Lab On a Chip System for Size-Based Gold and Polymer Nanoparticle Separation

Despite the increasing demand for nanoscale biomolecule analysis for point-of-care (POC) application, nanoparticle separation remains a challenge in many applications due to huge sample loss during separation, low throughput, large scale input materials requirement, and sophisticated technologies. As the separation eciency may affect the subsequent sample processing and analysis, a robust and reliable size-based separation technique is necessary. This study presents a lab on a chip system to enhance the separation performance by using rapid and straightforward polymer prototyping. In particular, the system consists of a microuidic network with embedded membrane lters with different pore size cut-offs and an ultrasonic transmitter for acoustic agitation. Using the novel system, we successfully demonstrate the fractionation of 15 nm Au NP from polydisperse nanoparticle solution in the presence of ultrasonic wave (28-40 kHz) generated by the transducer incorporated with the microuidic system during the separation. Ultrasonic irradiation helps in preventing cake formation and reversing the fouling process by acoustic agitation. The suggested system signicantly increases the ow rate during the separation process and improves the recovery of target size nanoparticles. This microuidic platform is expected to serve as a powerful tool for sample preparation and analytical methodology in POC applications. Twenty microliters of gold nanoparticle stock with 80 µL 0.1% F108 solution and placed into a loading reservoir. Filtration was performed with ultrasonic irradiation to investigate the eciency of ultrasonic treatment, and the UV-Vis spectroscopy results were compared with the results of ltration without irradiation. To obtain a monodisperse target size, maximizing the recovery rate while no detectable particles from the retentate solution are essential. Recovery rates and loss were calculated through normalization by the absorbance prole of the Signicant sample loss was observed after the ltration when ultrasonic irradiation was not performed. It to the adsorption of onto the membrane surface during the Peak indicates that recovered after yielding a prolonged recovery rate. From the peak absorbance of retentate, we assumed 17.9% of input could not penetrate the and the were by manual On the other hand, with ultrasonic a signicant enhancement in recovery rate was observed. The absorbance of the retentate few 60 nm were recovered in the It achieved a 70.6% recovery rate introducing ultrasonic treatment during the separation on a microuidic is 3.9


Introduction
With the growth in micro uidic technologies in the COVID-19 era, a substantial advance in the Lab-on-a-Chip (LOC) system has been presented, enabling miniaturization and integration of complex functions routinely performed by hand in the traditional analytical processes [1][2][3] . One of the most active applications in the LOC is observed in biomedical research with Point-of-care (POC) diagnostic devices. The POC diagnostic system can rapidly deliver diagnostic outcomes to patients providing the users with many bene ts over the traditional diagnostic methods such as straightforward and easy-to-use interface, portability, less sample and reagent consumption, etc 4 .
In biomedical research, size-based particle separation plays a signi cant role in the sample preparation process and the detection of target species 5,6 . As the size of analytes in biological research range from angstrom to millimetre scale, the separation mechanism of choice depends on their dimensions. During the past decade, many size-based separation techniques have been established for biomedical applications 5,7 . Enormous progress has been made, and size-based separation technologies move towards the miniaturized platforms to meet the requirements for point of care testing 8- 10 . Many detection techniques require isolating target species of interest from solutions during sample preparation to provide unbiased and accurate insights into the biological process or phenomena. Since the dimension of biological structures ranges from angstrom to millimetre scale, the separation method varies depending on the target size due to the growing need for robust and reproducible size-based particle separation techniques. Especially, signi cant efforts have been put into developing a miniaturized size-selective particle sorting method for point of care applications such as LOC particle sorters 8- 10 . However, the LOC based separation techniques are limited to the sub-micron scale (such as cells, viruses, bacteria), which results in signi cant sample loss, diagnostics misinterpretation, and poor detection limit [11][12][13][14] . Developing robust and reliable size-based particle separation techniques is emerging as an essential prerequisite for nanoscale measurement to obtain insights into the various biological systems on a nanometer scale.
For nanoscale biomolecule sorting, biomedical research has carried out several size-dependent nanoparticle separation techniques such as technology-based on external elds application, physical sieving and size-selective precipitation. Unfortunately, commonly used current technologies for NP separation were still not su ciently e cient in selectively sorting particles based on their size or molecular weight for proper POC use due to their complex hands-on processes and incompatibility with downstream analytical devices. A miniaturized, compact, cost-effective platform for size-based separation is more favourable than one in centralized laboratories to eliminate the need for bulky and sophisticated instruments and extensive input material and minimize the complexity of the device operation.
For nanoscale biomolecule separation, several size-dependent nanoparticle separation techniques have been employed. Based on the separation mechanism, these techniques fall into three categories: the use of external elds (centrifugal [15][16][17] , electrical 18-20 and magnetic 21,22 ), physical barriers for sieving 23 , and size-dependent precipitation 24 . One of the most widely used methods based on the application of external elds is eld ow fractionation (FFF). FFF is a particle separation technique taking advantage of the eld exerted to a solution owing through a channel. The eld is perpendicular to the uid ow, which results in the separation of particles suspended in the carrier solution due to the difference in travelling speed depending on their size and molecular weight. Despite its variations and versatility, it has not been widely used as a size based nanoparticle separation method due to its low throughput and extensive input material required. Alternatively, a size-dependent precipitation method has been devised. Sizedependent precipitation occurs when the stability, physical or chemical properties change depending on the surface chemistry of the nanoparticles. Sieving is another alternative for the size-selective fractionation of nanoparticles. Chromatography and membrane ltration employs physical barriers such as a column or physical hole to control elution and retention depending on the size of particles.
Unfortunately, these traditional size based nanoparticle separation techniques result in huge sample loss, therefore, large input material is required. In addition, bulky instrumentation is involved in generating external eld during the separation process. Due to the complexity of these technologies, there has been increasing demand for the development of miniaturized and compact platform for size-based separation, which can be used as a cost-effective preparative method in point of care application.
There have been many micro uidic approaches to bring the size-selective nanoparticle separation into miniaturized platforms. One of the most common separation mechanisms utilized in the micro uidic platform is membrane-based ltration because of its high resolution and straightforward procedure [25][26][27][28] .
Gaborski et al. 23 demonstrated nanoparticle fractionation using porous nanocrystalline silicon membranes with resolution as high as 5 nm, potentially applicable to Lab-on-a-Chip system for dead-end ltration. However, membrane ltration typically suffers from low e ciency associated with clogging and formation of cake layer near membrane lter 29 , leading to low ow rate, sample loss, and high uid resistance 30 . Also, the fabrication of nanopores involves a complicated and time-consuming microfabrication process due to nanoscale particle separation. To circumvent this limitation, " lter-free" micro uidic separation techniques have been developed. Micro uidic approaches based on electrophoresis, eld-ow fractionation (FFF) 31,32 and centrifugation 15,17 have been devised and implemented for size-based nanoparticle separation and fractionation. However, these techniques still rely on external elds which usually need the bulky instrument to manipulate particle motion.
Furthermore, most passive separation techniques require expensive fabrication to build microstructures. For a point-of-care application, lab on a chip system that can be realized with a simple fabrication process, providing a user-friendly interface enabling easy sample handling, yet with high resolution with minor sample loss is necessary. For these reasons, the lack of reliable size-selective nanoparticle separation is a bottleneck in analyses of nanoscale biomolecules such as ribosome pro ling, which requires size-selective separation of monosome from polysome molecules 33 .
As membrane ltration can provide the most straightforward processing and high resolution, it has many advantages over lter-free alternatives. Hence it can be an excellent size-selective nanoparticle separation tool as long as the problems associated with membrane fouling during the process is addressed.As a part of its effort, many publications and research works have improved membrane ltration by utilizing ultrasonic treatment of membrane lters. However, most results focused on bulk processing for industrial use, such as wastewater processing, food processing, and ultrapure water production 34,35 . Typically, they employed ultrasonic bath, which requires high energy consumption and extensive input material, and it may not be suitable for biological applications where available input material is low.
To realize simple and rapid size-selective nanoparticle separation for POC application, we have developed a LOC system based on dead-end ltration by incorporating commercially available track-etched membrane lters with different pore sizes cut-off. Sample containing polydisperse nanoparticles has been processed in the micro uidic platform with embedded membrane lters. The dead-end ltration may cause clogging and membrane fouling due to the formation of a cake layer near the membrane surface, leading to low ow rate and membrane and sample loss damage. We have introduced ultrasonic treatment by integrating ultrasonic transmitters with a membrane ltration system online to overcome this issue. Ultrasonic assisted membrane ltration was modi ed to bring it into a miniaturized micro uidic platform. The functionality of this novel LOC based technology was con rmed by demonstrating size based gold and polymer nanoparticle separation.
Materials And Methods

Basic principles of dead-end membrane ltration
Typically, the input uid ow in dead-end membrane ltration is perpendicular to the membrane surface, called the "feed" solution. Driven by pressure in the ltration system along the channel, a fraction of the uid passes through the membrane pores, called " ltrate" or sometimes "permeate". Moreover, a fraction of the input remains on the membrane surface, called "retentate". To quantify membrane-based ltration performance, permeates and retentates are collected separately and investigated by various analysis methods. When the targeted nanoparticle is a metal such as gold or silver, absorbance measurement is one way to quantify the performance. For small gold nanoparticles (~30 nm), the surface plasmon resonance effect induces the absorption maximum at 526 nm wavelength. For relatively large gold nanoparticles (60 nm), the absorption maximum is generated at 540 nm wavelength. Hence, the UV-Vis spectrometer is a valuable tool for identifying the size of the gold nanoparticles. Measuring the absorption from the collected ltrate and retentate samples can quantify the fraction of target size nanoparticles in each sample. From the absorption measurement data, the performance of the dead-end membrane ltration system can be determined.

Working principle of ultrasonic-assisted nanoparticle separation
The working principle of ultrasonic-assisted nanoparticle separation on a micro uidic device is illustrated in Fig. 1. The separation mechanism is straightforward using dead-end ltration. The micro uidic system for size-based nanoparticle separation consists of a micro uidic device with track-etched polycarbonate membrane lters with sharp pore size cut-off incorporated with an ultrasonic transducer. To isolate monodisperse nanoparticles from the mixture, membrane lters with a pore size larger and smaller than the particle size were used. As shown in Fig. 1, target (Green) particles penetrate the membrane with a greater cut-off value while larger particles settle onto the membrane. A membrane with a smaller cut-off was introduced to capture target particles enabling puri cation before collecting target particles.
Impurities and particles smaller than the target ow through the second membrane lter and proceed to the waste reservoir.
One of the critical limitations in dead-end ltration is sample loss and decrease in permeability during the separation process due to particle adsorption on the membrane surface and the formation of a cake layer near the surface. An ultrasonic wave was exploited to detach fouled particles on the surface and enhance the target recovery of particles. (Fig. 2) The transducer was located near the membranes for e cient ultrasound irradiation to minimize the attenuation as the wave propagates through the material. And the chamber was designed to t the active area of the transducer to maximize the coverage of ultrasound irradiation. Negative pressure was applied through the outlet to prevent leakage at the interface of the PDMS layer and the membranes.

Chip design and fabrication
The micro uidic nanoparticle separation device prototype was designed using AutoCAD software and constructed by a standard soft lithography process. The device consists of two circular reservoirs connected by a micro uidic network to ltrate nanoparticles, as illustrated in Fig. 3. The micro uidic network patterned onto the bottom layer, which connects the two reservoirs. The side channels connected to the reservoir for sample collection are patterned onto the middle layer. And the top layer serves as a top cover to form a closed micro uidic network while the left reservoir remains open for sample loading. The right reservoir has micro uidic barrier structures to guide the uid ow uniformly to cover the entire surface of the membrane, and it enables the even distribution of the particles to in ltrate the membrane surface.
The left reservoir serves as a sample loading well where a mixture of nanoparticles was loaded. The membrane lters with different pore sizes, including 15 nm and 30 nm, were integrated with three PDMS layers in the 20 nm NP separation process. First, the structures of each PDMS layer were patterned onto a transparency mask and supplied by Fineline Imaging (Colorado Springs, CO) at 32000 dpi resolution. To fabricate a master mould, SU-8 2025 was spun onto a silicon substrate using a spin coater at 2000 rpm for 30 sec to achieve a uniform thickness of 40 um. Then the master went through a soft bake process to evaporate solvent at 65 °C for 2 min and 95°C for 5 min followed by UV exposure (200 mJ/cm 2 ) for 30 sec. For Post-exposure bake (PEB), the master was cured at 65 min for 1 min and 95 min 3 min. Then the pattern was developed by immersing the master in the developer solution for 5 min and rinsing with Deionized water and isopropyl alcohol.
Excess water must be removed in a spin dryer. After the patterning process, the SU-8 mould was modi ed with trimethylchlorosilane (Sigma-Aldrich, USA) under vacuum. Next, the PDMS layers were fabricated by standard soft lithographic techniques. Brie y, PDMS was mixed with a curing agent at a ratio of 10:1, and it poured onto the SU-8 master mould to form a replica of the device. Degassing was performed to remove entrapped bubbles from the layer. Then the layer was baked in a convection oven at 65 °C for an hour for curing and peeled off from the master mould. Two reservoir regions were punched with biopsy punches, and holes at the inlet and outlet were punched for uidic access.
Membrane lters were cut using the laser cutting process into the desired size. Pore sizes of membrane lters varied depending on the target nanoparticle size. Uncured elastomer/curing agent mixture was applied along the reservoir's edge to integrate the membrane lter on the PDMS bottom layer. Then the membrane was aligned, placed at the centre of the reservoir, and pressed gently with a round tip tweezer to absorb the uncured mixture into the lter layer. The bottom layer and membrane lter were cured in an oven at 65 °C for at least 30 min. Three PDMS layers were oxygen plasma treated to activate the surface, aligned and brought into contact with each other to form a micro uidic platform. Tygon tubings were inserted into through holes at inlets and outlet regions. After integration and assembly, no signi cant leak was observed during operation. SU-8 master mould can be reused after washing with IPA and deionized water. The micro uidic device was coupled with ultrasonic transducers by using silicone at the interface of the transducer and PDMS layer to minimize the energy loss and attenuation during operation. 4. Ultrasonic assisted nanoparticle separation on a micro uidic device Before separation, the micro uidic network was rinsed with 0.1% Poly(ethylene glycol)-blockpoly(propylene glycol)-block-poly(ethylene glycol) (Pluronic F108)(Sigma-Aldrich) solution to remove residual impurities and treat the uidic PDMS surface, thereby preventing adsorption of nanoparticle on the membrane surface or uidic channel wall and formation of fouling on the membrane surface. Gold nanoparticles (Nanocs, NY, USA) with a varied diameter (15,30,50,80,100,200, 400, 600 nm) were used to demonstrate ultrasonic-assisted nanoparticle separation on micro uidic devices. In the case of 60 nm/ 100 nm mixture, 10 µl of the stock solution of 60 and 100 nm (1.9 10 9 / mL and 3.8 10 9 / mL respectively) Au nanoparticles in 0.1mM PBS were added into 1 PBS, and the nal volume was brought up to 100 µl. The mixture was loaded into a sample reservoir of an assembled micro uidic device with a membrane lter for separation. (Fig. 4) The mixture was driven into the micro uidic device by the negative pressure generated by the syringe pump operating on "withdraw" mode. Flow velocity was set to 0.1 mL/h. The micro uidic device was irradiated with ultrasonic waves generated by the ultrasonic transducer during the separation. Typically, a 40 kHz square wave was developed by a function generator and 50 times ampli ed through a high voltage ampli er (WMA-300, Falco systems) and used to excite the ultrasonic transducer. During the separation process, other inlets and outlets were closed than the waste port. Before collecting target particles, 0.1 % luronic F108 solution was injected at 0.1 mL/h for 10 min as a washing buffer. Backwashing with the ultrasonic wave was performed to retrieve 60 nm gold nanoparticle (target) from the chamber region between two membrane lters. 60 nm gold nanoparticles (Au NP) were collected in either 0.1 % 108 solution or PBS for downstream analysis. The retentate settled on the rst membrane was collected by pipetting up and down several times for measurement.

Characterization of Populations of Nanoparticles
UV-Vis spectroscopy was used to characterize the populations of nanoparticles before and after the ltration process by quantifying the absorption intensity of each solution. Populations of nanoparticles were characterized by measuring peak absorbance and intensity wavelengths on a multimode plate reader (Perkin Elmer). The wavelengths of peak absorbance depend on the sizes of gold nanoparticles.
Typically, peak absorbance occurs within 564 -574 nm and 538 -544 nm for 100 nm and 60 nm Au NP. Samples (Filtrates, retentates) collected from the micro uidic device after separation were analyzed within the 400 -700 nm wavelength range, and the results were compared with that of the stock solution. Based on concentration and volume loaded, the recovery rate and sample loss were calculated.
To validate the ltration e ciency uorescence image was obtained using a uorescent microscope. 350 and 450 nm uorescently labelled polystyrene nanoparticles were ltered on a micro uidic device incorporated with a 400 nm pore membrane lter to visualize NP adsorbed onto the membrane surface. The membrane was taken out of the device and investigated under a microscope. The 450 nm particles settled onto the membrane were considered sample loss, and the number of particles was counted automatically using Image J software.
In addition, particle size distribution was also determined by dynamic light scattering (DLS) using isolated ltrate and retentate prepared in 0.1% F108 solution. The distribution data were acquired and × × × compared with the distribution data of each stock solution. All experiments were carried out at least 3 times.
Results & Discussion

Characterization of micro uidic size based NP separation with ultrasonic irradiation
The e ciency of nanoparticle ltration using the micro uidic system with ultrasonic irradiation was evaluated using monodisperse 60 nm gold nanoparticles and a single track-etched polycarbonate membrane lter with 80 nm pore size cut-off. (Fig. 5) Twenty microliters of gold nanoparticle stock solution were diluted with 80 µL 0.1% F108 solution and placed into a loading reservoir. Filtration was performed with ultrasonic irradiation to investigate the e ciency of ultrasonic treatment, and the UV-Vis spectroscopy results were compared with the results of ltration without irradiation. To obtain a monodisperse target size, maximizing the recovery rate while no detectable particles from the retentate solution are essential. Recovery rates and loss were calculated through normalization by the absorbance pro le of the feed solution. Signi cant sample loss was observed after the ltration when ultrasonic irradiation was not performed. It is likely due to the adsorption of particles onto the membrane surface during the process. Peak absorbance of ltrate indicates that 21% of input was recovered after ltration, yielding a prolonged recovery rate. From the peak absorbance of retentate, we assumed 17.9% of input could not penetrate the membrane, and the particles were retrieved by manual pipetting. On the other hand, with ultrasonic irradiation, a signi cant enhancement in recovery rate was observed. The absorbance of the retentate solution was negligible, indicating few 60 nm particles were recovered in the solution. It achieved a 70.6% recovery rate introducing ultrasonic treatment during the separation on a micro uidic device, which is 3.9 times higher than without ultrasonic treatment.

Effect of surface treatment
In the initial attempt to determine the ltration e ciency of the micro uidic device, it was found that the ltration e ciency is greatly dependent on the membrane surface condition. Once the volume loaded into the sample reservoir passed through the membrane, and the membrane surface started to dry, Au NP settled on the membrane surface were compacted. And the foulants formed a cake layer, making it challenging to re-suspend particles from the membrane surface for downstream analysis regardless of vigorous pipetting. The membrane lter surface must be kept wet during and after the ltration to maximize e ciency and recovery. Monodisperse uorescently labelled polystyrene (PS) NPs demonstrated this phenomenon, visualizing it using a uorescent microscope. The hydraulic diameter of NP was 450 nm, and the pore size cut-off of the membrane lter used for ltration was 400 nm. Fig. 6 shows the comparison of removal rate between the cases with and without ultrasonic treatment, either on wet or dry surfaces. Fig. 6-1 showed the 450 nm particles on the 400 nm lter surface when 100 µl of sample passed through the membrane. Under wet surface conditions (Fig. 6-2 and 3), the effectiveness of ultrasonic cleaning was examined. As a result, most input particles were successfully removed from the lter surface by both back ushing or pipetting. After ltration without ultrasonic treatment, the loss was 5%, while no detectable quantities of nanoparticles were found from the surface after ultrasonic cleaning. This result indicates that nanoparticles largerthan the pore size of the membrane lter tend to stay in suspension under the wet surface condition. The cake layer formed during the ltration is easier to remove from the membrane surface during the cleaning process combined with ultrasonic irradiation. In contrast to a wet surface, severe fouling of membrane lter was observed once the membrane surface was dry, indicating the irreversible formation of cake layer with foulants on the surface under dry surface conditions. (Fig. 6-4 and 5) Ultrasonic irradiation exhibited an improvement in removal rate. However, signi cant losses were detected from both cases. As a result, 65% of input particles were successfully removed from the surface without ultrasonic cleaning, while 70% of input were removed by ultrasonic cleaning. Hence, we assumed that 30% of nanoparticles larger than pore size were presumably adsorbed on membrane surface irreversibly under dry surface conditions. Therefore, the membrane lter surface needs to remain wet during and after ltration before recovering target particles, and ultrasonic cleaning enhances the removal of particles from the surface.
To examine the utility of the micro uidic device in size based nanoparticle sorting, ltration of 100 / 400 nm NP mixture was performed, and the particle size distribution was investigated by dynamic light scattering (DLS) technique. (Fig. 7) A mix of two different sized polystyrene particles was driven into the micro uidic device and pressurized to initiate size based ltration. For 100 / 400 nm separation, 80 nm and 200 nm track-etched membrane lter was selected for separation. The ltrate was collected from the reservoir by back ushing with ultrasonic treatment. Fig. 7 shows particle size distribution pro les of 100 nm and 400 nm particle stock solution, ltrate and mixture. Size distribution of the polydisperse mixture was examined before ltration. As a result, the ltrate showed a size distribution pro le similar to the original 100 nm stock solution without a detectable shift in peak wavelength and spectrum broadening.
We compared the size distribution pro le measured by the light scattering technique to the size estimated by absorbance spectra. As a result, peak absorbance of the ltrate was ~70% of stock solution with slight variation in peak wavelength, indicating successful size discrimination and a high recovery rate.

Effect of orientation of the ultrasonic treatment
The effect of orientation of ultrasonic transmitter on the performance of the device in particle fractionation and recovery rate was investigated using 60 nm Au NP. Particle fractionation was carried out with track-etched membrane lters, with the mean pore sizes cut-off of 80 nm. The ultrasonic transmitter was incorporated into the device at the feed side or permeate side to examine the effect of the orientation of the ultrasonic transmitter on the fractionation capability. The 40 kHz square US waves were irradiated on the membrane from the permeate side or feed side. Fig. 8A shows the layouts of the micro uidic particle ltration system with the embedded ultrasonic transmitter. Output voltage through ampli er was 40 Vp-p. The 50 µl of 60 nm particle stock solution was diluted with 50 µl 0.1% F108 surfactant solution to investigate ultrasonic-assisted ltration performance. To initiate ltration, 100 µl of the sample was driven into the device. 100 µl of 0.1% F108 was reloaded when the rst sample passed so that the residual particles were collected in the collection reservoir to maximize the recovery. After the ltration, collected samples were examined using a UV-Vis spectrophotometer, and an absorbance spectra scan was performed within the 350 -700 nm wavelength range. As a result, the higher recovery rate of 60 nm Au NP was obtained from the sample ltered using feed side ultrasonic treatment at the input concentration studied. (57%) (Fig. 8B) In the case of ultrasound irradiation at the permeate side of the membrane, it yielded 48% of the original concentration using the same input sample.
In addition, the volumetric ow rate was monitored every 30 sec up to 20 min using each transmitter con guration under the same pressure and input concentration condition. (Fig.9) As a result, a similar trend was also seen in volumetric ow rate results. The higher volumetric ow rate was achieved when ultrasound was irradiated onto the membrane surface from the feed side of the membrane. A signi cant decrease in ow rate occurred after 5 min ltration when ultrasound was irradiated from the permeate side of the membrane, indicating critical fouling of the membrane. Despite the decrease in ow rate in ultrasound irradiation from the feed side, the ow rate was stabilized after 10 min initial separation process. On the other hand, ow rate consistently decreased in ultrasonic irradiation from the permeate side, showing a similar trend as with no ultrasonic treatment. The ow rate was double with ultrasonic irradiation compared to no ultrasonic treatment when the ultrasonic propagated from the feed side. According to the ow rate results, we assumed that the orientation of the ultrasonic transmitter affected the effectiveness of the ultrasonic treatment in preventing the formation of cake layer and fouling of membrane as well as recovery of the target particle Conclusions The LOC system for size-based NP separation using dead-end membrane ltration has shown a signi cant improvement in separation performance employing ultrasonic agitation. Even though many studies have been reported for ultrasonic-assisted ltration, most of them were practised in macroscale with sophisticated ultrasonic transducer system. The LOC system suggested here uses rapid and straightforward integration of ultrasonic transducer into the micro uidic system and enables nanoscale particles.
Using the novel method, we have fractionated 15 nm Au NP from polydisperse (15/30 nm) nanoparticle solution in the presence of ultrasonic wave (28-40 kHz) transmitted by the transducer integrated with the micro uidic system during the separation. The suggested technique has prevented the cake layer formation and foulant adhesion to the membrane surface by agitation. Typically, the processing time has been ~10 min by enhancing the ow rate in the presence of ultrasonic treatment during the process. In addition, it has achieved 70% of the recovery rate in Au NP separation, showing signi cant improvement in recovery rate compared to the case without ultrasonic treatment. The novel micro uidic lab on a chip has held great potential as a helpful tool that provides many advantages over other micro uidic sizebased techniques such as 1) simple fabrication by employing commercially available track-etched membrane lter with sharp pore size cut-off to bypass complicated fabrication process for building sieving microstructure, 2) easy sample handling utilizing automatic process, 3) high recovery rate (70% in case of 60 nm Au NP) of target size NP, 4) No additional chemical pretreatment required by using membrane ltration 5) Fast and straightforward processing, 6) Little dilution of the sample. This LOC approach is the rst nanoparticle separation method based on a combination of dead-end membrane ltration and ultrasonic treatment in a microscale environment to the best of our knowledge.
As these techniques are still at an early stage of development, they need system-level veri cation of feasibility for real-world application. The next step to further investigate the feasibility of this LOC system is to exercise size-selective separation using biosample. Monosome(~30nm) separation will be carried out using the LOC system as a preparative sample processing in ribosome pro ling to   Figure 1 Schematic of a micro uidic platform for nanoparticle separation. Track-etched polycarbonate membranes with different pore sizes were embedded to capture target molecules. Particles more extensive than the cut-off of the rst membrane will remain on the membrane (waste). In contrast, target molecules penetrate the rst molecule with large pores and capture in a microchamber underneath the second membrane with the lower cut-off. Captured molecules are collected in an elution buffer by applying negative pressure to micro uidic networks connected to a syringe pump. Ultrasonic transducers were incorporated with the micro uidic platform to enhance recovery by detaching particles on the fouled membrane surface.

Figure 2
Effect of ultrasonic treatment during NP separation   can go through the second membrane with a smaller cut-off (15nm), while the target size particle (20nm) stay in the chamber region between those two membrane lters.   Fractionation of 100nm polystyrene NP from polydisperse mixture solution by using ultrasonic-assisted micro uidic particle sorting. Dynamic light scattering (DLS) spectra were obtained from the ltrate (Green) and a mixture containing 1:1 polydisperse 100nm and 400 nm particles (Black). Light scattering spectra of a stock solution of 100nm and 400 nm were obtained for comparison. (Red and blue respectively) Size distribution of ltrate shows that size of particles collected in the ltrate is close to 100 nm stock solution.

Figure 8
Effect of the orientation of ultrasonic treatment on recovery rate. Ultrasonic wave with 40 kHz, 40Vp-p output voltage generated by ultrasonic transmitter was focused onto membrane lters either at the permeate side or feed side of the membrane. (A)Layouts of the micro uidic systems for particle ltration for ultrasonic treatment from the membrane lters' permeate side (left) and the feed side (right). The distance between the transducer and membrane remains 300 µm for the experiment. (B) Absorbance spectra of 60 nm Au NP collected in the ltrate solution by ultrasonic treatment from the permeate side (left) and feed side (right) of the membrane lters were compared with the absorbance using an original stock solution of 60 nm Au NP.

Figure 9
Volumetric ow rate of sample ltered using micro uidic particle separation system as a function of ltration time. Samples containing 60 nm Au NP were processed using the micro uidic system with ultrasonic irradiation from the green feed (green) and permeate (blue) or without ultrasonic treatment. As a result, a decrease in ow rate was observed in three cases. However, the highest ow rate was achieved with ultrasonic irradiation from the feed side, and the steady ow rate was observed after 600 sec of the ltration process. For the case of the ultrasound treatment from the permeate side, it showed the same level of fouling as the case of no ultrasound treatment, which indicates that the ultrasonic irradiation is not effective when the ultrasonic waves propagate from the permeate side of the membrane lter