Enrichment of microparticles in droplets using acoustophoresis

Enrichment of microparticles

in droplets using acoustophoresis

Klara Björnander Rahimi

Abstract

Enrichment of microparticles in droplets using acoustophoresis

Klara Björnander Rahimi

droppar

Mikrofluidik är det vetenskapliga fält som behandlar små volymer av vätskor inneslutna i kanalsystem i mikroskala. I storleksordningen en miljondels meter spelar yteffekter större roll än vad som är bekant i vardagliga förteelser, vilket gör att flöden inte blandas på det kao- tiskt turbulenta sätt som det vanligtvis gör. En kaffedrickare i normal storlek som dricker sitt kaffe med mjölk kan nöja sig med att röra några varv med en sked för att blanda ut mjölken i kaffet. Vore kaffekoppen i mikrometerskala istället skulle det bli ytterst krång- ligt att blanda de två vätskorna, eftersom de då flödar laminärt. Ett laminärt flöde innebär att vätskor som färdas i samma riktning flödar parallellt där den enda laterala rörelsen som sker är via diffusion. Att blanda saker via diffusion är en tidskrävande process och när drycken vore jämnt blandad skulle kaffet vara alldeles för kallt för att kunna åtnjutas.

För att blanda reagens i mikroskala är det istället väldigt fördelaktigt att använda sig av droppar. Dropparna är olösbara i den omslutande vätskan och de består ofta av en vatten- löslig vätska omgivna av olja. Gränsskiktet mellan vatten och olja skjuvas mot kanalväggen när droppen rör sig, vilket inducerar rörelse inne i droppen som gör att ämnena därinne blandas fort. Dropparna kan liknas vid enskilda provrör, i vilka reaktioner kan ske. Celler kan inneslutas i droppar och i ett mikrofluidiskt chip kan deras beteende studeras. Att studera celler på detta sätt är mer noggrant än att studera celler i cellkulturer. Det konven- tionella sättet att forska inom cellbiologi är att undersöka stora cellpopulationer samtidigt och ett problem med denna metod är att sällsynta celler och cellreaktioner lätt försvinner i den homogena massan av mer vanligt förekommande celler och reaktioner. Genom att studera isolerade celler i var sin droppe utmärks även de ovanliga händelserna vilket gör en sådan mätning mer noggrann.

Då droppen innehåller partiklar är det möjligt att manipulera dem, att exempelvis positionera dem på bestämda platser. Genom att försätta det mikrofluidala systemet i högfrekventa vibrationer uppstår ett tryckfält inuti mikrokanalen. Om kanalbredden är lika bred som halva våglängden på tryckvågorna uppstår stående vågor med en nodlinje i mitten mellan kanalväggarna. Längst nodlinjen råder tryckminimum och partiklar som har en positiv s k. kontrastfaktor kommer samlas där. Kontrastfaktorn beror på materiale- genskaperna hos partiklarna och det omgivandet mediet. Celler omgivna av vatten har en

positiv kontrastfaktor. Tekniken med vilken det går att styra mikropartiklar på detta sätt kallas akustofores.

Genom att förutsäga var partiklarna kommer att samlas kan det mikrofluidala systemet utformas för att påverka dropparna med olika processer. Det går att injicera ny vätska och reagens in i dropparna, de kan ledas genom kanalstrukturer för att stimulera blandning av reagens och de kan analyseras med olika mätredskap för att avgöra hur många partiklar eller levande celler som finns där i. För att utföra cellanalys krävs ofta att cellerna tvättas ett antal gånger, vilket kan ske genom att droppen injiceras med ren näringslösning. För att inte droppen ska bli för stor av dessa vätskebidrag krävs också att droppen efteråt delas upp i mindre storlekar. Detta kan göras genom att leda in den i en förgrening i kanalen, där droppen klyvs till två eller flera mindre droppar.

I detta projekt användes två kanalgeometrier i vilka vattendroppar skapades och sedan leddes in i en kanalförgrening där dropparna delades i tre dotterdroppar. I dropparna fanns mikropartiklar som först linjerades upp i nodlinjen i kanalen med hjälp av akustofores. När dropparna nådde droppdelaren var partiklarna linjerade och majoriteten av dessa hamnade i mittendroppen, vilken berikades på mikropartiklar.

De två designerna som testades var tillverkade på två sätt, det ena via kristallplans- beroende våtetsning och det andra via torretsning. Se figuren nedan i vilken konceptet är illustrerat för de två designerna.

Konceptet för projektet illustreras i denna schematiska bild. a) är den våtetsade designen och b) den torretsade.

Stabil droppdelning observerades för båda chipen för olika flödesinställningar. Partik- larna som samlades i mittendroppen räknades och det våtetsade chipet samlade som bäst 81% av de inkommande mikropartiklarna och det torretsade 93%.

Examensarbete 30 hp på

civilingenjörsprogrammet i Teknisk fysik med materialvetenskap Uppsala Universitet, januari 2018

Page

1 Introduction 1

2 Theory and background 2

2.1 Microfluidics . . . 2

2.1.1 Lab-on-a-chip . . . 3

2.2 Acoustophoresis . . . 4

2.2.1 Acoustic radiation force . . . 4

2.2.2 Secondary radiation force . . . 5

2.3 Other aspects regarding acoustophoresis . . . 6

2.3.1 Acoustic streaming . . . 6

2.3.2 Bulk and surface acoustic waves . . . 6

2.4 Droplet microfluidics . . . 7

2.4.1 Droplet generation . . . 7

2.4.2 Calculating droplet volume . . . 8

2.4.3 Droplets and cells . . . 9

2.5 Droplet microfluidics and acoustophoresis . . . 9

2.5.1 Goal of master’s thesis . . . 12

3 Materials and fabrication 14 3.1 Material requirements . . . 14

3.2 Fabrication methods . . . 14

3.2.1 Photolithography . . . 15

3.2.2 Wet etching . . . 15

3.2.3 Dry etching . . . 16

3.2.4 Actuation of acoustic waves and fluid setup . . . 16

3.3 Two designs used for this project . . . 18

4 Experimental 20 4.1 Experimental setup . . . 20

4.2 Acquiring data . . . 20

4.3 Comparison between the wet etched chip and the dry etched chip . . . 22

4.3.1 Constant volumetric flow . . . 22

4.3.2 Constant droplet speed . . . 23

4.4 Evaluation dry etched chip . . . 23

4.4.1 Influence of acoustic force at different throughput . . . 23

4.4.2 Concentration factor . . . 24

4.4.3 Manipulation of cells . . . 25

5 Results and discussion 26 5.1 Comparison between the wet etched chip and the dry etched chip . . . 26

5.1.1 Recovery . . . 28

5.1.2 Comparing two stable events . . . 35

5.2 Performance dry etched chip . . . 38

5.2.1 Influence of acoustic force at different throughput . . . 38

5.2.2 Concentration factor . . . 40

5.2.3 Manipulation of cells . . . 42

6 Conclusion 44 6.1 Comparison between the wet etched chip and the dry etched chip . . . 44

6.2 Evaluation of the dry etched chip . . . 45

6.3 Manipulation of cells . . . 45

6.4 Future work . . . 45

7 Acknowledgements 47 Appendix A LOT traveller for production of dry etched chip 52 Appendix B Experimental data 55 B.1 Comparison between wet etched chip and the and dry etched chip . . . 55

B.1.1 Volumetric flow rate 10.5 µL/min . . . 55

B.1.2 Length-to-width set to 3 . . . 55

B.2 Comparing two stable events . . . 56

B.3 Performance dry etched chip . . . 56

B.3.1 Influence of acoustic force at different throughput . . . 56

B.3.2 Concentration factor . . . 56

Appendix C Fibers in the microfluidic channels 57

Microfluidics is the name of the scientific field that deals with small volumes of fluids contained in channels in the micrometer range. Volumes that are handled are very small, typically nanolitres down to attolitres [1], and the channels where these volumes are kept are within a size range of a few hundred micrometers down to tens of micrometers. The channels can be etched in silicon or glass wafers, moulded in a polymer or even be made out of the space between the fibers in a piece of paper.

Droplet microfluidics is a field within microfluidics where micro scaled droplets are used for different applications. It is a rapidly expanding field as a result of its large potential in many different applications, such as material synthesis and biological analysis [1,2]. The ink jet printing technique is what started the interest in droplet microfluidics with its need to rapidly dispense small droplets with a high level of control and now droplet microfluidics is used in life science as cell vessels for diagnostics and biological research [3, 4]. When using droplets as containers for particles, a technique to control the particles is very useful.

To fit this requirement, acoustophoresis makes its appearance as a label free method for particle manipulation. It aligns particles in pressure nodes and antinodes in a microfluidic channel, using the intrinsic material properties of the particles themselves and has the ability to separate differently sized objects [5]. It is a well controlled method suitable for manipulating the position of cells and has been proven to do so on several occasions in single phase flows [6–9]. It has also been shown that 5-10 µm microparticles and cells can be manipulated inside aqueous droplets [10–12]. This Master’s thesis project is a step towards utilizing acoustophoresis together with droplet microfluidics in order to establish faster cell analysis processes with higher sensitivity than the conventional methods used today. In order to do so, microfluidic chip designs must be established and evaluated with regard to particle recovery and droplet behavior.

2 Theory and background

This chapter aims to present the reader with the theory necessary to understand the fields in which this project takes place. The following sections gives a description of microfluidics and the lab-on-chip concept, acoustophoresis, droplet microfluidics and finally a summary of the advantages one receives when combining acoustics in droplet microfluidics. The chapter is completed by a description of the goal of this project.

In this thesis, the word beads is sometimes used instead of the word microparticles and they refer to the 10 µm polystyrene particles suspended in the water droplets. The words system and device are sometimes used as synonyms for chip and refers to the two kinds of microchips fabricated and used for these experiments.

2.1 Microfluidics

In sub millimeter dimensions, fluids behave differently compared to the macro scale in which we find ordinary objects. In macro scale, inertia has a big impact on the dynamics of a fluid. Turbulence will occur which will influence the behavior of the fluid volume. In the micro scale however, viscous forces will dominate the intertial ones, and fluidic flows will not experience turbulence. Instead, a laminar flow will occur. [1,13,14]

The reason for this is because surface effects have a bigger impact than volume effects in the microscopical size range, illustrated in figure 2.1. The distinction between turbu- lent and laminar flows is defined by a dimensionless number called Reynolds number (Re), which is the ratio between inertial and viscous forces. A low value indicates the lam- inar regime, where two liquids can flow together without lateral blending other than by molecular diffusion. A high value represents the more commonly known domain in which turbulence and unpredictable mixing takes place. [14, 15]

Microscopic objects are so small the gravitational pull is smaller relative to surface ef- fects which now increases in importance. This is the reason why small droplets of conden- sation sticks to a glass window instead of dripping down to the ground and how in nature you can find water striders securely walking on the surface of the water.

Figure 2.1: A graphic illustration of the relationship between a quadratic and cubic behavior.

The physical effects in micro scale is dominated by those which relate to L² (such as viscocity, surface tension and evaporation), and in the macro scale the dominating effects (such as gravity and inertia) relate to L³.

2.1.1 Lab-on-a-chip

Microfluidic devices can be used for many applications, combining many fields of research.

Tools available in conventional laboratories such as pumps, mixers and analytical sensing and detection methods have been miniaturized and can be integrated on microfluidic chips.

When they are connected to liquid reservoirs, analytical tasks can be performed in the micro scale on these chips. Platforms able to accomplish miniaturized experiments are called lab-on-a-chip. [3, 15–17]

There are many positive outcomes from the miniaturization of a system, where the most obvious one is reduction in sample and reagent volume. Gained from the volume reduction is an increased temperature control over the reaction environment since the mi- crofluidic chips can be put in incubation chambers, and heat transfer is quick since the surface-to-volume ratio is large. Since the space for the fluidic pathway is small, paralleliza- tion of several channels and integration of several experimental tasks is possible while still remaining on a small surface. Parallelization will in the end result in a higher throughput of the microfluidic system. [16]

Examples of commercial lab-on-chip applications are capillary test strips in which preg- nancy status, blood glucose levels and malaria diagnosis can be obtained only from pro- viding urine or blood from a patient. In the test strips, fluid transport is made possible by capillary pressure which will force liquid into small, wettable capillaries. The capillary force is negligible in the macro size range, but quite noticeable in the micro range. The test strips has chemical reagents pre-dried into the capillaries which will react to the specific components of the bodily fluid and the result from the reaction is visible and interpretable by the patient. [3, 18, 19]

Acoustophoresis

2.2 Acoustophoresis

When applying high frequency vibrations to a microfluidic device, pressure waves will form inside the fluidic chamber and affect the medium within. Dealing with these pressure waves inside fluids is done in the interdisciplinary field of acoustofluidics and the concept of manipulating particles using the acoustic force is named acoustophoresis. The frequency range used is in the ultrasonic region, which starts at 20 kHz and continue to above several hundred megahertz in the ultrasonic range.

When the width of the microfluidic channel is a multiple of half the wavelength of the pressure wave, a standing wave will appear between the channel walls, with its node positioned in the middle of the channel and antinodes at the walls [20,21].

When n=1 in equation 2.1, this condition is fulfilled and the fundamental harmonic is acquired in the system.

n∗ λ

2 = v

2f (2.1)

The second harmonic is acquired when n=2, and a schematic illustration can be seen figure 2.2. λ is wavelength of the pressure waves, v is the speed of sound in the media and f is the frequency of the vibrations.

Figure 2.2: Schematic images depicting microparticles (green circles) in a single phase flow in a microfluidic channel. a) Acoustics is not applied and the microparticles are randomly distributed in the channel. b) Acoustics is applied (red lines indicate standing waves) and first harmonic is aquired. The microparticles are aligned in the pressure nodes of the ingoing and reflected pressure wave. c) Acoustics is applied and the second harmonic is aquired. The microparticles are aligned in two pressure nodes.

2.2.1 Acoustic radiation force

When the chamber is filled with a medium containing suspended particles, these particles will experience a force when the pressure waves are present in the system. This force is

called the acoustic radiation force, expressed in equation 2.2, and its size will be dependent on several parameters: a is the radius of the microparticle and k is the wave number which is defined by k = ²_λ^π. E_acis the acoustic energy density which is related to the amplitude p of the pressure waves in the channel which, in turn, is related to the voltage U applied to the piezoelectric actuator: E_ac∝ p²∝ U². φ is the acoustic contrast factor, β and ρ is compressibility and density, respectively. x is the distance from the channel walls. [22]

F_rad= 4πa³kE_acφ(β,ρ)sin(2kx) (2.2) Depending on the sign of the contrast factor φ, the particles will be driven to either the node or the antinodes of the pressure waves. When the contrast factor is positive particles will be driven towards the node, see figure 2.2. A negative contrast factor however, will make particles gather at the antinodes. [10, 23]

This factor is determined by the density and compressibility of the particles p and the surrounding fluid f, see equation 2.3. [5,20,22]

φ = 1

3(5ρ_p− 2ρf

2ρ_p+ ρ_f −β_p

β_f ) (2.3)

When the second harmonic is acquired in the system, two pressure nodes and three antinodes is created between the chamber walls. Particles with a positive contrast factor will align in the pressure nodes, which can be seen in figure 2.3 where this is applied in a flow with two phases.

Figure 2.3: Pressure field is generated with two nodes at ¹⁄⁴ and³⁄⁴in from the channel walls.

Particles are aligned in the nodes, close to the walls and are sorted into the side droplets. Scale bar is 150 µm. Picture by Fornell et al. [11]. Standing waves are indicated by added red dashed lines.

2.2.2 Secondary radiation force

Apart from the primary acoustic radiation force F_rad, there is also a secondary force present in the system (sometimes referred to as the Bjerknes force). This is an interparticle force,

Other aspects regarding acoustophoresis

present when more than one particle is involved, and may act as either an attractive force or repulsive force between the particles suspended in the fluid. However, the magnitude of the force is dependent on the distance between two particles influencing each other and its effect is mainly noticeable at very high particle densities. Otherwise, it is several magnitudes smaller than the primary radiation force F_rad. [21,24]

Results by Fornell et al. [11] indicate that the distribution of particles in the droplets are not affected by this secondary force in a dominating way, at least not up to a particle amount to 50-60 particles per ingoing droplet. In this report, the particle amount will not exceed this amount, hence the influence of the secondary radiation force will not be taken into account.

2.3 Other aspects regarding acoustophoresis

2.3.1 Acoustic streaming

It has been shown that particles with a radius 0.2 µm are predominately affected by a second effect caused by the pressure field in the chamber [25]. This effect is called acoustic streaming and will generate hydrodynamic drag forces on the particles suspended in the fluid, via rotational flows created in the chamber. The hydrodynamic drag forces related to these rotational flows will counteract the acoustic radiation force which focuses particles to the nodes (or antinodes) of the pressure waves present in the chamber [14,24,26]. Acoustic streaming can be used for mixing fluids in laminar flows [13] and to sort differently sized particles [27] and will mainly affect particles with a radius smaller than a micrometer.

When particles are larger than 1 µm the acoustic radiation forces will dominate [25].

2.3.2 Bulk and surface acoustic waves

In acoustofluidics there are two groups in which methods are divided into, based on the way the acoustic waves propagates in the system. When the chip is made out of silicon or glass the acoustic waves will propagate through the bulk of the material, originated from a piezoelectric crystal attached to the device. The fluidic chamber is the acoustic resonator, where resonance occurs between incoming waves and waves reflected at the channel walls.

When the wavelength of the acoustic waves matches the channel width in accordance with equation 2.1, there will be resonance and standing waves are the result. The waves created this way are named bulk acoustic waves. [11,28,29]

In microfluidic systems based on polydimetylsiloxane (PDMS) acoustic waves are in- troduced to the fluidic chamber through the underlying substrate which is a piezoelectric material, on top of which the PDMS is bonded. The actuator in this case are metallic electrodes deposited on the substrate on either side of the microfluidic chamber. Acoustic waves from the electrodes will propagate through the surface of the piezoelectric material

and interfere in the fluidic chamber. Waves propagating through the surface of the material are called surface acoustic waves. [5,9,29]

2.4 Droplet microfluidics

Droplet microfluidics is a field within microfluidics, in which monodisperse droplets are generated and handled inside a microfluidic system. The droplets consist of either air or an aqueous liquid contained by a second liquid in which the droplets are immiscible, typically water droplets in oil. In a microfluidic chip, droplets can be generated at a high frequency with high reproducibility and range between hundreds to tens of micrometers in radius.

They serve as containers for particles or reactants, well-defined micro chambers for the reactions to take place, similar to ordinary test tubes found in a chemical laboratory. The interface between water and oil is a diffusion boundary for large molecules, and while small molecules may penetrate the surface, cross contamination of larger particles will not occur.

In addition, reagent volume and dispersion of reagents may be reduced as the reaction is limited to a droplet and not the entire microfluidic chamber. [15, 17,30]

Another advantage one receives when using droplets as reaction vessels is rapid mix- ing. This is in contrast to laminar flows in which mixing will not happen apart from the slow process of molecular diffusion. Interaction between the stationary wall and the fluid boundary causes shearing and fluid movement, which induces mixing when flow is recirculated. It is also common to lead droplets in complex channel geometries to induce mixing further since shearing is extra effective when the droplet passes corners in the fluidic channel. [30, 31]

What is also possible when restricting reactions to a specific volume are time resolved analyses of reactions. By controlling the speed of droplets while letting them pass a channel with a known length, it is possible to analyze the reaction response at the end of this channel and easily define the time in which the reaction took place. It is also possible to etch incubation chambers for long time storage of molecules inside droplets. [4, 17]

2.4.1 Droplet generation

In this thesis droplets are generated in two ways; via T-junction and flow focusing, seen in figure 2.4. They are both well-established methods for producing monodisperse droplets at rates up 10 kHz [32]. Both are passive methods, meaning that they generate droplets via constant flow rates which may be supplied by syringe pumps. Active droplet generation requires an external source of energy to activate droplet generation. [33]

When using a flow focusing design, two streams of the continuous phase are flowing into a stream of the dispersed phase pinching it, figure 2.4 a)-c). Break up occurs when the shear stress from the continuous phase is larger than the interfacial tension holding the droplets attached to the stream, see figure 2.4 c). [2, 15,34].

Droplet microfluidics

In a T-junction, a continuous phase is interrupted by a second immiscible fluid via a perpendicularly connected channel. The continuous phase will act upon the intruding fluid by shear forces and break it into dispersed droplets, see figure 2.4 d)-f). [2]

A dimensionless number which can be connected to this technique is called the Capil- lary number, (Ca). The Ca-number is the ratio between viscous forces of the continuous phase and the interfacial tension between the continuous and the dispersed phase. Small values of Ca represent stable droplet generation regime, as opposed to larger Ca values which implies a very rapid and uncontrollable droplet generation in the T-junction. [15,34]

Figure 2.4: The three pictures in the top row is droplet generation process in a flow focusing system. The three pictures in the bottom row is a droplet generation process in a T-junction.

Direction of flow is from left to right. In the leftmost and middle images, fluid is starting to add volume and are starting to be squeezed off from the rest. In the rightmost images a new droplet has formed and is transported away.

2.4.2 Calculating droplet volume

The simplest way of calculating the droplet volume is to approximate its size with that of a rectangular block, finding the volume to be V = A_{c r os s}∗ L where the cross sectional area A_{c r os s} is found using the height H and width W corresponding to the channel geometry.

L is the measured droplet length which is measured by viewing the droplets from above using an optical microscope. This simple formula will not take the real shape of the droplet into account and since droplets in general as well as those confined within non-wetted

surfaces are not rectangular, a more precise expression can be used to find the droplet volume. Musterd et al. [35] presented a mathematical equation which takes into account the roundness of the two ends of the droplet, as well as the curvature of the droplets’

elongated cylinder shaped volume. Conditions for using this formula and estimating the volume of a model with this geometry within a 5% confidence interval are that the aspect ratio of the channel must be less than 1 and that the droplet length is more than three times its height. The formula can be applied to three types of channel geometries and in this report, the volume is calculated when droplets occur in a channel with a rectangular cross section. Hence, expression 2.4 is used:

V =



H∗ W − (4 − π) ∗

 2 H + 2

W

‹−2

∗ (L −W

3 ) (2.4)

π is the contact angle (in radians) which in the experiments was measured and found to be 155°. The aspect ratio of the channels are less than 1 and the dimensions of all systems can be seen in table 3.2.

2.4.3 Droplets and cells

Droplet microfluidics is a useful tool for making cell analyses. To earn more knowledge about biological processes and when looking for rare cells in a blood sample in diagnostics, droplets are very suitable to use as cell containers since their size match that of cells. Water droplets can be injected with nutrients or reagents such as enzymes and gene material via droplet fusion or droplet injection methods, while the permeability of respiratory gas and the removal of toxic matter is possible since the droplet interface is permeable for small molecules. The conventional method of using Petri dishes for bulk analyses of cells also provide these requirements for cell survival, however analysis results lack in resolution and cell response from rare cells or mutations are hidden within the homogeneous bulk. [4]

When instead inserting fewer, or even single cells in separate containers such as the droplets, a measurement of the content can be done and even very rare events are showed in the data [17], see figure 2.5. Single cell measurements have been acknowledged to be a promising tool in diagnostics, prognostics and cancer therapy [17, 32].

2.5 Droplet microfluidics and acoustophoresis

Acoustics has on many occasions been shown to manipulate particles and cells. For ex- ample, in single phase laminar flows it has been used to concentrate rare circulating tumor cells [36] and to differentiate between differently sized cell types for crime investigations of sexual assaults [37]. In droplet microfluidics, it has been used to manipulate whole droplets using acoustic streaming [38] and in 2015 Fornell et al. showed it was possible to manipu- late microparticles inside moving droplets using bulk acoustic waves [10]. In 2017 Park et

Droplet microfluidics and acoustophoresis

Figure 2.5: A schematic describing bulk cell analysis and single cell analysis. a) In a conventional cell test, the rare cells or events from rare reactions (red and blue dots) are hidden within the bulk of the more common cells or events (green and orange dots). b) In single cell analysis when each cell is trapped in a compartment and measured individually even small peaks from rare events are visible in the resulting data. Illustration is inspired by L. Söderberg [17].

al. used travelling surface waves to sort differently sized microparticles in droplets using a PDMS device [12].

Acoustophoresis can be used as a gentle manipulation technique. Since the fluid veloc- ity is at its maximum in the nodal line in a standing wave system where the pressure field is at its minimum, cells are not subjected to shearing from velocity and pressure gradients.

Cavitation is a process in which bubbles of air or vapor are formed due to ultrasonic pres- sure waves, and can be used for cleaning lab equipment and cell lysis devices. However, these devices operate at frequencies in the kHz region where cavitation is more likely to occur. Acoustophoresis operates in the ultrasonic frequency region where frequencies are much higher, which makes cavitation less likely to appear. [6]

Compared to magnetic particle manipulation methods (magnetophoresis) which re- quire magnetic particles as labels, acoustophoresis is a label free method reliant only on the microparticles’ own size and density. Also, both electrical (used in electrophoresis) and magnetic fields run the risk of altering the cells phenotype. Optical manipulation is done by lasers which risk thermal heating of the droplet to the point of cell lysis. Physical objects may as well be used for particle sorting, but may induce shear stress which also introduce the risk of cell death or phenotype alteration. [15]

An illustration of particle manipulation using acoustophoresis can be seen in figure 2.6.

The polystyrene particles have a positive contrast factor in a water-in-oil system and aligns in the nodes of the standing waves.

Figure 2.6: a) Acoustics turned off and microparticles randomly distributed in the droplet. b) Acoustics is turned on and microparticles are aligned in the center of the droplet.

Brouzes et al. [39] constructed a process in which droplet microfluidics is used in a cell analysis process, see figure 2.7. In it, droplets were generated with different optical labels where each label indicated a specific chemical reactant. These droplets were merged with droplets containing single cells, and later incubated when the cells were allowed time to react with the chemical reagent. After incubation, the optical label of each droplet was read and it was determined whether the cell was alive or not. In a procedure like this, there are several ways to integrate acoustophoresis. In the beginning of the process, acoustic waves could be implemented to activate droplet generation and droplets would form on-demand [40], contrary to the passive flow-focus method used by Brouzes. To merge droplets, Brouzes utilizes an electrical field to force droplets together, a process which Leibacher et al. demonstrated using acoustophoresis as well as the ability to sort droplets after their optical label has been read [41, 42]. Lastly, prior to cell encapsulation and before Brouzes process takes place, cells must undergo washing which also can be performed using acoustophoresis [43].

Ordinarily, fluorinated oils are used as the continuous media when handling cells in microfluidic assays. Oxygen is highly solvable in these oils which makes cells viable for a long period of time [44]. However, when testing the acoustic focus on particles suspended in water droplets in fluorinated oil, Fornell et al. found that the ability to focus particles was weaker compared to olive oil was used. They present theoretical results which also shows that the acoustic force on the particles is very low in the fluorinated system. [45]

Droplet microfluidics and acoustophoresis

Figure 2.7: A schematic image describing the process constructed by Brouzes et al. [39]. Droplets with different optical labels corresponding to different chemical reagents are generated. The labeled droplets were merged with droplets containing single cells and incubated, which is fol- lowed by incubation. Lastly, the optical label of the droplets were read and it was determined whether the cell was dead or alive.

2.5.1 Goal of master’s thesis

In this project, two different chip designs will be used to enrich water droplets in oil with microparticles, using acoustophoresis. The goals can be summarized as:

• Align particles inside water droplets for two different chip designs.

• Compare results from the two chips.

• Manipulate cells using acoustophoresis.

In the chips, water droplets will be generated with microparticles suspended in them.

The channel width is set so that a standing bulk acoustic wave is formed, with a nodal line in the center of the channel. The polystyrene particles have a positive contrast factor and will align in the nodal line when acoustics is applied. The water droplets will then enter a trifurcation, which divides the incoming droplets in three daughter droplets. A schematic image illustrates the concept in figure 2.8. The middle droplet will then be enriched as the majority of the microparticles will be collected in it, while some of its volume is removed.

The silicon chips were fabricated before the start of this project by two different fab- rication techniques; crystalline dependent wet etching and crystalline independent dry etching. Fabrication and images of the systems will be described and displayed in chapter 3.

Figure 2.8: Schematic images of the two microfluidic systems and the concept of using acoustophoresis to enrich particles. The red rectangle in each image represents the piezoelec- tric element which is used as an actuator to induce the bulk acoustic waves. Fluid movement is from left to right in the images. a) is the wet etched system. b) is the dry etched system.

The purpose of enriching droplets with microparticles is so that a droplet split later can be integrated with an injection system to produce a cell washing procedure. A droplet injector would inject fluid in the cell containing droplets in the system. The cells would be flushed with the freshly supplied fluid and the droplet would later be enriched once more in a droplet split using acoustophoresis and a suitable droplet split design. Knowledge regarding enrichment of particles in the systems could be incorporated when designing process schemes for single cell analyses with a high recovery for the cell washing process.

3 Materials and fabrication

In this chapter, the material requirements are listed and the two microfluidic systems are shown. It also includes a description of the fabrication processes involved for the manu- facturing the microchips. Assembly of the microchips involving gluing of actuators and soldering conducting wires was done as a part of this project, however the clean room fabrication was not. Hence, clean room fabrication is only briefly explained.

3.1 Material requirements

The microfluidic chip must consist of a material with a high acoustic impedance [5, 21]

which eliminated the choice of PDMS and instead, the channel geometry is etched in silicon wafers which was anodically bonded with a glass lid.

In earlier work it has been seen that the fluid phases inside the channels must possess similar acoustic properties, such as the density and speed of sound (which is related to compressibility) to generate a strong acoustic force on the particles.

In this project, deionized and filtered water (Milli-Q) is used as dispersed phase and olive oil (Classico, Zeta) which has shown good acoustic focusing [45] is used as continuous phase. The relevant material properties of these fluids are presented in table 3.1.

Table 3.1: Density and speed of sound of olive oil and water [10].

Olive oil Water Density [kg/dm^3] 0.91 1 Speed of sound [m/s] 1450 1481

3.2 Fabrication methods

The microfluidic chips evaluated in this report are both fabricated using photolithography followed by two different techniques of etching: wet etching and dry etching. Below, the two methods will be described and the process of fabrication will be presented. A schematic illustration of photolithography is found in figure 3.2.

3.2.1 Photolithography

The fabrication process starts with the creation of a pattern in a CAD software. The pattern corresponds to the shape that will be etched into the silicon wafer and needs to be transferred on to the surface of the wafer. The process of doing so is called photolithography.

From this computer generated pattern a mask is created, which consists of a glass wafer with the desired pattern shaped in chromium on top of it. When a mask is created, a coating of a resist is applied in an even layer on the wafer. The resist is a polymer which is dripped in liquid form on top of the wafer while the wafer is spinning and this evenly distributes the polymer on the surface. After, the wafer is baked in an oven to harden and the mask is aligned and placed on top of the wafer. If the resist is positive, illumination of an ultraviolet (UV) light source will weaken the polymer, making it dissolvable. After exposure with the UV light the exposed resist will be washed away with developing solution, creating a positive image of the mask pattern on the wafer. [46]

While if the resist is negative, the UV exposure cures the polymer making it harder and a negative copy of the mask is transferred to the resist. The choice of using a positive or negative resist needs to be done prior to designing a mask so that the correct pattern is transferred to the resist.

3.2.2 Wet etching

If the silicon wafer is untreated before the resist is deposited, a naturally occurring layer of SiO₂ will cover the surface. Before photolithography takes place, the wafer will be oxidized to increase the thickness of this SiO₂ layer. After photolithography the wafer is dipped in a HF solution, which will etch the bare SiO₂ . When etching of the oxide is complete, silicon from the bulk will be exposed in the shape of the desired pattern.

When the resist is washed away, the wafer is dipped in a potassium hydroxide (KOH) solution which will etch silicon while leaving the areas covered by the SiO₂. The etching is crystalline dependent, meaning that planes in different orientation within the silicon crystal etch in different rates. This will limit the way a design can be etched into the wafer, since it is not possible to etch vertical channel walls in any given direction relative to the crystallographic orientation. For this project, the wafer used was a {100} wafer with a primary flat in the <110>direction. To acquire vertical walls along the main channel, the pattern was inclined 45° angle relative the primary flat. The crystalline dependent etching is described in figure 3.1. Holes for inlets and outlets in the chip were etched during this process.

Fabrication methods

Figure 3.1: A schematic illustration describing the importance of alignment of a design before crystalline dependent wet etching. On wafer A the design is placed with its main channel ori- ented parallel to the primary flat. The etched channel (cross sections marked with a and a’) will have inclined walls. On wafer B the design is placed so that the main channel is angled relative to the primary flat. The etched channel (cross section marked with b and b’) will have vertical walls. The dotted lines (red) are the cross sections where the wet etched chip was diced and SEM images was acquired. 1 and 2 marks the position for 3.4 b) and c), respectively. However, the walls at the trifurcation are angled relative to the main channel and will be inclined after wet etching has taken place.

3.2.3 Dry etching

The wafer which was to be dry etched was first washed in preparation of photolithography.

A positive resist was applied, illuminated and developed by washing away the uncured resist. Contrary to wet etching process, the dry etching process is crystalline independent, meaning that it is indifferent to crystallographic planes in the bulk. Therefore, the mask needed not to be aligned relative to the primary flat. Etching was preformed using Deep reactive ion etching (DRIE) according to the Bosch process which is specifically developed for etching high aspect ratio features. The Bosch process is an iteration of (SF₆) plasma etching and deposition of a chemically inert polymer (C₄F₈), and these steps are repeated until the desired height is achieved. The etching was followed by several washing steps using plasma ashing and acetone to eliminate residual resist.

3.2.4 Actuation of acoustic waves and fluid setup

When the dry etched chip had been fabricated, holes were drilled for inlets and outlets. A glass wafer was anodically bonded on top of the wafers to seal the channels (and for optical transparency). The chips were then diced and separated. Silicone tubes were cut into

∼1-2 cm and glued on top of the inlet and outlet holes. Conducting wires were soldered onto a piezoelectric crystal which was used as a transducer, and was glued on the back of the chips. Both systems were lastly flushed with a hydrophobic solution (Repel-Silane, Pharmacia Biotech), and after waiting approximately 1 hour it was rinsed with oil.

Figure 3.2: An schematic illustration of the two different fabrication processes. Column a) de- scribes the progress of the wet etching. Step 1 is when oxidation of the wafer has taken place and a layer of SiO₂is covering the surface of the wafer. Step 2 is when the deposition of resist has taken place. In step 3, UV light is illuminating the wafer, through a chromium glass mask which shades specific parts of the surface. Step 4 is after the soluble resist has been removed and the mask is transferred to the resist. Step 5 is after the wafer has been dipped in HF which has etched the SiO₂ and transferred the pattern to the remaining SiO₂ . Step 6 is after KOH has etched the bare silicon and the pattern has been completely transferred to the wafer. Column b) describes the dry etching process. Step 1 is deposition of resist. Step 2 is when the resist is exposed to an UV light source through a mask which covers specific parts of the surface. Step 3 is when the uncured resist has been washed away. Step 4 is after DRIE and the pattern has been completely transferred to the wafer.

Two designs used for this project

3.3 Two designs used for this project

The result from the fabrication was two chips with different designs; a wet etched system and a dry etched system both pictured in figure 3.3. Outer dimensions are approximately 3 mm ∗ 35 mm for the wet etched chip and 9 mm ∗ 20 mm for the dry etched chip. The piezoelectric crystals are 1 mm thick with a fundamental frequency of 2 MHz. The wet etched chip was manufactured by Anna Fornell at Lund’s University. The dry etched chip was manufactured during a student project course at Uppsala University, participated by Klara Björnander Rahimi (author of this Master’s thesis project), Karolina Svensson, Anders Holmberg and Fredrik Ekström. A LOT traveler with the manufacturing process for the dry etched chip is attached in appendix A.

Figure 3.3: Four photographs of the two microsystems. a) is the view from above of the wet etched system and b) is the same system viewed from below. c) is the view from above of the dry etched system and d) is the dry etched system viewed from below. In the images one can see conductive wires soldered on to a piezoelectric crystal, which shaped as a half circle in a) and b) and shaped as a rectangle in c) and d). The silicon tubes connected to inlet and outlet holes are also visible.

Dimensions of the channels were measured on Scanning Electron Microscope (SEM) images using the software analysis tool ImageJ. The images can be seen in figure 3.4 as well as the drop split features for each chip. SEM images were supplied by Karolina Svensson and Martin Andersson at Uppsala University. The dimensions are presented in table 3.2.

Figure 3.4: a) is an optical photograph of the drop splitting feature in the wet etched chip. It is apparent that the walls angled relative to the main channels are inclined, a result from the crystalline dependency of the etching. b) and c) are SEM images of the cross sections of the main channel and the three channels following the trifurcation in which dimensions were measured for the wet etched chip (see figure 3.1). d) is the drop splitting feature in the dry etched chip. The red lines indicate where the chip was diced and SEM images was required of the cross section. 1 and 2 marks the position of e) and f), respectively. SEM images provided by Karolina Svensson and Martin Andersson, Uppsala University.

Table 3.2: Values of the dimensions of the wet etched and dry etched chip.

Width Heigth

Wet etched chip Main channel 406 µm 157 µm

Center and side channels 387 µm 164 µm

Dry etched chip Main channel 383 µm 97 µm

Center and side channels 120 µm 93 µm

4 Experimental

This chapter will describe the experimental setup and how the data was acquired from the experiments. All fluidic settings for the performed experiments are also specified here.

In the tables, the volumetric flow rate for the side channels are combined into one term, named Q_sides.

4.1 Experimental setup

The microchip was placed under a microscope (Olympus XM10) and Teflon tubes were connected to the silicone tube inlet and outlets. Oil was introduced to the system via a syringe and the outlets were also connected to syringes (Henke Sass Wolf 1 µL Soft-Ject), and the syringes were controlled via a syringe pump (NeMESYS). The dispersed phase was added to the system by a tube submerged in a 1.5 mL Eppendorf tube with 1 mL of a 5%

microparticle solution, consisting of 10 µm polystyrene microparticles (Sigma-Aldrich Co) and filtered water. The conducting wires were connected to a function generator (Agilent 33220A) connected to an amplifier (Amplifier Research 75A250), and the voltage and frequency were measured using an oscilloscope (Tektronix TDS 1004).

4.2 Acquiring data

The data that makes up the result of this thesis are of three types: the images captured in the microscope, the counted numbers of particles in a droplet and finally the measured length of droplets. For the latter two, ImageJ was used to acquire the data. By using a plug-in feature in ImageJ called "Cell counter" individual microparticles can be marked and added to a counter which summarized the particles found in either the side channels or in the center channel. Since this is a purely visual process, it is essential that the particles have enough contrast to be seen in the image. Since the droplets have a rounded a shape, the interface between the dispersed phase and continuous phase were visualized with a dark boundary. Microparticles close to the interface by the sides of the droplet would be hidden from view in this dark line and not accounted for in the analysis. This problem is related to the optical microscope, it was constant for all the experiment and hence, a systematical

Figure 4.1: Photograph of a microchip mounted in the microscope. The chip is taped on the transparent plastic stage with black tape and tubes sticking out from below.

error.

Since no particle filter was used it was also possible that grains of dust captured in the oil accidentally could be assumed to be microparticles. However, since the microparticles were very characteristic in size, shape and contrast this was a problem of minor influence when the beads were counted.

Also, in order to calculate the concentration factor in section 4.4.2, the volume of the droplets was needed. The volume was obtained by using formula constructed by Musterd et al. [35] which required the channel geometries and the length of each droplet, see section 2.4.2. In ImageJ a straight line was drawn between the two endpoints of a droplet, which length was assumed to be equal to the length of the droplet, see figure 4.2. This type of data gathering is highly dependent on good visual contrast between the channel, the continuous phase and the dispersed phase in the system in order to place the endpoints of the straight line as close to the boundaries of the droplet.

When converting from pixels to µm the scale used was 0.6211 pixels µm⁻¹which im- plies that a droplet measured to 800 µm would have an error of 1% if only one of the two measurement points are placed 5 pixels away from the droplet boundary. Worth noting however, is that it is not possible to find the actual volume of the droplet and this method in addition to the formula provided by Musterd et al. is only an estimation of unknown actual error margin.

Comparison between the wet etched chip and the dry etched chip

Figure 4.2: A capture of the data collection process. Using ImageJ, it is possible to keep track of microparticles collected in the center, and side droplets by coloring them in different colors. The yellow line indicates the straight line which is measured to find the length of a droplet.

4.3 Comparison between the wet etched chip and the dry etched chip

Different sets of fluidic settings were tested when comparing the wet etched and the dry etched systems. The voltage (V_pp) is directly proportional to the acoustic radiation force, see equation 2.2, and is held constant for all comparative experiments so that the same force is applied in all cases. The optimal frequency of the piezoelectric crystals is theoretically 2 MHz and in practice the best alignment occurred when the signal was set to 1.87 MHz for the wet etched chip and 1.90 MHz for the dry etched chip. When presenting fluidic settings in this section, a positive value indicates injection and a negative indicates withdrawal from the syringe pumps.

4.3.1 Constant volumetric flow

The volumetric flow rate and the fluidic settings were constant in this experiment, at two different settings: 10.5 µL/min and 18 µL/min. The fluid settings used in this experiment are presented in table 4.1:

Table 4.1: Fludic setup for experiments with the same volumetric flow rate.

[µL/min] Wet etched chip Dry etched chip

Q_tot Q_water Q_oil Q_center Q_sides Q_water Q_oil Q_center Q_sides

10.5 5 5.5 -3.5 -7 5 5.5 -3.5 -7

18 6 12 -6 -12 6 12 -6 -12

4.3.2 Constant droplet speed

Secondly, the recovery for the two systems was compared with respect to droplet speed in the main channel prior to droplet splitting. The volumetric flow rate is adjusted by the pumps injecting and extracting fluid into and from the microfluidic system, while the droplet speed is found by using the volumetric flow rate in combination with the cross sectional area of the channel. The conversion between volumetric flow rate and droplet speed can be seen in equation 4.1.

Q

A_{c r os s} → v : µL

mi n∗ m² → m

s (4.1)

Because of shearing between droplet boundary and side wall of the channel when the droplet is moving, the speed of the droplet is likely to influence turbulence inside droplets.

When the droplet progression speed is constant, turbulence occurring inside droplets might also be similar and alignment of particles will be equally disturbed. This might result in a fairer comparison between the two designs. This was done at two different settings where droplet speed was set to 0.47 cm/s. The droplets produced at these settings had two differ- ent length-to-width ratios, where one was small (L ≈ 3 W) and the other was long (L > 3).

No value is given to the longer droplets since their endpoints extended outside of the field of view of the microscope and was impossible to measure. Settings are presented in table 4.2

Table 4.2: Fluidic setup for experiments when droplet speed is constant and droplet length-to- width ratio is varied. All values are presented in µL/min.

l e n g t h

wi d t h Wet etched chip Dry etched chip

Q_water Q_oil Q_center Q_sides Q_water Q_oil Q_center Q_sides

3 8.5 9.5 -6 -12 5 5.5 -3.5 -7

>3 6 12 -6 -12 2.5 8 -3. 5 -7

4.4 Evaluation dry etched chip

Below are the experimental settings described for the evaluation of the performance of the dry etched chip. The influence of acoustic force was varied for three flow settings and different flow settings was tested to see how concentration were changed within a droplet.

4.4.1 Influence of acoustic force at different throughput

The throughput of the dry etched system was tested with three volumetric flow rates:

3 µL/min, 10 µL/min and 18 µL/min. Meanwhile, the voltage applied was varied from 0 V_pp, 10 V_ppup to 20 V_pp. Fluidic settings are presented in table 4.3.

Evaluation dry etched chip

Table 4.3: Fluidic settings for three different volumetric flow rates in the dry etched system. All values are presented in µL/min.

Q_tot Q_water Q_oil Q_center Q_sides

3 1 2 -1 -2

10.5 5 5.5 -3.5 -7

18 9 9 -6 -12

4.4.2 Concentration factor

Recovery data does not take volume reduction into account and a concentration factor was developed to see if this model would be suitable for optimizing the enrichment experiment.

This concentration factor was found dividing the enriched droplets concentration c_c, by the concentration of the incoming droplet c₀, i.e. the total number of particles n divided by the total volume v (which is the sum of the center volume v_c and side droplet volumes v_s) before the droplet is split. The concentration factor is expressed in equation 4.2.

c f = c_c c₀ =

ncvc nc+2ns vc+2vs

=n_c+ (^2vsnc_vc )

2n_s+ n_c (4.2)

The flow settings were set to stepwise minimize the withdrawal rate through the center outlet to create smaller and smaller center droplets while collecting particles in them. By decreasing volume of the center droplets while still enriching them a high concentration factor would ideally be acquired. The flow settings are presented in table 4.4. Voltage applied was 20 V_pp.

Table 4.4: Fluidic settings for concentration factor experiments. The leftmost column is the relation between the two side channels and the center channel. The withdrawal rate in the center channel is decreased 5 times, starting with a withdrawal rate which is half that of the two side channels down to a tenth of the two side channels’ withdrawal rate. All values presented are in µL/min.

side: center: side Q_water Q_oil Q_center Q_sides

1:1:1 5.5 5 -3.5 -7

1:¹/²:1 5.5 5 -2.1 -8.4

1:¹/³:1 5.5 5 -1.5 -9

1:¹/⁴:1 5.5 5 -1.17 -9.33 1:¹/⁵:1 5.5 5 -0.95 -9.55 1:¹/¹⁰:1 5.5 5 -0.5 -10

4.4.3 Manipulation of cells

Lastly, a cell experiment was performed. Yeast cells were extracted from baking yeast (KronJäst Söta Degar) by mixing a small volume of yeast with filtered water. The total volumetric flow rate in the main channel was set to 5 µL/min and a voltage of 40 V_ppwas applied.

5 Results and discussion

This chapter is dedicated to present the results from the experiments performed. Because of the practical nature of the project, a discussion regarding the presented data and aspects regarding the experimental accomplishment will follow immediately for every section in this chapter.

All data presented in this chapter as graphs are specified in appendix B.

5.1 Comparison between the wet etched chip and the dry etched chip

Common design features and functions

Magnitude of acoustic force

At the applied voltage of 25 V_pp the droplet generation in the wet etched chip was dis- rupted by the acoustics. Droplets entering the center channel of the droplet split coalesced into larger droplets, more frequently with acoustics applied compared with when acous- tics was not applied which made it impossible to retrieve recovery data for the individual droplets, see figure 5.1. In some experiments, while testing the effect of the acoustics, droplets seemed to be generated with a larger volume. Fornell et al. [10] used the same chip design and experimental setup as in this project and achieved particle alignment and droplet splitting at 25 V_pp, but this was not successfully repeated during the experiments in this project. Acoustics have been used to manipulate water droplets in oil [38, 42] but to what extent the acoustics is affecting droplet generation and coalescence was not further investigated in this report.

Voltage was set to 10 V_pp, since the acoustic radiation force generated at this setting was able to align particles in the wet etched chip while not disrupting the droplet splitting in a dominating way. It was noticed that droplets appeared to be stable during the droplet splitting when using water with food dye (red color, Ekströms) as a dispersed phase. It was therefore tested to see whether adding the food dye containing glycerol would act as a droplet stabilizer. However, when microparticles were added the instability of the droplet split appeared anew and the effect of the dye was not remnant. The dye was not used for

Figure 5.1: In the wet etched split it was observed that an applied acoustic field influenced the stability of the droplet split. a) Acoustics is not applied and the incoming droplets are evenly split in the droplet split. b) Acoustics is applied (25 V_pp) and the droplet split is not stable. The center droplets are coalescing into a stream. Particle alignment is visible in the photo.

the bead experiments.

Number of droplets used as data points

Even though the magnitude of the acoustic force was minimized to the point where droplets coalesced less often, there was still coalescence occurring in the wet etched system. Since it was desirable to collect data for the same amount of data points (which in this case is the number of droplets) for both systems, the comparative data is based on 20 consecutive droplets.

Number of microparticles in an individual droplet

It was desired to have a similar amount of microparticles inside every droplet. However, this was a challenging task to establish. The bead solution was added to the microfluidic system via a tube submerged in an Eppendorf tube. Sedimentation in the tube was a re- occurring problem, which made it necessary to shake the Eppendorf tube before the start of each experiments to keep the microparticles in suspension. A risk of beads sedimenting was also present inside the tube connected to the chip, which required the tube to be short so that there would be no kinks where particles could settle. The diameter of this tube was thinner than the other tubes (0.3 mm while the other was 0.8 mm), to increase the volumetric flow rate to prevent sedimentation.

Comparison between the wet etched chip and the dry etched chip

5.1.1 Recovery

Same volumetric flow rate 10.5µL/min

When the volumetric flow rate was held constant at 10.5 µL/min and the settings were identical for the two designs, stable droplet splitting was never observed in the wet etched chip. During a majority of the time the incoming droplets were divided into two daughter droplets, which were uneven in size, see figures 5.2, a) and b). When a daughter droplet passed through one of the side channels, a pressure inequality appeared between the side channels which withdrew the next incoming droplet through the other side channel.

During brief moments, three daughter droplets appeared. At those moments there was always a big size difference between the droplets, where the center droplet was significantly larger than the side droplets. However, these moments passed quickly and for most of the time, the droplet splitting was unstable.

Stable droplet splitting was observed for the dry etched chip at this setting, dividing the incoming droplet into three equally sized droplets, while retaining the acoustic focus during the entire droplet splitting event, see figures 5.2 c) and d).

The consequence of the uneven droplet splitting in the wet etched system was a com- plete loss of particle alignment during droplet splitting and the resulting recovery in the center outlet was less than one fourth compared to the dry etched system. For the dry etched system, particle alignment was successful and the center droplets were enriched with microparticles, figure 5.2 d).

Figure 5.2: Figures a) and b) are optical images of the wet etched droplet split where droplets are divided into two unevenly sized droplets (red dashed lines indicate the pair of daughter droplets that was split from the same incoming droplet). In a) acoustics is not applied, and in b) acoustics is applied. There is no visible alignment of the microparticles (marked with red rings) which are randomly distributed in the droplets. Figures c) and d) are optical images of the dry etched droplet split when droplets are divided in three equally sized droplets. In c) acoustics is not applied and the microparticles are randomly spread out. In d) acoustics is applied and the majority of the microparticles are collected in the center droplet.

The instability of the wet etched system is indicated from the high standard deviation in figure 5.3 which is very large compared to the standard deviation of the dry etched chip with the same setup.

Comparison between the wet etched chip and the dry etched chip

Figure 5.3: Recovery data of the wet etched chip (yellow striped) and the dry etched chip (green fill) when the volumetric flow rate is constant at 10.5_{mi n}^µL.

Same volumetric flow rate 18µL/min

Stable droplet splitting was observed and recovery data was collected for the wet etched sys- tem when the volumetric flow rate was 18 µL/min. However, it was not observed for the dry etched system. The two side channels split differently sized daughter droplets, where one of the side droplets was significantly smaller than the other two and the second side droplet was very long, see figure 5.4 a) and b). The longer side droplets expanded outside the field of view and made recovery of microparticles inside that droplet impossible to col- lect. This inequality in size seemed not to be dependent on a certain channel, since when the system was stopped and started anew the small and long daughter droplets switched sides. At one occasion this unequal daughter droplet production switched sides during an ongoing experiment, meaning that from one moment to the next the small daughter droplets occurred on one side and then the other. This indicates that the instability of the droplet splitting was not dependent on particular defects that might occur inside one of the microchannels but rather an instability of the geometry of the system in itself.

Alignment of particles was successful in the wet etched chip and it was also possible to focus particles during droplet splitting, see figure 5.4. In the dry etched chip however, par- ticle alignment did not occur at these fluidic settings. In section 5.2.1, the same volumetric flow rate is evaluated at a different fluidic setting with a higher acoustic force for the dry etched chip. Since the outer dimensions of the two chips are different, there might be a difference in the resulting acoustic force generated in the two chips when the same volt- age is applied. The dry etched chip has a larger total area compared with the wet etched chip which is thinner since it has been diced closely to the main channel. At present there is nothing but discussions regarding the bulk materials absorption of the transmitted en- ergy from the actuator and no literature has yet been published. Bulk material absorption

would reduce the magnitude of the pressure waves appearing inside the microchannels and in this case the acoustic radiation force would be larger in the wet etched chip since the device is smaller. In addition, since the microchannel in the dry etched chip is thinner than in the wet etched chip, the droplet speed is higher which would generate more movement inside the droplet which, in turn, also would counteract the alignment to a greater degree.

Since recovery data was not collected for the dry etched chip at this setting, a compar- ison is done between the chips in section 5.1.2. The same data for the wet etched chip is compared with data from the dry etched chip when its fluidic settings are adjusted so that droplet speed is constant in the comparison.

Figure 5.4: a) and b) is when the volumetric flow rate is 18 µL/min in the wet etched chip. In a) acoustics is not applied and particles are visible in all three daughter droplets. In b) acoustics is applied at 10 V_pp and aligned particles are visible in the center daughter droplet. In c) and d) the instability of the dry etched chip is visible. c) is without acoustics applied and d) is when acoustics is applied.

Comparison between the wet etched chip and the dry etched chip

Constant droplet speed

When droplet speed was the constant factor of the two systems, the relative size of droplets was varied. First droplet length-to-width was adjusted to approximately 3 and later longer droplets were generated with a smaller length-to-width ratio. Droplet speed was constantly held to 0.47 cm/s during both cases.

Length-to-width set to 3

Incoming droplet length-to-width ratio is set to 3 by adjusting the input of oil for the systems, see figure 5.5. When the droplets had this size, the dry etched chip achieved stable droplet splitting while the wet etched chip did not.

The instability of the wet etched chip is similar to when the volumetric flow rate was set to 10.5 µL/min in figure 5.2 a) and b) when droplets were divided into two daughter droplets instead of three. The recovery for both systems are retrieved and presented in figure 5.6. The interval spanned by the standard deviation is once more an indication of the unstable droplet splitting, similarly to when the volumetric flow rate was 10.5 µL/min in figure 5.3. When the incoming droplets are small, the dry etched system seem to be more successful.

Figure 5.5: In a) and c) incoming droplets are shown and measured to have a length-to-width ratio of approximately 3. a) is the incoming droplet in the wet etched system where microparticles are aligned, and b) is the droplet split in the wet etched split when droplet splitting is unstable and microparticles are no longer aligned (red dashed ovals indicate the pair of daughter droplets that was split from the same incoming droplet). c) is the incoming droplet in the dry etched system where microparticles are aligned. d) is the droplet split in the dry etched chip where majority of microparticles are in the center daughter droplet.

Comparison between the wet etched chip and the dry etched chip

Figure 5.6: Recovery data of the wet etched chip (yellow striped) and the dry etched chip (green fill) when the droplet velocity and length-to-width ratio is 3.

Length larger than width

When droplet length increased the opposite situation occurred from when length-to-width ratio was set to 3. Stable droplet splitting was observed in the wet etched chip while it was not observed for the dry etched chip, see figure 5.7 b) and d). Particle alignment was successful in the wet etched chip for which recovery data was retrieved, while it was not aligned in the dry etched chip, see figure 5.7 a) and c). It was not possible to retrieve recovery data from the experiments performed in the dry etched chip. Recovery data from the wet etched chip is presented and compared in section 5.1.2.

Figure 5.7: a) is the incoming droplet and b) is the droplet split in the wet etched system. Particles are aligned in a) and droplet splitting is stable and every incoming droplet is divided into three daughter droplets. c) and d) are the incoming droplet and the droplet split in the dry etched system. Particles are not aligned and droplet splitting is uneven and unstable. For both cases the incoming droplets are long compared their width, and acoustics is turned on in every image at 10 V.

5.1.2 Comparing two stable events

Stable droplet splitting was observed for both chips, at different volumetric flow rates and differently sized droplets. Constant factors was droplet speed and the voltage applied to generate the acoustic force. Recovery and standard deviation from the successful cases of droplet splitting are displayed in figure 5.8.