11/07/2024

Spectroscopic Insights into Biorefineries - A demand for process analytics in bioeconomy

Dr. Alexander Echtermeyer , S-PACT

The transition from a fossil-based chemical industry to a circular, sustainable bioeconomy is one of the crucial steps towards a sustainable future economy. Closed carbon cycles must be established, i.e., by chemical utilization of biogenic raw and residual materials.

This can be achieved with biorefineries that are among the key technologies to provide versatile molecular building blocks to the chemical industry. Especially lignocellulosic biomass serves as a great renewable resource as it is available in large quantities and naturally rich in complex molecular structures such as sugars and aromatic compounds.

To harvest these molecules from the strongly heterogeneous raw material of seasonally fluctuating quality while preserving their complex molecular structure is the aim of a lignocellulosic biorefinery.

In contrast to established petrochemical processes, lignocellulosic biorefineries are built on innovative processing concepts at rather mild reaction conditions, with the requirement of flexible operation to quickly adjust to different types and seasonal qualities of lignocellulosic raw materials.

To provide the necessary process data for a flexible and efficient biorefinery plant operation, inline process spectroscopy techniques, such as Raman, mid-infrared (MIR), or nuclear magnetic resonance (NMR) spectroscopy, are well suited because they yield time-resolved multi-variate data in fast, robust, reliable, and automated manner. Coupled with chemometric methods, these monitoring approaches acquire quantitative multi-component concentration profiles to inform process control units and operators.

Figure 1 illustrates the typical process steps in a lignocellulosic biorefinery on a simplified level to convert lignocellulosic raw and waste biomass into versatile building blocks for the chemical industry. Examples are highlighted for successful applications of process monitoring solutions which employ spectroscopic techniques combined with comprehensive chemometric methods to gain insight into the different process steps [1,2,6-10]. In the following sections, the respective examples for spectroscopy-based process monitoring are discussed in more detail.

Fig.1: Some application examples for process spectroscopy at different steps of the conversion of lignocellulosic raw and waste biomass to biorefinery products. Adapted from [1].

Pretreatment and hydrolysis monitored by MIR spectroscopy

In a first step, the shredded lignocellulosic biomass is pretreated and hydrolyzed to break the complex and resistant bio-matrix and to depolymerize the sugar polymers. This step often involves organic or mineral acids for catalysis, aqueous-organic solvent systems, and is usually performed at a temperature between 160 - 220 °C with respective pressures.

Figure 2 illustrates the application of fiber-optic MIR spectroscopy (Matrix MF, Bruker Optics, Germany) to monitor the sulfuric acid-catalyzed hydrolysis of residual wood with in-situ conversion of the biomass-derived C₆-sugars to levulinic acid (LA) in a stirred 50 L Hastelloy reactor for one to three hours at 180 °C [3]. The air-cooled MIR attenuated total reflection (ATR) fiber-optic probe (IFS Aachen, Germany) enables direct and time-resolved insight into the process despite a corrosive, dark, and turbid reaction suspension.

The obtained inline MIR spectra resolve the reaction progress and can be quantitatively evaluated using spectral Hard Modeling to track the concentration profiles of several main components such as glucose, xylose, 5-hydroxymethylfurfural (5-HMF), furfural, acetic acid, formic acid (FA), LA, and sulfuric acid. This allows to derive more detailed kinetic models and to determine the end of the reaction which limits product degradation and energy consumption.

Fig.2: MIR spectroscopy applied for inline monitoring of chemo-catalytic conversion of lignocellulosic biomass to levulinic acid. Adapted from [3].

(Bio-)chemical conversion

Fig.3: (Di-)Carboxylic acids as
renewable building blocks.

In a second step, the obtained C₅- and C₆-sugars are chemically or biologically converted to versatile building block components such as furfural, 5-HMF, or different types of carboxylic acids, cf. Figure 3. Several types of process spectroscopy techniques are suitable for monitoring these reactions.

¹H NMR for homogeneous chemo-catalytic processes

In case of the acid-catalyzed conversion of fructose (FRC) to LA, ¹H NMR spectroscopy is advantageous for process monitoring as, in contrast to Raman or MIR spectroscopy, the chemical components of interest show almost no peak overlap which renders chemometrics based on simple peak integration feasible. Using ¹H NMR spectroscopy (Spinsolve 43, Magritek, Germany), the conversion of FRC to the intermediate 5-HMF with consecutive conversion to FA and LA is tracked (Figure 4).

Fig.4: Sulfuric acid-catalyzed conversion of FRC to LA analyzed by ¹H NMR spectroscopy for the species concentration profiles. Adapted from [1].

Chemometric methods such as peak integration or, in case of overlapping and complex signals, spectral Hard Modeling provide the concentration profiles of those components that allow for determination of humin (HUM) side product formation based on mass balance calculation [1,3,8].

Raman spectroscopy in heterogeneous bio-catalytic processes

As an alternative to chemo-catalyzed conversion, bio-catalyzed processes like fermentations can be applied to convert the sugars, obtained after biomass pretreatment and hydrolysis, to versatile building block chemicals. Here, carboxylic acids such as lactic, malic, succinic, or itaconic acid are interesting precursors to a large variety of chemical products, cf. Figure 3.

As those dicarboxylic acids form a pH-dependent equilibrium in aqueous solution, process analytics targeting on monitoring those processes need to distinguish between the differrent dissociation states, meaning the dissociated and associated carboxylic acid species. A monitoring approach based on inline Raman spectroscopy (Rxn2, Endress+Hauser, Germany) coupled with spectral Hard Modeling has been developed to address this challenge as Raman spectroscopy is very suitable for analysis of organics in aqueous solution, while spectral Hard Modeling can resolve the strongly superimposed Raman signals of the structurally similar dissociation states [1,2]. Using a titration-based method, a spectral Hard Model of the different dissociation states in water is constructed and calibrated with minimal effort due to precise knowledge of the dissociation equilibrium pK_a values (Figure 5).

Fig.5: Spectral pure component (PCM) and mixture Hard Models for dissociating itaconic acid in aqueous solution monitored by inline Raman spectroscopy. Adapted from [1,2].

In case the pK_a values of the dissociating chemical system are not available, an alternative method is proposed that combines spectral analysis by Multivariate Curve Resolution and model-based estimation of the pK_a values to provide pure component spectra for spectral Hard Model construction and the required dissociation equilibrium model (pK_a values) for calibration. This approach has already been successfully transferred to other carboxylic and mineral acids, allowing precise acid species monitoring at different pH values [1,2,9].

The above outlined strategy for carboxylic acid monitoring with spectral Hard Modeling is employed as the basis for Raman-informed inline analytics in fermentation experiments, displayed in Figure 6, ranging from 2 L laboratory-scale to 100 L pilot-scale [4,5]. Thereby, the spectral Hard Model comprising water and itaconic acid species is extended by further component models for sucrose, glucose, and fructose to account for the substrates used during fermentation, as well as by erythritol as side product and by ethanol that is used during downstream processing but is introduced to the fermenter in small amounts by recycle streams.

Fig.6: Inline Raman spectroscopy used for monitoring of a fermentation process employing the fungus Ustilago cynodontis in 100 L pilot-scale to convert thick juice from sugar beet processing into itaconic acid. The zoom shows the Raman immersion probe (Rxn-10 + bIO, Endress+Hauser) mounted with a custom spacing adapter.

Product separation

In a third step, the platform chemicals need to be separated from the product mixture and further purified. To harvest carboxylic acids from the fermentation medium, membrane filtration for cell retention followed by chromatographic separation of the itaconic acid species from the fermentation medium with hydrophobic adsorbents highlights an efficient way of downstream processing as it is a very selective method with low-energy requirements and limited associated waste formation. In Figure 7, Inline Raman spectroscopy combined with chemometrics based on spectral Hard Modeling is successfully used to resolve the concentration profiles of the outlet streams from the chromatographic separation unit over time, thereby providing insight into the elution times of the different dissociation states which supports the development of novel adsorbent materials with improved performance [1,10].

Fig.7: Inline Raman spectra and concentration profiles of dissociating itaconic acid species at the outlet of a chromatographic column monitored in real-time. Adapted from [1].

An alternative option to chromatographic separation of the target carboxylic acids from the product mixture is electrochemically driven pH-shift reactive extraction. This technique is a novel and promising approach as it circumvents salt waste formation - a large issue of chemically induced pH-shift in conventional process concepts - and is powered by renewably sourced electricity, improving the ecological footprint of the purification method. Inline monitoring of both the aqueous and organic phases is possible, e.g., by Raman spectroscopy combined with spectral Hard Modeling, for quantifying the concentration profiles of the dissociating acid species [1].

The strength of this monitoring approach is demonstrated for the electrochemically driven pH-shift reactive extraction of itaconic acid from an aqueous solution of carboxylic acids and sodium sulfate as electrolyte. Employing the combination of inline Raman spectroscopy (Rxn2, Endress+Hauser, Germany) and quantitative spectral Hard Modeling, it is possible to track concentration profiles of the different itaconic acid species in the aqueous phase along with the extraction of fully associated itaconic acid in the organic phase, both in real time (cf. Figure 8).

Fig.8: Raman spectra and concentration profiles of dissociating itaconic acid and electrolyte species during electrochemically induced pH-shift reactive extraction, inline monitored in real time for both phases by Raman spectroscopy and spectral Hard Modeling. Adapted from [1].

The results match well with the HPLC reference analytics. Moreover, it is possible to track the sulfate electrolyte species to avoid sulfate protonation and extraction upon depletion of carboxylic acid buffer capacity, which otherwise causes impurification and capacity reduction of the organic phase, leading to tremendously diminish the process coulomb efficiency, and strongly impeded process economics. In subsequent process steps, the carboxylic acid in the organic phase is separated from the organic solvents and purified, e.g., by crystallization methods, while the fermentation medium in the aqueous phase is recycled in the overall production process.

Conclusion

Tab.1: Characteristics of spectroscopic
PAT for inline monitoring.

Biorefinery plants need to deal with strongly heterogeneous biogenic raw and waste materials that are subject to seasonal fluctuation in amount, composition, and quality.

Therefore, those plants require analytical methods that are fast, robust, and reliable, that can be automated, and that comprehensively capture a large number of analytes - a great match with process spectroscopy solutions, such as MIR, Raman, NMR, or others, which serve these requirements if coupled with powerful chemometrics methods. Exploiting the respective characteristics of the available techniques allows to address the challenges in each of the process steps adequately.

All techniques benefit from the application of spectral Hard Modeling, paving a cost-efficient way to quantitatively analyze the spectra of the multi-component mixtures based on available molecular knowledge of the processes. In contrast to data-driven approaches like Partial Least Squares (PLS) Regression, these models can easily be adapted or extended to new process conditions or additionally occurring components. Therefore, spectroscopy-based fast and direct process insight is possible that extends process knowledge and improves process efficiency, e.g., by avoiding or reducing side product formation, thus strengthening the economic competitiveness of biorefineries to support the transition toward a green future of the chemical industry.

Acknowledgement

The presented applications are developed in the Center for Next Generation Processes and Products (NGP²) of Aachener Verfahrenstechnik (AVT) at RWTH Aachen University, Aachen, Germany. Laboratory scale experiments are done at Fluid Process Engineering (AVT.FVT), Biochemical Engineering (AVT.BioVT), and Process Systems Engineering (AVT.SVT), while pilot-scale research is conducted at NGP² Biorefinery.

References

1. Echtermeyer. Doctoral Dissertation, RWTH Aachen University, 2022
   
2. Echtermeyer et int. Viell. Appl. Spectrosc., 2021, 75(5):506-519.
   
3. Echtermeyer & Viell. 6th European Conference on Process Analytics and Control Technology, Copenhagen, 2023
   
4. Saur et int. Jupke. Bioengineering, 2023, 10(6), 723 - 767
   
5. Pastoors et int. Büchs. Biotechnol. Biofuels Bioprod., 2023, 16(181)
   
6. Aigner et int. Jupke. J. Chem. Eng. Data, 2020, 65(3):993-1004
   
7. Karacasulu et int. Mitsos. Sustain. Energy Fuels, 2022, 6(11):2734-2744
   
8. Echtermeyer & Viell. ACS Omega, 2024, 9(6):6432-6441
   
9. Roth et int. Jupke. J. Chem. Eng. Data, 2023, 68(6):1397-1410
   
10. Biselli et int. Jupke. J. Chromatogr. A, 2022, 1675, 463140

» Download article (815 KByte)