164 Solid/Liquid Separation: Scale-up of Industrial Equipment

• Concentration and/or diafiltration of product or purification of high

molecular weight or solid contaminants;
• Establish molecular weights of product and contaminants;
• Fluid concentration and viscosity;
• Target final concentration factor;
• Number of diafiltration volumes required;
• Required processing time;
• Pre-filtration requirements;
• Temperature sensitivity;
• Target product yield.

Consideration of these points will enable the appropriate technology to

be selected.

4.3.1.6 Membrane selection

To concentrate in a molecule, the molecular weight must be deter-
mined and then a membrane selected that possesses a molecular weight
cut off (MWCO) that is 20% to 30% of that value. For example, to
concentrate a protein with a molecular weight of 100 kD, a membrane
rated at 30 kD would be appropriate.
Conversely, when the objective is to pass (transmit) the protein of
interest, then the MWCO should be 3-5 times greater than the
molecular weight of the protein.
Next the membrane format should be determined. For example, screen
cassettes should be chosen for filtered solutions to maximise the flux
through the device. Suspended screen or open channel cassettes, hollow
fibre or ceramics should be chosen when the process fluid contains
suspended solids or the viscosity is too high to run a screen cassette.
Cell harvest applications would usually require open channel or
suspended screen options.
Further key considerations to selecting the most appropriate format
will be the scale (process volumes) of the final process, compatibility
of the device and membranes with the process fluid and cleaning
and/or steam requirements.
Typically cassette technology is preferable for smaller scale processes,
where the target final volumes require low system hold-up volumes.
Open channel devices require higher crossflow rates per unit area, as
compared to cassettes, in order to generate the required shear rates at the
membrane surface. As a result, they tend to have higher hold-up volumes
 4 • Membrane filters - microfiltration and ultrafiltration 165

per unit area, and this aspect in combination with the higher flow
requirements can Umit their applications in small biotech processes.

43A. 7 System optimisation

The primary characteristics of TFF membrane systems are flux and
trans-membrane pressure (TMP). The flux is the volumetric permeate
flow rate per unit area of membrane and is often measured in litres/
mVh. It has become customary in the filtration industry to use the term
(or abbreviation) LMH for flux. Derivation of the Flux v^- TMP
relationship for the system is the key to optimisation and sizing.
As discussed above, TFF enables separation of solids where contaminants
exist at high concentrations. The technique also makes more complex
separation stages feasible. In such cases the TFF processes include a
number of steps that need to be performed. These can be broadly
classified into three distinct areas: Preconditioning, Processing and
Post conditioning. The flow schematic in Figure 4.23 below outlines
these three areas and the key steps within each.

1 Preconditioning |
1 Processing | 1 Post Use Conditioning |

M
1 Installation | Optimisation | 1 •{ Buffer Flush |
^ i i
1 Flushing | 1 Concentration | 1 CIP 1

1
i
Sanitisation 1 1
i
Diafiltration | 1
1
Flushing |

1
i
Flushing | 1
^
Product Recovery | 1
i
1 Clean Water Permeability |
i
1 Clean Water Permeability | 1
i
Storage |
i
1 Integrity Testing |
i
1 Buffer Conditioning | i

F i g u r e 4 . 2 3 Typical TFF process steps

It is important to note that the 'Optimisation' step is performed once at

the start of the process evaluation stage, to determine optimum
operating parameters. Generally, optimisation tests are run to deter-
mine the TMP that gives the best flux for any given process and device
combination (at a set crossflow rate). The logical process steps for the
derivation of the flux v^ TMP relationship is given in Figure 4.24. Also
included in the figure is the schematic equipment layout as a variant of
Figure 4.18 showing the return of the permeate to the feed tank. In this
way a constant contaminant loading is presented to the membrane
surface. The data collected would be tabulated as illustrated in Table 4.3.
166 Solid/Liquid Separation: Scale-up of Industrial Equipment

Optimisation
fpT^ /^^^^
1 Fill System (5 Mins) •J tssia
vV
1
Y
v y

i
i Jk 1 raJCgj

1 Measure Fp, Rp & Pp

t ^' ^r
1 Maintain System Vol
FEED
i
PERMEATE

1 Increase Rp by 0.25 bar ^ TANK

* (PT^ 1

1
i
Measure Fp, Rp & Pp PUMP A
(\ " V
1 ^
i
1 Repeat TMP Increase Step ^ Legend

i PR : Retentate Pressure
Pp : Permeate Pressure
1 Produce Flux vs TMP Curve Pp: Feed Pressure

F i g u r e 4 . 2 4 Process optimisation

Table 4 . 3 Typical optimisation values.

Feed Retentote Permeate TMP Retentote Permeote Permeote

pressure pressure pressure (bor) flow flow flux
PF. (bor) PR,(bar) Pp, (bor) (l/min) (ml/min) (Imh)

1.0 0 0 0.5 0.7 30 19.5

1.25 0.25 0 0.75 0.7 40 26
1.5 0.5 0 1 0.7 50 32.5
1.75 0.75 0 1.25 0.7 58 37.7
2.0 1 0 1.5 0.7 62 40.1
2.25 1.25 0 1.75 0.7 66 42.3

Other factors such as transmission of unwanted molecules and the

number of diafiltration volumes required to achieve the desired purity,
are also evaluated as necessary (see Section 4.3.1.9). It is also common
to take samples of the feed and permeate streams during optimisation,
and to analyse both to determine % transmission or % retention
(depending upon where the product is). The Table shows an example of
typical parameter values during the UF optimisation process (this data
relates to tests performed with a 0.1 m^ cassette).
During optimisation it is important to plot the relationship between flux
and trans-membrane pressure as the test develops, such that the optimum
TMP can be identified immediately from the graph. There is no gain in
any aspect of performance by allowing the curve to plateau. Filtrate flux
may decrease, percent transmission may decrease, operating costs will
increase (assuming any full scale system is subsequently operated at the
sub-optimal high TMP) and the membrane will be more difficult to clean.
 4 • Membrane filters - microfiltration and ultrafiltration 167

The graphical plots in Figure 4.25 illustrate the typical curve generated
during an actual optimisation run. After each setting plotting TMP
against flux as discussed will develop the type of relationship shown in
Figure 4.26.
At the point where the curve begins to plateau the optimum conditions
for maximum flux have been achieved at that given cross flow
velocity. If acceptable filtrate flux or transmission is not achieved, the
membrane can be cleaned and the optimisation process repeated using
a different cross flow velocity. Increasing cross flow will tend to give
increased filtrate flux at the same TMP.

4.3.1.8 Determination of the maximum gel concentration (CQ)

When the optimum conditions are established as determined above, the
system can be transferred into concentration mode by directing the
permeate line away from the feed vessel to a separate permeate tank
which may be graduated to facilitate volume measurement. This
configuration is shown in Figure 4.27 as a further minor modification
to the apparatus presented earlier.
At this stage the following data should be recorded:

• Time
• Pressures (Pp, PR and Pp)
• Filtrate flow rate
• Volumes (feed and permeate).

Readings should be taken at regular intervals throughout the trial up to

the end point of the test run, which is usually either the maximum
concentration factor required or achievable.
This data should be used to plot filtrate flux v^* log (volume con-
centration factor, VCF), Figure 4.28. The flux should decrease slowly
and linearly with increasing concentration factor. If it does not then it
may indicate that a component of the feed solution is changing (such as
precipitation of proteins) or that the crossflow rate is not sufficient
(causing excessive fouling of the membrane).
The curve can be extrapolated to the x-axis (flux = 0), the concentration
factor where the curve crosses the x-axis is the maximum concentration
achievable (CQ). CQ occurs when the bulk flow concentration is equal
to the gel layer concentration. In reality this concentration is
unachievable and the optimum point to end concentration and start
diafiltration if required is at a lower concentration.
1 6 8 Solid/Liquid Separation: Scale-up of Industrial Equipment

Flux Vs TMP
50.00 1.25
X
3 40.00 1.00
LL
(]")
30.00 0.75
CO -L 20.00 0.50
CD
^J
E 10.00 0.25
CD
Q_ 0.00 0.00
0.00 0.50 1.00 1.50
Transmembrane Pressure
-Process Flux -Differential Pressure

Flux Vs TMP
50.00 1.25
X
3 40.00 1.00
U-
CD 30.00 0.75
-•—• X 20.00 0.50
CD
0
_l
E
0)
10.00 0.25

Q- 0.00 0.00
00 0.50 1.00 1.50 2.00

Transmembrane Pressure
Process Flux -"—Differential Pressure!

Flux Vs TMP
1.25
X
_3 40.00 k—•—^ 1.00
LJL 30.00 0.75
0 ^-v
CD X^
-S 20.00 0.50
CD ^
E d. 10.00 0.25
CD
QL 0.00 0.00
0.00 0.50 1.00 1.50 2.00
Transmembrane Pressure
h-»- Process Flux - • — Differential Pressure

Flux Vs TMP

0.50 1.00 1.50 2.00

Transmembrane Pressure
J^ • Process Flux -Differential Pressure

Figure 4»25 Optimisation plots

 4 • Membrane filters - microfiltration and ultrafiltration 169

t MEMBRANE CONTROLLED
REGION
GEL LAYER CONTROLLED
REGION
< M H

(at constant CF)

OPTIMAL
TMP
RANGE

Trans-membrane Pressure
Figure 4«26 Classic overall optimisation plot

RETENTATE
csa*a-

FEED PERMEATq
TANK TANK

Legend
PUMP ' PR : Retentate Pressure
Pp: Permeate Pressure
PF : Feed Pressure

Figure 4«27 Concentration set-up

2 3 4 5 6 7 8 9 10
Concentration Factor

Figure 4.28 Classic plot of flux vs log (VCF)