Supercritical carbon dioxide extraction of crude tall oil: an alternative upgrading approach

2.1 Materials

Crude tall oil was provided by the pulp and paper producer Kemira Oyj (Figure 1) The crude tall oil had a highly viscous and crumbly consistency. It was dark brown in colour and had an unpleasant odour. The water content was determined by Karl Fischer titration to 2.0 wt.% and the density was 1,001 kg/m³ at 293.15 K. CO₂ (99.5 %) was obtained from Linde AG in 50 L dip-tube bottles and the CO₂ density was taken from literature data (Anwar and Carrol 2016) based on the pressure and temperature.

Figure 1:

Crude tall oil produced by Kemira Oyi.

2.2 Supercritical CO₂ extraction plant

An extraction plant from HDT Sigmar Mothes GmbH was used for the supercritical CO₂ extraction of the crude tall oil. The plant has already been used in modified ways for the extraction of pyrolysis liquids (Feng and Meier 2015, 2016, 2017; Möck et al. 2023). A simplified schematic structure is shown in Figure 2 and a more detailed structure is published in another study (Möck et al. 2023).

Figure 2:

Flow scheme of scCO₂ extraction plant (simplified).

The liquid CO₂ was compressed using a high-pressure pump (Figure 2; 2) and supercritical state was reached by heating (Figure 2; 5). The scCO₂ flow was measured using a Coriolis mass flow meter (Figure 2; 3) and adjusted via the pump stroke. The scCO₂ flowed through the extraction vessels (Figure 3), which were inserted into the heated extraction chamber. Two different extraction vessels (G1 and G2, Figure 3) with different volumes were used. The extraction vessels were inserted so that they were flush with the top of the extraction chamber. The vessels consist of stainless-steel tubes with an inner diameter of 30 mm. The vessels were sealed with a special plug. The plug was made out of PTFE and a stainless-steel tube that redirected the scCO₂ into the crude tall oil. The vessels, the plug and the assumed flow of scCO₂ through the vessels are shown in Figure 3. Vessel G1 has a volume of 473 cm³ and a length of 710 mm. Vessel G2 has a smaller volume of 185 cm³ and a length of 252 mm.

Figure 3:

Extraction vessels and theoretical flow of the scCO₂ through the vessels (G1 = vessel G1, G2 = vessel G2, P = plug for redirecting the scCO₂).

After filling in the crude tall oil, the upper 2–3 cm of the extraction vessels were stuffed with filter cotton to prevent undissolved tall oil from being discharged and thereby blocking the valve. It is assumed that mixing between scCO₂ and tall oil occurs in the entire volume of the extraction vessel due to the scCO₂ flow. Thereby, the headspace volume, i.e. the unfilled volume above the tall oil, has an influence on the contact time between scCO₂ and crude tall oil. The contact time is also dependent on the density of the scCO₂ and the scCO₂ flow. The initial contact time at the start of extraction can be calculated using Equation (1). During extraction, the headspace volume increases due to the discharge of extract, which slowly increases the contact time between scCO₂ and tall oil.

The scCO₂ and dissolved extract were depressurised to atmospheric pressure via a heated backpressure valve (Figure 2; 9). The CO₂ was separated from the extract in a separator system consisting of two bottles with an insert filled with 2 mm glass beads and a 22 ml vial. The use of a three-way valve made it possible to switch between the two separators and thus enable a simple sampling during extraction. A more detailed description of the separator system has been described by Möck et al. (2023).

2.3 Supercritical CO₂ extraction parameters

Before extraction, the tall oil was homogenised for about 30 min by a rolling board mixer. The extraction of 50 g crude tall oil in the preheated extraction plant was carried out using the two vessels G1 and G2 at two different pressure-temperature combinations. After filling the crude tall oil, the headspace volume, i.e. the unfilled vessel volume above the crude tall oil, was 423 cm³ for vessel G1 and 135 cm³ for vessel G2. The temperatures were 313.15 K and 333.15 K and the pressures (12.5 MPa and 20 MPa) were chosen so that similar scCO₂ densities were achieved (724 and 731 kg/m³). The extractions were carried out over 4 h with a CO₂ flow of 5 g/min at a solvent to feed ratio of 24 g/g. The resulting contact times between tall oil and scCO₂ are 61.3 min at 333.15 K and 61.8 min at 313.15 K for vessel G1. For vessel G2, the contact times were 19.5 min at 333.15 K and 19.8 min at 313.15 K.

In addition, 13.4 g tall oil were extracted with vessel G2 at scCO₂ density of 724 kg/m³ (20 MPa, 333.15 K) with a significantly lower contact time of 8.3 min and a much higher solvent to feed ratio of 269 g/g. The CO₂ flow was 15 g/min and the extraction time was 4 h. The headspace volume was 172 cm³. In total five extractions were carried out in duplicate and the parameters are summarized in Table 1.

2.4 Calculation of the yield, losses and vapor phase loading

A VWR LPW-2103i scale was used for weighing. The residue was determined by weighing the extraction vessel before and after extraction. The amount of extracts collected was determined by weighing the bottle in the separator system, including the insert filled with glass beads. The yield was calculated according to Equation (2).

The vapour phase loading (VPL) indicates how much extract was extracted from the feed with one kg of CO₂. The VPL was calculated according to Equation (3).

2.5 GC-MS/FID

The crude tall oil, the extracts and the residues were characterised by GC-MS/FID. In order to measure the fatty acid content, the samples were methylated online with trimethylsulfonium hydroxide (TMSH). Approximately 15 μl of sample was mixed with dichloromethane to obtain a concentration of 2.5 mg/ml. To 1 ml of this solution 100 μl of a mixture of acetone and fluoranthene (2,000 mg/ml) and 100 μl of TMSH was added. The samples were analysed using an Agilent 6890 GC equipped with an MS 7959 C detector (MSD), FID, and a VF 5-ms column. The exact parameters are given in Table 2.

The evaluation was carried out using Massfinder 4 and NIST. The FID signal was used for quantification. The response factors of α-pinene, camphene, δ-3-carene, palmitic acid, linoleic acid, oleic acid and stearic acid were calibrated against standards. The response factor of the remaining substances was estimated based on the substance group. The amounts of methylated substances were back calculated to the unmethylated state. A chromatogram of the tall oil with the identified peaks is included in the Supplementary material.

3.1 GC-MS/FID characterisation of the crude tall oil

The crude tall oil consists mainly of diterpenoids (70.5 wt.%) and fatty acids (25.0 wt.%). The fatty acids consist of oleic acid (8.9 wt.%), linoleic acid (8.2 wt.%) palmitic acid (2.1 wt.%), stearic acid (0.5 wt.%), behenic acid (0.5 wt.%) and small amounts of lignoceric acid (0.3 wt.%). In addition, five fatty acids (4.6 wt.%) are detected, but they could not be clearly identified. The diterpenoids consist mainly of abietic acid (30.6 wt.%), dehydroabietic acid (11.4 wt.%), palustric acid (8.6 wt.%), neoabietic acid (7.9 wt.%) and sandaracopimaric acid (6.5 wt.%). Furthermore, small amounts of monoterpenes (2.6 wt.%) and benzenes (<0.5 wt.%) are detected in the tall oil. The GC-MS/FID detectable composition of the substances in the crude tall oil is compared with the composition of the residue after extraction in Table 3.

Nogueira and Pereira (1995) were able to detect 37.5 wt.% fatty acids and 49.4 wt.% resin acids, which mainly consist of diterpenoids, in tall oil. In comparison, in the tall oil used herein, the content of resin acids is lower and the content of fatty acids is higher. Since tall oil is based on a natural product, changes in the composition are normal. The types of trees used, origin, storage and the pulping process all have an influence on the composition (Panda 2013; Sandermann 1960).

3.2 Influence of the temperature and extraction vessel on the yield

In contrast to the crude tall oil, the extract has a pleasant odour of pine and is cloudy yellow to transparent yellow and has a much lower viscosity, similar to that of rapeseed oil. The cloudy colouration of the extract did not occur in all samples and disappeared towards the end of the extraction time. Figure 4 shows two tall oil scCO₂ extracts.

Figure 4:

Tall oil scCO₂ extracts obtained with vessel G1 at 333.15 K (ρ_scCO2 = 724 kg/m³; q_mscCO2 = 5 g/min; 50 g crude tall oil; 4 h extraction time; A = cloudy yellow extract after 3 h, B = transparent yellow extract after 4 h).

The extraction curves of the extraction of 50 g crude tall oil with vessel G1 and G2 at 313.15 K and 333.15 K are shown in Figure 5.

Figure 5:

Extraction curves of the extraction of 50 g crude tall oil with vessel G1 and G2 at 313.15 K and 333.15 K over 4 h at a scCO₂ flow of 5 g/min.

Temperature is an important factor in increasing the yield at the same scCO₂ density. The highest yield (12.4 wt.%) is achieved at 333.15 K with vessel G1. At 313.15 K, the yield with vessel G1 is 8.2 wt.% and therefore is 4.2 % points lower than at 333.15 K. In addition to the temperature, the extraction vessel used and thus the contact time between scCO₂ and crude tall oil also have an influence. This can be seen particularly at the lower temperature of 313.15 K. The yield for vessel G2 (contact time = 19.5 min) at 313.15 K is only 3.8 wt.% and so 4.4 % points lower than for vessel G1 (contact time = 61.3 min). The VPLs of the collected extract are between 5.2 and 1.6 g extract/kg scCO₂, with the highest VPL achieved with vessel G1 at 333.15 K and the lowest with vessel G2 at 313.15 K. The VPLs are as follows.

1. G1, 333.15 K, 20 MPa, 724 kg/m³: 5.2 g extract/kg scCO₂

2. G2, 333.15 K, 20 MPa, 724 kg/m³: 4.1 g extract/kg scCO₂

3. G1, 313.15 K, 12.5 MPa, 731 kg/m³: 3.4 g extract/kg scCO₂

4. G2, 313.15 K, 12.5 MPa, 731 kg/m³: 1.6 g extract/kg scCO₂

Harvala et al. (1987) determined the solubility of crude tall oil in scCO₂ at 313.15 K and 323.15 K at different densities. The solubility at 313.15 K and 10 MPa (ρ_scCO2 = 629 kg/m³) was about 0.6 g extract/kg scCO₂ and increased to about 3.4 g extract/kg scCO₂ at 150 MPa and 313.15 K (ρ_scCO2 = 780 kg/m³). At 323.15 K the solubility was about 4.8 g extract/kg scCO₂ at a pressure of 20 MPa (ρ_scCO2 = 784 kg/m³). The VPLs determined in this study with vessel G1 are in a similar range to the given values.

Extraction with scCO₂ is a combination of dissolving and flushing out the extract (Verres 1994) and if the flow rates are too high, the contact time of the scCO₂ is too short to create an equilibrium (Gupta and Shim 2007). Consequently, it is assumed that the contact time between scCO₂ and crude tall oil in vessel G2 is not sufficient to create an equilibrium and therefore the VPL/yield is lower than with vessel G1.

The cumulative extraction curves obtained with vessel G1 are almost linear after a scCO₂ consumption of 300 g, respectively an extraction time of 60 min, is reached. The extraction curves of vessel G2, on the other hand, are almost linear from the beginning. After the consumption of 300 g scCO₂, the yield obtained with vessel G2 at 333.15 K (2.3 wt.%) is even higher than that of vessel G1 at 333.15 K (1.8 wt.%). It is assumed that the lower headspace volume of vessel G2 (135 cm³) leads to faster flushing out of the dissolved extract at the beginning compared to vessel G1. Furthermore, it is assumed that as soon as the scCO₂ and crude tall oil are mixed in the higher headspace volume of vessel G1 (423 cm³), the longer contact time provides the higher yields.

During scCO₂ extraction of natural substances, there is generally a decrease in yield after an almost linear extraction curve and from that the less accessible or soluble substances are extracted (Sovová 1994). From an economic point of view, it is advisable to continue extraction as long as the yield increases linearly with extraction time (del Valle 2015). Since the extraction curves obtained herein are almost linear, it can be assumed that even higher yields can be achieved if the extraction is continued.

After depressurisation, it was noticed that in the residue a high amount of CO₂ is dissolved. This resulted in a negative loss and the formation of bubbles in the residue. A decrease in the weight of the residue over several days was observed.

3.3 GC-MS/FID characterization of the extracts

The produced extracts are analysed by GC/MS-FID after an online methylation to evaluate the distribution of different substance classes due scCO₂-extraction. Five different substance groups are found in the crude tall oil and the extracts as well: free fatty acids (linoleic acid, oleic acid), high volatile monoterpenes (α-pinen, δ-3-caren), diterpenoids (abietic acid, palustric acid), benzenes and unknown fatty acids (see Figure 6). As shown before, the total yield of different substance groups varies with the influence of temperature and the volume of extraction vessel. In general, the total yield of extractives increases with extraction temperature. The influence of time during extraction is not as pronounced as the influence of extraction temperature. At higher extraction temperature (333.15 K), independent on vessel volume, higher amounts of GC-detectable extractives, especially free fatty acids and diterpenoids are found. At the first sampling after 1 h (333.15 K, vessel G1) an extractive composition showing mainly free fatty acids (28.1 wt.%), diterpenoids (34.7 wt.%) and monoterpenes (17.5 wt.%) is found. Compared to crude tall oil, the proportion of monoterpenes and free fatty acids increases, while the proportion of diterpenoids decrease. These amount decreases from 70.5 wt.% to 34.7 wt.% (333.15 K, vessel G1, 1 h). The composition shifts with increasing extraction time towards a relative increase in the proportion of free fatty acids and diterpenoids. The amount of free fatty acids increases from 28.1 wt.% (1 h) to 34.4 wt.% (4 h) and the amount of diterpenoids from 34.7 wt.% (1 h) to 43.3 wt.% (4 h), respectively (333.15 K, vessel G1). In contrast, the proportions of volatile monoterpenes decrease successively as the extraction time increases. The amounts decrease from 17.5 wt.% after 1 h to 5.3 wt.% after 4 h extraction time (333.15 K, vessel G1). The extracts after the pressure release differ only slightly in composition from the extracts obtained during extraction. The residue found after extraction is again very similar to the crude tall oil with a high proportion of diterpenoids (85.3 wt.%) and lower proportions of free fatty acids (20.4 wt.%) and monoterpenes (0.4 wt.%) (333.15 K, vessel G1). The extraction temperature has only a minor influence on this distribution of the named fractions. Although a reduction in temperature causes a slight decrease in the overall yield of the extracts, the composition does not change essentially.

Figure 6:

Depiction of the GC-detectable compounds of crude tall oil fractions after scCO₂ extraction at certain sampling times, pressure release and residue in dependence of extraction temperature and extraction vessel: (a) extraction with vessel G1 (V = 473 cm³, L = 710 mm), (b) extraction with vessel G2 (V = 185 cm³, L = 252 mm).

This behaviour can also be observed when using the extraction vessel G2 (see Figure 6b). The composition of the extracts is shifted in favour of free fatty acids and monoterpenes. The proportion of diterpenoids decreases. Furthermore, the behaviour during the extraction time is approximately the same, whereby the proportion of free fatty acids and high volatile monoterpenes also increases in vessel G2 compared to the crude tall oil. The amount of diterpenoids decreases in the same manner. With increasing extraction time, the amount of diterpenoids decreases as well. In contrast to vessel G1, the proportion of diterpenoids does not increase during the extraction time at 333.15 K but remains the same.

In principle, the distribution of extractable substances over the extraction time shows a similar pattern to the extraction of comparable intermediate products from technical wood processing. In comparison, the scCO₂ extraction of beech wood slow pyrolysis liquids also produces a characteristic composition of the extracts obtained at different densities or temperatures and extraction times. This also revealed the substance-specific dependence of solubility on scCO₂ density. When extracting crude tall oil and pyrolysis oil, it can be seen that the low molecular weight components are extracted first and the proportion of higher molecular weight substances increases as the extraction time increases. In addition, high molecular weight substances or components with a low solubility in scCO₂ remain in the residue (Möck et al. 2023). In the case of crude tall oil, the high molecular weight compounds have already been separated off due to further processing during the kraft process and are therefore no longer found during extraction. The occasional overestimation of the substance amounts, which is particularly noticeable in the quantification of residues, may be caused by assumed response factors in the quantification of the individual substances. Small individual deviations can quickly add up to larger deviations and ultimately lead to a content of over 100 wt.%. As no standards are available for many tall oil-specific substances, it is still only possible to work with similar substances with similar retention behaviour.

Compared to vacuum distillation, the scCO₂ extraction used in this study does not achieve as distinct a separation into different substance fractions. According to Nogueira (1996), vacuum distillation results in a tall oil fatty acids fraction containing at least 90 wt.% fatty acids. In contrast, the maximum fatty acid content in the scCO₂ extracts reaches only 45.0 wt.% (333.15 K, vessel G1, pressure release). In vacuum distillation, different temperatures are used for separation. In the experiments conducted in this study, extraction was performed in a very low scCO₂ density range (724–731 kg/m³). To achieve better fractionation, future work will involve the use of separators with varying scCO₂ densities, as proposed by Harvala et al. (1987) and applied by Möck et al. (2024) for the fractionation of pyrolysis liquids. Another option for improved fractionation could be multistage extraction using different densities and cosolvents, as demonstrated by Barbini et al. (2021) for the extraction of pine bark.

3.4 Maximising the yield by increasing the solvent to feed ratio

In order to find out the maximum amount that can be extracted from the tall oil and at what point the amount of dissolved extract decreases, an extraction with a significantly higher solvent to feed ratio of 269 (g/g) was carried out with vessel G2 at 333.15 K. For this purpose, the amount of tall oil used (13.4 g) was reduced compared to the previous experiments and the CO₂ flow was increased to 15 g/min. Figure 7 shows the yields and VPLs of the extraction.

Figure 7:

Yield and VPL of an extraction with a solvent to feed ratio of 269 g/g crude tall oil (333.15 K; vessel G2 (V = 185 cm³, L = 252 mm); q_mscCO2 = 15 g/min; 13.4 g crude tall oil).

By increasing the solvent to feed ratio, the yield increases up to 57 wt.%. However, the VPLs are between 1.7 g extract/kg scCO₂ and 2.2 g extract/kg scCO₂ and are thus only about half as high as in the previous experiments using a lower solvent to feed ratio. The VPL increases to 2.2 g extract/kg scCO₂ up to a scCO₂ consumption of 1.2 kg and then decreases to 2.0 g extract/kg scCO₂. The slight decrease in the VPLs suggests that far more extract can be extracted from the crude tall oil before a significant decrease in the extraction curve occurs. By increasing the scCO₂ flow to 15 g/min, the contact time is only 8.3 min. In the previous experiment with the same vessel, the contact time was 19.5 min and was thus more than twice as high. Based on the results, it is assumed that the contact time between crude tall oil and scCO₂ is crucial for an effective extraction. The residue was analysed by GC-MS/FID to find out which substances are preferentially extracted and which accumulate in the residue. Table 3 compares the substances contained in the residue with those in the crude tall oil.

In comparison, the content of fatty acids (including unknown fatty acids) decreases from 25.0 wt.% in crude tall oil to 17.2 wt.% in the residue. Monoterpenes and benzenes are no longer detectable in the residue. The proportion of diterpenoids in the residue is 82.6 wt.% compared to 70.5 wt.% in crude tall oil. The substance with the highest proportion in the residue is abietic acid, at 36.4 wt.%. In crude tall oil, abietic acid accounts for 30.6 wt.%. Consequently, there is a slight enrichment of diterpenoids in the residue. In the residue, no monoterpenes are present, as these have significantly higher solubility compared to fatty acids. For instance, α-pinene already has a solubility of 53.5 g/kg scCO₂ at a scCO₂ density of 358 kg/m³ at 333.15 K (Maheshwari et al. 1992). In contrast, the solubility of oleic acid at 333.15 K and a density of 730 kg/m³ is only 7.6 g/kg scCO₂ (Akgün et al. 1999). The higher solubility of monoterpenes thus results in their preferential extraction. For diterpenoids no solubility data could be found. However, as an enrichment occurs in the residue, it can be assumed that their solubility is lower than that of the fatty acids and monoterpenes.