Talanta 152 (2016) 112–118 Contents lists available at ScienceDirect Talanta http://d 0039-91 n Corr lès-Nan E-m jpinheiu journal homepage: www.elsevier.com/locate/talanta Towards field trace metal speciation using electroanalytical techniques and tangential ultrafiltration Adnívia Santos Costa Monteiro a, Corinne Parat b, André Henrique Rosa a, José Paulo Pinheiro c,d,n a UNESP, Departamento de Engenharia Ambiental, Sorocaba, SP, Brasil b Université de Pau et des Pays de l’Adour, CNRS UMR 5254, LCABIE, 64000 Pau, France c CNRS, LIEC, UMR 7360, Vandoeuvre-lès-Nancy F-54501, France d Université de Lorraine, LIEC, UMR 7360, Vandoeuvre-lès-Nancy F-54501, France a r t i c l e i n f o Article history: Received 14 October 2015 Received in revised form 22 January 2016 Accepted 25 January 2016 Available online 26 January 2016 Keywords: Trace metal speciation Cadmium Tangential ultrafiltration Stripping chronopotentiometry AGNES On-site analysis x.doi.org/10.1016/j.talanta.2016.01.053 40/& 2016 Elsevier B.V. All rights reserved. esponding author at: Université de Lorraine, L cy F-54501, France. ail addresses: jose-paulo.pinheiro@univ-lorrai algpt@gmail.com (J.P. Pinheiro). a b s t r a c t In this work we propose a trace metal speciation methodology to determine the total, free and ultra- filtered (o1 KDa) metal fractions using electrochemical methods (SCP and AGNES) and tangential ul- trafiltration (UF) experiments that can easily be carried out on-site. We tested our methodology spiking Cadmium ions into two natural waters samples from Itapanhau and Sorocabinha rivers in Sao Paulo State, Brazil. The limits of detection (LOD) was 1.6�10�9 M for the total Cd2þ determination performed by Stripping Chronopotentiometry (SCP) in the source and acidified ultrafiltered solution and 1.9�10�9 M for the free Cd2þ determination using Absence of gradients and Nernstian equilibrium stripping (AGNES), using a thin mercury film electrode. The total metal determination was performed by SCP in acidified samples and the results compared with graphite furnace atomic absorption spectroscopy (GF-AAS). The SCP results were adequate with a 96% of recovery from the known metal spike for the 12 samples tested. For the Itapanhau sample the free metal determined by AGNES and the ultrafiltered fraction are identical, while for the Sorocabinha the free metal in the source is significantly smaller than the ultrafiltered fraction, indicating that this sample must be rich in metal complexes with small in- organic ligands that are able to permeate the 1 kDa membrane. The proposed metal speciation metho- dology validated in the laboratory combining UF and SCP/AGNES is able to be used in on-site experiments providing valid information regarding the total and free metal concentrations and additionally some insight on the role of small inorganic ligands to the metal complexation. & 2016 Elsevier B.V. All rights reserved. 1. Introduction The interest of studying trace metal ions in natural waters stems from their toxicity towards microorganisms and accumu- lation through the food chain. Toxic metallic elements, like Pb, Cd, Zn and Hg are found in trace amounts in environmental waters mostly as a result from anthropogenic activity. Their impact on the environment depends on their concentration and speciation since metals may be present as hydrated cations, dissolved molecular species or as parts of complex mineral particles. The dissolved fraction has a much higher bioavailability than the particulate one, thus a higher toxicity. Currently the conventional notion in ecotoxicological studies is that the controlling species regarding trace metal bioavailability is IEC, UMR 7360, Vandoeuvre- ne.fr, the free hydrated cation (FIAM, BLM) [1]. Since the trace amounts of toxic metal present in natural water are difficult to measure; therefore the free cations are even more challenging, since their concentration often lies on the nanomolar range or below. Natural waters have a plethora of ligands able to bind the metal ions, like dissolved inorganic and organic ligands, colloids, suspended matter and the surfaces of organisms, hence only a small part of the total dissolved metal exists as free hydrated cations [2]. Environmental trace metal analysis is not a simple task due to the complexity of the environmental matrices. Additionally the collection and handling of natural waters samples can further in- duce changes in the distribution of species prior to laboratory measurements, thus on-site measurements are preferable since they avoid problems due to contamination during sampling and storage, metal adsorption to container walls and instability of metal species during transport and storage prior to measurements. Over the years an extensive effort has been made to design chemical sensors for on-site measurements and there are several books on chemical sensors for aquatic measurements that describe www.sciencedirect.com/science/journal/00399140 www.elsevier.com/locate/talanta http://dx.doi.org/10.1016/j.talanta.2016.01.053 http://dx.doi.org/10.1016/j.talanta.2016.01.053 http://dx.doi.org/10.1016/j.talanta.2016.01.053 http://crossmark.crossref.org/dialog/?doi=10.1016/j.talanta.2016.01.053&domain=pdf http://crossmark.crossref.org/dialog/?doi=10.1016/j.talanta.2016.01.053&domain=pdf http://crossmark.crossref.org/dialog/?doi=10.1016/j.talanta.2016.01.053&domain=pdf mailto:jose-paulo.pinheiro@univ-lorraine.fr mailto:jpinheiualgpt@gmail.com http://dx.doi.org/10.1016/j.talanta.2016.01.053 A.S.C. Monteiro et al. / Talanta 152 (2016) 112–118 113 various continuous flow analyzers as well as optical and electro- chemical sensors [3]. Nevertheless these methods focus mostly on the determination of total metal; hence at this point there are very few affordable on-site systems for free trace metal quantification at the concentration levels present in natural waters. In terms of portability for on-site trace metal detection system, electrochemical devices are the most advantageous due to their small size and low energy demand. Stripping electrochemical methods coupled with the correct electrode/flow systems present some of the lowest detection limits available for metal determi- nations, often as low as a few nmol L�1 [4]. Amongst the stripping electrochemical techniques one of the most suited for environmental analysis is Stripping Chron- opotentiometry (SCP) [5] especially because the signal is not af- fected by adsorption of organic species at the surface of the elec- trode [6]. SCP is a two-step technique composed of a deposition step, where metal ions are reduced at a constant potential and a chosen fixed time, followed by a stripping step, where the amalgamated metal is reoxidized through the application of a constant oxidizing current. When a thin mercury film electrode is used in SCP, the rapid metal transport inside the thin film during stripping means that measurements are performed almost always under conditions of complete depletion which was explored by Parat et al. [7] to use much higher stripping currents and therefore remove the oxygen competition in this step. AGNES (Absence of Gradients and Nernstian Equilibrium Stripping technique) is an electrochemical technique that mea- sures directly the free metal ion concentration [8]. In the deposi- tion step the amalgamation of the metal in the mercury electrode is allowed to continue until the equilibrium value between oxi- dized and reduced forms in solution and in the mercury electrode is reached. This equilibrium is set by the deposition potential and follows Nernst equation. When this situation is achieved no con- centration gradients will be present, either in solution or in the mercury amalgam and the amount of amalgamated metal will be proportional to the free metal ion concentration in the bulk so- lution. The main disadvantage of AGNES is the waiting time in the deposition step needed to attain equilibrium, however this might be mitigated by using a thin mercury film electrode to benefit from its larger area to volume ratio [9]. Worth mentioning is also the change in the stripping mode from a simple Cottrell decay to a stripping chronopotentiometry step [10]. Another commonly used approach to study the speciation of metals in natural waters is the fractionation of the sample based on tangential ultrafiltration. The procedure is generally carried out with a low pore size membrane (e.g. 1 kDa) to distinguish the free species from the colloidal ones. The permeate is generally defined as the free fraction (FF) although the metal cations here may also be complexed with small organic or inorganic ligands that are small enough to pass the membrane [11,,12]. It is interesting to compare the free metal ions and the small species able to permeate through the 1 kDa membrane which ef- fectively constitute the most mobile metal fraction in natural waters, thus in this work we propose a methodology to determine the total, free and ultrafiltered (o1 kDa) metal fractions using electrochemical methods and tangential ultrafiltration experi- ments that could easily be carried out on-site. In this work, this methodology was tested in laboratory by spiking cadmium ions into two natural waters samples from Itapanhau and Sorocabinha rivers in Sao Paulo State, Brazil and the results obtained by the electrochemical techniques were compared with GF-AAS in the source waters and ultrafiltered samples for validation of the methodology. 2. Materials and methods 2.1. Reagents Chemicals used in this work were of analytical reagent grade and used as received, unless stated otherwise. Ultrapure milli-Q water (resistivity 18.2 MΩ cm) was employed in all experiments. Nitrogen (499.99% pure) for the electrochemical experiments and argon (499.99% pure) for the atomic absorption experiments were provided by White Martins. Standard pH 4.0070.02 and pH 7.070.2 solutions were obtained from MS Tecnopon In- strumentação, Brasil. Sterile 0.45 mm and 47 mm diameter acetate of cellulose membranes were used to filtrate the samples. KH2P04, Na2CO3 and NaHCO3 all 1000 mg L�1 were obtained from Merck. Nitric acid 65% (p.a) and phosphoric acid 85% (p.a) were purchased from Merck. The standard stock solutions of mercury nitrate (100574 mg L�1), cadmium nitrate (100074 mg L�1), and multi element standard solution G16V ICP (100 mg L�1, NIST-USA rastreable) were purchased from Specsol. Cd(II) solution was prepared from dilution of the certified standard using milli-Q water. Palladium and Magnesium matrix modifiers (10.070.2 g L�1) used in GFAAS measurements were purchased from Merck. Sodium nitrate electrolyte solutions (0.1 M) and MES (2-(N-morpholino)ethanesulfonic acid) buffer (0.2 mol L�1) were prepared from the solids (Merck). Solutions prepared from nitric acid and sodium hydroxides (0.1 M standard, Merck) were used to adjust the pH when necessary. Ammonium acetate, hydrochloric acid, potassium thiocyanide, potassium chloride (all p.a. from Merck) was used to prepare the solution for the cleaning of the working electrode and re-dissolution of the mercury film. The working electrode was screen-printed using a carbon commercial ink Electrodag s PF-407A. Tangential ultrafiltration was performed using a regenerated cellulose acetate membrane of 1 kDa and 76 mm diameter (Millipore, Billerica, U.S.A). 2.2. Equipment An Ecochemie Autolab PGSTAT10 potentiostat (controlled by GPES 4.9 software) was used in conjunction with a Metrohm 663VA stand (Metrohm, Switzerland). A three electrode config- uration was used comprising a Hg thin film plated onto a home- made screen-printed electrode as the working electrode, a glassy carbon rod counter electrode (6.1247.000), and an Ag/AgCl re- ference electrode (6.0728.020) with a salt bridge (6.1245.010) all from Metrohm. A Marconi Instrument (model MA-522) and an Analyzer (model 2A14) combination pH electrode, calibrated with MS tec- nopon buffers (Merck) were used to measure pH in the laboratory. A Teflon home-made tangential ultrafiltration device was built according to Burba et al. [12] and used in conjunction with a peristaltic pump, Gilson Minipuls 3 with Agilent tygon tubing. The atomic absorption experiments were carried out in a Var- ian Zeeman Atomic Absorption Spectrometer model AA240z with a PSD 120 auto sampler, Zeeman correction and graphite furnace with integrated L’vov platform. The hollow cathode lamp was obtained from Agilent (λ¼228,8 nm (Cd) e i¼10 mA), and the argon flow used was 300 mL min�1. The signal was measured in peak integration mode. An inductively coupled plasma atomic emission spectrometer (ICP-OES) model 700 Agilent Technologies was used to quantify the following elements: Al, Co, Cr, Cu, Fe, Mn, Ni, and Zn. Dissolved organic carbon was determined using an Analytik Jena Carbon Analyzer (model N/C 3100). The field measurements of pH, con- ductivity and turbidity were performed using a Hanna multi- parameter probe model HI9829. UV/vis spectroscopic measures were performed in a spectrophotometer Varian model Cary 50. Fig. 1. Map of the state of São Paulo indicating the collection points in the Sorocabinha and Itapanhau rivers. A.S.C. Monteiro et al. / Talanta 152 (2016) 112–118114 2.3. Natural water collection and characterization experiments The natural water samples were collected in the Itapanhaú and Sorocabinha Rivers located in São Paulo state in Brazil (Fig. 1) in November 2014. The geographical coordinates of the sampling points are: latitude 24 °41’59” and longitude 47°35’43.68” for the Itapanhaú river and latitude 23°50’23” and longitude 46 °08’21” for the Sorocabinha river [16]. The samples were collected from the surface layer using poly- ethylene flasks which were rinsed thoroughly with the water sample before sampling. The samples were stored at 4 °C until the analysis. The pH, the conductivity and the dissolved oxygen were measured in situ. Roughly 10 L of water sample were taken to the laboratory where DOC, sulfate, chloride and dissolved metal were analyzed within 48 h. Details of ICP-OES analysis, Dissolved organic carbon (DOC) and UV/vis Spectroscopy experiments are given in the electronic sup- port information. 2.4. Graphite furnace atomic spectroscopy (GF-AAS) experiments The pyrolysis temperature and atomization temperature for the Cd measurements in the current samples were 700 and 1600 °C. These values resulted from an optimization procedure using 10 μg Cd2þ L�1 (prepared from dilution of the standard solution 1000 mg L�1) in HNO3 2% (v/v) in presence of the chemical modifiers 5 μL Pd(NO3)2þ3 μL Mg(NO3)2. The calibration plot was carried out for Cd2þ concentrations of 0.89; 1.78; 4.45; 7.12 and 8.90�10�8 mol L�1. All measurements were performed in tripli- cate and when necessary the samples were diluted using milli-Q water. The obtained LOD and LOQ were respectively 0.6 mg L�1 (0.5�10�8 mol L�1) and 1.9 mg L�1 (1.7�10�8 mol L�1). 2.5. Tangential ultrafiltration experiments Tangential ultrafiltration experiments were carried out using the collected samples from the rivers Itapanhau and Sorocabinha at the natural pH and ionic strength under constant stirring and pH monitoring. All measurements were carried out at room tem- perature (23–26 °C). Each experiment started with 1 L of sample water (source) and the system was equilibrated for 15 min for membrane condition- ing using a speed of 6.0 rpm. The valve controlling the return flow is then set at 1 mL min�1 while the exit valve of the ultrafiltration device was set at a flow of approximately 0.3 mL min�1. At this point a 25 mL aliquot was collected from the source and the col- lection of first ultrafiltered aliquot (also 25 mL) was initiated. After the first permeate is collected, four additions of standard solution of Cd(NO3)2 8.9�10�4 mol L�1 were carried out in order to obtain a total concentration of Cd(II) of 0.7, 1.5, 2.4 and 3.2�10�7 mol L�1 in the source. Following each addition the source pH is adjusted back to the original value using NaOH 1.0 mol L�1 due to the small variation induced by the acid present in the standard solution of Cd(II). The tests performed on the system showed that a waiting period of 20 min is necessary to achieve equilibrium prior to the collection of the 25 mL of sample. At this point the collection of ultrafiltered aliquots is initiated with 5 mL destined for GF-AAS measurements and 20 mL for electroanalytical quantification. All metal additions were computed taking into account the volume variation of the source solution. For the total metal determination using SCP in both, the source and ultrafiltered aliquots, the 20 mL of solution were acidified with 0.2 mL de HNO3 1.0 mol L�1 and subsequently named Source acidified and UF acidified. Table 1 Sorocabinha and Itapanhau rivers characterization. Itapanhau Sorocabinha pH 6.2 6.7 Conductivity (mS cm�1) 189 1043 DOC (mg L�1) 14.0 16.8 Turbidity (NTU) 3.7 6.1 Dissolved oxygen (mg O L�1) 1.8 1.4 Cl� (mol L�1) 3.6� 10�4 8.5�10�4 SO4 2� (mol L�1) 5.0� 10�6 2.2�10�6 Al (mol L�1) 4.9� 10�6 1.5�10�6 Ca (mol L�1) 4.2� 10�5 3.0�10�4 Cd (mol L�1) o LOD o LOD Co (mol L�1) o LOD o LOD Cr (mol L�1) o LOD 6.5�10�9 Cu (mol L�1) o LOD 3.7�10�8 Fe (mol L�1) 13.0�10�6 6.1�10�6 K (mol L�1) 5.5� 10�5 2.1�10�4 Mg (mol L�1) 6.7� 10�5 8.4�10�4 Mn (mol L�1) 1.3� 10�6 1.9�10�6 Ni (mol L�1) o LOD o LOD Pb (mol L�1) o LOD o LOD Zn (mol L�1) 2.3� 10�8 2.0�10�7 A.S.C. Monteiro et al. / Talanta 152 (2016) 112–118 115 2.6. Working electrode preparation The working electrodes were screen-printed on 1 mm-thick polystyrene support for serigraphy (Sericol) using inks purchased from Acheson Colloids. The electrode was screen-printed and al- lowed to dry (1 h at room temperature) and cure (1 h in the oven at 60 °C). After completion of the screen-printing process, an in- sulating layer was spread manually over the conductive tracks leaving a working disk area of 9.6 mm2 [10]. Prior to deposition of the Hg film the screen-printed carbon electrode was conditioned using an electrochemical pre-treatment consisting on fifty successive cyclic voltammetry scans between �0.8 and þ0.8 V at 0.1 V s�1 in NH4CH3COO (1 mol L�1)/HCl (0.5 mol L�1) solution [13]. The thin Hg film was prepared ex-situ in 0.24 mmol L�1 Hg(II) nitrate in nitric acid 0.73 mmol L�1 (pH 1.9) by electrodeposition at �1.3 V for 700 s using a rotation rate of the stirrer of 1000 rpm. After each working day the charge as- sociated with the deposited Hg (QHg) was determined to assess the state of the mercury film. This was carried out by electronic in- tegration of the linear sweep stripping peak of Hg with a scan rate v¼0.005 V s�1 in 5 mmol L�1 of ammonium thiocyanide (pH 3.4) using a stripping step from �0.15 V to þ0.4 V [9]. 2.7. Electrochemical experiments The measurements were performed in triplicate in the la- boratory. The solution was purged for 15 min at the beginning of every experiment and the electrochemical cell was continuously purged over the solution with nitrogen gas during the measure- ment. All measurements were carried out at room temperature (23–26 °C). The experiments were carried out in 40 mL disposable polystyrene (PS) beakers. A three or four point calibration is daily performed prior to the sample measurements using known metal concentrations in the interval between 3�10�9 and 3�10�7 mol L�1. The limits of detection (LOD) and quantification (LOQ) were estimated from the calibrations curve using the following equation: = ( )kS mLOX / 1y where k¼3 for the LOD and 10 for the LOQ, b is the intercept of the regression line, Sy is the standard deviation of the residuals and m is the slope of the calibration graph. 2.8. Stripping chronopotentiometry experiments in acid media (total metal in the source acidified and UF acidified) Stripping chronopotentiometry measurements were carried out in 20 mL acidified source solutions of Cd(II). The experimental conditions used were: deposition potential �0.78 V (vs Ag/AgCl), deposition time (td) 45 s, stripping current (Is) 2.5�10�6 A ap- plied until the potential reached �0.30 V, rotation speed 1000 rpm (only during the deposition step). The SCP signal is the characteristic transition time τ* which is given by [5]: τ α *= * = * ( ) I t I c t I 2 d d s M,T d s where *Id , the limiting diffusion current, is directly proportional to the metal concentration in the bulk *cM,T. 2.9. AGNES experiments (free metal in the source) Previous to the measurements, the solutions were set at the desired pH using MES (2-N-morpholino-ethanesulfonic acid) buffer. According to Soares et al. [14] MES does not bind Cd(II). In the case of the Sorocabinha we used the natural ionic strength but for the Itapanhau water sample, the ionic strength was set at 0.010 mol L�1 adding NaNO3 and following the conductivity of the solution. This is necessary since a sufficiently high ionic strength (Z10 mmol L�1) and a well buffered system are required to ob- tain accurate results in AGNES [15]. The free metal ion concentration was determined by the AGNES method according to the procedure developed by Parat et al. [10] where the quantification of the metal amalgamated in the equili- brium step is quantified using a chronopotentiometric measure- ment. The measurements were performed applying a deposition potential E1 of �0.72 V for a period of time t1 of 240 s. The stripping step was performed by SCP using an oxidizing current Is of 2.5�10�6 A applied until the potential reached �0.30 V. The rotation speed used was 1000 rpm. In this implementation the AGNES signal is the charge corre- sponding to the amalgamated metal in the mercury electrode Q which is given by: τ= * = * ( )Q I hc 3s M where τ* is the reoxidation time, Is is the reoxidation current ap- plied in the stripping step, h is a constant and *cM is the free metal concentration in the bulk. 3. Results and discussion The proposed methodology for on-site analysis was tested in the laboratory by spiking Cd(II) ions into two natural waters samples from Itapanhau and Sorocabinha rivers in Sao Paulo State- Brazil. These rivers were chosen due to their different character- istics especially regarding conductivity, since the Sorocabinha sampling is relatively close to the sea.. 3.1. Natural river waters characterization Table 1 shows the major parameters that allow us to char- acterize both rivers. Their pH and DOC are quite close, while the conductivity of the Sorocabinha river is much higher than the Itapanhau one due to its proximity to the ocean. This is also re- flected in the higher contents of major cations like Kþ , Mg2þ and Ca2þ and anions like Cl� . Cadmium, Pb, Co and Ni are below the detection limit in both rivers, a small amount of both Cu and Cr Table 2 Ratio A254/A436 for the Itapanhaú and Sorocabinha water samples. Sorocabinha Itapanhaú Absorbance 254 nm 0.81 0.79 436 nm 0.07 0.07 Ratio A254/A436 11.9 11.3 A.S.C. Monteiro et al. / Talanta 152 (2016) 112–118116 appears in the Sorocabinha, while Zn appears in both. Table 2 reports the UV/vis spectroscopic ratio (A254/A436), i.e. the absorption values taken at 254 and 436 nm in the UV/vis spectra, which is a semi-quantitative method to characterize hu- mic substances in natural waters. This ratio is used as an indicator of the origin and the modifications of natural organic matter in aqueous systems. For instance, absorbance at 254 nm is typical for aromatic groups. Due to the absorption of visible light by humic acids, this ratio rises with increasing contents of humic acids in solution. These two samples showed very similar values, which means that the origin of their dissolved organic matter is likely similar. 3.2. Proposed analytical methodology for on-site analysis The methodology proposed is depicted in Fig. 2. A sample is collected and filtered through a 0.45 mm filter to separate the particulate and dissolved fractions. Three different treatments are applied on the filtered fraction: (a) the filtered fraction is acidified with HNO3 before being analyzed by SCP to get the total dissolved metal concentration, (b) the filtered fraction is directly analyzed by using the AGNES method to get the free metal concentration and (c) the filtered fraction is ultrafiltered at 1 kDa then acidified with HNO3 before being analyzed by SCP to get the metal concentration including both, the free metal ion and the metal complexed to small ligands. For each fraction, the pH, the conductivity and the temperature are measured and aliquots are then taken for differ- ent analyses. Total metal determination: First the electroanalytical system is calibrated to obtain the value of the proportionality constant; α (Eq. (2)) using three or four metal additions measured in triplicate. The calibration solution is prepared using 20 mL NaNO3 at the same conductivity as the sample concentrations, which is then acidified by addition of 0.1 mL of HNO3 1 mol L�1. Following the calibration procedure the electrodes are Fig. 2. Schema of the proposed analytica thoroughly cleaned using milli-Q water and the polystyrene bea- ker is changed. One aliquot of 20 mL of sample is placed directly in the PS beaker and acidified with 0.1 mL of HNO3 1 mol L�1 after which the total metal is measured in triplicate. Ten calibration curves were measured in the range 3�10�9 –1�10�7 mol L�1 for which the LOD obtained was in the interval 0.8 to 2.4�10�9 and seven calibration curves were taken in the range 3�10�9 to 1�10�7 mol L�1 to 3�10�7 mol L�1 yielding LOD in the interval 3.6–9.8�10�9 which is of the same order of magnitude than the LOD obtained for GF-AAS (5.0�10�9 mol L�1). Note that these values were obtained for a deposition time of 45 s, hence they can still be improved by a significant margin since traditionally deposition times used in anodic stripping voltammetry range between 60 and 600 s, as long as the blank solution has no metal impurities. Free metal determination: the first step is the verification of the equilibrium condition for AGNES measurements for the chosen deposition potential. This should be done regularly due to possible variations in the potential of the reference electrode. A serie of measurements is done using different deposition times, usually from 30 s to 360 s and the equilibrium conditions are reached when the reoxidation time becomes the same for increasing de- position time [8]. After this step a calibration is performed to obtain the value of the proportionality constant; h (Eq. (3)) using three or four metal additions measured in triplicate. The calibra- tion solution is prepared using 20 mL NaNO3 at the same ionic strength as the sample concentrations (being the minimum value used of 10 mmol L�1) and adding 1 mmol L�1 of a suitable buffer. Following the calibration procedure, the electrodes are thor- oughly cleaned using milli-Q water and the PS beaker is changed. One aliquot of 20 mL of sample is placed directly in the PS beaker, then the buffer is added (set to a final concentration of 1 mmol L�1) and if necessary the ionic strength is set to 10 mmol L�1 by addition of NaNO3 1 mol L�1 and then the free metal is measured in triplicate. Nine calibration curves were measured in the range 0.65�10�9–1�10�7 mol L�1 for which the LOD obtained was in the interval 0.7–3.6�10�9 mol L�1, which is a noteworthy result since this is the detection limit for the free metal and should be compared with the values obtained for ion selective electrodes (ISE) which are usually two to three orders of magnitude higher. In contrast this value will not be easier to improve ever since we are already operating at a deposition potential close to the limiting current. l methodology for on-site analyses. Fig. 3. Comparison of the SCP and GFAAS total measured metal with the real spiked concentration. Table 3 Tangential UF free Cd and small complexes metal measured by SCP and GFAAS in acidified solutions. Spiked total metal source (10�7 M) UF SCP (M) (10�7 M) UF AAS (M) (10�7 M) Total metal Source (10�7 M) UF SCP (M) (10�7 M) UF AAS (M) (10�7 M) Itapanhau 1 Itapanhau 2 0.7 0.3 0.2 0.7 0.3 0.2 1.5 0.5 0.5 1.5 0.4 0.5 2.4 1.0 0.9 2.4 0.9 0.9 3.2 1.3 0.9a 3.2 1.3 1.0a Sorocabinha 1 Sorocabinha 2 0.7 0.4 0.5 0.7 0.5 0.5 1.5 0.9 0.9 1.5 0.8 1.1a 2.4 1.4 1.4 2.4 1.3 1.5 3.2 1.8 1.9 3.2 1.8 2.0 a Points in disagreement. Fig. 4. Comparison of the total (spiked) metal with both the metal determined in the ultrafiltered samples and the free metal determined directly in the Itapanhau (a) and Sorocabinha (b) rivers samples. A.S.C. Monteiro et al. / Talanta 152 (2016) 112–118 117 Tangential ultrafiltration (1 KDa): free metal and small complexes: The membrane is conditioned for 15 min with the source solution and the return and ultrafiltrated flow valves are set to 1 and 0.3 mL min�1 respectively. 25 mL of ultrafiltered solution are col- lected and 5 mL are taken for the GF-AAS measurements. The re- maining 20 mL are acidified with 0.1 mL of HNO3 1 mol L�1 after which their metal concentrations are measured in triplicate by SCP. The determination of total metal and ultrafiltered metal per- formed by SCP is validated by comparison with the samples measured by GF-AAS. 3.3. Total metal determination in the source solution Due to the absence of Cd in the natural water samples, four spikes of Cd of approximately 0.7, 1.5, 2.4 and 3.2�10�7 mol L�1, were carried out in each experiment. Fig. 3 shows the concentra- tions, percentage of recovery and differences obtained by SCP and GFAAS in two spiked natural water solutions from Sorocabinha and one from Itapanhau rivers. The % of recovery obtained for SCP is quite good averaging 96% in a range varying from 85% to 109% while for GF-AAS are slightly worse with an average of 91% in an interval from 85% to 97%. These percentages of recovery are quite good since these environmental matrices are often difficult to measure reproducibly (Table B in Electronic Support information). The slope values in Fig. 3 show that SCP (0.953) is slightly more accurate than GFAAS (0.892) while the correlation coefficient va- lues show that SCP is less precise than GFAAS (0.976 vs 0.995). These results confirm the SCP ability to accurately determine the total metal in these spiked natural samples. In the electrochemical measurements, no destruction of the organic matter was necessary for the total metal determination which is probably due to the weak binding of Cd(II) to the organic matter, which are mostly from the fulvic acid fractions. For other metals like Cu(II) or Pb(II) this simplification of the methodology assumed here must be validated. 3.4. Ultrafiltered (o1 kDa) metal determination Table 3 shows the metal concentrations measured by SCP and GFAAS in the filtered solution for four experiments carried out in the Sorocabinha and Itapanhau rivers. First the results obtained for the two replicas are in good agreement. A second observation is that the results obtained by SCP and GFAAS are mostly concordant, with the exception of the three GFAAS points marked with aster- isks. Again the results obtained by SCP in an acid media are in good agreement with the GFAAS confirming the good performance of this technique. An interesting result is that although the DOC contents and pH are quite similar for both samples, the metal measured in the ul- trafiltrated samples is significantly higher for the Sorocabinha than for the Itapanhau samples. 3.5. Free metal determination in the source solution Fig. 4a and b shows the comparison of the total (spiked) metal with both the metal determined in the ultrafiltered samples and A.S.C. Monteiro et al. / Talanta 152 (2016) 112–118118 the free metal determined directly in the Itapanhau (a) and Sor- ocabinha (b) river samples. The free metal is determined directly in the river solution by AGNES and upon comparison with the free plus small metal complexes measured in the ultrafiltered samples. Three interesting results are evident: i) for the Itapanhau the free metal fraction determined by AGNES and the ultrafiltered fraction are quite similar; ii) for the Sorocabinha the free metal is sig- nificantly smaller than the metal measured in the ultrafiltered fraction; iii) the metal concentration in the ultrafiltered fraction of the Sorocabinha is roughly twice that of the Itapanhau while its free metal in the source is actually smaller. Starting from the first observation the agreement between the free and ultrafiltered fraction for the Itapanhau River is quite re- markable indicating that apart from the natural organic matter, represented by the DOC there are few small inorganic ligands in solution that are able to permeate the 1 kDa membrane. Regarding the methodology, this is an interesting observation since the agreement between ultrafiltered and free metal determinations implies a negligible contribution from small inorganic ligands for the metal binding. The second observation means that the Sorocabinha River must be more concentrated in small inorganic ligands that are able to permeate the 1 kDa membrane, so that the metal content mea- sured in the ultrafiltered fraction is higher than the free metal content measured in the source. This is supported by the much higher conductivity of the Sorocabinha (1043 mS cm�1) as com- pared with the Itapanhau (189 mS cm�1). The third observation suggests that the Cd speciation is dif- ferent in both river samples being controlled mostly by the organic matter in the Itapanhau River while in the Sorocabinha river the small inorganic ligands and the competition from other cations, namely a roughly 10� higher concentration of calcium and magnesium ions in solution must be taken into account. Note that for Cd binding with DOC is mostly electrostatic so the competition of Ca and Mg is very effective. Regarding the DOC even if the concentration is similar in both rivers the nature of the organic matter present might be different, namely in the relative amounts of fulvic and humic fractions present. The higher ionic strength and the cations present in the Sorocabinha mean that the humic fraction of the DOC will be less stable or already coagulated where in the Itapanhau this humic fraction can still be significant. In future works we plan to investigated this issue in detail. To confirm these observations we carried out an AGNES mea- surement in the non-acidified ultrafiltered fraction of two Sor- ocabinha river samples (spiked at 2.4 and 3.2�10�7 mol L�1) to determine the free Cd proportion in the ultrafiltered solution. The results showed that only 37–45% of the total Cd of this fraction is free, thus confirming the metal is indeed complexed in the non- acidified ultrafiltered sample, most likely by small inorganic ligands. We validated both the SCP total Cd2þ determination in the source and acidified ultrafiltered solution with a LOD of 1.6�10�9 M and the AGNES free Cd2þ determination with a LOD of 1.9�10�9 M. We demonstrated that the proposed metal spe- ciation methodology is able to provide valid information regarding the total and free metal while being able to additionally afford some insight on the contribution of small inorganic ligands to the metal complexation in the source. We plan to follow this proof-of-concept with the development of the same methodology for other metals like Pb and Zn and their mixtures, to further investigate the nature of the DOC in both rivers and finally to apply this methodology on-site and obtain experimental data to model the binding of metal ions in these rivers. Acknowledgment The authors acknowledge the support of FAPESP and CNPq (Brazilian Agencies) and the support from the French National Research Agency ANR Species (ANR) and the Portuguese National Funding Agency For Science, Research and Technology (FCT-ANR/ AAG-MAA/0065/2012 and Blanc International II-SIMI 6-Système Terre environnement risques SPECIES). A.S.C. Monteiro acknowl- edges a PhD grant from FAPESP/Brazil and A.H. Rosa/J.P. Pinheiro acknowledges a visiting grant from CNPq/Brazil (PVE 400572/ 2013-3). Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.talanta.2016.01. 053. References [1] D.M. Di Toro, H.E. Allen, H.L. Bergman, J.S. Meyer, P.R. Paquin, R.C. Santore, Biotic ligand model of the acute toxicity of metals. 1. Technical basis, Environ. Toxicol. Chem. 20 (10) (2001) 2383–2396. [2] J. Buffle, Complexation Reactions in Aquatic Systems: An Analytical Approach, Ellis Horwood, Chichester, 1988. [3] J. Buffle, G. Horvai, In Situ Monitoring of Aquatic Systems: Chemical Analysis and Speciation, John Wiley & Sons, New York, 2000. [4] A.M. Mota, J.P. Pinheiro, M.L. Simoes Goncalves, Electrochemical methods for speciation of trace elements in marine waters. Dynamic Aspects, J. Phys. Chem. A 116 (25) (2012) 6433–6442. [5] R.M. Town, H.P. van Leeuwen, Fundamental features of metal ion determina- tion by stripping chronopotentiometry, J. Electroanal. Chem. 509 (1) (2001) 58–65. [6] R.M. Town, H.P. van Leeuwen, Effects of adsorption in stripping chron- opotentiometric metal speciation analysis, J. Electroanal. Chem. 523 (1–2) (2002) 1–15. [7] C. Parat, A. Schneider, A. Castetbon, M. Potin-Gautier, Determination of trace metal speciation parameters by using screen-printed electrodes in stripping chronopotentiometry without deaerating, Anal. Chim. Acta 688 (2) (2011) 156–162. [8] J. Galceran, E. Companys, J. Puy, J. Cecilia, J.L. Garces, AGNES: a new electro- analytical technique for measuring free metal ion concentration, J. Electroanal. Chem. 566 (1) (2004) 95–109. [9] L.S. Rocha, E. Companys, J. Galceran, H.M. Carapuca, J.P. Pinheiro, Evaluation of thin mercury film rotating disk electrode to perform absence of gradients and Nernstian equilibrium stripping (AGNES) measurements, Talanta 80 (5) (2010) 1881–1887. [10] C. Parat, L. Authier, D. Aguilar, E. Companys, J. Puy, J. Galceran, M. Potin- Gautier, Direct determination of free metal concentration by implementing stripping chronopotentiometry as the second stage of AGNES, Analyst 136 (20) (2011) 4337–4343. [11] P. Burba, J. Rocha, D. Klockow, Labile complexes of trace-metals in aquatic humic substances-investigations by means of an ion exchange-based flow procedure, Fresenius J. Anal. Chem. 349 (12) (1994) 800–807. [12] P. Burba, J. Van den Bergh, D. Klockow, On-site characterization of humic-rich hydrocolloids and their metal loading by means of mobile size-fractionation and exchange techniques, Fresenius J. Anal. Chem. 371 (5) (2001) 660–669. [13] S.C.C. Monterroso, H.M. Carapuca, J.E.J. Simao, A.C. Duarte, Optimisation of mercury film deposition on glassy carbon electrodes: evaluation of the com- bined effects of pH, thiocyanate ion and deposition potential, Anal. Chim. Acta 503 (2) (2004) 203–212. [14] H. Soares, P. Conde, A. a N. Almeida, M. Vasconcelos, Evaluation of n-sub- stituted aminosulfonic acid pH buffers with a morpholinic ring for cadmium and lead speciation studies by electroanalytical techniques, Anal. Chim. Acta 394 (2–3) (1999) 325–335. [15] D. Aguilar, C. Parat, J. Galceran, E. Companys, J. Puy, L. Authier, M. Potin- Gautier, Determination of free metal ion concentrations with AGNES in low ionic strength media, J. Electroanal. Chem. 689 (2013) 276–283. [16] D. Goveia, J.C. Rocha, L.C. de Oliveira, L.C. de Morais, V. de Campos, L.F. Fraceto, A.H. Rosa, Structural caracterization of aquatic humic substance extracted from Itapanhau and Ribeira De Iguape Rivers, Quimica Nova 34 (5) (2011) 753–U31. http://dx.doi.org/10.1016/j.talanta.2016.01.053 http://dx.doi.org/10.1016/j.talanta.2016.01.053 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref1 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref1 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref1 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref1 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref2 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref2 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref3 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref3 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref3 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref4 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref4 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref4 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref4 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref5 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref5 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref5 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref5 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref6 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref6 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref6 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref6 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref7 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref7 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref7 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref7 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref7 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref8 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref8 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref8 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref8 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref9 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref9 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref9 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref9 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref9 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref10 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref10 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref10 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref10 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref10 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref11 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref11 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref11 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref11 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref12 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref12 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref12 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref12 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref13 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref13 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref13 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref13 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref13 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref14 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref14 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref14 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref14 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref14 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref15 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref15 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref15 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref15 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref16 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref16 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref16 http://refhub.elsevier.com/S0039-9140(16)30053-4/sbref16 Towards field trace metal speciation using electroanalytical techniques and tangential ultrafiltration Introduction Materials and methods Reagents Equipment Natural water collection and characterization experiments Graphite furnace atomic spectroscopy (GF-AAS) experiments Tangential ultrafiltration experiments Working electrode preparation Electrochemical experiments Stripping chronopotentiometry experiments in acid media (total metal in the source acidified and UF acidified) AGNES experiments (free metal in the source) Results and discussion Natural river waters characterization Proposed analytical methodology for on-site analysis Total metal determination in the source solution Ultrafiltered (lt1kDa) metal determination Free metal determination in the source solution Acknowledgment Supplementary material References