Determination of Biodiesel Adulteration with Raw Vegetable Oil from ATR-FTIR Data using Chemometric Tools

Transcription

1 Article J. Brz. Chem. Soc., Vol. 22, No. 7, , Printed in Brzil Sociedde Brsileir de Químic $ A Determintion of Biodiesel Adultertion with Rw Vegetble Oil from ATR-FTIR Dt using Chemometric Tools Itâni P. Sores,*, This F. Rezende, b Rit de Cássi. C. Pereir, b Cláudio G. dos Sntos c nd Isbel C. P. Fortes b Embrp Agroenergi, Prque Estção Biológic, Av. W3 Norte s/n, Brsíli-DF, Brzil b Lbortório de Ensio de Combustíveis, Deprtmento de Químic, Universidde Federl de Mins Geris, Av. Antônio Crlos, Cmpus Pmpulh, Belo Horizonte-MG, Brzil c Deprtmento de Químic, Universidde Federl de Ouro Preto, Ru Diogo de Vsconcelos 122, Ouro Preto-MG, Brzil Três origens diferentes de biodiesel (lgodão, mmon e plm) form dulterds com óleo de soj cru, em concentrções vrindo de 1 40% (m/m). Ests mostrs form nlisds por espectrometri de infrvermelho médio (MIR) e os seus espectros form estuddos em três diferentes fixs espectris: espectro inteiro ( cm -1 ), e ns fixs de cm -1 e cm -1. Pr determinr origem do biodiesel utilizdo no sistem dulterdo, os ddos espectris form nlisdos pel ferrment nálise de componentes principis (PCA) e melhor segregção ds origens foi obtid pr o espectro inteiro ( cm -1 ). A vriânci explicd foi de 99%, pr os três primeiros componentes. Pr quntificr o óleo de soj cru foi plicd ferrment mínimos qudrdos prciis (PLS). Os melhores resultdos form obtidos pr região espectrl de cm -1, com vlores de RMSEP (erro médio qudrático de previsão) vrindo de 1,10 1,47% (m/m). Three different biodiesel sources (cotton, cstor nd plm) were dulterted with rw soyben oil t concentrtions rnging from 1-40% (m/m). These smples were nlyzed by infrred spectrometry (MIR) nd their spectr were studied t three different spectrl rnges: full spectrum ( cm -1 ), nd the spectrl rnges of cm -1 nd cm -1. To determine the source of biodiesel used in the dulterted system, the dt were nlyzed by PCA (principl components nlysis) spectrl nlysis tool nd the best segregtion of the sources ws obtined t the rnge of cm -1. The explined vrince ws of 99% for the first three components. To quntify the rw soyben oil prtil lest squres (PLS) tool ws pplied. The best results were obtined for the spectrl rnge of cm -1, with vlues of RMSEP (root men squre error of prediction) vrying from 1.10 to 1.47% (m/m). Keywords: biodiesel dultertion, multivrite clibrtion, ATR-FTIR Introduction Biodiesel is usully produced by the trnsesterifiction of vegetble oil or niml ft with short lcohol chin in the presence of ctlyst. 1,2 Since the CO 2 relesed during combustion is cptured by the oleginous plnt, biodiesel represents n importnt fuel lterntive. This contributes to the reduction in the emission of CO 2, which is the min responsible for the greenhouse effect. Combustion of *e-mil: itni.sores@embrp.br biodiesel lso reduces prticulte mteril nd SO x emission when compred to conventionl fossil fuel. 3 Biodiesel is minly produced from rpeseed oil in Europe nd other countries in the world. In Brzil, there is lrge number of oleginous plnts, which could be used for biodiesel production nd, becuse of groclimtic zoning, some of these oleginous cultures concentrte in specific regions. Plm, for exmple, is more common in the northern re while cstor is esier to find in the northest re, nd soyben culture develops better in the southern nd south estern res. This shows the gret potentil of Brzil s

2 1230 Determintion of Biodiesel Adultertion with Rw Vegetble Oil J. Brz. Chem. Soc. world producer nd exporter of this commodity. A common use of biodiesel is in blends with conventionl minerl diesel fuel. In Brzil, diesel hs been commercilized with the ddition of 5% volume of biodiesel since Jnury One of the biggest problems of the current fuel (gsoline, ethnol nd biodiesel) scenrio in Brzil is dultertion, 4 prt from the tx-evsion involved with this prctice. Adultertion lso results in incresing environmentl pollution, s well s, in consumer hrm, since the product does not meet the regulr specifictions, with potentil to cuse severl problems to cr engines. In the prticulr cse of biodiesel, government subsidy is different when compred to other fuels. This type of differentited subsidy cn led to unrel declrtions of the biodiesel source nd, consequently, to tx evsion. Another problem tht cn occur in the process of biodiesel production is the ddition of rw oil to B100 (pure biodiesel), since the process costs re still very significnt. Thus, it is impertive to solve or try to minimize these problems, by developing methodologies which llow both the identifiction of biodiesels source nd the determintion of its dultertion. One of the nlyticl techniques which hve been mostly used to monitor qulity of biodiesel nd petrodiesel blends is infrred (IR) spectroscopy, due to its mny dvntges. It is non-destructive, very relible nd llows direct nd fst determintion of severl properties without smple pretretment. 1,5,6 In recent yers number of reports hs ppered on the use of multivrite nlysis pplied to ner infrred spectroscopy (NIR) nd Fourier-trnsform infrred spectroscopy (FTIR). By using this pproch, Pereir et l. 7 hve determined gsoline dultertion; Che Mn nd Setiowt 8 hve determined ftty cid in plmitolein using clibrtion for prtil lest squres (PLS), which ws lso pplied by Knothe 5 to monitor the completion of the trnsesterifiction rection of biodiesel. Clibrtion methods bsed on FTIR, MIR nd NIR spectroscopy hve lso been developed for the determintion of the methyl ester content in biodiesel blends 9 nd the content of biodiesel in diesel fuel blends, tking the presence of rw vegetble oil into ccount. 1 The ppliction of multivrite models to the nlysis of biodiesel is vluble becuse the IR spectr of vegetble oils nd their corresponding esters re very similr, resulting in n overlpping bnd. 10 Nevertheless, this methodology hs not been ble to segregte biodiesel from different sources or to quntify the biodiesel dultertion with rw vegetble oil. Oliveir et l. 11 used FTIR nd NIR spectroscopy to design clibrtion models for the determintion of the methyl ester content in biodiesel blends (methyl ester + diesel). Other nlyticl techniques hve lso been utilized for the chrcteriztion of biodiesel profile. Monteiro nd co-workers 12 obtined good results by H 1 NMR technique to determine the biodiesel/diesel proportion using smples of soy nd cstor derived biodiesel mixed with diesel form three different btches. Cthrino et l. 13 fingerprinted severl origins of biodiesel using electrospry ioniztion mss spectrometry (ESI-MS). Also using ESI-MS, Eide nd Zhlsen 14 fingerprinted biodiesel origins nd mixtures of diesel, clssifying them with multivrite nlysis. In previous work, 15 the dultertion of biodiesel with vegetble oil ws determined using FTIR with the ccessory of ttenuted totl reflectnce (ATR) nd PLS clibrtion, with vrible selection. However, when this method is pplied there is need to optimize the model ccording to ech biodiesel origin. In the present work, n even simpler lterntive is presented, which uses spectrl rnge where there is high correltion between the IR bsorbnce nd the grde of dulternt tht cn be pplied to biodiesel produced from ny origin. Principl components nlysis (PCA) ws used to clssify the biodiesel origin lso using different spectrl rnges. Experimentl Smples Rw soyben oil ws purchsed from locl mrket. Biodiesel used in the experiments were from compnies nd/ or Universities, which hve lredy produced them for the mrket or on bench scle, kindly donted. Cstor oil ester ws supplied by Snt Cruz Stte University (Bhi Stte, Brzil), plm oil ester by Agroplm Compny (Prá Stte, Brzil), nd cotton oil ester ws obtined from Soymins (Mins Geris Stte, Brzil). The smples were produced by clssic bsic trnsesterifiction rection using methylic route Smples were chrcterized ccording to current prmeters estblished by the Ntionl Agency for Petrol, Nturl Gs nd Biofuels (ANP), Resolution 07/ The ssys were done t the Lbortory for Fuel Assys (LEC) of Federl University of Mins Geris (Mins Geris Stte, Brzil). A totl of 120 smples were prepred by mixing biodiesel from different sources with rw soyben oil in percentges vrying from 1 to 40% m/m with 1% m/m increments. These smples were used s clssifying for PCA nd clibrtion set. The externl vlidtion set comprised other 15 smples of ech source which were prepred in the sme wy s the clibrtion set, but the percentge of rw soyben oil dded ws rndomly chosen, resulting in 45 smples. ATR-FTIR nlysis ATR-FTIR spectr were obtined from n ABB Bomen IR spectrometer model MB 102 equipped with n

3 Vol. 22, No. 7, 2011 Sores et l ATR smpling ccessory with deuterted triglycerine sulfte detector. All spectr were collected t 16 ± 1 ºC using n verge of 16 scns, with spectrl resolution of 2 cm -1. The bckground spectr were obtined using clen ATR ccessory with n verge of 100 scns. After recording ech spectrum, the cell ws clened by successive tretments with heptne. The verge spectr in the rnge cm -1 from triplicte nlysis were treted chemometriclly using MINITAB softwre, version 14. Modeling nd dt nlysis PCA is well-known tool in multivrite dt nlysis for visulizing informtion from lrge dt sets. PCA relies on the liner trnsformtion of the originl set of mesurements into substntilly smller set of uncorrelted vribles while retining s much informtion present in the originl dt set s possible. 20,21 The originl dt set is substituted by two mtrices tht contin informtion bout the weight of the originl vrible in the PC spce (loding mtrix) nd the scttering of the smples in this spce (score mtrix). Thus, grphicl presenttion of the pir-wise components llows the nturl grouping of the smples to be observed indicting the similrity between smples nd llowing different groups of smples to be identified. In this work, PCA ws employed to verify the possibility of clssifying smples biodiesel from cotton, cstor nd plm oils, dulterted by different levels of rw soyben oil. Since ll vribles considered in this study were within the sme scle, PCs were obtined from the covrince mtrix. PLS regression is populr multivrite clibrtion method iming to ssess the degree of reltionship between set of x-predictor vribles nd set of y-outcome vribles. 1 It hs been widely pplied to multicomponent spectrl nlysis, especilly in IR, NIR nd Rmn spectroscopy. PLS is full spectrl clibrtion method nd hs built-in cpcity to del with specific problems of full spectrum clibrtion. 22 However, the selection of wvelength or wvenumber region is still very importnt. 23,24 An importnt gol is to serch for informtive spectrl regions for multicomponent spectrl nlysis. Informtive regions men tht they contin useful informtion for building PLS model nd re helpful to improve the performnce of the model. 20 PLS is powerful pproch for the nlysis of mixtures nd ws employed to determine the concentrtion of soyben oil in biodiesel, using leve-one-out cross vlidtion method. Predictive residul error sum of squres (PRESS) is commonly used criterion for LVs number selection. 23 For every set of dt from ech biodiesel source, PLS model with selected LV number is built, nd root men squre error of clibrtion for cross vlidtion (RMSECV). Once the externl vlidtion is mde, the root men squre error of prediction (RMSEP) cn be clculted. Results nd Discussion Physicl-chemicl ssys In order to simulte n dulterted system it is fundmentl to begin with smples tht re considered within specifiction, ccording to Brzilin legisltion (Resolution 07/2008). Thusly, the physicochemicl prmeters were previously obtined for ll the biodiesel smples used in this study. The ssys were performed ccording to the ntionl nd interntionl stndrds (ABNT/ASTM /EN), s presented in Tble 1. The specified vlues by the regultory gency nd the results obtined of the biodiesel smples from three distinct sources re showed in Tble 1. The results show tht the smples used meet the requirements of current legisltion. It is note worthy tht ll smples presented grdes of ether higher thn 96.5%, which is considered threshold for it to be mrketed. ATR-FTIR nlysis MIR spectr of cstor, plm nd cotton biodiesel re very similr to tht of non-esterified soyben oil (Figure 1), showing bsorption bnds in the regions 3700 to 3000 cm -1, 1900 to 1500 cm -1 nd 1800 to 800 cm -1. Another importnt feture in these spectr comes from the distinctive bnd t 3333 cm -1 in the cstor oil spectrum, which cn be ssigned to xil stretching vibrtions of hydroxyl O H bond in ricinoleic cid. 24 Bnds round 1200 cm -1 my be ssigned to the ntisymmetric xil stretching vibrtions of CC(=O) O, bonds of the ester group, while those round 1183 cm -1 my be ssigned to symmetric xil stretching of O C C bonds. Crbonyl bsorption of sturted liphtic esters usully ppers from 1750 to 1735 cm -1, while tht for α, β-unsturted esters from 1730 to 1715 cm -1. In monomers nd dimers of crboxylic cids crbonyl bsorptions pper on 1760 cm -1 nd from 1720 to 1706 cm -1, respectively. 24 Crboxylic cids show in-plne bending of C O H bond in 1408 cm -1 nd xil deformtion for dimer C O bond in 1280 cm -1. Crboxylic cid dimer shows n intense nd brod O H xil stretching in the region 3300 to 2500 cm 1, usully centered in 3000 cm This bsorption my be due to the hydroxyl of ricinoleic cid from cstor oil, ftty cids, glycerin nd mono-nd diglycerides. In Figure 1, the

4 1232 Determintion of Biodiesel Adultertion with Rw Vegetble Oil J. Brz. Chem. Soc. Tble 1. Physicochemicl ssys in the biodiesel smples Chrcteristic Unit Limit vlue Obtined vlue Method ABNT NBR ASTM D EN/ISO Aspect - (1)* Cotton, Cstor, Plm (Cler for ll) - Specific grvity t 20 º C kg/m³ * Cotton (860), Cstor (878), Plm (859) 4052 Kinemtic viscosity t 40 ºC mm²/s * Cotton (3.4), Cstor (5.2), Plm (3.7) 445 Flsh point, min. (2) ºC 100.0* Cotton (131.1), Cstor (142.6), Plm (132.3) 93 Ester grde, min mss % 96.5* Cotton (96.8), Cstor (97.2), Plm (97.5) EN Crbon Residul mss % 0.050* Cotton (0.039), Cstor (0.041), Plm (0.028) 4530 Sulfted sh, mx. mss % 0.020* Cotton (0.015), Cstor (0.018), Plm (0.012) 874 Sodium + Potssium, mx. mg kg -1 4* Cotton (0.8), Cstor (1.02), Plm (1.3) Clcium + Mgnesium, mx. mg kg -1 4* Cotton (0.9), Cstor (0.7), Plm (0.7) Phosphorus, mx. mg kg -1 10* Cotton (5.6), Cstor (6.3), Plm (6.02) Corrosivity to copper, 3 h t 50 ºC, mx. - 1* Cotton (1), Cstor (1), Plm (1) 130 Point of clogging, cold filter, mx. ºC 19 Cotton (4), Cstor (3), Plm (3) 6371 *Estblished limit ccording to the Brzilin legisltion; (1) Cler nd free of impurities with the ssys temperture noted; (2) When the nlysis of the flsh point exceeds 130 ºC, the nlysis of ethnol nd methnol content is discrted. ester. These uthors used multivrite clibrtion of the bnds in the region 1800 to 1700 cm -1, corresponding to xil stretching vibrtions of crbonyl groups to distinguish soyben from its ester. Clssifiction of biodiesel groups by PCA Figure 1. MIR infrred spectr of () cotton biodiesel, (b) cstor biodiesel, (c) plm biodiesel nd (d) rw soyben oil. hydroxyl bsorption is observed only in cstor oil biodiesel (spectrum b). The overlpped bnds in the fingerprint region (1300 to 900 cm -1 ) indicte tht univrite clibrtion models my cuse significnt prediction error in the quntifiction of biodiesel smples with different concentrtion when rw oil is present. Those models re lso indequte for identifying the presence of rw oil in spoiled blend either becuse of incomplete conversion during esterifiction rection or the illegl ddition of rw oil. Zgonel et l. 10 lso observed overlpped bnds in the MIR spectr of soyben oil nd its corresponding PCA ws used in n ttempt to evlute if biodiesel from different sources (cotton, cstor or plm oil) exhibited distinguishing fetures tht could mke the identifiction of these sources esier, even though they were spoiled by rw soyben oil. In n ttempt to obtin models with more efficient segregtion of groups the whole spectrum s well s some of its regions ws considered. According to literture, the best region ws ssigned to the xil deformtion of crbonyl group ( cm -1 ). Zgonel et l. 10 hve shown tht there is displcement of crbonyl bnd of biodiesel nd the rw oil when first derivtive spectrum is obtined in this region. Thus, from literture informtion nd visul nlysis PCA models were built up considering the full spectrum ( cm -1 ) nd two spectrl rnges ( nd cm -1, encompssing both the crbonyl s CC(=O)-O vibrtions. These spectrl rnges were lbeled s follows: model 1, 2 nd 3, respectively.

5 Vol. 22, No. 7, 2011 Sores et l The vrince explined by the first ten PCs for ech model is shown in Figure 2. For model 2 nd 3, the first three principl components cptured round 98% of the totl dt while for model 1 the response is 99%. The results suggest four independent vrition sources, indicting tht ech biodiesel source contributes differently to IR dt. Figure 3. Principl component nlysis for: ) MID full infrred spectrum ( cm -1 ), b) spectr rnge cm -1 nd c) spectr rnge cm -1. Figure 2. Vrince cptured for the first principl components. Figure 3 shows three-dimensionl plot of the first three principl components for ech set of dt studied. It is possible to distinguish, in ech model, three different groups of smples (cotton, cstor nd plm biodiesel). This demonstrtes unequivoclly tht segregtion between smples ws very efficient nd confirms tht the IR dt relly contins enough informtion to ggregte the smples ccording to its biodiesel source. However, generl trend to dispersion mong smples belonging to the sme group cn be noticed. Figure 3 shows the smllest dispersion for given group, nd the lrgest distnce between ech group. The best results for clssifiction nd identifiction of the biodiesel source were obtined from the full spectrum ( cm -1 ). The region cm -1, corresponding to crbonyl vibrtions, presented greter dispersion thn the full spectrum, which lso proved to be good region to work with multivrite nlysis. According to Figure 3c, the dt from spectrl rnge cm -1 shows to be inefficient for source seprtion when the three spectrl rnges were compred. It demonstrted close proximity between smples from cstor nd cotton biodiesel. PLS models In order to quntify the mount of rw soyben oil dded to the different biodiesel sources, multivrite clibrtion models were built, using the sme MIR spectr employed in PCA nlysis by PLS. Thus, one model ws built: for ech spectrl rnge studied: model 1 (full spectrum, cm -1 ), model 2 ( cm -1 ) nd model 3 ( cm -1 ). Tble 1 lists the root men squre error of clibrtion for cross vlidtion (RMSECV) of the model, correltion coefficient (R), ltent vrible (LV) nd root men squre error of prediction (RMSEP) of the PLS model, considering ll set of smples (cotton, cstor nd plm biodiesel) nd different spectrl rnges studied. Becuse of n lgorithm limittion in the softwre relted to the numbers of vribles tht cn be used, the RMSEP vlues were only clculted for model 2 nd 3. The mximum numbers of vrible is 1000 for set of clibrtion nd vlidtion, flling short from the 1488 vribles needed for the clibrtion nd prediction in model 1, precluding the externl vlidtion nd clcultion of RMSEP for this spectrl rnge. According to Tble 2, considering both the spectrl rnge s the biodiesel source, slight vrition from to ws observed for the R vlues. A similr behvior ws observed for the LV vlues, which fluctuted between 5 nd 7. While the R nd LV vlues did not vried noticebly between the models, the spectrl rnge, cm -1, gve the smllest vlues of RMSECV for ll of the three sources of biodiesel. A better picture of the clibrtion results cn be seen in Figure 4, resulting from the models developed for the three sources of biodiesel s function of the predicted vlues for model 3 ( cm -1 ). This spectrl rnge ws chosen becuse of the smllest vlues of RMSECV nd RMSEP presented. The plot shows only very smll dispersion which is comprble to ll three sources. Cross vlidted results were quite close to the clculted vlue, if not coincident. These findings strongly support region cm -1 s the most relible to detect dultertion of biodiesel by non esterified oil. One wy to observe qulittively the linerity of model is through the chrt of residuls versus concentrtion of the smples. Residuls should be rndomly distributed long

6 1234 Determintion of Biodiesel Adultertion with Rw Vegetble Oil J. Brz. Chem. Soc. Tble 2. PLS clibrtion results for biodiesel smples mixed with rw soyben oil Model Rnge Biodiesel R RMSECV (%, m/m) RMSEP (%, m/m) Vribles for ech spectrl rnge LV cm -1 cotton cstor plm cm -1 cotton cstor plm cm -1 cotton cstor plm Number of columns exceeded the softwre processing cpcity. the clibrtion curve. The residuls generted by the models were quite similr. Figure 5 shows the grph of residuls for biodiesel clibrtion from cstor oil for the spectrl rnge of cm -1. As displyed in Figure 5 the residuls re distributed rndomly, indicting the linerity of the model. Figure 5. Chrts of residuls produced in the PLS model of cstor oil biodiesel. The dvntge of this methodology in comprison to the previous one 15 is the speed nd simplicity for building the models. The spectrl rnge, cm -1, generted prediction errors comprble to those obtined in the previous work 4 using the selection of vribles with best results pproch, with RMSEP vrying from 1.10 to 1.47% (m/m) nd from 0.65 to 1.40% (m/m) respectively. Conclusion Figure 4. PLS clibrtion model for spectrl rnge cm -1 : ) cotton biodiesel, b) cstor biodiesel nd c) plm biodiesel. PCA hs shown tht MIR spectr dt contin informtion to differentite smples of biodiesel, ccording to their source, even if there is some mount of rw soyben oil present. This chemometric tool showed to be suitble to clssify blends of biodiesel/rw soyben distinguishing the groups from different sources very

7 Vol. 22, No. 7, 2011 Sores et l well. The best result ws obtined when the full MIR infrred spectrum ws used. PLS model bsed on MIR spectr developed in this work proved to be suitble s prcticl nlyticl method to predict rw soyben content in biodiesel blends from 1 to 40% m/m. The spectrl rnge cm -1, showed to the best region to develop PLS clibrtion model for quntifiction of rw oil in biodiesel smples giving the lowest vlues for RMSECV nd RMSEP. The dvntge of using this spectrl rnge to build models is tht these models cn be pplied to biodiesel of different sources. In contrst, the vrible selection method must be optimized for ech source of biodiesel. The dvntge of this methodology is tht it is very fst in determining the origin of the biodiesel, nd whether or not it is dulterted, if so the level of the dulternt. Acknowledgements The uthors would like to thnk Agênci Ncionl de Petróleo, Gás Nturl e Biocombustíveis (ANP) nd Fundção de Ampro à Pesquis do Estdo de Mins Geris (FAPEMIG) for the finncil support nd to LEC (Lbortório de Ensios de Combustíveis, UFMG) nd its stff for the technicl support. References 1. Pimentel, M. F.; Ribeiro, G. M. G. S.; Cruz, R. S.; Strgevitch, L.; Pcheco Filho, J. G. A.; Teixeir, L. S. G.; Microchem. J. 2006, 82, Zullikh, S.; Li, C.; Vli, S. R.; Ju, Y.; Bioresour. Technol. 2005, 96, Crbbe, E.; Nolsco-Hipolito, C.; Kobyshi, G.; Sonomoto, K.; Ishizki, A.; Process Biochem. 2001, 37, Sores, I. P.; Fonsec, T. R.; Fortes, I. C. P.; Energy Fuels 2009, 23, Knothe, G.; J. Am. Oil Chem. Soc. 2001, 78, Knothe, G.; J. Am. Oil Chem. Soc. 2000, 77, Pereir, R. C. C.; Skrobot, V. L.; Cstro, E. V. R.; Fortes, I. C. P.; Ps, V. M. D.; Energy Fuels 2006, 20, Che Mn, Y. B.; Setiowty, G.; Food Chem. 1999, 66, Oliveir, J. S.; Montlvão, R.; Dher, L.; Surez, P. A. Z.; Rubim, J. C.; Tlnt 2006, 69, Zgonel, G. F.; Perlt-Zmor, P.; Rmos, L. P.; Tlnt 2004, 63, Oliveir, C. C.; Brndão, C. R. R.; Rmlho, H. F.; d Cost, L. A. F.; Surez, P. A. Z.; Rubim, J. C.; Anl. Chim. Act 2007, 587, Monteiro, M. R.; Ambrozin, A. R. P.; Lio, L. M.; Ferreir, A. G.; Fuel 2009, 88, Cthrino, R. R.; Milgre, H. M. S.; Sriv, S. A.; Grci, C. M.; Schuchrdt, U.; Eberlin, M. N.; Augusti, R.; Pereir, R. C. L.; Guimrães, M. J. R.; de Sá, G. F.; Cixeiro, J. M. R.; de Souz, V.; Energy Fuels 2007, 21, Eide, I.; Zhlsen. K.; Energy Fuels 2007, 21, Sores, I. P.; Rezende, T. F.; Silv, R. C.; Cstro, E. V. R.; Fortes, I. C. P.; Energy Fuels 2008, 22, Schuchrdt, U.; Sercheli, R.; Vrgs, R. M.; J. Brz. Chem. Soc. 1998, 9, Fukud, H.; Kondo, A.; Nod, H.; J. Biosci. Bioeng. 2001, 92, Meher, L. C.; Sgr, D.V.; Nik, S. N.; Renew. Sust. Energ. Rev. 2006, 10, 298. Wold, S.; Esbensen, K.; Geldi, P.; Chemom. Intell. Lb. Syst. 1987, 2, ccessed in November Mrtens, H.; Nes, T.; Multivrite Clibrtion nd Clibrtion, John Wiley & Sons: Chichester, Du, Y. P.; Ling, Y. Z.; Jing, J. H.; Berry, R. J.; Ozki, Y.; Anl. Chim. Act 2004, 501, Spiegelmn, C. H.; McShne, M. J.; Goetz, M. J.; Motmedi, M.; Yue, Q. L.; Coté, G. L.; Anl. Chem. 1998, 70, Xu, L.; SchechterI, I.; Anl. Chem. 1996, 68, Silverstein, R. M.; Webster, F. X.; Spectrometric Identifiction of Orgnic Compounds, 6 th ed.; Wiley: New York, Submitted: September 10, 2010 Published online: Februry 24, 2011