Automated Dispersive Liquid−Liquid Microextraction−GasChromatography−Mass SpectrometryLiang Guo and Hian Kee Lee*

Department of Chemistry, National University of Singapore, 3 Science Drive 3, Singapore 117543, Singapore

*S Supporting Information

ABSTRACT: An innovative automated procedure, low-densitysolvent based/solvent demulsification dispersive liquid−liquidmicroextraction (automated DLLME) coupled to gas chroma-tography−mass spectrometry (GC/MS) analysis, has beendeveloped. The most significant innovation of the method isthe automation. The entire procedure, including the extractionof the model analytes (phthalate esters) by DLLME from theaqueous sample solution, breaking up of the emulsion afterextraction, collection of the extract, and analysis of the extractby GC/MS, was completely automated. The applications oflow-density solvent as extraction solvent and the solventdemulsification technique to break up the emulsion simplifiedthe procedure and facilitated its automation. Orthogonal arraydesign (OAD) as an efficient optimization strategy was employed to optimize the extraction parameters, with all the experimentsconducted auotmatically. An OA16 (41 × 212) matrix was initially employed for the identification of optimized extractionparameters (type and volume of extraction solvent, type and volume of dispersive solvent and demulsification solvent,demulsification time, and injection speed). Then, on the basis of the results, more levels (values) of five extraction parameterswere investigated by an OA16 (4

5) matrix and quantitatively assessed by the analysis of variance (ANOVA). Enrichment factors ofbetween 178- and 272-fold were obtained for the phthalate esters. The linearities were in the range of 0.1 and 50 μg/L and 0.2and 50 μg/L, depending on the analytes. Good limits of detection (in the range of 0.01 to 0.02 μg/L) and satisfactoryrepeatability (relative standard deviations of below 5.9%) were obtained. The proposed method demonstrates for the first timeintegrated sample preparation by DLLME and analysis by GC/MS that can be operated automatically across multipleexperiments.

In the past twenty or so years, several microextractionmethods, such as solid-phase microextraction (SPME)1−3

and liquid-phase microextraction (LPME),4−7 have beendeveloped for sample preparation. They are efficient,miniaturized, convenient, and environmentally benign com-pared to conventional liquid−liquid extraction (LLE) and solid-phase extraction (SPE) which are sometimes considered to bewasteful due to usage of a significant volume of potentially toxicorganic solvents as well as being time-consuming and labor-intensive.SPME, combining extraction and preconcentration in a

single step, is an effective and solvent-free technique. However,SPME fibers are relatively expensive, generally fragile, and havea limited lifetime, especially for some direct immersionextraction for complex sample matrices. Moreover, SPME hasbeen reported to suffer from sample carry-over problems.8

With the advantages of being economical, effective, andsolvent-minimized, LPME has gained considerable applicabilityto various compounds and has been developed in a variety ofconfigurations and modes, such as single drop microextraction(SDME),9 headspace LPME,10 dynamic LPME,11−13 hollowfiber protected LPME,14−17 continuous flow LPME,18 andsolvent bar microextraction,19−21 among others.

As a relatively new mode of LPME, dispersive liquid−liquidmicroextraction (DLLME) was first introduced in 2006 byRezaee et al.22 In this procedure, a mixture of extraction solventand dispersive solvent is rapidly injected into an aqueoussample solution to form an emulsion. Due to the extractionsolvent being highly dispersed in aqueous phase in finedroplets, extraction can be achieved very fast. This is the mostsignificant advantage of DLLME. Organic solvents with higherdensity than water are typically used in order to convenientlycollect the extract as the sedimentation phase aftercentrifugation. Since its introduction, DLLME has seen aconsiderable number of applications.23−27 However, the use ofa solvent with higher density than water limits the widerapplicability of DLLME and has been a main disadvantage ofthis method. To overcome this problem, many recent researchefforts have been focused on the use of organic extractionsolvents with lower density than water in DLLME.28−35 Morerecently, an alternative of DLLME, termed as solvent

Received: August 13, 2013Accepted: March 25, 2014Published: March 25, 2014

Article

pubs.acs.org/ac

© 2014 American Chemical Society 3743 dx.doi.org/10.1021/ac404088c | Anal. Chem. 2014, 86, 3743−3749

demulsification DLLME, has been developed.36,37 In thistechnique, after extraction, a demulsification solvent is usedto break up the emulsion, avoiding the use of centrifugation.However, to the best of our knowledge, none of these

developed DLLME procedures, including the original modesand variations, involves automation of sequential, multipleextractions integrated with analyses. Extending the DLLMEprocedure to automated operation would therefore be a naturalprogression.In the present work, for the first time, a novel automated

approach, low density solvent based/solvent demulsificationDLLME (automated DLLME) coupled to analysis by gaschromatography−mass spectrometry (GC/MS), was devel-oped. In this procedure, all extraction steps and analysis byGC/MS were continuously carried out repetitively, completely,and automatically with the use of a CTC CombiPalautosampler. Optimization of the extraction conditions wereperformed using orthogonal array design (OAD) and analysisof variance (ANOVA), which were also conducted automati-cally. The optimized conditions were applied to the analysis ofphthalate esters from genuine water samples as a demonstrationof the feasibility of the approach.

■ EXPERIMENTAL SECTION

Chemicals and Materials. The phthalate ester (PE)standards were purchased from Supelco (Bellefonte, PA,USA) as a reagent kit. They include benzylbutyl phthalate(BBP), di-(2-ethylhexyl) phthalate (DEHP), dibutyl phthalate(DBP), diethyl phthalate (DEP), dimethyl phthalate (DMP),and di-n-octyl phthalate (DNOP).High-performance liquid chromatography (HPLC)-grade

methanol, acetone, acetonitrile (ACN), and n-hexane werepurchased from Tedia Company (Fairfield, OH, USA).Toluene was from Fisher (Loughborough, UK) while 1-octanolwas bought from Merck (Darmstadt, Germany). The o-xylenewas obtained from Sigma-Aldrich (St. Louis, MO, USA).

Ultrapure water was produced on a Nanopure waterpurification system (Barnstead, Dubuque, IA, USA).A Hamilton 100 μL GC microsyringe (Reno, Nevada, USA)

affixed to an autosampler was used both for the extractionprocedure and for injecting the extract into the GC/MS system.The home-designed and modified glass vials with narrow necks,modified from conventional autosampler vials, were used forthe procedure (Figure 1, which also shows the vial dimensions).Vials were made in the glass-blowing workshop in theDepartment of Chemistry, National University of Singapore.The narrow neck facilitated the collection of the extract foranalysis after the automated DLLME (see below).37

Sample Preparation. A stock solution was prepared bydiluting the mixed PE standard reagent kit with methanol andwas stored in the refrigerator at 4 °C until use. Water samplesused for studying the extraction performance and optimizingthe extraction conditions were prepared daily by spikingultrapure water with the standard solution at different knownconcentrations.Genuine river water samples were collected from a local river

using precleaned glass bottles. The bottles were transported tothe laboratory immediately under cool conditions. Genuine tapwater samples were directly collected from our laboratory. Allcollected water samples were kept in the dark at 4 °C until use.The genuine water samples were filtered with a Millex-HN 13mm filter with 0.45 μm pore size (Millipore Corporation,Billerica) prior to the automated DLLME procedure.

Automated DLLME Procedure. The entire automaticprocess, including the DLLME procedure and injection of thefinal extract into the GC/MS system, was carried out on a CTCAnalytics (Zwingen, Switzerland) CombiPAL autosampler(coupled to a GC/MS system) with the aid of the CycleComposer software (CTC Analytics). The sample vial wasfilled with an optimized volume of 12.6 mL of sample solution,which was then placed in the autosampler tray for theextraction process (see Figure 1).

Figure 1. Schematic of automated DLLME. (a) Injection of extraction and dispersive solvent; (b) formation of emulsion; (c) injection ofdemulsification solvent to clear emulsion; solution level is raised to the narrow neck of vial; (d) collection of extract. At left: Design and dimensionsof a modified vial. For clarity, the schematic is not to scale.

Analytical Chemistry Article

dx.doi.org/10.1021/ac404088c | Anal. Chem. 2014, 86, 3743−37493744

Figure 1 shows the schematic of the automated DLLMEprocedure. Briefly, the optimized DLLME processes were asfollows, all conducted automatically: 80 μL of a solvent mixture(6:130, v/v) consisting of the extraction solvent (toluene) andthe dispersive solvent (acetonitrile) was withdrawn into themicrosyringe. The microsyringe was then moved to the samplevial, and the mixture was dispensed into the sample solution ata rate of 200 μL/s. This cycle was repeated 17 times (takingabout 340 s duration). The repetitive step was necessary toallow the transfer of the desired total volume (1.36 mL) of thesolvent (consisting of approximately 60 μL of toluene and 1.3mL of acetonitrile). An emulsion of the extraction solvent,dispersive solvent, and aqueous sample was formed in thesample vial. After a 2 min extraction, the microsyringe wasconveyed to a vial containing the demulsification solvent(acetonitrile) to withdraw 70 μL of solvent from it. Themicrosyringe was then moved to the sample vial again to injectall 70 μL of the demulsification solvent at a speed of 200 μL/sinto the solution to break up the emulsion. The demulsificationstep was repeated 20 times (about 400 s to yield 1.4 mL). Aftera 3 min demulsification time, the mixture cleared up andseparated into two phases. Coincidentally, under the optimizedextraction conditions, the upper organic layer (extract), toluene,moved to the narrow neck of the vial, which facilitated thecollection of the extract (∼35 μL) by the syringe for analysis.Lastly, 10 μL of the extract was collected by the syringe andinjected into the GC/MS system. The entire DLLME processwas then repeated automatically for processing the next andsubsequent samples sequentially.GC/MS Analysis. Analysis was performed on a Shimadzu

(Kyoto, Japan) QP2010 GC/MS system equipped with aprogrammable temperature vaporizer (PTV) injector and a DB-5 MS (J&W Scientific, Folsom, CA, USA) fused silica capillarycolumn (30 m length × 0.25 mm internal diameter, 0.25 μmfilm thickness). Helium (purity 99.9999%) was employed as thecarrier gas at a flow rate of 1.2 mL/min. Thirty seconds afterinjection, the vaporizer was heated from 90 to 250 °C at 250°C/min, with the temperature maintained for 1 min. The splitvent ratio was set at 50:1 for 30 s and changed to 10:1 forinjection. The interface temperature was maintained at 270 °C.The GC oven was initially held at a temperature of 90 °C for 3min and was then programmed to 180 °C at 25 °C/min,followed by another increase to 230 °C at 15 °C/min. Finally,the temperature was increased to 290 °C at 25 °C/min andheld for 4 min at the final value. The solvent cutoff time was 6.5min. PE standards and samples were analyzed in selective ionmonitoring (SIM) mode for quantitative determination of theanalytes. The monitored ions of the analytes were selected onthe basis of good selectivity and high sensitivity and were set asfollows: DMP, m/z 77, 163; DEP, m/z 149, 177; DBP, m/z149, 205; BBP, m/z 91, 149; DEHP, m/z 149, 167; and DNOP,m/z 149, 279.

■ RESULTS AND DISCUSSIONOrthogonal Array Design (OAD) Optimization. In this

work, OAD was employed to evaluate and optimize theautomated DLLME conditions. The enrichment factor (EF)was calculated and used as the response for all trials. On thebasis of the OAD experiments, the optimized extractionconditions were obtained.First, seven factors that affect automated DLLME including

the type of extraction solvent, the volume of extraction solvent,the type of dispersive solvent, the volume of dispersive solvent

and demulsification solvent, demulsification time, the injectionspeed, and six possible interactions, were assigned in a mixed-level OA16(4

1 × 212) matrix. Details of the assignment of factorsand their level settings are given in Table S-1, SupportingInformation, and results of the OAD experiments are given inTable S-2, Supporting Information.According to the results from the first mixed-level OA16 (4

1

× 212), toluene was used as the extraction solvent andacetonitrile was chosen as the dispersive and demulsificationsolvent. (Details of the analyses are provided as SupportingInformation.) Five other factors (the volume of the extractionsolvent, the volumes of the dispersive and demulsificationsolvents, the demulsification time, and the injection speed)were further determined using a four-level OA16 (45) matrix.Detailed assignment of factors and their levels of screening andresults according to the OA16 (45) matrix are provided asSupporting Information (Tables S-3 and S-4). Figure 2 shows

the OAD-calculated average responses (r) of the sum ofenrichment factors of the analytes for the four assigned levels(r1, r2, r3, and r4) of significant factors of the four-level OA16(45) matrix.

Volume of the Extraction Solvent. The volume ofextraction solvent plays an important role in DLLME efficiency.Different volumes of toluene (at levels of 60, 70, 80, and 90 μL)were considered. It can be seen from Figure 2 that responses(r1, r2, r3, and r4) (refer to the Supporting Information, Table S-4, r1, r2, r3, and r4 in column A) decreased when the volume oftoluene increased from 60 to 90 μL, especially with lowervolumes, conceivably due to the dilution effect. At lowervolumes of toluene, as expected, higher EFs were obtained. It isworth mentioning that a lower volume of extraction solvent(e.g., below 60 μL) might lead to difficulty in collecting theextract by microsyringe automatically after the demulsification.Additionally, the repeatability of results was reduced whendealing with a lower volume of extraction solvent. Inconsideration of a suitable balance of the extraction efficiencyand ease of collecting the extract, a volume of 60 μL of toluenewas applied for subsequent experiments.

Volume of the Dispersive and DemulsificationSolvents. A series of volumes (1200, 1300, 1400, and 1500

Figure 2. The effect of significant factors of the four-level OA16 (45)

matrix (refer to the Supporting Information, Table S-4) on the averageresponses (r). The arrows indicate the final chosen conditions.

Analytical Chemistry Article

dx.doi.org/10.1021/ac404088c | Anal. Chem. 2014, 86, 3743−37493745

μL) of acetonitrile was considered to investigate the effect ofthe volume of dispersive and demulsification solvent onextraction efficiency. The r values in Figure 2 (refer to theSupporting Information, Table S-4, column B) indicate that,when the volume of dispersive solvent increased from 1200 to1300 μL, the extraction efficiency was enhanced. This may beaccounted for by the fact that the emulsion resulting from thecombination of extraction solvent, dispersive solvent, and theaqueous sample solution could not be completely formed withan insufficient volume of dispersive solvent. A further increaseof dispersive solvent up to 1500 μL had no significantadditional influence on extraction efficiency. The latter alsoincreased considerably with the increase of demulsificationsolvent volume, up to 1400 μL; after that, the responsesflattened out (refer to Supporting Information, Table S-4, rvalues in column C). Therefore, in this work, the volumes ofdispersive solvent and demulsification solvent were selected tobe 1300 and 1400 μL, respectively.Demulsification Time. In DLLME, the interval between

the completion of injection of the demulsification solvent andthe start of the collection of the extract is defined as thedemulsification time. Different demulsification times (1, 3, 5,and 8 min) were investigated to study their effect on extraction.As can be seen from Figure 2 (refer to the SupportingInformation, Table S-4, r values in column D), the extractionefficiency increased significantly when the demulsification timewas increased from 1 to 3 min, after which, there was only aslight improvement. Considering the efficiency of the extractionprocess and in the interest of having a reasonable totalextraction time, the demulsification time adopted was 3 min.Injection Speed of the Mixture of Extraction Solvent

and Dispersive Solvent. In this present work, all extractionsteps were automatically carried out. Thus, it is worthwhile toinvestigate the injection speed in detail. The extractionefficiency in Figure 2 (refer to the Supporting Information,Table S-4, r values in column E) improved with the increase ofthe injection speed. This could be explained by the fact that thefaster the injection speed, the greater is the extent of theformation of the emulsion. This is a common observation inDLLME. When the mixture of extraction solvent and dispersivesolvent was injected rapidly into the aqueous sample, theextraction solvent could undergo efficient dispersion in thesample in the form of miniscule droplets, which facilitated theextraction of analytes by the extraction solvent. Thus, aninjection speed of 200 μL/s was used for the injection of themixture, this being the maximum injection speed of the CTCsystem.Overall, the optimized automated DLLME conditions were

as follows: use of 60 μL of toluene as extraction solvent, 1300

μL of acetonitrile as dispersive solvent, and 1400 μL ofacetonitrile as demulsification solvent, demulsification time of 3min, and injection speed of 200 μL/s. A 100 μL syringe wasused here to carry out the procedure. Thus, the injection of themixture of extraction solvent and dispersive solvent (in a ratio6:130) was repeated 17 times with 80 μL each time, and theinjection of demulsification solvent was repeated 20 times with70 μL each time.ANOVA was used to assess the OAD16 (4

5) results. From theresults shown in Table 1, it is indicated that all the factorsevaluated, including factor A (volume of extraction solvent),factor B (volume of dispersive solvent), factor C (volume ofdemulsification solvent), factor D (demulsification time), andfactor E (injection speed), were statistically significant at p <0.001. Furthermore, according to the percentage contribution(PC) values in Table 1, it can be concluded that, of the fiveinvestigated factors, the most significant one contributing to theextraction efficiency was the volume of extraction solvent(49%), followed by the volume of dispersive solvent (20%) andthe volume of demulsification solvent (19%), the injectionspeed (6%), and the demulsification time (3%).In the procedure, extraction time is defined as the interval

between the completion of injection of the mixture ofextraction and dispersive solvent and the start of the injectionof the demulsification solvent. To investigate the effect ofextraction time, experiments were carried out considering 1, 2,5, 8, and 10 min. The extraction time profiles are shown inFigure 3. The results showed that the extraction efficiency wasindependent of the extraction time, as has been previouslydemonstrated.22,32 Prolonged extraction time did not contrib-ute to an increase in extraction efficiency, as expected. Fastextraction is the main advantage of DLLME. Thus, extractiontime was not selected as a screening extraction parameter in theOAD matrix. A selected time of 2 min was set to ensurecomplete extraction and also in consideration of a reasonableDLLME processing time.

Method Evaluation. Under the optimized conditions, theperformance of the automated DLLME with regard to thelinearity, limits of detection (LODs), and repeatability wasinvestigated to validate the proposed method. The resultsobtained are given in Table 2.Good linearity of the method was observed in the range of

0.1−50 μg/L and 0.2−50 μg/L, respectively, depending on theanalytes, with regression coefficients (r2) higher than 0.9903 forall analytes. The repeatability was evaluated for 21 intradayreplicates and 21 interday replicates (7 replicates per day for 3consecutive days) of automated extractions and analyses inrelation to spiked ultrapure water samples (spiked at aconcentration of 10 μg/L of each analyte) at the same

Table 1. ANOVA for Experimental Responses in the OA16 (45) Matrixa

source SS d.f. MS Fb SS′ PC (%)

A 551 202.73 3 183 734.24 249.70 548 995.23 49.27B 224 085.40 3 74 695.13 101.51 221 877.90 19.92C 209 524.90 3 69 841.63 94.92 207 317.40 18.60D 40 974.40 3 13 658.13 18.56 38 766.90 3.48E 64 904.23 3 21 634.74 29.40 62 696.73 5.63error 23 546.67 32 735.833 34 584.17 3.10total 1 114 238.3 47 1 114 238.31 100

aA = volume of extraction solvent; B = volume of dispersive solvent; C = volume of demulsification solvent; D = demulsification time; E = injectionspeed; SS =sum of squares; d.f. = degrees of freedom; MS = mean squares; SS′ = purified sum of squares; PC = percentage contribution. bCriticalvalue is 6.96 (p < 0.001).

Analytical Chemistry Article

dx.doi.org/10.1021/ac404088c | Anal. Chem. 2014, 86, 3743−37493746

operational conditions for all samples. The relative standarddeviations (RSDs) were below 5.9% for all the analytes,showing good repeatability of the method. The enrichmentfactors ranged from 178 to 272.The LODs for the PEs, calculated at a signal-to-noise (S/N)

ratio of 3, were in the range of 0.01 and 0.02 μg/L. The resultsobtained by the present procedure are better than orcomparable with several previous reports, such as polypyr-role-coated magnetic particle based micro-SPE-GC/MS,38

DLLME-HPLC−variable wavelength detection,39 ultrasound-assisted emulsification microextraction-GC−flame ionizationdetection,40 solidification floating organic micorodrop basedLPME-GC/MS,41 ionic liquid cold-induced aggregationDLLME-HPLC,42 and magnetic stirring-assisted DLLME-HPLC.43 However, the current method, integrating DLLMEand GC/MS seamlessly, is automated. All the DLLME stepscould be automatically carried out for consecutive multipleextractions conveniently. The operator was only needed to bepresent at the beginning of the first experiment to initiate thesequence.Analysis of Genuine River Water and Tap Water. The

practical suitability of the proposed DLLME was evaluated bydetermining PEs in genuine river water samples collected froma local river and tap water samples directly collected from ourlaboratory. Each water sample was divided into three portionsand processed and analyzed in parallel. However, no targetanalytes were found in both the river water and tap watersamples, indicating they were not contaminated by the target

analytes or the concentrations of the target analytes were belowthe LODs of the procedure.The PE standards (10 μg/L of each analyte) were spiked into

the river water samples and the tap water samples and wereextracted and analyzed using the developed method to studymatrix effects. The relative recoveries, used to evaluate matrixeffects and defined as the ratios of the peak areas of the analytesin river water extracts to peak areas of the analytes in ultrapurewater extracts spiked at the same concentrations, werecalculated and are summarized in Table 3. The relative

recoveries varied between 84.1% and 112.8%, with RSDs (n= 5) of <6.2%, suggesting that the matrix had a negligible effecton the developed method. The results demonstrated that thepresent procedure was suitable for application to the analysis ofgenuine environmental water samples. It is, however, expectedthat more complex environmental water samples would bemore challenging to the procedure. Figure 4 shows the GC/MSchromatograms of extracts from spiked and unspiked riverwater samples.

■ CONCLUSIONFor the first time, an automated low-density solvent based/solvent demulsification dispersive liquid−liquid microextraction(automated DLLME) procedure coupled to gas chromatog-raphy−mass spectrometry (GC/MS) has been developed. Themost interesting aspect of the proposed procedure is theautomation of every step of DLLME, including the injection ofthe mixture of extraction solvent and dispersive solvent to forman emulsion, extraction of analytes, breaking up the emulsionafter extraction, collection of the extract, and injection of theextract into the GC/MS system for analysis. This cycle could berepeated many times. This allowed automation of the process,which would otherwise be impossible with centrifugation. Nohuman intervention was required in the entire extractionprocedure, except at its commencement.

Figure 3. Extraction time profiles (spiked at a concentration of 10 μg/L of each analyte and performed five times).

Table 2. Linear Range, Regression Data, Limits of Detection (LODs), Relative Standard Deviations (RSDs), and EnrichmentFactors of Automated-DLLME-GC/MSa

analyte linear range (μg/L) r2 LOD (μg/L) RSD (%) intraday (n = 21) RSD (%) interday (n = 21) enrichment factor

DMP 0.2−50 0.9945 0.02 5.4 5.9 193DEP 0.2−50 0.9916 0.02 4.1 5.0 178DBP 0.1−50 0.9978 0.01 2.3 4.9 272BBP 0.1−50 0.9903 0.02 3.5 3.7 211DEHP 0.1−50 0.9927 0.01 4.8 3.5 235DNOP 0.1−50 0.9959 0.01 5.6 5.8 251

aSpiked at 10.0 μg/L of each analyte.

Table 3. Analysis of PEs in Spiked Genuine Water Samplesby Automated-DLLME-GC/MSa

spiked river water spiked tap water

analyterelative

recovery (%)RSD (%)(n = 5)

relativerecovery (%)

RSD (%)(n = 5)

DMP 88.5 6.0 101.9 4.1DEP 91.7 3.2 84.1 5.8DBP 99.2 4.5 93.0 2.3BBP 105.4 3.9 107.6 3.7DEHP 95.6 5.1 102.7 4.4DNOP 112.8 5.6 97.3 6.2

aSpiked at 10.0 μg/L of each analyte.

Analytical Chemistry Article

dx.doi.org/10.1021/ac404088c | Anal. Chem. 2014, 86, 3743−37493747

It is also noteworthy that some interesting features werecombined in this method. First, an organic solvent with a lowerdensity than water was used as the extraction solvent inDLLME, making the retrieval of extract as the upper layerabove the sample convenient. Second, solvent demulsification,instead of centrifugation, was used to break up the emulsionafter extraction. Third, a home-designed glass vial was used asthe extraction vessel to facilitate the retrieval of extract. Allthese features contribute to the feasibility of the automatedDLLME procedure.Two orthogonal array design (OAD) matrixes were applied

to evaluate the extraction parameters. In the first stage, an OA16

(41 × 212) matrix was used to study the effects of seven factors.Toluene was selected as extraction solvent, and acetonitrile wasselected as dispersive and demulsification solvent. Then, anOA16 (4

5) matrix was applied to study five other factors in amore rigorous way. By using ANOVA, the significance andcontribution of each factor was statistically evaluated.Under the optimized conditions, the developed method

exhibited satisfactory analyte enrichment factors. Goodlinearity, limits of detection, and repeatability were alsoachieved. Overall, this study demonstrated that automatedDLLME was a convenient and efficient approach for samplepreparation. Most importantly, the proposed technique offeredgreat convenience; all the extraction steps could be conductedcompletely automatically. At the end of each extraction, theextract was automatically introduced into GC/MS system, in aseamless operation. The extraction-analytical cycle was thenrepeated. It is believed that the proposed method is the firstrealization of automated DLLME integrated with GC/MSanalysis. This opens up an innovative and interesting way toexpand the convenience and applicability of DLLME, as apotential onsite integrated sample preparation-analysis system,that can be operated unattended across multiple experiments.An automated extraction procedure based on a dual-syringeautosampler system will likely facilitate the extraction and GC/MS analysis, with a high-volume syringe for liquid transfersduring DLLME and a separate low-volume syringe for extractinjection, for improved chromatographic performance. Ourlaboratory is currently evaluating such an autosampler systemto further enhance automated DLLME-GC/MS operations.

■ ASSOCIATED CONTENT

*S Supporting InformationAdditional information as described in text. This material isavailable free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATION

Corresponding Author*Phone: +65 6516 2995. Fax: +65 6779 1691. E-mail:chmleehk@nus.edu.sg.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

The authors are grateful for the financial support of this workby the Singapore National Research Foundation under itsEnvironmental & Water Technologies Strategic ResearchProgramme, administered by the Environment & WaterIndustry Programme Office of the Public Utilities Board. Theauthors also acknowledge the support of the NationalUniversity of Singapore.

■ REFERENCES(1) Arthur, C. L.; Pawliszyn, J. Anal. Chem. 1990, 62, 2145−2148.(2) Lord, H.; Pawliszyn, J. J. Chromatogr., A 2000, 885, 153−193.(3) Nerín, C.; Salafranca, J.; Aznar, M.; Batlle, R. Anal. Bioanal. Chem.2009, 393, 809−833.(4) Liu, H. H.; Dasgupta, P. K. Anal. Chem. 1996, 68, 1817−1821.(5) Jeannot, M. A.; Cantwell, F. F. Anal. Chem. 1996, 68, 2236−2240.(6) Jeannot, M. A.; Cantwell, F. F. Anal. Chem. 1997, 69, 2935−2940.(7) Xu, L.; Basheer, C.; Lee, H. K. J. Chromatogr., A 2007, 1152,184−192.(8) Yang, Y.; Miller, D. J.; Hawthorne, S. B. J. Chromatogr., A 1998,800, 257−266.(9) Chen, H.; Chen, R. W.; Feng, R.; Li, S. Q. Chromatographia2009, 70, 165−172.(10) Zhang, J.; Su, T.; Lee, H. K. Anal. Chem. 2005, 77, 1988−1992.(11) Jiang, X. M.; Oh, S. Y.; Lee, H. K. Anal. Chem. 2005, 77, 1689−1695.(12) Hou, L.; Lee, H. K. Anal. Chem. 2003, 75, 2784−2789.(13) He, Y.; Lee, H. K. Anal. Chem. 1997, 69, 4634−4640.(14) Pedersen-Bjergaard, S.; Rasmussen, K. E. Anal. Chem. 1999, 71,2650−2655.(15) Shen, G.; Lee, H. K. Anal. Chem. 2002, 74, 648−654.(16) Basheer, C.; Jayaraman, A.; Kee, M. K.; Valiyaveettil, S.; Lee, H.K. J. Chromatogr., A 2005, 1100, 137−143.

Figure 4. GC/MS chromatograms of genuine samples obtained from automated DLLME under the optimized conditions: (a) spiked river watersample, (b) blank river water sample. Peaks: (1) DMP, (2) DEP, (3) DBP, (4) BBP, (5) DEHP, and (6) DNOP.

Analytical Chemistry Article

dx.doi.org/10.1021/ac404088c | Anal. Chem. 2014, 86, 3743−37493748

(17) Ratola, N.; Alves, A.; Kalogerakis, N.; Psillakis, E. Anal. Chim.Acta 2008, 618, 70−78.(18) Liu, W. P.; Lee, H. K. Anal. Chem. 2000, 72, 4462−4467.(19) Jiang, X. M.; Lee, H. K. Anal. Chem. 2004, 76, 5591−5596.(20) Wang, L.; Wang, L. L.; Chen, J.; Du, W. J.; Fan, G. L.; Lu, X. H.J. Chromatogr., A 2012, 1256, 9−14.(21) Liu, W.; Zhang, L.; Fan, L. B.; Lin, Z.; Cai, Y. M.; Wei, Z. Y.;Chen, G. N. J. Chromatogr., A 2012, 1233, 1−7.(22) Rezaee, M.; Assadi, Y.; Hosseini, M. M.; Agnee, E.; Ahmadi, F.;Berijani, S. J. Chromatogr., A 2006, 1116, 1−9.(23) Pusvaskiene, E.; Januskevic, B.; Prichldko, A.; Vickackaite, V.Chromatographia 2009, 69, 271−276.(24) Negreira, N.; Rodríguez, I.; Rubí, E.; Cela, R. Anal. Bioanal.Chem. 2010, 398, 995−1004.(25) Leong, M. I.; Chang, C. C.; Fuh, M. R.; Huang, S. D. J.Chromatogr., A 2010, 1217, 5455−5461.(26) Campillo, N.; Vinas, P.; Cacho, J. I.; Penalver, R.; Hernandez-Cordoba, M. J. Chromatogr., A 2010, 1217, 7323−7330.(27) Karimi, M.; Sereshti, H.; Samadi, S.; Parastar, H. J. Chromatogr.,A 2010, 1217, 7017−7023.(28) Saleh, A.; Yamini, Y.; Faraji, M.; Rezaee, M.; Ghambarian, M. J.Chromatogr., A 2009, 1216, 6673−6679.(29) Farajzadeh, M. A.; Seyedi, S. E.; Shalamzari, M. S.; Bamorowat,M. J. Sep. Sci. 2009, 32, 3191−3200.(30) Yiantzi, E.; Psillakis, E.; Tyrovola, K.; Kalogerakis, N. Talanta2010, 80, 2057−2062.(31) Zacharis, C. K.; Tzanavaras, P. D.; Roubos, K.; Khima, K. J.Chromatogr., A 2010, 1217, 5896−5900.(32) Shi, Z. G.; Lee, H. K. Anal. Chem. 2010, 82, 1540−1545.(33) Guo, L.; Lee, H. K. J. Chromatogr., A 2012, 1235, 1−9.(34) Leong, M. I.; Huang, S. D. J. Chromatogr., A 2008, 1211, 8−12.(35) Anthemidis, A. N.; Ioannou, K. I. G. Talanta 2009, 79, 86−91.(36) Chen, H.; Chen, R. W.; Li, S. Q. J. Chromatogr., A 2010, 1217,1244−1248.(37) Guo, L.; Lee, H. K. J. Chromatogr., A 2011, 1218, 5040−5046.(38) Meng, J. R.; Bu, J.; Deng, C. H.; Zhang, X. M. J. Chromatogr., A2011, 1218, 1585−1591.(39) Liang, P.; Xu, J.; Li, Q. Anal. Chim. Acta 2008, 609, 53−58.(40) Yan, H. Y.; Cheng, X. L.; Liu, B. M. J. Chromatogr., B 2011, 879,2507−2512.(41) Farahani, H.; Ganjali, M. R.; Dinarvand, R.; Norouzi, P. Talanta2008, 76, 718−723.(42) Zhang, H.; Chen, X. Q.; Jiang, X. Y. Anal. Chim. Acta 2011, 689,137−142.(43) Ranjbri, E.; Hadjmohammadi, M. R. Talanta 2012, 100, 447−453.

Analytical Chemistry Article

dx.doi.org/10.1021/ac404088c | Anal. Chem. 2014, 86, 3743−37493749