Professional Documents
Culture Documents
Introduction
Goal is to improve sparging system
Efficiency
O2 delivery vs. CO2 stripping
Shear effects
Foam generation
Agitation
Relative gas entrance distribution
Liquid column height
Gas partial pressure concentration
Gas flow rate
Pore material & design
What is gained?
Better cell culture environment
Lower operating cost
2
What is kLa?
Comes from simplified gas liquid film theory equation:
NL = kLa(CLi CL )
kL is the transfer coefficient
a is the area of flux
CLi is the interface concentration
CL is the liquid bulk concentration
Liquid
Driving factors
Differential partial pressure
Surface area available for transfer vs.
volume
Gas bubble liquid film thickness
O2
CO2
100
% Gas
Saturation 80
60
40
20
0
81
83
85
87
89
Minutes
Figure 3: CO2 Stripping Data Vertically Mirrored On O2
Solubility
Operating partial pressure
Note: CO2 sensor limits exceeded causing signal saturation near 20%.
Table 1: Driving Partial Pressure Delta Available to Strip CO2 and Add O2 in a Typical Animal Cell Culture Bioreactor (in atm
assuming 1 atm ambient pressure) and Corresponding Liters of Dissolved Gas in DI H2O at 37C2,3
Reactor dissolved
O2 set point
30% air saturation
50% air saturation
91
sparging with O2
CO2
O2
0.06
0.147
0.937
47.4
3.3
0.06
0.11
0.89
47.4
5.6
Why do we care?
Dissolved CO2 Carbonic acid
Lowers vessel pH
Base is added to regulate pH
Vessel osmolality increased
Variable
Membrane/mesh (polymer, metal)
Can generate small or large bubbles
depending on mesh density and material
Macro sparger
Open pipe (polymer, metal)
Sintered
<1
wide bell
low
high
low
Membrane
< 1 to > 3
narrow to wide bell
low
low to high
low to high
Open Pipe
>2
wide bell
low
low
mid to high
Drilled Hole
>1
uniform to wide bell
low to high
low to mid
mid to high
Time
Bulk solution changes during culture operation
Dual spargers allow adaptive load balancing between micro and macro
for best CO2 stripping verses O2 kLa
7
Small bubble
High surface/vol. + high residence time = high
efficiency (generally high kLa)
PROs
CONs
Large bubble
Plateau
Region
0.6
12
0.5
10
0.4
0.3
0.2
0.1
0
0
Normalized Distribution
Reduced effectiveness
Skewed CO2 stripping / O2 delivery ratio
kLa performance after volume and kLa
skewing applied
Original bubble size standard deviation
Volume skew (4/3*pi*d/2^3)
kLa Skew 7.954*d^2.48 + diminishing
returns near 0.078 mm Diameter
Figure 6: kLa and Volume Shifts of Bubble Diameter Bell Curve at 250L With Target Ave. Bubble Dia. of 2.4mm
Note: Area under the green curve is reduced over 20%; additionally skewing based on actual data, bell curve having normal distribution is
an assumption purely for demonstrative purpose
Impeller shear
Sparge systems dependent on high agitation may require damaging impeller
speeds
10
Native polyethylene polymers are low energy; favors air (contact angle > 90 )8
Pores will hold onto a bubble more aggressively
Same size pores in polyethylene spargers generate larger bubbles than in metal
Pre-distribution
Stainless vessels = wide separation between pores
SUS vessels = sparge mechanism in small area
Results in loss of efficiency
Matching sparger pore size and quantity from stainless steel to single-use may
not give equivalent results
11
Initial Investigation
Single pore analysis using existing film disc tooling
Laser-generated pores from 0.02 to 6mm diameter
Study behavior with precision mass flow controller (MFC) and high speed
camera
Determine relative bubble size generated by a given pore size
12
Macro sparger
Initial testing with two boundaries
Mid laser capability 0.175mm pore diameter
Pore saturation estimate (0.1vvm, 500 pores)
Pulse
modulated
Pore
saturation
Bubble size
increase
0.05
0.1
Chaos
begins
16
12
8
4
0
0
0.15
DHS AIR VVM
0.2
0.25
Figure 7: 250L Drilled Hole Sparger 500 Pores 0.203mm dia. O2 kLa, 139rpm, 6.67cm dia. Disk
13
0.3
VVM
0.005
0.005
0.005
O2 Delivery
kLa
kLa/cm
9.91
0.190
9.56
0.105
9.77
0.055
CO2 Stripping
kLa
kLa/cm
1
0.019
.96
0.011
.89
0.005
%CO2/O2
kLa
10.1%
10.0%
9.1%
Note: Frit slight CO2 stripping reduction with larger vessels likely attributed to buildup in headspace and hold-up gasses
VVM
0.0125
0.0125
0.0125
O2 Delivery
kLa
kLa/cm
2.05
0.039
3.54
0.039
5.71
0.040
CO2 Stripping
kLa
kLa/cm
1.06
0.020
1.59
0.017
2.08
0.015
%CO2/O2
kLa
51.4%
44.8%
36.5%
0.838mm pore size shows bubbles not approaching equilibration (scale with column ht.)
Table 5: kLa Scaling With Liquid Column Height Using 0.838mm Pore DHS: Pre-Saturation Flow Rates
Vessel Vol
(L)
1000
2000
VVM
0.05
0.05
O2 Delivery
kLa
kLa/cm
7.49
0.053
9.98
0.056
CO2 Stripping
kLa
kLa/cm
3.95
0.028
4.82
0.027
%CO2/O2
kLa
52.7%
48.3%
Note: 2000L vessel gets slight efficiency bump from using 2 discs instead of one; better pre-distribution
14
Refined Analysis
Pre-distribution study performed
Size discs as large as possible
pore saturation
20
kLa 1/hr
15
10
Pulse Modulated
Bubble Size
Chaos
5
0
0
0.05
0.1
0.15
0.2
VVM
Figure 9: 250L Drilled Hole Sparger 760 Pores 0.233mm Dia. kLa
15
0.25
0.3
16
17
18
19
20
Results Characterization
Reduced or eliminated base
consumption in test vessels
25
175
140
105
70
20
15
10
39
36
33
30
27
24
21
18
kLa Ratio
0.6
0.5
0.4
0.3
0.2
0.1
0
180
135
90
45
0 3 6
9 12 15
18 21 24
0
27 30 33
36 39
Micro Sparger sLPM
0.7
15
Consistent character
12
0
9
0
6
35
3
5
0
kLa 1/hrs
30
Promising performance
21
Open pipe
6.0
Open pipe
frit
5.0
25
4.0
20
3.0
15
2.0
10
1.0
0.0
0
0
0.01
0.02
0.03
0.04
0.05
0.01
VVM
0.02
VVM
0.03
0.04
22
0.05
kLa 1/hr 14
CO2
12
10
8
6
4
2
0
0
0.01
0.02
0.03
VVM
Key takeaways:
Linear correlation, ratio near 50%
DHS O2 delivery increase is not excessive
Open pipe has similar ratio but lacks effectiveness
23
0.04
0.05
Pauline M. Doran. Bioproces Engineering Principles. Elsevier Science & Technology Books, 1995,
ISBN:0122208552
(chapter 9 is of primary focus), linearizing equation used for data sets is:
Data used for calculations is from 20 to 80% of air saturation values
Delay applied between sparge up and down settings to reduce holdup-gas influence
O2 delivery kLa sparge down with pure N2, sparge up with air
CO2 stripping kLa sparge down with pure CO2, sparge up with air
Finesse TruFluor DO, Polestar DO (ppm class), CO2 sensors used to measure dissolved gas
concentration
Typical ambient atmospheric pressure of 865mbar, data normalized against lab pressure changes and
drift prior to kLa calculations
sensor
Test data clearly shows that CO2 sensor data, when mirrored, under the noted conditions, is identical to O 2
sensor data. Further CO2 sensing after confirming this fact has been carried out using O 2 sensors only.
2)
Estimates based on Henrys constants. NIST Chemistry Webbook (2005). Retrieved September 25, 2013 from
http://webbook.nist.gov/chemistry/
3)
Henrys constant values were adjusted by temperature dependence according to: Francis L. Smith and Allan H.
Harvey (September 2007). Avoid Common Pitfalls When Using Henry's Law. CEP (Chemical Engineering
Progress). ISSN 0360-7275 in conjunction with enthalpy of solution values from CRC Handbook of Chemistry and
Physics 87th Ed.
24
In this case the mass*g is buoyancy and directly proportional to volume, substituting volume of a sphere and area of a circle and simplifying constants
leads to:
Relative Terminal Velocity = sqrt [ Bubble_diameter / Drag_coeff ]
The drag coefficient is:
Drag_coeff = (2Drag_Force) / (Fluid_densityvelocity2Cross_section_area)
Drag force assuming approximate laminar flow is a constant multiplies by velocity and fluid density is assumed constant leading to:
Relative Terminal Velocity = sqrt [ Bubble_diameter^2Relative Terminal Velocity ]
Further simplification leads to:
Relative Terminal Velocity = Bubble_diameter^2
And so an increase in bubble diameter yields roughly a squared increase in averaged vertical terminal velocity.
25
In practice actual bubble size impact scaling is further modified by changes in partial pressure equilibrium during
bubble rise, deformation of assumed bubble sphere, differences in single pore testing measured bubble diameter
vs. actual bubble diameter influenced by vessel agitation mechanism, and possible other unknown factors.
The following figure shows actual kLa values (adjusted for column height ratio differences) and normalized with
respect to 250L vessel data. Estimated kLa performance ratios based from single pore bubble size data analysis
and equations predicting cubic scaling effect. We can see in practice that actual scaling was closer to an exponent
of ~2.5 though much of the skewing from predicted appears to occur in the smaller bubble size estimates. Further
analysis of bubble size behavior in actual vessel must be carried out to help determine principle source of
inaccuracy.
0.5
y = 13.876x-3
0
1.5
26
y = 8.387x-2.416
2.5
y = 7.954x-2.476
3.5
Ying Zhu et all. NS0 Cell Damage by High Gas Velocity Sparging in Protein-Free and Cholesterol-Free Cultures.
Biotechnology and Engineering, 2008, DOI 10.1002/bit.21950
7)
Sachiro Kakinoki et all. Surface Modification of SUS 316L Stainless Steel with Tartaric Acid Derivative-Crosslinked
Human Serum Albumin Matrices. The Open Biotechnology Journal, 2008, 2, 143-147
8)
27