Speciation of Chromium in Water Samples after Dispersive Liquid-Liquid Microextraction, and Detection by Means of High-Resolution Continuum Source Atomic Absorption Spectrometry

A newly analytical method has been developed to determine total chromium and speciation of this element in water samples through dispersive liquid-liquid microextraction combined with a high-resolution continuum source flame atomic absorption spectrometry. The most significant variables affecting complexation and extraction were optimized by using response surface methodology and univariate optimization. The best conditions for both the complexation and extraction elements in this study were: complexing agent ammonium pyrrolidine dithiocarbamate (APDC 6.0 mmol L); pH at 2.0 (Cr) and at 7.0 (Cr total); NaCl (5% m/v); 1-undecanol (50 μL) and ethanol 300 (Cr) and 275 μL (total Cr). Under optimal conditions, this method resulted in a 20-100 μg L linear range for Cr and total chromium, detection limits of 0.35 (Cr) and 6.7 μg L (total Cr), as well as enriching factor of 26 (Cr) and 19 for total Cr. The method accuracy was carried out by using certified water reference material (NIST CRM 1643e), and the results achieved were in agreement with the certified value (t-test at a confidence interval of 95%). The method developed was applied in samples of mineral water, tap water (the recovery values ranged from 88 to 115%) and seawater.


Introduction
Chromium is an element commonly found in rocks, animals, plants, soil, rivers and seawater, dusts and volcanic fumes, as well as in Earth's crust at a concentration of approximately 100 mg kg -1 . 1 Chromium in the environment can be of natural origins, such as rock and soil erosion, volcanic eruptions, or of anthropogenic origins, through the use of this chemical element in a range of industrial activities such as metallurgy (steel, aluminum and alloys), refractories (cement, glass and clay) and chemical industries (leather tanning, wood preservation and pigments). 2The second source is significantly responsible for Cr III and Cr VI in the environment.
Trivalent chromium, Cr III , is essentially found in the mechanisms for metabolism of carbohydrates, lipids and proteins, with a lower level of toxicity and less mobility than hexavalent chromium in the environment, due to the fact that the element is present in mineral structures, in the form of precipitates with some elements as iron, aluminum and/or manganese, and complexed with organic matter. 3In contrast, hexavalent chromium, Cr VI , is said to be carcinogenic as it is highly permeable in biological membranes, this being related to the fact that the chromate ion is a predominant specie of Cr VI in physiological pH, whose tetrahedral structure is similar to other ions, such as sulfate and phosphate, leading it to make a path into cells by carrying ions. 4,5Since chromium's toxicological properties are dependent on the element's oxidation state, it is, therefore, important to carry out the correspondent chemical speciation for quantification purposes of these species. 5n the other hand, considering the low concentrations of chromium found, for example, in natural water samples as µg L -1 it is important to follow a preconcentration procedure prior to the detection process. 6This procedure will allow the concentration and separation of the matrix analyte, aiming to further detectability and eliminate any expected effects of the matrix and foreign ions. 7he so-called dispersive liquid-liquid microextraction (DLLME) was created by Rezaee et al. 8 in 2006 and consists of quickly inserting, with the aid of a microsyringe, a mixture of extractant and disperser solvents with an aqueous solution containing the analyte.The build-up of fine and disperse droplets in the extractant solvent allows for extraction the analyte.Among other advantages of DLLME, an important one is that can enhance the use of reduced volumes of organic solvents, at low cost, with quickness, good values of recovery and enrichment factors. 9DLLME has been widely used to determine organic and inorganic analytes, 10 considering that for the second analyte a complexing agent is usually added, thus resulting in a hydrophobic complex for an extraction at a later stage. 11Dithiocarbamates (DTCs) are mostly used as complexing agents, forming stable and neutral complexes with a wide range of metals. 12odium diethyldithiocarbamate (DDTC) and ammonium pyrrolidine dithiocarbamate (APDC) are two of the most widely used in DLLME. 13enerally, the detection of metals after DLLME is carried out using absorption and atomic emission techniques.Some of the techniques most widely applied are flame atomic absorption spectrometry (FAAS) and graphite furnace absorption atomic spectrometry (GFAAS). 13Other detection methods have been combined with DLLME including inductively coupled plasma optical emission spectrometry (ICP OES), 14 laser-induced breakdown spectrometry (LIBS) [15][16][17][18][19] and tungsten coil atomic emission spectrometry (WCAES). 20][23][24][25] In this context, this work aims at developing an analytical method for determination and speciation of Cr in water samples (tap water, mineral water and seawater) by using the DLLME procedure combined with HR-CS FAAS.To do so, the authors of this study assessed two complexing agents widely used, DDTC and APDC.Those variables encountered in complexation and microextraction were optimized by using response surface methodology and univariate optimization, which makes it possible to look into the influences of such variances, as well as the interaction effects between them on the analytical response.

Reagents and solutions
All the solutions used in this work were prepared with analytical-grade reagents and ultrapure water (Milli-Q Plus system, Millipore, Bedford, MA, EUA).The standard solutions of chromium were then prepared by diluting standard solutions of K 2 CrO 4 10% (m/v) and Cr III 1000 mg L -1 (Specsol, Jacareí, SP, Brazil).
DDTC and APDC (Sigma-Aldrich, St. Louis, MO, USA) were two complexing reagents used.For adjusting the values of pH, solutions of sulfuric acid and ammonium hydroxide were used in different concentrations that resulted from diluting sulfuric acid at 95-97% (Merck, Darmstadt, Germany) and ammonium hydroxide at 28-30% (Sigma-Aldrich).
To study the effect of foreign ions, solutions containing aluminum, barium, cadmium, calcium, lead, chlorite, cobalt, copper, iron, manganese, nickel, nitrate, potassium, sodium, sulfate and zinc were prepared by diluting the standard solutions with a concentration of 1000 µg mL -1 (Ultra Scientific, USA).
To assess the method accuracy, a certified reference material (CRM) of water was used (NIST 1643e trace elements in water) from the National Institute of Standards and Technology (Gaithersburg, MD, USA).Tap water (from our lab in the city of Santo André, SP, Brazil), mineral water (bought locally) and seawater (from city of Santos, SP, Brazil) were used as real samples.The tap and mineral water samples were used without any previous treatment and to remove particle material from the seawater the sample was filtered using a 0.45 µm pore size membrane filter.

DLLME procedure
For the complexation and extraction of chromium species by means of DLLME, different volumes of samples or reference solutions were transferred to a glass tube of 10 mL.

For determining total chromium
The pH value was adjusted with ammonium hydroxide (pH 7.0) then 600 µL APDC (90 mmol L -1 ) and 1800 µL NaCl 25% (m/v) solutions were added.The resulting solution was heated in a water bath at 80 ºC for 15 minutes 26 and cooled at room temperature, then the mixture 50 µL of 1-undecanol and 275 µL of ethanol was added using a glass syringe.The separation of phases was carried out in a centrifuge at 2000 rpm for 3 minutes.The organic phase was collected with a micropipette, then diluted in 200 µL of ethanol and then all volume introduced into HR-CS FAAS, with the aid of a pipette tip.
For the determination of Cr VI After acidification with sulfuric acid 1 mol L -1 (pH 2.0), 600 µL APDC (90 mmol L -1 ) and 1800 µL NaCl 25% (m/v) were added.By using a glass syringe, the mixture 50 µL of 1-undecanol and 300 µL of ethanol were added.The other steps followed the same procedure used for total chromium.

For the determination of Cr III
It was evaluated from the difference between the total Cr and Cr VI concentrations found using the above mentioned procedures.

Instrumentation
A pH meter model 780 Metrohm (Herisau, Switzerland) equipped with a combined glass electrode was used to measure pH, a hotplate model IKA C-MAG HS 7 (Germany) was used to heat to 80 °C, and a Quimis centrifuge model Q222T (São Paulo, Brazil) for speeding up the separation of phases.The stirred extracting and dispersing solvents were added to the sample through a glass syringe with a capacity of 1000 µL (SGE Analytical Science, USA).
The detection of chromium by HR-CS FAAS was carried out through a ContrAA® 300 (Analytik Jena AG, Jena, Germany) equipped with a xenon short-arc lamp (XBO 301, 300 W, GLE, Berlin, Germany) operating in hotspot mode.The measurements were made at 357.9 nm, the main line of chromium.The other instrumentation conditions used were: air flow 490 L h -1 , acetylene flow 110 L h -1 , burner height 13 mm, sample flow rate at 6 mL min -1 and spectral resolution 0.2 nm.

Effect assessment of sample volume in analytical signal
In the proposed DLLME procedure, the volume produced within the organic phase, containing the analytes extracted in 1-undecanol, was 45 µL.Even after dilution, the final volume with ethanol did not enable the continuous nebulization process in the detection.As result, it was then decided to carry out discrete nebulization.Discrete volumes (50 to 200 µL) of a solution containing Cr 50 µg L -1 obtained from the DLLME procedure were manually added to a pipette tip (20 to 200 µL) connected to the nebulizer aspiration tube. 27By keeping the instrumental parameters as above mentioned and a reading time of 10 s, transient signals were produced and then measured at peak area.Taking into account that the manual insertion of the sample can lead to a higher standard deviation of measurements, the sample volume selected was based on the peak area and corresponding relative standard deviations (n = 3).The intensity of peak area increased with the sample volume, and the relative standard deviations reached 15, 11, 7.0 and 7.7% for volumes of 50, 100, 150 and 200 µL, respectively.A sample volume insertion of 200 µL was selected for further experiments.
Oviedo et al. 28 achieved analytical calibration curves by performing continuous and discrete nebulization of Mo in an aqueous and organic medium.The ratio of angular coefficients of analytical calibration curves of samples in an organic medium obtained by discrete and continuous nebulization was around 10, displaying an increased sensitiveness level.That is, another advantage from discrete nebulization in addition to a reduced sample volume.

Optimization of flame composition in chromium detection
A study was carried out for flame composition in chromium analytical signal.Five different flame compositions were assessed with acetylene flow rate at 40, 100, 105, 110 and 120 L h -1 and by also keeping the air flow rate at 490 L h -1 (C:O 0.40; 0.97; 1.0; 1.1 and 1.2, respectively).Figure 1 shows that the stoichiometric flame with a C:O 0.40 ratio has the lowest sensitiveness level.On the other hand, in an air-acetylene flame formed by composition of 490 and 110 L h -1 air-acetylene, respectively (C:O 1.1), it was observed the highest sensitiveness level.A strategy widely used to increase sensitiveness in determining chromium consists of using a reducing flame; 29,30 however, as noticed in Figure 1, in higher flow rates of acetylene at 120 L h -1 , for example, the reduction of sensitiveness can be the result of carbides formation which makes the atomization process difficult.Thus a flame composition of 490 and 110 L h -1 of air and acetylene, respectively, was selected for further experiments.Vol. 28, No. 2, 2017

DLLME procedure optimization
The DLLME procedure was optimized by using multivariate designs in an Excel spreadsheet.The integrated absorbance signal achieved after microextraction was based on Origin®, version 9.0 (Copyright © 1991-2012, OriginLab Corporation) and Statistica®, version 10 (Copyright © 1984-2011, StatSoft).The optimization study was carried out independently for Cr III and Cr VI .
In order to perform the first screening of assessed variables, a fractional factorial design 2 v 5-1 was carried out.Table 1 shows all the factors studied and their respective levels, Table S1 (APDC complexant) and Table S2 (DDTC complexant) (Supplementary Information section) complete matrix of factorial design and integrated absorbance values.
Along with the results displayed in Tables S1 and S2, Pareto charts are presented in Figures 2a-2d, making it possible to assess the influence of variables and interaction between them on the analytical response.The horizontal bars represent the absolute values of estimated effects, while the vertical bar is known as a reference line.The effects that go beyond the reference line are significant for the analytical response within the confidence interval of 95% (p < 0.05). 31igures 2a and 2b show that the two chromium species have the most significant effects for variables, pH value, complexant concentration and interaction between them, thus enhancing the importance of both variables in the DLLME procedure.At low pH values, Cr III is complexed with water molecules forming aqua complexes which are kinetically inert.3][34][35][36][37] The pH value is one of the variables allowing the speciation of chromium, as different from Cr III , the hexavalent specie is immediately reduced to the trivalent form in the presence of the complexant agent, thus quickly forming the complexes.
Figure 2c indicates the same occurrence to variables, pH value, complexant concentration and interaction between them show significant.As for Cr VI , Figure 2d, the only variable seen as significant was the disperser solvent (ethanol) volume.
Based on the results from the fractional factorial design and considering: (i) the need to keep the concentration of complexants, APDC and DDTC, in excess related to the chromium species in order to ensure the quantitative complexation in the presence of other metal ions; (ii) the value of pH is one of the variables allowing the speciation of chromium; (iii) the increase of extractant solvent volume (1-undecanol) results in a reduced analytical signal due to lower enrichment factor; and (iv) the increase of electrolyte concentration displayed a positive effect on complex Cr III /APDC it was then decided to carry out a factorial design 3 2 with variables, pH value and disperser solvent in complexation and extraction of Cr III with APDC and DDTC.The remaining variables such as 6.0 mmol L -1 of APDC and DDTC (highest level), 50 µL of 1-undecanol (lowest level), 5.0% (m/v) of NaCl (highest level) for Cr III /APDC and 0.0% (no addition) of NaCl (lowest level) for Cr III /DDTC were kept constant.
The design experimental matrix generated a total of 11 experiments, and the values of integrated absorbance are shown in Table S3 (Supplementary Information section).Table 2 shows statistical results calculated to 3 2 factorial design.
Taking into account the results presented in Table 2, it can be observed that just the average was considered significant within 95% confidence interval (p-value < 0.05).The factors investigated did not present significance within 95% interval because all p-values were higher than 0.05.The lack of fit test returned F calculated (MS lack of fit /MS pure error ) equal to 4.33 for Cr III /APDC and 0.57 for Cr III /DDTC, lower than F critical0.05,v1=3,v2=2= 19.16,which indicates absence of lack of fit within 95% confidence interval (Tables S4 and S5, Supplementary Information section).The regression significance test returned the value of F calculated (MS model /MS residual ) equal to 0.91 for Cr III /APDC and 2.84 for Cr III /DDTC, lower than F critical 0.05,n1=5,n2=5 = 5.05 which indicates a no significant regression within 95% confidence interval (Tables S4 and S5, Supplementary Information section). 31However, taking into account that the percentage of variation explained by regression and maximum percentage of variation explained can be considered acceptable and the models residue for Cr III /APDC and Cr III /DDTC presented normal behavior through Shapiro-Wilk test for 95% confidence (p-value > 0.05), the response surface was built from the experiments.
Figures 3a and 3b, these graphs show that by using APDC, optimal conditions for DLLME were disperser solvent volume 275 µL and pH value 7.0, and also using  DDTC the best conditions for DLLME corresponded to the central point, that is, 275 µL of disperser solvent and pH value of 6.5.
For Cr VI based on the results obtained from the fractional factorial design, Figures 2b and 2d, and considering: (i-iii) as previously mentioned for Cr III ; (iv) that the increase of concentration of electrolyte displayed a positive effect on complex Cr VI /APDC and (v) that the only significant variable for Cr VI /DDTC was the disperser solvent volume, it was also decided to assess the disperser solvent (ethanol) volume influence on microextraction of Cr VI with DDTC.The other variables were maintained fixed: 6.0 mmol L -1 of DDTC (highest level), pH value 2.0 (lowest level), 0.0% (no addition) of NaCl (lowest level) and 50 µL of 1-undecanol (lowest level).
Figure 4 shows an increased integrated absorbance up to 280 µL of disperser solvent.In DLLME, the disperser solvent promotes the extractant solvent dispersion in aqueous solution, but an excess of disperser solvent should be avoided due to an increase in solubility of hydrophobic complex in aqueous phase, thus its reduced quantity in the organic phase. 19Even though the integrated absorbance displayed a slight increase at 310 and 320 µL of disperser solvent (Figure 4) the analytical signal was kept below the initial volumes, 250 and 260 µL.As the extraction efficiency was higher between 270 and 280 µL of disperser solvent, we decided then to keep the volume of 275 µL for remaining part of this study.
In summary, using DDTC the selected experimental conditions as follows: value of pH 2.0 for Cr VI and pH 6.5 for total Cr, concentration of DDTC 6.0 mmol L -1 , with no addition of NaCl, 50 µL for 1-undecanol and 275 µL of ethanol.By using APDC, conditions as follows: value of pH 2.0 for Cr VI and pH 7.0 for total Cr, concentration of APDC 6.0 mmol L -1 , 5.0% (m/v) NaCl, 50 and 275 µL of 1-undecanol and ethanol, respectively, for total Cr and 300 µL of ethanol for Cr VI .
The reaction between Cr III and APDC is then accelerated by heating the solution prior to the extraction procedure. 26,33,35,36o assess the influence of temperature and time on formation of the Cr III complex with APDC after adjusting pH value (7.0) and addition of APDC (6.0 mmol L -1 ) to an aqueous solution containing 50 µg L -1 of Cr III , the resulting solution was heated to 80 °C for 15 minutes 26 and after cooling at room temperature, microextraction took place.Integrated absorbance increased by 32 times in compared with the absorbance achieved without the heat step, thus showing the importance of the latter in higher sensitivity.
To select the type of complexant, the DLLME procedure was followed (n = 3), using experimental conditions previously selected and a heating step using APDC at pH 7.0, in aqueous solutions containing a mixture of Cr III and Cr VI 50 µg L -1 .The selection of complexant was based on both the peak area and the relative standard deviations (RSD).At pH 2.0 (Cr VI ), integrated absorbance corresponded to 0.129 (RSD 21%) and 0.257 (RSD 3.2%) for DDTC and APDC, respectively.For total chromium-pH 6.5 (DDTC) and pH 7.0 (APDC)-the values of integrated absorbance were respectively 0.469 (RSD 9.3%) and 0.418 (RSD 0.98%).Apart from the lowest relative standard deviation using APDC for Cr VI and total chromium, it was possible to notice that the analytical signal was twice as high for determining Cr VI with APDC.It was then decided to use APDC as the complexant in the DLLME procedure proposed for determination and speciation of chromium.Table 3 shows optimal conditions for the DLLME procedure as well as for detection by HR-CS FAAS.

Analytical features of the proposed method
The analytical curves were achieved with and without the DLLME procedure (DLLME-HR-CS FAAS and HR-CS FAAS).Both curves were also obtained in triplicate and for the analytical curve without the DLLME procedure, the standards solutions of Cr were prepared in a medium of ethanol.Table 4 summarizes the analytical features for DLLME-HR-CS FAAS and HR-CS FAAS.
By comparing the angular coefficient of the analytical curves with and without the DLLME procedure, an increased level of sensitivity was obtained (26 times for Cr VI and 19 times for total Cr).The combination of DLLME and HR-CS FAAS made it possible to reach quantification limits considerably below the acceptable levels of Cr established for drinking water by the World Health Organization (WHO), 38 Brazil's Ministry of Health 39 and European Drinking Water Directive of the European Union Council (EUC) 40 of 50 µg L -1 and the Environmental Protection Agency (EPA) of 100 µg L -1 . 41

Effect of foreign ions
The study of interfering elements is carried out, in general, in an univariate way, where only one ion is assessed at a time, although this type of study requires a  long time frame, as the ions are simultaneously present in the sample.In this paper, the study of foreign ions was conducted in a multivariate way by means of a two-level Plackett-Burman design allowing the calculation of the main effects of an experimental set through a reduced number of experiments. 42,43he foreign ions assessed were the following: aluminum, barium, cadmium, calcium, lead, chloride, cobalt, copper, iron, manganese, nickel, nitrate, potassium, sodium, sulfate and zinc.Except for chloride anions, nitrate and sulfate, the selection for the concentration of foreign ions was based on the value stated in the certified reference material NIST CRM 1643e (trace elements in water).The concentrations of foreign ions and levels assessed in the Plackett-Burman design are shown in Table S6 (Supplementary Information section).
Based on the design used, a matrix was designed with 27 experiments, out of which three correspond to the triplicate in the central point.For the DLLME procedure, the solution used was one containing 50 µg L -1 Cr VI and Cr III in the presence of foreign ions as well as in the conditions previously set (Table 3).The entire matrix and integrated absorbance values are displayed in Tables S7 and S8 (Supplementary Information section).Such values were used to design Pareto charts, Figures 5a and 5b, where it can be noticed that in the conditions as selected for the DLLME procedure, all the foreign ions did not interfere in the complexation, microextraction and detection of chromium, due to the fact that the variances encountered did not deviate any further from the reference line.As a result, they were not seen as significant for the analytical response within the confidence interval of 95% (p < 0.05).

Assessment of accuracy evaluation and applicability of the method to water samples
Table 5 shows the results obtained for the addition and  recovery trials of Cr III and Cr VI in different water samples and for determining Cr III on certified reference material NIST CRM 1643e (trace elements in water).Recovery values ranged from 88 to 115%, suggesting that the proposed method is free from matrix effect.According to the t-test carried out (at a confidence interval of 95%) there, were no significant differences between the values set and certified in the water reference material.
The proposed method was also applied to the detection of Cr in seawater, but both species of Cr were not detected.In order to compare the obtained results, total Cr was determined for mineral water 3 and seawater samples using the graphite furnace absorption atomic spectrometry (GFAAS) method.Table S9 (Supplementary Information section) presents information about the heating program using GFAAS, and some analytical features provided by the method.As obtained for the proposed method, Cr was not detected in both samples using GFAAS (LOD: 0.03 and LOQ: 0.10 µg L -1 ).

Comparison of the proposed method with those previous reported
5][46] The proposed method leads to detection limits higher than or comparable to the ones obtained by using different chelating reagents and extractant solvents.The proposed DLLME procedure uses a reduced volume of non-chloride solvents and low toxicity by combining 1-undecanol as an extractant solvent and ethanol as a disperser solvent.Moreover, the speciation of Cr was performed without adding any other reagent such as KMnO 4 for the oxidation of Cr III to Cr VI or Na 2 SO 3 to reduce Cr VI to Cr III , and the organic phase was introduced without any previous evaporation of chloride solvent.
The results presented in the study of influence from other ions, Figures 5a and 5b, as well as those presented in Table 5 showed the good method selectivity without the need to add masking agent such as ethylenediamine tetraacetic acid (EDTA), that are at times necessary to the elimination of the interference of other ions, Co, Fe and Ni, for the detection of Cr by FAAS. 26,35,36The combination of DLLME with HR-CS-FAAS to detect Cr presents, in addition to the advantages previously showed, the possibility to a simultaneous correction of the background within the proximities of the analytical line, enabling to view the spectral region around the analytical line and automatic correction covering every continuum events. 24

Conclusions
The development of analytical methods that allow speciation and determination of potentially toxic chemical Value ± standard deviation, n = 3; b Cr III concentration was evaluated from the difference between total Cr and Cr VI concentration, (value ± standard deviation, n = 3); c certified value 20.40 ± 0.24 µg L -1 Cr; At 95% confidence limit (n = 3) t teorical = 4.303 and t experimental = 3.939; LOQ: limit of quantification.Vol. 28, No. 2, 2017   elements has become of paramount importance, mainly when toxicological characteristics are dependent on the oxidation state, which can be seen for Cr III and Cr VI .The DLLME procedure proposed in this paper made it possible to carry out the determination and speciation of chromium in different water samples by using reduced volumes of non-chloride solvents, 1-undecanol and ethanol, and a consequent reduction in residue produced.The combination of DLLME and HR-CS FAAS also made it possible to reach quantification limits, 1.1 and 22 µg L -1 for Cr VI and total Cr, respectively, considerably below the acceptable levels of Cr established for drinking water.

Figure 5 .
Figure 5. Pareto chart achieved from Plackett-Burman factorial design aiming at the effect of interfering ions in optimized conditions of extraction and detection at (a) pH 2.0; (b) pH 7.0.

Table 1 .
Factors and levels assessed to fractional factorial design 2 v

Table 2 .
Statistical results for 3 2 factorial design

Table 3 .
Commitment conditions used for chromium speciation and detection

Table 5 .
Determination of Cr III and Cr VI in water samples and certified reference materials using the proposed DLLME-HR-CS FAAS method