Target quantification and semi-target screening of halogenated carboxylic acids in drinking water using ultra-high performance liquid chromatography-quadrupole orbitrap high-resolution mass spectrometry

a b s t r a c t
To monitor the existing and emerging halogenated carboxylic acids (HCAs) in drinking water, a sensitive and rapid ultra-high performance liquid chromatography-quadrupole orbitrap high-resolution mass spec- trometry method for simultaneous target quantification of 10 haloacetic acids (HAAs) and semi-target screening of 19 HCAs was developed. After filtration, drinking water samples were injected into the in- strument. HCAs were separated on an HSS T3 column and detected by a type of non-target scan in the electrospray ionization negative mode. For target quantification of 10 HAAs, good linearity was obtained and the correlation coefficients were higher than 0.995. The limits of detection were in the range of 0.050–2.0 μg/L. The recoveries were in the range of 89.7%-108%, 83.4%-121%, 77.1%-116% and 80.2%-104% at levels of 2.5, 5.0, 10 and 20 μg/L, respectively, with relative standard deviations of 1.26%-16.9%. For semi-target screening of 19 HCAs, several criteria including accurate m/z, predicted retention time, de- duced fragment ions and simulated isotope pattern were used for identification. The method was applied to analyzing 41 drinking water samples successfully. Five HAAs were detected by target quantification, with dichloroacetic acid and trichloroacetic acid exceeding the limits suggested by the U.S. Environmental Protection Agency and the World Health Organization. Eight HCAs were preliminary identified by semi- target screening, and three of them were further confirmed with reference standards purchased later.

1.Introduction
Halogenated carboxylic acids (HCAs), a group of known or emerging disinfection by-products (DBPs) in drinking water, are generally formed by disinfection with chlorine, chloramines, chlo- rine dioxide, ozone, etc. [1] In recent years, HCAs especially haloacetic acids (HAAs) have become one hotspot for research on DBPs because of their potential genotoxicity and carcinogenicity. Monochloroacetic acid (MCAA), brominated acetic acids (BAAs) and iodinated acetic acids (IAAs) were reported to cause DNA dam-age in mammalian cells and induce mutation in bacteria, while dichloroacetic acid (DCAA) and trichloroacetic acid (TCAA) were shown weakly mutagenic in bacteria [2]. To date, DCAA and TCAA have been classified as the Group 2B (possibly carcinogenic to hu- mans) and the Group C (not classifiable as to its carcinogenicity to humans) by the Integrated Risk Information System, respectively [31,32]. For the purpose of ensuring safety of drinking water and reducing exposure of the population to DBPs, the U.S. Environ- mental Protection Agency (EPA) has set a maximum contaminant level (MCL) for the sum of five HAAs (HAA5), which include MCAA, DCAA, TCAA, monobromoacetic acid (MBAA) and dibromoacetic acid (DBAA). According to the stage 2 of Disinfectants and Disin- fection Byproducts Rules, the MCL of HAA5 is 0.060 mg/L. More- over, DCAA should not be detected and the concentration of TCAA cannot exceed 0.2 mg/L [3].

The World Health Organization (WHO)has also established guideline values for three HAAs, i.e. MCAA, DCAA and TCAA, and their MCLs have been set at 0.02, 0.05 and0.02 mg/L, respectively [4]. Although there are only five regulated HAAs, it is still necessary to comprehensively monitor other known and particularly emerging contaminants of HCAs due to their po- tential health hazards.A series of methods have been established for determination of HCAs, including gas chromatography (GC), ion chromatography (IC), reversed-phase liquid chromatography (RPLC) and these chro- matographic methods coupling to mass spectrometry (MS) and/or inductively coupled plasma (ICP). GC is the method earlier used to detect HCAs. For example, three EPA standard analytical meth- ods for determination of HAAs were GC with electron capture de- tection [5–7]. However, complicated derivatization was needed for GC methods to achieve good sensitivity. IC was usually used for analysis of HCAs since it could directly separate and detect them without derivatization [8,9]. To obtain high accuracy and sensitiv- ity, MS has been gradually applied to determination of HCAs. GC- MS methods have been developed by many researchers, but also need derivatization. IC-electrospray ionization-tandem MS (IC-ESI- MS/MS) is one of the EPA standard methods (NO.557) and nine HAAs in water samples can be detected directly without any sam- ple preparation. However, an organic solvent (e.g. acetonitrile) is needed to add post-column to assist desolvation in the ESI source, which means that an auxiliary pump is needed [10]. Because of its element-selective detection, ICP-MS can be applied to HCA detec- tion after chromatographic separation [11]. An IC-ICP-MS method was developed to detect BAAs, IAAs and some anions except chlo- rinated acetic acids (CAAs) because ICP-MS did not discern well the chlorine atom [12].

HCAs are hydrophilic and acidic, making it dif- ficult to retain and separate them on columns with traditional non- polar stationary phases for RPLC [13]. While using an HSS T3 col- umn bonded a trifunctional C18 alkyl phase, Luo et al. [14] estab- lished a reversed-phase ultra-high performance liquid chromatog- raphy (RP-UPLC)-MS/MS method for determination of 13 HAAs. The method obtained successful separation within 8 min, which was much faster than those IC methods in published literatures [12,15,16], and sub-μg/L sensitivity was obtained for most of the HAAs. It was shown that UPLC was an alternative for separation of HAAs.However, the MS methods described above performed well only in quantitation of limited numbers of HAAs with a target screening strategy by triple quadrupole (QqQ)-MS, which usually operated in multiple reaction monitoring (MRM) to acquire analyte-specific information with high sensitivity [17]. These low-resolution MS sometimes, however, could not obtain accurate qualitative results owing to sample interference. High-resolution MS (HRMS) could provide a possible solution. Orbitrap-MS is an HRMS technique which can offer high resolution of 140,000 (full-width at half- maximum) at m/z 200 and low mass error (<5 ppm). Besides, the combination of mass analyzers of an orbitrap and a quadrupole can realize data acquisition by a type of non-target scan, i.e. the full MS scan companied further with the data-dependent MS/MS scan, which can meet the requirements of target, semi-target and non- target screening. Although Gallidabino et al. [18] have developed a retrospective IC-HRMS method for screening HCAs in drinking wa- ter, only a single-stage orbitrap-MS was used, which meant that the confirmation for targets and suspects was lack of the informa- tion of their fragment ions. Furthermore, the sample preparation and IC separation in the method were somewhat time-consuming. In this study, we developed a rapid, sensitive and comprehen- sive UPLC-quadrupole orbitrap MS method without any sample preparation but filtration using both target and semi-target screen- ing strategies to monitor the existing and emerging HCAs in drink- ing water. This method was applied to target quantification of 10common HAAs and semi-target screening of some other HCAs.

2.Material and methods
MCAA (99.5%), MBAA (99.5%), DBAA (99.5%) and tribromoacetic acid (TBAA, 99.5%) were obtained from Dr. Ehrenstorfer (Augs- burg, Germany). Monoiodoacetic acid (MIAA, 97%) and diiodoacetic acid (DIAA, 98%) were purchased from Toronto Research Chemicals (Toronto, Canada). Bromodichloroacetic acid (BDCAA) and chlorodi- bromoacetic acid (CDBAA) stock solutions (1000 μg/mL in methyl tert–butyl ether) were both obtained from Fisher Scientific (Pitts- burgh, USA). TCAA and DCAA stock solutions (1000 μg/mL in methanol) were purchased from Agro-Environment Protection In- stitute of the Ministry of Agriculture (Tianjin, China). LC-MS grade acetic acid and acetonitrile were obtained from Thermo Fisher Scientific (Pittsburgh, USA). Ultrapure water (18.2 M▲•cm) was prepared with a Milli-Q Purification System (Millipore, MA, USA). Pierce ESI negative ion calibration solution was purchased from Thermo Fisher Scientific (Rockford, USA).Stock standard solutions (1000 μg/mL in ultrapure water) were prepared in-house for MCAA, MBAA, DBAA, TBAA, MIAA and DIAA. All of the ten stock solutions were stored at 4 °C in darkness. A series of calibration standard solutions of increasing concentration for each HAA were prepared freshly by ultrapure water before us- ing.Several types of drinking water, i.e. treated water from water- works, pipe network water, tap water and bottled water, were col- lected and stored at −80 °C.

After being filtered through a 0.22 μm polyethersulfone membrane filter, each water sample was directly injected into the instrument for analysis.The UPLC separation was performed on a Dionex UltiMate 3000 RSLC system and an Acquity UPLC HSS T3 column (2.1 mm×100 mm, 1.8 μm, Waters, USA). The injection volume was 10 μL. The autosampler and the column were kept at 6 and 10 °C, re- spectively. The mobile phase was composed of acetonitrile (A) and ultrapure water (B), both containing 0.001% (v/v) acetic acid. The flow rate was 0.42 mL/min. The gradient elution procedure was as follows: 0–10 min,5%−95% A, held by 95% A for 5 min, then equi- librated by 5% A for 5 min before the next injection.The Q-ExactiveTM Plus mass spectrometer was equipped with an ESI source (HESI-II, Thermo Fischer Scientific, Bremen, Ger- many) and the negative mode was operated in the experiment. The parameters of the HESI-II source were set as follows: sheath gas flow rate at 45 arbitrary units (a.u.), auxiliary gas flow rate at 15 a.u., sweep gas flow rate at 0 a.u., S-lens RF level at 55, spray volt- age at 3 kV, capillary temperature and auxiliary gas heater temper- ature both at 250 °C. A data-dependent (dd), non-target acquisition technique, i.e. full MS/dd-MS2, was applied for the data acquisi- tion. The inclusion list was set to “on”, ensuring priority of dd-MS2 acquisition for the concerned target and semi-target analytes over sample matrix. For the full MS acquisition, the resolution was set to 70,000 in the m/z range of 62–320 with the automatic gain con- trol (AGC) target at 3 × 106 and the maximum injection time (MIT) at 100 ms.

For the dd-MS2 scan, parameters were set as follows: resolution at 17,500, AGC target at 105, MIT at 50 ms, loop count at 10, topN at 10, isolation window at 1.4 m/z and stepped normalized collision energy (NCE) at 10, 30 and 60 for semi-target analytes.Table 1Experimental retention time, molecular formula and theoretical m/z of precursor and fragment ions and optimized NCE for target analytes.CompoundExperimental retention time of referenceThe customized NCE for target analytes was given in the inclusion list (Table 1). And other parameters for the dd-MS2 scan were ap- plied as follows: intensity threshold at 1 × 103, apex trigger at 2 to 4 s and dynamic exclusion at 12 s. The mass calibration was car- ried out on the instrument every three days using the Pierce ESI negative ion calibration solution. It should be noted that the cus- tomized calibration also needed to be done, following the conven- tional calmix calibration, for achieving higher mass accuracy. Con- sidering the low molecular weight for HCAs, three negative ions, whose m/z (59.01385, 265.14790 and 514.28440) were closest to the analytes, were chosen for performing the customized calibra- tion. The Xcalibur 4.0 software was used for instrument control and data acquisition, and the TraceFinder 3.3 software was used for semi-target screening and target quantitation analysis.It is impossible to obtain all the reference standards, which may result in missing some emerging or potential DPBs. To solve this problem, a semi-target screening method was established for iden- tification of HCAs without reference standards. After the data were acquired according to Sections 2.3 and 2.4, data analysis was per- formed on TraceFinder. First, a suspect list of 19 reported HCAs without reference standards was made based upon previous stud- ies [15,18]. Second, a compound database containing the infor- mation (formula, adduct species, fragment ions, polarity, retention time, etc.) of the 19 HCAs was built based upon theoretical or pre- dicted data.

Third, several parameters (m/z of precursor and frag- ment ions, retention time, isotope pattern) were set as detection criteria for identification or confirmation of suspects.The theoretical m/z of the precursor ion was calculated by its molecular formula and charge. The molecular formulas of frag- ment ions were predicted by the loss regulation observed from the ten HAAs with reference standards in our experiment and the homologues in other studies [15,16]. Consequently, the the- oretical m/z of the fragment ions could be calculated by their predicted molecular formulas. The retention time was predicted based upon linear solvation energy relationship (LSER) [19]. The isotope pattern of the precursor ion was simulated via TraceFinder.Each suspect was first identified by matching the retention time within the specified window (±2 min) and the precursor ion (peak area threshold>5000, signal-to-noise (S/N) ratio threshold>5, mass error<5 ppm). Then, it was further identified by matching the isotope pattern (fit threshold>60%, mass error<5 ppm, intensity deviation<20%) and fragment ions (at least one fragment present, intensity threshold>1000, mass error<5 ppm). If successfully iden- tified, the suspect could be further compared with the available reference standards to decide whether to be moved to confirmed status or not.The ten target analytes were identified and confirmed by the similar procedure as the semi-target screening described above. The differences were that some parameters (adduct species, frag- ment ions and retention time) were obtained from experimental rather than predicted data.

The confirmation criteria were as fol- lows: retention time within ±0.3 min, mass error of the precursor and fragment ions <5 ppm. The target quantification method for10 HAAs was validated by a series of performance characteristics, including linearity, linear range, limits of detection (LODs), lim- its of quantification (LOQs), precision and recovery. Linearity was evaluated by constructing a calibration curve at increasing con- centration, and the correlation coefficient was calculated by least- squares linear regression analysis between peak area and concen- tration. LOD and LOQ were usually calculated by measuring S/N ra- tios of 3 and 10, respectively. However, the baseline noises of some analytes were too low to be detected in the extracted ion chro- matogram obtained from HRMS, LOD and LOQ should be deter- mined by visual evaluation provided by the International Council for Harmonization (ICH) [20]. Intra- and inter-day precisions were determined by analyzing six replicate spiked tap water samples at concentration of 2.5, 5, 10 and 20 μg/L within the same day and over three separate days. Recoveries were calculated by detecting six replicate tap water samples per concentration spiked at four levels of 2.5, 5, 10 and 20 μg/L.

3.Results and discussion
All of the chromatographic conditions were optimized us- ing the standard mixture solution. Reverse-phase columns have been reported for separation of HAAs commonly such as Waters HSS T3, BEH C18 and Restek IBD columns [13,14,21,22]. In this study, Waters BEH C18 (2.1 mm×100 mm, 1.7 μm), CSH C18 (2.1 mm×100 mm, 1.7 μm) and HSS T3 columns were compared. As shown in Fig. 1, there is poor retention for HAAs on the BEH C18 column, and the obviously broad peak shape was exhibited on the CSH C18 column. Only on the HSS T3 column, both satisfactory peak shape and good resolution for all the ten HAAs were ob- tained. Compared with the conventional C18 columns, the HSS T3 column had a lower surface carbon density [14,21], which could promote the retention of polar compounds, and thus obtained the better resolution. Therefore, the HSS T3 column was finally se- lected. Based on previous reports [13,14], the mobile phase of ace- tonitrile/water containing formic acid or acetic acid was investi- gated in our experiment, and acetic acid was chosen as the addi- tive. Good separation and better sensitivity were achieved, because acetic acid had weaker inhibition effect on ionization during the electrospray process. The addition ratios of acetic acid at 0.0005%,0.001%, 0.002% and 0.003% were also investigated. Finally, 0.001% acetic acid was selected to obtain effective separation and good sensitivity (Fig. S1-S2, Table S1, Supplementary material).Column temperature was tested at 5, 10, 15 and 20 °C, as HAAs were reported to incline to degrade at high temperature [23].

The temperature of 5 and 10 °C showed almost the same sensitivity. However, lower temperature produced higher backpressure, and the backpressure at 5 °C approximated the upper limit (15,000 psi) of the instrument. Therefore, 10 °C was finally selected (Fig. S3, Ta- ble S2, Supplementary material).The injection volume was investigated in the range of 1–25 μL using a spiked sample (50 μg/L).Peak deformation was found when the injection volume was above 10 μL, which might be due to the overloading of the column by sample matrix. Therefore, a 10-μL in- jection volume of drinking water samples was finally set as a com- promise of high sensitivity and good peak shape.The parameters of the HESI source including spray voltage, sheath gas, auxiliary gas, sweep gas, S-lens RF level, auxiliary gas heater temperature and capillary temperature were tuned to im- prove sensitivity by directly infusing the standard mixture of the ten HAAs into the mass spectrometer. As a result, a compromise of these common parameters was reached to fit as many analytes as possible for stable and sensitive signals.Two types of ions, i.e. deprotonated molecule [M−H]− and de- carboxylated molecule [M−COOH]−, were mainly observed in the full MS scan. To obtain better sensitivity, the type with higher sig-nal intensity was selected for quantification and further higher- energy collisional dissociation.

The experiment results showed that heavily substituted HAAs (e.g. BDCAA, CDBAA) were inclined to form the [M−COOH]− ion, while most mono- and di-HAAs except DBAA were the [M−H]− ion (Table 1). Among all the isotopologions of the same [M−H]− or [M−COOH]− species, the most intense one was chosen as the quantitative ion. The typical examples were BAAs. For di- and tri-bromo-containing ions, M+2 was the most intense isotopolog ion rather other M (Fig. 2(a)). While in other cases, the monoisotopic ion M was the most intense one (Fig. 2(b)). In dd-MS2 scan, the NCE for target analytes was optimized in- dividually. The results were shown in Table 1. Deduced from the optimized NCE results of the target analytes, a stepped NCE was applied to obtaining the mixed fragment ions at different collision energies (10, 30 and 60), which might be suitable for the fragmen- tation of semi-target analytes. It was observed that mono-HAAs and HAAs with the intense [M−COOH]− ion preferred to disinte- grate into halogen ions, and di-HAAs with the intense [M−H]− ion were inclined to disintegrate into the [M−COOH]− ion. However, as a result of the mass range limit of the mass spectrometer in the study, ions with m/z below 50 could not be detected, such as the fragment ion (Cl−) of MCAA and TCAA. Experimental retention time, molecular formula and theoretical m/z of precursor and frag- ment ions and optimized NCE for target analytes were shown inTable 1.The validation of target quantification method was based on the data acquired from precursor ions in the full MS scan. Good linear- ity was obtained for each of the ten HAAs in the target analysis as shown in Table 2, with the correlation coefficient (r) higher than 0.995.

The LODs and LOQs were in the range of 0.050–2.0 μg/L and 0.17–6.7 μg/L, respectively. The recoveries were in the range of 89.7%−108%, 83.4%−121%, 77.1%−116% and 80.2%−104% at levelsof 2.5, 5.0, 10 and 20 μg/L, respectively. Intra- and inter-day RSDswere in the range of 1.26%−16.9% and 3.52%−12.7%, respectively. The proposed method was applied to the analysis of 41 drink-ing water samples, and 5 target analytes (MCAA, DCAA, DBAA,TCAA and BDCAA) were detected and quantified, with concentra- tions ranging from 0.356 to 24.9 μg/L. It should be pointed out that DCAA in 18 samples and TCAA in 3 samples exceeded the limits suggested by the EPA and the WHO, respectively, while the concentrations of BDCAA in 19 samples were in the range of 1.17–2.75 μg/L between the LOD and the LOQ. The chromatograms of a standard solution and a drinking water sample were shown in Fig. 3.The target analytes were identified and confirmed by comparing with the reference standards, according to the procedure and the criteria described in Section 2.6. The ion chromatograms of target analytes were extracted from the total ion current chromatograms, and the mass error of most analytes in samples were below 3 ppm. The experimental retention time differentials between reference standards and real samples were observed below 0.22 min. The re- tention time shift might be influenced by the matrix in samples. In spiked samples at low concentration (2.5 μg/L), isotope pattern scores were in the range of 60%−99% with 5 ppm of the allowed mass error and 20% of the allowed intensity deviation.

The scores for some analytes were lower than the default value 90% set by the software, due to the low concentration level and rich isotope fine structures of chloro– or bromo–containing ions. Therefore, the isotope pattern score was set at 60% as a criterion for semi-target screening.According to previous studies [18,21], the semi-target screening list was established, including not only the common HAAs but also some other HCAs (Table 3). Among these HCAs, although some of them, such as 2,2-DCPA (dalapon) and MCPA, are monitored and reported conventionally [10,18], their reference standards were un- available in this experiment, and thus they were taken as semi- target analytes. The addition of these HCAs could expand the ap- plication scope of the screening method.3.5.2.Identification and confirmation procedureAccording to previous studies [10,18] and our experimental re- sults of the target HAAs, two types of precursor ions, [M−H]− and [M−COOH]−, were both monitored for semi-target HCAs. The pre-cursor ions whose mass error was within 5 ppm were extracted from the total ion current chromatogram for further identification. At least one chromatographic peak was found for each semi-target analyte except seven analytes, i.e. MFAA, TFAA, BDFAA, FDCAA, FD- BAA, FCBAA and MBBA. For DCPA, MBPA and DBBA, even two peaks were observed, which might result from the existence of isomers of halopropionic acids and halobutyric acids.In order to improve the reliability of identification, it was nec- essary to use some additional parameters, such as predicted re- tention time, which could narrow down the scope of the candi- dates. Retention time could be predicted based on some models, such as linear-free energy relationships and hydrophobic subtrac- tion models [24]. Christine et al. [25] used retention time pre- dicted by a LSER model to exclude 60% candidates successfully.

However, to our knowledge, there have been no previous reports for the application of predicted retention time in the screening of HCAs, and in the best case, expected elution order was uti- lized only [18]. In our experiment, the LSER model was applied to predicting retention time of semi-target analytes. At first, 10 target analytes were selected to build up a training set and the lipophilicity parameter ϕ0 was calculated via the method proposed by Valkó [26]. Good linearity was obtained between the reten- tion time values in linear gradient elution and the determined ϕ0 values in isocratic elution for 10 target analytes (ϕ0 = 11.358tR − 13.781,r = 0.994, Eq. 1). The chromatographic hydrophobicity in- dex (CHI) values of them were calculated by means of the slope and intercept of Eq. 1. The CHI values obviously behaved signif- icant linear relationship with linear gradient retention time val- ues for 10 target analytes (CHI = 11.358tR − 13.781, r = 1.00, Eq. 2). Solute descriptors A, B0, S, E and V [27] of 10 target ana- lytes were taken from ACD/I-Lab (https://ilab.acdlabs.com/iLab2/), and a LSER model was established by the stepwise selection regression method (CHI = −60.58A − 22.66S − 24.95E + 162.40V − 24.55, R2 = 0.999, F = 1108.84, Eq. 3). It should be noted that there was no statistical significance for B0 in this model (P > 0.05). Therefore, CHI values of semi-target analytes could be calculated by Eq. 3. The predicted retention time of semi-target analytes could be obtained via Eq. 2. And the error window for the predicted re- tention time was set at ±2 min in order to balance the require- ment of reliability and error-tolerant rate.

The predicted retention times might be closed to each other for isomers because of the similar structures, which resulted in the overlap between different error windows, so it was difficult to set up a reliable corresponding relationship between the experimental and the predicted retention time of semi-target analytes for the isomers. However, compared to application of m/z alone for identification, m/z combined with predicted retention time could narrow down the scope of the can- didates greatly.Subsequently, the candidates were further identified by isotope pattern and fragment ions after accessing them with both m/z and predicted retention time successfully. Actually, isotope pat- tern was regarded as important identification information as frag- ment ions. The fit between the expected pattern and the measured one was represented as the isotope pattern score (as a percentage value), which was calculated by TraceFinder using an isotope pat- tern matching algorithm considering isotope mass error and rela- tive isotopic abundance. Finally, based upon these identification in- formation, identification confidence according to the common level system proposed by Schymanski et al. [28] was applied for the gradation of candidates to ease the communication of identifica- tion results, and there were five levels for identification. Among 19 semi-target analytes in the suspect screening list, 9 ones passed the criteria of m/z and predicted retention time, and thus would start gradation at Level 3 (tentative candidate) [29]. BCAA in 17 of 41 samples passed all the criteria, so it should obviously be moved to Level 2 (probable structure). Although there was absent of part of identification information for BIAA, DFAA, CDFAA, MBPA and BCAA in some samples, the five tentative candidates could still move to Level 2, because the lack of isotope pattern for BIAA in 3 samples and the absence of fragment ions for DFAA, CDFAA, MBPA and BCAA in some samples were due to low concentration in sam- ples. Five analytes (CIAA in all samples, BCAA, CDFAA, BIAA and DBBA in some samples) moved to Level 5 (exact mass) for the fail- ure of matching with both fragment ions and isotope pattern.

For the remaining detected semi-target analytes, i.e. MCPA and DCPA, there was evidence for possible structures but not sufficient for ex- act structures, and therefore they were finally identified at Level 3. Besides, the candidates (MCBA, DCBA in all samples and DBBA in 2 samples) passing m/z but failing in predicted retention time also moved to Level 5.To further confirm the existence of the semi-target analytes preliminarily identified at Level 2 and 3, several reference stan- dards (2-MCPA, 3-MCPA, 2, 2-DCPA, 2, 3-DCPA and BCAA) werepurchased later due to their easy availability. For the peaks taken as the suspect of MCPA, they were confirmed to be 2-MCPA by comparison of the retention time under the same analytical condi- tions in 14 drinking water samples, which meant the identification confidence of 2-MCPA could be reclassified to Level 1, although the absence of fragment ions of 2-MCPA in both samples and standardsolution might be due to the inclination of disintegrating into chlo- rine ions. There were three isomers of DCPA, but just two peaks were detected in the experiment. According to the elution order and retention time of reference standards, the peaks eluting firstly in 10 samples were 2, 3-DCPA, and the latter ones in 38 samples were 2, 2-DCPA. The positive confirmation of BCAA was obtained by matching the retention time. These results demonstrated the validity of the semi-target screening strategy, and eight compounds were preliminary identified, companying with the further confir- mation of the four (2-MCPA, 2, 2-DCPA, 2, 3-DCPA and BCAA) by comparing with the reference standards. The results of identifica- tion and confirmation of semi-target analytes are shown in Table 4.

4.Conclusion
A rapid and sensitive UPLC-Q-Orbitrap-HRMS method was de- veloped and applied to target quantification of 10 HAAs and semi- target screening of 19 HCAs. Compared with other studies us- ing solid phase extraction or dispersive solid phase extraction as the sample preparation method [22,30], LODs of the developed method were slightly higher than theirs, but could meet the need of drinking water monitoring. Sample preparation was sometimes used to enhance the detection sensitivity, but more preparation steps would bring more errors and longer analytical time. There- fore, the samples were injected after filtration in the study to save time and reduce reagent consumption. As a result, good recover- ies and sensitivity (μg/L) were realized for the target quantification method. While the semi-target screening method was applied to the identification of suspect compounds via matching with accu- rate m/z, predicted retention time, deduced fragment ions and sim- ulated isotope pattern, and the validity of the semi-target screening strategy were shown successfully by four reference standards purchased after preliminary identification. Screening of 41 drink- ing water samples in Sichuan province of P. R. China revealed the occurrence of 13 out of a list of 29 target and suspect HCAs, and 5 HCAs were detected by the validated quantification method in the range of 0.356–4.90 μg/L. In addition to the benefits mentioned above, full MS/dd-MS2 Sodium dichloroacetate data acquisition mode could provide retro- spective analysis of existing data by adding the emerging or inter- esting HCAs into the screening compound database.