(309d) Recovery of Organic Acids Produced by Acidogenic Fermentation: Modeling of the Separation Process

Acetic, propionic, butyric and lactic acids are largely used by industry, as solvents, preservatives or biofuels. These molecules are mainly derived from oil. However, as oil resources are decreasing and the energy issues are motivating the use of green chemicals, it is essential to find an alternative way to produce these molecules from renewable raw materials.

Using all fermentable, very diversified and cheap biomass with few pretreatments, acidogenic fermentation leads to the production of these acids in the fermentation broth, but only in low concentration (+/- 1 %), which constitutes the main issue.

Many separation techniques can be proposed but no efficient process has been identified yet. The global aim of this research is to separate acids from the fermentation broth by designing a process which is technically, environmentally and energy efficient.

Process

The most promising separation technique seems to be the liquid-liquid extraction, considering that operating parameters are optimized, and especially the solvent composition. However, few data are available for similar cases.

The separation process that we investigate is schematically presented in Fig. 1. The fermentation step produces a broth made of water and volatile fatty acids. This broth gets in contact with a solvent which extracts the organic acids. The organic acids present in the solvent phase are then back extracted with glycerol. Glycerol is available as a by-product of biodiesel production and would benefit from new uses [1]. This separation process is thus composed of two steps: a first extraction with an intermediate solvent and a re-extraction with glycerol. The most appropriate intermediate solvent has to be identified.

Aim

The aim of the present study was to develop the mathematical model of this complete separation process and to use this model with experimental distribution coefficients and the solubilities in order to identify the influence of various parametres on the performance of the whole process. These are preliminary steps towards the optimisation of the process.

Materials and methods

Experimental distribution coefficients

1 %_v/v butyric acid model solutions were prepared by dissolving 50 µl butyric acid (Fluka Chemica, >99.5%) in 5 ml distilled water. The model solution were equilibrated at 25°C with 5 ml of each tested solvent. The phases were separated by centrifugation and 4 ml of the solvent phase was equilibrated with 4 ml glycerol (Janssen Chemical, 99%) before a new separation by centrifugation. The butyric acid concentrations in the aqueous and glycerol phases were determined by gas chromatography and the concentration in the organic phase was calculated by mass balance. These data were used to calculate the distribution coefficients defined as:

K_D= x_i^ext/ x_i^raf

where x_i^ext et x_i^raf represent the mass fraction of the compound i respectively in the extract and in the raffinate.

Five solvents, based on tri-n-octylamine, were tested:

1. Tri-n-octylamine (Acros, 98%)
2. 30 %_v/v Tri-n-octylamine + 70 %_v/v Dodecanol (Janssen Chemical, 98%)
3. 30 %_v/v Tri-n-octylamine + 40 %_v/v Dodecanol + 30 %_v/v Hexane (Lab-scan, 99%)
4. 50 %_v/v Tri-n-octylamine + 50 %_v/v Tributylphosphate (Vel s.a., 97%)
5. 50 %_v/v Tri-n-octylamine + 50 %_v/v Tripentylamine (Aldrich, 98%)

Figure 1: TOA-based extraction - glycerol-based re-extraction process for the separation of volatile fatty acids produced by acidogenic fermentation of biomass.

Process modeling

In order to simulate the global process, a program was developed. The algorithm calculates the equilibrium composition of each phase on a given stage, based on mass balance, distribution coefficients and solubility limits.

As limit condition, our model imposes that all the glycerol present in the recycled solvent phase is transferred into the aqueous phase in the first extraction column. Likewise, all the water present in the solvent phase is transferred into the glycerol phase in the re-extraction column.

Results and discussion

Experimental section

The distribution coefficients determined experimentally are presented in Fig. 2.

Figure 2: Distribution coefficients determined for the extraction of the butyric acid at 25°C and a pH 2.7 for the initial aqueous solution

(a) for the first column (K_{D solvent/water}) (b) for the second column (K_{D glycerol/solvent})

Large differences between the values of the distribution coefficients are clearly observed for the different solvents. An opposite trend is observed for the distribution coefficient of the two extraction steps. Figure 3 presents the correlation between the two distribution coefficients or each solvent tested, together with the best fit of the equation: KD_{solvent/water} *KD_{glycerol/solvent}= C.

The best fit value of C is 0.675 +/- 0.005 standard error.

Figure 3: Distribution coefficients for the re-extraction (KD_{glycerol/solvent}) as a function of the distribution coefficients obtained for the first extraction (KD_{solvent/water}), and best fit of the equation: KD_{solvent/water} *KD_{glycerol/solvent}=C

Modeling section

The performance of the process was first assessed with the extraction degree represents the ratio between the quantity of acid finally transferred to the glycerol phase and the quantity of acid initially in the aqueous phase. Fig. 4 presents the influence of the number of plateau (same number for both extraction steps), of solvent recycling and of the distribution coefficient on the extraction degree. In all cases, KD_{solvent/water} *KD_{glycerol/solvent} = 0.675. The recycling has no effect on the extraction degree for KD_{solvent/water} = 0.5. A small difference of the values of the extraction degree between the simulation of the process with or without recycling is observed for KD_{solvent/water} = 1. Recycling is seen to have a stronger effect for KD_{solvent/water} = 1.5. When the distribution coefficient is less favourable to the transfer to glycerol in the reextraction step (case KD_{solvent/water}= 1.5), recycling of the solvent has a higher influence to improve the overall efficiency of the process. All further simulations were performed with recycling.

Figure 4: Extraction degree as a function of the number of stages for KD_{solvent/water} with KD_{solvent/water} *KD_{glycerol/solvent} = 0.675. Simulation of the complete process with and without solvent recycling.

Figure 5: Extraction degree as a function of KD_{solvent/water} with KD_{solvent/water} *KD_{glycerol/solvent} = 0.675 Simulation of the complete process with solvent recycling and columns of 30 stages. Flow rates: Glycerol: 100 kg/h. Feed: 100 kg/h and 1% wt butyric acid. Solubilities: 1g/kg

Fig. 5 compares the extraction degree as a function of KD_{solvent/water} for different solvent mass flow rates (keeping constant the feed and glycerol flow rates). All conditions tested reach the same maximum extraction degree. Then, a compromise has to be found between all parameters on basis of all performances criteria.

Conclusion

An extraction - re-extraction process for the recovery of acetic, propionic, butyric and lactic acids from the fermentation broth was investigated. The distribution coefficients of the different solvents tested experimentally appear to fit a common relationship with butyric acid: KD_{solvent/water} *KD_{glycerol/solvent} = 0.675 With the program we developed, it is now possible to investigate the influence of various parameters on the performances of the process and to identify optimised conditions that are worth to be tested experimentally.

The full communication will address the influence of parameters like the number of stages of each extraction column, the distribution coefficients, the solubilities and the feed, solvent and glycerol mass flow rates on the performances of the process. The latter will be assessed with extraction degrees, the concentration of acids in the glycerol phase, the solvent consumption and the selectivity of the extraction against the transfer of water

Acknowledgements

The authors would like to acknowledge the Belgian Fonds pour la formation à la Recherche dans líIndustrie et dans líAgriculture (F.R.I.A) for the financial support.

References

[1]C. O'Driscoll (2007) Biofuels, Bioprod., Bioref., 1, 6-7