MICROFLUIDIC FILTER This invention relates to improvements in microfluidic systems particularly in filters for use in micro fluidic processes.

Background to the invention Microfluidic systems are of interest in biological and chemical reaction methodology and analysis. In PCR processes and in the analysis of DNA micro beads are used.

Microparticle separations have the advantage in fluidic separations of providing a platform for tailored surface chemistry allowing for the easy manipulation of molecules from solution and simplifying processing requirements. They have been used for various applications, particularly in the areas of biology and healthcare where a large array of different particles, affinity mechanisms, and processes have been developed for techniques including immunoassay, cell separation and molecular biology. They are regularly employed in isolation and purification processes by binding the target molecule to the microparticle and thereby allowing the target to be transported and flexibly manipulated in various reagents.

The separation of the beads is an essential step in the progress of the reaction in the micro fluidic system.

WO 01/38865 discloses an on-chip packed reactor bed which includes a weir arrangement in the micro fluidic channel to create a trapping zone for the microparticles. This has the disadvantage of limited flux in larger channel to particle dimensions and design issues for dead volumes.

WO 01/85341 discloses a trapping device comprising a slotted wall. It is a high cost silicon based solution to the problem.

US 6,221, 677 uses a laminar flow control method where mixing is controlled by the diffusion of the different particles between two flowing streams. So the liquid contact time can be used to determine which particles get removed from solution.

USA patent application 2002/0185431 discloses a multilayer microfluidic device which incorporates a separate filter compressively restrained between device layers to promote a tight seal.

USA patent application 2003/0136451 discloses a method of fabricating a restriction in a microfluidic channel by polymerizing a prepolymer in situ.

It is an object of this invention to provide a low cost simple physical means of achieving micro particle separation in microfluidic systems.

Brief Description of the Invention To this end the present invention provides a microfluidic system which includes a) a first substrate which incorporates a first microfluidic channel b) a second substrate overlying said microfluidic channel which has at least one region that has been microperforated and which overlies said first microfluidic channel c) a third substrate overlying said second substrate and incorporating a second microfluidic channel or port adapted to communicate with said first microfluidic channel through said micro-perforated region.

This invention provides a method for integrating a filter onto a microfluidic platform using lamination based technologies. It is applicable to micro-fluid, including gas, particle separation in either in-line or cross-flow configurations. In particular it is useful for microchannel inlet protection or as a particle collector, such as in microsphere based biological assays. Thus the general range of the filters of this invention is primarily in the microfiltration range and the upper ultra filtration range.

The microfluidic devices of this invention may be formed by laminating a pre- perforated layer between two microchannel layers to form a filter membrane sized to allow fluid to pass and retain the microparticles. The membrane may be perforated by excimer ablation or by stamping or embossing.

This invention is particularly applicable to polymer microfluidic devices rather than silicon based micro fluidic devices. Polymer based microfluidic devices offer cost advantages. The technique of this invention easily integrates into existing manufacturing processes.

Any of the film forming polymers suitable for microfluidics may be used including polyethyleneterephalate (PET), polycarbonate, high density polyethylene and polymethylmethacrylate (PMMA).

These devices may constructed by laminating layers together using various bonding techniques, such as adhesive, thermal and solvent bonding.

Detailed Description of the invention Some preferred embodiments of the invention will be described with reference to The drawings in which Figure 1 illustrates schematically a number of arrangements for using the filter of the invention; Figure 2 illustrates an Excimer ablated filter membrane in 12pm PET a) Entrance holes 8pm b) exit holes 1 p ; Figure 3 illustrates a scanning electron microscope image of laser cut PET before a), and after b) bonding; Figure 4 illustrates a) a layered approach to microfluidic chip construction and b) fabricated device with insert showing filter chamber; Figure 5 illustrates a silica particle retention a) tightly packed column, b) image of microspheres in a channel with the capping layer removed; Figure 6 illustrates a membrane with 8pm entrance pores a) before filtration, and b) after filtration of methylene blue stained white blood cells.

With reference to figure 1 by arranging the micro channels and the flow of fluid through the micro channels various filter configurations may be achieved.

The perforated membrane 15 is laminated between two polymer films 12 and 13 incorporating micro channels 14. The filter 15 may be placed across the inlet 11 to the micro channel system as shown in figure 1A.

In a cross flow filter arrangement as shown in figure 1 B the filter prevents the micropartilces from moving into the permeate or outlet microchannel 17.

In an inline filter as shown in figure 1 C or a bead trap as shown in 1 D, the filter 15 prevents passage of the beads 16 into the outlet channel 17 and allows accumulation of the beads 16 to occur.

Typically the micro beads are of 0.1 to 100 microns in size usually about 5 microns.

The polymers used may be polycarbonate, high density polyethylene, polyethylene terephthalate (PET), or other suitable polymers. The same material may be used for the microfluidic substrates as for the membrane.

Example Devices were fabricated using PET film with a multilayer technique, chosen for its compatibility with web-based plastic film lamination systems for large-scale low cost manufacturing.

Channel Formation. The PET film was cut using a frequency tripled Nd : YAG laser (AVIA 3. 0W, Coherent Inc, USA) incorporated in a computer controlled 2-D laser cutting system (Lasertec, The Netherlands). The Q-switched, pulsed output of the laser generates a pulse duration of <30ns, a pulse energy of-180uJ, and for the purpose of this work a pulse repetition frequency of 10kHz was used. An f-theta lens having a focal length of 160mm was used to focus the beam to 30pm in diameter at the work piece. Computer Aided design (CAD) drawings were used to control the temperature compensated, x-y, galvo-scanner and synchronise the firing of the laser. The beam velocity at the work piece is determined by the scanner parameters, and was set to 0.11 m/s. This combination of parameters results in a pulse being fired every 11 pm, or a shot overlap equivalent to 2.7 shots per area.

The devices were constructed from layers of Luminar S10 (Toray Industries, Inc.) PET film. Individual layers were cut entirely through by passing over the same tool path many times. Cleaning of the samples was performed by sonication in a 1: 1 mixture of ethanol and water followed by multiple rinses in isopropanol and deionised water.

Membrane Fabrication. An Exitech S8000 excimer laser system equipped with a Lambda Physik LPX2101 laser source was used to perforate the 12pm PET film using standard mask projection lithography. The system illuminated a chrome-on- quartz mask with a homogenized 10mm square beam having an intensity deviation of approximately 5% RMS. A 10x projection lens with a NA of 0.3 was used giving a diffraction limited resolution of 0. 8pm.

Surface Modification. Chemical and UV surface modification were performed on both the 100pm and 12pm thick samples. The chemical treatment was achieved by placing the samples in a solution of 5M NaOH (Merck, Australia) at 80C for 1min to etch, as previously reported17, followed by rinsing under flowing isopropanol then deionised water. The UV exposure was performed by placing the samples in a custom made light box with four 14watt metal halide lamps (Amber

Lamps, GPH369T5VH). The samples were placed 60mm directly below the lamps and measurements were taken for timed exposures of 10,30, 60,150, 360 mins.

Bonding. An in-house embossing system was used to thermally bond the substrates. The unit consisted of a regulated hydraulic press (REXROTH, Australia) that applies pressure to the embossing chamber, with the temperature controlled (SHINKO, Australia) heating elements and forced air cooling. The samples were sandwiched in-between 1 mm of paper and flat plates of steel to ensure even pressure distribution. The chamber was evacuated (-80ka) and preheated to 190°C before being compressed to 7MPa for a period of 45 mins.

Cut profile characterisation and sample visualization was performed with a BX60 Olympus optical microscope. Detailed images were taken on a JEOL, JSM35, scanning electron microscope using gold coated samples, prepared with a Polarin Equipment sputtering unit. Height measurements were made using the Olympus OLS1200 He: Ne laser scanning confocal microscope.

Surface characterization was performed using XPS and contact angle measurements using a goniometer (Rame-hart, Inc). The samples were prepared by sonication in a 1: 1 mixture of ethanol and water followed by multiple rinses in isopropanol and deionised water. Samples were stored in air and analysed by XPS within two days. The XPS data was collected on a Kratos AXIS Ultra imaging XPS instrument using a monochromated Al Ka X-ray source at a power of 1486.6 eV.

The energy scale of the instrument was calibrated using the Au 4f7/2 photoelectron peak at binding energy (EB) = 83.98 eV. Charge correction was performed using the C 1s photoelectron component peak corresponding to C-C species at EB = 284.7 eV. As the samples were insulating, charge neutralisation was performed using low energy electron bombardment.

Flow and leakage tests were performed using custom computer controlled dual syringe pumps and pressure sensors. Flow rates from 10p1/min to 1 ml/min were used with back pressures up to 50psi. The 5pm silica particles used for packing the channel were purchased from Bangs Laboratories. Whole peripheral blood was collected by finger prick of a healthy Caucasian male, and solutions of methylene blue, triton x and phosphate buffer were all prepared in dionised water.

Electrosmotic flow was measured by filling a single channel with 25mM Phosphate buffer, emptying the reservoir at one end and replacing it with a 50mM solution of

the same buffer. A HP5r high voltage supply (Applied Kilovolts, UK) applied a potential of 200V/cm across the reservoirs, while the current was monitored across a 10k Ohm resistor between one of the reservoirs and ground using a AT-MIO- 16E-10 (National Instruments) data acquisition card. The current was monitored until it reached a plateau giving the time taken for the channel to fill with the less concentrated solution, dividing this by the length of the channel gave the electroosmotic velocity.

The filters were fabricated by excimer laser machining of 12pm PET film, with pore dimensions from 50pm down to 1 pm.

Fig. 2 shows a hole array of 1 Opm pitch with 8pm entrance and 1 pm exit holes.

For the filter to maintain a consistent pore size under increasing pressure the wall angles of the pores needed to be minimized to maintain good structural stability.

The fluence versus wall angle for excimer ablation has been well characterized and it has been shown that the higher fluences produce steeper wall angles. To this end the laser was operated in constant energy mode with a fluence of 1.5 J/cm2 at the workpiece, giving ablated wall angles of approximately 10 degrees.

The device was patterned using the 355nm YAG laser operating with a beam diameter of approximately 30pm, with a focal depth less than 100um. The cut width before bonding varied according to the substrate thickness, 1001um substrates gave a cut width of 40 5pm whereas the 12pm film gave a much larger cut width (120 20 pm) unless a fluid, such as isopropanol, was used to help dissipate the heat, giving a cut width much closer to the beam's diameter (27 2 pm). The chip layers were then cleaned with isopropanol in an ultrasonic bath removing most of the smaller pieces of ejected material that had redeposited around the cut, leaving only the larger fragments behind, as shown in Figure 3a).

The debri and ridges formed from the laser cutting process, which were as large as 201um high, were evenly compressed back into the bulk giving steeper wall angles and a more even surface (Figure 3b)).

The diffusion bonding method used to prototype these devices requires elevated temperature and pressure to allow the like substrates to diffuse into one another.

Therefore when bonding a multilayer structure like this it was necessary to use a

multistep bonding method to ensure that the membrane layer was fully sealed above the exit channel. In this case, layers 3 to 5 were firstly bonded together before layers 1 and 2 were added. Using the laser in the cross sectioning process can produce rough edges at the ends of the channel.

The cutting process increases the hydrocarbon component on the surface, decreasing the surface charge. In some applications this increase in hydrophobicity can aid in electrophoretic separations by reducing the electroosmotic flow and thereby helping to retain the sieving matrix and increase the separations. However, protein non-specific binding is increased with the more hydrophobic surface, and capillary action effects are also reduced making it difficult to fill micron sized channels with aqueous solutions. Therefore it is desirable in many biochip applications to increase the surface charge and hence decrease the polymer hydrophobicity. These charges can also be reacted with an amine, such as N-methyl-1, 3-propane diamine, to form amine terminated groups that can undergo further reaction to bind oligonucleotides to the surface for DNA hybridisation based experiments.

To increase the surface energy two techniques of surface modification were investigated; saponification or base hydrolysis, and UV exposure. The base hydrolysis was found to have an optimal exposure time after 1 min and gave a contact angle reduced from 75° to 16°, whereas the UV modification required 1hr to give a contact angle down to 25°. Furthermore the UV contact angle increased to around 35° upon washing, which may be attributed to the removal of low molecular weight fragments.

XPS surface analysis was performed on standard PET and modified PET samples with exposure times of 1 min for NaOH modified and 60mins for UV modified. Table 1 shows their relative percentage atomic concentrations, showing there is a significant reduction in the level of carbon and an increase in the oxygen concentration for both treated samples. As expected from the lower contact angle the chemically modified sample has significantly higher oxygen content in addition to a large amount of sodium contamination due to residual NaOH.

Table 1: Relative % atomic concentration of the carbon, oxygen and sodium elements present on the surface of base hydrolysed, UV modified and clean PET as determined by XPS.

Sample Relative % Atomic Concentration

O C Na Clean PET 26. 8 73. 2 UV treated PET 32. 9 67. 1- NaOH treated PET 37. 5 45. 1 16.4 A simple in-line filter design was used to demonstrate particle filtering using a multilayer laminate construction. The device fabricated, shown in figure 4b), provides multiple test structures, two of which have dual inlets for reagent delivery, an extended channel to allow for greater time for diffusive mixing, and a large filter chamber for particulate retention. The third structure provides two intersecting channels via a long narrow filter region with the same porosity as the other test structures. The device construction, as outlined in Fig 4a), consists of 5 layers of PET film ; a bottom sealing layer 21 of 100microns, an exit channel layer 22 Of 100microns, the filter layer 23 of 12 microns, the inlet channel layer 24 of 100microns, and a top-sealing layer 25 of 100microns with entrance and exit ports.

Figure 4 b) shows an assembled device with a close-up of the filter chamber.

To maximize flux for a given pore size the pitch needed to be minimized without reducing the membrane's strength to the point of failure. The filters were designed with 4pm exit holes at a pitch of 10pm giving rise to an effective porosity of approximately 13%. Reducing the pitch further resulted in the inlet of the pores overlapping, thinning the membrane and decreasing its strength greatly.

The filter's surface energy plays an important role when initially filling the device or when air bubbles are introduced into the system. The pressure required to initially pass the liquid through the membrane is the sum of the applied pressure and the effective pressure introduced from the surface tension If the membrane is more hydrophilic then small pores like those used here will fill by capillary pressure. For surfaces where the contact angle of the fluid with the surface is greater than zero an applied pressure is required to pass fluid through the pores. The membranes fabricated in unmodified PET with pore dimensions of 4pm gave a bubble pressure of 1.8 0.16 psi, which is below the calculated value

of 9.8 psi. This may be attributed to the combination of factors from the pore exits having slightly rounded geometries and varying in size, a variation of surface tension, and the bubble point pressure method being limited to measuring the largest pore hole present.

The electrosmotic flow velocities were measured for the untreated and chemically modified samples. In both cases the direction of flow was from the anode to the cathode, confirming the presence of negatively charged surface groups. The unmodified PET gave an electroosmotic mobility of 0.76 x 10-4 0.12 x 10-4 cm2/Vs and after a 5 min NaOH treatment this was increased to 1.2 x 10-4 0.14 x 10-4 cm2Ns. These values indicate that YAG laser cut channels in unmodified PET give a comparable result to embossed polycarbonate, as demonstrated by Liu et al who achieved mobilities of 0.7 x 10-4 cm2/Vs21. Likewise, the 3 hour UV treatment produced a similar result to the 5min base hydrolysis treatment with an increase in mobility to approx 1.25 x 10-4 cm2Ns.

Bead Trapping. Silica microspheres were introduced by pipetting an aqueous solution of 10% beads and 1% Triton X into the device and applying a vacuum (10psi) to the channel outlet. Under this vacuum the beads took approximately 30mins to pack the column tightly, leaving only a small trailing amount of loosely packed particles, as can be seen in Figure 5a) which shows a 225pm x 100pm packed channel. Upon introduction the beads immediately filled the length of the channel covering the filter, increasing the backpressure and reducing the flow rate.

The image in figure 5b), of a channel with the top layer removed, clearly shows the end of a column packed with the microspheres. Increasing the applied pressure speeds the packing process, however when using 5pm beads with only a 4pm porous filter then greater forces cause the microspheres to deform the membrane and pass through the filter. This was not a problem for larger bead to pore ratios.

White Blood Cell Filtering. The difficulty with whole blood filtration lies within the elasticity and structural strength of the biological cells. The Leukocytes, or white blood cells (WBC), we are isolating from peripheral human blood are highly deformable and range in diameter from 6-8 pm for the smaller lymphocytes to 12- 20pm for the monocytes, with the most abundant being the neutrophils that comprise approximately 60-75% of the WBC present and are around 10-15pm in size. Key to the successful isolation of DNA for PCR is the removal of the red

blood cells containing hemoglobin that inhibits the amplification process. Red blood cells are a relatively rigid discoidal shape 7pm in diameter and 2pm thick.

Blood filtration in silicon based devices has been demonstrated with weir and pillar filters, at a flow rate of 0. 035pl/min, and found that the red blood cells easily pass through gaps as small as 3pm, whereas the much larger and highly deformable white blood cells can pass through gaps as small as 7pm.

A mix of whole blood 2ul, in 20ul of 0.005% methylene blue (for nucleus staining) was injected into the device followed by a wash with 200ul PBS. The clear red blood cells and smaller stained white blood cells passed easily through the membrane with 4pm pores, leaving a cake of mostly stained leukocytes (Figure 6).

With the filter clogged the flow rate was greatly reduced using the unmodified PET membrane, and back-flushing met with only limited success.

Modifying the membrane with the 1min chemical treatment before the device was assembled produced a much more hydrophilic membrane that had less irreversible binding of particles to its surface and successfully allowed back-flushing to occur.

In forming polymer based microfluidic devices the use of web based manufacturing techniques to emboss or stamp the channels, laminate the layers is the key to cost savings. The present invention is entirely compatible with such web based manufacturing techniques as used in the. printing and packaging industries.

The advantages of this invention include: Compatible with polymer lamination technologies lower cost and ease of manufacture.

Can be made of same material as rest of substrate.

Can allow for greater flux than weir type filters for larger channel to bead geometries.

Those skilled in the art will realize that the principle of this invention can be expressed in a variety of embodiments without departing from the invention.