Introduction

The gradual depletion of fossil fuels and moreover, the gradual increase in energy demand have stimulated interest in alternative sources of energy over the last few decades [1]. Moreover, the consumption of fossil fuels has been causing serious environmental pollution by generating carbon dioxide and leading to greenhouse effects [1]. Waste cooking oil (WCO) is a cheap and renewable source for fuel production, and hence its utilization makes it a prudent source for production of biofuel [2]. In biofuel production, more than 80% of the overall cost belongs to feedstock [3]. Hence, utilization of waste vegetable oils as a feedstock not only resolves the economics of biofuels, but also contributes a lot to the reduction in waste disposal [4]. Biofuels are liquid or gaseous fuels that can be generated from renewable sources [5]. The production of biofuels follows numerous routes depending on the type of feed and final product [2]. These routes can be thermochemical conversions (pyrolysis/thermal cracking, combustion, gasification) and biochemical methods [6]. Thermal cracking is one of the likely routes which produces biofuels that are analogous to petroleum derivatives [7, 8].

To understand the features of thermal cracking and to compute kinetic parameters, kinetic studies become a vital tool [9,10,11,12,13,14]. Currently, the application of thermo-analytical techniques such as thermogravimetric analysis (TGA) and differential thermogravimetry (DTG) under non-isothermal condition are popularizing in this respect. TGA gives mass loss as a function of time and temperature. The mass loss data generated from TGA are the important source to investigate the thermal degradation of different materials [15,16,17,18,19]. The rate of mass loss of a sample is predicted from the first derivative of TG and is proportional to the area under the curve [20]. The most important parameters that affect the degradation behavior of oil samples are heating rates, amount of sample, atmospheric condition (reactive or inert) and final decomposition temperature. Moreover, different kinetic models have been developed by several researchers [21,22,23,24,25,26,27] to compute kinetic parameters based on TGA data. In the present study, thermal degradation kinetics of WCO sample was investigated using various iso-conversional methods such as Friedman (FM), Flynn–Wall–Ozawa (FWO) and Kissinger–Akahira–Sunose (KAS) methods which are widely used to calculate the activation energy of degradation of different vegetable oils [28]. Further, thermal cracking of the sample in a laboratory-scale semi-batch reactor was studied to understand the effect of process parameters on the yield of organic liquid product (OLP).

Materials and methods

Materials

WCO was collected from Indian Institute of Technology Guwahati student cafeteria, Assam, India. After simple filtration process, the WCO passed through anhydrous sodium sulfate to remove the traces of water. The pretreated WCO showed the acid value of 3 mg KOH g−1 and density 752 kg m−3. The main fatty acids present in WCO are myristic acid (C14:0): 0.5%, palmitic acid (C16:0): 20.4%, stearic acid (C18:0): 4.8%, oleic acid (C18:1): 52.9%, linoleic acid (C18:2): 13.5%, linolenic acid (C18:3): 0.8% and eicosanoic acid (C20:0): 0.3% [29]. Chemicals such as deuterated chloroform (CDCl3) and methanol [CH3OH (HPLC grade)] were purchased from Sigma-Aldrich Pvt. Ltd., India.

Thermal cracking

Thermogravimetric analysis (TGA) of WCO was conducted using thermogravimetric analyzer (Model no: TGA 851e/LF/1100). Approximately, 10 mg of WCO sample was used for each experiment. The temperature was raised from 30 to 800 °C at various heating rates (10, 20, 30, 40 and 50 °C min−1) under an inert atmosphere. The flow rate of the gas (99.99% purity N2) was maintained at 60 mL min−1.

Further, laboratory-scale thermal cracking of WCO was carried out in a semi-batch reactor (250 mL capacity). For each experiment, 30 g of WCO was placed inside the reactor under a constant flow of gaseous nitrogen with a flow rate of 60 mL min−1. The reactor was heated externally by an electric furnace with PID controller to achieve final temperature of 400, 425, 450, 475 and 500 °C, at heating rates of 10 and 20 °C min−1 and kept under isothermal conditions for the duration of 60–300 min. A condenser connected with flask was assembled to collect condensable vapor (i.e., OLP) in the flasks (Fig. 1). The yield of product was calculated using (Eq. 1)

$${\text{Yield}}\;{\text{of}}\;{\text{OLP}}\; ( {\text{mass\% )}}\; = \;\left( {\frac{{{\text{Mass}}\;{\text{of}}\;{\text{OLP}}\;{\text{collected}}}}{{{\text{Mass}}\;{\text{of}}\;{\text{feed}}}}} \right)\; \times \;100\%$$
(1)
Fig. 1
figure 1

Thermal cracking setup; Notations: A: condenser, B: OLP yield, C: feed, D: furnace, E: reactor temperature, F: nitrogen gas

Full size image

Characterization of WCO and OLP

Some of the physicochemical properties such as density, kinematic viscosity and calorific value of WCO and OLP were estimated using standard ASTM methods. Specific gravity bottles of 25 mL were used to measure the volume of sample, and density was calculated as the ratio of mass to volume of a sample. On the other hand, the viscosity of the fuel is an important property that affects spray properties of the fuel which in turn affect engine performance. Kinematic viscosity was measured using U-tube viscometer. It was calculated by multiplying measured average time with a constant of U-tube viscometer. The average time is the time at which a known volume of liquid flowing under gravity to pass through a calibrated glass capillary viscometer tube [30]. The calorific values of WCO and OLP were determined using bomb calorimeter in which 0.5–1 g of samples was placed in the bomb calorimeter crucible. When the temperature becomes constant, the bomb was fired and the rise in temperature was recorded to calculate calorific value according to ASTMD 2015-85. Flash point was determined by pouring about 100 mL of the samples into the flash cup, and the lighted test flame was passed along the circumference of the cup. When a flash appeared at any point on the surface of the sample, the temperature was recorded, and that temperature was considered as a flash point according to ASTM D 92.

Fourier-transform infrared spectroscopy (FTIR) (Shimadzu Model: IRAffinity1) was used for identification of functional groups present in WCO and OLP with the aid of attenuated total reflectance (ATR). Approximately, 10 mg of the samples was injected on ATR crystal and the spectra were generated by the computer software in the range of wave number 400–4000 cm−1. FTIR spectra of the samples were used for further analysis.

Nuclear magnetic resonance spectrometer analyses (NMR) of WCO and OLP were performed on Bruker Ascend ™ 600 MHz. Approximately, 10 mg of each sample was mixed with deuterated chloroform (CDCl3), and the mixture was placed in the NMR tube. The OLP sample used in the chromatographic analysis was obtained under optimum reaction parameters of temperature 475 °C, residence time 120 min, catalyst loading 4 mass% and heating rate 10 °C min−1. The chemical shifts are reported in ppm, using solvent proton signal as standard. The chemical shift of CDCl3 observed at 7.2 ppm was taken as a reference. Further, gas chromatography–mass spectroscopy (GC–MS) (Model no: 450-GC, 240-MS; Make: M/s Varian) was used for identification of organic fractions present in the OLP sample. A capillary column (VF-5 ms) of 50 m × 0.2 mm (0.33 μm film thickness) dimension was used. About 1 µL of the sample was introduced manually using a syringe into a gas chromatograph for analysis. Nitrogen was used as a carrier gas, and the following temperature program was set up for analysis. The oven was heated from temperature 70 °C (2 min) to 300 °C (7 min) at a heating rate of 10 °C min−1 to an injector temperature of 250 °C and holding time of 3 min. The identification of chemical compounds was performed on GC–MS, and the generated mass spectrum (peaks) was matched with a National Institute of Standards and Technology (NIST) research library.

Kinetic parameter calculation

The thermal decomposition rate of the sample could be expressed by considering Arrhenius temperature dependency at a constant heating rate using Eq. 2 [31]. The activation energy (E) for thermal decomposition of WCO was calculated using a model-free approches such as Friedman (FM) (Eq. 5) [27], Flynn–Wall–Ozawa (FWO) (Eq. 7) [22] and Kissinger–Akahira–Sunose (KAS) (Eq. 6) [21] methods. For all methods, numerous TGA curves at different heating rates for the same value of conversion were used to determine the activation energy.

$$\frac{{{\text{d}}\alpha }}{{{\text{d}}t}} = \frac{A}{\beta }\;{ \exp }\;\left( { - \frac{E}{\text{RT}}} \right)f(\alpha )$$
(2)

where β, f(α), A, α, E, T and R refer to a heating rate (°C min−1), reaction mechanism model, pre-exponential factor (1 min−1), degree of decomposition of the samples, activation energy (kJ mol−1), temperature (K) and gas constant (8.314 J mol−1 K−1), respectively. The value of α at appropriate time/temperature was calculated using TG data (Eq. 3)

$$\alpha \; = \;\left( {\frac{{M_{\text{o}} - M_{\text{t}} }}{{M_{\text{o}} - M_{\infty } }}} \right)$$
(3)
$$\beta = \frac{{{\text{d}}T}}{{{\text{d}}t}} \,$$
(4)

where \(\alpha\) is the mass loss fraction, \(M_{\text{o}}\) is the initial mass of the sample, \(M_{\text{t}}\) is the mass of a sample at time t, \(M_{\infty }\) is the final, non-decomposable mass of the sample after the complete pyrolysis reaction.

FM is a differential iso-conversional method and is directly based on (Eqs. 2, 4) and its natural logarithm to obtain Eq. 5.

$$\ln \;\left( {\frac{{\text{d}\alpha }}{{\text{d}t}}} \right)_{\text{i},{\text{j}}} \; = \;\ln \;\left( {A \, f(\alpha )_{\text{i},{\text{j}}} } \right)\; - \;\frac{{E_{\text{i}} }}{{\text{RT}_{\text{i},{\text{j}}} }}$$
(5)

where i, j represent α and β, respectively. Different degrees of conversion (α) from 0.1 to 0.9 with 0.1 increments were considered to estimate E from the slope of \(\ln \left( {\frac{{\text{d}\alpha }}{{\text{d}t}}} \right)_{\text{i},{\text{j}}}\) versus \(\frac{1}{{T_{\text{i},{\text{j}}} }}\) curve.

KAS method is generated from Coats–Redfern approximation [21]. The general expression for nth-order thermal decomposition of sample is as follows (Eq. 6),

$$\ln \left( {\frac{\beta }{{T^{2} }}} \right)_{\text{i},{\text{j}}} \; = \;\ln \;\left( {\frac{{\text{AR}}}{{E_{\text{g}} (\alpha )_{\text{i},{\text{j}}} }}} \right)\; - \;\frac{{E_{\text{i}} }}{{\text{RT}_{\text{i},{\text{j}}}}}$$
(6)

The plot \(\ln \;\left( {\frac{\beta }{{T^{2} }}} \right)\) versus \(\frac{1}{T}\) at the value of \(\alpha\) was a straight line with slope \(- \frac{E}{R}\).

FWO method is a model-free iso-conversion method. The fundamental expression (Eq. 2) was also used by an integral method of FWO (Eq. 7) using Doyle approximation [22].

$$\ln \;(\beta_{\text{i},{\text{j}}} )\; = \;\ln \;\left( {\frac{{{\text{AE}}_{\text{i}} }}{{R_{\text{g}} (\alpha )_{\text{i},{\text{j}}} }}} \right)\; - \;5.331\; - \;1.052\frac{{E_{\text{i}} }}{{\text{RT}_{\text{i},{\text{j}}} }}$$
(7)

Similar to FM, for different conversion (α) the plot \(\ln \,(\beta )_{\text{i}.{\text{j}}}\) versus \(\frac{1}{{T_{\text{i},{\text{j}}} }}\) is a straight line with a slope for a variable value of \(\alpha\) which is a straight line whose slope is \(- \;1.052\frac{{E_{i} }}{R}\) and E was calculated from the slope of a line.

Results and discussion

Thermogravimetric analysis (TGA)

The thermal decomposition process of WCO was investigated in the temperature range of 30–800 °C at heating rates of 10, 20, 30, 40 and 50 °C min−1 (Fig. 2a). The thermal decomposition of WCO at various heating rates (10, 20, 30, 40 and 50 °C min−1) showed similar degradation pattern with less than 1% residue remained at around 530 °C (Fig. 2a). The main thermal degradation of the sample occurred in the temperature region of 318–500 °C and essentially completed in this zone. The behavior of a thermogravimetric curve with the corresponding DTG peaks gives a clear indication of the number of stages of thermal degradation [28]. Considering a single heating rate (i.e., 10 °C min−1), the thermal decomposition of WCO occurred in the temperature range of 318–500 °C (96.5 mass%). With an increase in the heating rate from 10 to 50 °C min−1, the rate of mass loss (5.3 mass% min−1) was proximately same in the temperature range 318–500 °C. It can also be noticed that increase in the heating rate did not alter the pattern of the thermal decomposition; rather, it shifted the TG curve of WCO degradation to a higher temperature. The slope change on TG and DTG curve indicates the beginning of new step for a thermal degradation sample. Hence, it can be observed that the thermal decomposition of WCO took place in a single stage with similar degradation trends as that of at multi-heating rates (Fig. 2). But, peak temperatures (the rate at which maximum degradation occurred) for the thermal degradation of the samples were varying in their position. This may be because the heating rate is the significant factor on the thermal decomposition of the sample [28]. With an increase in the heating rate from 10 to 50 °C min−1, the maximum rate of mass loss was increased from 17.56 to 96.15% min−1 and also the temperature at which maximum mass loss fraction occurred was shifted from 418.5 to 452 °C (Table 1). Similarly, the initial temperature (T onset) and the final temperature (T offset) for the thermal decomposition of WCO sample were shifted from 318.7 to 362.4 and 478.7 to 496.6 °C, respectively. R50 (the rate at which 50% mass loss) is the measure of the reactivity of samples for the thermal decomposition of the WCO. The values were found to be 17, 38.5, 54 and 73 and 96.5 mass% min−1 at temperatures of 420, 432, 439, 445 and 452 °C at heating rate of 10 °C min−1, 20, 30, 40 and 50 °C min−1, respectively. Generally, the temperature at which rate of mass loss is maximum (M max) was found to be proximity similar to that of R 50.

Fig. 2
figure 2

TGA (a) and DTG (b) curve of WCO decomposition at heating rates of 10, 20, 30, 40 and 50 °C min−1

Full size image
Table 1 Onset, endset and peak temperatures and their respective maximum mass loss of WCO
Full size table

Kinetic analysis

Figure 3a–c shows the iso-conversional plots of FM, KAS and FWO at heating rates of 10, 20, 30, 40 and 50 °C min−1, respectively. Figures indicate that the fitted lines are parallel at different conversions (0.1–0.9) and heating rates (10–50 °C min−1). From the plots, the activation energies for the selected conversion (α) were calculated from the slope of the line (Fig. 3a–c) and obtained values are summarized in Table 2. The values of activation energies obtained were within the range of 212.5–295 kJ mol−1 (FM), 161–255 kJ mol−1 (KAS) and 172–267 kJ mol−1 (FWO). Higher correlation coefficients (R 2) obtained for all methods showed fitness of the model. The activation energy values (Fig. 3d) found to be increased with an increase in the conversion for KAS and FWO methods, but the fluctuation in the activation energy value was observed in FM method which might be because of mathematical errors associated with the method. The mean values of activation energies were found to be 243.7 kJ mol−1 (FM), 211.3 kJ mol−1 (KAS) and 222 kJ mol−1 (FWO). It can be seen from (Table 2) that the average activation energy values calculated for all kinetic methods (FM, KAS and FWO) were in good agreement. Thus, the models might be reliable for fitting the thermal degradation of WCO.

Fig. 3
figure 3

Kinetic fitting of WCO degradation by FM (a), KAS (b), FWO (c) and activation energy distribution (d)

Full size image
Table 2 Values of activation energies obtained using FM, FWO and KAS methods
Full size table

Thermal cracking of waste cooking oil (WCO) using semi-batch reactor

TGA profile of WCO sample (Fig. 2a) confirmed that the active decomposition of WCO occurred in the temperature range of 318–500 °C. Therefore, the laboratory-scale cracking of WCO in a semi-batch reactor was conducted in the temperature range of 400–500 °C with 25 °C increment. The effect of cracking temperatures (400–500 °C), heating rates (10–20 °C min−1) and residence time (60–300 min) on the yield of OLP was investigated. From Fig. 4, it can be observed that as the residence time increases from 60 to 180 min the yield of OLP increased for all the temperature ranges (400–500 °C) and heating rates (10–20 °C min−1). Further, increase in the residence time to 300 min caused the increase in the yield of OLP for 400, 425 and 450 °C temperatures. However, increase in residence time to 300 min did not show significant enhancement in the yield of OLP at higher temperatures of 475 and 500 °C at both heating rates of 10 and 20 °C min−1. This is may be due to secondary cracking of products at higher temperatures [32]. Figure 4a shows the effect of residence time on the yield of OLP for various temperatures at a heating rate of 10 °C min−1. The OLP yield found to be increased from 8 to 29 mass% (400 °C), 12 to 51 mass% (425 °C), 18 to 64 mass% (450 °C), 28 to 74 mass% (475 °C) and 38 to 65 mass% (500 °C) with an increase in the residence time from 60 to 300 min at a heating rate of 10 °C min−1. From Fig. 4a, it can be seen that the OLP yield was increased continuously from 8 to 29, 12 to 51 and 18 to 63 mass% at cracking temperature of 400, 425 and 450 °C, respectively, with an increase in the residence time from 60 to 300 min at a heating rate of 10 °C min−1. However, the OLP yield was increased from 28 to 60 mass% (475 °C) and 38 to 61 mass% (500 °C) for the residence time of 60–120 min and further the OLP yield slightly increased in case of longer residence time and higher cracking temperatures (475 and 500 °C). Similarly, for 20 °C min−1 the yield of OLP was increased from 4 to 26 mass% (400 °C), 9 to 44 mass% (425 °C), 14 to 57 mass% (450 °C), 27 to 72 mass% (475 °C) and 46 to 65 mass% (500 °C) as residence time increases from 60 to 300 min. Furthermore, from this study, it is concluded that for all studied cracking temperatures (i.e., 400–500 °C), the higher yield of OLP was obtained at a heating rate of 10 °C min−1 as compared to 20 °C min−1. These results attribute that high heating rates favor the yield of non-condensable gases [33]. The values of the yield of OLP obtained during thermal cracking of WCO were comparable with thermal cracking of soybean oil [34, 35] and rapeseed oil [36]. Overall, the optimum yield of 72 mass% of OLP was obtained at a residence time of 180 min and thermal cracking temperature of 475 °C at a heating rate of 10 °C min−1.

Fig. 4
figure 4

Effect of time on yield of OLP for heating rates of 10 °C min−1 (a) and 20 °C min−1 (b)

Full size image

Characterization of WCO and OLP

During thermal cracking of WCO, several reactions such as decarbonylation (CO removal), decarboxylation (CO2 removal) and dehydration (H2O removal) were observed which resulted in the formation of CO2, CO and H2O as evolved products [37]. As a result, the ratio of carbon to oxygen atoms is higher in OLP as compared to WCO. For this reason, the higher calorific value was attained for OLP (39.25 MJ kg−1) as compared to WCO (38.9 MJ kg−1). The density of WCO and OLP was found to be 0.96 and 0.86 g cm−3, respectively. The decrease in density (0.86 g cm−3) and viscosity (3.79 mm2 s−1) of the produced OLP were due to the removal of oxygen and cracking of higher molecular weight triglyceride of WCO into smaller molecules. The viscosity values were found to be in the range of those reported in the literature [38, 39]. The density of OLP obtained from thermal cracking of WCO was similar with OLP obtained from carinata oil (0.86 g cm−3) [40]. However, the density of OLP (0.86 g cm−3) estimated in the present study was lower than OLP (0.88 g cm−3) reported by Li et al. [39]. Moreover, the flash point is the lowest temperature at which the sample evaporates enough fluid to form a combustible gas. It indicates easy burning of the sample in air. It can be claimed that the flash point of OLP obtained in this study was measured by ASTM standards and found to be 97 °C.

After the collection of condensable liquid during the cracking reactions, the collected OLP was subjected to GC–MS and FTIR analysis for identification of hydrocarbon fractions depending on their boiling range, i.e., gasoline (60–80 °C), kerosene (80–120 °C) and diesel (120–200 °C) fractions [41]. From GC–MS analysis of OLP shown in Table 3, it can be observed that the identified fraction contains hydrocarbon chains compounds between C6 and C19. Majority of hydrocarbon fractions found in this study ranges from C7 to C14. This study confirms the absence of aromatic hydrocarbons, olefinic and paraffinic hydrocarbons which were found to be the dominating fractions, and the results are comparable to those reported in the literature [42].

Table 3 Main chemical compounds present in OLP
Full size table

From Fig. 5, the FTIR spectra of WCO and OLP are found to be approximately similar which could be due to the similar nature of the chemical structure of functional groups present in the sample. But, only the characteristic peak at 1142 cm−1 which represents a stretching vibration of C–O ester functional group in WCO sample. Symmetric (C–H) and asymmetric stretching of the saturated C–C bond was observed at wavenumber 2915–2918 and 2847–2855 cm−1 for both samples, respectively. The stretching vibration at 1454 cm−1 corresponds to bending vibrations of CH2 and CH3 aliphatic groups. The band at 719–720 cm−1 belongs to the overlapping of methylene (–CH2) rocking vibration and to the out-of-plane vibration of cis-disubstituted olefins [43]. Ester carbonyl functional group of the esters of glycerol (triglycerides) in WCO and fatty acid of esters in OLP sample were represented at 1705–711 cm−1. Similar observation was reported for waste palm cooking oil and rapeseed oil [43, 44]. From the overall FTIR spectra, it was observed that OLP contains essential functional groups which confirm that the product comprises aliphatic hydrocarbons.

Fig. 5
figure 5

FTIR analysis of WCO and OLP

Full size image

The chemical shifts and the relative hydrogen distributions in WCO and OLP obtained from 1H NMR spectroscopy of the samples (Fig. 6) are listed in Table 4. The chemical shift regions for 1H NMR can be grouped into three major regions: aromatic (9–6 ppm), olefinic (6–4 ppm) and aliphatic (3–0.5 ppm) fractions [45, 46]. However, the aliphatic proton section has been further subdivided into Hα (2–3 ppm), Hβ (1–2 ppm) and Hγ regions (0.5–1 ppm). Table 4 signifies that both WCO and OLP did not contain fractions which belong to the aromatic range. Most of the characteristics peaks were in the range of 5–1 ppm for WCO and 2.5–0.76 ppm for OLP. In the 1H NMR spectrum of OLP, chemical shifts between 0.76 and 1.38 ppm indicate hydrogen bonded to carbon, which confirmed the existence of aliphatic hydrocarbons, as well as the chemical shifts between 2.12 and 2.40 ppm indicate the presence of hydrogen and unsaturated fatty acids (=CH– methine) groups, respectively. From the overall study of 1H NMR, the dominating chemical shifts for OLP indicated that all the compounds in the sample correspond to aliphatic hydrocarbons and some traces of aliphatic carbonyl hydrocarbons.

Fig. 6
figure 6

1H NMR of WCO and OLP (Notations for A–J shown in Table 4)

Full size image
Table 4 Relative hydrogen distributions in OLP
Full size table

Conclusions

The application of WCO for biofuel production would be a promising option to secure future energy demands. Moreover, the thermal kinetics and thermal cracking are the complementary studies in the present work. The TGA graph clarifies that with an increase in heating rates, the maximum mass loss rate increased and moved toward higher temperatures. From the graph, the sample described decomposition peak profiles that confirmed single decomposition stage and the main decomposition temperature was established between 318 and 500 °C in which more than 96.7 mass% of the sample was completely decomposed. Results obtained from TGA were used for activation energy calculation by non-isothermal, model-free kinetic methods such as FM, KAS and FWO within the conversion range of 0.1–0.9 from the slope of the lines. The activation energy obtained for WCO by FM, KAS and FWO was 243.7, 211.23, and 222 kJ mol−1, respectively. From the result, it can be claimed that these methods were in good agreement and can be used to estimate the activation energy of WCO degradation. Further, thermal cracking of WCO oil was performed at a temperature range of 400–500 °C in which the optimum yield of 72 mass% OLP was obtained at a temperature of 475 °C, with a heating rate of 10 °C min−1 and residence time of 180 min. It has been observed from 1H NMR to GC–MS that the majority of fractions present in OLP corresponds to aliphatic hydrocarbons.