and only the scan records that include all five ions will be searched in the defined domain. The tolerance range for an m/z error between the measured and reference data was set at ±0.05 for HRTOFMS analysis. The m/z error was set at ±0.02 for OCDD and OCDF, because several fragment ions were not able to be distinguished at the ±0.05 setting for 13 C labeled OCDF and native OCDD that elute at very similar times (e.g., measured m/z 459.757 for 13 C labeled OCDF (M + 8) and measured m/z 459.731 for OCDD (M + 4)). In the qMS analysis with T-SEN, nominal m/z values were used in the search. When an ion count setting is set too high, the detection limit deteriorates because low-intensity ions are included in the search. In this research, for calculation of the signal intensity values of the compounds, fragmentation patterns in the database and measurements on the five ions were compared. If the intensities of fragment ions were higher than that expected from the fragmentation patterns in the database even by 0.1, overlapping of

Y. Zushi et al. / Analytica Chimica Acta 778 (2013) 54–62

57

Fig. 1. Example of calculation of ri min and yi .

matrix-derived signals was thought to have occurred. The intensities of these overlapping signals were calculated and subtracted from the measured mass spectrum according to the calculation formula as follows: ri min = min

m 1 m 2 m 3 m 4 m 5 i i i i i

yi = ri min ˛1

˛1

,

˛2

,

˛3

,

˛4

,

˛5

(1) (2)

where mi is the measured intensity of the ion in scan record i, ˛ is the relative intensity of the ion listed in the database and yi is the calculated intensity of the compound in scan record i. For example, the mi 1 to mi 5 are the measured intensities of the 5 most intense peaks for a compound in the user database in scan record i. The ˛1 to ˛5 are the relative intensities of the respective ions and these are stored in the database. Eq. (1) finds an ion with the smallest margin between the expected and the measured intensity, with the calculation factor (ri min ), converting the relative intensity value to an artificial absolute intensity. The product of the conversion factor and the relative intensity of ion 1 become the artificial absolute intensity of the compound. Taking account of the margin of error for several ions by the equation can suppress the overlapped matrix-derived signals and result in avoiding unreasonable overestimations. Fig. 1 shows the calculation procedure. The number of ions can be changed by the user up to the number of ions stored in the user database (approximately 30 in this study). The search ranges of compounds in the 2D chromatogram were determined using the RTs of GC1 and GC2 in the user database. The width of the range was adjusted by the parameter setting to avoid peak finding errors caused by RT shifts. This search range definition is also advantageous for peak finding. The peak shape is interpreted, based on simple increases and decreases in the signal intensities in the chromatogram and a fitting technique was not used. Also, the settings of minimum peak width threshold and minimum peak height threshold were not used for peak detection and peak shape determination. A conceptual diagram of the peak sentinel is shown in Fig. 2. A further key modification of the software was the addition of a function for automatic RT correction of target compounds. This function is important for automatic peak finding in real samples with complex matrices. RT1 and RT2 in the database were adjusted for each homolog (e.g., PCDD/Fs, PCBs, POPs, PBDEs) in the first step of the peak sentinel using the IS. More specifically, a number of congeners of PCDD/Fs, PCBs, PBDEs exist in the environment and these have the same/similar fragmentation patterns and accurate ion masses. These congeners sometimes partially overlap

or co-elute even in GC × GC and accurate RT correction was required in this study. Therefore, we developed an algorithm to correct the RTs of individual compounds listed in the database using an IS that could be easily detected in the chromatograms without erroneous detection of their isomers. The RTs of each native PCDD/F or PCB were identified and corrected automatically based on the RT of its IS, which was identified before searching for the peaks of the native compound. A wider range was used for peak finding of each IS than was used for the corresponding native compound. This is because a known and limited number of ISs were spiked into the sample and, therefore, analysts can set a wide range for finding them without the possibility of false detection of their isomers. A wide range should also be used for peak finding of native compounds where isomer peaks are not close by. Owing to the function of retention time correction and peak finding in defined search range, in other words the function of peak sentinel, narrow setting for peak finding of native compounds should be allowed. T-SEN can be used with a 64-bit computer (CPU: Xeon 5160, 3.00 and 2.96 GHz, physical memory: 48 GB RAM, 4 GB × 12 with 667 MHz) for processing the large data sets of GC × GC–HRTOFMS. The reference calculation speed was 10–20 min for 170 compounds (ion count, 5), which includes the processing time for the sequential RT correction and processing of 1–2 GB of centroid data (HRTOFMS) in a network common data form, termed the netCDF file. The calculation time depended on the ion count, the search range in the peak sentinel process and the data size. A demonstration version of the database and the T-SEN source code are available as the R software package, “TSEN”. Sample data for demonstration purposes are downloadable, in accordance with the help note, “TSEN”. The software R is available free from http://www.r-project.org/. The version of R should be R-2.10.1 or later. The following script should be executed in R: install.packages(“TSEN”,repos=“http://cran.md.tsukuba.ac.jp/”) library(TSEN) data(database.dat) download.file(url=“http://www.nies.go.jp/analysis/downloads/ demo data stdPCDD DFs.cdf”,destfile = “demo data stdPCDD DFs.cdf”,mode=“wb”) cdf.dat 0.95 was considered acceptable for automatic rapid screening of chemicals. In this study, the R2 values were >0.99 for most compounds and all R2 values were >0.97 by automatic calculation. Details of calibration curves by T-SEN with GC × GC–HRTOFMS are provided in the supplementary material (Table S-3). Calibration curves for GC × GC–qMS with T-SEN were also prepared in the same way for evaluating analytical performance (see next section). 3.3. Quantification by GC × GC–MS with T-SEN Automatic quantification of PCDD/Fs and PCBs was conducted using GC × GC–MS with T-SEN, and results were compared with the certified values of the CRM (Fig. 5). The certified values were obtained in an interlaboratory study [20]. A clean sample, a magnetic sector-type high-resolution mass spectrometer (HRsectorMS) and manual peak checking with signal integration were used for quantification of the compounds in the CRM. The concentrations for the compounds determined by GC × GC–qMS with T-SEN were much higher than the certified values, especially for the PCDD/Fs, which had relatively low certified values. The bias toward high

concentrations was due to incorrect assignment of chromatographic peaks that were higher in intensity than the target ions measured at nominal mass-to-charge ratio. The concentrations of several PCBs such as PCB77, PCB118 and PCB105, obtained by GC × GC–qMS with T-SEN, were lower than the certified values. There were two possible causes for these results. First, there could be mass errors in the target ions because of the measurement of interfering ions with high intensities that were in close proximity to the target ions. These interfering ions could not be distinguished from target ions in the process of centroid conversion of mass spectra. In such cases, the mass spectra of the target ions might be shifted by the interfering ion(s), which would result in inaccuracies with peak identification. Second, the IS may be falsely assigned as a peak with high intensity and this would translate into a low concentration for the corresponding native compound. A number of false positives and overestimated concentrations were obtained by GC × GC–qMS and T-SEN for the crude sediment sample. With the exception of PCB123, GC × GC–HRTOFMS with T-SEN performed well and did not produce any false negatives or positives. For PCB123, co-elution with PCB107 and 109 was confirmed using a HT-8 PCB column. This column is better for PCB separation than the InertCap 5MS/Sil column (Fig. S-1). The

60

Y. Zushi et al. / Analytica Chimica Acta 778 (2013) 54–62

Fig. 4. Comparison between HRTOFMS and qMS with T-SEN. The m/z ranges were 50–700 for HRTOFMS and 150–530 for qMS. (a) TIC of standard solution (PCDD/Fs and PCBs) measured by HRTOFMS. (b) TIC of the lake sediment (CRM) measured by HRTOFMS. (c) T-SEN results with an ion count of one and 0.05 Da wide window for target compound searching from the lake sediment data of HRTOFMS. (d) T-SEN results with an ion count of five and 0.05 Da wide window for target compound searching from the lake sediment data of HRTOFMS. (a ) TIC of standard solution (PCDD/Fs and PCBs) measured by qMS. (b ) TIC of the lake sediment (CRM) measured by qMS. (c ) T-SEN results with an ion count of one and 0.05 Da wide window for target compound searching from the lake sediment data of qMS. (d ) T-SEN results with an ion count of five and 0.05 Da wide window for target compound searching from the lake sediment data of qMS.

quantification of partially separated peaks with overlapped fragment ions, such as in the case of PCB123 in this study, is problematic for automatic calculation. Although a few peaks co-eluted even in the GC × GC analysis of this study, the column combination can be tailored to the compounds of interest, such as PCDD/Fs, regulated substances, and substances of very high concern (SVHC) with the result that satisfactory separation can be achieved for automatic screening. The accuracy of quantification was low for compounds with concentrations below the limit of quantification (LOQ) (73.6–1262 pg g−1 for 17 PCDD/F congeners and 7.1–54.2 pg g−1 for 12 PCB congeners). For compounds with concentrations higher than the LOQ, the difference between the measured and CRM data

ranged from 7.3 to 36.9% (Table 1). These results were considered acceptable for automatic rapid quantitative screening. In this study, samples with complex matrices, such as the crude lake sediment extract used, can be accurately analyzed. Even with a enhanced S N−1 in GC × GC compared with 1D GC due to the cryofocusing of target compounds, a 10% reduction of mass spectra in a scan record was observed (1113 spectra in GC × GC, 1238 spectra in 1D GC). We applied T-SEN with slight modification to 1D GC analysis of the lake sediment data to evaluate the matrix separation effect in GC × GC. Although all of the spiked ISs of the PCBs (24 species) were detected correctly in GC × GC, only 19 species were detected in 1D GC. This was because a number of matrix components co-eluted and overlapped with the mass spectra of the PCBs.

Y. Zushi et al. / Analytica Chimica Acta 778 (2013) 54–62

61

Table 1 Comparison of measured value obtained by GC × GC–HRTOFMS with T-SEN and certified values of the lake sediment with limit of quantification.

2,3,7,8-TCDF 2,3,7,8-TCDD 1,2,3,7,8-PeCDF 2,3,4,7,8-PeCDF 1,2,3,7,8-PeCDD 1,2,3,4,7,8-HxCDF 1,2,3,6,7,8-HxCDF 2,3,4,6,7,8-HxCDF 1,2,3,4,7,8-HxCDD 1,2,3,6,7,8-HxCDD 1,2,3,7,8,9-HxCDD 1,2,3,7,8,9-HxCDF 1,2,3,4,6,7,8-HpCDF 1,2,3,4,6,7,8-HpCDD 1,2,3,4,7,8,9-HpCDF OCDD OCDF PCB81 PCB77 PCB123 PCB118 PCB114 PCB105 PCB126 PCB167 PCB156 PCB157 PCB169 PCB189

Difference (%)

CRM (WMS-01) (pg g−1 )

HRTOFMS and T-SEN (no cleanup) (pg g−1 ) Average (n = 3)

2 SD

Reference value

2 SD

27 2.4 NS NS 0.7 25 15 5.1 NS NS NS NS 210 220 NS 1760 430 36 1940 1500 8990 250 3360 75 400 530 120 NS 66

28 4.7 – – 2.6 44 27 10 – – – – 46 16 – 420 190 60 310 980 670 84 1770 33 77 91 8.2 – 6.4

52.5 17.7 12.6 18.5 7.96 67.3 20.3 16 8.66 20.8 17.3 2.68 299 293 15.1 1899 509 75 1717 209 8115 207 3998 84.9 330 715 186 7.97 85.2

16 5.6 5 6.1 2.8 24 8.7 8 2.7 4.8 8 4 73 63 4.6 456 157 79 520 191 1663 128 951 35 85 248 81 5.3 17.8

LOQ in HRTOFMS and T-SEN pg g−1

49.3 (