\
LACTIC ACID RECOVERY BY TANGENTIAL FILTRATION
1
2
Júlio Carvalho , José Roseiro , José Santos
1
1
Instituto Superior Técnico, Lisbon, Portugal and Instituto Nacional de Engenharia, Tecnologia e Inovação, I.P.,
2
Lisbon, Portugal
ABSTRACT
Ultrafiltration of lactic acid fermentation broths was studied using a cross f low m em brane system . T he m olecular weight cut -offs (MW CO) of the tested
hollow f iber (polys ulfone) m em branes were 10 k Da and 100 kDa. T he
experiments were conducted at three levels of transmembran e pressure (42,
105 and 149 kPa), and four tangential velocities (3.9, 15.9, 34.4 and 46.4
-1
cm ). The permeate flow-rate decreased with time due concentration
polarization and to membrane fouling. Higher transmembrane pressure and
tangential velocity promoted higher permeate flux. Transmembrane pressure
and tangential velocity had no significant effect on lactic acid recovery and
protein retention.
The model of resistance in series was applied to determine the intrinsic
12
-1
resistance
of
the
membrane
(6.2x10 m ),
the
resistance
due
to
concentration polarization and to fouling.
D ynam ic m odels where applied to describe the decrease of the perm eate
flow over operation time. The models that best fit the experimental results
were found to be: complete blocking, int ermediate block, block pattern
models and Mehta and Ho Zidney models .
Keywords: Lactic Acid; Recycled Paper Sludge; Ultrafiltration; Fouling; Concentration Polarization;
Dynamic Models.
1. Introduction
Organic
acids
constitute
an
important group of additives having
extensive uses in the f ood industr y.
They function as antioxidants and
stabilize the pH, thus preserving
the
organolectic
properties
of
1
foods . In particular, lactic acid is
2
widely applied in foods , chemical
stocks,
medicines
and
environmentally friendly packaging
3
materials .
Lactic acid can be manufactured by
chem ical synthesis or carbohydrate
fermentation.
The chem ical syn thesis m ethod
produces a racemic mixture of
lactic acid, whereas fermentative
synthesis originates optically pure
lactic
acid.
However,
the
biotechnological
production
of
lactic
acid
using
lactic
acid
bacteria is limited by the high cost
of the fermentation medium, mainly
associated
to
the
C-source.
W astewater treatment facilities of
recycling
paper
mills
generate
thousands of tons of concentrated
sludge per year. This waste is
usually disposed on landfills, an
expensive
and
environmentally
prejudicial end solution, and thus it
raises a serious disposal problem
requiring urgent solution. Since
this recycling paper sludge has a
high content of polysaccharides
(mainly cellulose), it constitutes an
alternative interesting substrate,
with negative cost, for lactic acid
fermentation.
Its
conversion
previously
requires
the
polysaccharides on sludge to be
broken down into the constitutive
monomers.
During the fermentation process,
lactic
acid
inhibits
the
microorganisms growth. This effect
can be avoided by continuously
removing it from the fermentation
broth, using membrane separation
processes such as ultrafiltration.
The filtration unit operation is one
of
the
oldest
for
solid/fluid
separation. It is used in several
wa ys, ranging from the use of
1
porous media not consolidated, in
the form of powder or granules, to
consolidated porous media, such
as filter paper or filter screens.
These consolidated media filters
can
be
called
membranes.
A membrane can be defined as a
selective barrier, solid or liquid,
which separates two phases and
restricts the transport of one or
more specific chemical species.
This selective process can occur
either by diffusion or by convection,
and it is driven by a gradient of
chemical
potential
(pressure,
temperature and concentration) or
electric potential.
All these membrane operations
work
mainly
due
to
the
transmembrane
pressure
(TMP),
the pressure being higher at one
side (retentate side) than the other
side (permeate side). Due to this
TMP, the particles that are smaller
than the membrane pore size are
pushed through the pores and are
recovered in the permeate flow.
The flow through the membrane
can be dead-end flow or cross-flow,
the
latter
is
also
known
as
tangential flow. Dead-end is less
efficient and more expensive. The
t a n g e n t i a l f l o w i s m o r e a t t r a c tive as it
allows less deposition of particles
on the membrane surface and
presents the additional benefit of
sweeping any particles that are
adsorbed onto the surface.
The changes in the performance of
the membrane can be caused by
three
different phenomena: the
deterioration of the membrane, the
concentration
polarization
and
fouling. The latter phenomenon
includes adsorption processes, the
gel layer f orm ation, the pores
blocking and deposits of solid
particles on the surface of the
membrane.
2. Materials and methods
2.1. Feedstock
The present study used pressed
recycled
paper
sludge
(RPS)
consisting of the solids resulting
from
the
wastewater
treatment
f acilit y of a local paper rec ycling
mill
(Renova,
Torres
Novas,
Portugal).
The
composition
of
neutralised RPS was determined to
be (in mass percentage of the
oven-dried sludge): 29.3% ash,
3.5% fat, 4.8% protein, 20.4%
lignin, 34.1% cellulose and 7.9%
xylan.
2.2. Enzymatic hydrolysis
RPS h ydrolysate was obtained by
enzym atic saccharification with a
m ixture of two comm ercial enzym e
preparations
(cellulolytic
and
xylanolytic,
from
Novozym es,
®
Denmark):
Celluclast
1.5L,
applied on a dosage of 10 U
-1
(FPase)
g
carboh ydrate;
and
®
-1
Novozym
188,
0.4
mL
g
carbohydrate on sludge. A sam ple
of RPS, after neutralisation with
h ydrochloric acid (to reduce the
carbonate content) was suspended
in 0.1 M sodium acetate buffer pH
5.5, for an initial consistency of
7.5% (w/v), expressed in terms of
total polysaccharides m ass, and it
was
sterilised
by
autoclaving
(121ºC, 2 atm, for 15 min). This
sludge suspension was incubated
with
the
enzyme
solution
(previously sterilised by filtration
through a 0.22 m membrane filter),
with an orbital shaking of 150 rpm,
at
35ºC
for
5
da ys.
Aseptic
conditions
were
assured
throughout.
The
sludge
hydrolysate,
after
residual solid removal by filtration,
was analysed, regarding the sugars
composition, by HPLC and filter sterilised
for
use
as
culture
medium for lactic acid production.
2.3. Lactic acid production
2.3.1. Microorganism and culture medium
Lactobacillus
rhamnosus
ATCC
7469, obtained from the American
T ype Culture Collection, was used
in this work. The microbia l strain
was grown in RPS h ydrolysate
supplemented
with
all
the
components (except glucose) of
Man Rogosa and Sharpe (MRS)
broth and calcium carbonate (at a
30 g/L concentration, for buffering
effect).
2.3.2. Culture conditions
Cultivations were carried o ut for 6
2
da ys in Erlenmeyer f lasks, at 37ºC
with orbital shaking (150 rpm). The
obtained
broth
was
used
in
filtration experiments for lactic
acid recovery.
2.4. Filtration studies for lactic acid recovery
where
(L/(m
J
2
2.4.2. Performance equations
Lactic acid recovery was expressed
as follows:
R (%)
V p .C p
V0 .C 0
100
(1)
For
the
calculation
of
protein
rejection (%), Eq. 2 was used:
CC
CC
Cp
100
(2)
Membrane flux is a measure of the
permeate flow rate taking into
account the active surface area of
the membrane and it is calculated
using Eq. 3:
1
A
dV
dt
flux
2
The transmembrane press ure (TMP)
is calculated using Eq. 4:
TPM
P0
PS
Pp
2
(4)
where Po and Ps (kPa) is the inlet
and
outlet
pressure
of
the
retentate,
respectively;
and
Pp
(kPa) is the permeate pressure,
which
corresponds,
in
this
experimental
set,
to
the
atmospheric
pressure
(corresponding
to
a
relative
pressure of zero).
Using Darcy´s law as a theoretical
basis, the flux decline can be
simply determined by:
J
TMP
.R t
(5)
where
TMP
(kPa)
is
the
transmembrane pressure drop; µ is
the viscosity; and R t is the total
resistance. The total resistance in
series can be obtained by summing
all of the resistances caused by an
individual fouling mechanism:
where
Rm
is
the
(6)
m embrane
resistance; Rrev is the reversible
resistance
(caused
by
a
concentration polarization and the
reversible fouling); and Rirrev is
the
irreversible
resistance
(referring
to
the
irreversible
fouling phenomenon).
2.5. Analytical methods
where Cp and Cc (g/L) are the
protein
concentrations
in
the
permeate
and
concentrate,
respectively.
J
permeate
Rt = Rm+Rrev+Rirrev
where Cp and Co are the lactic
concentrations (g/L) in permeate
and f eed, respectively; V p and Vo
(L) are the permeate and initial
volumes, respectively.
Rej (%)
the
A is the membrane area
(m ); V is the filtrate volume (L);
and t is the unit time (h).
2.4.1. Membrane filtration system
The
ultrafiltration
membrane
system consisted of a recirculation
pump, a cross-flow ultrafiltration
module
H 1 P 1 0 - 4 3 ( Amicon)
and
thermostatised bath 35 ºC. The
concentrate was recycled to the
feed tank while permeate was
collected into a feed tank placed
on an analytical b alance.
Transmembrane
pressure
and
tangential velocity were adjusted
by a manual valve and pump
controller.
Polysulf one
asym m etric
membranes, with 10 and 100 kDa
o f MW CO, we re u s e d in t he se
experiments. The surface area of
2
the membrane was 0.03 m .
is
·h));
(3)
Filter paper activity (FPase), which
describes the overall cellulol ytic
activity,
was
determined
using
W hatman number 1 filter paper as
substrate, at the conditions used
f or the enzym atic h ydrolysis of
RPS.
Enzym e
activity
was
expressed in international units (U),
as the am ount of enzym e required
to release 1
mol of glucose
reducing equivalent
per minute
under
the
assay
conditions.
Reducing sugars were estimated by
3
3. Results and discussion
3.1. Characterization of the fermentation broth
The chemical composition of the
culture broth was determined to be,
in average: lactic acid, 37.3 g/L;
cellobiose, 1.6 g/L; xylose, 4.3 g/L;
acetic acid, 20.6 g/L; and succinic
acid
and
ethanol,
vestigial
concentrations. This composition
demonstrates
that
glucose
was
totally metabolised.
3.2. Membrane permeability
As
expected,
the
results
demonstrated that the 100 kDa
membrane
presented
higher
-6
permeability (Lp=1.49x10
m/s·kPa)
12
-1
and resistance (Rm= 75.01x10
m )
than
the
10
kDa
membrane
-6
(Lp=0.18x10
m/s·kPa
and
12
-1
Rm=6.20x10
m ).
3.3. Membrane selection
The
performance
of
polysulfone
membranes,
molecular weight exclusion
two
with
limits
of 10 and 100 kDa, was evaluated
using
the
permeate
flow,
the
recovery and the selectivity as
criteria.
The
membranes
were
tested under the same conditions
(transmembrane pressure of 105
kPa, tangential velocity of 15.9
cm/s and temperature of 35ºC).
It was verified (Figure 1) that both
membranes
exhibited
similar
average fluxes.
20
10 kDa
J (L/hm2)
16
100 kDa
12
8
4
0
1
2
3
4
5
CF
Figure 1 – Permeate flux versus
concentration factor (vs=15.9 cm/s
and P0=180 kPa).
Taking
into
account
the
experimental results, the 10 kDa
membrane was selected for further
studies because it exhibited higher
rejection of total soluble protein.
3.4. Operation parameters
3.4.1. Tangential velocity
From Figure 2, it is observed that
the permeate flux increases with
increasing tangential velocity. For
low tangential velocity (3.9 and
15.9 cm/s), the increase in shear
stress promotes an increase in the
permeate flux, due to the reduction
on the concentration polarization
and consequent decrease in the
thickness of the gel la yer.
10
Average Flux (L/hm2)
the
dinitrosalycilic
acid
(DNS)
4
method .
Sugars (glucose, xylose, cellobiose
and xylobiose), alcohols (ethanol
and isopropanol) and organic acids
(lactic,
acetic,
propionic
and
succinic acids) were quantified by
high-performance
liquid
chromatography (HPLC) using a
W aters LC1 module 1 plus (Millford,
LA,
USA)
instrumentation.
An
Aminex HPX-87H column (Bio-Rad
Laboratories, Hercules, CA, USA)
was used, operating at 50ºC with
0.005 M H2SO4 as mobile phase at
a flow rate of 0.4 mL/min.
The
dry
matter
content
was
determ ined by oven drying (at
105°C till constant weight, for
about 16 hours) the solid sediment
obtained by centrifugation (13000
rpm, 4ºC, 5 minutes).
The
total
soluble
protein
was
determined
by
the
method
5
described by Bradford .
The average size of particles and
their distribution was determined
by
"D ynam ic
Laser
Light
Scattering" using a Zeta sizer 1000
HSA (Malvern Instruments, UK)
equipment.
8
6
4
2
0
0
20
40
60
Surface velocity (cm/s)
Figure 2 function
of
Flux behaviour as
surface
velocity
4
(TMP=105 kPa, 10 kDa and T=35
ºC).
Table 1 – Results obtained for the permeate
flux for different surface velocities (initial lactic
acid concentration = 41.1±2.7 g/L; initial
protein concentration = 0.60±0.17 g/L).
Surface
velocity
(cm/s)
3.9
15.9
34.4
46.4
Shear Lactic acid Protein
stress recovery Rejection
(Pa)
(%)
(%)
0.3
1.0
2.2
3.0
71.2
70.8
71.5
71.5
96.9
93.7
94.0
97.3
3.4.2. Transmembrane pressure
In Figure 3, it is observed that the
average
permeate
flux
linearly
varies up to 100 kPa. Afterwards,
it tends to an asym ptotic region of
mass transfer control, in which the
increase on pressure does not
cause a significant increase in the
flux.
<J> (L/hm2)
For high tangential velocity, shear
stress promotes a reduction of the
boundary
la yer,
not
only
by
reducing polarization, but also by
removing material from membrane
surface. The larger particles are
first withdrawn, leaving sma ller
and more compact particles on the
membrane, and thus with lower
m em brane perm eabilit y. T his effect
results on a reduction in the
permeate flux, due to the fact that
the
membrane
becomes
less
permeable.
From Table 1, it is observed that
the recovery values obtained for
lactic
acid
are
very
similar,
demonstrating that recovery is no
influenced
by
the
tangential
velocity. T he protein rejection also
did not vary singnificantly with this
operation parameter.
8
6
4
2
0
0
50
100
150
200
Transmembrane pressure (kPa)
Figures 3 – Behaviour of average
permeate
flux
versus
transmembrane pressure (v s=15.9
cm/s).
For low pressure values, the initial
flux
causes
a
decrease
in
macromolecules
adsorption
and,
consequently, there will probably
be
a
reduction
of
sub -layer
formation in the membrane surface,
thus
reducing
concentration
polarization and fouling, increasing
the permeate flux.
From Table 2, it can be concluded
that transmembrane pressure does
not influence lactic acid recovery
and protein rejection.
Table 2 – Results obtained for transmembrane
pressure versus permeate flux. (initial lactic
acid concentration = 39.1±0.20 g/L; initial
protein concentration = 0.70±0.18 g/L).
TMP
(kPa)
Lactic acid
recovery
(%)
42
105
149
72.4
70.8
72.4
Protein
rejection
(%)
95.7
93.7
96.5
The
plot
of
resistance
versus
transmembrane pressure, in Figure
4, demonstrates that irreversible
and
reversible
resistance
increases
with
increasing
transmembrane
pressure.
These
results were already expectable:
for higher pressure, the viscosity
near the membrane surface is
higher and thus mass transfer
resistance is also higher, with the
corresponding
increase
in
the
reversible resistance.
5
5E+13
4E+13
3E+13
2E+13
Rm
Rrev
Rirrev
1E+13
0
25
50
75
100
125
150
components
present
in
the
fermentation
broth.
This
effect
might
result
from
the
protein
adsorption
on
the
membrane
surface and from possible blocking
inside their pores.
3.6. Dynamic Models Application
All the models used indicate that
the initial water flow increases with
the transmembrane pressure and
the surf ace velocity.
175
Transmembrane pressure (kPa)
Figure 4 – Plot of resistance
versus transmembrane pressure.
High
pressures
increase
irreversible adsorption of colloids
and
other
particles
and
the
particles
inclusion
inside
the
membrane
pores,
causing
the
obstruction,
which
justifies
the
increase in irreversible resistance.
14
Supernatant
12
Suspended
solids
10
J (L/hm2)
Resistance (m -1)
7E+13
6E+13
8
6
4
2
0
1
3.5. Influence of the broth components in the
behaviour of tangential filtration
Two different experiments were
carried
out:
one
with
the
supernatant and the other with the
solids (bacterial cells, possible
cell fragments and other particles
in
suspension)
obtained
by
centrifugation. The solids were
suspended in 0.1 M sodium acetate
buffer. The initial composition of
the fermentation broth is presented
in Table 3.
Table 3 – Initial concentrations of processed
supernatant and suspended solids.
Concentration (g/L)
Biomass Lactic Protein
acid
39.24
0.26
Supernatant
Suspended
solids
32.15
0.97
0.09
In Figure 5, it can be observed that
the average permeate flux obtained
for the suspended solids filtration
2
(9.22 L/h·m ) is much higher than
the
one
obtained
for
the
2
3
CF
4
5
Figure 5 – Permeate flux versus
Concentration Factor (vs=15.9 cm/s,
T=35 ºC and TMP=105 kPa).
The
complete
blocking
and
intermediate
blocking
models
predict that the filtration constant
increases with increasing surface
velocity,
resulting
from
the
increase
in
transmembrane
pressure. This fact would not be
expected, since the increase in
surface
velocity
decreases
the
concentration
polarization
and
fouling of the membrane.
The
blocking
standard
model
indicates
that
the
filtration
constant
increases
with
transmembrane
pressure
and
surface
velocity.
This
increase
leads to reduction of fouling and
concentration polarization, which
in turn causes a decrease in the
volume
of
mass
fraction
of
particles deposited per unit volume
of permeate.
Mehta and Ho Zidney models well
adjust to the experimental data.
2
supernatant filtration (6.37 L/h·m ).
Theses results seem to indicate
that the flux passage through the
membrane is limited by the soluble
4. Conclusion
Of the two membranes tested, the
10 kDa ultrafiltration membrane
6
was selected since it demonstrated
higher selectivity.
In all the tests performed, there
was a significant decrease in the
permeate flow throughout the time
(rapid phenomenon that occurs in
the starting period, with a sharp
decrease in flow), followed by a
slow decline that has continued
throughout the period of operation.
It
follows
that
the
tangential
velocity does not influence the
recovery of lactic acid neither the
rejection of protein, and this means
that high permeate flows do not
necessarily imply higher values of
recover y.
The
reversible
and
irreversible
resistance
increase
with
the
tangential velocity.
W hen applying the m odels that
best fit the experimental results
(complete
blocking,
standard
blocking,
intermediate
blocking,
and Mehta and Ho Zidney models),
the values predicted for the initial
permeate flow (Jvo) are lower than
the
results
experimentally
measured. This observation results
from the fact that in the initial
period the experimental setup fails
in the measurements for the high
f low due to the high system hold
up.
Principle
of
Protein -dye
Binding.
Analytical Biochemistry. 72: 248-54.
5. References
1.
Frederick,
J.F.
(2000)
Encyclopedia of Food Science and
Technology. John W iley & Sons, New
York.
2. Boyaval, P., Corre, C., Terre, S.
(1987)
Continuous
Lactic
Acid
Fermentation
with
Concentrated
Product Recovery by Ultrafiltration
and
Electrodialysis.
Biotechnology
Letters. 9: 207-12.
3. Hitomi, O., Hiyama, K., Yoshida, T.
(1992) Kinetics of Growth and Lactic
Acid Production in Continuous and
Batch Culture. Applied Microbiology
and Biotechnology. 37: 544-8.
4.
Miller,
G.L.
(1959)
Use
of
Dinitrosalicylic
Acid
Reagent
for
Determination
of
Reducing
Sugar.
Analytical Chemistry. 31:426-8.
5. Bradford, M.M. (1976) A Rapid and
Sensitive
Method
for
the
Quantification
of
Microgram
Quantities of Protein Utilizing the
7