a n a l y t i c a c h i m i c a a c t a 6 2 9 ( 2 0 0 8 ) 92–97

avai lab le at www.sc iencedi rec t .com

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

Application of robust NiTi–ZrO2–PEG SPME fiberin the determination of haloanisoles incork stopper samples

Dilma Budziak, Edmar Martendal, Eduardo Carasek ∗

Departamento de Química, Universidade Federal de Santa Catarina, Florianópolis, SC 88040-900, Brazil

a r t i c l e i n f o

Article history:

Received 8 May 2008

Received in revised form

7 August 2008

Accepted 12 September 2008

Published on line 21 September 2008

Keywords:

Sol–gel

Poly(ethylene glycol)

NiTinol

a b s t r a c t

In this study, a novel solid-phase microextraction (SPME) fiber obtained using sol–gel tech-

nology was applied in the determination of off-flavor compounds (2,4,6-trichloroanisole

(TCA), 2,4,6-tribromoanisole (TBA) and pentachloroanisole (PCA)) present in cork stopper

samples. A NiTi alloy previously electrodeposited with zirconium oxide was used as the

substrate for a poly(ethylene glycol) (PEG) coating. Scanning electronic microscopy showed

good uniformity of the coating and allowed the coating thickness to be estimated as around

17 �m. The optimization of the main parameters influencing the extraction efficiency, such

as cork sample mass, sodium chloride mass, extraction temperature and extraction time

were optimized using a full factorial design, followed by a Doehlert design. The optimum

conditions were: 20 min of extraction at 70 ◦C using 60 mg of the cork sample and 10 mL of

water saturated with sodium chloride in a 20 mL amber vial with constant magnetic stirring.

Zirconium oxide

Haloanisoles

Cork

Solid-phase microextraction

Satisfactory detection limits between 2.5 and 5.1 ng g−1 were obtained, as well as good preci-

sion (R.S.D. in the range of 5.8–12.0%). Recovery tests were performed on three different cork

samples, and values between 83 and 119% were obtained. The proposed SPME fiber was com-

pared with commercially available fibers and good results were achieved, demonstrating its

applicability.

[5,6].

1. Introduction

The quality of a wine is highly dependent on its aroma. Manyresearchers have evaluated compounds responsible for theoriginal wine odor as well as the occurrence of unpleasantodors, called off-flavors [1]. One of the most serious prob-lems in the wine industry is the undesirable odor caused bya contaminated cork stopper (also called cork taint) [2]. Suchcontamination is responsible for high economic losses in the

wine and cork industries [3].

Several substances, such as geosmin, 2-methyl-isoborneol,guaiacol, 1-octen-3-one, 1-octen-3-ol and chlorophenols have

∗ Corresponding author. Fax: +55 48 37216845.E-mail address: carasek@qmc.ufsc.br (E. Carasek).

0003-2670/$ – see front matter © 2008 Elsevier B.V. All rights reserved.doi:10.1016/j.aca.2008.09.038

© 2008 Elsevier B.V. All rights reserved.

been reported to be responsible for the organoleptic defect [1].However, the main compounds responsible for this unpleas-ant odor are some haloanisoles: 2,4,6-trichloroanisole (TCA),2,3,4,6-tetrachloroanisole (TeCA), pentachloroanisole (PCA)and 2,4,6-tribromoanisole (TBA) [4].

Chlorophenols have been frequently used as fungicides,herbicides or wood preservatives, and are used in wine-growing and by some manufacturers of cork stoppers

The sensorial threshold limit for TCA in wine varies con-siderably in the literature (1.4–210 ng L−1) and for TBA a valueof 7.9 ng L−1 has been reported [6].

c t a

aqt

ait[

b(

sr(hrsi

sttsud

2

2

CGitot

gptI3

o(

2

H(cUaacws

a n a l y t i c a c h i m i c a a

These off-flavor compounds are present in cork stopperst low concentration (nanograms per gram of cork). Conse-uently, it is necessary to develop suitable analytical methodso determine their presence in cork samples [7].

Solid-phase microextraction (SPME) has been successfullypplied in the determination of haloanisoles and halophenolsn wine and cork samples [1,3,8], presenting great advan-ages compared to the traditional techniques that use solvents9,10].

Its main disadvantage is the fragility. However, this haseen overcome through application of different substrates

mainly metallic wires) to replace the fused silica rod [11–16].A NiTi alloy, which has interesting characteristics, such as

hape memory effect, superelasticity, durability, wear and cor-osion resistance [17], electrodeposited with zirconium oxideZrO2), which offers strong adhesion onto metallic surfaces,igh thermal stability, biocompatibility, wear and corrosionesistance [18,19], is applied in this study as a new substrate forol–gel reactions. These advantages are important for improv-ng physical and chemical stability of the coating [20–22].

The aforementioned advantages, combined with those ofol–gel coating technology, indicate that zirconium oxide elec-rolitically deposited onto a NiTi alloy coated using sol–gelechnology may represent a promising SPME fiber. In thistudy, poly(ethylene glycol) (PEG) was coated onto NiTi–ZrO2

sing the sol–gel technology and this fiber was applied in theetermination of TCA, TBA and PCA in cork stoppers.

. Experimental

.1. Instrumentation

hromatographic analyses were performed with a ShimadzuC-14B gas chromatograph, equipped with split/splitless

njector. An electron capture detector was used for the detec-ion of haloanisoles. Chromatographic separation was carriedut in an OV-5 capillary column (30 m × 0.25 mm, 0.25 �m filmhickness; OV Specialty Chemical, Marietta, OH).

Ultrapure nitrogen was used as the carrier and make-upas at 1.0 and 35 mL min−1, respectively. Column oven tem-erature was 60 ◦C (1 min), 20 ◦C min−1 to 130 ◦C, 10 ◦C min−1

o 175 ◦C, 1 ◦C min−1 to 181 ◦C and 40 ◦C min−1 to 280 ◦C (2 min).njector and detector temperatures were fixed at 290 and00 ◦C, respectively.

Morphology and composition of the fibers with and with-ut coating were evaluated by scanning electron microscopy

SEM), using a Philips XL-30 microscope.

.2. Chemicals

aloanisole standards, including TCA, TBA and PCASigma–Aldrich, Milwaukee, WI, USA), were gravimetri-ally prepared in HPLC-grade methanol (Tedia, Fairfield, OH,SA) at a concentration of 2000 mg L−1. From this solution,diluted solution at a concentration of 2000 �g L−1 of each

nalyte was also prepared in methanol and used to spike theork samples. Sodium chloride (Merck, Darmstadt, Germany)as used for the modification of the ionic strength of the

amples.

6 2 9 ( 2 0 0 8 ) 92–97 93

Sodium hydroxide (Vetec) and hydrochloric acid (Merck)were both used for activation of the NiTi–ZrO2 surface beforesol–gel coating.

The sol–gel materials were methyltrimethoxysilane(MTMS) (UCT, Bristol, PA, USA), trimethylmethoxysi-lane (TMMS) and PEG methyl ether (Sigma–Aldrich) andtrifluoroacetic acid (TFA) (Vetec).

2.3. Preparation of sol–gel fibers

Wires of NiTi with 0.2 mm thickness and approximately 2 cmlength were used as the support for electrodeposition ofZrO2. The optimum coating conditions were obtained fromGiacomelli et al. [23]. More details on the electrodepositionprocedure employed for the fiber manufacturing are presentedin our previous studies [24,25].

The NiTi alloy previously electrodeposited with ZrO2 wassubmerged in a 1.0 mol L−1 NaOH solution for 1 h. The fiberswere then rinsed with 0.1 mol L−1 HCl for 30 min and the fiberswere washed with water and used within 5 h.

The PEG sol-solution was prepared by mixing 400 �L ofMTMS, 800 mg of PEG and 500 �L of TFA (95%) in glass cap-illary tubes. One centimeter of the fiber was dipped into thesol-solution for about 40 min. This procedure was repeated fivetimes, always with a freshly prepared sol-solution. Fibers weremounted in commercial SPME devices. Three fibers withoutthe ZrO2 film were also coated in the same way as those withZrO2. All fibers were thermally conditioned in the GC injectorport at 250 ◦C for 4 h.

2.4. Preparation of cork samples

Cork samples were ground and particles smaller than 1 mmwere selected for the analysis. The samples were kept in ambervials at 5 ◦C until analysis. For optimization of the method, thesamples were spiked with 300 ng g−1 of TCA and 200 ng g−1 ofTBA and PCA, and kept in a closed amber vial for interactionof the analytes with the sample matrixes for at least 72 h.

2.5. Optimization of the method for extraction ofhaloanisoles

Factors influencing the extraction of the analytes were initiallystudied by means of a full factorial design. Four factors werestudied at two levels: extraction temperature (40 and 70 ◦C),sodium chloride mass (0 and 3 g), extraction time (10 and30 min) and sample mass (100 and 200 mg). A triplicate centerpoint was included in the design (55 ◦C, 1.5 g of NaCl, 20 minof extraction and 150 mg of sample) to estimate the linear-ity of the response between the levels chosen for the factorialdesign. From the results obtained, extraction temperature andsample mass were evaluated through a Doehlert matrix. Sam-ple mass was evaluated at five levels (10, 35, 60, 85 and 110 mg)and extraction temperature at three levels (60, 70 and 80 ◦C).The data were processed using the Statistica 6.0 computer pro-gram. Each sample mass required for the optimization was

weighed directly in a 20 mL amber vial in which 10 mL of waterwas added. Magnetic stirring was used throughout the extrac-tion procedure. All extractions were carried out in headspacemode.

94 a n a l y t i c a c h i m i c a a c t a 6 2 9 ( 2 0 0 8 ) 92–97

roce

Fig. 1 – SEM micrographs of fibers coated using the sol–gel p270×.

3. Results and discussion

3.1. Characterization of the NiTi–ZrO2–PEG fiber

Fig. 1 shows the images obtained by SEM for two NiTi alloyfibers, prepared with and without ZrO2 coating, and thencoated with PEG through the sol–gel process. Both fibers werepreviously thermally conditioned in the GC injector port beforeSEM analysis.

As can be observed in Fig. 1, the surface of the coatedNiTi–ZrO2 fiber (Fig. 1A) is much more uniform and homoge-neous than that of the fiber without ZrO2 film (Fig. 1B). This ismost probably due to the presence of zirconia film on the NiTisurface, which can provide a much more suitable surface (con-taining Zr–OH groups) where the sol–gel reactions can takeplace, leading to a greater adhesion of the polymeric coating tothe substrate. Thus, the electrodeposition of ZrO2 on the NiTialloy was shown to be a very important step in the fiber man-ufacturing procedure. The thickness of the polymeric coatingwas evaluated by comparing micrographs with and withoutcoating, resulting in approximately 17 ± 3 �m.

3.2. Desorption temperature of the NiTi–ZrO2–PEGfiber

Desorption temperature and thermal stability of the PEGcoated fiber was evaluated in the range of 230–320 ◦C usingthe GC injector port. Extraction conditions were set at: 30 minof extraction at 40 ◦C with analytes in pure water at 0.5 �g L−1.

Five replicates were performed at each temperature, and thedesorption time was fixed at 10 min. No loss of peak area wasobserved on increasing the desorption temperature, suggest-ing that the polymeric coating is chemically bonded to the

ss using NiTi–ZrO2 (A) and NiTi only (B). Magnification of

substrate. High thermal stability is a highly desirable featurein an SPME fiber, as it allows an increase in the desorptiontemperature, improving the peak resolution, lowering the des-orption time and eliminating the carry-over effects. For thesubsequent studies, the desorption time and temperaturewere set at 5 min and 290 ◦C, respectively, with no carry-overeffects being observed.

3.3. Optimization of the haloanisoles extractionconditions for cork samples using the NiTi–ZrO2–PEG fiber

The parameters sodium chloride mass, sample mass, extrac-tion temperarure and extraction time were multivariatelyoptimized, firstly by a full factorial design, to estimate themain and interaction effects on the analytical response. Theresponse used for each experiment required by the design wasthe geometric mean of the peak area of the three analytes. Theresults of the factorial design are given in Fig. 2. Effects thatcross the vertical line are statistically significant at the 95%confidence level.

The variable with the highest significance is sample mass,which has a negative sign, indicating that a lower sample massleads to a higher response. Due to the lower density of thesample, most of the cork particles stay at the surface of theaqueous solution, even with efficient stirring. It is assumedthat high sample masses create a barrier that hinders thetransfer of the analytes from the sample to the headspace.The variables extraction temperature and ionic strength werealso significant, but with positive values, indicating that betterresponses are obtained at the higher levels evaluated for these

variables. Interestingly, two interaction effects were signifi-cant: (i) temperature and sodium chloride mass (1 × 2) and (ii)sodium chloride mass and sample mass (2 × 4). This indicatesthat these variables cannot be evaluated using a univariate

a n a l y t i c a c h i m i c a a c t a 6 2 9 ( 2 0 0 8 ) 92–97 95

Fig. 2 – Pareto chart obtained for the full factorial design foroptimization of the variables and their interactions forh

aualnbwcctiftsa

teatruo

m1mcpTeca

3

TteLt

Three cork samples were analyzed using the proposedmethod. The analytes were not detected in any of the sam-ples. Thus, to evaluate the accuracy of the method, the cork

Table 1 – Linear range, correlation coefficients, detectionand quantification limits obtained for the method forhaloanisole determination in cork samples using theNiTi–ZrO2–PEG fiber

Compound Linearrangea

Rb LODc

(ng g−1)LOQd

(ng g−1)

TCA 20–400 0.9989 5.1 17.0TBA 20–300 0.9993 6.2 20.6PCA 10–300 0.9954 2.5 8.3

aloanisole determination. R2 = 0.93975.

pproach if a real optimum is to be found. In other words, anivariate procedure would probably lead to a false optimum,nd some of the sensitivity in the method would likely beost. The interpretation of the curvature parameter (which wasot significant, indicating that the response is roughly linearetween the levels chosen for each variable), in combinationith the main and interaction effects of the salt addition, indi-

ates that the higher level evaluated for this variable is the bestondition, giving a better extraction efficiency. As this concen-ration is close to the sample saturation, it is not possible tonvestigate higher concentrations of sodium chloride. Thus,or simplicity, this variable was set at 3.0 g per 10 mL waterhroughout the study. The variable extraction time was notignificant at the levels studied, and, for simplicity, it was fixedt 20 min.

For the final optimization of the sample mass and extrac-ion temperature, a response surface methodology wasmployed, using a Doehlert matrix. Sample mass was studiedt five levels (10, 35, 60, 85 and 110 mg) and extraction tempera-ure at three levels (60, 70 and 80 ◦C). Once again, the analyticalesponse was taken as the geometric mean of the individ-al responses for the three analytes. The response surfacebtained is shown in Fig. 3.

The data obtained were fitted to a second-order polyno-ial: R = − 3832633.8 + 114204.7x − 820.2x2 + 4464.1y − 46.4y2 +

8.1xy. From this equation (which has a satisfactory deter-ination coefficient of 0.9253) it is possible to estimate the

ritical point of the surface: 70.3 ◦C and 61.8 mg. This criticaloint is clearly a maximum point, as can be seen in Fig. 3.hus, the results of the optimization are: 20 min of headspacextraction at 70 ◦C using 60 mg of the cork sample which is inontact with 10 mL of saturated sodium chloride in a 20 mLmber vial with constant magnetic stirring.

.4. Performance of the NiTi–ZrO2–PEG fiber

he performance of the NiTi–ZrO2–PEG fiber in the extrac-

ion of haloanisoles in cork samples after optimization wasvaluated through obtention of the main analytical features.inear range, correlation coefficient, and detection and quan-ification limits are presented in Table 1. Good correlation

Fig. 3 – Response surface obtained from the Doehlertdesign.

coefficients higher than 0.9954 were obtained. Detection limitsare between 2.5 and 6.2 ng g−1, being comparable with someother studies reported in the literature. Lorenzo et al. [26]using headspace sorptive extraction and detection by GC–MSobtained LOD values of 2.67, 30.65 and 11.01 ng g−1 for TCA,TBA and PCA, respectively. Carasek et al. [2] employing HS-SPME coupled to GC–TOF-MS obtained detection limits of 1.2and 1.6 ng g−1 for TCA and PCA, respectively.

Repeatability of the method was assessed through sevenconsecutive extractions of independently prepared solutionscontaining 60 mg of cork sample spiked with 25 ng g−1 of TCAand TBA and 15 ng g−1 of PCA. Relative standard deviationswere in the range of 5.8–12.0%. Fiber-to-fiber reproducibilitywas also checked with five different fibers, and the R.S.D. waslower than 13%.

3.5. Application to cork samples

a Studied linear range in ng g−1.b Correlation coefficient of calibration curve.c Limit of detecton.d Limit of quantification.

96 a n a l y t i c a c h i m i c a a c t

Table 2 – Recovery tests for the three cork samplesanalyzed by the proposed method

Spike (ng g−1) Recovery (%)

Sample I Sample II Sample III

TCA84.8 86.3 ± 3.6 93.0 ± 5.0 83.7 ± 2.9254.4 90.5 ± 3.4 109.6 ± 2.9 92.2 ± 5.3

TBA68.6 98.4 ± 3.8 101.6 ± 1.8 100.9 ± 5.5205.6 119.0 ± 4.2 113.0 ± 2.5 117.8 ± 3.8

PCA63.4 103.6 ± 2.6 106.2 ± 4.3 100.4 ± 1.5190.1 102.5 ± 3.1 106.9 ± 5.5 102.5 ± 3.0

Fig. 4 – Chromatographic profile of a spiked cork stopper

r

sample: (1) TCA; (2) TBA; (3) PCA.

samples were spiked at two levels of concentration: 84.8 and254.4 ng g−1 for TCA, 68.6 and 205.6 ng g−1 for TBA and 63.4and 190.1 ng g−1 for PCA. The recovery values are shown inTable 2. Satisfactory values in the range of 83.2 and 119.0%were obtained, taking into account the complexity of the sam-

ple. The chromatographic profile of an spiked cork sample canbe seen in Fig. 4.

Fig. 5 – Comparison of the NiTi–ZrO2–PEG 17 �m fiber withcommercially available fibers PA 85 �m and CW/DVB 65 �m.

a 6 2 9 ( 2 0 0 8 ) 92–97

3.6. Comparison of the NiTi–ZrO2–PEG fiber withcommercial fibers

The extraction efficiency of the proposed fiber was com-pared with two commercially available SPME fibers: car-bowax/divinylbenzene (CW/DVB) and polyacrylate (PA). Theresults are presented in Fig. 5.

The extraction profile of the proposed NiTi–ZrO2–PEG fiberwas comparable to the other fibers evaluated. The PA 85 �mfiber had a better sensitivity toward the analytes, possibly dueto the higher coating thickness. The proposed NiTi–ZrO2–PEGfiber, even with a thinner coating of around 17 �m, had ahigher affinity for the analytes compared to that of the 65 �mCW/DVB fiber.

4. Conclusions

The proposed method using the new NiTi–ZrO2–PEG SPMEfiber showed good selectivity and sensitivity toward the ana-lytes, with detection limits at the ng g−1 level. Good accuracyand precision were obtained, demonstrating the suitabilityof the method. The chemometric tools used herein allowedus to obtain the optimum conditions quickly and reliably.The sensitivity of the NiTi–ZrO2–PEG fiber was comparable tothose of the commercially available CW/DVB 65 �m and PA85 �m fibers. This fiber was used for more than 400 extrac-tion and desorption cycles with no loss of efficiency. Thus, itis a promising alternative for use in SPME, being easily andinexpensively prepared.

Acknowledgments

The authors thank CNPq for financial support and NanoEndo-luminal Company for supplying the NiTi wires.

e f e r e n c e s

[1] S. Insa, V. Salvadó, E. Anticó, J. Chromatogr. A 1047 (2004) 15.[2] E. Carasek, E. Cudjoe, J. Pawliszyn, J. Chromatogr. A 1138

(2007) 10.[3] P. Vlachos, A. Kampioti, M. Kornaros, G. Lyberatos, Food

Chem. 105 (2007) 681.[4] C. Pizarro, N. Pérez-del-Notario, J.M. González-Sáiz, J.

Chromatogr. A 1166 (2007) 1.[5] M. Riu, M. Mestres, O. Busto, J. Guasch, Anal. Chim. Acta 563

(2006) 310.[6] S. Jönsson, T. Uusitalo, B. van Bavel, I.-B. Gustafsson, J.

Chromatogr. A 1111 (2006) 71.[7] Ó. Ezquerro, Á. Garrido-López, M.T. Tena, J. Chromatogr. A

1102 (2006) 18.[8] A. Martínez-Urunuela, J.M. González-Sáiz, C. Pizarro, J.

Chromatogr. A 1056 (2004) 49.[9] H. Lord, J. Pawliszyn, J. Chromatogr. A 885 (2000) 153.

[10] T.P. Gbatu, K.L. Sutton, J.A. Caruso, Anal. Chim. Acta 402(1999) 67.

[11] H. Bagheri, A. Mir, E. Babanezhad, Anal. Chim. Acta 532(2005) 89.

[12] D. Djozan, Y. Assadi, S.H. Haddadi, Anal. Chem. 73 (2001)4054.

[13] Dj. Djozan, S. Bahar, Chromatographia 59 (2004) 95.

c t a

a n a l y t i c a c h i m i c a a

[14] M. Azenha, P. Nogueira, A. Fernando-Silva, Anal. Chem. 78(2006) 2071.

[15] M.A. Farajzadeh, N.A. Rahmani, Talanta 65 (2005) 700.[16] H. Minjia, T. Chao, Z. Qunfang, J. Guibin, J. Chromatogr. A

1048 (2004) 257.

[17] L. Tan, W.C. Crone, Acta Mater. 50 (2002) 4449.[18] I. Zhitomirsky, A. Petric, Mater. Lett. 46 (2000) 1.[19] S.K. Yen, M.J. Guo, H.Z. Zan, Biomaterials 22 (2001) 125.[20] J. Yu, L. Dong, C. Wu, L. Wu, J. Xing, J. Chromatogr. A 978

(2002) 37.

6 2 9 ( 2 0 0 8 ) 92–97 97

[21] L. Yun, Anal. Chim. Acta 486 (2003) 63.[22] Z. Zeng, W. Qiu, Z. Huang, Anal. Chem. 73 (2001) 2429.[23] F.C. Giacomelli, C. Giacomelli, A.G. Oliveira, A. Spinelli,

Mater. Lett. 59 (2005) 754.[24] D. Budziak, E. Martendal, E. Carasek, J. Chromatogr. A 1164

(2007) 18.[25] D. Budziak, E. Martendal, E. Carasek, Anal. Chem. Acta 598

(2007) 254.[26] C. Lorenzo, A. Zalacain, G.L. Alonso, M.R. Salinas, J.

Chromatogr. A 1114 (2006) 250.