Aeration in fed batch cultivations of Saccharomyces cerevisiae

Aeration in fed batch cultivations of Saccharomyces cerevisiae

Master of Science thesis within Industrial Biotechnology

Jonas Bjarre

Supervisor, Sven Risén Process Development Engineer

at Jästbolaget AB

Examiner, Professor Gen Larsson

Head of the Industrial Biotechnology Division

Abstract

Jästbolaget AB was founded in 1983 and is today the Nordic’s largest producer of baker’s yeast. The production process is a series of fed batch cultivations with the largest reactors at 215 m³. Aeration is essential in baker’s yeast production in order to get the yield as high as possible. Air is supplied through fixed spargers, and is to volume the biggest raw material.

The compressors working up pressure of air stand for about 20 % of the factory’s energy consumption, and over aeration also have a negative impact on the cultivations. There have been problems with under aeration, especially during summertime, but normally it is believed that excess of oxygen is transferred to the cultivations. If there is such an excess the size is however unknown. Jästbolaget also strives towards producing a variety of special yeasts, which in turn might have different oxygen demands. This general problem can be thought of occurring in many other cell cultivation industries as well, and is of extra importance at production of a low value product when maximum yield from every raw material is essential.

From this background this work’s goal was to optimize aeration profiles in fed batch cultivations of Saccharomyces cerevisiae, with the Jästbolaget’s production process used as a model. In other words to suggest profiles where the oxygen supply matches the oxygen consumption by the cells, giving a steady state in the oxygen concentration in the reactor.

If aeration is insufficient the yeast produces ethanol as a fermentation product, which traditionally are used as verification of that sufficient air is supplied in baker’s yeast factories. But ethanol can also be formed from overfeeding via overflow metabolism. This was via lab fermentations that were ran at different growth rates concluded not to be an issue for the normal yeast strain use. The normal growth rate in the production process was showed to be below the growth rate for overflow metabolism. It is however likely that other strains that are grown might experience this problem.

With this settled the main goal of this work could be continued, with ethanol formation as a response for under aeration. Aeration profiles were created and evaluated using simulations, and verified in the lab scale reactors. A suitable profile was determined to generally follow

𝑄(𝑘_𝐿𝑎(𝑡)) = 𝑄 (𝑐𝑒^{𝑆𝐹𝑅∙𝑡} 𝑉(𝑡))

While the general form of the aeration could be determined, the level where to start the aeration was harder to determine. It was however concluded that in order to introduce the profile to the production the exact level have to be evaluated anyway, and that one cannot solely rely on scaled results from the lab.

One part of this problem can be circumvented if the profile is set according to kLa or OTR rather than the gas velocity. For Jästbolaget this requires extra investment in gas analyzers for the reactors, preferably on both the ingoing and outgoing gas, so that kLa and OTR can be measured online. However with this in place, the aeration profiles impact in the cultivation also can be monitored online via a recalculation of the outgoing air concentration. With further work there is a possibility to differentiate between overflow and anaerobic ethanol formation via online estimation of the current cell mass and oxygen uptake rate in the bioreactor.

With the assumed numbers used in this work a saving potential for Jästbolaget was estimated to be at least 10 % reduction in air volume from the current profile.

Sammanfattning

Jästbolaget AB grundades 1983 och är idag Nordens största tillverkare av bagerijäst. Tillverkningsprocessen är en serie fed batch odlingar med de största reaktorerna på 215 m3. Luftning är viktigt i bagerijästproduktion för att få ett så högt utbyte som möjligt. Luftningen sker genom fasta spridare i botten av reaktorn, och är till volym den största råvara.

Kompressorerna som arbetar upp trycket står för ungefär 20 % fabrikens energiförbrukning, och överluftning har en negativ inverkan på odlingarna. Det har förekommit problem med under luftning, särskilt under sommaren, men normalt tros det att ett överskott av syre överförs till odlingarna. Om det finns ett sådant överskott är storleken emellertid okänd. Jästbolaget strävar också mot att producera ett flertal specialjäststammar, som i sig kan ha olika syrebehov. Detta allmänna problem kan också tänkas förekomma i många andra cellodlingsindustrier, och är av extra vikt när det kommer till produktion av en lågvärdeprodukt, då maximalt utbyte från varje råvara är av extra vikt.

Med denna bakgrund var målet med detta arbete att optimera luftningsprofiler i fed batch odlingar av Saccharomyces cerevisiae, med Jästbolagets produktionsprocess som modell. Med andra ord för att föreslå profiler där syretillförseln matchar syreförbrukningen av cellerna, vilket ger ett stabilt värde i syrekoncentrationen i reaktorn.

Om luftningen är otillräcklig producerar jästen etanol som fermentationsprodukt, vilket traditionellt har används som kontroll av att tillräcklig mängd luft tillförs i odlingen. Men etanol kan även bildas från övermatning via overflow metabolism. Detta var genom laboratorieskalaodlingar i olika tillväxthastigheter bevisat inte vara ett problem för den normala jäststammen som används. Den normala tillväxthastigheten i produktionsprocessen är under den kritiska tillväxthastigheten för overflow metabolism. Det är dock troligt att andra stammar som odlas ge upphov till detta problem.

Med detta utrett kunde det huvudsakliga målet med detta arbete kunde fortsätta, med etanol bildning som respons på underluftning. Luftningsprofiler skapades och utvärderas med simuleringar och kontrollerades i laboratorieskala reaktorer. En lämplig profil bestämdes att i allmänhet följa

𝑄(𝑘_𝐿𝑎(𝑡)) = 𝑄 (𝑐𝑒^{𝑆𝐹𝑅∙𝑡} 𝑉(𝑡))

Även om den allmänna formen av luftningsprofilen kunde fastställas, var det svårare att bestämma den exakta nivån. Det konstaterades emellertid att för att införa profilen till produktion måste den exakta nivån utvärderas i alla fall, och att man inte enbart kan förlita sig på skalade resultat från labbet.

En del av detta problem kan kringgås om profilen skulle sättas enligt kLa eller OTR istället för gashastighet.

För Jästbolaget kräver extra investeringar i gasanalysatorer för reaktorerna, helst på både ingående och utgående gas, så att kLa och OTR kan mätas i realtid. Men med detta på plats, kan också luftningsprofilernas inverkan på odlingen också följas online via en omräkning av den utgående luftkoncentrationen. Med ytterligare arbete finns det också möjligheter att skilja mellan overflow och anaerob etanolbildning via online uppskattning av den aktuella cellmassan och syreupptagningshastighet i bioreaktorn.

Med de antagna siffrorna som använts i detta arbete uppskattades en besparingspotential för Jästbolaget till minst 10 % minskning av luftvolymen från den aktuella profilen.

Contents

1 Introduction ... 1

2 Background ... 3

2.1 Basic metabolism ... 3

2.2 Aeration in bioreactors ... 6

2.3 About Jästbolaget’s production process. ... 9

3 Problem description and strategy ... 11

4 Methodology ... 13

4.1 Cultivations ... 13

4.1.1 Reactors ... 13

4.1.2 Scaling ... 13

4.1.3 Sampling and analysis ... 14

4.2 Overflow ethanol formation ... 14

4.3 k

a measurement ... 15

4.3.1 Calculation ... 15

4.4 Anaerobic ethanol formation ... 16

4.5 Simulations ... 17

5 Results ... 19

5.1 Overflow ethanol formation ... 19

5.2 k

a measurement ... 24

5.3 Anaerobic ethanol formation ... 28

5.3.1 Scaling the results ... 33

5.4 Simulation... 34

5.5 Introduction to large scale production ... 37

5.5.1 Online measurement of the metabolic state ... 37

5.5.2 Verification of the aeration profile ... 39

6 Discussion ... 43

6.1 Conclusions ... 44

6.1.1 General remarks ... 44

7 References ... i

8 Appendix ... ii

Nomenclature

Abbreviation Explanation Unit (if not otherwise

specified)

𝑨𝑯 Absolute humidity 𝑔

𝑚³

⁄

𝑪 Carbon content 𝑔𝑐𝑎𝑟𝑏𝑜𝑛

⁄ 𝑔

𝒄 Concentration 𝑔

⁄ 𝑙

𝑪𝑷𝑹 Carbon dioxide production rate 𝑚𝑜𝑙

⁄ ℎ

𝑪𝑫𝑾 Cell dry weight 𝑔

𝑫𝑶𝑻 Dissolved oxygen tension %𝑎𝑖𝑟 𝑠𝑎𝑡.

𝑭 Feed rate 𝑚³

⁄ℎ 𝒌_𝑳𝐚 Volumetric oxygen mass transfer

coefficient 1

⁄ ℎ

𝒌_𝑯 Henrys constant 𝑚³∙ 𝑏𝑎𝑟

⁄𝑚𝑜𝑙 𝒌 the ratio of specific heat

𝝁 Specific growth rate 1

⁄ ℎ

𝑴𝑸 Molar flow rate 𝑚𝑜𝑙

⁄ ℎ

𝑴 Molar mass 𝑔

⁄𝑚𝑜𝑙

𝑴𝑺𝑬 Monosaccharide equivalents 𝑔_𝑀𝑆𝐸

𝑔𝑚𝑜𝑙𝑎𝑠𝑠𝑒𝑠

⁄

𝑶𝟐 Concentration 𝑉𝑜𝑙%

𝑶𝑻𝑹 Oxygen transfer rate 𝑚𝑜𝑙𝑒

𝑚³∙ ℎ

𝑶𝑼𝑹 Oxygen uptake rate 𝑚𝑜𝑙

⁄ ℎ

𝑶𝑪𝑹 Oxygen consumption rate 𝑔𝑂₂

⁄ ℎ

𝒑 Pressure 𝑏𝑎𝑟

𝑸 Gas velocity 𝑚³

⁄ ℎ 𝒒𝒐 Specific oxygen consumption

rate

𝑔_{𝑜𝑥𝑦𝑔𝑒𝑛}

𝑔_{𝑐𝑒𝑙𝑙𝑠}∙ ℎ

⁄

𝒒_𝒔 Specific substrate uptake rate 𝑔𝑠𝑢𝑏𝑠𝑡𝑟𝑎𝑡𝑒

𝑔_{𝑐𝑒𝑙𝑙𝑠}∙ ℎ

⁄

𝑹 Gas constant 𝑚³∙ 𝑏𝑎𝑟

°𝐾 ∙ 𝑚𝑜𝑙

⁄

𝑹𝑸 Respiratory quotient 𝑚𝑜𝑙/𝑚𝑜𝑙

𝑺 Substrate concentration 𝑔

⁄ 𝑙

𝑺𝑭𝑹 Specific feed rate 1

⁄ ℎ

𝑻 Temperature °𝐾

𝒕 Time ℎ

𝑽 Volume 𝑚³

𝒘 Weight 𝑔

𝑿 Cell concentration 𝑔

⁄ 𝑙

𝒀_𝒙𝒐 Biomass yield on oxygen 𝑔_{𝑐𝑒𝑙𝑙𝑠}

𝑔𝑜𝑥𝑦𝑔𝑒𝑛

⁄

𝒀_𝒙𝒔 Biomass yield on substrate 𝑔_{𝑐𝑒𝑙𝑙𝑠}

𝑔𝑠𝑢𝑏𝑠𝑡𝑟𝑎𝑡𝑒

⁄

Index Explanation

𝑿_𝟎 Initial value

𝑿^𝒊𝒏 Inlet value

𝑿^𝒐𝒖𝒕 Outlet value

𝑿^∗ Saturated value

𝑿^𝜽 Value at STP

𝑿_𝒊 Value in inlet

1 Introduction

Jästbolaget AB was founded in 1983 and have since then produced baker’s yeast in the same location in Rotebro, in northern Stockholm. Today they are the Nordic’s largest producer, with a yearly production of about 20 000 tons of yeast. The ready product ranges all from the small 50 g packages in Figure 1, to 20 kg bags of granulated yeast, to big tanker trucks with yeast solution for the industry bakeries. Besides baker’s yeast Jästbolaget also produces a variety of special yeast, e.g. wine and beer yeasts.

The production process is fed batch cultivations of the microorganism Saccharomyces cerevisiae, is a single celled, eukaryotic fungi. It is used in baking for its ability to raise the dough by producing carbon dioxide from the available sugars. They are larger than bacteria, around 8 µm. Reproduction can occur sexually, changing genetic material, or asexually by budding. This gives rise to exponential growth when one cell divides into two, they into four etc. (Madigan et al. 2012)

Air is supplied in large amounts during the cultivations, and is to volume the biggest raw material. Without aerobic conditions the yeast only utilizes a fraction of the energy in the added substrate, the rest ends up in the fermentation product ethanol. But adding oxygen comes with a high operational cost, and is desired to be added in just the right amount. This general problem can be thought of occurring in many other cell cultivation industries as well, especially in production of low value products when maximum yield from every raw material is of extra importance.

This work’s goal was to optimize aeration profiles in fed batch cultivations of Saccharomyces cerevisiae, with the Jästbolaget’s production process used as a model. In other words find profiles that matched the oxygen transfer rate to the consumption rate, giving a steady state in the oxygen concentration in the reactor.

Figure 1 The ready yeast product, in Jästbolagets characteristic yellow 50 g package © Jästbolaget 2016

2 Background

2.1 Basic metabolism

To be able to grow yeast needs energy, that is taken from the chemical energy stored in sugar compounds via various oxidation and reduction reactions. The main yield are the energy rich compounds adenosine triphosphate (ATP) and Nicotinamide adenine dinucleotide (NADH). A general overview is shown in Figure 2. This proceeds through glycolysis in the cytoplasm converting incoming sugar molecules in a series of reactions to pyruvate. After the uptake of the sugars into the cell, the starting molecule of this reaction series is glucose. However other a variety of different molecules, e.g. fructose, can be converted into glucose and follow the same metabolism. (Cooper and Hausman 2009)

In the glycolysis a net of two molecules of ATP per molecule of glucose are generated via substrate level phosphorylation, meaning that ADP is phosphorylated via the energy from energy rich intermediates. Also two molecules of NADH is formed. (Madigan et al, 2012). Glycolysis can occur in absence of oxygen, and still provide these energy compounds to the cell (Cooper and Hausman 2009).

From pyruvate two main metabolic routes can be taken that have a great influence on the oxygen consumption. Either pyruvate proceeds into the cells mitochondria, through the citric acid cycle (TCA- cycle), or it stays in the cytosol to form ethanol. This metabolic route way is decided by the conditions that the cells are currently subjected to, causing a balance of different compounds that the chemical laws of equilibrium tries to even out. (Feldmann 2012).

The TCA-cycle is a metabolic circle with nine compounds producing two carbon dioxide molecules, which are replenished by the incoming pyruvate through oxidative decarboxylation to aceltyl-CoA, closing the carbon balance. One cycle from one molecule of pyruvate produces four molecules of NADH, one flavin adenine dinucleotide (FADH2) and one molecule of guanosine triphosphate (GTP), which is an energy carrier just as ATP. NADH and FADH2 are then used to produce ATP via respiration, or the electron transport chain, at the inner mitochondrial membrane. The phosphorylation energy is here taken from a proton gradient over the inner membrane, yielding roughly three molecules of ATP per NADH and two per FADH2. This brings a yield of roughly 36 moles ATP per mole of ingoing glucose. (Cooper and Hausman 2009).

To balance the oxidation of NADH and FADH2 oxygen is reduced to water and becomes the final acceptor of electrons. The oxygen/water redox couple has the strongest potential known is living systems, at +0.82 V, which put against the CO2/glucose potential at -0.43 V gives the possibility of the high formation of ATP.

This in total brings us to the well familiar formula for cell respiration. (Madigan et al. 2012).

𝐶₆𝐻₁₂𝑂₆+ 6 𝑂₂→ 6 𝐻₂𝑂 + 6𝐶𝑂₂

At the absence of oxygen this final acceptor of electrons is lacking, and respiration cannot take place. The TCA-cycle is not oxygen demanding in itself, and becomes a sort of horseshoe in order to still form the necessary metabolites and building blocks to grow. But without respiration NADH and FADH2 is not consumed in the same extent, and would accumulate if the TCA-cycle continued in the normal pace (Larsson 2015).

Instead the metabolic flux is pushed to ethanol formation from pyruvate, in order to balance the NADH formation and use a fermentation product as final acceptor of electrons. Pyruvate decarboxylated in the cytoplasm to acetaldehyde, which by oxidation of NADH is reduced to ethanol, becoming the new final acceptor. In this way the reduction of NAD⁺ during glycolysis balanced and no accumulation occurs, all following the general principle of fermentation. (Larsson 2015). During this metabolic route, the overall metabolisms follows the general formula:

𝐶₆𝐻₁₂𝑂₆→ 2 𝐶₂𝐻₅𝑂𝐻 + 2 𝐶𝑂₂

Figure 2 Energy metabolism of one half mole glucose in yeast (Pham et al. 1998, Cooper and Hausman 2009) During these conditions ADP phosphorylation only occurs via substrate level phosphorylation in the glycolysis, reducing the ATP yield from glucose to 2 mol/mol. The acetaldehyde/ethanol redox potential is only at -0.20 V, 0.23 V difference from the glucose that are supposed to be oxidized. As a result only a fraction of the energy in glucose is utilized, and the remainder ends up in the formed ethanol, in turn giving a low yield of cells. (Madigan et al. 2012).

Since the energy yield is lower, the metabolic flux has to increase in order to maintain the same growth, something that yeast cells are able to do. This gives rise to the so called Pasteur Effect, meaning glycolysis proceeds at a higher rate during anaerobic conditions, which was first observed by Louis Pasteur in the middle of the 19^th century (Feldmann 2012, Pasteur 1861)

If oxygen becomes available after that ethanol is already formed, this can be consumed again. This proceeds through reformation of acetaldehyde, which enters the mitochondria, is oxidized to acetate, and via acetyl- CoA enters the TCA-cycle. The formation of acetyl-CoA from acetate requires the energy from dephosphorylation of ATP, thus slightly reducing the overall yield of energy. At ethanol consumption the general formula looks like following: (Enfors et al. 1990).

2 𝐶₂𝐻₅𝑂𝐻 + 6 𝑂₂→ 4 𝐶𝑂₂+ 6 𝐻₂𝑂

Yeast cells also exhibit an overflow type ethanol formation, which occurs at high glycolytic fluxes at fully aerated conditions. To understand this one can think of it as a respiratory bottleneck, or a maximum rate of oxygen consumption, that if exceeded directs the additional part of the flux towards ethanol formation. The sugar concentration and uptake rate at which this occurs is often denoted “crit”. This is commonly also referred to as the Crabtree effect, from Herbert Crabtree that in early 19-hudreds observed that tumour tissue exhibited a reduced oxidative capacity after prolonged exposure to high carbohydrate concentrations.

However the effect here is better attributed to an overflow type metabolism (Larsson 2014, Crabtree 1929).

The exact location of this metabolic limitation is unknown. It has been proven respiratory capacity in the mitochondria can handle higher throughputs, and should not cause this limitation. The limitation may instead be among a TCA-cycle enzyme. (van Urk et al. 1988). The ability of oxidative ethanol formation may however be a well formulated strategy from the yeast. This gives rise to a “make-accumulate-consume” type of degradation, what via high glycolytic flux accumulates ethanol to poison the surrounding, efficiently fighting of other microorganisms that competes for the sugar. When the sugars later run low, the yeast can re-assimilate the ethanol with a minor yield loss. (Hagman et al. 2013).

Not all the sugars that are taken up by the cells follows the energy metabolism, this is also the main sources of components that are going to be converted into biomass. About half of the total flux takes other routes and exits from the one suggested in Figure 2, to be formed into building blocks for the cell. This fraction of the total uptake rate, qs, is called qs, anabolism. The other half goes to energy formation, according to Figure 2, called qs, energy. This can again be divided into the energy required for maintenance of the existing cell, and energy for growth of new ones. Illustrated in Figure 3. (Enfors 2011).

Figure 3 General sugar and oxygen fluxes in a cell (Enfors 2011)

The consumption of oxygen is almost entirely dependent on the energy metabolism, in order to perform cell respiration. A neglect able part is from the cell anabolism. (Enfors 2011).

2.2 Aeration in bioreactors

When Jästbolaget was founded in 1893 the approximate yield was around 10 % from the grains that were put into the process. Naturally the goal of baker’s yeast production is to maximize the yield of the cell mass, and avoid that the energy ends up in ethanol (Reed and Nagodawithana 1991). Because yeast is a low value product the yield is extra important great importance (Georgre et al. 1998). The previous section have made it clear how the oxygen influence yeast growth and that it needs to be added to achieve the best possible yield. How do we achieve oxygenation in production?

Oxygen have ha poor solubility in water compared to other nutrients in cell cultivations, meaning that a continuous supply is necessary to support higher cell concentrations. Continuous aeration of yeast cultivations started in Great Britain in 1886, and towards the end of the 19^th century this was a well- established method. However they were still ran in as batch cultivations, meaning that all the raw materials was dumped into the reaction vessel and left to grow. (Reed and Nagodawithana 1991).

This technique however had its drawbacks. Since carbohydrates are available in excess this will give a strong overflow formation of ethanol as discussed above, even though oxygen is available. The aeration however helps. At start when the cell concentration is low the bioreactors transfer capacity of oxygen will be greater than the cells consumption, meaning that all cells are fully aerated. But when cell concentration raises, so does the need for oxygen, according to

𝑂𝐶𝑅 = 𝑞_𝑜∙ 𝑋 ∙ 𝑉

The transfer capacity of the reactor however remains the same. This results in a limit where all the oxygen added is immediately consumed, and the remaining cells are left no oxygen. (Larsson 2014).

Between 1910-1920 German and Danish engineers developed another cultivation concept to circumvent this problem, the fed batch cultivation, Figure 4. The general problem was that the cells overall grew too fast, first beyond their respiration capacity, and later that the cell growth couldn’t be lowered when the bioreactors maximum transfer capacity had been reached. By instead of adding all the sugar at the start, feeding them during the cultivation these problems could be solved. This way the initial growth could be controlled by just adding an amount of sugars that allowed a certain growth, and when the transfer capacity had reached its maximum a constant feed could give an even lowered growth. The result here is that a feed profile is used, shown in green in the figure below. This is the type of cultivations that are currently ran in almost all commercial processes now, including at Jästbolaget. (Reed and Nagodawithana 1991).

Figure 4 General principle of a fed batch cultivation in yeast production. The raise in DOT is due to the increase of volume when using diluted feeds, as in yeast production.

With lower growth rate the oxygen demand is lowered according to 𝑂𝐶𝑅 = 𝑞_𝑜∙ 𝑋 ∙ 𝑉

𝑞𝑜= 𝜇 𝑌𝑥𝑜

→ 𝑂𝐶𝑅 = 𝜇

𝑌_𝑥𝑜∙ 𝑋 ∙ 𝑉

(Larson 2014) Since this is achieved through the fed batch type cultivation, it becomes unnecessary to aerate at the maximum capacity from the start, instead an aeration profile like the feed profile can be used to match the oxygen transfer to the consumption. The air that is added is a raw material like any other, so over aeration firstly comes with a cost, but also causes foaming in a higher extent. (Larson, 2014). With this background given it becomes clear that it is a must to aerate the reactor, but it is also desired to not over aerate.

If any calculations are to be made about the suitable aeration profiles like this, first the oxygen transfer capacity have to measured and quantified, so it can be matched to the oxygen consumption. What happens when oxygen is transferred to the cultivation media is that the oxygen entering in gas form in small bubbles diffuse into the media, becoming dissolved oxygen in the liquid phase. From here the oxygen must first enter the cell walls, and diffuse all the way into the site of reaction in the cell. This flow of oxygen can be estimated from Frick’s law of diffusion, which states that the molar flow through a film MQ, with the surface A and thickness δ, is dependent on difference in concentration over the film Δc and a specific diffusion constant D, according to

𝑀𝑄 [𝑚𝑜𝑙𝑒

ℎ ] = 𝐷 [𝑚²

ℎ ] ∙ 𝐴[𝑚²]∆𝑐 [𝑚𝑜𝑙𝑒 𝑚³ ] 𝛿[𝑚]

(Enfors 2011) We introduce what’s called the mass transfer coefficient, kL, which is the sum of all partial mass transfer resistances from the gas bubble to the reaction site in the cell. The denotation L comes from that the liquid side has the highest resistance and is the most important to consider (Larson 2014).

𝑘_𝐿[𝑚 ℎ] =

𝐷 [𝑚² ℎ ] 𝛿[𝑚]

The equation can now be rewritten as 𝑀𝑄 [𝑚𝑜𝑙𝑒

ℎ ] = 𝑘𝐿[𝑚

ℎ] ∙ 𝐴[𝑚²] ∙ ∆𝑐 [𝑚𝑜𝑙𝑒 𝑚³ ]

In a bioreactor we are more interested in the total flux into the volume, rather than over just one bubble area. We introduce the specific interface area a, the total bubble area A, over the whole reactor volume

𝑎 [1

𝑚] =𝑡𝑜𝑡 𝐴 [𝑚²] 𝑉 [𝑚³] And get the equation OTR

𝑂𝑇𝑅 [𝑚𝑜𝑙𝑒

𝑚³∙ ℎ] = 𝑘𝐿[𝑚 ℎ] ∙ 𝑎 [1

𝑚] ∙ ∆𝑐 [𝑚𝑜𝑙𝑒

𝑚³ ] = 𝑘𝐿𝑎 [1

ℎ] ∙ ∆𝑐 [𝑚𝑜𝑙𝑒 𝑚³ ]

(Enfors 2011) The equation constituents of two parts, kLa and Δc. The difference in concentration, often referred to as the driving force, is the saturated oxygen concentration in the media, minus the current. The consequence of this correlation is that at high oxygen concentration the transfer is slow, and at low concentrations high.

This parameter is mostly affected by the biomass consumption of the oxygen, but also different media have different levels of oxygen saturation. It can also be given in g/l or %air sat. (Garcia-Ochoa and Gomez 2009).

The kLa term is more a measurement of the bioreactors specific capacity of transferring oxygen. The media formulation again play a role, parameters as viscosity has a clear impact. The main factors however comes down to the bioreactor design, like efficient spargers, giving small bubbles increasing the interface area or height of the reactor, increasing the gas hold up that in turn lead to a higher interface area. Besides the factors that are fix in the reactor, the gas velocity in and the stirring speed increases kLa. (Garcia-Ochoa and Gomez, 2009).

Since these are the factors that we are able to influence and change during a cultivation, it becomes important to know how kLa is affected by different settings of these parameters. If a non-calescent media is used, meaning that bubbles not have the tendency to merge together into larger bubbles, bubble column reactors have a kLa(Q) that is linear. Stir tank reactors (STR) have a faster increase at low gas velocities, but becomes linear at higher. Illustrated in Figure 5. (Larsson 2014).

Figure 5 kLa(Q) for different reactors (Larsson 2014)

There are different ways to follow the aeration in a bioreactor. DOT can be measured online, and give an indication of the oxygen concentration in the reactor. For yeast ethanol can also be measured, showing which metabolism the cells are currently following. One cannot on the other hand distinguish between ethanol formed from oxygen limitation from overflow ethanol. Also, since there are great gradient differences in the reactor, measuring the concentration at one place does not say anything about the whole reactor conditions. (George et al. 1998).

From the metabolic equations under the previous section we can use what’s called the respiratory quotient (RQ) to see what metabolism the cells are currently running. If we consider the oxygen consumption and carbon dioxide formation in the equations we see that the ratio is different in each equation. If fully aerated metabolism the ratio is 1:1, giving a RQ of one mol/mol. If ethanol is formed, carbon dioxide is formed while no oxygen is consumed, giving and RQ of above one mol/mol. At ethanol uptake, more oxygen is consumed than carbon dioxide produced, giving a RQ of below one mol/mol. This gives a quick response of if ethanol is formed or consumed, but cannot differentiate between anaerobic or overflow ethanol. (Franzén 2003, Reed and Nagodawithana 1991)

0 0 kLa

Q

STR Bubble column

2.3 About Jästbolaget’s production process.

The production process in at Jästbolaget is a series of cultivation steps, from shake flasks in the lab to the final bubble column reactors at 215 m³, 17.5 m high and 4 m diameter. The first large scale reactor is run as batch in order to get up cell mass for inoculation of the bigger reactors. The two final steps are however fed batches, the first at 80 m³ and the final at 215 m³. These reactors are located in the highest part of the middle building, and goes from ground level all the way to the top Figure 6.

Air is supplied through fixed spargers in the bottom of the reactors, with maximum gas velocity of 7600 m³/h at 1.2 bar pressure. About 85 000 m³ of air is added during the final cultivations. This powerful aeration also serves as stirring in the reactors. Start water volume in the reactors are 80 m³, with an end volume of around 120 m³, however towards the end the aeration raises the cultivation broth to an almost full reactor. Temperature is monitored and controlled with cool water in a jacket around the reactors. pH increases during the cultivations, and are kept constant at 5 using sulfuric acid. Ethanol concentration are also monitored online.

The source of sugar are molasses containing about 30 mass percent sucrose, a disaccharide of fructose and glucose. Hydrolysis are done by the yeast enzyme invertase, resulting in 15 mass percent fructose and 15 mass percent glucose in the feed. Glucose is preferred over fructose, and consumed before fructose, however not to the extent that fructose accumulate. Both sugars are consumed in a reasonably rate, and the total sugar concentration are kept low throughout the cultivation unless extensive overfeeding occurs. (Enfors et al 1990).

Ammonia are the main source of nitrogen and are added according to a profile just as the feed of molasses, to give the yeast different characteristics. Phosphoric acid are used as source for phosphorus, and magnesium sulfate as magnesium, but also large contribution to the trace element also comes from the molasses. Biotin and thiamin are also added as vitamins. To control the foaming foam reducing substance are added.

Figure 6 Overview of Jästbolaget production site © Jästbolaget 2016 Permission to publish from Jästbolaget, Anders Wallin

3 Problem description and strategy

Aeration is essential in baker’s yeast production in order to get the yield as high as possible and avoid ethanol formation. This is extra important since this is a low value product with the raw materials as the biggest production cost, so loss of yield can’t be accepted if the factory are to stay in business. Aeration is however also expensive, the compressors working up the pressure before adding the air stand for about 20 % of the factory’s energy consumption. Over aeration also causes additional foaming which is a well-known problem in cell cultivations, and the foam reducing substance added to control this have a negative impact on the cultivation. This general problem can be thought of occurring in many other cell cultivation industries apart from baker’s yeast production, and is of extra importance at production of a low value product.

For Jästbolaget’s specific case there have been problems with ethanol formation, especially during summer time, but if this comes from overflow metabolism or oxygen limitation is partially unknown. It’s believed that an excess of oxygen is present during the cultivations, but the size of this is unknown. Jästbolaget also strives towards producing a variety of special yeasts other than the old traditional strain, meaning that a greater understanding of the aeration is necessary.

From this background the hypothesis formed in this work was the following; That an aeration profile can be ran in a production process so that from a non-ethanol producing oxygen supply match the increase to the increase in oxygen consumption by the cells, giving a steady state in the oxygen concentration in the reactor.

Looking at the equations for oxygen uptake and oxygen transfer the solution was believed to be when the following condition is fulfilled

𝑞𝑜∙ 𝑋 ∙ 𝑉 = 𝑘𝐿𝑎 ∙ ∆𝑐

The goal was then consequently to adjust aeration profiles in such a way, using Jästbolaget’s fed batch cultivations of Saccharomyces cerevisiae as a model. To quantify oxygen limitation it is a good strategy to attribute this to formation of the fermentation product ethanol. But due to the yeast’s double nature of ethanol formation the strategy in turn was divided into two main partial goals. The first was to determine at which sugar uptake rates overflow ethanol formation takes place, to be used as a negative in the second goal to determine at which oxygen uptake rates ethanol is formed. Without the former it is impossible to say anything when ethanol is formed during the latter.

To scale results between different reactors and perform any kind of calculations the oxygen transfer capacity of the reactors had to be measured. To also be able to design profiles the capacity’s relation to an operating parameter had to be investigated. The main parameter that comes to mind in of course the gas velocity.

The strategy then consisted of determination of:

1. When overflow ethanol is formed, qEtOH(qs)

To be used as a negative in the oxygen limitation experiments 2. Oxygen transfer capacity of the reactors, kLa(Q)

To be able to scale results and calculate profiles

3. When anaerobic ethanol is formed, qEtOH(qo, kLa, Q)

To design aeration profiles where OTR=OCR and no ethanol is formed.

4 Methodology

4.1 Cultivations 4.1.1 Reactors

The production scale reactor investigated at Jästbolaget was a bubble column reactor at 215 m³. Data for calculations were taken from the control system. The outgoing concentrations of oxygen, carbon dioxide and humidity along with temperature and pressure was measured with a temporary installation of a BlueInOne gas analyzer (BlueSens, Germany).

The lab scale reactors used in the experiments were two Lars 15 liter STR (Belach, Sweden). The start volume was 7 liters. To monitor and control the cultivations the reactors were equipped with the standard measuring equipment (temperature, pH, DOT, etc.) but also alcocontrol ethanol sensors (Cetotech, Germany) and tandem gas analyzers (Magellan instruments, UK)

4.1.2 Scaling

The recipe and feed profiles from the commercial process was scaled down to lab scale in according to volume, meaning a scale factor of approximately 11430. Inoculum, molasses, chemicals and vitamins used in the lab were taken from the production, except from ammonia and sulfuric acid that were taken from the laboratory (Merck, Germany). Other cultivation conditions such as temperature and pH were the same as in the commercial process.

In the production process investigated the molasses are firstly fed exponentially at approximately a SFR of 0.2 for six hours, followed by a constant feed for eight hours, in a traditional fed batch type cultivation. In the exponential phase the SFR are initially slightly above 0.2 and towards the constant phase slightly below.

For simplicity the molasses were fed at a constant SFR throughout the whole exponential phase in the lab scale. The goal were also to reach steady state in the cultivations, to ensure that the data that were to be obtained would have stabilized and reached a constant values.

The constant phase in a fed batch is initiated due to that the maximum OTR have been reached, and the oxygen consumption instead needs to be controlled by reduced feeding. Since the goal with this project was to optimize the gas flow profile according to the given feed profile, the constant phase was disregarded from the experiments. Instead the focus was put on the exponential phase. This also had the advantage of bringing the cultivations to around six hours, so that they could be performed during one day.

The feed profile in the experiments then followed:

𝐹(𝑡) = 𝐹₀𝑒^{𝑆𝐹𝑅∙𝑡} with

𝐹0=µ ∙ 𝑋₀∙ 𝑉 𝑌𝑥𝑠∙ 𝑆𝑖

(Enfors 2011)

4.1.3 Sampling and analysis

CDW was taken with half hour interval to follow the cultivations. Around 7 ml of cultivation broth was taken and weighed in centrifuge tubes. The tubes were centrifuged for 3500 RPM for 10 minutes, the supernatant removed, washed with deionized water, centrifuged again and the supernatant once again removed. The remaining pellet was dried overnight at 105 °C, let cool down in presence of silica, and weighed. The cultivation broth was assumed to have the same density as water, and the cell concentration calculated as

𝑋 = 1000 ∙𝑤𝑑𝑟𝑖𝑒𝑑 𝑡𝑢𝑏𝑒− 𝑤𝑒𝑚𝑝𝑡𝑦 𝑡𝑢𝑏𝑒

𝑤𝑠𝑎𝑚𝑝𝑙𝑒

The total cell dry weight was calculated from

𝐶𝐷𝑊 = 𝑋 ∙ 𝑉

To measure the sugar concentrations during the cultivation the method described by Larsson and Törnkvist (1996), was used. About 2 ml sample was taken fast with a pre weighed syringe containing 2 ml of 0.66 M perchloric acid, stopping the metabolism of the remaining sugar in the sample. The syringe was weighed again to get the exact sample volume, and centrifuged at 4500 RPM for 5 minutes. The supernatant was collected and frozen.

On the analysis day the samples were neutralized with 250 µl of concentrated potassium carbonate (500 g/l), put on ice for 15 minutes and centrifuged again. The supernatant was collected, centrifuged again, and the supernatant analyzed with HPLC (Bio-rad Aminex HPX-87H column). The measured response was compared with standard solutions for sucrose, glucose and fructose, and a concentration in g/l was obtained.

This concentration was compensated according to the dilution from the acid and potassium carbonate according to

𝑐𝐶𝑢𝑙𝑡𝑖𝑣𝑎𝑡𝑖𝑜𝑛= 𝑐_{𝐻𝑃𝐿𝐶}2.25 + 𝑤𝑠𝑎𝑚𝑝𝑙𝑒

𝑤_{𝑠𝑎𝑚𝑝𝑙𝑒}

4.2 Overflow ethanol formation

To be able to draw any conclusions when ethanol was formed during the oxygen limitation experiments the ethanol formation from overflow metabolism needed to be determined. More precisely qEtOH(qs) needed to be determined. This was done with a series of cultivations with increasing SFR, from 0.18 to 0.23. These were performed with no oxygen limitation, to be sure that ethanol formed comes from overflow metabolism.

CDW and sugar concentration samples were taken every half an hour. µ and CDW0 was calculated by fitting the CDW data in excel to the equation

𝐶𝐷𝑊(𝑡) = 𝐶𝐷𝑊₀𝑒^𝜇∙𝑡 From this the yield was calculated from

𝑌_𝑥𝑠= 𝐶𝐷𝑊 − 𝐶𝐷𝑊0

𝑤𝑡𝑜𝑡 𝑚𝑜𝑙𝑎𝑠𝑠𝑒𝑠∙ 𝑆_𝑖 qs was calculated from

𝑞_𝑠=𝐹 ∙ 𝑆_𝑖−∆(𝑐𝐶𝑢𝑙𝑡𝑖𝑣𝑎𝑡𝑖𝑜𝑛∙ 𝑉)

∆𝑡 𝐶𝐷𝑊

The feed assumed to contain only sucrose, and for glucose and fructose the MSE factor was divided by two to get the correct inlet flow.

4.3 k

a measurement

To scale the results from the experiments to the production scale kLa for both reactors had to be determined.

Also kLa(Q) is not necessary linear, and also needed to be investigated in the two reactors to be able to design the appropriate aeration profiles.

For the production scale reactor data was collected from one cultivation and kLa calculated. Firstly cultivations were performed in the lab scale reactors to find a fix stirring speed that would be able to give the whole span of kLa as in the production scale reactor at different gas velocities. This way a profile could be created by just warrying the gas velocity. To find this, the whole span of kLa for the production reactor was measured, and a stirring speed for the lab scale reactor that was in the middle of this span was chosen by varying the stirring speed over a medium gas velocity, 10 l/min. With this stirring speed fixed, the gas velocity was then increased from the lowest setting to determine kLa(Q) for the lab scale reactors.

4.3.1 Calculation

Q^out and Qⁱⁿ was calculated from a mass balance on nitrogen, with the assumption that no nitrogen is consumed or added during the process,

𝑁2𝑖𝑛 = 𝑁2𝑜𝑢𝑡

For the lab scale reactors Qⁱⁿ was logged by the control system, and Q^out needed to be estimated. The nitrogen balance gives the correlation

𝑄^𝑜𝑢𝑡= 𝑄^𝑖𝑛 100 − 𝑂₂^𝑖𝑛− 𝐶𝑂₂^𝑖𝑛 100 − 𝑂₂^𝑜𝑢𝑡− 𝐶𝑂₂^𝑜𝑢𝑡

(Enfors 2011, Larsson 2014) where the concentration of the components in the gas are volume percent. The inlet values were taken from the initial value from the cultivations by running air through the cultivation media before inoculation. The outlet gas for the lab scale reactors was condensated, and assumed completely dry.

For the production scale reactor Q^out was instead logged, and Qⁱⁿ needed to be calculated. The absolute humidity, temperature and pressure was logged via the gas analyzer, which were converted to volume percent water vapor by assuming ideal gases

𝐻₂𝑂^{𝑖𝑛/𝑜𝑢𝑡} =𝐴𝐻^{𝑖𝑛/𝑜𝑢𝑡}∙ 𝑅 ∙ 𝑇 𝑀_𝐻2𝑂∙ 𝑝 Qⁱⁿ was then calculated from the correlation

𝑄^𝑖𝑛= 𝑄^𝑜𝑢𝑡100 − 𝑂₂^𝑜𝑢𝑡− 𝐶𝑂₂^𝑜𝑢𝑡− 𝐻₂𝑂^𝑜𝑢𝑡 100 − 𝑂₂^𝑖𝑛− 𝐶𝑂₂^𝑖𝑛− 𝐻2𝑂^𝑖𝑛

From the inlet and outlet gas flow the molar flow rate of oxygen for the inlet and outlet of gas was calculated from

𝑀𝑄_𝑂

2 𝑖𝑛/𝑜𝑢𝑡

= 𝑄^{𝑖𝑛/𝑜𝑢𝑡}𝑂₂^{𝑖𝑛/𝑜𝑢𝑡} 100

𝑝 𝑇 ∙ 𝑅

(Garcia-Ochoa and Gomez 2009) assuming ideal gasses. For the production scale reactors the pressure for the inlet gas was calculated as the set point value 1.2 bar, and for the outlet the logged values from gas analysis were used. Temperature for the inlet was calculated as 80 °C in accordance to what was believed at Jästbolaget, and for the outlet the logged value from the gas analyzer. This increase of temperature is due to the compression of the air. For the lab scale the pressure was set to 1 bar and the temperature to the cultivation temperature 30 °C

The difference in the inlet and outlet molar flow rates gives the oxygen consumption rate according to 𝑂𝑈𝑅 = 𝑀𝑄_𝑂^𝑖𝑛₂− 𝑀𝑄_𝑂^𝑜𝑢𝑡₂

(Garcia-Ochoa and Gomez 2009)

The solubility of oxygen was determined with henrys constant which was calculated from

𝑘𝐻= 1 𝑘_𝐻^𝜃𝑒⁽

−∆𝑠𝑜𝑙𝐻 𝑅 (1

𝑇−1 𝑇^𝜃))

(Sander 2015) Values for –ΔsolH and 1/kHθ for pure water were given by Sander (2015).

For the lab scale reactors the driving force was then calculated from the DOT value according to

∆𝑐𝑂₂= 𝑐𝑂∗₂− 𝑐𝑂₂=

𝐷𝑂𝑇^∗ 100

𝑂₂^𝑖𝑛 100

1 𝑘𝐻

−𝐷𝑂𝑇 100

𝑂₂^𝑖𝑛 100

1 𝑘𝐻

(Enfors 2011 and Larsson 2014) To account for the gradient difference of oxygen concentration in the production scale reactor the driving force was estimated from the logarithmic mean between the inlet and outlet concentration of oxygen.

𝑐_𝑂^{𝑖𝑛/𝑜𝑢𝑡}₂ = 𝑝 𝑘_𝐻

𝑂₂^{𝑖𝑛/𝑜𝑢𝑡} 100

∆𝑐𝑂₂ = 𝑐_𝑂^𝑖𝑛₂− 𝑐_𝑂^𝑜𝑢𝑡₂ 𝑙𝑛(𝑐_𝑂^𝑖𝑛₂) − 𝑙𝑛(𝑐_𝑂^𝑜𝑢𝑡₂ )

(Garcia-Ochoa and Gomez 2009) The pressure for the inlet gas was calculated as the set point value 1.2 bar, and for the outlet the logged values from gas analysis were used.

With the driving force and OUR estimated kLa for the reactors was calculated from 𝑘𝐿𝑎 = 𝑂𝑈𝑅

𝑉 ∙ ∆𝑐_𝑂2

(Garcia-Ochoa and Gomez 2009)

4.4 Anaerobic ethanol formation

The goal with this section, and more or less the whole project was to find an aeration profile that would give a steady state in the DOT and without any ethanol formation. To do this different models for the profile was derived and evaluated in the lab scale reactors. The first model was an exponential increase of kLa to account for the increase of cell mass according to:

𝑘_𝐿𝑎(𝑡) = 𝑘_𝐿𝑎₀∙ 𝑒^{𝑆𝐹𝑅∙𝑡}

This model was tested, and from the results the model was optimized. This is later found under results.

qEtOH(qo) was estimated by following qo during the cultivations. This was calculated from OUR and CDW 𝑞_𝑜 = 𝑂𝑈𝑅

𝐶𝐷𝑊 (Pham et al 1998)

To verify the ethanol formation that the ethanol sensors indicated the RQ was also calculated. In similar way as the OUR was calculated the CPR was calculated

𝐶𝑃𝑅 = 𝑀𝑄_𝐶𝑂^𝑖𝑛₂− 𝑀𝑄_𝐶𝑂^𝑜𝑢𝑡₂

(Pham et al 1998) By dividing CPR by OUR the RQ value was obtained

𝑅𝑄 =𝐶𝑃𝑅

𝑂𝐶𝑅 (Pham et al 1998)

4.5 Simulations

To verify the results from the cultivations with theory simulation was used. At the end this was also used to test how the new aeration profiles would behave in the production process. For this the MATLAB based SimuPlot4 was used (Enfors 2015). The general mass balances for the fed batch model were

𝑑𝑆 𝑑𝑡 =𝐹

𝑉𝑆𝑖−𝐹

𝑉𝑆 − 𝑞𝑠∙ 𝑋 𝑑𝑋

𝑑𝑡 = 𝜇 ∙ 𝑋 −𝐹 𝑉𝑋 𝑑𝑉

𝑑𝑡 = 𝐹 𝑑𝐷𝑂𝑇

𝑑𝑡 = 𝑘_𝐿𝑎 ∙ ∆𝐷𝑂𝑇 − 𝑞_𝑜∙ 𝑋 ∙ 𝐻 where H is derived from henrys constant according to

𝐻 =100 ∙ 𝑘𝐻

𝑝_𝑜2

(Enfors 2011 and Larsson 2014) The basic equations used to obtain values for the mass balances used where the following. For simplicity the maintenance was excluded.

𝐹 = 𝐹0𝑒^{𝑆𝐹𝑅∙𝑡} 𝑞_𝑠 =𝑞_{𝑠 𝑚𝑎𝑥}∙ 𝑆

𝑆 + 𝐾_𝑠 𝑞_{𝑠 𝑎𝑛}= 𝑞_𝑠∙ 𝑌_𝑥𝑠𝐶𝑥

𝐶_𝑠 𝑞_𝑜= 𝑌_𝑜𝑠∙ (𝑞_𝑠− 𝑞_{𝑠 𝑎𝑛})

𝜇 = 𝑌_𝑥𝑠∙ 𝑞_𝑠

(Enfors 2011 and Larsson 2014) Constant values were taken from Pham et al 1998.

To simulate an aeration profile that would result in a steady state of the oxygen supply additional equations were added. At steady state the change in DOT is zero, or the consumption and transfer of oxygen is equal, which gives following expression for kLa:

𝑑𝐷𝑂𝑇

𝑑𝑡 = 0 → 𝑘𝐿𝑎 =𝑞_𝑜∙ 𝑋 ∙ 𝐻

∆𝐷𝑂𝑇

To then model Q, the correlation between kLa and Q had to be taken into account:

𝑄 = 𝑄(𝑘_𝐿𝑎) This is later derived from the experiment results under results.

To simulate the concentration of oxygen in the outlet the following expression was used

𝑂2𝑜𝑢𝑡=(𝑉𝑜𝑙 𝑜𝑥𝑦𝑔𝑒𝑛 𝑖𝑛 − 𝑉𝑜𝑙 𝑐𝑜𝑛𝑠𝑢𝑚𝑒𝑑 𝑜𝑥𝑦𝑔𝑒𝑛)

𝑄^𝑜𝑢𝑡 =

(𝑄^𝑖𝑛∙ 𝑂₂^𝑖𝑛− 𝑂𝐶𝑅100 ∙ 𝑅 ∙ 𝑇 𝑝 ∙ 𝑀𝑤 𝑂₂ ) 𝑄^𝑜𝑢𝑡

with

𝑂𝐶𝑅 = 𝑞_𝑜∙ 𝑋 ∙ 𝑉 and

𝑄^𝑖𝑛 = 𝑄^𝑜𝑢𝑡= 𝑄 giving

𝑂₂^𝑜𝑢𝑡= 𝑂₂^𝑖𝑛−𝑂𝐶𝑅 𝑄

100 ∙ 𝑅 ∙ 𝑇 𝑝 ∙ 𝑀_{𝑤 𝑂2}

(Enfors 2015)

5 Results

The goal of this work was to design aeration profiles that matched the oxygen transfer from the aeration to the oxygen consumption by the cells, so that no excess of oxygen is built up in the reactors. This was done via lab scale cultivations using ethanol formation as a sign of oxygen limitation, simulations of theoretical cultivations, and measurements from Jästbolaget’s commercial process. The main parts were to investigate where the limit for overflow ethanol were, to measure kLa for the reactors and to determine when anaerobic ethanol formation occurs.

5.1 Overflow ethanol formation

To be able to draw any conclusions when ethanol was formed during the oxygen limitation experiments the ethanol formation from overflow metabolism needed to be determined, to be used as a negative in the later experiments. More precisely qEtOH(qs) needed to be determined. Six cultivations were performed with increasing SFR, from 0.18 to 0.23 under fully aerated conditions in order to find the critical qs where ethanol is formed from overflow metabolism. A typical cultivation is shown in Figure 7, with its corresponding concentrations of sugars in Figure 8.

Figure 7 Cultivation at 0.2 SFR

A short ethanol peak was always present during the first hour of the feed before the cells adjusted and consumed the ethanol. For the remainder of all the cultivations the no ethanol was formed.

Figure 8 Sugar concentrations for the cultivation at 0.2 SFR

As seen in Figure 8 glucose was consumed immediately, while the levels of fructose and sucrose built up during the cultivations. At steady state these concentrations should level out and be constant, and since this is not the case a perfect steady state was never reached during these cultivations. However a tendency to a steady level was seen toward the end.

Towards the end feeds were at 400 g/h which gave a maximum sucrose concentration of approximately 0.8 g/l at 9 l, meaning that the feed was consumed in about 3:40 minutes.

𝑡_{𝑐𝑜𝑛𝑠𝑢𝑚𝑡𝑖𝑜𝑛}=𝑐_{𝑠𝑢𝑐𝑟𝑜𝑠𝑒}∙ 𝑉

𝐹 ∙ 𝑆_𝑖 = 0.8 𝑔

⁄ ∙ 9𝑙𝑙 400

60 ∙ 60

𝑔⁄ ∙ 0.295 𝑔 𝑔𝑠 ⁄

= 219.66 𝑠 ~3 min 40 𝑠

Scrit have in literature been estimated to 0.110 g/l glucose plus fructose at growth on molasses (Enfors et al.

1990). This limit was then almost instantaneously reached in the cultivations, and the results from this work points to a higher Scrit. This variance might be due to that different strains were used or faulty measurements.

Table 1 Summary of the cultivations for overflow ethanol formation

SFR /h 0.18 0.19 0.20 0.21 0.22 0.23

µ /h 0.20 0.23 0.22 0.25 0.26 0.27

qsuc g/g,h 0.27 0.36 0.31 0.42 0.44 0.46

qo g/g,h 0.11 0.16 0.13 0.19 0.18 0.20

qs/qo mol/mol 0.23 0.21 0.23 0.21 0.23 0.22

CDW0 g 90 59 89 61 62 64

Yxs g/g 0.67 0.57 0.66 0.54 0.55 0.55

In Table 1 a summary of all the cultivations can be seen. Values are the average of the data points towards the end, when a more or less constant phase had been reached. What can be seen is that µ was consequently higher than SFR throughout all cultivations, illustrated in Figure 9.

Figure 9 µ(SFR) in the experiments

Two main factors can be offered as explanation why this is, looking at the equation for F0. 𝐹0=µ ∙ 𝐶𝐷𝑊₀

𝑌𝑥𝑠∙ 𝑆𝑖

First the yield was higher than the yield assumed in the calculations. The calculated yield was at 0.5 g/g, but we see from the results that this was more around 0.55 g/g. For SFR 0.18 and 0.20 this value landed on the unrealistic 0.67 and 0.66, which is probably due to errors in measurement. Higher yields have been noted in lab scale compared with production scale in in molasses fed yeast cultivations (George et al 1998).

Secondly CDW0 was lower than expected. The calculated value was 87.5 g cells, but in some cases landed on values around 60 g. For SFR 0.18 and 0.20 the about right inoculum size was achieved, and as a result we see that the difference between SFR and µ for these cultivations were 0.02 /h, while for the others 0.04 /h.

These then only had the increased yield as contributing factor to the increased growth rate. Inoculum for the cultivations SFR 0.18 and 0.20 were taken from one batch and the others from another, and the error here is suspected to be dry weight value for the second batch that were given from the production.

As a consequence the yeast grew faster than intended. However as shown in Figure 10 the cell still grew at an almost constant growth rate, although higher than intended.

0,19 0,21 0,23 0,25 0,27

0,17 0,18 0,19 0,2 0,21 0,22 0,23 0,24

µ [/h]

SFR [/h]

y = 88,644e^0,2176x R² = 0,9959

0 1 2 3 4 5 6 7

0 400

Cultivation time [h]

CDW [g]

If the higher yield and lower inoculum are put into simulation in Figure 11 we see that the growth rate should decrease slightly during the span of the experiment. But no drastic jumps should be expected, and a more or less constant growth can be expected toward the end. However we see that the growth rate indeed should be higher than intended, and should be accounted for when conclusions are made.

Figure 11 Simulation of a to high F0

If this is taken into account and the data is plotted against µ rather than SFR we end up with Figure 12.

Figure 12 q0(µ) and qs(µ)

We can see that there is a clear relation between qs and qo, at approximately 0.22 mol/mol. The general equation for aerobic growth on carbohydrates gives this value to 0.33 mol/mol (note that sucrose contains two MSE, and half the flux goes to cell anabolism), which is the largest part of the oxygen consumption. This lowered value can be contributed errors in measurement, and indicates there are some error in the values.

0,00 0,50

0,00 0,25

0,19 0,21 0,23 0,25 0,27

qsuc[g/g,h]

qo[g/g,h]

µ [/h]

One of the main conclusions from this sections is that the overflow ethanol formation starts at higher molasses uptake rates than originally believed. The whole span that was investigated turned out yield no ethanol. Literature for industrial production puts this at approximately growth rates above 0.2 /h (Reed and Nagodawithana 1991). However in continuous cultivation experiments values of up to 0.38 /h have been reported (Postma et al. 1989, Rieger et al. 1982). One explanation for this variation is that this value is very strain dependent, and depending on the stain used in the experiments different values can be expected (Reed and Nagodawithana 1991).

To still get a rough estimation where this critical limit lays for this strain data from a test run before the experiments was used. This had a SFR of 0.25, with otherwise the same conditions as the other cultivations, and gave strong ethanol formation, Figure 13.

Figure 13 Cultivation at 0.25 SFR

No samples were sadly taken during this trial run. Instead linear approximations were made from Figure 9 and Figure 12 up to the current set point to get value for µ, qo and qsuc, shown in Table 2.

Table 2 The theoretical data for SFR 0.25

SFR /h 0.25

µ /h 0.30

qsuc g/g,h 0.54

qo g/g,h 0.23

This very rough estimation gives the result that the overflow ethanol comes somewhere after a qsuc of 0.5 g/g,h, with a corresponding µ of 0.30 1/h at fully aerated conditions. Regardless of the errors in the experiment it can be clearly stated that growth rates around 0.20 1/h that are ran in the production process should not give any overflow ethanol.

Since this experiment was performed at fully aerated conditions we also get the required qo at different growth rates and corresponding qsuc. We see that for µ of 0.20 1/h with qsuc at 0.27 g/g,h qo is about 0.11 g/g,h. Values below this should then give ethanol formation. However since some errors are expected in oxygen uptake rate values as discussed above this value should not be trusted entirely.

5.2 k

a measurement

To scale the results from the experiments to the production scale kLa for both reactors had to be determined.

Also kLa(Q) is not necessary linear, and also needed to be investigated in the two reactors to be able to design the appropriate aeration profiles. Data was collected from one cultivation in the production scale reactor, shown in Figure 14. Following in Figure 15 the kLa calculated from this data plotted against Q

Figure 14 Raw data from the production scale reactor. Qⁱⁿ in light blue and Q^out in dark blue, CO2out in light green and O2out in dark green

Figure 15 kLa(Q) for the production scale reactor

We see that kLa(Q) seems to have a linear correlation, which is also supported in literature for bubble column reactors (Larson 2014). This is the correlation that was used for calculations for the remainder of this work.

0 0

4

concentration [Vol%]

Q [m3/h]

time [h]

y = 0,122x - 55,727 R² = 0,9511

0 1000

0 2000 4000 6000 8000

kLa [/h]

Q [m³/h]

To get a better fit one could also have taken the correlation from kLa plotted against Q/V, to get the units right. This is illustrated in Figure 16. As shown many of the outliers from the previous plot becomes more concentrated in this method. The correlations are however similar, and should give roughly the same values.

Figure 16 kLa(Q/V) for the production scale reactor

No measurement was available of the inlet temperature, these plots were calculated with 80°C fixed as inlet temperature, according to what was believed to be at Jästbolaget. This factor influences the inlet flow of oxygen greatly, according to

𝑀𝑄𝑂𝑖𝑛₂ = 𝑄𝑂𝑖𝑛₂𝑂₂^𝑖𝑛 100

𝑝 𝑇 ∙ 𝑅

meaning that the inlet concentration is inversely proportional to temperature, higher kLa at lower temperatures. If we instead assume 60 and 40°C the correlations end up as in Figure 17.

Figure 17 kLa(Q) for the production scale reactor at 40, 60 and 80 °C 0

1000

0 20 40 60 80

kLa [/h]

Q/V [/h]

0 1000

0 2000 4000 6000 8000

kLa [/h]

Q [m³/h]

To offer another way of thinking this inlet temperature can be calculated. Assuming adiabatic conditions, no loss of heat to the surroundings, the temperature increase of compression can be described as

𝑇₂= 𝑇₁(𝑝2

𝑝₁)

𝑘−1 𝑘

(Atkins and De Paula, 2009) where k is the ratio of specific heat. This is then theoretical maximum, with no heat loss. Before compression the air led through a pipe system, so say that the ingoing air is room temperature. This gives an outgoing temperature of:

𝑇2= 𝑇1(𝑝₂ 𝑝1

)

𝑘−1

𝑘 = 298 °𝐾 (1.2 𝑏𝑎𝑟 1 𝑏𝑎𝑟 )

1.4−1 1.4

= 313.93 °𝐾 ~41 °𝐶

Of course there might be other factors affecting the temperature, raising the temperature even more.

However this illustrates that it is reasonable that the actual value might differ from 80°C, and that the temperature needs to be measured if the aeration is to be calculated with more certainty.

To think further, since the air is taken from outside it is likely that this factor changes during season, and also during night and day, meaning that this might have saving potential on a daily basis. The equation for the molar flow of ingoing oxygen shows that the concentration indeed decrease during higher temperatures.

This is also the general experience at Jästbolaget. However without the temperature the effects of this just becomes guesses, and are hard to quantify.

To circumvent this whole problem, kLa and OTR could be measured online, and the aeration intensity be regulated to a profile be set according to the actual kLa or OTR instead of gas velocity. This requires gas analyzers on the outgoing air for all the reactors, and preferably on the ingoing air as well. However with this in installed the aeration intensity would be immediately regulated and corrected to any temperature differences, potentially reducing to energy consumption.

For the work at hand 80°C inlet temperature was assumed, and from Figure 15 it was concluded that to model the production scale reactors in the lab the lab reactors had to cover a kLa interval of roughly 200- 1000 /h. At medium aeration (10 l/min) the rpm was increased from 500 to 1500 and the kLa calculated, Figure 18.

Figure 18 kLa(rpm) at 10 l/min 0

1200

500 700 900 1100 1300 1500

kLa [/h]

Stirrer speed [RPM]