Flow-injection determination of total organic ﬂuorine with off-line deﬂuorination reaction on a solid sorbent bed

(1)

a v a i l a b l e a t w w w . s c i e n c e d i r e c t . c o m

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / a c a

Flow-injection determination of total organic ﬂuorine with off-line deﬂuorination reaction on a solid sorbent bed

Jacek Musijowski

^a

, Marek Trojanowicz

^a^,∗

, Bogdan Szostek

^b

, Jos´e Luis Fontes da Costa Lima

^c

, Rui Lapa

^c

, Hiroki Yamashita

^d

, Toshio Takayanagi

^d

, Shoji Motomizu

^d

aLaboratory of Flow Analysis and Chromatography, Department of Chemistry, Warsaw University, Pasteura 1, 02-093 Warsaw, Poland

bDuPont Haskell Laboratory for Health and Environmental Sciences, 1090 Elkton Road, Newark, DE 19714-0050, USA

cLaboratory of Physical Chemistry, Faculty of Pharmacy, University of Porto, Rua Anibal Cunha 164, 4099-030 Porto, Portugal

dDepartment of Chemistry, University of Okayama, Okayama 700-8530, Japan

a r t i c l e i n f o

Article history:

Received 8 December 2006 Received in revised form 13 March 2007

Accepted 29 March 2007 Published on line 6 April 2007

Keywords:

Perﬂuorocarboxylic acids Solid-phase extraction Flow-injection analysis Total organic ﬂuorine Sodium biphenyl

a b s t r a c t

Considering recent reports on widespread occurrence and concerns about perfluoroalkyl substances (PFAS) in environmental and biological systems, analysis of these compounds have gained much attention in recent years. Majority of analyte-specific methods are based on a LC/MS/MS or a GC/MS detection, however many environmental or biological studies would benefit from a total organic fluorine (TOF) determination. Presented work was aimed at developing a method for TOF determination. TOF is determined as an amount of inorganic fluoride obtained after defluorination reaction conducted off-line using sodium biphenyl reagent directly on the sorbent without elution of retained analytes. Recovered fluoride was analyzed using flow-injection system with either fluorimetric or potentiometric detection.

The TOF method was tested using perfluorocarboxylic acids (PFCA), including perfluorooc- tanoic acid (PFOA), as model compounds. Considering low concentrations of PFAS in natural samples, solid-phase extraction as a preconcentration procedure was evaluated. Several carbon-based sorbents were tested, namely multi-wall carbon nanotubes, carbon nanofibres and activated carbon. Good sorption of all analytes was achieved and defluorination reaction was possible to carry out directly on a sorbent bed. Recoveries obtained for PFCAs, adsorbed on an activated carbon sorbent, and measured as TOF, were 99.5±1.7, 110±9.4, 95±26, 120±32, 110±12 for C4, C6, C8, C10 and C12-PFCA, respectively. Two flow systems that would enable the defluorination reaction and fluoride determination in a single system were designed and tested.

1. Introduction

Interactions of halogenated organics, especially chlorinated and brominated, with biological and environmental systems

∗Corresponding author. Tel.: +48 22 822 35 32; fax: +48 22 822 35 32.

E-mail address:trojan@chem.uw.edu.pl(M. Trojanowicz).

have forced strict control of their production and release to the environment. Fluorinated chemicals initially did not gain much attention considering high stability of C F bond and consequently an assumption of negligible activity. Currently,

doi:10.1016/j.aca.2007.03.068

(2)

however, it is commonly acknowledged that the fluorinated organics, especially the perfluoroalkyl substances (PFAS) do impact living organisms in numerous ways, show a tendency of accumulation in the environment and many of PFAS are classified as persistent organic pollutants[1].

Organic fluorine has been present in the environment in the form of simple hydrochlorofluorocarbons (HCFCs) of volcanic origin and biologically produced compounds containing single fluorine atom [2,3]. However, in recent decades numerous synthetic fluoroorganic compounds have been produced and widely employed in various branches of industry, in agriculture as pesticides and domestic products as surfactants or oliophobic agents [1]. Extensive use of synthetic products containing fluorinated organics has significantly increased its content in the environment.

The spectrum of anthropogenic fluorinated chemicals vary from fluorocarbons containing from one to four fluorine atoms, through abundant family of trifluoromethyl derivatives up to fully fluorinated polymers like polytetrafluoroethy- lene (PTFE) [1]. However, PFAS is a group of fluorinated chemicals that gained particular attention in recent years.

Carbon chain entirely substituted with fluorine is excep- tionally inert and thus PFAS have very little environmental activity [1,4]. Commonly analyzed are perfluorocarboxylic acids (PFCAs) with 4–15 carbon atoms in the alkyl chain [4], fluorotelomer alcohols [4,5], perfluorosulfonates (especially perfluoroocatnesulfonate—PFOS)[5-7]and perfluorosul- fonamides (especially perfluorooctanesulfonamide—PFOSA) [5,7]. PFCAs and PFOS are not metabolized in biological systems[8,9], however the partially fluorinated compounds are usually degraded to neighboring perfluorinated homologues and defluorination in biological conditions is possible only for compounds with individual fluorine atoms[1].

All the compounds that do not undergo environmental degradation will contribute to the total organic ﬂuorine (TOF).

Considering variance of chemical entities containing fluorine, it is virtually impossible to develop a single analytical method, which would allow determination of them all. Therefore, a TOF measurement provides valuable information about the TOF load in a sample and completeness of detection of all fluorinated organic compounds present in a sample, when contrasted with the results of analyte specific methods e.g.

LC/MS/MS.

TOF determination methodology consists of breaking C F bonds and quantification of obtained inorganic fluoride. There are several ways of carrying out the first stage of the procedure, however they are all based on treatment of fluoroorganics with high temperature or aggressive chemicals. Although open ash- ing is practical for samples with low content of perfluorinated organics[10], it may produce incomplete defluorination in case of fully fluorinated compounds[11]. Moreover, it is believed to be easily exposed to external contamination[12,16]. More resistant to interferences and generally more reliable, however also more cumbersome, are procedures employing confined combustion such as oxygen-bomb [13,14]. Defluorination using a Wickbold oxyhydrogen-flame combustion is one of the most efficient ways of decomposition of fluorinated organics [15,16]. Nevertheless, handling hydrogen–oxygen mixture may be hazardous for inexperienced operator. One of the most efficient defluorination procedure utilizing a chemical agent has

been demonstrated by Venkateswarlu[17]conducting reduc- tive cleavage of C F using commercially available sodium biphenyl reagent (SBP). Considering high efﬁciency of ﬂuo- ride recovery and simple execution at room temperature, the reaction with SBP seems to be the most suitable for this appli- cation. Challenging aspect of the reaction is the fact that SBP is immediately deactivated in contact with water; therefore it requires strictly non-aqueous conditions.

Second stage of the procedure, determination of obtained inorganic fluoride, can be carried out using numerous, well known, sensitive and reliable methods. One of most fre- quently employed techniques is the potentiometric detection of fluoride with an ion-selective electrode (F-ISE)[18–20]. It allows very selective and straightforward fluoride determination with a limit of detection (LOD) of 0.02 mg L⁻¹F⁻[21].

Fluorimetric methods of fluoride determination are indirect and usually very sensitive. Proportional to fluoride concentration changes in fluorescence of Zr–quercitin complex allowed determination at LOD level of 0.06 mg L⁻¹[22]. LOD of 10␮g L⁻¹ is achievable using changes in the rate of formation of fluorescent complex of Al³⁺–Eriochrome Red B [23]. Linear correlation between the enhancement of chemiluminescence of luminol–periodate system and fluoride concentration, allowed its determination with remarkable LOD of 20 ng L⁻¹ [24]. Very low LOD of 0.8␮g L⁻¹ was reported using electro- spray mass spectrometry[25], however serious drawback of the method is its susceptibility to matrix effects. More recently, techniques such as capillary electrophoresis (CE) and ion chromatography (IC) with different detection methods were employed. Determination of fluoride in pharmaceutical for- mulations using CE with conductometric detection (CD) was carried out with LOD of approximately 20␮g L⁻¹[26]. CE with indirect UV detection allowed reaching LOQ of 0.04 mg L⁻¹ in toothpaste[27]. Transient isotachophoretic preconcentration and the same type of detection produced LOD at level of 0.2␮g L⁻¹[28]. Employing IC–CD, LOD of 0.040 mg L⁻¹in battery electrolyte was reported[29]. In excess of aluminum ions, fluoride can be determined as aluminum monofluoride (AlF²⁺).

Employing fluorescence detection and molecular absorp- tion spectrometry LODs of approximately 0.2␮g L⁻¹ [30]and 30␮g L⁻¹[31]were obtained, respectively. When transformed into volatile fluorosilanes, fluoride can be determined by means of gas chromatography[32]at the level of 0.01 mg L⁻¹. Finally, series of methods employing nuclear reactions are possible like photon or neutron activation analysis[33–35].

The aim of this research was to design a flow-injection system that would allow determination of TOF employing off-line defluorination with SBP directly on a sorbent bed and subsequent fluoride detection. Minimization of manual operations potentially would provide improvement in precision, elimina- tion of aggressive chemicals handling as well as reduction of possible sources of contamination. This research was divided into three stages.

First one was to obtain a selective and sensitive tool for fluoride detection after defluorination reaction. Potentiomet- ric and fluorimetric methods present certain advantages. They are mature, well developed techniques, easily adaptable into a flow mode, provide satisfactory sensitivity with limit of detection at the level of tens of␮g L⁻¹and require relatively simple and inexpensive instrumentation.

(3)

The second stage originated from a concern about the low level concentrations of ﬂuorochemicals present in natural samples. Because the analytes, expressed as TOF, are found at␮g L⁻¹and sub␮g L⁻¹ levels of ﬂuoride, the procedure requires a preconcetration step. The most commonly employed method of preconcentration, employed both for organic and inorganic species in modern trace analysis is a solid-phase extraction (SPE) with the use of solid sorbents.

In ﬂow systems such steps are employed on-line mostly for determination of trace metals, but numerous applications have also been described for preconcentration of organic species such as carbaryl and 1-naphthol with FIA with FTIR detection[36], or metal carbamates on octadecyl sorbents[37]

or fullerenes[38].

In this study two approaches were assessed. Conventional SPE using reversed-phase sorbents, where analytes are first retained, then eluted and subjected to defluorination. Second approach is based on conducting the chemical reaction with SBP directly on a solid sorbent, followed by elution of inorganic fluoride.

The final stage involved designing and building a flow system, which would combine all the components including preconcentration, defluorination reaction carried out directly on a sorbent bed and fluoride detection. Two different approaches were developed and evaluated: a FIA system with the constant flow of reagents and a SIA (sequential-injection analysis) system with an open reactor.

2. Experimental

2.1. Instrumentation

Potentiometric measurements were conducted using a ﬂuo- ride ion-selective electrode and an Ag/AgCl single junction reference electrode, both purchased from Thermo-Orion (Witchford, England). The electrodes were connected to an EMF-16, data acquisition system from Lawson Labs Inc.

(Malvern, PA, USA). Fluorimetric measurements were conducted using a Shimadzu (Tokyo, Japan) ﬂuorimetric detector model RF-550 with a chromato-integrator Hitachi (Tokyo, Japan) model D-2000. Solutions in the manifold with ﬂuori- metric detection were propelled using double-plunger pumps DMX-2000 from SNK (Tokyo, Japan).

Solid-phase extraction procedures were all carried out on a Vac Elut 20 Extraction Manifold obtained from Varian (Walnut Creek, CA, USA). Polypropylene frits and tubes for solid- phase extraction were obtained from Supelco (Bellefonte, PA, USA).

Minipuls 2 peristaltic pump used in the potentiometric system was obtained from Gilson (Middleton, WI, USA). Tygon silicone tubing of 1.42 mm I.D. used for pumping and PTFE tubing of 0.82 mm I.D. for connections were obtained from Cole Parmer (Chicago, IL, USA). A two-position injection valve was obtained from Rheodyne (type 7125) (Rohnert Park, CA, USA). Injection loops of 150␮L (ﬂuorimetric system) and 100␮L (potentiometric system) were used throughout the experiments.

Figs. 1 and 2present diagrams of manifolds employing ﬂu- orimetric and potentiometric detection, respectively.

Fig. 1 – Manifold of FIA system with ﬂuorimetric detector.

Measurements carried out at excitation wavelength = 422 nm and emission wavelength= 491 nm.

Fig. 2 – Manifold for FIA system with potentiometric detector. FISE—ﬂuoride-ion selective electrode,

REF—reference electrode. Carrier solution consisted of 1 M acetic buffer of pH 5.5, 200 mM NaCl and 5␮M F⁻.

2.2. Reagents

All chemicals used were of analytical grade. Sodium biphenyl reagent (irritant and highly ﬂammable!), sodium ﬂuoride, zir- conium(IV)sulfate (hydrate) and quercetin (dihydrate) were purchased from Sigma–Aldrich (Steinheim, Germany). Acetic acid, sodium hydroxide and sodium chloride used for preparation of buffer solution were obtained from Chempur (Piekary Slaskie, Poland). Methanol and acetonitrile were obtained from Merck (Darmstadt, Germany). Diethyl ether and isopropanol were acquired from POCh (Gliwice, Poland).

Hydrochloric acid was obtained from Fluka (Buchs, Switzer- land). All water solutions were made using deionized water from Milli-Q system (Millipore, Bedford, MA, USA). Argon used for sorbent drying was obtained from BOC (Warsaw, Poland).

The analytes were all perfluorocarboxylic acids (PFCAs) namely perfluorohexanoic acid (C6-PFCA) obtained from Fluka (Buchs, Switzerland), perfluorobutyric acid (C4-PFCA), per- fluorooctanoic acid (C8-PFCA, PFOA), perfluorodecanoic acid (C10-PFCA) and perfluorododecanoic acid (C12-PFCA) obtained from Aldrich Chemical (Milwaukee, WI, USA).

C18 reverse phase extraction cartridges (500 mg) were obtained from J.T. Baker (Deventer, Holland). Envi-Carb graphitized non-porous carbon (250 mg cartridges) was obtained from Supelco (Bellefonte, PA, USA). Carbon based sorbents included activated carbon purchased from Wako (Osaka, Japan), multi-wall carbon nanotubes and nanoﬁbres obtained from Sigma–Aldrich (Steinheim, Germany). Extraction car-

(4)

tridges were prepared manually by weighing 200 mg of sorbent and packing it into polypropylene extraction tubes.

Acetic buffer was prepared from glacial acetic acid and pH adjusted to 5.5 using solid sodium hydroxide. One molar buffer used in the potentiometric ﬂow system as an ionic strength adjuster additionally contained 200 mM NaCl and 5␮M ﬂuo- ride. Quercetin solution was prepared by dissolution 10 mg of quercetin in 100 mL of 1:1 water:methanol.

2.3. Procedures

Defluorination reaction was initially evaluated in a batch mode using the following procedure. One millilitre of diethyl ether standard sample of PFCA was placed in a polypropylene tube and 1 mL of SBP was added. The tube was then capped for 10 min and afterward 1 mL of deionized water was added for extraction of fluoride. The organic phase was aspirated and discarded. Approximately 60␮L of glacial acetic acid was added in order to adjust pH of the sample to about 5.5 in the presence of methyl red. The volume was brought up to 5 mL with acetic buffer. The solution was then injected into the flow system. Initial calibration of potentiometric system and blank were done using the same procedure with pure ether and with fluoride standard solution instead of 1 mL of pure water. Comparison of calibration curves obtained using the procedure and obtained using fluoride solutions prepared in acetic buffer showed no significant difference. Therefore, all following calibrations in the potentiometric system were done using standard solutions prepared by subsequent dilutions of 0.1 M fluoride solution in 500 mM acetic buffer.

Sorption of ﬂuoride on activated carbon was tested by pass- ing the buffer solution of ﬂuoride standard (50␮M) through the sorbent and then immediate measurement of collected solution.

Conventional SPE procedure with a C18 sorbent was per- formed as follows. The bed was activated with two volumes of the extraction solvent (methanol, ca. 8 mL) then conditioned with the same amount of deionized water. Next, 100 mL of solution containing mixture of standards (C6 to C12-PFCAs) each in concentration of 50␮M was introduced onto the sorbent. The bed was washed twice with 500␮L of methanol to elute the analytes. The solution was then evaporated to approximately 250␮L and brought up to 1 mL with a background electrolyte used for analysis with capillary electrophoresis (CE) in order to evaluate preconcentration efﬁciency. Procedure of CE measurement was described else- where[39].

Evaluation of sorption on carbon based sorbents (200 mg each) was carried out in a similar manner as that for the C18 sorbent. The bed was activated with 5 mL of methanol and conditioned with 5 mL of water. Then 5 mL of solution of standard mixture was introduced (C6, C8 and C12-PFCAs) in concentration of 20␮M each. Next, the bed was washed with 1 mL of water and finally 5 mL of methanol was used for elution. All fractions (from loading, washing and elution steps) were collected for analysis. Samples were first prepared by evaporation to approximately 50␮L, addition of 200␮L methanol and bringing up to 500␮L final volume with the background electrolyte for compatibility with CE determination.

Combination of preconcentration using SPE with deﬂu- orination reaction on a sorbent bed was carried out using the following procedure. First, the sorbent bed was activated with 5 mL of methanol and rinsed with 5 mL of deionized water. Next, 5 mL or 500 mL of standard solution of PFCA was introduced. Considering that SBP is very sensitive to water, the sorbent had to be dried with argon for 15 min.

After that time, 1.5 mL of SBP was added, so that the entire bed was homogenously filled and the defluorination reaction was carried out for 10 min. Then 2×2 mL of 500 mM acetic buffer pH 5.5 was passed through the sorbent in order to extract obtained fluoride. The organic phase was then discarded and the pH adjusted with glacial acetic acid in the presence of methyl red. The volume was brought up to 5 mL with acetic buffer and final solution was injected into the flow system.

3. Results and discussion

3.1. Comparison of ﬂow-injection systems for ﬂuoride detection

Two different flow-injection systems have been designed and examined for determination of inorganic fluoride released from fluorinated organic compounds.Table 1presents comparison of operational parameters obtained for systems with fluorimetric and potentiometric detection. Manifold diagrams are shown inFigs. 1 and 2. The potentiometric system is much simpler and therefore easier to optimize. Parameters like flow rate and carrier buffer composition were optimized. Flow rates between 0.6 and 2.5 mL min⁻¹per channel were tested and value of 1.1 was chosen as a compromise between peak height and signal duration (peak width). Next, the influence of buffer concentration on the peak height was tested. Chang- ing acetate concentration from 50 to 1000 mM showed minor changes in the signal magnitude, therefore the 1000 mM buffer was accepted, considering the need for high ionic strength of carrier solution. Moreover, addition of low concentration of fluoride into the carrier buffer significantly influenced the peak height and the calibration curve. Fluoride presence decreased peak height, however, it allowed faster stabilization of electrode base potential as well as much faster response.

Fluoride concentration between 0 and 5␮M was tested and 5␮M was chosen as a compromise between the peak width

Table 1 – Comparison of operational parameters of flow systems for fluoride determination using fluorimetric and potentiometric detection

Parameter Potentiometric

detection

Fluorimetric detection

LOD (␮g L⁻¹) 103 40

Precision (R.S.D., %)^a 1.9 2.8

Throughput^b(h⁻¹) 54 60

Sample loop volume (␮L) 100 150

aObtained for 10 injections of 20␮M F⁻.

bCalculated for injections of 50␮M F⁻.

(5)

and height. At higher concentration much larger decrease of the analyte peak hight can be expected, hence, it was not further studied. Also, two detection cell designs were evaluated, namely standard wall-jet assembly and Perspex home-made thin-layer cell. The latter was chosen because of its better signal dynamics as well as overall ruggedness. Disadvantages of the potentiometric detection are high sensitivity to elec- trostatic disturbances, requirement of precise degassing of ﬂowing solutions and the need of concentrated buffer as the signals are highly dependant on the ionic strength. In order to equalize the ionic strength, the sample and standards are injected into a water stream and mixed with 1 M acetic buffer.

Chelating agents usually added to the carrier solution, were not included at this stage of project mainly because of use of standard solutions, and secondly because the SPE procedure should sufﬁciently eliminate interferences caused by metal ions complexing the ﬂuoride ions.

The advantage of ﬂuorimetric detection, for which conditions were adapted from earlier work[22], is a linear relation of response versus concentration and better limit of detection observed. Drawbacks of this detection method include: substantial sensitivity to changes in ﬂow rates as well as necessity of fresh solutions preparation for every measurement cycle.

Fig. 3a and b present signal recordings obtained using systems with potentiometric and fluorimetric detection, respectively, while Table 1 shows functional parameters obtained for FIA systems with both detections. As for the system with potentiometric detection a signal for injection of blank was observed, in this case LOD was determined as signal corresponding to triple value of standard deviation of blank signal divided by the slope of calibration plot. In case of fluorimetric detection where no blank response was observed, LOD was calculated as the concentration of fluoride corresponding to S/N = 3, where S is signal magnitude and N amplitude of the peak-to-peak noise.

3.2. Optimization of deﬂuorination reaction with the SBP

Parameters of the defluorination reaction were examined in the batch mode, carried out using SBP and standard solutions of perfluorinated substances (C6, C8, C10 and C12-PFCAs at concentrations 7.2 mg L⁻¹, 7.2 mg L⁻¹, 7.8 mg L⁻¹, 7.1 mg L⁻¹, respectively). Taking into account reasonable throughput of future flow system, one of the optimized parameters was the reaction time. Therefore, the influence of reaction time on fluoride recovery was assessed. Reactions were carried out for 10, 7, 5 and 3 min, using solution of C8-PFCA. Results showed that a 3-min reaction produced not lesser amount of fluoride than that at a longer reaction time. The observed differences in signals a–l inFig. 3a can be then considered as insignifi- cant, within the precision of the measurement. Ratio of SBP to sample volume was also evaluated. Increasing SBP to sample ratio from 1:1 to 2:1 did not have any significant effect on the reaction efficiency.

A study was also carried out for PFCAs with various length of the carbon chain in the PFCA molecule. Again, no consider- able difference in the ﬂuoride recoveries between tested PFCAs were found.

Fig. 3 – (a) Signal recordings obtained using a FIA system with potentiometric detection. Calibration and sample signals obtained after a batch procedure of deﬂuorination of PFCAs with different chain length using 10 min reaction time (samples a–h) and different reaction times for C8-PFCA (samples i–l); (B) blank, samples a, e—C6; b, f—C8; c, g—C10; d, h—C12; i: 3 min; j: 5 min; k: 7 min; l: 10 min. (b) Signal recordings of calibration obtained using FIA system with ﬂuorimetric detection.

3.3. Preconcentration of PFCAs on different sorbents

Several commercially available sorbents were tested for the preconcentration capability and then also for possibility of performing the deﬂuorination reaction directly on a sorbent bed. Initially, conventional C18 reverse phase sorbent was tested. However, because shorter chain PFCAs, especially C6- PFCA, were poorly retained on 250 mg of sorbent, quantitative retention from aqueous samples and then consequently the recovery was not possible. Similar observations were made by other authors[40,41].

Interaction between C18-sorbent and SBP was tested by visually observing the sorbent bed and cartridge weighing.

No visible reaction took place and only 0.1% of weight loss between fresh cartridge and after contact with SBP was noted.

Nevertheless, because of very small particle size, SBP would not penetrate entire volume of the sorbent bed, even when substantial vacuum was applied.

(6)

Table 2 – Sorption efﬁciency and recovery (%) for PFCAs on carbon based sorbents, using methanol as eluate

Sorbent Analyte

C6-PFCA C8-PFCA C12-PFCA

Retained Eluted Retained Eluted Retained Eluted

MWCNT 100 21.2 100 0 100 0

Carbon nanoﬁbres 100 34.8 100 53.9 100 40.8

Envi-Carb 100 0 100 0 100 0

MWCNT: multi-wall carbon nanotubes; Envi-Carb: commercial cartridges with graphitized carbon.

Considering chemical inertness and high affinity toward organic matter, carbon based sorbents were then tested. Multi- wall carbon nanotubes (MWCNT), carbon nanofibres as well as commercial sorbent Envi-Carb were investigated.Table 2 shows recoveries obtained for three PFCAs using tested sorbents with methanol as eluent and CE detection. None of the analytes was detected in the aqueous loading or washing solutions and only partial recovery was achieved using elution with methanol. Employing other organic solvents such as isopropanol and acetonitrile did not improve quantitative recovery of analytes. This clearly indicates that carbon based materials retain PFCAs very effectively. Moreover, it allows washing carbon sorbents with methanol solution without significant analytes loss, what is beneficial considering the need of sorbent drying before reaction with SBP.

3.4. On-sorbent deﬂuorination procedure

In order to simplify the entire procedure of TOF determination, an attempt was made to carry out the defluorination reaction directly on the sorbent bed. Preliminary experiments of conducting defluorination directly on a solid sorbent included all tested carbon-based materials, however, because of very small grains of nanotubes, nanofibres and Envi-Carb beds and therefore difficulties in homogenous penetration by viscous SBP, granulated (mesh 40–230) activated carbon from Wako was finally chosen and used throughout the experiments. Fluoride sorption tests showed that 92.7–93.6% of inorganic fluoride passed through the sorbent when loaded as aqueous solutions.

Table 3presents ﬂuoride recoveries obtained in the deﬂu- orination of C4, C6, C8, C10 and C12 PFCAs. Preconcentrations were made from 5 mL and 500 mL of aqueous sample. Con- centrations were 4.48␮M, 3.62␮M, 3.39␮M, 2.60␮M for C6, C8, C10, C12-PFCA, respectively preconcentrated from a 5 mL

sample volume and 44.8 nM, 36.2 nM, 25.2 nM, 14.7 nM for C6, C8, C10 and C12-PFCA, respectively preconcentrated from a 500 mL sample volume. For samples preconcentrated from a 500 mL sample volume, concentrations of PFCAs expressed in␮g L⁻¹are: 14.0, 15.0, 12.9, 9.03 for C6, C8, C10, C12-PFCA, respectively. Recoveries varied from 74% to 144% with mean and standard deviation values shown inTable 3. Relatively low precision, especially for longer chain acids, is most proba- bly due to incomplete deﬂuorination of these analytes, when absorbed on the sorbent bed, as well as inherent difﬁculties with extraction repeatability.

3.5. FIA system with on-line deﬂuorination reaction

Having a method to release fluorine from an organic com- pound and a tool to determine the released fluoride, two concepts of combining these steps in a flow-injection system with on-line defluorination were investigated.

The first approach was a simple FIA system, where injection of ether sample solution (e.g. eluent from preconcentration column) without on-line preconcentration step was projected. Its advantages would be simplicity and continuous flow of reagents. Defluorination reaction was designed to take place in a reaction coil in conditions of constant flow.Fig. 4 shows a diagram of the test manifold, where injection valve is placed in the buffer line for testing purposes. The main fea- ture of this system was an in-flow phase separator, necessary for undisturbed functioning of the potentiometric detector, incompatible with organic solvents. Construction of the separator required small volume in order not to increase system dispersion. On the other hand, very small volume required precise control over flow rates of organic and aqueous phases in order to maintain constant ratio of phases in the separator.

The phase-separator was made of two Perspex blocks, glued together, where on one of them a gravity separation compart-

Table 3 – Recovery of ﬂuoride obtained in the deﬂuorination reaction of PFCAs on carbon based sorbents

Sorbent C4 C6 C8 C10 C12

MWCNT – – 26±16^b – –

Envi-Carb – – 67±14^b – –

Cact 99.5±1.7^a 110±9.4^b 94.4±3.4^a 120±32^b 110±12^b

95±26^b

MWCNT: multi-wall carbon nanotubes.

a Results obtained using ﬂuorimetric detection.

b Results obtained using potentiometric detection.

(7)

Fig. 4 – FIA system design with the use of on-line deﬂuorination reaction with SBP, phase separator and potentiometric detection.

ment was engraved (seeFig. 4) with an inlet for mixture of organic/aqueos solution and two outlets for separated solutions. Satisfactory precision of flow rate control, required for operation of such a system, was very difficult using a standard peristaltic pump. The result of using a standard peristaltic pump was either a loss of sample in organic waste or inflow of organic solvent into the detection channel, producing unex- pected behavior of the measured potential.

Another problem, which arose during the system testing was formation of a precipitate (GC–MS analysis indicated non- aromatic derivatives of biphenyl) upon contact of SBP with aqueous solution. The phenomenon was not observed in the batch-mode, where water was added to 1:1 sample (ether):SBP solution. The adsorption of precipitate inside of the system caused changes in ﬂow rates and intermittently clogged the system (see example recording of signal from the system inFig. 5with injection of 1 mM ﬂuoride standard solution).

Additional problem occurred in the organic solvent pumping channel. Standard silicone tubing looses ﬂexibility and becomes stiff in contact with diethyl ether. Rubber tubing, which was used instead, although remained elastic, was wear- ing out relatively fast.

In spite of simple construction and advantageous constant ﬂow of reagents, the drawbacks of the FIA system made it difﬁcult to operate and caused high maintenance with little promise of a success as a rugged and routine system.

3.6. SIA system with an open reactor

The second approach was based on SIA system incorporating two three-way valves and one multiposition valve and a persi- taltic pump, controlled using a computer. The manifold used for tests is presented inFig. 6. The most important part of the system is an open reactor that allows avoiding the reaction using SBP inside of a tubing and consequently deposition of the precipitate. The characteristics of the SIA procedure per- mits straightforward addition of a washing step necessary for the reactor cleaning. The reactor can also serve as an effec- tive phase separator and it is much easier to integrate with an SPE cartridge. Although the construction of the SIA system is more complicated than the FIA system, the former is easy to operate considering computer based control.

Fig. 5 – Example of FIA signal obtained in the system with on-line deﬂuorination and phase separator. Peaks a–f correspond to injections of 100␮L of aqueous 1 mM ﬂuoride standard solution. Parts of recordings marked with

asterisks correspond to malfunction of the system during continuous measurements because of precipitate appearance or inﬂow of ether to detection channel.

Regardless of all the advantages, this SIA system pos- sesses a critical hindrance concerning SBP transport to the reactor using common holding coil. If some water droplets were left in the holding coil from previous cycle, SBP was immediately deactivated. Rinsing with ether was not efﬁcient enough, whereas water miscible solvents such as methanol also reacted with SBP. This essentially means that a redesign of the system is necessary, where additional transporting channel for SBP would be included.

(8)

Fig. 6 – SIA system with the on-line deﬂuorination reaction in an open reactor and potentiometric detection.

4. Conclusions

This project goal was to employ a flow system in determination of TOF using SPE and the defluorination reaction using SBP reagent. This was partly achieved by a novel approach of on-sorbent defluorination using carbon based sorbents, which allowed very efficient preconcentration of analytes, did not interact with SBP, allowed reaching low detection limits and simplified the entire procedure. Acceptable recoveries of fluoride from different PFCAs were demonstrated. It should be noted, however, that currently efforts are directed to overcome relatively low precision of the procedure.

Troublesome requirement of non-aqueous reaction conditions as well as precipitate formation rendered the use of a constant ﬂow system impractical and impeded SBP transport in SIA mode. Nonetheless, SIA system with an open reactor seems to be the most suitable strategy to follow.

Although relatively high concentrations of model perfluorinated compounds were used (␮g L⁻¹ level), lower concentrations are accessible not only from the fluoride detection point of view, but also by increasing the preconcentrated sample volume. The future developments will focus on improvement of control over fluoride recoveries from longer chain acids, further validation of procedure and TOF determination in the environmental (water and air) samples. Finally, the continuation of SIA system construction including the implementation of SPE reactor and the transporting channel for SPB is planned.

r e f e r e n c e s

[1] B.D. Key, R.D. Howell, C.S. Criddle, Environ. Sci. Technol. 31 (1997) 2445.

[2] G.W. Gribble, J. Nat. Prod. 55 (1992) 1353.

[3] D.B. Harper, D. O’Hagan, Nat. Prod. Rep. 11 (1994) 123.

[4] K. Prevedouros, I.T. Cousins, R.C. Buck, S.H. Korzeniowski, Environ. Sci. Technol. 40 (2006) 32.

[5] S. Taniyasu, K. Kannan, M. Ka So, A. Gulkowska, E. Sinclair, T. Okazawa, N. Yamashita, J. Chormatogr. A. 1093 (2005) 89.

[6] J.P. Giesy, K. Kannan, Environ. Sci. Technol. 35 (2001) 1139.

[7] Z. Kuklenyik, J.A. Reich, J.S. Tully, L.L. Needham, A.M.

Calafat, Environ. Sci. Technol. 38 (2004) 3698.

[8] E. Kissa, Fluorinated Surfactants and Repellants, second ed., Marcel Dekker, New York, 2001.

[9] B.D. Key, R.D. Howell, C.S. Criddle, Environ. Sci. Technol. 32 (1998) 2283.

[10] L. Singer, R.H. Ophaug, Anal. Chem. 49 (1977) 38.

[11] G.J. Kakabadse, J.B. Manohin, M. Bather, E.C. Weller, Nature 229 (1971) 626.

[12] P. Venkateswarlu, Adv. Dent. Res. 8 (1994) 80.

[13] P. Venkateswarlu, Anal. Biochem. 68 (1975) 512.

[14] J. Belisle, D.F. Hagen, Anal. Biochem. 87 (1987) 545.

[15] P.B. Sweester, Anal. Chem. 28 (1956) 1766.

[16] V. ˇC ´apka, C.P. Bowers, J.N. Narvesen, R.F. Rossi, Talanta 64 (2004) 869.

[17] P. Venkateswarlu, Anal. Chem. 54 (1982) 1132.

[18] M.S. Frant, J.W. Ross Jr., Science 154 (1966) 1553.

[19] K. Itai, H. Tsunoda, Clin. Chim. Acta 308 (2001) 163.

[20] J.F. van Staden, R.I. Stefan, Talanta 49 (1999) 1017.

[21] Orion Ion-Plus Fluoride Electrode Instruction Manual, Thermo Electron Co., 2003.

[22] M. Garrido, A.G. Lista, M. Palomeque, B.S. Fern ´andez Band, Talanta 58 (2002) 849.

[23] V. Marco, F. Carrillo, C. Per ´ez-Conde, C. C ´amara, Anal. Chim.

Acta 283 (1993) 489.

[24] Z. Song, N. Zhang, Spectrosc. Lett. 36 (2003) 117.

[25] D.A. Barnett, G. Horlick, J. Anal. At. Spectrom. 12 (1997) 497.

[26] R.C. Wiliams, R. Boucher, J. Brown, J.R. Scull, J. Walker, D.

Paolini, J. Phram. Biomed. Anal. 16 (1997) 469.

[27] P. Wang, S.F.Y. Li, H.K. Lee, J. Chromatogr. A. 765 (1997) 353.

[28] Th. Ehmann, K. B ¨achmann, L. Fabry, H. R ¨ufer, S. Pahlke, L.

Kotz, Chromatographia 45 (1997) 301.

[29] P.L. Buldini, A. Mevoli, J.L. Sharma, Analyst 123 (1997) 1109.

[30] P. Jones, Anal. Chim. Acta 258 (1992) 123.

[31] K. Tsunoda, K. Fujiwara, K. Fuwa, Anal. Chem. 49 (1977) 2035.

[32] J.A. Fresen, J.H. Cox, M.J. Witter, Pharm. Weekblad. 103 (1968) 909.

[33] H. Sakurai, M. Fukuda, Y. Hayashibe, Y. Sayama, K.

Masumoto, T. Ohtsuki, J. Radioanal, Nucl. Chem. 217 (1997) 267.

[34] T.P. Cheng, H.D. Anderson, D.S. Mills, V.L. Spate, C.K. Baskett, J.S. Morris, J. Radioanal. Nucl. Chem. 217 (1997) 171.

[35] K. Wagner, W. Gomer, M. Hedrich, P. Jost, Chr. Segebade, Fresenius J. Anal. Chem. 362 (1998) 382.

[36] Y. Daghbouche, S. Garrigues, M. de la Guardia, Anal. Chim.

Acta 314 (1995) 203.

[37] J. Ruzicka, A. Arndal, Anal. Chim. Acta 216 (1989) 243.

[38] M. Gallego, Y.P. de Pena, M. Valcarcel, Anal. Chem. 66 (1994) 4074.

[39] L. W ´ojcik, B. Szostek, W. Maruszak, M. Trojanowicz, Electrophoresis 26 (2005) 1080.

[40] C.-L. Tseng, L.-L. Liu, C.-M. Chen, W.-H. Ding, J. Chromatogr.

A. 1105 (2006) 119.

[41] C.A. Moody, W.C. Kwan, J.W. Martin, D.C.G. Muir, S.A.

Mabury, Anal. Chem. 73 (2001) 2200.