Agitation-Aeration in Submerged Fermentation

I. A Comparative Study of the Sulfite and Polarographic Methods for MeasuringOxygen Solution Rates in a Fermentor

R. STEEL1 AND M. R. BRIERLEY2

Indu4strial Biochemistry Section, College of Science and Technology, M1anchester, England

Received for puiblication July 21, 1958

Although the submerged aerobic cultivation ofmicroorganisms has become an established featureof the present-day fermentation industry, equipmentdesign and operation have been based largely uponiempirical studies. Attention has been directed to theproblem of supplying sufficient oxygen to microbialcultures and to methods for defining aeration efficiencyin fermentors (reviewed by Finn, 1954). The sulfiteoxidation method for determining oxygen solutionrates in a fermentor has been used to test the efficiencyof equipment design (Chain and Gualandi, 1954;Roxburgh et al., 1954; Bowers, 1955) as a parameterfor increase in scale (Cooper et al., 1944; Karow et al.,1953) or as a measure of the actual amount of oxygentransferred to a microbial culture (Maxon anid Johnson,1953; Pirt, 1957).

In certain cases, measurements of oxygen solutionrate determined by the sulfite method have agreedwith those obtained by the polarographic technique(Bartholomew et al., 1950; Chain and Gualandi, -1954)whereas in others (Wise, 1950), a large disparity hasbeen noted between the results obtained by the twomethods. Because of the lack of agreement betweenthe results of different investigators, the present workwas undertaken to compare the sulfite and polarographictechniques. Oxygen absorption-coefficients were meas-ured over a fairly wide range of agitator speeds andair flow rates and correlations between absorptioncoefficients and operating variables were determined.The results indicate the limitations of the sulfite oxi-dation technique as applied to fermentations.

MATERIALS AND METHODS

Theoretical. The theory of oxygen absorption in afermentor liquid has been adequately discussed byBartholomew et al. (1950), Wise (1951), Finni (1954),and Elsworth et al. (1957).

Fermentor and accessories. Fermentor design was ofthe conventional type (Finn, 1954) consistinig of a

1 Present address: The Upjohn Company, Kalamazoo,Michigan.

2 Present address: Department of Biological Sciences, WyeCollege, near Ashford, Kent, England.

6-in. diameter glass tank fitted with 4 diametricallyspaced baffles. An open-pipe sparger was locatedcentrally below an 8-bladed turbine impeller. Theratio of the tank diameter, impeller diameter, bladelength, and blade width was 15:5:1.25: 1. The impellerwas situated one impeller diameter from the base ofthe tank; the operating liquid volume was 2.5 L.The fermentor was kept in a temperature controlledwater bath (25 ± 1 C). Air flow rates were measuredwith capillary orifices calibrated at 25 C and atmos-pheric pressure. Agitator speed was determined fromthe current produced by a tachometer-generator drivenby the impeller shaft, the ammeter scale-deflectionhaving been calibrated against impeller speed with astroboscopic flash unit.

Sulfite-oxidation experiments. The method of Cooperet al. (1944) was used. The recommendations of Finn(1954) have been followed with regard to the use ofunits; the absorption coefficients (KLa values) areexpressed as mmoles of oxygen per L per hr (unitconcentration difference), or hr-1.

Polarographic equipment and techniques. A manualpolarograph3 was used in all experiments. Preliminarystudies dealt with the design and operating char-acteristics of "axial" and "radial" rotating platinum

3 Electrochemical Laboratories, Mlanchester, England.

CONTACT WIRE

-RUBBER SLNG

PVC BE-ARING

LEAD TO BRASS CYLINDER 3ODPGM

-GLASS CUP1 -RUBBER BUNG

K- tMERCURY -WIRE LEADPVC BEARINGSA'KC

--GLASS TUBE-RUBBER STOP MERCURYGLASS CAPILLARY

RA S SNGLIOD d' UBRBN

-z- -IIIGLASS CASING

PLATINUM '-SILVER FOILBEAD 14

U SINTERED DISC

RADIAL ROTATING AXIAL SILVER-SILVER CHLORIDEPLATINUM CATHODE CATHODE ANODE

Figllre 1. Construction of electrodes

51

on April 22, 2020 by guest

http://aem.asm

.org/D

ownloaded from

R. STEEL A'ND R.R. BRIERLEY

cathodes used in conjunction with saturated calomeland silver/silver chloride reference electrodes. Thedesign and construction of electrodes are shown infigure 1. Satisfactory operation of the electrode systemwas found to be related to the tip velocity of the cathode(see later); rotation at 1500 rpm, using a synchronousspeed motor4 (150 hp), was adopted in the final experi-ments. The silver/silver chloride anodes were madefrom discs of silver foil (assay grade) 0.25 in diameterwhich were silver soldered to silver wire. This was thenchloridized in 0.1 N hydrochloric acid by passing acurrent of 2 ma for 30 min. The electrode was removedfrom the bath, and rubbed with fine emery paper andthen chloridized for a further 30 min. During this timethe silver became a purplish-brown color. The assemblyof this electrode is shown in figure 1. Under actualconditions of operation, the electrodes entered thefermentor liquid from the upper surface and wereinserted to a depth of about 2 in. The peak of the oxygenwave was found to be at -0.75 and at -0.90 v usingstandard calomel and silver/silver chloride referenceelectrodes, respectively, as reported previously byChain and Gualandi (1954).Oxygen absorption coefficients were determined by

the gassing-out method of Bartholomew et al. (1950)and Wise (1951), the polarograph having been cali-brated with salt solutions containing a known oxygenconcentration (Hixson and Gaden, 1950). Briefly,the salt solution in the fermentor (0.5 M KCl) wasgassed-out with oxygen-free nitrogen and the residualcurrent reading noted; then the agitation and aerationwere started at constant levels and galvanometerreadings were taken every 5 sec until saturation wasreached. The results of duplicate determinations werealmost identical when the experiments were conductedwithin a short period of time (see remarks on vari-ability given later).To check the operation of the electrode system,

the rate of oxygen uptake by a yeast suspension (wetbrewers' yeast, 300 g; Na2HPO4, 4.75 g; KH2PO4,18.22 g; glucose, 25 g; KCl, 37 g; final volume 2.5 L;pH 6.2; olive oil added as required to combat foaming),in the fermentor was compared with that in a Warburgapparatus. Using the fermentor, the values of KLa,C and CL were determined for the relationship.

Q..2KLa = (Q0-

azide. From these data the value of Qo2 was calculated.The Warburg experiments were conducted by themethod of Umbreit et al. (1949) with undiluted yeastsuspension (3 ml) removed from the fermentor. Otherexperiments with yeast showed that the oxygen supplyrate to the Warburg vessels was not limited by in-adequate shaking.

Scope of the work. With the sulfite system, oxygensolution rates were measured over a range of air flowrates from 0.5 to 4.5 L per min and agitator speedsfrom 385 to 1150 rpm. Because of differences in theresults obtained by the two methods, the polarographicwork was extended to cover a slightly wider range ofthese variables.

RESULTS

Sulfite experiments. Oxygen absorption-coefficientsobtained for all conditions of agitation and aeration

TABLE 1Absorption coefficients obtained by the sulfite method

for various conditions of agitation and aeration

KLa, hr-'

Air Flow, Agitator speed, rpm

385 600 800 925 1150

0.5 34 75 192 288 3601.0 32 85 252 432 4901.5 33 105 300 608 7352.5 39 131 376 638 9503.5 36 128 385 708 10324.5 44 140 386 726 1030

000

600

600

T-

-JY.

4001

200-

(1)

where KLa is the oxygen absorption coefficient, Qo,is the oxygen uptake rate in mmoles per L per hr, C

is the saturation concentration of dissolved oxygen3

and CL is the actual concentration of dissolved oxygen

in the fermentor liquid. Values of K,a were determinedfor the fermentor after poisoniing the yeast with sodium

4Metropolitan Vickers Electrical Company, Manchester,England.

1001_

OA 1.0 2.0

AIR FLOW RATE £G] L/MIN3.0 40 0

Figure 2. Effect of air flow rate on oxygen absorption bysodium sulfite solution. Agitator speed as parameter, logarith-mic plot; the value of X in the relationship KLa = K Gx is

given where G is air flow rate L per min.

AGITATOR SPEED RPM

S

.1 op

6OE0

0 a . . X. 0

52 [VOL. 7I on A

pril 22, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

AGITATION-AERATION IN SUBMIERGED FERAIENTATION. I

are given in table 1. Over the range of variables ex-

amined, the absorption coefficients varied from 32 to1032. At all air flow rates tested the absorption co-

efficients increased with increase in agitator speedwhereas at constant agitator speed the absorption co-

AIR RATE L/MIN 4.5

4000

So -

600 -

Y=3.2

400- 05

5

100

800

60-

40-

400 600 K=AGITATOR SPEED [N] RPM

Figure 3. Effect of agitator speed on oxygen absorption bysodium sulfite solution. Air flow as parameter, logarithmicplot; the value of Y in the relatonship KLa = K N" is givenwhere N is agitator speed, rpm.

TABLE 2

Effect of impeller speed (liquid turbulence) on thegalvanometer deflection using various electrode

arrangements

Galvanometerdeflection, mm

Electrode System Impellerspeed, rpm in deflec-500 1500 tion, %

Axial platinum cathode rotated at 1100rpm (1.5 ft per sec), saturated cal-omel anode; applied potential,-0.75 v .... ... ..... ...... 87 111 25.7

Radial platinum cathode rotated at1500 rpm (9.1 ft per sec), saturatedcalomel anode; applied potential,-0.75 v .......... 92 96 4.3

Radial platinum cathode rotated at1500 rpm (9.1 ft per sec), Ag/AgClanode; applied potential, -0.90 v 87 89 2.3

efficients increased with increase in air flow rate upto the point at which the impeller became "loaded,"that is, the rotational speed of the impeller was in-sufficient to completely disperse the air supplied. Atthe loading point a change was observed visually inthe character of the air dispersion. In addition to smallbubbles circulating with the bulk flow-pattern, acomplement of large bubbles was produced and theseescaped around the impeller and passed through theliquid at a relatively high velocity. This effect wasmore pronounced as the air flow rate was furtherincreased beyond the loading point. At the lowestagitator speed (385 rpm) tested, the magnitude ofincrease of absorption rate over the range of air flowtested was slight, thus indicating that loading occurredunder all conditions of operation. This was also sub-stantiated by visual observations of bubble dispersion.When absorption coefficients were plotted against

air flow rate (G), on logarithmic scales, a linear re-lationship was noted below the loading point for eachimpeller speed (figure 2). However, the value of theexponent on G was greater with higher agitator speeds.The correlation between absorption coefficient andagitator speed (N) chaniged slightly as the air flowrate was varied (figure 3) but averaged about 3.0.

Polarographic Experiments

Performance of the electrode systems. The performanceof the electrode systems was found to be affected bythe turbulence of the liquid in the fermentor (table 2).When the tip velocity of the cathode was increasedfrom 1.5 to 9.1 feet per second, the magnitude of thestirring effect brought about by turbulence of the fer-mentor liquid was decreased considerably. Further,the silver/silver chloride reference electrode appeared tobe more stable than the calomel electrode. Con-sequently, all subsequent experiments were made withthe "radial" platinum cathode (tip velocity 9.1 feetper second) and silver/silver chloride reference elec-trode.

Variability of absorption coefficients. It was observedthat the maximum galvanometer deflection, cor-responding to an oxygen-saturated solution, tended todecrease over a period of hr. This did not seem to affectthe calculated values of the absorption coefficientsdetermined under similar conditions of agitation andaeration, the largest difference being about 2 per cent.

TABLE 3Oxygen uptake by a yeast suispension in the fermentor

Agitator Air Flow C CL K La, hr-' Qo 2Speed Rate

rpm L/min mtmttoles/L/hr605 1.3 0.155 64 9.9635 0.7 0.143 86 12.3

1959] la"3

on April 22, 2020 by guest

http://aem.asm

.org/D

ownloaded from

R. STEEL AND M. R. BRIERLEY[

< 200

100

80

60

40

AAIR FLOW RATE (GI L/MINFigure 4. Effect of air flow rate on oxygen absorption by

potassium chloride solution. Agitator speed as parameter,logarithmic plot; the value of X in the relationship KLa =K Gx is given where G is air flow rate, L per min.

8000

600

400-

Il

. 200

100

80

60

400 600 800 1000AGITATOR SPEED tN) RPM

2000

Figure 5. Effect of agitator speed on oxygen absorption bypotassium chloride solution. Air flow as parameter, logarithmicplot; the value of Y in the relationship KLa = K N' is givenwhere N is agitator speed, rpm.

Further, the silver/silver chloride electrode decom-posed with extended use which necessitated a replace-ment every 3 to 4 weeks. For this reason, absorptioncoefficients were determined under standard operatingconditions (air flow rate, 2.5 L per min; agitator speed,800 rpm) at the start of daily operations. The absorp-tion coefficients determined for 30 runs yielded a meanvalue of 305 with a coefficient of variation of 5.6 percent. It should be noted that this variability arose fromthree main sources, (a) the use of different "radial"platinum cathodes, (b) the use of different silver/silverchloride anodes, and (c) variations in the performanceof these electrodes with age.

Check on electrode operation. Determinations of oxygenuptake by a yeast suspension in the fermentor yieldedthe data as given in table 3. The average value of thetwo determinations of oxygen uptake was 11.1 mmolesper L per hr which compared with a value of 10.8obtained by the Warburg method.

Correlation of absorption coefficients with operatingvariables. The data obtained in this part of the workcovered air flow rates from 0.5 to 5.5 L per min andagitator speeds from 500 to 1500 rpm. The results aregiven in the form of appropriate plots. Figure 4 showsthe correlation of absorption coefficient with gas flowrate; between 600 and 1500 rpm the data follow therelationship, KLa = K G053. A logarithmic plot ofabsorption coefficient against agitator speed with airflow! rate as parameter (figure 5) followed the relation-ship KLa = K N' ° for agitator speeds above about 800rpm. Below 800 rpm the exponent on N varied withair flow rate from 4.7 at 0.5 L per min to 8.1 at 5.5L per miii. The marked change in the correlation atagitator speed 800 rpm shown in figure 5 probablyreflects the change from laminar to turbulent flowin the fermentor liquid.

DISCUSSIONFor the sulfite system, the correlations between

oxygen absorption coefficients and operating variables(KLa = K G0 40-0.70 and KLa = K N3 0) agree reason-ably wA-ell wvith those reported by previous investigators.Further, the correlation between absorption coefficientand air flow rate in the polarographic system is ingeneral agreement writh that found by the sulfitemethod. However, the correlation, KLa = K N' 0,obtained polarographically, does not seem to havebeen.reported previously. The different values obtainedfor the exponent on N in the two systems probablyreflects differences in the oxygen transfer mechanism.

It is noteworthy that the sulfite technique maygive a false impression of equipment performance,a fact which has been noted previously by Wise (1950),Schultz and Gaden (1956), and Carpani and Roxburgh(1958). In the present work, the results obtained bythe two methods agreed reasonably well at 800 rpm

AIR RATE,L,M4IIN

YI-lO"

[VOL. 754

on April 22, 2020 by guest

http://aem.asm

.org/D

ownloaded from

AGITATION-AERATION IN SUBMERGED FERAIENTATION.

at air flow rates between 0.5 and 3.5 L per min (figure6) whereas agreement outside this narrow range ofoperating conditions was rather poor.From a comparison of the results (compare figures

2 and 4 and figures 3 and 5, also arithmetic plots infigures 6 and 7), it may be noted that the polarographiccoefficients exceeded the sulfite values under conditionsof equipment operation where the impeller speed wasinsufficient to disperse completely all the air supplied,that is, "loading conditions." Under these conditions,bubbles escaped around the impeller, their velocityrelative to an element of fluid was high (as indicatedby "boiling" at the liquid surface), and flow aroundthese bubbles was probably turbulent. Examples ofthese operating conditions are as follows: in figure 6,at 600 rpm at all gas flow rates; at 800 rpm at airflow rates above 3 L per min; in figure 7, below about800 rpm at air flow 1.0 L per min and below about850 rpm at air flow 4.5 L per min. Since visual ob-servations indicated no marked differences in bubbledispersion between the two systems, it seemed thatthe explanation for this difference in results was relatedto the oxygen absorption mechanism.

It is noteworthy in this regard that an increase inbubble velocity relative to an element of fluid (that is,a change in flow conditions around bubbles, eitherfrom laminar to turbulent flow or increase in the degreeof turbulence) in the two systems may influence theliquid-film oxygen-transfer coefficient (KL) in a differ-ent manner in each case. Coppock and Mleiklejohn(1951) reported that an increase in bubble velocity inwater resulted in a corresponding increase in K,presumably because of the increase in turbulenceadjacent to the liquid film and a consequent reductionin the liquid-film resistance (thickniess). In contrast,the results of Schultz and Gaden (1956) showed that

1000[

800a

Ii

-Z

-i6001

400

200

0 1 2 3 4AIR FLOW RATE [ L/M IN

5 6

KL for sulfite solution decreased with increase in inter-facial turbulence, presumably because of a reductionin the "time average" oxygen concentration and itseffect on the oxidation of S03= to S04=. Accordingly,it would appear that the loading air flow rate in thesulfite system (see figure 6) is a function of the chem-istry of the reaction rather than of equipment per-formance.

Similarly, the higher sulfite coefficients obtainedat agitator speeds above 800 rpm (figure 6) may also beindicative of changes in boundary or film conditionis.As the agitator speed is increased beyond 800 rpm(figure 5) turbulent flow is produced in the liqjuid bulk,a greater bubble area is incorporated within the bulkflow-pattern and these bubbles tend to follow the pathof fluid flow. Consequently, there is less bubble areawith motion relative to an element of fluid. Underthese conditions the increase in absorption coefficientin the sulfite system measures the increase in surfacearea plus any increase in apparent KL brought aboutby the increased stirring rate and the correspondingincrease in the "time average" oxygen concentration.The increase in absorption coefficient in the polaro-graphic system measures the increase in transfer areaminus the effect brought about by the faster stirringspeed and the corresponding increase in liquid-filmresistance. Hence with increase in agitator speed thesulfite coefficients increase at a faster rate than thepolarographic values (figure 7). It should be noted,however, that this explanation may in fact be an over-simplification because the transfer area was not meas-ured in the two systems.

Differences between oxygen absorption by sulfitesolution and by oxygen-free water have been notedby other workers considering temperature effects.For example, Schultz and Gaden (1956) calculatedfrom the data of other workers that the activationenergy for the sulfite reaction was characteristic of

500 700 900 1100 1300 1500AGITATOR SPEED (N) RPM

Figutre 6. Effect of air flow rate on oxygen absorption;agitator speed as parameter, arithmetic plot.

Figulre 7. Effect of agitator speed on oxygen absorption; airflow rate as parameter, arithmetic plot.

AGITATOR SPEED,RPM

1959] 1 55

o SULPH ITE

on April 22, 2020 by guest

http://aem.asm

.org/D

ownloaded from

5R. STEEL AND Mi. R. BRIERLEY

control by a chemical reactioni, whereas Bartholomewet al. (1950) reported that oxygen absorption intowater was controlled by diffusioni through a liquidfilm.The present results indicate that absorption coeffi-

cients obtained by the sulfite method should not beused as a measure of the actual amount of oxygentransferred to a microbial culture. Nevertheless, theyare useful for assessing improvements in equipmentdesign as long as the impeller speed is sufficient todisperse the air supplied. For the equipment used inthe present work the highest air flow rate which couldbe used at any particular agitator speed was expressedby the relationship G = K N0O83 where G was the airflow rate at which the loaded condition occurred(results not given in the text).

ACKNOWALEDGMENTS

The authors are grateful to Professor T. K. Walkerfor his interest in this work. One of us (R. S.) acknowl-edges receipt of an Imperial Chemical Industries Re-search Fellowship, and the other (M. R. B.) thanksthe Manchester Education Committee for the grantof funds.

SUMMAIRY

Mleasurements were made of oxygen absorption ratesby catalyzed (Cu++) sodium sulfite solution and bypotassium chloride solution (polarographic method)using a fermentor of conventional design. Correlationsobtained between oxygen absorption-coefficients (KLa,hr-1) and operating variables were as follows: for thesulfite system, KLa = K G0 40-0.70 and KLa = KN2 5-3 2 where G and N are air flow rate and agitatorspeed, respectively; for the polarographic system,KLa = G K0 53 and KLa = K N' °. Absorption coeffi-cients obtained by the two methods were essentiallysimilar only over a very narrow range of operatingconditions, namely, at agitator speed of 800 rpm withair flow rates between 0.5 and 3.5 L per min. At lowagitator speeds (below 800 rpm) the sulfite coefficientswere lower than the polarographic values at all airflow rates tested whereas at high agitator speeds(above 800 rpm) the sulfite coefficients were higherthan the polarographic values. At constant agitatorspeed (for example 600 rpm) the "loading" air flowrate was lower for the sulfite system (2.0 L per min)

than for the polarographic system (4.5 L per min).The discrepancies between the results obtained by thetwo methods are explained in the terms of the differ-ences in oxygen absorption mechanism in the twosystems.

REFERENCES

BARTHOLOMEW, W. H., KAROW, E. O., SFAT, M. R., ANDWILHELM, R. H. 1950 Mass transfer of oxygen in sub-merged fermentation of Streptomyces griseus. Ind. Eng.Chem., 42, 1801-1809.

BOWERS, R. H. 1955 The mechanics of bubble formation.J. Appl. Chem., 5, 542-548.

CARPANI, R. E. AND ROXBURGH, J. M. 1958 Studies onfermentation aeration. I. The oxygen transfer coef-ficient. Can. J. Chem. Eng., 36, 73-77.

CHAIN, E. B. AND GUALANDI, G. 1954 Aeration studies.Rend. ist. super. sanita (English ed.), 17, 5-60.

COOPER, C. M., FERNSTROM, G. A., AND MILLER, S. A. 1944Performance of agitated gas-liquid contactors. Ind.Eng. Chem., 36, 504-509.

COPPOCK, P. 0. AND MEIKLEJOHN, G. T. 1951 The behaviourof gas bubbles in relation to mass transfer. Trans. Inst.Chem. Engrs., 29, 75-86.

ELSWORTH, R., WILLIAMIS, V., AND HARRIS-SMITH, R. 1957 Asy-stematic assessment of dissolved oxygen supply in a 20litre culture vessel. J. Appl. Chem., 7, 261-268.

FINN, R. K. 1954 Agitation-aeration in the laboratory andin industrv. Bacteriol. Revs., 18, 254-274.

HIXSON, A. W. AND GADEN, E. L., JR. 1950 Oxygen transferin submerged fermentation. Ind. Eng. Chem., 42, 1792-1801.

KAROW, E. O., BARTHOLOMEW, W. H., AND SFAT, M. R. 1953Oxygen transfer and agitation in submerged fermentations.J. Agr. Food Chem., 1, 302-306.

1IAXON, W. D. AND JOHNSON, M. J. 1953 Aeration studies orDpropagation of bakers yeast. Ind. Eng. Chem., 45, 2554-2560.

PIRT, S. J. 1957 The oxygen requirement of growing culturesof an Aerobacter species, determined by means of thecontinuous culture technique. J. Gen. Microbiol., 16,59-75.

ROXBI-RGH, J. iI., SPENCER, J. F. T., AND SALLANS, H. R.1954 Factors affecting the production of ustilagic acid byUstilago zeae. J. Agr. Food Chem., 2, 1121-1128.

SCHULTZ, J. S. AND GADEN, E. L. 1956 Sulfite oxidation as ameasure of aeration effectiveness. Ind. Eng. Chem.,48, 2209-2212.

UMBREIT, W. W., BURRIS, R. H., AND STAUFFER, J. F. 1949M1anornetric methods and tissue metabolism. BurgessPublishing Co., Minneapolis, Minnesota.

WISE, W. S. 1950 The aeration of culture media: a com-parison of the sulphite and polarographic methods. J.Soc. Chem. Ind. (London), Suppl. No. 1, 69, 40-41.

WISE, W. S. 1951 The measurement of aeration of culturemedia. J. Gen. MIicrobiol., 5, 167-177.

56 [VOL. 7

on April 22, 2020 by guest

http://aem.asm

.org/D

ownloaded from