Grupo BIOSEL supplied the residual oil (RO) used in this research.

FAME from residual oil were obtained by esterification (with or without previous triglyceride hydrolysis) of free fatty acid and by transesterification of the remaining triglycerides (TG).

Hydrolysis. In order to hydrolyze and transform the oil triglycerides into free fatty acids, lipases from Candida rugosa were employed. The hydrolyses were conducted in a 500-cm³ conical flask with vigorous shaking in a standard orbital shaker. The procedure is summarised as follow: 100 cm³ of oil sample were heated at 30 °C for 30 min. After the addition of the desired amount of lipase dissolved in 10 cm³ of deionized water (milli-Q), the mixture was vigorously shaken. The conditions assayed included an enzyme concentration between 0.6 and 1.2 g dm⁻³ acting at pH 7, 30 °C temperature, 1 atm pressure, water-to-oil molar ratios from 1:2 to 4:1 (v/v) (Table 2) and 24 h reaction time. These conditions were chosen because they are in the range of more suitable conditions for TG hydrolysis pointed out by other authors (Chowdhury et al., 2016). Subsequently, the samples were centrifuged and the oil phase was separated from the aqueous phase and the acidity index of the oil phase was determined. The yield of the hydrolysis reaction (η_H) was defined as follow:

Where AI_i is the initial acidity index (60.5 %) and AI_f stands for the acidity index at the end of the hydrolysis process.

Esterification. The esterification reaction of FFA contained in 100 cm³ sample was carried out in a 500-cm³ stirred tank reactor, provided with a Heidolph RZR 2052 stirrer and a coolant to avoid losses in methanol. The working conditions were 60 °C temperature, 700 rpm stirring and 2 h reaction time. Different types of homogeneous (HCl and H₂SO₄) and heterogeneous (commercial Amberlyst-15 supplied by The Dow Company, Midland, Michigan, United States, heavy-metal-contaminated pyrolyzed roots from Jatropha curcas L., and pyrolyzed roots from Jatropha curcas L. treated with H₃PO₄, KOH or CO₂) catalysts were assayed. Different weight percentages of catalyst (3, 5, 8, 10 and 15% wt.) and various methanol-to-oil molar ratios (2:1, 8:1, 10:1, 15:1 and 20:1) were also assayed. After esterification, the acid catalyst used in the esterification was neutralized with a NaOH in ethanol solution. The esterification yield (η_E) was calculated as follow:

where AI_i is the initial acidity index (60.5 %) and AI^f stands for the acidity index at the end of the esterification.

Transesterification. For the transesterification of the resulting mixture from the esterification reaction, a 1-dm³stirred tank reactor was used. The initial working conditions were: 60 °C temperature, 700 rpm stirring, 1 h reaction time, 8:1 methanol-to-oil molar ratio and 1% (wt.) NaOH in relation to the oil as catalyst. These conditions were chosen based on our previous research (Pereda Marín et al., 2003; García-Martín et al., 2018, 2019a). After the transesterification reaction, a decanting step was carried out to separate impurities from FAME. For this purpose, the mixture was transferred to a separating funnel and left to decant until a complete separation into two phases was observed. In the upper phase, the biodiesel was found together with methanol, water and soaps, while the lower phase contained a mixture of glycerine, methanol, soaps, salts and water. Once the two phases were separated, the methanol was removed from the biodiesel phase using a Rotavapor-R and a vacuum pump Diaphragm Vacuum Pump GM100 at room temperature. Finally, a filtration step was carried out using a mesh filter containing oak chips with alumina, as described elsewhere (García-Martín et al., 2018, 2019a), which allows the passage of the methyl esters while retaining the rest of the soaps, glycerol, salts and water.

Biocatalysts were the resulting biochars from the pyrolysis of roots from Jatropha curcas L. plants grown in non-contaminated soils and treated with H₃PO₄ (Biocatalyst 1), KOH (Biocatalyst 2) or CO₂ (Biocatalyst 3), and from the pyrolysis of roots from J. curcas L. Plants, used for the phytoremediation of contaminated soils with different heavy metal concentrations (Biocatalyst 4 and Biocatalyst 5).

Biocatalysts from non-contaminated J. curcas L. roots. Biocatalysts 1 and 2 were chemically activated by means of a treatment with an acid or a base, respectively, prior to the carbonization process. In this process, the root samples were impregnated with H₃PO₄ (Biocatalyst 1) or KOH (Biocatalyst 2) in a 1:2 mass ratio and dried in an oven at 60 °C for 22 h. In order to eliminate the excess of acid or base, the samples were washed with water until reaching a pH value close to 7. The solids obtained after washing were again dried in an oven at 60 °C for 22 h (Bastidas et al., 2010). Subsequently, the pyrolysis of both samples was carried out at 600 °C for 2 h with a heating rate of 10 °C·min⁻¹ and an argon flow rate of 150 cm³·min⁻¹.

Biocatalyst 3 was obtained by direct physical activation of the plant roots, combining the process of thermal carbonization and the activation by oxidation using CO₂ as oxidizing agent in the same step. This thermal treatment of reduction was accomplished by the treatment of the sample with CO₂ at 600 °C. The oxidizing agent reacts with the carbon atoms generating a porous structure. The oxidizing atmosphere brings about changes in the chemical and physical properties of coal. Plasticity is physically eliminated, the formation of a porous texture is favored, and the oxygenated groups are chemically increased with a decrease in the aliphatic groups that are the most vulnerable structures in the oxidation (Uribe et al., 2013). Subsequently, pyrolysis was carried out for 2 h at the same temperature (600 °C) under argon atmosphere at a flow rate of 150 cm³·min⁻¹ with the aim of improving the adsorbent capacity of the biochars.

Biocatalysts from heavy-metal-contaminated J. curcas L. roots. The dried and sieved roots of J. curcas L. plants that were previously used for the phyto-remediation of 2 heavy-metal contaminated soils (Álvarez-Mateos et al., 2019) were subjected to pyrolysis to obtain 2 biocatalysts. Biocatalyst 4 was obtained from roots containing 31.8, 39.2, 1118.3, 13.7, 16.4, 44.7 and 22.4 mg·kg⁻¹ Cr, Mn, Fe, Ni, Cu, Zn and Pb, respectively, while the roots used for the synthesis of Biocatalyst 5 contained 45.9, 69.7, 1841.6, 21.9, 22.4, 124.1 and 57.1 mg·kg⁻¹ Cr, Mn, Fe, Ni, Cu, Zn and Pb, respectively.

Pyrolysis of 0.5 g of each root took place inside a 25 × 300 quartz tube placed in a Carbolite Tube Furnace MTF 12/38/250 equipped with a Eurotherm 2416CG temperature controller. The outlet of the quartz tube was connected to two bubblers submerged in ice to condense the flue gases (bio-oils), with an extractor hood to eliminate the uncondensed gases. The N₂ flow was set to 6 dm³·h⁻¹, and the heating rate to 30 °C·min⁻¹ until reaching 550 °C, which was maintained for 2 h. The resulting biochars had graphite-like structure based on Raman spectra analysis (Álvarez-Mateos et al., 2019) and have shown a great performance in the catalysis of similar reactions such as glycerol esterification with acetic acid or acetic anhydride to obtain oxygenated fuel additives (García-Martín et al., 2019b).

Free fatty acids were quantified in the starting residual oil and after each reaction using the acidity index (AI) parameter (UNE-EN 140140). The AI was determined by titration with an ethanolic 0.1 N KOH solution and expressed as oleic acid percentage (García Martín et al., 2019c). All measurements were performed in triplicate.

The composition of the starting RO and the samples resulting from the hydrolysis, esterification and transesterification reactions, in terms of glyceride compounds, was analyzed by high performance size exclusion chromatography (HPSEC). This technique consists of the separation of the compounds according to their molecular size. The elution order was as follows: Polymers of triglycerides (trimers and dimers), triglycerides (TG), diglycerides (DG), monoglycerides (MG) and finally FFA or FAME. For their quantification, a liquid chromatograph Hewlett Packard 1050 working with an isocratic flow rate of 0.7 cm³·min⁻¹ of tetrahydrofuran (THF) was used. The equipment was provided with a manual rheodyne injector with a 20 μL loop, a column Agilent PL-gel 3 μm (size of pore 100 Å) and it was connected to a refractive index detector Merck L-7490. The sample concentration was 50 mg·cm⁻³ THF. The data was processed using the 32 Karat program (Beckman Coulter, Inc.). The total time of the chromatographic analysis was 15 min.

The FAME content in samples was obtained following UNE-EN 14103:2011 and UNE-EN ISO 12966-1:2015 standards. Briefly, 50 mg of WCO were dissolved in 2 cm³ heptane and then transesterified using a 0.3 cm³ 2 N methanolic KOH solution. After decanting, the supernatant was collected and FAME percentages were analyzed in an HP 5890 series II gas chromatograph equipped with a SP2380 capillary column (60 m × 0.25 × 0.25 µm). The initial column temperature was 185 °C and the ramp rate was 3 °C min⁻¹ until reaching 220 °C. The injection was operated in splitless mode, the injector and detector temperatures were 225 °C and 250 °C, respectively. FAME were identified by mass spectrometry, comparing the spectra with those in a database for this type of compound (Wiley, NIST), and their concentration was calculated using methyl heptadecanoate as internal standard.

The texture characterization of the biochars was carried out using the BET method to determine the specific surface area and by the adsorption/desorption isotherms to study the pore structure. To do this, a Micrometrics Gemini V-2365//V.1.00 adsorption equipment as well as a Micrometrics TriStar II adsorption/desorption equipment, respectively, were used. The biochars were degassed at 250 °C for 24 h under N₂ atmosphere.

Hydrolysis was carried out to increase the number of free fatty acids in the residual oil by breaking down the triglyceride molecules, so a larger amount of FFA can be subsequently converted into FAME by esterification. If the TG hydrolysis were complete, the subsequent production of biodiesel would be performed via esterification of FFA without the need to resort to TG transesterification. For the hydrolysis of triglycerides to be carried out it is necessary to use a catalyst. In our case, lipases from the yeast Candida rugose were used. The experimental parameters assessed were the amount of lipase added (0.06, 0.08 and 0.12% wt.) and the H₂O:RO (v/v) ratio (from 4:1 to 1:2). In previous studies it was found that the maximum TG conversions were 93% for 1:2 (v/v) ratio after 10 h reaction, 70% for 4:1 (v/v) ratio after 24 h reaction, and 92% for 4.12:1 (v/v) ratio after 30 h reaction (Chowdhury et al., 2016). Based on these results, we assayed H₂O:RO ratios (v/v) within a similar range (from 4:1 to 1:2), fixing the reaction time to 24 h. To assess the most suitable conditions for the hydrolysis reaction, the acidity index was chosen as the parameter to be followed.

Table 2 illustrates all the hydrolysis reactions carried out and the different reaction conditions. It shows the amount of lipase added as catalyst, the H₂O:RO ratio used, the final acidity of the oil and the yield (η_H) of the hydrolysis reaction.

The hydrolysis with 1:1 H₂O:RO (v/v) ratio, for the same lipase concentration (0.08% wt.) was the one that achieved the highest acidity index (85.4%). Subsequently, lipase concentrations above and below 0.08% wt. were assayed (0.12 and 0.06% wt., respectively), obtaining average percentages of acidity of 72.2 and 70.7% wt., respectively. Therefore, the most suitable hydrolysis conditions were 0.08 percentage by weight of lipase with respect to the oil and 1:1 H₂O:RO (v/v) ratio, under which an acidity index of 85.4% and reaction yield of 62.9%, respectively, were achieved. The resulting sample contained 85.4% FFA and 14.7% TG, DG and MG.

Similarly to hydrolysis reactions, esterification reactions of FFA were monitored by the acidity index. The experiments were carried out with homogeneous and heterogeneous catalysts. The homogeneous catalysts used were sulphuric acid and hydrochloric acid since other acids such as nitric acid have been reported to achieve lower yields and are economically more expensive (Su, 2013). The heterogeneous catalysts used were Amberlyst-15 and the 5 biocatalysts synthetized from J. curcas L. roots.

Homogeneous catalysts. Different concentrations of H₂SO₄ (wt.) were tested for the same 15:1 methanol:residual oil (MeOH:RO) molar ratio (Table 3). The best performance was achieved with 5% (wt.) H₂SO₄ (run E1). This result is in agreement with that reported by Marchetti and Errazu (2008), who studied the esterification of FFA with methanol at different concentrations of catalyst (1.0–5.2 % wt.), and concluded that 5% (wt.) H₂SO₄ was the optimum catalyst dosage for this type of reaction (Marchetti and Errazu, 2008).

The difference in yield between sulphuric acid and hydrochloric acid was remarkable (Table 3). This is because the hydrochloric acid used was 37% wt. (63% wt. water). As aforementioned, water is the product of esterification and causes equilibrium to shift towards the formation of reactants which inhibit the reaction (Botton et al., 2018).

The reaction products of the three esterifications with sulphuric acid (runs E1, E2 and E3) were analyzed by GC. The percentages of individual methyl esters were similar to those of the initial RO (Table 3). Therefore, and similarly to glyceride transesterification, fatty acid esterification does not alter the fatty acid profile in the resulting methyl esters mixture (García-Martín et al., 2019a). Furthermore, the percentages of FAME in the reaction products obtained by GC analyses (66.0% for run E1, 63.4% forrun E2 and 56.7% for run E3) were in agreement with the esterification yield calculated through the AI parameter and the initial RO composition provided by HPSEC (Table 1).

When combining the most suitable conditions for TG hydrolysis with lipases (run H1, Table 2) and FFA esterification (run E1, Table 3), in spite of producing more methyl esters, the resulting yield of the two-process scheme (run HE1, Table 3) was lower than that of esterification alone (E1). This is probably due to the water generated in the previous hydrolysis because in order to hydrolyze the triglycerides, water was added in a ratio of 1:1 (v/v) with respect to RO, which was not removed from the reaction medium to minimize operational costs, thus reducing the yield of the subsequent esterification by 3%. In a study conducted on the influence of water on the esterification of fatty acids, it was observed that when the concentration of water was 5% wt., the conversion in the esterification reaction decreased by 30% (Botton et al., 2018). On the contrary, when the percentage of water in the reaction medium was greater, about 10–30% wt., the reduction in conversion was lower. These authors pointed out that this phenomenon could be explained by the formation of second phase in the reaction medium at the beginning of the reaction, and in this case both alcohol and water are partially eliminated from direct contact with the enzyme, thus increasing the yield (Botton et al., 2018). In our case the reaction medium contained 50% wt. water, so two phases were probably formed, hence the reduction in yield was only 3%. In this case, a small change in the fatty acid profile was detected (Table 3), mainly towards the increase in saturated fatty acids.

Heterogeneous catalysts. Acid homogeneous catalysts, such as H₂SO₄ and HCl, are corrosive liquids. Alternatives to this type of acid catalysts are heterogeneous catalysts, which minimize environmental problems and can reduce biodiesel production costs (Borges and Díaz, 2012), hence the esterification with Amberlyst-15 and biocatalysts from the pyrolyzed roots of J. curcas L. was assayed. Several works can be found in the literature dealing with the heterogeneous catalysis of FFA esterification reactions of oils with relative low acidity (5–15% wt. FFA) to avoid saponification in the subsequent TG transesterification (Kastner et al., 2012; Hidayat et al., 2015). However, there is a lack of papers concerning the FFA esterification of oils with high acidity such as the raw material used in this research. The main objective of this part of the work was to compare the performance of Amberlyst-15 with that of the five biocatalysts, whose main characteristics are shown in Table 4. As expected, the catalyst treated with CO₂ (Biocatalyst 3) presented the highest specific surface area. The characteristics of our biochars were similar (in the same order of magnitude) to those of other biocatalysts, mainly sulfonated biochars, and ion-exchange resins used for FFA esterification (Örbay et al., 2008; Kastner et al., 2012). In this sense, BET surface area and average pore diameter of coconut shell biochar were reported to be 244.2 m²·g⁻¹and 4.86 nm, respectively (Hidayat et al., 2015).

Firstly, we assessed the esterification performance of Amberlyst-15. It was observed that the greater the amount of catalyst, the higher the yield, i.e. AI decreased, as pointed out by other authors (Özbay et al., 2008). These authors concluded that the best esterification conditions were 15% wt. Amberlyst-15 at 60 °C. Then, using this optimal amount of catalyst, the influence of the MeOH:RO molar ratio was studied. In this case, the higher the molar ratio, the higher the yield. No improvement was observed when increasing the molar ratio from 15:1 to 20:1, so the former one was selected for trials with biocatalysts. Among the biocatalysts, the one that yielded the most was the pyrolyzed roots activated with H₃PO₄ (Biocatalyst 1), which was the one with the highest surface area and the smallest pore size (Table 4), followed by the pyrolyzed roots from phytoremediation of the contaminated soil with the greatest concentration of heavy metals (Biocatalyst 5). One last heterogeneous esterification was assayed with this latter biocatalyst with increased methanol to residual oil molar ratio of 20:1 (Table 5). The esterification performance of Biocatalyst 5 increased, producing a 37.8% reaction yield, yet a little lower than that achieved over Amberlyst-15 under the same conditions (44.4%). All in all, although the yields obtained in the esterification over biocatalysts were lower than those obtained over homogeneous catalysts, they were similar than those achieved with Amberlyst-15 commercial catalyst, which accounts for the potential of these biochars to catalyze reactions.

With the aim of obtaining a complete conversion of FFA and TG into FAME, the RO was consecutively subjected to TG hydrolysis, FFA esterification and transesterification of the remaining TG in the same reactor under the most favourable conditions found for each stage. The conditions for TG hydrolysis and FFA esterification were obtained in this work, while those for TG transesterification were selected based on previous works (Pereda Marín et al., 2003; García-Martín et al., 2018, 2019a). To assess the extent of the reaction, mainly the conversion of TG and DG into FFA, HPSEC was selected as the most suitable technique to monitor the reactions.

The starting RO glyceride content (trimers + dimers + TG + DG) was roughly 38% based on HPSEC data (Table 1). The hydrolysis of these compounds with lipases decreased the glyceride content to 27%, increasing the FFA concentration from 62.4 to 72.8% (Table 1). In spite of the fact that trimers and dimers were completely hydrolyzed, the results indicated that the TG hydrolysis was not effective, since a large quantity of TG (12.6%) and DG (14.9%) remained in the RO without releasing their fatty acids. This result is in contrast with the data obtained in Table 2, in which the AI increased from 60.5% to 85.4% for run H1, and this latter AI value does not match the FFA obtained by HPSEC (72.8%). The subsequent FFA esterification led to a FFA+FAME concentration of 74.4%. Based on the AI data from Table 3, AI for run HE1 was 4.2%, which could indicate that the residual oil contained roughly 70% FAME. Finally, the final transesterification stage (HET1) completely converted the TG (Table 1), but not totally into FFA, since the concentration of DG was significant (18.8%). The incomplete glyceride transesterification and the increase in DG concentration could have occurred because of an inhibition by-product, since the reaction medium contained at least 74% FAME at the beginning of the transesterification, thus displacing the reaction toward the formation of intermediate reaction products such as DG. It could be also due to the remaining FFA from the esterification stage, which form soaps and inhibit the transesterification (García-Martín et al., 2018). The final FAME concentration provided by HPSEC at the end of the 3-stage scheme was 81.2% wt., which was in agreement with the value obtained by GC-MS (84.0 ± 2.1%).