USO DE LA CROMATOGRAFÍA GASEOSA BIDIMENSIONAL COMPLETA PARA EVALUAR EL TIEMPO DE VIDA MEDIA DEL ACEITE DE SOYA, SU EFECTO EN LA ELABORACIÓN DE BIODIESEL

...

In this context, since its introduction, the Comprehensive Two-dimensional Gas Chromatography (GC×GC) has been readily recognized as a tool that offers increased separation and detectability without extending analysis time (Liu et al., 1991; Dallüge and Beens, 2003; Gorécki et al, 2004; Marriot et al., 2012). Essentially, comprehensive GC consists of a modified GC with two serially coupled columns, intermediated by a modulator. The primary criterion is that the separation mechanisms of the two stationary phases must differ, to some degree, in selectivity (Dallüge and Beens, 2003; Adahchour et al., 2006). Secondly, the dimensions of the capillary columns must have an appropriate sampling frequency (or modulation period), in order to minimize the band broadening and maximize peak production rate in the secondary column (Stoll et al., 2007; Marriot et al., 2012). Clearly, this technique has revealed many trace-constituents present in otherwise common samples, hence leading to a rediscovery of new markers in many different areas. Apart from the theoretical considerations, an advantage of comprehensive GC arises when the dimensionality of the techniques resembles that of the sample (Jalali-Heravi et al., 2011). Consequently, a clear relationship between chemical structure and retention coordinates is readily perceived, commonly refered to as chromatographic structure or structuration, which often can be relied on as an additional identification tool.

Because of the higher dimensionality of the technique, an increased amount of information is inherently present in the GC×GC chromatograms. As a consequence, the interpretation of these results is often a difficult and cumbersome task. In this sense, high order chemometric tools such as Multiway Principal Component Analysis (MPCA) is often recommended (Wold et al., 1987). MPCA allows grouping and further classification of these samples according to their chemical similarities through the descomposition in score and loadings of a multi-way array with ordinary properties of two-way PCA, the score vectors show the classification between samples and the loading the compounds responsible by this classification, in this manner MPCA is able to determine the relationship between samples complex of complex chemical mixtures. It is worth mentioning that for proper use of this algorithm it is necessary for the retention time to be highly reproducible in both dimensions, in this context various post-data processing methods are commonly used (Wold etl al., 1987; Brereton, 2003).

More recently, multivariate analysis with GC×GC data has proved a technique promising data analysis technique (Pierce et al., 2006; Ventura et al., 2011; Mispelaar et al., 2003; Mispelaar et al., 2005; Hantao et al., 2013). Multiway Principal component analysis and GC×GC–FID were used to identify new compounds responsible for the classification of biodiesel from several raw materials, which suffered from coelution during conventional analysis (Mogollón et al, 2016). Ventura et al., 2011, used the MPCA and GC×GC–FID for the chemical classification of eight maltene fractions from crude oils, this tecnique allows to identify large numbers of new components in this fraction. Hantao et al., 2013, in the same manner used unfolded-partial least squares discriminant analysis (U-PLS-DA) as mathematical algorithm for the tentative identification of 40 compounds useful as disease biomarkers for the Eucalyptus plant. However, the consequences of this increased detectability and peak capacity upon the results obtained by these exploratory analysis or MPCA is still incipient.

So, the objective of this study is to evaluate the influence of the different column combinations on the exploratory results and identify the compounds responsible for the differentiation between the biodiesel prepared from regular and aged soy oil by GC×GC-FID and GC×GC-QMS.

- Materials and Methods

- Chromatographic methods

GC× GC-FID

The GC×GC-FID equipment was used for the preliminary optimization of the chromatographic profiles, and the generation of the raw data used for the exploratory analysis. This prototype with cryogenic modulator is based on a HP 6890 Series GC-FID coupled to a model 7263 liquid auto-sampler (Hewlett-Packard, Wilmington, DE) with a split-splitless injector, the cryogenic fluid was N2 cooled in liquid nitrogen (LN2) N2, the flow was toggled by two three-way Asco solenoid valves (Florham Park, NJ) controlled by a NI USB 6008 12 bit AD/DA board controlled by lab-made software developed using the LabView 8.2 programming environment (National Instruments, TX, USA) (Pedroso et al., 2008).

The evaluated combinations of capillary columns is shown in the Table 1. A modulation period of 6.0 s and a frequency of acquisition of 100 Hz, was used for all runs.

The injection and detection ports were held at 250 °C and 275 °C, respectively. Hydrogen at 0.6 mL min-1 was used as carrier gas. With the aim of increasing the detectability of the minor compounds, 2 µL of biodiesel were injected with a concentration of 10 % in the system GC×GC-FID. The chromatographic analysis were carried out at least in duplicates.

Table 1

GC×GC-QMS

The GC×GC-QMS prototype consisted of a modified commercial GC-QMS QP2010Plus (Shimadzu, Tokyo, JPN) fitted with a split-splitless injector (split ratio 1:10) where 1 µL was injected, Furthermore, this prototype also uses a two-staged cryogenic modulator. The command to these valves was controlled by a NI USB 6009 14 bit AD/DA board. The column set consisted of a HP-5 (30 m length, 0.25 mm i.d. and 0.25 µm film thickness) column connected to a SupelcoWax 10M (1 m length, 0.10 mm i.d. and 0.10 µm film thickness) column. The oven’s temperature programming was set from 170 °C to 240 °C at a rate of 3 °C min−1, held at 240 °C for 15 min. The temperature of the injection port and the transfer line was set to 275 °C and 250 °C, respectivelly. The ion source temperature was set to 230 ºC. The ion count was performed by a continuous dynode electron multiplier, whose conversion dynode was set to 750 V. However, for the elution region of the majoritary compounds, such as C16:0 and C18:0, the ionization filament was turned off to avoid damages. The scan range was from 40 to 283 Th, yielding an acquisition rate of 25 Hz. For all runs, the modulation period was set to 6.0 s. For identification, the minimum accepted similarity between the reference and sample mass spectrum was 80%. For the retention index, a tolerance of 5 units was tolerated. The identification of the compounds was performed