Beyond biofuels: economic opportunities, recent advances and challenges in property modeling for vegetable oils

4.1 Direct use of vegetable oils

This can be of two types. The first one is to find new applications for known VOs. For example, soybean oil, the second most produced VO, is traditionally used in foods and personal care applications among others and has also a great potential to be used in lubricants [9]. Many VOs have also been applied as corrosion inhibitors and polymeric binders in coatings [10].

The second type consists in finding alternative VOs/fats for a given application. The search for cocoa butter equivalents (CBE) is an example of this approach, in which the desired physical properties of cocoa butter are sought among other fats/oils mixtures [11].

Considering the large variety of feedstocks, it is reasonable to think that tailored blends could well provide alternatives solutions. However, the number of possible blends is extremely large and, as it will be discussed in the next sections, computational tools can be useful to cope with the large combinatorial problems for product design.

4.2 Modification of the triacylglycerols profile

VOs are mainly composed by TAGs. More than the fatty acid composition, it is the TAG composition (how the fatty acids are distributed in the glycerol structure) that controls the properties of VOs [12]. Figure 3 shows the typical molecular structure of fatty acids and TAGs. The demand for more sophisticated products increases and the simple physical mixture of VOs is not sufficient for matching the desired properties [13]. Other physical (fractionation) and chemical (hydrogenation, interesterification) processes can be used to change the TAG composition and then modify the physical properties of the initial raw materials.

Figure 3

Molecular representation of vegetable oils chemical species (triacylglycerols, fatty acids and glycerol).

In the food industry, the hydrogenation process has been associated with negative nutritional effects on consumers’ health due to the formation of trans fatty acids. In this scenario, the chemical interesterification emerged as a useful technique to change physical properties of VOs blends for food development. This reaction uses catalysts (sodium methoxide, sodium metal) and heat to promote a random distribution of fatty acids among the three positions of the glycerol backbone (sn-1, sn-2 and sn-3, Figure 3). On the other hand, enzymatic interesterification uses regioselective lipases under mild reaction conditions to affect only specific positions of the glycerol backbone. Table 1 shows some different scales of study concerning vegetable oils and the corresponding research topics with some references. Not being exhaustive, this short list gives just a few examples for further reading.

Structured lipids are tailored TAGs, restructured by chemical or enzymatic reactions to change the stereochemical positions of fatty acids on the glycerol structure. The aim is to improve the nutritional or therapeutic properties for functional foods, while keeping the physical characteristics within the desired range [33]. Medium-long-medium (MLM) and conjugated linoleic acids (CLA) TAGs are recognized for their improved nutritional characteristics. MLM has medium chain fatty acids (MCFA) (C6–C12) in stereo positions sn-1 and sn-3 of the glycerol backbone and functional long-chain polyunsaturated fatty acids (PUFA) in position sn-2. The CLA is a general term for the positional and geometric isomers of the octadecadienoic acid with conjugated double bonds at carbon atoms 9/11 or 10/12 [36].

4.3 Use of glycerol as a platform for green chemicals

Glycerol (a by-product in the production of some biofuels) can be a platform chemical for a great variety of specialty chemicals. One example is the synthesis of propanediol, epichlorohydrin and acrylic acid. These compounds, classically synthesized by petrochemical based routes (using propene), are used, respectively, for the production of polymers, epoxy resins and various plastics, coatings, adhesives and elastomers. These chemicals can also be synthesized using glycerol as the starting material [26, 27]. Currently, propanediol can be synthesized from glycerol in one reaction step, whereas the petrochemical one uses two reaction steps. The glycerol-based route also avoids the carcinogenic intermediate propylene oxide [26]. Glycerol carbonate has also a large potential as a novel component of gas-separation membranes, solvent and biolubricant [28, 29]

4.4 Use of fatty acids as a platform for green chemicals

The fatty acids present in VOs can be also used as starting materials for the synthesis of value-added chemicals. Double bonds present in some fatty acids represent suitable reactive sites. Fatty acids are a rich source of monomers that can be chemically functionalized to make biosourced polymers, such as diacids, polyacids, diols, polyols, diamines, polyamines, hydroxy acids, amino acids, polyaldehydes, vinyl compounds and polyurethanes [30]. Indeed, polyurethanes are one of the most promising area of interest of functionalized VOs/fatty acids and there are different synthesis routes for polyols at research or commercial stage, such as transesterification, hydroformylation, ozonolysis and thiol-ene coupling [30, 31]. There is a great variety of fatty acids (and consequently, a great variety of possible monomers) and this opens opportunities to polymers having new and enhanced properties. Besides polymers, coating for wood surfaces, PVC plasticizers, lubricants, surfactants and personal care products are also examples of products where fatty acids can be used.

5.1 Synthesis routes and performance of new chemicals

Apart from the eco-friendly sound of bio-based chemicals, important issues related to synthesis and performance must also be considered. Regarding synthesis of VO- based products, toxic and hazardous co-reactants are still needed in some cases (e.g., isocyanate in the production of polyurethanes from VOs [31]). The use of hazardous solvents, the energy consumption, the cost and the by-products generation must be considered as well. In this context, the use of catalysts (enzymes) offers a potentially cleaner and more energy-efficient production than conventional chemical processes in oleochemistry [32]. The enzymatic route can lead to fewer by-products, simpler product recovery, less waste generation and higher yields. However, further developments in suitable enzymes that are not commercially available are necessary. This can involve genetic modifications on the organism that produces the enzyme to improve enzyme activity and stability under the process conditions [32].

The performance of the new chemicals at usage conditions is another important issue. Although VOs have advantageous properties (e.g., biodegradability, low volatility, high flash point), there are also drawbacks that must be overcome. The main issues limiting the use of VOs as industrial fluids are their low thermal and oxidation stability and poor low-temperature properties (as fatty acids and TAGs crystallize easily). Current studies aim to improve oxidation and low temperature stability; however, a generalized approach improving the overall performance of VOs as industrial fluids is still lacking [37]. In this context, the use of additives and the modification of the chemical structure of VO are an alternative. As an example, ethylene-vinyl acetate copolymer and ethyl cellulose have been tested to improve the thermal stability and to increase the viscosity range of different VOs [38].

5.2 Experimental measurements

The availability of experimental data is critical for process and/or product modeling, simulation and design. If the computational tools traditionally used in the chemical/petrochemical industry is to be used in the oil and oleochemical industry, the physico-chemical properties of lipid compounds present in VOs must be known.

Natural VOs can be formed by dozens of different chemical species, such as acylglycerols (mono, di and triacylglycerols), free fatty acids, tocopherols, sterols, carotenes and minor compounds. The accurate separation, purification and identification of all lipid compounds are a difficult task. Bulk physical properties can be measured (e.g., viscosity, density, solid fat content) but the measured data of the pure components are still scarce when compared to other chemicals. If we consider just the TAGs part of the mixtures, uncertainties arise even from HPLC analysis. According to Lida et al. [18], many TAGs cannot be identified due to the lack of standards and the fact that TAGs with equivalent carbon numbers are very similar, thereby making their identification (based on relative retention times), difficult and uncertain. The lack of data is even more pronounced for temperature dependent properties (e.g., viscosity, vapor pressure, density, heat capacity and superficial tension). In such cases, it is difficult to obtain pure component data in a large range of temperatures.

Recent efforts have been made in order to compile and classify the chemical structures, physical and biological properties of different classes of lipids in a computational database. Examples are the LIPID BANK [39], the LIPIDAT [40, 41], the LIPID MAPS Structure Database (LMSD) [42] and the CAPEC Lipids Database [43]. Experiments are costly and time-consuming but experimental data form the basis of prediction models and parameter regressions.

5.3 Property models

Traditional group contribution (GC) methods, connectivity index and UNIFAC methods have been recently adapted to cover lipid systems [44–48]. These improvements rely on the revision of GCs parameters using a lipid database containing different classes of lipids (fatty acids, fatty esters, acylglycerols) and their experimental thermophysical properties. Therefore, they are strongly related to the previous experimental challenge, in particular for TAGs. As commented by Prausnitz et al. [49], we cannot get something from nothing, and predictive methods will continue to be dependent on reliable experimental data.

These new predictive models have the advantage that their parameters have been regressed using only lipids properties and they are expected to perform better than former general GC methods. Figure 4 shows a general methodology for depicting a molecule into functional groups, allowing to develop and use GC models.

Figure 4

Molecular representation and computational representation of group contribution models for property predictions.

This computational representation was implemented by some of the authors [50] to deal with bio-solvent design. In this approach, the SMILES notation of a molecule is automatically converted into a matrix (molecular graph) in which the atoms and bonds are identified.

Though the origins of current GC methods are over 50 years, lipid-based models have only recently been developed, as mentioned above. These improvements on thermophysical properties and mixture models have so far been mainly focused on improving the design of biorefinery processes through computational simulations. However, they are also useful for the design of products using VOs. Flöter [51] highlighted the importance of the melting point, the heat of fusion and of the solubility for food product development using fats. Crystallization (polymorphisms in solid state, kinetics of polymorphic transitions) and the resulting crystal habit and size are also of great importance for product development. In this context, modeling can help to design better experiments during the search for product solutions [51].

For product design, mixture models must include equilibrium models, in particular solid-liquid equilibrium (SLE), since many end-user properties (texture, creaminess, plasticity) depend on the careful control of the solid fat content (SFC).

In VOs, many properties are determined not only by the molecular structure but also by the crystalline arrangement between molecules, since TAGs can solidify in different crystals packing (polymorphism). The problem of SLE modeling has been addressed using different chemical species normally found in VOs (Figure 5). Even non-equilibrium (time-dependent) properties play an important role in products using VOs. This is the case of the kinetic of crystallization which influences the crystals shape and size that ultimately set the sensorial performance of fatty food products.

Figure 5

The solid-liquid equilibrium problem in some lipids and examples of strategies to cope with it.

The polymorphism (crystals morphology) in the solid state (not covered by conventional GC methods) must be considered when designing tailored fat-based foods. The so-called topological indices can offer opportunities to improve description at this level, as they are able to predict properties with higher accuracy than simple GC techniques, taking into account three-dimensional steric effects [52]. Some powerful molecular dynamics models can also be found in literature [53], but these models remain time consuming and not suitable for large multi-component problems.

Thermodynamic models can be found for pure, binary and ternary mixtures [54–56] and for multi-component mixtures [12, 57, 58]. The solid-liquid diagrams of various fatty mixtures are still unknown or were published with limited information, although such phase diagrams are a powerful tool for product development [59, 60]. More empirical or semi-empirical methods, such neural networks [61] and different correlations, have also been used [62–64] aiming to predict the solid-liquid behavior in such mixtures. A broader and deeper review of the SLE problem can be found in Maximo et al. [60].

5.4 Life cycle analysis for new chemicals and raw materials

The life cycle analysis (LCA) has already been applied to biofuels production [65] and to the VO production step [66]. This holistic approach should be also extended for evaluating alternative uses of VOs. The goal is a broader knowledge of their impact on the environment and health. LCA must be used in the VOs crop as well as in the industrial production of the emerging chemicals derived from them. Indeed, when the whole chain of the processes involved is considered, unexpected conclusions can emerge. Törnvall and Hatti-Kaul [32]studied a comparative LCA of various wood coating products; they concluded that a fossil feedstock-based coating was the most favorable with respect to many environmental categories (global warming, acidification, eutrophication and photochemical ozone). The lower environmental performance of the wax esters coatings based on VOs was mostly due to the negative effects of the rapeseed oil cultivation stage. This example demonstrates that a holist approach (like LCA) is necessary to measure the real impacts when comparing renewable and fossil feedstocks. The agricultural phase also imposes environmental constraints that must be considered (use of fertilizers, pesticides, steel for machines, land use, among others). As different VOs have different agricultural characteristics, it is important to consider that various crops can lead to diverse environmental performances. The diversity of raw materials and the multiple ways to combine them is a driver to computer-aided tools, as discussed in the following section.