This invention relates to a process for modifying gas separation membranes and to the resultant gas separation membranes.

The use of membranes comprising a polyimide discriminating layer to separate gases is known in the art. The known membranes rely on differences in the relative permeability of the gases through the discriminating layer. Typically a mixture of gasses is brought into contact with one side of the membrane and at least one of the gases permeates through its discriminating layer faster than the other gas(es). In this way, the initial gas stream is separated into two streams, one of which is enriched in the selectively permeating gas(es) and the other of which is depleted.

One of the problems with gas separation membranes is that the discriminating layer can become plasticized, reducing its ability to discriminate between different gases and reducing its selectivity. Furthermore, defects such as craters are often present in the discriminating layer and these can result in surface leaks.

Covalent crosslinking of the polyimide discriminating layer of gas separation membranes was investigated in WO 2006/009520 and also by Tin et al in the article entitled "Chemical cross-linking Modification of Polyimide Membranes for Gas Separation", Journal of Membrane Science 189 (2001 ) 231-239.

US 5,286,280 describes the preparation of composite membranes comprising a porous support, an intermediate gutter layer and a discriminating layer made from what is known as a "6FDA" type polyimide.

The present invention provides a process for modifying gas separation membranes comprising a polyimide discriminating layer under mild, cost-effective conditions in order to improve their properties. The process can be performed quickly and provide membranes having a very thin discriminating layer.

According to a first aspect of the present invention there is provided a process for modifying (or preparing) a gas separation membrane comprising a polyimide discriminating layer having acid groups, the process comprising the step of crosslinking the acid groups with an organic polyamine, CHARACTERISED IN THAT the organic polyamine has a weight average molecular weight (Mw) of at least 400 KDa and the process is performed such that said crosslinking is predominantly non-covalent crosslinking.

The acid groups are preferably selected from sulphonic, sulphinic, phosphoric and phosphonic acid groups and especially carboxyl groups. The acid groups may be all the same (e.g. all are carboxyl groups) or the acid groups optionally comprise two or more types of acid groups (e.g. two or more of the aforementioned acid groups, for example sulphonic acid groups and carboxyl groups). Preferably the acid groups comprise carboxyl groups and/or sulphonic acid groups, more preferably the acid groups are carboxy groups or sulphonic acid groups.

The acid groups may be in any form, for example the free acid or salt form, e.g. in the form of a salt with a metal, ammonia or an amine (e.g. a primary, secondary or tertiary amine, preferably comprising six or less carbon atoms).

Preferred carboxyl groups are of the formula -CO2H or a salt thereof (e.g. one of the aforementioned salts).

Typically the acid groups of the polyimide discriminating layer are crosslinked by bringing the polyimide discriminating layer into contact with a composition comprising the organic polyamine. In order to achieve crosslinking which is predominantly non-covalent crosslinking, one will generally choose relatively mild conditions for such contact. The precise conditions chosen depend to some extent on the acidity of the acid group, the basicity of the organic polyamine and the (undesirable) reactivity of the organic polyamine towards the polyimide. Generally one will select milder conditions for organic polyamines which have higher reactivity with the polyimide in order to avoid or reduce the extent of covalent bond formation between the polyimide and the organic polyamine.

Typically the process is performed entirely at a temperature not exceeding

49 ^° C, more preferably not exceeding 45 ^° C, especially at a temperature in the range 5 to 40 ^° C, more especially 10 to 30 ^° C. Low temperatures such as these are preferred because they reduce the chances of covalent crosslinking between the polyimide support and the amine groups of the organic polyamine.

The process is preferably performed such that the polyimide discriminating layer having the acid groups is in contact with the composition comprising the organic polyamine for 1 to 45 minutes, more preferably 2 to 10 minutes, preferably at a temperature above room temperature (e.g. at least 25 ^° C).

One may determine whether the crosslinking is predominantly non-covalent crosslinking by repeating the process (i.e. performing a Comparative Process) under identical conditions using a membrane which is identical except that the polyimide layer lacks the acid groups and determining whether the resultant membrane has any crosslinking. If crosslinking is detected when the process is performed using a membrane comprising a polyimide discriminating layer having acid groups, but repetition under identical conditions using a membrane comprising a polyimide discriminating layer which is free from acid groups shows no crosslinking (i.e. in the Comparative Process), then the conditions under which the process of the invention was performed were such that the crosslinking was predominantly (or entirely) non-covalent crosslinking. One may detect the presence or absence of crosslinking using energy- dispersive X-ray ("EDX") mapping, e.g. using a Jeol JSM-6335F field emission scanning electron microscope.

A further technique for determining whether the crosslinking is predominantly non-covalent crosslinking is to observe the infra red spectrum of the polyimide before and after crosslinking. If the infra red spectrum at about 1718, 1783 and 1351 cm ^"1 remains substantially the same after crosslinking as before this indicates that the imide ring has remained intact, even after the crosslinking, and hence the crosslinking is predominantly (or entirely) non-covalent.

Furthermore, while ionically bonded and hydrogen bonded organic polyamine crosslinkers may be removed from the membrane by adjusting its pH, covalent bonding is more permanent and the crosslinker is much more strongly bound to the discriminating layer. Thus the removability of the organic polyamine crosslinker by pH adjustment also indicates whether or not the crosslinking is non- covalent.

One may use the following calculation to determine the ratio of non- covalent to covalent crosslinking (i.e. the NCC%) caused by the organic polyamine: NCC% = (Mremovable/Mtotal) X 100% wherein:

Mtotai is the total mass per cm ² of organic polyamine which is present on the modified membrane before it is stirred as described in M _re movabie below; and

Mremovabie is the mass per cm ² of organic polyamine which is removed when the modified membrane is stirred at 20 ^° C with ten times its dry weight of 0.1 M NaOH for 10 minutes.

One may determine M _to tai by measuring the increase in mass per cm ² of membrane resulting from the crosslinking reaction. For example, one may weigh a dry sample of the membrane before and after crosslinking reaction and the increase in weight per cm ² is M _to tai-

One may use gas-liquid chromatography to determine the concentration of organic polyamine present in the 0.1 M NaOH and, knowing the volume of 0.1 M NaOH, and the area of the membrane one may calculate M _re movabie-

If the NCC% is >50% then the crosslinking is predominantly non-covalent crosslinking. Preferably the NCC% is >75%, more preferably >85%, especially >95%, more especially about 100%.

In ionic crosslinking, two or more of the acid groups are in ionised form (e.g. a carboxyl group is in the form -CO2 ^" and a sulphonic acid group is in the form - SO3 ^" ) and are linked by two or more protonated amine groups (e.g. -NH ₃ ⁺ groups) present in a molecule of the organic polyamine.

In hydrogen bonding, typically =0 groups (e.g. carbonyl groups (C=0) of two or more carboxyl groups) in the polyimide are linked by hydrogen bonding to the hydrogen atoms present in the amino groups (e.g. -NH ₂ groups) of the organic polyamine. Ionic crosslinking and hydrogen bond crosslinking are illustrated below in a purely schematic, non-limiting manner, where the substantially vertical line connecting the carboxyl groups is the polyimide backbone and R is the part of the organic olyamine linking amine groups:

ionic crosslinking hydrogen bond crosslinking

Typically the non-covalent crosslinking is a combination of both ionic crosslinking and hydrogen bonding. In contrast, in covalent crosslinking amino groups present in the organic polyamine condense with the polyimide discriminating layer to form amide bonds therewith, e.g. the acid groups (e.g. - CO2H) of the polyimide discriminating layer condense with an amino group (e.g. - NH ₂ ) of the organic polyamine to form an amide and typically water (H ₂ 0) (e.g. - C0 ₂ H + H ₂ N-→ -CONH- + H ₂ 0).

Typically covalent crosslinking occurs at high temperatures and involves the organic polyamine ring-opening the imide rings present in the polyimide discriminating layer to form a covalent bonds therewith.

When the acid groups of the polyimide discriminating layer are crosslinked by bringing the polyimide discriminating layer into contact with a composition comprising the organic polyamine, the composition preferably comprises at least 0.1 wt% organic polyamine, e.g. 0.1 to 20wt%, more preferably 0.25 to 15wt%, especially 0.5 to 10wt% of the organic polyamine. The identity of the remainder of the composition is not particularly critical, although typically it comprises water and optionally one or more water-miscible organic solvents (e.g. ethanol, methanol, isopropanol and/or n-methyl pyrrolidinone).

One may prepare a polyimide discriminating layer having acid groups by, for example, the reaction of a dianhydride and a diamine, at least one of which has an acid group. Typically the dianhydride and the diamine each comprise an aromatic (e.g. phenylene) group.

Examples of diamines include: 2,3,5,6-tetramethyl-1 ,4-phenylenediamine; 4,4'-[1 ,4-phenylenebis(1 -methyl-ethylidene)]bisaniline; 2,4,6-trimethyl- 1 ,3- phenylenediamine; 2,2-bis[4-(4-aminophenoxy)-phenyl]propane; 2,7-bis(4- aminophenoxy)-naphthalene; 4,4,-methylene-bis(2,6-diisopropylaniline); 1 ,4-bis(4- aminophenoxy)benzene; 4,4'-bis(4-aminophenoxy)-biphenyl; 1 ,3-bis(4- aminophenoxy)benzene; 4,4'-(methylethylidene)bisaniline; 4-isopropyl-1 ,3- diaminobenzene; 1 ,5-diaminodiphenylether; diaminonaphthalene; 4,4'- diaminodiphenylether; metaphenylenediamine; paraphenylenediamine; Ν,Ν'- metaphenylenebis(m-aminobenzanilide); and 3,3 -diaminobenzanilide and the aforementioned compounds having an acid group, e.g. a sulphonic, sulphinic, phosphoric or phosphonic acid group or especially a carboxyl group (e.g. 3,5- diaminobenzoic acid).

Examples of dianhydrides include: 3, 4,3', 4'- diphenyldi(trifluoromethyl)methanetetracarboxylicdianhydride (also called 6FDA); pyromellitic dianhydride; 2,3,4,3',4'-diphenylsulfone tetracarboxylic dianhydride; 3,4,3',4'-benzophenone tetra-carboxylic dianhydride; pyrazinetetracarboxylic dianhydride; 3,4,3',4'-diphenyldimethylmethane tetracarboxylic dianhydride; 3,4,3',4'-diphenyldi(trifluoro-methyl) methanetetracarboxylic dianhydride; 2,3,6,7- naphthalenetetracarboxylic dianhydride; 3,4,3',4'-diphenyl tetracarboxylic dianhydride; 3,4,9, 10-perylenetetracarboxylic dianhydride; 3, 4,3', 4'- diphenylethertetra carboxylic dianhydride; 1 ,2,4,5-naphthalenetetracarboxylic dianhydride; 1 ,4,5,8-naphthalenetetracarboxylic dianhydride; 1 ,8,9, 10- phenanthrene tetracarboxylic dianhydride; 3,4,3',4'-diphenylmethane- tetracarboxylic dianhydride; and 2,3,4,5-thiophenetetra-carboxylic dianhydride.

The dianhydride and diamine may be reacted together by any of the known means for forming organic polymers. Discriminating layers in film form may be prepared by melt pressing, melt extrusion, solution casting, and the like. When the discriminating layer is formed from polymer solution in organic solvent, it may be desirable to incorporate up to 100% by weight of soluble salt, based on the total weight of dianhydride and diamine, e.g. LiCI, LiBr, L1N O3 and/or CaC etc..

Preferably the polyimide discriminating layer comprises groups of the Formula (1 ) wherein R is an acid group:

Formula (1 ).

Preferably R is a carboxyl group or a sulphonic acid group.

US 5,286,280 describes the preparation of composite membranes comprising a porous support, an intermediate gutter layer comprising poly(dimethylsiloxane) groups and a polyimide discriminating layer made from what is known as a "6FDA" type polyimide. In order to prepare membranes used in the present process (i.e. pre-modification), one may follow the general method of US 5,286,280 except that during formation of the polyimide discriminating layer one also includes a monomer which provides an acid group in the resultant polyimide discriminating layer (e.g. one may include a 3,5-diaminobenzene compound having an acid group such as 3,5-diaminobenzoic acid or 3,5-diamino benzene sulphonic acid).

The average thickness increase of the discriminating layer as a result of crosslinking with the organic polyamine is preferably in the range 1 to 100nm, more preferably 2 to 90nm, especially 3 to 80nm, more especially 4 to 60 nm, particularly 5 to 50 nm.

The average thickness of the discriminating layer after it has been crosslinked is preferably in the range 50nm to 20pm, more preferably 50nm to 1 pm, especially 50nm to 200nm.

The organic polyamine may be any organic polyamine of Mw 400 KDa or higher which has amine groups capable of forming non-covalent bonds with the acid groups of the polyimide discriminating layer. The organic polyamine is free from inorganic groups, e.g. free from aluminosilicates. Preferably the organic polyamine is free from silicon atoms. In a particularly preferred embodiment, the organic polyamine contains only carbon, hydrogen, nitrogen and optionally oxygen atoms.

Preferably the organic polyamine contains amine groups which are quite close together, for example at least two amine groups of the organic polyamine are 2, 3, 4, 5, 6 or 7 bonds apart, e.g. 3, 4, 5, 6 or 7 bonds apart. For example, in 1 ,2-diaminoethane groups the amine groups are 3 bonds apart (-N-C-C-N-) and in a hexamethylene diamine groups the amine groups are 7 bonds apart (-N-C-C-C- C-C-C-N-). Preferably at least two amine groups of the organic polyamine are 7 bonds apart, more preferably 6 bonds apart, especially 5 bonds apart, more especially 4 bonds apart, particularly 3 bonds, more particularly 2 bonds apart.

The amine groups present in the organic polyamine may be, for example, in-chain (e.g. as in -CH ₂ -NH-CH ₃ ), terminal (e.g. as in -CH ₂ -NH ₂ or -C6H ₄ -NH ₂ ), or form part of a ring (e.g. as in -CH(CH ₂ CH ₂ ) ₂ NH or pyridinyl (-CshUN)), or a combination of two or more of such groups.

Examples of suitable organic polyamines include poly(allylamine) (PAA), polyethyleneimine (PEI), polyvinylamine, poly (1 -vinylpyrrolidone) (PVP), PVP-co- 2-dimethylaminoethylmethacrylate, poly(4-vinyl)pyridine (P4VP), polyethyleneorganic polyamine, amine-terminated polyethyleneglycol, amine- terminated polypropyleneglycol or polybutyleneglycol, in each case having an Mw 400 KDa or higher, and mixtures comprising two or more thereof.

The amine groups may be primary amine, secondary amine, tertiary amine, quaternary amine or a combination of two or more thereof. Preferably the organic polyamine comprises primary amino groups (-NH ₂ ) because this can result in membranes having particularly good durability.

The Mw of the organic polyamine is preferably >600 KDa, more preferably >700 KDa, especially >800 KDa, more especially >950 KDa. (KDa means thousand Daltons).

The process optionally further comprises the step of washing the modified membrane to remove free organic polyamine, e.g. using deionised water. The process may also comprise the step of drying the modified membrane, e.g. using a known technique such as a dryer, blower or by infra-red heating.

Using the present process one may form a very thin discriminating layer with good selectivity and a very low level of surface defects.

The process of the present invention optionally further comprises the steps of providing a porous support, forming a gutter layer on the support and forming the discriminating layer on the gutter layer. Typically the gutter layer and the discriminating layer are formed by a coating technique, for example one of the coating techniques mentioned below.

The primary purpose of the porous support is to provide mechanical strength to the discriminating layer without materially reducing the flux. Therefore the porous support is typically open pored, relative to the discriminating layer.

The porous support may be, for example, a microporous organic or inorganic membrane, or a woven or non-woven fabric. The porous support may be constructed from any suitable material. Examples of such materials include polysulfones, polyethersulfones, polyimides, polyetherimides, polyamides, polyamideimides, polyacrylonitrile, polycarbonates, polyesters, polyacrylates, cellulose acetate, polyethylene, polypropylene, polyvinylidenefluoride, polytetrafluoroethylene, poly(4-methyl 1 -pentene) and especially polyacrylonitrile. One may use, for example, a commercially available, porous sheet material as the support. Alternatively one may prepare the porous support using techniques generally known in the art for the preparation of microporous materials. One may also use a porous support which has been subjected to a corona discharge treatment, glow discharge treatment, flame treatment, ultraviolet light irradiation treatment or the like, e.g. for the purpose of improving its wettability and/or adhesiveness.

The porous support preferably possesses pores which are as large as possible, consistent with providing a smooth surface for the subsequent gutter layer (when present) or discriminating layer.

The porous support preferably has an average pore size of at least about 50% greater than the average pore size of the discriminating layer, more preferably at least about 100% greater, especially at least about 200% greater, particularly at least about 1000% greater than the average pore size of the discriminating layer.

The pores passing through the porous support typically have an average diameter of 0.001 to 10pm, preferably 0.01 to 1 pm. The pores at the surface of the porous support will typically have a diameter of 0.001 to 0.1 pm, preferably 0.005 to 0.05pm. The pore diameter may be determined by, for example, viewing the surface of the porous support by scanning electron microscopy ("SEM") or by cutting through the support and measuring the diameter of the pores within the porous support, again by SEM.

The porosity at the surface of the porous support may also be expressed as a % porosity, i.e.

% porosity = 100% x (area of the surface which is missing due to pores)

(total surface area)

The areas required for the above calculation may be determined by inspecting the surface of the porous support by SEM. Thus, in a preferred embodiment, the porous support has a % porosity >1 %, more preferably >3%, especially >10%, more especially >20%.

The porosity of the porous support may also be expressed as a CO2 gas permeance (units are m ³ (STP)/m ² .s.kPa). When the membrane is intended for use in gas separation the porous support preferably has a CO2 gas permeance of 5 to 150 x 10 ^"5 m ³ (STP)/m ² .s.kPa, more preferably of 5 to 100, most preferably of 7 to 70 x 10 ^"5 m ³ (STP)/m ² .s.kPa.

Alternatively the porosity may be characterised by measuring the N ₂ gas flow rate through the porous support. Gas flow rate can be determined by any suitable technique, for example using a Porolux™ 1000 device, available from Porometer.com. Typically the Porolux 1000 is set at the maximum pressure (about 34 bar) and one measures the flow rate (L/min) of N ₂ gas through the porous support under test. The N ₂ flow rate through the porous support at a pressure of about 34 bar for an effective sample area of 2.69 cm ² (effective diameter of 18.5 mm) is preferably >1 L/min, more preferably >5 L/min, especially >10 L/min, more especially >25 L/min. The higher of these flow rates are preferred because this reduces the likelihood of the gas flux of the resultant membrane being reduced by the porous support.

The above pore sizes and porosities refer to the porous support before it has been converted into a gas separation membrane.

The porous support preferably has an average thickness of 20 to 500 pm, preferably 50 to 400 μιτη, especially 100 to 300 pm.

In view of the foregoing, the gas separation membrane optionally comprises a porous support and a gutter layer, wherein the gutter layer is located between the support and the discriminating layer. The gutter layer usually has the function of providing a smooth and continuous surface for the discriminating layer.

Preferred gutter layers comprises poly(dimethylsiloxane) groups.

The gutter layer preferably has an average thickness 25 to 400nm, preferably 30 to 350nm, especially 50 to 300nm, e.g. 70 to 120nm, or 130 to 170nm , or 180 to 220nm or 230 to 270nm .

The thickness of the gutter layer may be determined by cutting through the gas separation membrane and examining its cross section by SEM. The part of the gutter layer which is present within the pores of the support is not taken into account.

The gutter layer is preferably non-porous, i.e. any pores present therein have an average diameter <1 nm.

Thus the present process optionally further comprises the step of forming the gutter layer between the support and the discriminating layer, preferably by a process comprising radiation curing of a radiation-curable composition.

The radiation curing may be performed using any source which provides the wavelength and intensity of radiation necessary to cause the radiation-curable composition to polymerise. For example, electron beam, UV, visible and/or infra red radiation may be used to cure the radiation-curable composition, the appropriate radiation being selected to match the components.

Preferably radiation curing of the radiation-curable composition used to form the optional gutter layer begins within 7 seconds, more preferably within 5 seconds, most preferably within 3 seconds, of the radiation-curable composition being applied to the porous support.

Suitable sources of ultraviolet light include mercury arc lamps, carbon arc lamps, low pressure mercury lamps, medium pressure mercury lamps, high pressure mercury lamps, swirlflow plasma arc lamps, metal halide lamps, xenon lamps, tungsten lamps, halogen lamps, lasers and ultraviolet light emitting diodes. Particularly preferred are ultraviolet light emitting lamps of the medium or high pressure mercury vapour type. In addition, additives such as metal halides may be present to modify the emission spectrum of the lamp. In most cases lamps with emission maxima between 200 and 450nm are particularly suitable.

The energy output of the irradiation source is preferably from 20 to 1000W/cm, preferably from 40 to 500W/cm but may be higher or lower as long as the desired exposure dose can be realized.

The radiation-curable composition used to prepare the optional gutter layer preferably comprises:

(1 ) 0.5 to 50 wt% of radiation-curable component(s);

(2) 0 to 5 wt% of a photo-initiator; and

(3) 50 to 99.5 wt% of inert solvent.

The radiation-curable component(s) typically have at least one radiation- curable group. Radiation curable groups include ethylenically unsaturated groups (e.g. (meth)acrylic groups (e.g. CH ₂ =CR ¹ -C(0)- groups), especially (meth)acrylate groups (e.g. CH ₂ =CR ¹ -C(0)0- groups), (meth)acrylamide groups (e.g. CH ₂ =CR ¹ -C(0)NR ¹ - groups), wherein each R ¹ independently is H or CH ₃ ) and especially epoxide groups (e.g. glycidyl and epoxycyclohexyl groups).

The preferred ethylenically unsaturated groups are acrylate groups because of their fast polymerisation rates, especially when the irradiation uses UV light. Many compounds having acrylate groups are also readily available from commercial sources.

Photo-initiators may be included in the radiation-curable composition and are usually required when the curing uses UV radiation. Suitable photo-initiators are those known in the art such as radical type, cation type or anion type photo- initiators.

Cationic photo-initiators are preferred when the radiation-curable component(s) comprises curable groups such as epoxy, oxetane, other ring- opening heterocyclic groups or vinyl ether groups.

Preferred cationic photo-initiators include organic salts of non-nucleophilic anions, e.g. hexafluoroarsinate anion, antimony (V) hexafluoride anion, phosphorus hexafluoride anion, tetrafluoroborate anion and tetrakis(2,3,4,5,6- pentafluorophenyl)boranuide anion. Commercially available cationic photo- initiators include UV-9380c, UV-9390c (manufactured by Momentive performance materials), UVI-6974, UVI-6970, UVI-6990 (manufactured by Union Carbide Corp.), CD-1010, CD-101 1 , CD-1012 (manufactured by Sartomer Corp.), Adekaoptomer™ SP-150, SP-151 , SP-170, SP-171 (manufactured by Asahi Denka Kogyo Co., Ltd.), Irgacure™ 250, Irgacure™ 261 (Ciba Specialty Chemicals Corp.), CI-2481 , CI-2624, CI-2639, CI-2064 (Nippon Soda Co., Ltd.), DTS-102, DTS-103, NAT-103, NDS-103, TPS-103, MDS-103, MPI-103 and BBI- 103 (Midori Chemical Co., Ltd.). The above mentioned cationic photo-initiators can be used either individually or in combination of two or more.

Radical Type I and/or type II photo-initiators may also be used.

Examples of radical type I photo-initiators are as described in WO 2007/018425, page 14, line 23 to page 15, line 26, which are incorporated herein by reference thereto.

Examples of radical type II photo-initiators are as described in WO 2007/018425, page 15, line 27 to page 16, line 27, which are incorporated herein by reference thereto.

Type I photo-initiators are preferred, especially alpha- hydroxyalkylphenones, such as 2-hydroxy-2-methyl-1 -phenyl propan-1 -one, 2- hydroxy-2-methyl-1 -(4-tert-butyl-) phenylpropan-1 -one, 2-hydroxy-[4 ^' -(2- hydroxypropoxy)phenyl]-2-methylpropan-1 -one, 2-hydroxy-1 -[4-(2- hydroxyethoxy)phenyl]-2-methyl propan-1 -one, 1 -hydroxycyclohexylphenylketone and oligo[2-hydroxy-2-methyl-1 -{4-(1 -methylvinyl)phenyl}propanone], alpha- aminoalkylphenones, alpha-sulfonylalkylphenones and acylphosphine oxides such as 2,4,6-trimethylbenzoyl-diphenylphosphine oxide, ethyl-2,4,6-trimethylbenzoyl- phenylphosphinate and bis(2,4,6-trimethylbenzoyl)-phenylphosphine oxide, are preferred.

Preferably the weight ratio of photo-initiator to radiation-curable components present in the radiation-curable composition is between 0.001 and 0.2 to 1 , more preferably between 0.01 and 0.1 to 1 . A single type of photo-initiator may be used but also a combination of several different types.

When no photo-initiator is included in the radiation-curable composition, the composition can be advantageously cured by electron-beam exposure. Preferably the electron beam output is between 50 and 300keV. Curing can also be achieved by plasma or corona exposure.

The function of the inert solvent (3) is to provide the radiation-curable composition with a viscosity suitable for the particular method used to apply the curable composition to the porous support. For high speed application processes one will usually choose an inert solvent of low viscosity. The number of parts of component (3) is preferably 70 to 99.5wt%, more preferably 80 to 99wt%, especially 90 to 98wt%.

Inert solvents are not radiation-curable. In a specific embodiment there is no solvent present.

The radiation-curable composition may contain other components, for example surfactants, surface tension modifiers, viscosity enhancing agents, biocides and/or other components capable of co-polymerisation with the other ingredients.

The radiation-curable composition may be applied to the porous support by any suitable coating technique, for example by curtain coating, meniscus type dip coating, kiss coating, pre-metered slot die coating, reverse or forward kiss gravure coating, multi roll gravure coating, spin coating and/or slide bead coating.

Conveniently the radiation-curable composition may be coated onto the porous support by a multilayer coating method, for example using a consecutive multilayer coating method, optionally along with the components used to form the discriminating layer

In a preferred consecutive multilayer process a layer of the radiation- curable composition and the discriminating layer (or the chemicals used to prepare the discriminating layer) are applied consecutively to the support, with the radiation-curable composition being applied before the discriminating layer (or the chemicals used to prepare the discriminating layer).

In order to produce a sufficiently flowable composition for use in a high speed coating machine, the radiation-curable composition preferably has a viscosity below 4000 mPa s when measured at 25°C, more preferably from 0.4 to 1000 mPa s when measured at 25°C. Most preferably the viscosity of the radiation-curable composition is from 0.4 to 500 mPa.s when measured at 25°C. For coating methods such as slide bead coating the preferred viscosity is from 1 to 100 mPa.s when measured at 25°C. The desired viscosity is preferably achieved by controlling the amount of solvent in the radiation-curable composition and/or by the conditions for preparing the radiation curable polymer.

In the multi-layer coating methods mentioned above, one may optionally apply a lower inert solvent layer to the porous support followed by applying the radiation-curable composition.

With suitable coating techniques, coating speeds of at least 5m/min, e.g. at least 10m/min or even higher, such as 15m/min, 20m/min, or even up to 100m/min, can be reached. In a preferred embodiment the radiation-curable composition (and also the discriminating layer) is applied to the support at one of the aforementioned coating speeds.

The thickness of the cured gutter layer on the support may be influenced by controlling the amount of curable composition per unit area applied to the support. For example, as the amount of curable composition per unit area increases, so does the thickness of the resultant gutter layer. The same principle applies to formation of the discriminating layer.

While it is possible to practice the invention on a batch basis with a stationary porous support, to gain full advantage of the invention it is much preferred to perform the process on a continuous basis using a moving porous support, e.g. the porous support may be in the form of a roll which is unwound continuously or the porous support may rest on a continuously driven belt. Using such techniques the radiation-curable composition can be applied to the porous support on a continuous basis or it can be applied on a large batch basis. Removal of the inert solvent from the radiation-curable composition membrane can be accomplished at any stage after the radiation-curable composition has been applied to the support, e.g. by evaporation.

Thus in a preferred process, the radiation-curable composition is applied continuously to the porous support by means of a manufacturing unit comprising a radiation-curable composition application station, curing is performed using an irradiation source located downstream from the radiation-curable composition application station to form a gutter layer, the discriminating layer is formed on the gutter layer by a discriminating layer application station, the crosslinking is then performed by contacting the discriminating layer with a composition comprising the organic polyamine and the resultant gas separation membrane is collected at a collecting station, wherein the manufacturing unit comprises a means for moving the porous support from the radiation-curable composition application station to the irradiation source and to the discriminating layer application station and into contact with the composition comprising the organic polyamine and to the gas separation membrane collecting station.

Optionally the polyimide discriminating layer is formed on the gutter layer by a radiation curing process. Under such circumstances, the manufacturing unit preferably further comprises an irradiation source or a heater located downstream from the discriminating layer application station, thereby radiation- or thermally- curing the components used to form the discriminating layer.

The radiation-curable composition application station may be located at an upstream position relative to the irradiation source and the irradiation source is located at an upstream position relative to the discriminating layer application station.

While it is preferred for the gutter layer to be pore-free, the presence of some pores usually does not reduce the permselectivity of the final membrane because the discriminating layer is often able to fill minor defects in the gutter layer.

According to a second aspect of the present invention there is provided a gas separation membrane obtained by a process according to the first aspect of the present invention.

The following materials were used in the Examples (all without further purification): PAN is a porous support (polyacrylonitnle L10 ultrafiltration membrane from GMT Membrantechnik GmbH, Germany),

UV9300 is SilForce™ UV9300 from Momentive Performance Materials

Holdings. This is curable copolymer comprising reactive epoxy groups and linear poly(dimethylsiloxane) chains and is used to prepare gutter layers.

UV9390C is SilForce™ UV-9390C - a cationic photo-initiator (a solution of a bis(4-alkylaryl)iodonium hexafluoroantimonate salt and photosensitizer in a glycidyl ether reactive diluent).

Ti(OiPr) ₄ is titanium (IV) isopropoxide from Dorf Ketal Chemicals.

n-heptane is n-heptane from Brenntag Nederland BV.

MEK is 2-butanone from Brenntag Nederland BV.

MeOH is methanol.

CH is cyclohexanone from Brenntag Nederland BV.

PI1 : is 6FDA-TeMPD _x /DABAy,x/y=80/20; obtained from FUJIFILM

Corporation, having the following structure:

PI2: is 6FDA-TeMPD; obtained from FUJIFILM Corporation, having the following structure:

PI3 (6 FDA- TeMPD)n-(6FDA-DABS)y n/y = 80/20 obtained from

FUJIFILM Corporation, having the following structure:

PEI: is a 50 wt% solution of polyethylene imine in water, from Aldrich having the Mw specified in Table 2.

P4VP: is poly (4-vinylpyridine) of Mw 160 kDa from Aldrich.

PAA: is a 20 wt% solution of poly(allyllamine) of Mw 65 kDa in water, from

Aldrich.

PDMS: is poly(dimethylsiloxane), bis(3-amino-propyl)-terminated of Mw 2.5 kDa from Aldrich.

PVP-co: is a 19wt% solution of poly(1 -vinylpyrrolidone-co-2- dimethylaminoethyl methacrylate), Mw 1000kDa, in water, from Aldrich.

PVP: is a 50wt% solution of poly(l -vinylpyrrolidone) having the Mw specified in Table 2 in water, from Aldrich.

Evaluation of Gas Flux, Selectivity, Crater Defects (leaks), Water-contact Angle and Crosslinking Type

(A) Gas flux

The flux of CH ₄ and CO2 through the membranes was measured at 40°C and gas feed pressure of 6000 kPa using a gas permeation cell with a measurement diameter of 3.0 cm and a feed gas composition of 13 v/v % CO2 and 87 v/v % CH ₄ .

The flux of each gas was calculated based on the following equation:

Qi ⁼ {0perm ^' Xperm,i)/( ^' (PFeed ^' XFeed ~ Pperm ^' Xperm,i))

Where:

Qi = Flux of each gas (m ³ (STP)/m ² kPa s)

9perm = Permeate flow (m ³ (STP)/s)

Xperm = Volume fraction of each gas in the permeate

A = Membrane area (m ² )

PFeed = Feed gas pressure (kPa)

XFeed = Volume fraction of each gas in the feed Pperm = Permeate gas pressure (kPa)

STP is standard temperature and pressure, which is defined here as 25.0°C and 1 atmosphere (101 .325 kPa). (B) Selectivity

The selectivity (a _C o2 cH ) for the membranes was calculated from Q _C o2 and QCH4 calculated above, based on following equation:

(C) Crater-defects (leaks)

The membranes were examined for crater defects by dyeing the top- surface of the membrane, washing the membrane with water (5 litres) at room temperature and then visually examining the membrane. The presence of blue spots indicated the presence of craters. The dye was a 5wt% solution of tetrapotassium 2-(4-{5-[1 -(2,5-disulfonatophenyl)-4,5dihydro-3- ethylmethanoate - 5-oxopyrazol-4-ylidene]-1 ,3-pentadienyl}-3-ethylmethanoate-5-hydroxypyrazol-1 - yl)benzene-1 ,4-disulfonate in water.

(D) Water Contact Angle (WCA)

The WCA of the membranes was determined by using a VCA 2500 XE

(video contact angle analysis system) instrument from AST by the sessile drop method. On several locations at the top of each sample, 1 μΙ_ deionized water was injected and the images were recorded using a video camera system and the surface contact angles were calculated and averaged based on the recorded images of the water drops.

(E) Crosslinking Type

In order to determine whether the crosslinking was predominantly non- covalent, the discriminating layer was inspected for the reduction of intensity of peaks at approximately 1718cm ^"1 (attributed to the C=O symmetric stretch of the imide group), 1783 cm ^"1 (attributed to the C=O asymmetric stretch of imide group) and 1351 cm ^"1 (attributed to C-N stretch of imide group) indicating a imide ring cleavage and for the presence or appearance of two peaks at 1647 cm ^"1 (C=O stretch of the amide group) and 1521 cm ^"1 (a C-N stretch band of the C-N-H group) before and after crosslinking. In all Examples of the invention, no changes in intensities and no new IR peaks were observed after crosslinking, indicating that there was no discernible imide ring opening and therefore no discernable covalent crosslinking. Analogous techniques were used for detecting the presence of sulphonamide groups. The infra red measurements were done using a PerkinElmer Frontier FT-IR spectrophotometer using Attenuated Total Reflection (ATR) accessory on a germanium top plate from 4000 to 650 cm ^"1 in % transmission (T). Examples

Stage 1 - Preparation of a Support Carrying a Gutter Layer

A composition C1 having a viscosity of 64,300 mPas at 25°C (at 0.0396 s ^"1 ) was prepared by mixing the components described in Table 1 at 95°C for 105 hours:

Tablel

Viscosity was measured using a Brookfield LVDV-II + PCP viscosity meter, using either spindle CPE-40 or CPE-52 depending on viscosity range.

Radiation-curable composition RCC1 was prepared by cooling the above composition C1 to 20°C, diluting with n-heptane to a polymer concentration of 5wt %, filtering the resultant solution through a filter paper of 2.7 pm pore size and adding a photo-initiator (UV9390C, 0.50 wt%).

Radiation-curable RCC1 composition was then applied to a porous support (PAN) at a speed of 10 m/min by a meniscus dip coating and irradiated using a Light Hammer LH 10 from Fusion UV Systems fitted with a D-bulb with an intensity of 16.8 kW/m (70%). This resulted in a porous support having a gutter layer of dry thickness of about 200nm. Stage 2 - Application of the Discriminating Layer

A polyimide discriminating layer having acid groups was formed on the gutter layer prepared in Stage 1 by applying thereto a composition comprising PI1 (2wt%), CH (6wt%) and MEK (92 wt%) at 10 m/min by a meniscus type dip coating. After the cross-linking the membrane was washed to remove the excess of non-bound organic polyamine. The resultant discriminating layer had an average dry thickness of about 100nm.

The thickness of the gutter layer and discriminating layer were measured by cutting through the gas separation membrane and measuring the thickness from the surface of the porous support or the surface of the gutter layer outwards by SEM

Stage 3 - Non-covalently Crosslinking the Discriminating Layer

The acid groups of the polyimide discriminating layer arising from Stage 2 were then non-covalently crosslinked by applying thereto a solution of the organic polyamine indicated in Table 2 at the specified wt% in the solvent indicated in Table 2, at 10 m/min, again by meniscus type dip coating. The resultant, modified gas separation membranes had the gas flux, selectivity, crater visibility and water contact angle indicated in Table 2. The resultant discriminating layer had an average dry thickness of about 200 nm.

Examples 2 to 8 and Comparative Examples 1 to 17

Examples 2 to 7 and Comparative Examples 1 to 8 were prepared in an analogous manner to Example 1 except that Stage 1 (preparation of the gutter layer) was omitted and in Stage 2 the discriminating layer was formed directly on the PAN porous support.

Example 8 was performed exactly as described for Example 1 , except that

PI1 was replaced by an identical amount of PI3 to give a discriminating layer having different acid groups to those in Example 1 .

The organic polyamine (when used), solvent and results are shown in Table

2 below.

In Comparative Examples CEX1 , CEX3 to CEX6 and CEX8 the organic polyamines used had a Mw < 400 KDa. In Comparative Example CEX2 and CEX 7, Stage 3 (crosslinking) was omitted.

The results are shown in Table 2 below.

Table 2

Ingredient Example

CEX1 1 2 3 4 5 CEX2

Organic polyamine PEI PEI PEI PEI PEI PEI None

(wt% in solvent) (5) (2.5) (5) (0.1) (2.5) (50)

Organic polyamine Mw (KDa) 2 750 750 750 750 750 None

Solvent H ₂ 0 H ₂ 0 H ₂ 0 H ₂ 0 H ₂ 0 H ₂ 0 None

Gas Flux - 70 72 105 80 15 -

Selectivity (a _C0 2 cH4) - 19 18 18 18 19 -

Craters visible Yes No No No No No Yes

Water contact angle 39.7 32.8 39.6 40.5 39.2 37.2 85.8

Did IR analysis indicate No No No No No No No covalent crosslinking? means not measured due to craters

Table 2 Continued

Comparative Examples CEX9 to CEX17 were prepared in an analogous manner to Examples 2 to 7 above (no gutter layer), using the ingredients and amounts shown in Table 3, except that in place of PI1 (an acid-functional polyimide used to provide the discriminating layer having acid groups) there was used an identical amount of PI2 (a polyimide which lacked acid groups, resulting in a discriminating layer which lacked acid groups). Table 3

- means not measured due to craters

Table 3. continued

- means not measured due to craters

(In Comparative Example CEX17 the organic polyamine was omitted)