Microwave-assisted esterification (MAE) was carried out with sulfuric acid (10%, w/w H2SO4/FFA) as catalyst. The catalyst was dissolved in methanol (methanol to FFA mole ratio of 40) and was stirred for 5 min to ensure complete mixing. According to the American Oil Chemists’ Society (AOCS) Official Method (Ca 5a-40), FFA of OO was 2.36% which increased to 2.5, 5.0, and 10.0% by pure oleic acid (AOCS, 2000). The catalyst-methanol solution was mixed with 100 g high FFA OO. The reaction was carried out for 9 min in a domestic microwave oven (ME3410W, Samsung Malaysia Electronics, Kuala Lumpur, Malaysia) at 500 W with a wave frequency of 2450 MHz, equipped with a condenser (to condense back the methanol escaping from the reaction mixture). Also, the magnetic stirrer operated at 60 ºC for 120 min at 600 rpm as a conventional-assisted esterification (CAE). At the end of the reaction time, the reaction mixture was immediately cooled down to room temperature. The water-soluble components were separated from the product in a separatory funnel. The esterification yield was calculated according to the following equation (Chai et al., 2014):

KOH was dissolved in methanol and the mixture was stirred for 10 min to ensure complete mixing. Then, the catalyst-methanol solution was mixed with 100 g esterified HFFAOO. The reaction was carried out using the microwave oven and the conventional magnetic stirrer.

Microwave-assisted transesterification (MAT) of esterified HFFAOO was carried out at 500 W, methanol-to-oil mole ratio of 9, KOH concentration of 1.2%, and reaction time of 9 min. Conventional-assisted transesterification (CAT) was performed using a magnetic stirrer at 60 ºC for 120 min at 600 rpm. To evaluate the effect of esterification pre-treatment on transesterification yield, OO samples (with 2.5, 5.0, and 10.0% FFA) were also transesterified without esterification pre-treatment.

At the end of the reaction time, the reaction mixture was immediately cooled down to room temperature and transferred to a separating funnel where it was left overnight to separate into two phases. The crude methyl ester remained in the upper phase, while the catalyst and unreacted methanol were situated in the lower glycerol phase, meaning that small amounts of catalyst, methanol, and glycerol were present in the upper phase. The excess methanol in the ester phase was distilled off with a magnetic stirrer equipped with a condenser at 70 ºC for 30 min at 450 rpm. Hot distilled water (60 ºC) was sprayed over the surface of ester phase to remove the impurities and catalyst. Washing was performed 5 times to remove all dissolved catalysts and glycerol in the ester phase. The lower phase was discarded and a yellow-colored phase containing FAMEs was finally isolated. FAMEs were then dried using a magnetic stirrer at 80 ºC for 30 min at 250 rpm (Kanitkar et al., 2011; Patil et al., 2011). Margaric acid methyl ester (Methyl margarate; Methyl heptadecanoate) was added to the crude FAMEs as an internal standard (Figure 1) (Atapour and Kariminia, 2011). The purity of FAMEs was determined using the GC/FID method according to the method described by Golmakani et al., (2012a; 2012b). The weight of yield, methyl ester purity, and final yield were calculated according to the following equations (Patil et al., 2011; Hsiao et al., 2010):

Kinematic viscosity, refractive index, and density of purified FAMEs were analyzed according to the method of the American Society for Testing Materials (ASTM; D445), AOCS Cc7-25 Official Method, and AOCS 1a-64 Official Method, respectively (AOCS, 2000; ASTM, 2013). The fatty acid composition of FAMEs was determined according to the method described by Golmakani et al., (2012a; 2012b). Color attributes (L*, lightness; a*, greenness to redness; b*, blueness to yellowness) of FAMEs were analyzed using the Habibi et al. (2016) method.

All experiments were performed in triplicate and the data were reported as mean values of the measurements. Standard deviation values are presented in all tables. A general linear model (GLM) procedure from SAS (Statistical Analysis Software, version 9.1; SAS Institute Inc. Cary, NC) was used for comparison of mean values.

The effect of different esterification and transesterification methods on the fatty acid composition of the biodiesel produced are shown in Table 3. There were no significant differences among the fatty acid compositions of different initial FFA concentrations or different esterification and transesterification methods. Therefore, transesterification rate is not dependent on the chain length or saturation degree of fatty acids. This phenomenon can be related to the speed and completeness of the transesterification reaction. In contrast to our finding, Stavarache et al. (2007) noted that saturated fatty acids, which are commonly located in the 1 and 3 positions of triacylglycerols, had a higher transesterification rate.

The effect of different esterification and transesterification methods on the physical properties of the biodiesel produced are shown in Table 4. The kinematic viscosity of FAME is one of the most important properties which can affect storage condition, transportation, and biodiesel operation. HFFAOO samples transesterified without esterification had higher kinematic viscosity than those of esterified samples (both CAT and MAT methods). The kinematic viscosity of unesterified samples with 5.0 and 10.0% FFA after transesterification was similar to untreated OO (i.e. without esterification and transesterification; 36.806 mm²/s). This phenomenon can be related to the fact that although most triacylglycerols participated in the saponification reaction instead of transesterification, transesterification was not affected by saponification in the case of OO with 2.5% FFA. Also, in spite of the fact that the transesterification method had no significant effect on kinematic viscosity, the MAT of OO with 2.5% FFA had significantly higher kinematic viscosity than the CAT method.

According to ASTM D6751 Official Methods, the kinematic viscosity of FAMEs should be 1.9-6 mm²/s. Except for the FAMEs of OO with 5.0 and 10.0% FFA transesterified without esterification, all the FAMEs fall within the standard range indicated by ASTM.

In the case of esterified samples, MAT samples had lower kinematic viscosity than those of the CAT method. By increasing the initial FFA concentration from 2.5 to 10.0%, the kinematic viscosity of FAMEs increased due to soap formation. There was a significant negative correlation between kinematic viscosity and final yield (R² = 0.972; kinematic viscosity = (-0.065 × Final yield) + 10.413). During the transesterification process, triacylglycerols are converted to lower molecular weight FAMEs (Talebian-Kiakalaieh et al., 2013). Hence, by increasing the biodiesel final yield, kinematic viscosity decreased accordingly. The sample with the highest final yield (98.93%) showed the lowest kinematic viscosity (4.02 mm²/s). In both esterification and transesterification steps, the microwave heating method was more effective than the conventional heating method due to its higher yield of FAME production and accordingly lower kinematic viscosity.

For both CAT and MAT methods, HFFAOO samples transesterified without esterification had lower density in comparison with esterified samples. In the case of unesterified samples, although the transesterification method of OO with 5.0 and 10.0% FFA had no significant effect on density, microwave-assisted transesterified OO with 2.5% FFA had significantly higher density than the CAT method.

In the case of transesterified samples, the MAT of OO showed higher density than that of the CAT method. By increasing the initial FFA concentration, the densities of FAMEs were decreased. There was a significant positive correlation between density and final yield (R² = 0.98; Density = (0.299 × Final yield) + 857.720). The lower molecular weight of FAMEs in comparison with their corresponding triacylglycerols plays a key role in determining the density of the product (Motasemi and Ani, 2012). Therefore, by increasing final yield (i.e. higher production of lower molecular weight FAMEs), the density of the produced biodiesel increased. Using MAT led to higher final yield and hence density.

The refractive indices of HFFAOO samples transesterified without esterification were higher than those of esterified samples using both CAE and MAE methods. Also, different transesterification methods had no significant effect on refractive indices of OO with 2.5, 5.0, or 10.0% FFA.

In the case of esterified HFFAOOs, microwave-assisted transesterified samples had lower refractive indices. By increasing initial FFA concentration, the refractive indices of FAMEs increased. There was a significant negative correlation between refractive index and final yield (R² = 0.982; Refractive index = (-0.0001 × Final yield) + 1.465). By increasing final yield, refractive index decreased due to the conversion of triacylglycerols to lower molecular weight FAMEs. Also, microwave-assisted transesterified samples showed lower refractive indices according to their higher final yield.

Color is a qualitative parameter for evaluating the effects of the transesterification process on the biodiesel produced. HFFAOO samples transesterified without esterification had lower L*, a*, and b* values than those of esterified samples using either conventional or microwave methods. Regarding unesterified samples, although different transesterification methods of OO with 5.0 and 10.0% FFA had no significant effect on L*, a*, and b* values, microwave-assisted transesterified OO with 2.5% FFA had significantly higher L* value and lower b* value than those of the CAT method. Also, the a* value for OO with 2.5% FFA was similar in both MAT and CAT methods.

In the case of esterified samples, microwave-assisted transesterified samples had higher L* values and lower a* and b* values than those of conventional-assisted transesterified samples. Although there were no significant differences among different initial FFA concentrations in terms of L* and a* values, microwave-assisted transesterified OO with 10.0% FFA had significantly higher b* values than those of OO with 2.5 and 5.0% FFA. There was a positive correlation between the L* value and final yield (R² = 0.948; L* value = (0.152 × Final yield) + 46.597). Also, there was a negative correlation between the a* value and final yield (R² = 0.803; a* value = (-0.044 × Final yield) - 4.889) and between the b* value and final yield (R² = 0.985; b* value = (-0.4175 × Final yield) + 77.45). Therefore, although the L* value increased by increasing final yield, both a* and b* values decreased. The microwave heating method showed the highest L* value (61.50) and at the same time the lowest a* and b* values (9.25 and 36.50, respectively).

The physical properties of the FAMEs produced are meaningfully related to their transesterification final yield. Therefore, the physical properties of FAMEs can be used as reliable indices for predicting transesterification yield.