CO₂ gas–liquid mass transfer and k_La estimation: Numerical investigation in the context of airlift photobioreactor scale-up

Highlights

• Airlift CFD simulations are performed including gas–liquid mass transfer and CO₂ dissociation.

• Dynamic gassing-in experiments with O₂ and CO₂ are reproduced numerically and compared.

• Dynamic method with CO₂ underestimates k_La.

• Gas phase CO₂ depletion limits mass transfer in the upper part of large scale airlifts.

Abstract

This paper deals with gas–liquid mass transfer in an airlift via CFD simulations in the context of photobioreactor (PBR) scale-up. Two aspects are emphasized. Firstly, since carbon uptake by microalgae is of crucial importance as part of PBRs, CO₂ transfer is in focus, and numerical simulations are developed to take into account CO₂ gas–liquid transfer and dissociation in the aqueous phase. Secondly, since estimating k_La is of crucial importance when scaling-up PBRs, different ways to evaluate k_La are discussed using numerical experiments. Firstly, k_La may be estimated as the volume average value of the local mass transfer coefficients calculated from steady-state hydrodynamics and Higbie penetration model. Secondly, k_La can be deduced from classical dynamic gassing-out/gassing-in experiments. This second method is simulated for O₂, as commonly performed experimentally, and also with CO₂ since it is the transferred species in PBRs. Results show that k_La field is strongly heterogeneous, as expected in airlifts where gas is mainly present in the riser. Performed with O₂, the gassing-in method leads to quite accurate estimation of the spatial average value of local k_La. But, gassing-in methods performed with O₂ and CO₂ lead to discordant results. In fact, CFD shows that the CO₂ depletion in the gas phase has to be accounted for to predict k_La from CO₂ gassing-in method, especially at large scale. This study also puts into evidence the potentialities of CFD which allows to get detailed image of local gas–liquid mass transfer, depending on two-phase hydrodynamics, gas phase distribution and transferred species solubility.

Introduction

Microalgae or cyanobacteria cultivation has been the subject of growing interest in recent years since the potential applications are numerous and promising: production of high added value products for the chemical, cosmetic, food or pharmaceutical industries (pigments, vitamins, antioxidant molecules, proteins, polyunsaturated fatty acids, molecules with antibacterial or antiviral properties, etc.), or of bioenergy (biodiesel or biohydrogen). Microalgae are also envisaged to be used for carbon capture from flue gas.

Currently, large scale cultivation of microalgae is mainly limited to aquaculture or food supplement production. To develop microalgae production processes, technological advances are still required to increase energy efficiency and to reduce environmental impact and production costs especially at large scale. These advances have to be realized at the cultivation, harvesting and biorefinery steps. For a better control of cultivation conditions, microalgae culture operated in closed systems, named photobioreactors (PBR), is the most attractive.

In these PBRs, light but also carbon limitations have to be avoided (Ugwu et al., 2008). Contreras et al. (1998) noted that during the exponential growth phase CO₂ gradients could lead to limitations, and that during the linear growth phase CO₂ and light could be limiting. In a bubble column, Nauha and Alopaeus (2013) noticed algal growth inhibition due to low carbon dioxide levels in poorly mixed part of the reactor. Inorganic carbon is commonly provided by injection of CO₂ enriched air into the culture medium and, since microalgae cultivation is not yet developed in large scale PBRs, CO₂ is often used in excess at lab scale to avoid transfer limitation. In fact, for industrial applications such as pharmaceutical products or human food supplements where CO₂ from flue gas or industrial gaseous effluents cannot be employed, CO₂ supply could represent a significant production cost, especially since undissolved CO₂ is lost by degassing. The challenge is then to control the CO₂ amount to be injected: CO₂ must be injected in sufficient quantity for growth conditions while minimizing the global gas intake. In order to operate efficiently microalgae production systems, a deep understanding of CO₂ mass transfer from the gas phase to the microorganisms present in the liquid phase is needed (Langley et al., 2012).

This study proposes an investigation of CO₂ mass transfer capability and of overall volumetric mass transfer coefficient k_La estimation for CO₂ in the context of an airlift PBR scale-up. Airlifts are commonly considered as photobioreactors (Contreras et al., 1998, Sánchez Mirón et al., 2000, Merchuk et al., 2000). In fact, they offer good mixing, mass and heat transfer capabilities as well as easy maintenance and low energy cost. Besides they generate low physical stress and are easy to operate axenically. Despite their simple construction, complex interacting phenomena are occurring when sparging air enriched with CO₂ into airlifts, these phenomena having different orders of magnitude both in time and space. In this context, CFD is a powerful tool to simulate in detail the two-phase hydrodynamics, gas–liquid mass transfer, dissociation of CO₂ in the liquid phase, for a better characterization of all the involved phenomena. Understanding the coupling at different scales is one key to achieve successful PBR scale-up (Garcia-Ochoa and Gomez, 2009).

Only few papers deal with hydrodynamics and CO₂ gas–liquid mass transfer aspects in PBRs (Kordac and Linek, 2008, Khoo et al., 2016, Chen et al., 2016, Valdes et al., 2012). The most detailed investigations of CO₂ mass transfer in bioreactors concern mammalian cell cultures mixing tanks (Gray et al., 1996, Puskeiler et al., 2012, Sieblist et al., 2011) since high levels of pCO₂ and related acidification – or increased osmolality in case of pH control – may become a major problem when scale-up is to be achieved. Thus, in that case, efficient CO₂ transfer rate is also looked for to improve stripping (Nienow, 2006).

In gas–liquid mass transfer experimental studies, CO₂ transfer is often looked as analog to O₂ gas–liquid mass transfer. In fact k_La is easier to measure experimentally with O₂, using an oxygen probe, than with CO₂. The overall gas–liquid transfer coefficient of CO₂ is then often deduced from experiments where O₂ transfer is measured with a correction by O₂ and CO₂ diffusion coefficients (Babcock et al., 2002, Langley et al., 2012, Khoo et al., 2016, Fernandes et al., 2014, Garcia-Ochoa and Gomez, 2009). Some other authors (Hill, 2006, Sieblist et al., 2011, Valdes et al., 2012) followed the pH evolution to estimate the inorganic carbon concentration in the liquid phase and to deduce the overall volumetric mass transfer coefficient of CO₂ inside a mixing tank or a bubble column. CO₂ sensors have also been used to estimate directly CO₂ gas–liquid transfer during CO₂ stripping or absorption experiments (Matsunaga et al., 2009, Puskeiler et al., 2012). These experiments allow a direct determination of gas–liquid mass transfer for CO₂ without any assumption about a possible similarity between O₂ and CO₂ transfers.

These different methods to experimentally characterize k_La for CO₂ can lead to some contradictory or unexplained results in the literature. In fact, Kordac and Linek (2008) found results in contradiction with those of Hill (2006) concerning the k_La of CO₂ in a salt solution, which they explain by the fact that Hill's model does not take into account gas depletion: Kordac and Linek (2008) pointed out that considering a constant driving force could entail an underestimation of k_La for CO₂. Matsunaga et al. (2009) determined experimentally k_La for CO₂ in the context of scale-up of cell cultures bioreactors. As Hill (2006), they considered a constant gas phase composition and found that, during scale-up, the evolution of the k_La for O₂ and CO₂ were different. Sieblist et al. (2011) have taken the gas phase residence time into account as well as its varying composition: they established that for large bioreactors, when CO₂ stripping is looked for, bubbles were saturated with CO₂ before leaving the culture. Puskeiler et al. (2012), when measuring the simultaneous absorption of O₂ and CO₂ in a stirred tank, observed that the ratio $k_L{a}_{({\text{CO}}_2)}/k_L{a}_{(O_2)}$ decreased when increasing the impeller speed, which is contradictory to theory.

In this paper, CFD is used to characterize numerically CO₂ mass transfer in two airlifts, one at lab scale, the other at semi-industrial scale. In particular, the commonly used dynamic method for k_La determination will be reproduced numerically with O₂ and CO₂. These classical experiments will be numerically investigated to explain the discrepancies sometimes observed in the literature. It will be pointed out that CFD allows a detailed interpretation of local and dynamic phenomena occurring in such a complex situation where two-phase hydrodynamics is coupled to gas–liquid transfer and chemical reaction.