The Intersection of Flow Cytometry with Microfluidics and Microfabrication

a. Optimization of particle concentration

Particle concentration optimization in flow cytometry is performed to: a) avoid coincidences in the analysis volume for concentrated sample, b) reduce the event rate within the acceptable limits of the data acquisition system, c) shorten dilute sample processing time.

The most prominent example of an application that is too concentrated is white blood cell (WBC) population analysis in whole blood. This analysis is impossible in a typical flow cytometer due to the presence of about 5 × 10⁹ red blood cells (RBCs) per ml of blood.³⁰ To reduce the concentration of RBCs, the most common approach is osmotic lysis of RBCs, which leaves the 5 × 10⁶ WBCs per ml of blood largely intact for analysis.^{2, 3} The WBC concentration may be further adjusted via dilution and wash steps may also be used to remove RBC debris (see below).^{2, 3} Simple dilution is also possible for many applications, but this requires larger volumes of sample to be analyzed to measure the same number of WBCs. However, dilution is easy and alleviates concerns that osmotic lysis disproportionately affects rare cells.^{30, 31} Nonetheless, whole blood must be diluted 100 fold to obtain about 10⁷ cells per ml to avoid particle coincidences in the analysis volume and to match the maximum analysis rate of flow cytometers that deliver sample at about 10 μl/min. Emerging rare cell applications, such as counting circulating tumor cells (CTCs), will require detection of less than 100 cells per ml. This will require analysis of several ml of sample to ensure accurate counts of small numbers of cells. Therefore the dilution approach will require greater than 100 ml of diluted sample to be analyzed for very rare cell applications. This is not practical using conventional flow cytometers that deliver at rates of 10to 100 μl per minute. These concerns are the same for all rare cell applications, which include detection of CTCs, endothelial cells, hematopoetic stem cells (HSCs), and fetal cells in blood samples.^32–34

Thus, for truly rare cell analysis in blood samples, it is an important goal to develop technologies that can remove the RBC population, yet collect the cells of interest, such as WBCs, CTCs, or HSCs, without loss or modification. For flow cytometry, magnetic immunocapture techniques that enrich cells of interest based on the display of specific cell surface markers followed by analysis are most commonly used.^{3, 35} Though effective, this relies on specific cell markers, which may be lost during normal disease progression.^{31, 36} As such, label free microfluidic methods that specifically enrich cells of interest prior to flow cytometry analysis remains of interest. These have been well reviewed,³⁷ but of specific interest are flow through methods that could be included at the front end of a flow cytometry analysis. Such approaches include: a) flow in microfluidic structures such as chevrons or posts. ³⁸ b) inertial migration in narrow microfluidic channels.³⁹ c) acoustophoresis in microchannels.⁴⁰ d) magnetophoretic approaches that separate RBCs using the intrinsic magnetic properties of hemoglobin.⁴¹ Regardless of the method of choice, the overall throughput of the RBC clearing method must allow for processing of many ml of whole blood in a short time frame.

An interesting example that integrates several of the above approaches, is the CTC-iChip, which enables 8 ml/hr and 10⁷ cells/second to be processed (Fig. 4A).⁴² RBCs are first depleted using size dependent lateral displacement induced via a field of microposts. Larger cells move into a flow path for inertial focusing, and, finally, the magnetically labeled CTC. Also of great interest are acoustophoretic field flow fractionation techniques in microchannels that enrich based on size, density, and compressibility.⁴³ These have demonstrated the ability to support ml/minute flow rates and have been used to enrich for prostate tumor cells in blood (Fig. 4B).⁴⁴ The flow through nature of both of these approaches give them clear potential for integration with flow cytometry analysis.

Another alternative that may offer flow through cell separation is the use negative acoustic contrast particles that migrate to antinodes in an acoustic standing wave ^{45, 46}. Such particles are effective for the isolation of both proteins and cells, and are envisioned to be potentially applicable to inline flow cytometry analysis in the future. This concept is also of use for applications that have cells embedded largely in low-density lipid solutions, such as detection of cells in milk ^{47, 48}. Both of the above approaches drive cells to the node of a standing wave, while driving the negative contrast particles or lipids to the antinode.

Flow cytometry detection of particles of cells in dilute samples, such as pathogen detection, monitoring of algal populations, and detection of blood cells in milk, also would benefit from online concentration approaches. Of the methods mentioned above, acoustic concentration in microchannels is the simplest approach to concentration of particles as most cells and particles are easily concentrated to a node that can be positioned at the center of a microfluidic channel or the center of a cylindrical microcapillary for either direct analysis by flow cytometry or collection for later analysis.^{49, 50}

b. Rapid solution exchange

Though the small analysis volume of a flow cytometer often renders wash steps unnecessary, applications that use fluorescent probes at concentrations greater than about 500 nM, which is necessary for low affinity receptor studies, or have large concentrations of small particulate matter that scatters light, such as detection of rare cells after a lysis step, can cause excessive background that requires a wash step.^{3, 14} Wash steps using centrifugation are tedious and slow. Moreover, low affinity ligand receptor measurements require a very fast wash step otherwise the off rate of the bound ligand will prevent accurate cellular measurements.¹⁴ Thus, there is a need for rapid solution exchange around particles.

Stopped flow liquid handling systems have been developed that rapidly mix solutions and deliver them to a flow cytometer with a dead time as short as 120 ms.^51–53 Coaxial mixing capillaries have also been developed that can mix samples inline with flow cytometry analysis with dead times as short as 60 ms.^{54, 55} While both of these systems have been used for kinetic assays and rapid washing of samples, the use of microfluidics for direct particle manipulation offers the potential for significant improvements in terms of mixing times and device simplicity. Acoustic standing waves offer the ability to directly move particles from one streamline to another simply by the positioning the acoustic node of the wave in a separate flow stream with the microfluidic devices ^{43, 56, 57}. These approaches will have clear value for rapid mixing and washing in flow cytometry. It is also possible to use microstructures, as described above, to selectively move particles into new stream lines, but these approaches are not rapid. Therefore, alternative approaches using inertial focusing to rapidly switch particles between streams have been developed.⁵⁸ This approach uses wall driven lift to move particles across flow streams in under a ms and will be useful in flow cytometry analysis.⁵⁸

a. Microfluidic sample pumping

For microfabricated flow systems, integration of the above off chip pumps is a common solution.¹¹ However, towards the creation of a fully integrated microflow cytometer, several efforts have used electrophoresis or electro-osmotic flow, on-chip peristaltic pumps, or pneumatic channels to drive sample and sheath for analysis. ^61–64 These systems can generate flow rates as high as about 100 μL/min. As there are many microfluidic pump configurations, future efforts for flow cytometry analysis could explore non-pulsatile systems that support μl/min flow rates. Such systems include microfabricated disc pumps, high frequency piezo activated peristaltic pumps, and miniature rotary pumps.^{65, 66} It has been demonstrated that these pumps could generate few μL/min up to about 100 μL/min. An interesting alternative on-chip pumping of fluid is the transportation of cells in an electric field, where E. coli was transported through the analysis volume of a flow cytometer ⁶⁷. As the field lines intersected the interrogation volume the pumping simultaneously moved the cells and provided positioning for analysis.

Each microfluidic pumping mechanism has its own strengths and weaknesses. Absence of moving parts, ease of fabrication and integration, pulse-free flow, low power consumption, low-flow rates, high-flow rates, and ability to use in high back pressure conditions are some of key strengths in various pumping mechanisms mentioned here. On the other hand, limitations due to non-steady flow, need of controlling fluid properties, extra fabrication steps, limited flow rate ranges, and presence of moving parts can hinder the use of a wide array of microfluidic pumps in flow cytometry systems. Thus, the method of sample delivery will need to be carefully matched with the given flow cytometry application.

b. High throughput sampling

Sample delivery to analysis on the flow cytometer is a time consuming process that is typically performed manually at rates of about one sample analyzed per minute. To automate the processing of multiple samples, current commercial flow cytometry systems routinely employ tube handling systems, plate feeders, and other laboratory robotics. These approaches make handling hundreds of samples a relatively simple and hands free operation.³ Though such systems speed sampling to an extent, conventional automated sampling remains too slow to sample the hundreds of thousands of samples that are often required for high throughput screening applications. Nonetheless, the high analysis rates of flow cytometry make it an excellent choice for high throughput screening of reactions in or on cells and microspheres.⁶⁸ Therefore, high throughput “plug flow” systems capable of at rates of two samples per second have been developed ^68–70. These systems move a sampling capillary between samples arrayed in microtiter plates. The sampling tip is continuously driven or drawn by a peristaltic pump and as it enters each sample and it draws up as little as 2 μl before exiting the sample. As the sampling tip moves to another sample it draws in a small bubble of air that separates each sample. In this way a sampling line delivers a continuous stream of samples to the flow cytometer for analysis (Fig. 5A). Recent iterations of this approach have sampled from up to four 1536 well plates simultaneously. Use of multiplex microspheres for bioassays to screen for activities such as protease cleavage are expected increase screening rates to greater than 1000 experiments per minute and have been routinely used by a single person to screen 60,000 wells per day.^70–72 A typical flow cytometer’s maximum sample flow rate of a few hundred μL/min limits the ability of high throughput flow cytometry to process large volume samples. However, the emergence of new sample focusing approaches, such as acoustic focusing, that enable higher sample volume rates (as discussed in section B.3) may allow for screening of larger sample volumes in reasonable time frames. This would make new screening applications, such as screening for response of are cell populations to lead compounds, possible.

The ability of microfluidics to process small volumes in parallel has long been recognized to provide potential benefit for high throughput screening ⁷³. This approach has been used to provide image analysis of cells for high content screening in 384 microfluidic flow channels (Fig. 5B) ^{74, 75}. For this system, laser scanning is used for parallel analysis across each channel. Notably, it demonstrates the concept of increasing flow cytometry sample-to-sample analysis rates through parallelization. Beyond simply increasing sample throughput for screening, the use of microdroplet encapsulation approaches enables the further reduction of reaction volumes for screening and direct inline flow analysis. This can automate many steps of screening, including dose response analysis and sample loading to create multiplex assays. ^76–80 The combination of a long capillary line and a microfluidic droplet generation system offers the potential for the generation, storage, mixing and inline analysis of droplets for high throughput screening in a flow format (Fig. 5C) ⁸¹.

a. Particle focusing

There are many potential methods to focus particles for flow cytometry analysis in microfabricated systems, which include simple constraint by the walls of a flow channel,^{7, 9, 10, 82} use of sheath fluid for 2-D hydrodynamic focusing ^89–91, bulk acoustic standing waves (BAW) ^92–94, standing surface acoustic waves (SSAW) ^{95, 96}, dielectrophoresis (DEP) ^97–99, inertial focusing ^100–103, use of microstructures to induce flow path changes ^104–106, or a combination of these approaches. These have been well reviewed recently ^107–109. Therefore, we simply discuss the advantages and disadvantages of the various approaches and recent examples of each type.

The most straightforward flow system to fabricate is clearly the non-tapered microfluidic channel, where the sample is simply pumped through the channel for particle positioning. This offers flexibility with regards to channel placement and dimensions to optimize optical or other detection approaches, but it does not precisely restrict the flow path of the particle or the surrounding sample. This results in variations in flow velocity through the analysis volume, as most pressure driven flow systems have parabolic flow profiles ³⁹. This results in loss of analysis precision, but use of velocity tracking techniques can improve precision ^{12, 82}. This approach also does not constrain the sample stream, which results in optical interrogation of more of the sample stream and reduces the ability to resolve free vs. bound fluorescent probe.

Hydrodynamic focusing was integrated into flow cytometry in large part to improve the above concerns. However, while it is simply implemented to provide focusing in one-dimension (across the width of the stream), by flowing in a sheath stream in a channel or channels that intersects the sample stream on one or both sides, more complex fabrication approaches must be undertaken to provide focusing in two dimensions (side to side and top to bottom). Multilayer microfluidic systems have made this approach successful in many flow cytometry development efforts ¹⁰⁷. However, recently simple structures have been used to adjust the flow profile, such as chevron structures in PDMS microchannels, have been used to create two dimensional focusing without the need to apply sheath above and below the sample stream ^{104, 109}. This system uses the chevrons to drive sheath flow from the sides to also focus from the top and bottom. Similarly, a two dimensional focusing system in glass was created by etching micro-weirs above and below the sample stream, which results in tight vertical positioning of the particles in the sample stream ¹⁰⁵. These approaches are promising in that they do not require complex layered fluidics to achieve two dimensional particle focusing.

Despite the advantages of hydrodynamic focusing, it requires liquid sheath and accelerates the particles during analysis. This reduces instrument portability and the increased linear velocity increases the demands on the detection and data acquisition systems ^{17, 50, 110}. Furthermore, the use of sheath can lead to sample dilution after analysis, though microfluidic demixing approaches show promise in the recovery of sample ¹¹¹. For these reasons, sheathless particle focusing approaches have been pursued.

The high flow rates supported by particle focusing using BAW made it a sheathless focusing approach that was shown to be effective for flow cytometry ^{49, 50}. In this method, particles and cells are driven to the node of a BAW as they are more dense than the surrounding sample fluid ¹¹². Use of cylindrical capillary makes it possible to place the node of the standing wave axially within the capillary, due the structural excitation of the capillary walls ¹¹³. However it is not simple to microfabricate cylindrical microchannels, but rectangular microchannels are not only easily fabricated, more than one node of the standing wave can be created in transverse direction of the channel. This enables generation of multiple parallel focused streams of particles in a single channel ^{92, 93}. A single particle stream can be split into many focused streams in a split channel format (Fig. 6A) or in a single channel format ⁹³ without using multiple sample injectors or flow controls. This makes it simple to generate many flow streams that will enable parallel analysis in the future. Using drive frequencies that match harmonics across the width of the channel and the depth of the channel enables two-dimensional focusing for precise positioning of particles ^{47, 48}. Acoustic focusing can be implemented in wide range of flow rates, ranging from few μL/min to several mL/min ^{84, 92, 93}. Conventional silicon microfabrication technologies are well suited to create BAW driven microfluidic channels in silicon and glass with appropriate dimensions required for acoustic manipulation ^{11, 43}. If polymer materials are the preferred fabrication approach, SSAW based methods can be implemented ^114–117. These approaches create standing acoustic waves analogous to BAW but are not limited to rigid substrates as the standing waves a generated via integrated interdigital transducers (IDTs) (Fig. 6B, 6C) ¹¹⁷. For example PDMS microchannels with IDTs have been utilized in a SSAW based continuous cell focusing system ^{96, 117}.

Inertial particle focusing in microfluidics has gained interest due to its independence from external forces and its simplicity. In inertial focusing, lift forces manipulate particles in to distinct equilibrium positions along a microchannel ³⁹. Additionally, channel curvature can be used to induce Dean flow that results in repositioning of inertial focusing equilibrium positions. This enables single streams of particles to be created precisely in microchannels, which has been shown to be effective for flow cytometry analysis ¹⁰¹. Moreover, this approach induces inter-particle spacing, which eliminates particle coincidences and might allow for higher throughput analysis in the future ¹⁰¹. As in acoustic focusing, inertial focusing can create multiple focused streams of particles. Recent work demonstrated inertial focusing of cells in 256 parallel microchannels (Fig. 6D, 6E) ¹⁰². Inertial focusing based devices are simple to fabricate in PDMS substrate and are likely to be very valuable in future flow cytometry analysis approaches. However, this technique requires that the particle make up a significant portion of the channel cross-section and can result in size dependent se paration of particles to different equilibrium positions. For these reasons, care needs to be taken when designing inertial focusing flow channels for flow cytometry.

Dielectrophoresis (DEP) is widely used for manipulation of particles. In DEP, particles with different dielectric properties are separated in an inhomogeneous electric field (Fig. 6F and 6G). DEP devices can be fabricated in silicon, glass, and PDMS. However, inherent DEP properties like strong dependence of electrical conductivity of the medium and heat generation in conductive media limits DEP as a universal approach for particle focusing. Nonetheless, several devices have been developed that focus particles for effective flow cytometry analysis using DEP as the focusing force ^{98, 99, 118, 119}. High throughput flow cytometric analysis in clinical diagnostics is in high demand. Current flow cytometric analysis has a limit of about 50,000 cells/s. This speed is inadequate when it comes to the detection of rare cells like CTCs. Simply increasing the linear velocity in a conventional flow cytometer to increase analytical rates is not possible, as this causes pressures and shear effects that result in cell death. Additionally, this increases demands for highly sensitive and fast response electronics and detectors. Parallel analysis of multiple streams is an attractive alternative. However, it is not easy to generate parallel streams using hydrodynamic focusing, as such systems will require multiple sample inputs and flow control systems. Furthermore, multiple streams in sheath flow focusing will consume increased volume of reagents and generate more waste. Sheathless focusing methods like acoustic focusing and inertial focusing have the potential to generate multiple focused streams in parallel without significant increase in sample flow rate and with minimum device complexity.

Finally, all these focusing techniques have unique advantages as well as limitations. Combining two focusing techniques can compensate some limitations in one technique to some extent. For example, a sheath flow can be introduced into acoustic, DEP, or inertial devices to eliminate orthogonal dispersion of particles ^{103, 118, 120, 121}

b. Consideration of optical detection in a flow cytometer

Flow cytometers almost invariably collect forward scatter (FSC), side scatter (SSC), and at least a few colours of fluorescence. This basic format is very valuable as bivariate plots of light scatter at multiple angles and accurately differentiates WBCs into three or more distinct cell populations ^{122, 123}. While FSC is very roughly proportional to a particles relative size and side scatter is proportional the internal granularity of the particle, these parameters are not sufficient for cellular identification in many applications. Addition of fluorescence enables the use of immunophenotyping to further discriminate cell populations. A notable example of this approach is the use of an anti-CD45 fluorescent antibody to discriminate lymphocytes combined with an anti-CD4 fluorescent antibody to discriminate CD4+ T-cells, which is the basis of the Panleucogating method for CD4 cell counts in HIV analysis (Fig 7A) ^{124, 125}. Using this basic approach, cells can optically split into a variety of populations for diagnostics (Fig. 7A) or for larger scale use of multiple antibodies can determine many cell populations for use in immunological studies (Fig. 7B) ^{2, 21, 124}. In addition to antibodies for cellular markers, use of other fluorescent dyes such as intercalating dyes and lipophilic dyes enable flow cytometry to measure nucleic acid content, membrane potential, and membrane integrity, among many other cell properties ². The collection of multiple parameters of fluorescence is also the basis of suspension microsphere arrays, where up to 500 populations of microspheres are discriminated by a combination of their size, which alters the scatter parameters, and/or their level of fluorescence staining with one or more colours of fluorescence (Fig. 7C) ^126–129. Using the surface of a such microsphere, it is possible to perform multiplexed ELISA like immunoassays, single nucleotide polymorphism detection, and a variety of enzymatic activity assays ^{126–128, 130, 131}. In these assays, the type of assay is identified by the microsphere position in the array, and the assay result is read out in a separate fluorescence channel that has been gated to only measure the values from the microsphere of interest.

Sensitive and accurate analysis of fluorescence on a flow cytometer requires both calibration of the intensity and spectral response of the instrument. Detection of multiple fluorophores on a single instrument requires consideration of fluorescent crosstalk. In flow cytometry this is generally accomplished using a process known as compensation, which is a relatively complex mathematical process to deconvolve the individual contribution of each fluorophore in each detector that has been extensively discussed elsewhere, but must be considered for accurate flow cytometry measurements ¹³². As sensitivity and resolution is also critical to instrument performance, there have been several attempts to standardize evaluation of flow cytometers for these parameters. A well accepted approach is to use a set of calibrated microsphere standards that allows for precise estimation both the sensitivity and resolution of a flow cytometer’s fluorescence measurements ^{19, 133}. Using this approach to estimate sensitivity, many commercial flow cytometers now detect as few as 50 fluorophores per particle ¹⁸. This extremely sensitive performance has required the use of high numerical aperture collection optics and stable laser sources, but systems with performance sufficient for some flow cytometry applications have been built with very simple components such as laser pointers or LEDs for excitation and fiber optics for collection ^{110, 134, 135}.

Beyond fluorescence intensity, there have been several efforts to extract additional optical information such as fluorescence lifetime, Raman spectra, and complete images ^136–138. These approaches require the addition of high-speed modulation, CCD spectrometers, or digital cameras, but they provide significant additional information content that makes them valuable for applications such as apoptosis studies or highly multiplexed analysis. High speed flow based imaging has been widely adopted and current commercial systems can image bright field organisms flowing in a capillary at high volumetric flow rates as well as fluorescently image hydrodynamically focused cells at rates of 2000 cells/s ^{136, 139}. Even faster imaging systems that use time encoded signals to achieve imaging rates and inertial focusing as high as 100,000 s⁻¹ in a flow cytometer ¹⁴⁰. Another promising approach is the addition of Raman spectral collection to flow cytometers. The extremely narrow Raman spectral features in combination with surface enhanced Raman scattering nanoparticles and advanced data analysis approaches is anticipated to dramatically increase multiplexing and decrease optical probe cross-talk ¹⁴¹.

c. Optical detection using microfabricated systems

While there have been many microfluidic flow cytometers built, most have used stand off optical detection akin to a typical flow cytometer. Due to the similarity of these systems in principle to a conventional flow cytometer, they will not be reviewed here. There also has been extensive work on the integration of other micro optical components, such as filters from dye doped PDMS substrates, liquid micro-lenses with tunable focal length and transmission properties, and optical waveguides into microfluidic systems. ^142–144 Though there are many promising efforts in this area, here we will specifically discuss microfabrication approaches that may offer improved simplicity and/or affordability without use of complex fabrication approaches.

One of the simplest approaches to optical detection proposes a patterned mask integrated into a flow channel wall ¹⁴⁵. As the particle flows through the channel it is directly excited and its emission pattern is recorded in time as it passes the mask, which provides a known temporal modulation to the emission signal ¹⁴⁵. The known modulation is used to selectively deconvolve the particle-based fluorescence from background. Use of coloured optical masks enables the detection of multiple fluorophores. This approach will dramatically reduce the need for alignment and use of complex spectral selection elements.

Though direct fiber optic collection and delivery have been long proposed for microfluidic flow cytometry ¹⁴⁶, the performance of such systems still significantly pales to that of a conventional flow cytometer. This performance lag is shown in the > 5 fold increase in the coefficient of variation seen between fiber optic systems and a modern flow cytometer ¹⁴⁷. Though fiber systems will offer more robust performance due to their simplicity and rigid construction, the low numerical aperture of most such systems is currently highly limiting to fluorescence detection. Nonetheless, the robust alignment afforded by this approach has made it possible to create simple flow cytometers that effectively perform bead based detection of bacteria and toxins ^{148, 149}. However, the inherent optical limitations of the numerical aperture of most fibers has been recognized by the microfabrication community, which has fuelled the development of on chip lenses for light collection or use of lenses incorporated directly into fiber systems. These systems show improved light collection for flow cytometry ¹⁴⁴, but they add complexity to the microfabrication process.

One of the areas of opportunity for microfluidic flow cytometry is expected to be parallel analysis of samples. To that end technologies that faithfully reproduce the flow cytometry analytical volume are of interest. One microfabricated approach that can be applied over many parallel channels is the use of a microball lens array, which has been shown to enable sensitive collection over many flow channels and provides a tight analytical volume for each analysed stream ¹⁵⁰. This or similar approaches may be highly effective for high throughput flow cytometers.

d. Impedance detection

Some of the earliest flow cytometers incorporated Coulter volume sensing as a measurement parameter ¹⁵¹. Coulter volume sensing works by measuring the impedance change across a small orifice as a cell or particle passes through the orifice ¹⁵². The measurement of particle impedance as it displaces a volume of solution is extremely sensitive and it is commonly used to measure particles ranging from submicron to millimetres in diameter ¹⁵³. In addition to simply measuring the impedance change, which is proportional to the volume of the particle as it passes through, many microfluidic cytometers are pursuing the use of impedance spectroscopy as a method of particle analysis ¹⁵⁴. This approach has been demonstrated to successfully differentiate leucocyte populations in flow and to successfully count CD4+ cells ^155–157. As these approaches do not require optics they may be both effective in low cost and high throughput application areas.

a. Microvalve sorting

A common approach to creation of a microfluidic flow sorter is the use of on-chip microvalves, which was used in early work to selectively sort particles at rates of 20 per second ⁶¹. Though many valve approaches have been developed, they all share the principle that a valve is opened or closed to divert a particle into collection channel. These systems are inherently limited by the activation rate of their valves and the displacement volume of the valve, but recent microfabricated magnetic valve based sorting systems can achieve kHz activation frequencies with extremely small displacement volumes, which makes it a promising approach for both parallel and microfluidic sorting approaches ¹⁶¹.

b. Field based sorting

The use of acoustics has a long standing role in flow sorting, including the use of piezo drive to launch acoustic pressure waves into a y-shaped microchannel, where pressure wave diverts particles of interest at rates approaching 1000/s ²⁷. In addition, to this approach SSAW has been used to sort particles into multiple collection streams in microfabricated chips ¹⁶². This approach may enable direct incorporation of sorting in a variety of microfabricated flow cytometers and will be parallelizable, but will require increased linear velocities beyond the current 2 mm/s to achieve high sorting rates. Alternatively, surface acoustic waves can be used to induce acoustic streaming that directly imparts energy into a liquid. This approach has been used in a microfabricated sorting system that combines hydrodynamic focusing to position the flow stream as it enters the sorting region ¹¹⁵. The flow stream is then programmatically oscillated at frequencies of several kHz by the imposition of SAW waves, which results in the movement of the stream across the intersection of two collection channels. This has enabled accurate and gentle sorting of cells at rates as high as 2000 s⁻¹.

c. Droplet sorting

It is also possible to encapsulate cells in aqueous droplets within an oil stream, which enables flow cytometry measurements on the droplets as they flow past. In this case SAW can directly move the aqueous droplets in oil and has been used to sort cell containing droplets at rates of 2000 s^{−1 115, 163}. Dielectrophoresis has also been used to effectively sort encapsulated cells at similar rates ¹¹⁹. Though these approaches add the complexity of encapsulating cells in droplets, they enable examination of enzymatic activities external to the cell and obtain kHz sorting rates.