IN A LIQUID

The present invention relates to a method and apparatus for monitoring particles in a liquid. It is particularly concerned with the development of a direct observation technique as an online tool for the

visualisation of the transport of biopolymers or small particles during membrane filtration operations . It may offer real-time, dynamic assessment of problematic particles during filtration.

Membrane bioreactors (MBRs) represent the fastest growing advanced wastewater treatment technology with a global annual market value of around $1 bn. However, membrane energy demand has been constrained by the high particle concentrations promoted in MBR which require substantial energy input to minimise the formation of concentrated fouling layers at the membrane surface.

Membrane fouling impedes the passage of water through the membrane and demands large amounts of pumped air for control which constitutes one-third of the total process energy demand. The fouling layers gradually compress with time, exacerbating their tendency to restrict flow.

Controlling foulant layer formation demands an understanding of particle deposition kinetics, which is in turn linked to system hydrodynamics, bioreactor particle characteristics and, to a lesser extent, the membrane surface physiochemistry. Together, these parameters determine the structure of the foulant layer, and its potential ease of removal. The particles are heterogeneous both in size distribution and origin, but largely comprise (by weight or volume fraction) microbial cells formed into aggregates of >25 μm. However, studies determining the relative contribution of specific size fractions to membrane fouling have shown the sub-micron fraction (<1 um) , comprising protein and polysaccharide biopolymers, to contribute most to fouling. The preferential deposition of sub-micron biopolymers occurs because the large particles (>25 um) are readily transported away from the membrane by the hydrodynamic forces imposed to limit fouling, whereas the effect of the hydrodynamic shear force on smaller particles is limited. However, this generic perception of deposition cannot account for the complexity and heterogeneity of a particle suspension and the significance of the

hydrodynamic conditions on the transport of specific particle groups to the membrane surface. Understanding of these phenomena is limited, since current diagnostic methods do not allow a correlation between the particles in suspension and those deposited in the foulant layer.

To evaluate the contribution of the sub-micron foulant layer to fouling, invasive sampling methods are commonly employed which involve (i) physically scraping foulant from the membrane surface which compromise the physical structure, or (ii) extracting a sacrificial segment of membrane material to preserve the foulant layer for structural and compositional analysis, which necessarily compromises membrane integrity. The samples are then subject to advanced microscopic analysis which provides high resolution topographic description of the foulant layer or the speciation of biopolymers into proteins and polysaccharides. However, these techniques are limited by their requirement for dry conditions, which deforms the gel-like fouling layer. Consequently, the environmental relevance of the structural and biopolymer compositional data produced from these techniques is questionable.

Non-invasive direct observation (DO) methodologies have been developed which use a cross-flow filtration cell to control the hydrodynamic conditions at the membrane surface during foulant layer formation. Direct observation allows real-time measurement of foulant layer formation under representative conditions, critically without disturbing the deposit structure. Furthermore since the foulant layer is dynamic in nature, comprising both stagnant and fluidised regions, data from both areas can be captured in real-time to describe particle motion characteristics (back transport, diffusive, lateral migration) and under different hydrodynamic conditions

(i.e. effective shear) . The outcomes can subsequently be applied to model the specific resistance of the foulant layer formed.

Whilst DO methods permit the intricate measurement of bacterial cell mass transport to and from the membrane surface, the technique is currently only effective for suspensions of particles of diameter >3 μm. As such, DO methods are currently unable to detect those particles, namely the sub-micron biopolymers that represent the critical component in the formation of the most tenacious foulant layers .

According to the present invention, there is provided a method for monitoring particles in a liquid comprising providing a flow cell having a viewing window, providing reflected light fluorescence ("RLF") equipment including a source of exciting radiation and a detector, the RLF equipment being located in relation to the flow cell so that the exciting radiation can be passed into the cell through the viewing window, and fluorescence radiation can pass out of the cell through the viewing window and be received by the detector; passing a liquid containing particles through the flow cell while the source emits radiation capable of causing fluorescence of at least some of the particles, and using the detector to detect at least some of the resulting fluorescence radiation.

This method can enable the monitoring of particles of diameters below 3μm, even below lum (as well as larger particles) .

The method may employ a flow cell which includes a membrane extending in the main flow direction and separating a retentate flow path from a permeate flow path such that liquid tends to pass into the permeate flow path and to deposit particles on the membrane.

The retentate flow path may pass adjacent the viewing window.

The membrane may be a hollow fibre membrane.

The method may include monitoring of transmembrane pressure and/or permeate flux.

The RLF equipment preferably includes a microscope. This may be focused on or adjacent the surface of the membrane .

The method may include a preliminary step of coupling at least some particles to chromophores . There may be at least two chemically distinct types of particles which are coupled to respective different chromophores such that they are excitable by exciting radiation of respective different frequencies and/or produce fluorescence radiation with respective different frequencies. The chemically distinct types of particles may comprise (i) protein particles and (ii)

polysaccharide particles.

In another aspect the invention provides a method of operating a membrane bioreactor including monitoring particles generated thereby by means of a monitoring method as outlined above.

In a further aspect the invention provides a membrane bioreactor including apparatus for carrying out a monitoring method as outlined above, comprising said flow cell having a viewing window and said RLF equipment.

In a further aspect the invention provides apparatus for carrying out a monitoring method as outlined above comprising said flow cell and said RLF equipment which is located in relation to the flow cell so that the exciting radiation can be passed into the cell through the viewing window, and fluorescence radiation can pass out of the cell through the viewing window and be received by the detector.

A preferred embodiment of the invention may offer a new direct observation methodology which is capable of tracking sub-micron particles in real-time, on-line and dynamically during membrane filtration operation. Use of a microscopic technique employing reflected light fluorescence (RLF) may enable single particle resolution down to 0.10 um in diameter in static conditions and 0.25 μ in diameter in dynamic conditions . This makes it possible to capture the transport of the important sub- micron biopolymers range. An additional benefit of this technique is that through the use of reflected light fluorescence in parallel with dynamic observation, the sub-micron particles can also be speciated into proteine and polysaccharides, providing new insight into these polymeric groups .

This information is critical in both the

investigation of the underlying mechanisms of fouling and in the assessment of the effect of additives in the control of fouling where RLF-DVO can be used to diagnose additive dosing requirements necessary to improve fouling from both an operational and economic perspective when dosing at full scale (either on-site or in rhe

laboratory) and in testing the improved additive products for the amelioration of fouling.

Some embodiments of the invention will now be described in greater detail by way of example, with reference to the accompanying drawings. In these:

Fig. 1 is a schematic side view of an experimental set-up for carrying out a method embodying the invention for bench-scale testing.

Figs 2a to 2d are enlarged views of details: Fig. 2a is a top view of the viewing window region of a flow cell; Fig. 2b is a corresponding side view. Fig. 2c is similar to Fig. 2b but also shows a microscope lens and arrows indicating water flow. Fig. 2d is similar to Fig.

2b but the space above the hollow fibre membrane is exaggerated to facilitate the illustration of the forces on a particle.

Fig. 3 is a graph showing the relationship of crossflow velocity to particle concentrations for a range of different particle sizes.

Fig. 4 is a graph of particle velocity versus distance from the membrane over different filtration times . Fig. 5 is a graph of particle velocity versus distance from the membrane at different filtration times.

Fig. 6 is a graph of cake height versus filtration time .

Fig. 1 is a schematic representation of a reflected light fluorescence - direct observation flow cell set-up. This includes a flow cell 10 traversed by a tubular hollow fibre membrane 12. The flow cell 10 has a viewing window 14 in its upper face. There is a flow path through the cell 10 (via inlet 15 and outlet 15)

externally of the hollow membranes (i.e. on the retentate side) . There is also a flow outlet 18 from the

membrane's interior lumen. In this example, the system is fed by a 200 mL solution pumped through the retentate side of the cross flow cell using a peristaltic feed pump 20. The hollow fibre membrane (0.04 urn nominal pore size, Zeeweed, GE Power and Water) is mounted in the flow cell. Permeate (effluent) is extracted from the membrane using a peristaltic permeate pump 22 which delivers the permeate to a container 24 placed on a balance 26 for accurate control of permeate flux. The flow cell 10 is shown in more detail in Fig. 2. It is constructed of Perspex with an engineered viewing window 14 crafted into the top of the cell that enables high resolution imaging to be undertaken. Typical flow cells used in the direct observation technique are exclusively made of Perspex as the limit of detection for conventional techniques is strongly dependent on visible light penetration into the cell. However, the material thickness required to support a channel of sufficient mechanical integrity which can further provide an ppropriate environment for a range of cross flow velocity and shear conditions, generally dictates a reasonably thick Perapex

construction, which limits resolution. In this

invention, this limit to resolution is remedied by: (i) correct sizing of lens and window arrangement; and (ii) dependency on fluorescence rather background light to enable high resolution (discussed later) . Of course, since we use reflected light fluorescence ( LF) , we do not actually rely on penetration of light through the cell walls, so we could use cells of materials other than Perspex, e.g. of aluminium or other metal.

The window 14 is in a recess in the upper wall of the cell. An opening 30 is surrounded by a resilient 0- ring 32 to provide a watertight seal to a coverslip 34 which is held down by an annular clamp 36, suitably of aluminium.

The viewing window area is specifically engineered to: (i) enable direct contact {if needed) between a microscope lens 38 and the window; and (ii) enable sufficient strength to withstand the shear created by fluid pumping within the cell whilst keeping the viewing window sufficiently thin to provide high resolution visualisation. To resolve these issues: (i) , the recessed section is included in the Perspex 'top-side' with a diameter slightly wider than the microscope lens to ensure that the microscope lens can move to within a close range of the viewing window without forcing the full construction to be made of a thin (and therefore less mechanically strong) Perspex layer. Viewing windows of various thickness were tested for durability. The selected window (or coverslip) is sealed into the base of the recessed section and secured with an O-ring 32 and thin aluminium coverslip clamp (less than 1mm thick) . The flow cell is designed to accept hollow fibre

membranes interchangeably (although other membranes, e.g. flat sheet membranes, could also be studied under LF- DVO) . To date we have tested hollow fibre membranes 40 with an outside diameter of 0.0019 m, yielding an active membrane surface area of 0.00125 m ² .

As an example, in the lens and window arrangement used, a lens is selected (lens used, HC PL FL 10X/0.30) that provides a wide range a free working distance

(around 11mm) sufficient to incorporate the depth of field below the viewing window to the membrane surface (the surface of study) . The active region of interest (i.e. the fluidised region for study) must work within this range. The viewing window is constructed of a n = 1.5 coverslip (around 0.17mm thickness) which is

sufficiently thin that the impact on resolution is minimal, but sufficiently strong to withstand the forces associated with fluid pumping. This coverslip thickness provides resolution close to what can be obtained with an optical microscope in static mode. In this example, from the outer diameter of the hollow fibre membrane to the lower side of the viewing window is a distance (of moving fluid) of 2 to 3 mm which is sufficient to allow active fluid transport around the fibre. The total working distance described is then around 6.17 mm which includes some freeboard at the lens, glass window thickness and water depth above the membrane fibre - and clearly falls within the theorised working range of the lens.

The flow cell operates in analogous fashion to full- scale membrane operation whereby there is a driving force facilitating water permeation through the membrane to deliver clean water. This induces convective drag on the particle and so a cross-flow is introduced to impose lateral migration effects sufficient to counter-balance this effect and provide greater particle stability in the flow (i.e. to limit particle deposition on the membrane surface) . This is illustrated schematically in Fig. 2d. Good resolution can only be obtained when the

hydrodynamic environment is carefully controlled. This is facilitated by selecting appropriate cell dimensions (channel width of 0.012 m, length of 0.21 m and height of 0.006 m) and shape (rectangular with tapered ends for distribution) . These example dimensions provide control of both permeate and cross-flow conditions to within environmentally relevant hydrodynamic conditions and have been shown to enable the necessary resolution. For the assessment of real matrices, fluorescent microspheres can be included to act as a robust indicator of fouling, with enhanced resolution provided by their fluorescent capability. Key parameters that can be varied include the cross flow velocity on the retentate side of the fibre (the dimensions of the cell will affect the shear stress applied to the fluid) and the permeation rate through the fibre wall. These conditions can be

adjusted, depending on the shear stress wanted and the size of the particles of interest.

Optical microscopes use visible light, e.g. to analyse immobilised samples on glass slides. However, for more complex dynamic systems, this method is not powerful enough: (a) to provide refined resolution for small particles; and (b) to enable resolution for small particles that are in dynamic motion. For example, visible light (sometimes supported with additional

UV/light sources) is ueed with existing direct observation methods. However: (i) the thickness of the Perspex construction is such that insufficient light penetrates the Perspex material, limiting resolution of bodies that are observed; and (ii) visible light is often insufficient for discrimination of particles due to the limiting wavelength of visible light. Fluorescence microscopy, on the other hand, can use specific filters to provide light at a specific excitation wavelength. The light is absorbed by a particle (which, may have been labelled with a target specific fluorochrome) , and then emits light at a longer specific wavelength. This can enable recognition of the small particles and biopolymers present in solution. Fluorescence can therefore give enhanced resolution through two mechanisms: (a) the addition of specific fluorochromes to label and target specific particles of interest; and (b) the addition to the system of particles of known size and density which are already fluorescence enabled.

The example particle suspension, a real biological matrix that is to be filtered to separate clean water from the particles, can be broadly characterised as comprising bacteria (either as independent entities or aggregated) and colloidal compounds including the key groups : proteins and polysaccharides . In order to enable resolution of these compounds and also to enable

differentiation, it is necessary to first label the compounds. This can be achieved in-situ or ex-situ since binding occurs rapidly. Fluorochromes (the labelling agents) are selected according to their chemical

affinities with the biopolymers. For example, FITC (excitation/emission ("ex/em") = 490/525 nm) and

calcofluor (ex/em = 440/510 nm) can be used as fluorochromes specific for the labelling and

differentiation of proteins and a-polysaccharides . To differentiate broader characteristics of the

polysaccharide group, conA-TRITC (ex/em = 555/580 nm) can also be included as this can enable one to distinguish a- from ^-polysaccharides . In this method, appropriate filter cubes (and sufficient filter cubes) must then be provided for the microscope that are capable of

excitation/emission at appropriate wavelengths

corresponding to those of the selected fluorochromes .

The subsequent differentiation is given by differences in colour emitted (green, blue and red respectively) that can be seen at the microscope either in isolation or collectively by the overlaying of images. The intensity of light provided at each specific wavelength must be selected to match the required live feed resolution (commonly quantified in frames per second) .

This use of fluorochromes further enables the visualisation of particles of a small size that would be unobservable using standard light methods - particularly in real matrices which can involve high particle numbers and have proven problematic in previous direct

observation studies using visual light. This introduces a secondary application of the RLF-DVO technique as a diagnostic method in which a particle of known particle shape, diameter, charge, density and excitation/emission (e.g. a fluorochrome enabled latex microsphere,

manufactured to certified standards) is added to a real matrix to ascertain hydrodynamic behaviour within the real matrix. The advantages are: (i) the expected hydrodynamic behaviour of this particle can be absolutely described; (ii) high particle recognition can be ascertained in a highly concentrated particle matrix. The advantage of this method of analysis could be, for example, to ascertain the influence of a change to matrix behaviour by the inclusion of a chemical modification to the particle matrix such as a coagulant or descalant.

The system can be used for the direct visual observation of transport and deposition kinetics of sub- micron (and greater) particles, (as well as for

determination of the mass deposited on the membrane and/or the thickness of the foulant layer. The system may also be equipped with sensors (P2, P2, P3) to monitor pressure and permeate flux at various positions in the system. This can be used for direct measurement of permeability and transmembrane pressure (TMP) as common surrogates of fouling.

Experimental Results

Using the illustrated apparatus, fouling experiments were conducted for around 3h under fixed hydrodynamic conditions, for example, at liquid velocities in the retentate channel of 0.01 m s ^-1 and a fixed flux of 75 L m ^-2 h ^-1 . Transmembrane pressure was determined using pressure transducers P1,P2 placed upstream and downstream of the retentate channel and a vacuum pressure transducer P3 on the permeate side. Both retentate and permeate were fed back to the feed reservoir to keep a constant volume throughout fouling experiments . The focus of the microscope was made on the surface of the membrane and the above fluidised region to capture both the deposition of particles on the membrane and their transport in the bulk fluid. Photos and videos were recorded at 15 min intervals using the software Leica Application Suite (Leica Microsystems). The instrument's resolution was characterised using fluorochrome doped latex microspheres (ex/em = 468/508 nm) with a strictly controlled particle diameter and shape.

Latex microspheres ranging in diameter from 0.10 μm to 10 μm were tested in both static and dynamic

conditions. Under static conditions individual particles of all sizes could be distinguished. Method resolution was then tested under dynamic conditions and was found to be dependent upon particle concentration and liquid velocity (see Figure 3) . At lower concentrations, the fluorochrome doped latex microspheres were clearly visible as individual entities. This is particularly advantageous for use of the fluorochrome doped latex microspheres as a surrogate particle tracker in real matrices as only small volumes are required to ascertain system fouling.

Resolution of fluorochrome labelled biopolymers was also tested. Under the same hydrodynamic conditions (whilst also controlling particle-particle interactions through solution chemistry) , both protein and

polysaccharide biopolymer molecules were distinguishable as individual entities. Previous DVO techniques have not been able to speciate biopolymers into the groups of interest (critically in the sub-micron region these are generally observed to be protein and polysaccharide) . In the following example, in a complex mixture comprised of 50 mg 1 ^-1 of polysaccharide and 50 mg 1 ^-1 of protein, both biopolymers are detected using the RLF-DO technique. To be fluorescently detected, biopolymers have to be stained with appropriate fluorochromes . For instance, FITC, calcofluor white and conA-TRITC can be used to stain proteins, α-polysaccharides and β-polysaccharides respectively which by using the appropriate filters will be seen green, blue and red. Furthermore, the images are generated from two separate filter cubes . When these images are overlaid, the proximity of these biopolymers is evidenced and identified that the protein and

polysaccharide have formed aggregates through attractive interaction. These interactions within the bulk flow have not previously been seen before and provide

significant new insight into particle-particle behaviour before fouling commences. The same staining procedure can be used with real wastewater where the fluorochromes will only stain the biopolymer of interest. This

evidences that both of these important groups can be characterised and discriminated in real-time. This provides one of the first opportunities to characterise the particle deposit formed for protein and

polysaccharide content at the membrane during filtration without the need for invasive sampling techniques which are known to disturb samples and create significant measurement error.

Only direct visual observation methods have the ability to discriminate and quantify particle back transport velocity. However, no previous method has been able to measure back transport velocity of particles with diameters below 3μ . In an experiment, the back

transport of sub-micron particles was measured during dynamic cross-flow conditions, both next to and within the fluidised boundary layer. This is important as: (i) it is the sub-micron particle size range which largely controls fouling under cross-flow conditions; and (ii) it is from within this concentrated fluidised region of particles where particles subsequently arrive at the surface, causing surface fouling. As shown in Figure 4 , a monodispersion of sub-micron particles was used. It is evident that in this monodispersion, particle velocity near to the surface does not change - these conditions did not promote significant fouling.

Fig. 5 shows the results of an experiment in which a complex particle mixture was used to test particle back transport measurement. This time series experiment clearly indicates that, following a period of filtration (from 10 to 180 minutes), the particle back transport velocity will decrease in this concentrated fluidised layer. This is important as it is the force balance on the particle which keeps the particle fluidised.

Consequently, as particle velocity declines, there is a greater risk of surface deposition - or membrane fouling. Such measurements can therefore be used to predict incipient membrane fouling. This information from complex mixtures in critical in understanding: (i) the fouling mechanism; (ii) in establishing what impact additives have on particles kinetics during fouling; and (iii) in providing information which would enable an appropriate additive dose to be selected to ameliorate fouling. This progressive reduction in particle velocity is indicative of the onset of fouling following

concentration polarisation. Therefore, another example of the techniques utility in process control is where tracker particles (with fluorochrome embedded) can be monitored in the dynamic boundary layer to signal the onset of fouling before tenacious fouling is initiated.

Other in-situ measurements that can be facilitated by this method include cake height which provides a measure of particle accumulation at the membrane surface (Figure 6) . The resulting data can provide a direct measure of fouling progression and can be used as input data into supportive complex modelling to determine parameters regarding foulant layer properties such as porosity and specific resistance. These measurements are complementary to conventional DVO, however, further speciation of the developed cake into protein and polysaccharide composition is possible through the use of fluorescence.