Note: Descriptions are shown in the official language in which they were submitted.
cA.153
1053~17
This invention relates to membrane filtration
processes and to improved semi-permeable membranes for use
in such proces~es.
Semi-permeable membranes used in membrane
filtration processes enable the separation to be effected
of material down to molecular dimensions, usually from
aqueous systems. According to the selectivity of the
membranes, otherwise expressed as the rejection character-
istics, they ~ind widespread application for example in
desalinating brine, purifying effluent and concentrating
milk protein, particularly in whey.
In hyper~iltration processes in which small solute
molecules of molecular weight le~s than about 100 can be
separated, membranes of fine pore size are employed in
conjunction with filtration pressures of 1,000 psi or more
which are necessary to overcome the considerable osmotic
pressure generated by the small molecules. The selective
rejection of much larger molecules, eg proteins, of
molecular weight generally over 1,000, is effected on the
other hand by membranes of more open pore structure, in
ultrafiltration processes in which osmotic pressure is
negligible and in which therefore substantially lower
filtration pressures are adequate, generally about 100 psi
or even less.
The present invention provides a semi-permeable,
ultrafiltration membrane, suitable for use in ultra-
filtration processes, in which a minor amount of an inert,
impervious, water-insoluble, preferably inorganic, solid
material in the form of finely-divided non-colloidal
particles is dispersed wholly within the membranes so as to
-- 2 --
cA.153
1053417
swell or otherwise change the structure of the membrane,
thereby increasing its flux rate.
It has been found that the membranes of the
invention can ~xhibit up to 2-3 times the flux, at a
given pressure and temperature of otherwise identical
membranes without the added particles. On the other hand,
the rejection characteristics of the membranes towards
protein and other large organic molecules which can
normally be separated by ultrafiltration remain substantially
unaffected. The flux rate is the flow rate that can be
treated by unit area of membrane, and i9 commonly expressed
in gallons per 24 hours per ft2, either US or Imperial gallons.
The exact mechani~m by which the inorganic material
improves the membranes is not known. The filtration of
proteins in milk or other aqueous systems is adversely
affected by the build-up of proteins on the surface of the
membrane, forming a second "filter" having a poor flux rate.
Without wishing to diminish the scope of the invention
described by any expression as to its mechanism, it is
believed that filtration through the membranes of this
invention is improved by charged groups carried by the
particles, effective amounts of which thus repel milk
proteins from the membrane surface. This leads to higher
permeation rates of water and dissolved salts through the
membrane, by lining the surface of passages through the
membrane skin to give these areas a negative charge and
thereby allowing effusion of neutrally-charged molecules
through the membrane while rejecting charged molecules such
as p~otein. It should be emphasised that the membranes of
the present invention being of open pore structure,
cA.153
lQ534~7
exercise no selective filtration action on aqueous
solutions of small solutes, eg brine solutions 9 capable
of exerting a strong osmotic pressure and are thus
distinguished from reverse osmosis membranes which do so.
Their application in ultrafiltration processes lies in their
selective rejection of comparatively large molecules, for
example, proteins. The limits of their effectiveness for
this purpose, that i9, the minimum size of molecules which
they are capable of rejecting, depends largely upon the
effective pore size of the membrane and hence upon the
conditions and materials of its preparation but also upon the
conditions under which it is used, particularly the operating
pressure, increased pressure often effectively decreasing
pore size. This i9 particularly observed where, as in
milk and whey concentration using ultrafiltration methods, a
wide spectrum of solute molecular sizes is present, providing
a build-up of the bigger rejected solute molecules on the
membrane in a layer which itself exercises a filtration
action in the smaller molecules to which the membrane itself
is non-rejecting. Thus, lactose solutions may be found to
be filtered unchanged through a membrane which will however
at least partially reject the lactose in milk or whey in the
presence of the protein molecules, and the degree of rejection
may then be enhanced with increased pressure above that
customarily adopted for ultrafiltration.
Semi-permeable membranes are generally cast from a
solution, usually referred to as dope, of a film-forming
polymer in an organic solvent, the membranes used in hyper-
filtration processes being cast from volatile solvents, for
example acetone. The ultrafiltration membranes of the
_ ~ _
cA. 153
~0534~7
present invention, however, are conveniently prepared from
dope comprising non-volatile organic solvents having a
boiling point substantially in excess of 100C. Suitable
solvents include dimethyl formamide, dimethyl sulphoxide and
triethyl phosphate. It is surprising that membranes cast
from volatile solvents show no improvement when particles
are incorporated but on the contrary often exhibit flaws and
are then wholly unsuitable for use. An important feature
of the present invention is in the preparation of cellulose
acetate ultrafiltration membranes, particularly from
solutions in dimethyl formamide. These membranes can be
used at elevated temperatures up to approximately 80C,
enabling ultrafiltration processes to be carried out at
temperatures at which, for example, milk or whey may be
pasteurised.
The concentration of polymer in the dope is not
critical. Solutions from about 5% to 50% and above may be
used if desired, up to the limits of solubility of the
polymer. Preferably, however, a solution of 10-30% by weight
concentration is used. Very dilute solutions tend to form
very fragile membranes, while those prep~red from very
concentrated solutions may be tough but are often slow in
use.
The invention also provides a method of preparing
improved ultrafiltration membranes in which a solution of
film-forming polymeric material, for example a cellulose
ether or ester, is dissolved in a non-volatile solvent
and an inert, finely-divided inorganic water-insoluble
material is added having a mean specific surface area of preferably
30 at least 50 m2/g, and distributed throughout the solution,
cA.153
1053417
a film is cast and the solvent is leached by contact with
a miscible solvent in which the polymer is insoluble.
The particles should be small compared to the
molecular size of the membrane material. Bigger particles
5 exceeding the cellular dimensions tend to form gaps in the
membrane. Only a comparatively small concentration is
needed to provide effective cover for all the membrane
interfaces with the liquid to be filtered. The particles
should preferably constitute at least 1% of the membrane
casting solution, preferably 1-4% for carbon particles and
from 4-25% is particularly preferred for metal particles.
These amounts are expressed in the specification by weight,
as grammes per cc of solution, 1% therefore representing
1 gramme as additive per 100 cc of solution. Greater
15 amounts, up to 50%~ may enhance the membrane flux still
further, but some loss may then occur of selectivity, to
- give a lower rejection factor towards molecules of specified
size. The concentration at which this occurs is dependent
upon the nature of the casting dope, including the size and
nature of the active material. With these greater
quantities the membranes may then become selective only
towards the bigger molecules such as bacteria, while passing
even milk protein, or defects in the membrane may develop.
However, as much as 50% may be acceptable of some material,
25 eg metal particles, without loss of milk protein rejection.
In general also a greater change is effected using dimethyl
sulphoxide than dimethyl formamide as the casting solution.
Suitable material to be added to the membrane in
the form of particles in accordance with the invention
include lamp black, carbon black and soot. Other inorganic
cA.153
1053~7
materials which may be used include iron and ferrous alloys
including steel, metals generally if these are stable,
both elements and their alloys, particularly nickel, cobalt,
aluminium and their oxides, silica, silicon, sulphur and
alumina. The materials should be substantially insoluble
in water and the casting solvent, and exert no hydrolytic,
catalytic, oxidative, reductive or other chemical change
likely to lead to deterioration of the casting solvent or
membrane material. They should be impermeable and should
not penetrate the membrane when this is formed. The
particles ~hould not form suspensions in water.
A wide range o~ particle sizes may be adopted, but
the best size range may vary from one material to another.
Thus, for carbon particles a range of 10-30 millimicrons
is preferred, whereas for silica and metal particles the
individual particle size should preferably be within the
range 1-5 micron. Particles of metals, for example
stainless steel, exhibit a tendency to aggregation and may
be used in aggregated form, up to 200 microns in size, or
even more, and selected ranges of aggregates may show
improved behaviour compared with the rest, according to the
nature of the dope solvent and the concentration of the
added particles. While the coarser fractions of aggregated
metal particles may exercise a greater effect, they may
alternatively lead to membrane defect.
The particles themselves exhibit no permeability and
when slurried in water they should give a pH of 3-7.
The membranes of the invention may be prepared from
a variety of polymers. These are preferably cellulose-
based, preferably lower esters or ethers, eg acetate,
cA.153
1053~17
propïonate and butyrate, or methyl, ethyl or propylcellulose. Other polymers which may be used to prepare
the membrane include poly-ion polymers, prepared by
reaction of poly-anions with poly-cations, polyvinyl
chloride, polyacrylonitrile, poly-olefins and polyacrylic
esters, particularly of lower alcohols. Apart from the
addition of the inorganic particles, the membranes of the
invention may be prepared by methods which are conventional
for the preparation of ultrafiltration membranes. Thus,
the casting solution after the addition and distribution
therein of the inorganic particles, is cast as film in
flat, tubular or other convenient form, for example as
fibres, preferably at room temperature, but if desired at
other temperatures~ and i9 preferably contacted less than a
minute afterwards, in a leaching bath, for example water,
where the solvent diffuses out through the membrane into the
leaching bath while the water from the bath passes through
the membrane.
The membrane may also be cast directly onto a porous
backing providing adequate mechanical support for the
membrane. In any case, preferably the membrane when
completed is between 5 and 25 mils in thickness, ie 0.012-
0.0625 cms, but membrane thicknesses up to 1 mm or even more
may be suitable. Thicker membranes are more robust, but
show a corresponding decrease in flux rate. As in
conventional procedure in the preparation of semi-permeable
membranes, the thickness of the membrane may be controlled
by the method of applying the dope to the support on which
the film is prepared, and its concentration.
cA.153
~053417
In contrast to membranes cast from volatile solvents,
which form an active layer at the air interface that
perl'orms the selective filtration function and must be
exposed to the solution side of the filtration system for
best effect,membranes cast from non-volatile solvents form
a corresponding skin serving the same purpose at the
interface with the surface on which the film is cast, and
this must be exposed with the skin on the filtrate side,
remote from the solution undergoing filtration, to exhibit
a high flux while being selective to larger molecules in an
ultrafiltration capacity. In the preparation of a
membrane according to the invention it is found that the
greatest flux improvement effected by a given quantity of
additive particlesoccur~ when they settle in the membrane
casting, concentrating near the interface with the support
material in the active layer and thus providing an ani~trnpic, ie
asymmetric distribution. To this end, aggregated particles
are preferred which settle rapidly in the dope. Sufficient
time should be permitted for this to occur, but the membrane
should in any event be leached to remove solvent, within
five minutes of completing the casting. ~n asymmetric
distribution may however be encouraged in internal membranes
supported on tubes, by rotating these to apply centrifugal
force to the particles in the casting. In use, the
prepared membrane is mounted in a suitable test cell or
similar arrangement providing adequate mechanical support
for the membrane and the milk or other liquid system to be
filtered is supplied under pressure to the contact surface
of the membrane. The liquid is usually circulated
continuously until the degree of concentration required is
g
cA.153
~053~17
obtaincd.
In the following Examples a series of membranes
was prepared using a variety of particulate additive
materials, all of which exhibited a specific surface area
o~ at least 502m /gm and a pH of about 5. This was
measured by immersing an electrode oi a pH meter in the
supernatant liquor obtained by slurrying about 5~ of the
material under test in water.
The stainless steel particles used were of 316L
stainless steel, containing 14~ nickel, 17~ chro~ium and
2 . 5% molybdenum. They were nominally 5 microns diameter
but aggregated. In Examples 5 and 6 the particles were
sieve-graded and the fractions obtained were used in
separate tests to demonstrate the effect of the extent of
15 aggregation between the particles upon the flux rate.
About three-quarters of the aggregate was of mesh sieve
size 60-90 microns.
In each case the same grade of secondary cellulose
acetate was used to prepare the membrane, which was cast
at about 15C.
~LE 1 '
A semi-permeable membrane was prepared from a casting
solution having the following composition:-
20 gms cellulose acetate
100 mls dimethyl formamide
4 gms carbon black, particle size 14 millimicrons, pH 5Ø
The carbon was added to a solution of the cellulose
acetate in the dimethyl formamide, giving a viscous liquid
which was cast into a membrane by spreading the solution
with a doctor blade onto a plate of optical glass permitting
I ~
cA.153
lOS3q~17
a blade opening 0.2 cms. After 1 minute, the plate was
immersed in water to provide a membrane 0.018 cms in
thickness .
A similar membrane, prepared without the introduction
of carbon black particles was also prepared. The two
membranes were then compared by filtering skim milk having
a pH of 6.8, at a working pressure of 200 psig at 50C.
The results of the comparison are set out in Table 1, from
which it will be apparent that a marked improvement in flux
10 rate and selectivity results from the presence of carbon
black in the membrane. `
TABLE I
Example Control
Temp. C 15 50 15 50
Flux rate usgfd* 10.5 13.5 4.9 5.8
Lactose in permeate wt % 2.5 1.9 3.7 3.3
* US gallons per ft2 of membrane per 24-hour day.
From this data the rejection characteristics of the
membrane according to the invention, with respect to lactose,
was calculated as 10/, wbile that for the control was 8 ~ .
EXAMPLE 2
Membranes were prepared as described in Example 1,
except that instead of carbon black, particles of micronised
stainless steel were used having a particle size range of
1-5 microns. The thickness of the resulting membranes was
8 mils in each case, ie 0.020 cms.
The results of test runs carried out at 16C and
200 psig, on skim milk with a pH 6.8, are given in Table II.
cA.153
1~5341'7
TABLE II
Wt /0 Flux Rate Wt % lactose
Steel usgfd in permeate
0 5.46 ~ 3.3
12 9.1 ~ 3.0
18 9.8 ~ 2.8
24 10.5 2.2
EXAMPLE 3
*
A In this Example a micronised silica (Gasil 200) of mean
particle size 5 microns was incorporated in a series of
membranes otherwise prepared as described in Example 1. The
membranes were tested at various temperatures as previously
described, and the results appear in Table III.
TABLE III
oh Silica Operatin~ Temp.C Flux (usgfd)
0 16 5.46
0 50 5.95
6 16 9.1
6 50
12 16 9.8
12 50 12.6
18 16 11.~
24 16 10.5
EXAMPLE 4
The effect of changing the solvent on the properties of
the membranes according to the invention was examined in this
Example. Membranes were prepared otherwise as described in
Example 1, from dimethyl formamide and dimethyl sulphoxide,
using as the particulate matter Supercarbovar carbon, of
particle size 14 millimicrons and pH 5. The membrane
A~rK
- 12 -
cA. 153
1053~7
thickness in each case was 8 mils, and the membranes were
tested for the concentration of skim milk, against similarly
prepared control membranes containing no inert particles, at
room temperature (20C) and 200 psig, with the results
reported in Table IV.
TABLE IV
Solvent Dimethyl formamide Dimethyl sulphoxide
I .
Film thickness 8 15 8 15
I_ l_ _ _ . _ .
Wt % carbon 02 ¦ 40 2 4 0 2 4 0 2 4
I_ l _ _ _ . _
Flux usgfd ~ 9.1¦10.5 _ _ ~ 9.5 11.9 14.7 7.7 _ 11.8
EXAMPLE 5
Cellulose acetate dopes were prepared containing 20
grammes of cellulose acetate powder, 100 mls of dimethyl
formamide as solvent and powdered steel. A control dope of
the same proportions of cellulose acetate and solvent was also
prepared. The cellulose acetate was of grade E 3983, as
supplied by Eastman Kodak Ltd, containing 39.8% acetyl groups
and a viscosity No. 3. The dopes containing powdered steel
were vigorously shaken to give a uniform mixture before being
cast.
Casting was carried out in tubular modules mounted
vertically, each consisting of a fibreglass porous support tube
4 ft long and 4 inch in internal diameter, with walls about ~
inch thick. A plug of the dope was drawn through the tube in
each case, at about 2 feet per minute from the bottom of the
tube to the top at about 15C on a stainless steel, conical
casting bob about 3 inches in length, by means of which the
plug of dope was pushed up the tube to apply a uniform
layer of the dope about 1 ml thick.
-- 13 --
1053~7
The tubes containing the bigger aggregates in the
range examined were rotated by hand before admitting water into
the tube, to centrifuge the particles toward the membrane-support
tube interface.
Finally the tubes were emptied and mounted in a tube
separation unit for testing, which was carried out as follows:-
Pasteurised milk of zero fat content and 3-3.5 wt %
protein was circulated throught the unit at 50C and at measured
pressures and circulation rates. The protein rejection of each
membrane was determined by examining the filtrate, using a Pro-
milk Analyser (Ross Electric Co.), by a dye-binding detection
method using an amido-black dye.
From an examination of the results, it appeared that
a substantial increase in flux rate was provided by the membranes
according to the invention, from as little as 8 wt % (grammes
per cc of solvent) being sufficient to double the flux rate
compared with the control membrane. Greater quantities provided
a corresponding increase in flux rate, up to 50O/o which was the
maximum concentration measured. On the other hand, the complete
protein rejection of the membranes was sustained for all particle
si~es, from 50-200 microns, except at greater concentrations
than 25% with the smallest particles. Further particulars appear
in Tables V and VI, obtained at an operating pressure of 60 psig,
an operating temperature of 50C and a milk circulation throughput
of 700 gals per hour. In regard to Table V, protein rejection
failure was observed at other pressures and flow rates using more
than 25% of the steel particles.
- 14 -
Further tests esta~lish 3 that the flux remained
substantially higher for all the membranes made in accordance
with the invention by comparison with the control membrane,
whatever the extent to which the milk was concentrated, the
flux falling progressively at similar rates with increase in
the extent to which the milk was concentrated, both for the
control and for the membranes according to the invention. A
comparison with commercially-available tubular modules showed
that these exh~bited similar flux rates at the same
operational pressure, as the control membranes prepared in
the Example. Details of these further tests are given in
Table VII.
TABLE V
Concentration of Flux Rate Rejection to
steel of particle sizeIG/ft2 daY protein
15120-150 micron %
%
0 15 lO0
8 30 100
lO0
lS 45 lO0
. lO0
42 lO0
200 nil
300 nil
450 nil
- 15 -
cA. 153
1053'~7
TABLE VI
( 100% protein rejection throughout)
Particle size Flux
microns IG/ft2 d
(20 g/100 ml solvent) ' ay
60-90 30
90-120 30
120-150 50
150-200 50
TABLE VII
10Circulation Rate De~rl ,e of Concn. (X-fold)
Gals/hr. 1 2 l 3 4
60 psi~
700 42 23 1 16 11
100 psi~
700 60 395 25 16
(16)(10) (6) (4 5)
1200 78 50 33
In Table VII the flux obtained from the control
tubes, supporting membranes free from these steel particles,
is given in brackets. It will be observed that the flux
obtained using tubes according to the invention is
substantially higher, even at the 60 psig operating pressure,
than that obtained with the control at the higher operating
pressure of 100 psig for the same circulation rate and at all
degrees of concentration at which the measurements were made,
the flux falling at substantially the same rate in all cases
as the degree of concentration is increased.
- 16 -
cA.153
1053~17
EXAMPLE 6
In this Example a comparison is made of the efiect
of diiierent solvents on the rejection/flux rates of membranes
prepared in accordance with the invention. A similar method
was used for preparing tube-supported membranes as described
in Example ~, with 45~ in each case oi similar steel
particles. One pair oi tubes containing unfractionated
aggregates and made irom dope with dimethyl sulphoxide (DMSO)
and dimethyl iormamide (DMF) as solvent. Both gave a three-
fold improvement in ilux rate over corresponding control tubeswith no additive, in the concentration oi skim milk, protein
rejection remaining at 100~. ~he control made using DMSO
had a substantially higher ilux rate itseli than that made
using DMF,
lS When these comparative tests were repeated, using
however the 60-90 micron sieve fraction oi the steel
aggregate, a complete 1099 of selectivity ior milk protein
was observed with the DMSO tube, which however showed lOOyo
rejection of milk bacteria with no ilaws, enabling the milk
to be cold-sterilised by iiltration through the membrane
without change in composition.
In the case oi the DMF membrane, protein re~ection
remained at 100~, but the flux rate increased iurther, to a
similar rate to that obtained using the whole aggregate with
~5 DMSO.
In comparable tests in which in all respects the
particulars oi Example 1 were iollowed to obtain closely
similar membranes with the exception that the dope solvent
was acetone, the membrane was found to be wholly unacceptable
when particles oi any oi the additive material~ were present.
~ cA.153
~053'~17 `
EXAMPLE 7
This Example illustrates an alternative method
o~ preparing the membranes of the invention, in which instead
of incorporating the particulate matter in the dope before
the film is cast on the support surface, it is previously
spread on the latter and the film is cast over it. The
particles should not of course embed into the support layer,
so that the dope is free to percolate and thus substantially
wholly incorporate the particles.
A slurry of the particulate matter, comprising the
aggregated stainless steel particles described in the
preceding Examples in a 20~ gms/cc. concentration in
acetone, was painted on a glass plate and the solvent
evaporated, leaving a deposit of the particles on the plate
oi about 20 mgms per cm2.
A cellulose acetate film was formed on the plate as
described in Example 1, over the coating of steel particles.
On testing the film as described in Example 1 a
three-fold improvement in flux rate was observed over a
control membrane prepared under similar conditions but with
no particles, in the concentration of skim milk at a
circulation of 6 gallons per hour and a pressure of 50 psig,
through a membrane 112 inches in diameter.
A comparable improvement was obtained using similar
particles to those described in Example 1. It was also
found that the amount of the particles deposited could be
varied at least between 5 and 500 mgms. per cm2 and could be
applied directly to porous support means for the film, on
which the dope was cast in situ, either in plate or tubul~r
form.
