Analytical standards used were 1-hexanol, 3-methyl-1-butanol, 2-phenylethanol, isoamyl acetate, hexyl acetate, ethyl lactate, diethyl succinate, ethyl butanoate, ethyl hexanoate, ethyl octanoate, ethyl decanoate, hexanoic acid, octanoic acid, decanoic acid, nerol, linalool, α-terpineol, α-ionone and β-ionone, purchased from Sigma- Aldrich (Saint Louis, MO, USA), with purity ≥ 99%. A synthetic model wine was prepared with water previously purified in a Milli-Q^® system (Millipore, Bedford, MA, USA), 12% (v/v) of ethanol HPLC grade (JT Baker, Xalostoc, México) and 2 g∙L⁻¹ of tartaric acid (Merck, Darmstadt, Germany). The pH was adjusted to 3.2 using sodium hydroxide (NaOH) 1M. Sodium chloride (NaCl) was purchased from Vetec (Rio de Janeiro, Brasil).

Samples of Chardonnay wine were obtained in local market in Campinas, São Paulo, Brazil, and four bottles of a same production lot were used in experiments. Wines were produced in Andradas, Minas Gerais, Brasil (22˚04'04''S 46˚34'08''W) in 2011 vintage. For analysis, 10 mL aliquots of wine were pipetted into a 40 mL SPME vial, 3.0 g of sodium chloride was added and complete with screw-top caps and PTFE/silicon septa (Supelco Inc., Bellefonte, PA, EUA). During the sampling time, sample was constantly stirred with a small magnetic stirring bar. SPME fiber (Supelco Inc., Bellefonte, PA, EUA) used in this study was 50/30 µm with divinylbenzene/carboxen/ polydimethylsi- loxane (DVB/CAR/PDMS) coating conditioned before use according to the manufac- turer’s instructions. DVB/CAR/PDMS fibers were chosen according to the range of polarity and different functionalities of the mixture of molecules analyzed in this study: alcohols, esters, fatty acids, C₁₃-norisoprenoids and monoterpenes. Fiber was exposed to the sample headspace after equilibrium time of 10 minutes. The factors optimized were time of fiber exposure and temperature of sample, due to their influence in equilibrium system. After extraction, fiber was introduced into gas chromatography injector for desorption of the analytes at a temperature of 270˚C, in splitless mode for 15 minutes.

The GC-MS analysis were performed on a Agilent 7890A gas chromatograph (Agilent Technologies, Palo Alto, CA, EUA) equipped with a Agilent 7000 Triple Quad mass detector (Agilent Technologies, Palo Alto, CA). Liner used was specific for SPME analysis purchase from Sigma Aldrich (Saint Louis, MO, USA), with 0.75 mm of internal diameter. Chromatographic separation was achieved using a capillary column Supelcowax^® 10 (100% polyethyleneglycol) (Supelco Inc., Bellefonte, PA, EUA.) with following dimensions: 30 m × 0.25 mm × 0.25 µm. Carrier gas was high purity Helium at a constant flow of 1.0 mL∙min⁻¹ in splitless injection mode. The injector temperature was 270˚C and oven temperature program initialize with 30˚C, was held for 2 minutes and then increasing 4˚C min⁻¹ to 130˚C (2 minutes) followed to increase 8˚C min⁻¹ to 250˚C (5 minutes) [21] . Solvent delay used was 2.5 minutes.

Mass spectras were obtained by using electron impact (EI) as ionization mode and −70 eV as electron energy. Temperatures of interface, source and quadrupoles (Q1, Q2 and Q3) were 250˚C, 260˚C and 150˚C, respectively. Nitrogenium and Helium were used in collision cell (Q2) at 2.25 mL∙min⁻¹ and 1.5 mL∙min⁻¹ flows, respectively. Acquisition was performed in multiple reaction monitoring (MRM) mode. Precursor ions were used as qualifiers and product ions were as identifiers (Table 1). Mass range analyzed was from 30 to 400 m/z and 50 milliseconds of acquisition time. Resolution of MS1 and MS2 was set in wide mode. Collision energy (CID) was chosen to different analytes and are showed in Table 1. Dwell time was set in 1 milliseconds for all analytes.

Data were acquired and processes using Agilent Mass Hunter software (version B.05.00, Agilent Technologies). The compounds identification was achieved by com- paring the retention time and mass spectra obtained from sample with standards compounds presented in a model synthetic wine injected at same conditions. Qualifier and identifier ions were considered positive when they showed similarity of at least 75% with the standards prepared and analyzed as well as comparing the MS fragmentation with the mass spectras present in the National Institute of Standards Mass Spectral Library (NIST 2011).

Optimization of the HS-SPME conditions was carried out using a 2² factorial central composite rotational design (CCRD) with four axial points (α = 1.4142) and tree central points [22] . Variables chosen were the temperature (T, ˚C) and extraction time (t, minutes) and other parameters (amount of NaCl, equilibrium time, velocity of stirring, sample volume) were arbitrarily established by the authors. The levels of each variable can be seen in Table 2. The values of the factors are adjusted to better control of experiments. Twelve experiments were carried out at random. The software Statistica® v.7 (Statsoft Inc., Tulsa, OK, EUA) was used for statistical analysis.

Volatiles compounds monitored in this study were chosen because they represent the major chemical classes of aroma compounds in wines: alcohols, esters, fatty acids, monoterpenes and C₁₃-norisoprenoids [24] . The fiber used, DVB/CAR/PDMS, was performed in accordance with the interest of this work: cover a wide range of polarity, volatility and functionality represented by selected volatile compounds. Several authors have reported that DVB/CAR/PDMS fiber as the most selective and efficient for the detection of volatile compounds in wines [16] [23] [24] . The amount of salt added is intended to increase the ionic strength and promote “salting out” of the volatile compounds from their matrix by increasing the partition coefficient and also, concen- tration of the analyte in the headspace before extraction [14] [24] . Theoretically, any inorganic salt may be used but most commonly used salts are sodium sulphate (Na₂SO₄) and sodium chloride (NaCl) due to their high solubility in aqueous medium (wine) [24] . The concentration used is, usually, 30% of the sample volume [16] [25] . Sample volume used must be established according to the experimental procedures considering the volume of the vial used, headspace volume, size and depth of fiber exposure to the sample headspace.

^†: with α = 1.4142; ^‡: expressed in arbitrary units; ^§: central point repetition.

Figure 1 shows a Pareto diagram where the data obtained in experiments can be observed. Based on the analysis of the effects of the factors involved in the design, with 95% confidence level, it can be see that only the temperature factor (T, ˚C) was significant in the models evaluated.

Table 2 shows results obtained in the experiments conducted by DCCR where the response, total area of the chromatographic peaks of the nineteen compounds selected and monitored, is expressed in arbitrary units. Moreover, the table also presents the levels of the factors time (t, min) and temperature (T, ˚C) used in the execution of the experiments. Experiment number 5, with extraction temperature at 30˚C and 45 minutes of extraction time (fiber exposure) showed greater chromatographic response with area values on 2.54E+07.

Analysis of variance (ANOVA) with 95% of confidence was used to determine which factors significantly affect the response of the HS-SPME procedure and validate the mathematical model that describes the response surface of DCCR. Table 3 shows the values obtained by ANOVA. Based on the regression results which shows the existence, or not, of lack of fit of the mathematical model predictions can be based on this model [15] [22] .

The statistical significance of regression given by the quadratic means of the residues (MQ_R/MQ_r) or F_calculated was 395.97. When comparing, at the level of 95%, the values of F_calculated and F_{tabulated (5%, 6%, 95%)} which value is 4.39 can be observed that F_calculated > F_tabulated about 90.2 times, indicating that the correlation between variables can be considered adequate to this model.

^†: F_calculated; ^‡: F_tabulated..

Based on the obtained quadratic model was generated response surface to the experiment (Figure 2). Quadratic model generated the equation R = 2.49E + 07 − 3.45E + 05 × T − 7.59E + 04 × T² − 7.40E + 03 × t − 2.56E + 04 × t², where T is the variable 1 (extraction temperature, ˚C), t were variable 2 (extraction time, min) and R is the response (total area of the chromatographic peaks).

The process of extraction by HS-SPME involves the partition of the analytes in matrix, in headspace and fiber coating. In the equilibrium, amount of extracted sample is proportional to the partition coefficient and concentration of the analytes in the headspace. Extraction can be considered optimal when the concentration of the analytes reaches equilibrium distribution between the extraction phase (fiber coating) and headspace [8] [9] [12] [26] .

In HS-SPME method for extraction of volatile compounds, temperature has great influence on efficiency of the process. The kinetics of the extraction process is directly affected by temperature, as it acts in determining the vapor pressure of the analytes in the matrix [27] . Furthermore, temperature affects directly the partition coefficient of the analyte. The temperature increase, due to thermodynamic conditions, reduces the partition coefficient and hence decreases the amount of analyte extracted [25] [26] [27] . For analysis of volatile compounds in wine, using HS-SPME, temperatures have been reported in literature between 35˚C and 55˚C. However, these values can be vary depending on the type of wine analyzed, the matrix constituents and concentrations of the analyzed compounds [14] [16] [25] .

Extraction time, or fiber exposure to the sample headspace, influences the equilibr- ium between the phases involved and thus the extraction efficiency. Compounds with a lower partition coefficient, the time required to reach equilibrium must be increased. Compounds with higher partition coefficient require less time to reach equilibrium [12] [25] [28] .

The exposure of the fiber for shorter periods of time, or before reach the equilibrium, can make the concentration of the extracted compounds become underestimated. In

addition to exposure of the fiber for very long periods of time makes the compounds starts to compete for the active site in the fiber and also affects the final concentration [12] [14] [16] [24] . For the determination of volatile compounds in wines, the extraction time optimized ranged from 30 to 60 minutes [14] [24] [25] . For HS-SPME extraction in wines Chardonnay produced in Rio Grande do Sul, Brazil, and used as a basis for sparkling wine, Welke and co-workers [16] used 45 minutes as the optimum condition of extraction. In this study, the same extraction time (45 minutes) was observed by multivariate analysis, being determined as optimum equilibrium time.

Analyzing the lack of fit of the model generated, based on the values of quadratic mean and pure error of each response where F_calculated was 2.23 times F_tabulated. However, for there to be considered a good fit of the model, F_calculated < F_tabulated [22] Based on this result, this model could not be used to make predictions about the response. However, considering the optimum extraction point is based in real results, obtained experim- entally and with coefficient of variation calculated for the experiments (repetitions) of the central point was considerably low, 0.14%, this model indicate adequate repeatability of the method developed in this condition. Thereby, the coefficients of the model equation even showed lack of fit, can be used for the systematic understanding of response values as a function of factor levels studied [29] .

After the HS-SPME extraction process, volatile compound were separated and ident- ified using gas chromatography coupled to tandem mass spectrometry (GC-MS/MS). Confirmation of identity of each analyte was performed comparing the spectra obtai- ned by the injection of analytical standards and between the analytes present in the sample. Use of GC-MS/MS provides a high degree of selectivity, sensitivity and security in identification of compounds [30] .