Micromachining of microfluidic systems using a nanosecond laser

(1)

UPTEC Q 19011

Examensarbete 30 hp November 2019

Micromachining of microfluidic systems using a nanosecond laser

Process optimization and application

Per Söderbäck

(2)

Teknisk- naturvetenskaplig fakultet UTH-enheten

Besöksadress:

Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0

Postadress:

Box 536 751 21 Uppsala

Telefon:

018 – 471 30 03

Telefax:

018 – 471 30 00

Hemsida:

http://www.teknat.uu.se/student

Abstract

Micromachining of microfluidic systems using a nanosecond laser

Per Söderbäck

Microfluidics is a field of research that enables the manipulation of fluids in the submillimetre length scale. The technology allows the development of lab-on-a-chip devices, which are miniaturized systems for chemical and biological analysis. Currently, the conventional manufacturing methods for these systems require multiple time-consuming steps. Therefore, focus has shifted towards laser micromachining as an alternative method. Direct laser writing would circumvent many of the steps required for the conventional methods, drastically reducing the process time.

In this Master thesis project, it was shown that microfluidic chips can be manufactured using a Nd:YVO4 (532 nm) nanosecond laser system. The process was optimized for silicon and

borosilicate glass substrates. Acoustic focusing of polystyrene beads was demonstrated for a system etched in silicon. The optimized process used a power of 50%, a frequency of 10 kHz, a scan speed of 60 mm/s with triple lines as fill type and it had an etch rate of 4.3 µm/pass. Processed wafers were cleaned in buffered HF and bonded using anodic bonding as well as adhesive bonding. Processing of glass proved unpredictable, resulting in cracks and chippings. However, in- and outlets were successfully etched through thin glass wafers. It was found that crucial factors for the process were to control the focus, positioning of structures, structure orientation and the pulse separation for a uniform distribution of pulses. Based on the results, it is

estimated that the manufacturing process could be done in two to three days using the laser micromachining process.

Examinator: Åsa Kassman Rudolphi Ämnesgranskare: Maria Tenje

Handledare: Milena Moreira, Anna Fornell

(3)

Tillverkning av mikrofluidiksystem med laserskärare

Mikrofluidik är ett forskningsfält som handlar om att hantera små volymer av en vätska i mycket små system. Tekniken möjliggör bland annat utvecklingen av så kallade lab-on-a-chip system. Dessa system kan användas för att utföra komplexa processer som exempelvis kemisk och biologisk analys, vilka annars skulle kräva tillgång till ett labb.

Mikrofluidiksystem tillverkas konventionellt med fotolitografiska processer kombinerat med våt- eller torretsning. Dessa metoder utvecklades ursprungligen för halvledarindustrin och lämpar sig väl då man vill tillverka stora mängder av en struktur. Dessvärre är metoden omständlig och kräver många olika steg.

Vidare kräver metoden att man har en mask, vilket är en slags schablon, av mycket hög kvalitet för att överföra sitt mönster till substratet. Denna mask måste ofta beställas från en leverantör vilket är en extra kostnad rent ekonomiskt, men även tidsmässigt då den måste levereras. För att kringgå dessa begränsningar har mycket forskning gjorts om möjligheten att använda laserskärare för att skära ut dessa strukturer. Detta skulle potentiellt kunna ersätta en lång och dyr tillverkningsmetod med en snabbare och mer flexibel metod.

Lasertekniken används redan för många olika applikationer inom industrin, men tillverkning av mikrofluidiksystem har främst gjorts med mycket dyra lasersystem då dessa ger högre kvalitet på de skurna struk- turerna. I denna studie undersöktes möjligheten att använda en billigare Nd:YVO₄-laser för att skära ut strukturer ur kisel och glas. Denna typ av lasersystem har dock längre laserpulser vilket medför vissa ut- maningar för processen. När lasern smälter och förångar materialet kommer smältan att pressas ut och bilda vallar kring den skurna strukturen. Vidare har dessa strukturer ofta väggar som sluttar och är run- dade samt en bottenyta som är ojämn till följd av var laserpulserna träffat. För att skära ut fungerande mikrofluidiksystem krävs att dessa problem minimeras. Detta görs genom att optimera ett antal parametrar för lasersystemet exempelvis effekt, pulsfrekvens och svephastighet.

I denna studie optimerades processen för att skära strukturer i både kisel och glas, vilket är vanliga material inom mikrofluidiken. De optimerade processerna testades sedan genom att skära ut ett mikrofluidiksystem, designat för anrikning av partiklar i en vätska. En skiss av systemet kan ses i Figur 1. (a). Anrikning sker genom en teknik som kallas akustofores, vilken gör det möjligt att styra mikropartiklar i en vätska.

Styrningen sker genom att systemet utsätts för högfrekventa vibrationer. Om bredden på mikrokanalen är lika med halva våglängden hos den pålagda frekvensen genereras ett tryck- och hastighetsfält i kanalen.

När partiklarna i vätskan utsätts för tryck- och hastighetsfältet flyttas de till mitten av kanalen där de sedan kan anrikas. Anrikningen finns illustrerad i Figur 1. (b).

Resultatet av denna studie visar att det billigare lasersystemet går att använda för att skära fungerande mikrofluidiksystem i kisel. Genom att utsätta det tillverkade mikrofluidiksystemet för högfrekventa vibrationer kunde partiklarna i vätskan anrikas, se Figur 2. (a) och (b).

Vidare visar studien att den framtagna processen för glas ej är lämplig att använda för tillverkning av mikrofluidiksystem. Processen var oförutsägbar och resulterade i strukturer med bland annat sprickbildning och ojämn bottenyta.

(4)

(a) (b)

Figur 1: (a) System för anrikning av partiklar i vätska. (b) Schematisk representation av anrikningspro- cessen. Partiklar i vätska (I), system utsätts för högfrekventa vibrationer (II), partiklar flyttas mot kanalens centrum (III) varpå de fokuserade partiklarna kan extraheras (IV).

(a) (b)

Figur 2: Flöde av vätska med partiklar (a) utan vibrationer och (b) med vibrationer.

Examensarbete 30 hp

Civilingenjörsprogrammet i Teknisk fysik med materialvetenskap Uppsala Universitet, November 2019

(5)

Abbreviations

LOC Lab-on-a-chip

PDMS Polydimethylsiloxane LMM Laser micromachining UV Ultraviolet radiation IR Infrared radiation

LIBEW Laser-induced backside wet etching PMMA Polymethylmetacrylate

HAZ Heat affected zone

SEM Scanning electron microscope buffered HF Buffered hydrofluoric acid IPA Isopropyl alcohol

DI water Deionized water KOH Potassium hydroxide

(6)

1 Introduction

Microfluidics is a field of research that enables the manipulation of fluids in the submillimetre length scale.

The technology allows the development of lab-on-a-chip (LOC) devices, which are miniaturized systems capable of performing complex chemical and biological analysis [1, 2]. Currently, the conventional manufacturing methods for these systems, in glass or silicon, utilises photolithography techniques developed for the semiconductor industry while elastomeric micromoulding can be used to form systems in polydimethylsiloxane (PDMS) [1].

Photolithography is a multi-step process for transferring a pattern onto a substrate. The process involves steps such as oxide growth, application-, development- and removal of photoresist before the pattern can be etched onto the substrate. Furthermore, development of the photoresist requires a high-quality mask, similar to a stencil, to transfer the desired pattern [3]. For prototyping, this means that a new mask must be manufactured for each iteration, often by a third-party, resulting in increased costs both monetary and time-wise. Etching of the pattern is usually done using wet or dry etching which both are top-down surface structuring techniques, i.e. that structures are etched at the surface of the substrate and that the system must be sealed after etching.

With recent advances in laser technology, focus have shifted towards laser micromachining (LMM) as an alternative method for the manufacturing of microfluidic systems. The method can be used for drilling, cutting, welding and ablation for a wide range of materials and it is routinely used in areas such as the automobile and medical industries, production of semiconductors and a variety of microfabrication tasks [3, 4, 5]. The use of LMM for direct writing of structures would circumvent many of the steps required for photolithography as there is no need for transfer of pattern or development of photoresist. Furthermore, no mask is required for new iterations of a design. This could drastically reduce the process time, making it an attractive tool for research. Additionally, LMM using a femtosecond laser enables etching of 3D structures in the bulk of a glass material circumventing the sealing of structures [6]. However, the more advanced femto- and picosecond lasers are expensive.

This Master thesis project investigated the ability to LMM microfluidic system on silicon and borosilicate glass substrates using a more economic Nd:YVO₄laser system with pulses in the nanosecond regime.

One of the greater challenges in LMM is to achieve a constant depth with smooth surfaces of the etched structures [7]. Therefore, the LMM process was optimized to achieve well defined microfluidic structures with an uniform bottom surface as well as minimal taper, recast layer, cracking and striations while dis- regarding the process time. The process was then used to manufacture and evaluate three types of chips designed for acoustophoresis, a method of acoustically manipulating particles in microfluidics often used for enrichment, filtration/removal or separation of particles [8]. This type of system was selected as a proof- of-concept since it has simple structures while still highly dependent on the channel profile and dimensions.

Additionally, it offers a clear visual evaluation.

2 Laser micromachining

This chapter addresses some fundamental knowledge regarding laser technology and LMM using different types of laser systems as well as a comparison of the characteristic etching profiles between conventional processing methods and LMM.

2.1 Fundamentals of lasers

Laser is an acronym of Light Amplification by Stimulated Emission of Radiation. A laser is a source of coherent, monochromatic radiation. This means that the radiation is of the same wavelength, usually in the range of ultraviolet (UV) to infrared (IR), and that the waves have a constant phase difference with the same frequency [9].

Laser generation requires three processes, namely populated inversion, stimulated emission and amplification [9]. A simplified model of a laser system is illustrated in Figure 3.

Populated inversion refers to a non-equilibrium distribution of electrons in a lasing medium (Fig- ure 3. III), where higher energy states are more populated than the lower states [9]. This is achieved by pumping (Figure 3. IV) the lasing medium. During pumping, electrons are excited to higher energy states by applying an electric current (electrical pumping) or by absorption of radiation from a secondary source (optical and diode pumping). Optical and diode pumping are used for solid-state lasers while electrical pumping is used for gas lasers [9].

(9)

Figure 3: Schematic representation of a laser system. Mirrors in the laser cavity (I, V), Q-switch (II), lasing medium (III), pumping (IV) and the beam guiding method using deflector mirrors (VI) or by moving the work table (VII).

Laser radiation is generated from the excited electrons in the lasing medium through a process called stimulated emission. Here, an incoming photon propagating through the pumped laser medium triggers the excited electrons in the material to return to the lower energy state while releasing radiation with the same frequency, direction and phase as the incoming photon [9].

Before the laser radiation is emitted from the system, it is amplified in the laser cavity. The laser cavity usually consists of two mirrors (Figure 3. I, V), enclosing the lasing medium. By reflecting the radiation back and forth through the lasing medium, the radiation is amplified [9]. Usually one of the mirrors (Figure 3. V) in the cavity is partially transparent which allows laser radiation to be emitted as a continuous wave (CW) or as a pulsed laser.

A common way of generating a pulsed laser is by using a Q-switch (Figure 3. II). The Q-switch can be compared to a shutter which controls the ability to store energy within the cavity [9, 10]. This is done by deflection or diffraction of the light, in order to prevent amplification in the cavity. When the light is allowed to amplify in the pumped cavity, the stored energy can be rapidly discharged [9, 10]. By using this method, the lasing medium can reach high gain before lasing is initiated which results in a short and very intense burst of light [10].

There are usually two methods for guiding the laser beam during processing. Either the work table (Figure 3. VII) is moved in the XY-plane, or the beam is directed by beam deflectors (Figure 3. VI) which sweeps the laser across the substrate [4].

The laser systems are classified by the type of lasing medium that is used. The lasing medium determines the active wavelength of the laser. Laser systems used for micromachining usually have pulse durations, i.e. the time during which the laser/material interaction takes place, in the nano- to femtosecond regime with frequencies in the order of kHz [11]. Some typical systems used to generate laser radiation are excimer, solid-state and CO2lasers where solid-state and excimer lasers are the most common types of lasers used for micromachining in medical and electronic industries [5, 11].

2.1.1 Excimer lasers

Excimer lasers are pulsed gas lasers that produce radiation with wavelengths ranging from 193 to 351 nm.

These kinds of systems are used for patterning, cutting and structuring materials such as ceramics, glasses and polymers. Structuring can be done with a depth resolution of 0.1 µm and a spatial resolution of 1 µm [11]. Pulse durations for excimer lasers are usually in the range of 5 to 500 ns with low repetition frequencies below 1 kHz [12].

2.1.2 Solid-state lasers

Solid-state lasers broadly use crystalline solids and glasses as lasing medium and emit radiation with wavelengths ranging from 400 nm to 3 µm [11, 12]. Some examples of solid-state lasers are Nd:YAG and Nd:YVO₄. These types of lasers have high ablation efficiencies and are able to ablate a wide range of materials such as metals, glasses, diamond, silicon and rubber [11, 13]. Solid-state lasers are typically pulsed with pulse durations ranging from nano- to femtoseconds [9, 14]. Femtosecond solid-state lasers also enable machining of structures in the bulk of a transparent material [13].

(10)

Processing of transparent glass materials using longer pulse durations have proven to be challenging as the radiation is not absorbed well by the materials [14]. However, it has been shown that materials that are transparent to the active wavelength of the laser can be etched using a process called laser-induced backside wet etching (LIBEW). For this process, the laser is focused on the opposite side of the substrate, through the bulk, and it is absorbed at the glass/liquid interface. The process has been illustrated by Cheng et al., using a solution of Rose Bengal organic dye and acetone to etch soda-lime glass [15] and Yen et al. using liquid metallic absorbers such as Ga and eutectic In/Ga demonstrated for soda-lime glass and quartz [16].

2.1.3 CO₂lasers

CO₂ lasers uses a gas lasing medium that is a combination of carbon dioxide, nitrogen and helium to produce radiation with a wavelength in the range of 9.3 to 11 µm [14]. This wavelength is well absorbed by polymers, ceramics and glasses, making it suitable for processing said materials [3, 11]. Typical pulse durations are in the microsecond regime [11, 14].

CO₂lasers have been used to successfully etch microfluidic structures in Polymethylmethacrylate (PMMA) as shown by Snakenborg et al. [17].

2.2 Laser induced ablation

When laser radiation interacts with a material, at sufficiently high laser intensity, energy is absorbed by electrons in the material and ablation is initiated. The ablation process is schematically displayed in Fig- ure 4. (a). Depending on the interaction time, i.e. pulse duration, the mechanisms of material removal varies [11, 12]. If the pulse duration of the laser is greater than the heat diffusion time in the material, typically in the pico- to nanosecond regime, the material is heated to boiling temperatures. The material is then removed by vaporization (Figure 4. (a). II) as well as boiling, where globs of melted material are ejected from the heated area (Figure 4. (a). III) [3, 9, 18]. The heat diffusion results in reduced accuracy as material outside of the laser spot is also affected by melting. Around the etched feature, there is a large zone called the heat-affected-zone (HAZ) (Figure 4. (a). IV)) where the heat diffusion can cause micro or macro cracks due to thermal stresses [3].

(a) (b)

Figure 4: (a) Schematic representation of the ablation process using long pulse durations. Incident pulsed laser (I) interacts with the surface. Radiation is absorbed and the substrate is melted and vaporized (II).

Melted material is displaced or ejected (III). Heat is diffused into parts of the bulk material, referred to as HAZ (IV). (b) Characteristic etch profile acquired from LMM using long-pulsed laser. A lip of recast material outlines the structure (I), the profile is rounded with tapered walls (II) and surface debris can be found surrounding the structure (III).

If the pulse duration is shorter than the heat diffusion time, typically femtosecond or shorter, the heat does not diffuse away from the laser spot and the temperature instantly rises into the plasma regime [3]. As heating is contained within the targeted area, structures etched using these short pulsed lasers do not suffer from any HAZ or ejected material.

Displayed in Figure 4. (b). is a schematic representation of a typical etch profile for material processed using a long-pulsed laser. Etch profiles associated with long-pulsed LMM are characterised by a lip of recast material (Figure 4. (b). I), tapered walls (Figure 4. (b). II) and condensed or splattered surface debris (Figure 4. (b). III) surrounding the structure.

(11)

This can be compared to etch profiles acquired from conventional processing methods such as isotropic wet etching (Figure 5. (a)) where material is etched at the same rate in all directions, anisotropic wet etching (Figure 5. (b)) where the etch rate is dependent on the crystal orientation of the substrate and dry etching using the deep reactive etching (DRIE) method (Figure 5. (c)) [3].

(a) (b) (c)

Figure 5: (a) Characteristic profile of wet-etched structures using isotropic and (b) anisotropic etching. (c) The profile of a dry-etched structure using the Bosh process.

3 Microfluidic system for acoustic manipulation of particles

This chapter covers the concept of acoustophoresis as well as the function and design of the three chip types developed for this study.

3.1 Acoustophoresis

Acoustophoresis is a label free method for manipulating cells and particles in fluids using acoustics. The method is used for applications such as enrichment, filtration and particle separation [19].

In this study, a bulk acoustic wave (BAW) system was designed and fabricated using the LMM process.

This type of system is usually fabricated with materials that have a high acoustic impedance, such as glass or silicon [20]. Acoustic focusing is done by generating a standing wave field inside a cavity or a channel using a piezoelectric transducer attached to the chip, causing it to resonate with the frequency of the applied sound waves [19, 21]. If said cavity or channel is filled with a fluid and has a width that is a multiple of half the wavelength of the sound, the resonating walls of the cavity or channel generates a standing wave through the fluid. This condition is displayed in Equation 1. where W is the width, n is an integer and λ is the wavelength of the applied sound. The wavelength is given by Equation 2. where c is the speed of sound in the fluid and f is the applied frequency [20].

W =nλ

2 (1)

f = c

λ (2)

When a particle in the fluid is introduced to said channel, sound waves are scattered off the particle, resulting in acoustic radiation forces acting on the particle [20]. The applied forces pushes the particle either towards a node or a anti-node, and the direction of motion is determined by the acoustic material properties of the particle in relation to the fluid [19, 20]. The polystyrene micro beads used in this study focus towards the node. For an efficient focusing of particles, it is optimal to have a resonance channel with well defined dimensions and vertical walls in order to align the particles at the nodal plane within the channel.

(12)

3.2 Chip types

The three chip designs developed for this study are displayed in Figure 6.

The first type (Figure 6. (a)), hence referred to as the Silicon chip, is designed so that both structures and in- and outlets are etched in a 525 µm thick silicon wafer. Etching is done in one session and the system is sealed by bonding it to a glass wafer.

The second type (Figure 6. (b)), hence referred to as the Glass chip, is similar to the Silicon chip.

Structures are etched in a 1000 µm thick borosilicate glass wafer in order to give fully transparent structures.

However, in- and outlets are etched through a thin glass wafer that is bonded to the thick glass wafer. This requires two sessions of etching, one for structures and one for in- and outlets.

The third type (Figure 6. (c)), hence referred to as the Sandwich chip, is designed for a perfectly flat and transparent bottom and top of the system. This is achieved by etching the structures straight through a 300 µm thick silicon wafer. The silicon wafer is then sandwiched between two glass wafers. In- and outlets are etched through one of the glass wafers after bonding. This chip type requires two sessions of etching as well as bonding of two wafers.

(a) (b) (c)

Figure 6: Illustration of the three chip types. (a) The Silicon chip, consisting of a structured silicon wafer with glass cover. (b) The Glass chip consisting of a thick structured glass wafer and a glass cover with etched in- and outlets. (c) The Sandwich chip consisting of a structured thin silicon wafer sandwiched between two glass wafers. In- and outlets in the top glass wafer.

3.3 Design of the system

The microfluidic system developed for this study was designed for enriching a suspension of polystyrene micro beads. The system consists of a resonance channel that split into three smaller outlet channels. The concept of the acoustic focusing and enrichment of the particles is schematically illustrated in Figure 7.

The particle solution is injected into a resonance channel (Figure 7. (I)), ultrasound is applied at the first harmonic resonance frequency of the resonance channel (Figure 7. (II)), the beads align in the centre of the resonance channel (Figure 7. (III)) and are then extracted through the middle outlet channel (Figure 7. (IV)).

Figure 7: The function of the designed system. Particles enter the system (I), a resonance frequency is applied (II), the particles are focused (III) and extracted at the intersection (IV).

The width of the resonance channel was set to 380 µm. That corresponds to a theoretical resonance frequency of 2 MHz. The frequency was calculated using equation 1. and 2. for the first harmonic (n = 1)

(13)

and the speed of sound in water (c_WA = 1497 m/s) [20]. The width of the three outlet channels were set to 127 µm.

The total length of the chip had to be less than 27 mm in order to allow for the etching of multiple structures with correct positioning in relation to the centre of the work table, as explained in Section 5.2.

Additionally, silicon etching was restricted in the centre of the work table. In- and outlets were separated enough for tubing to be attached as well as allowing for clear imaging of the intersection without shadowing from tubing or glue residue.

The Silicon and the Glass chip both used the design illustrated in Figure 8. (a), while the design for the Sandwich chip is illustrated in Figure 8. (b). As the structures in the Sandwich chip were etched all the way through the wafer, each of the side outlet channels had their own outlet in order to connect the area between the outlet channels with the rest of the wafer.

The structures etched in silicon were reduced in width by 25 µm and the structures in glass were reduced by 30 µm in order to compensate for the structure broadening.

The dicing marks and chip labels differed between the chips. This can be seen in the Sections 4.4.4 and 4.4.5, covering the etching processes.

(a)

(b)

Figure 8: CAD drawing of the microfluidic structures used for both (a) the Silicon and (b) the Glass chips and the Sandwich chips. The width of the channels differed between the two materials in order to compensate for structure broadening.

(14)

4 Materials and methods

This chapter covers the laser marking system, the materials used, the method developed for evaluating etched structures, details on all the tests performed as well as etching, bonding and assembly of the chips.

An overview of the manufacturing process of the chips is illustrated in Figure 9.

Figure 9: A schematic representation of the process from design to assembled chip.

4.1 Laser marking system

The equipment used in this research was a Laser marking system AIO (Östling). The system uses a class 4 Nd:YVO₄green laser with a wavelength of 532 nm to ablate material. The laser is diode-pumped and pulses are generated by an electro-optical Q-switch. Properties of the system are listed in Table 1.

Table 1: Properties of the laser marking system.

Wavelength 532 nm Max pulse energy 400 µJ Peak power 200 kW Frequency 10 - 150 kHz Pulse duration 2 - 10 ns Minimum spot diameter 16 µm Focal length 94 ± <1 mm Work area 58 by 58 mm

The laser marking system has a stationary work table and uses beam deflectors to guide the laser. As the two motors controlling the deflector unit have a certain response time, it is necessary to ensure that the deflector unit is in position before the laser is activated. This is done by adding waiting times that allow the motor to reach the desired position before the marking process is continued. For high quality markings it is important that these waiting times are fine tuned to avoid any artefact features.

4.1.1 Laser parameters

The laser marking system is controlled by a set of parameters. These parameters are listed below together with a brief explanation.

Focus - Focus was adjusted manually by turning a wheel with 2 mm adjustment per full turn. The wheel had no markings and focusing was done by visual and acoustical inspection. During processing, it is essential to control the focus as this determines the laser spot size. A larger spot size, as a result of inaccurate focusing, can result in excessive taper of the feature. This however, can be mitigated by using a lens with a wide depth of focus, which means that there is a larger interval where the spot size remains in focus [3].

Power - The percentage value of the maximum power of the laser used. A higher power is expected to result in a higher etch rate which means increased processing speed [11]. However, high power settings are also expected to increase recast material, surface debris, HAZ and potential damages such as cracks.

(15)

Frequency - Refers to the pulse frequency of the laser with which the marking is done. The frequency is directly controlled by a Q-switch. A low frequency gives shorter pulse durations with higher peak powers and pulse energy [10]. Therefore, the etch rate is expected to be higher for lower frequencies.

Scan speed - Refers to the speed with which the laser is swept across the substrate, in mm per second. A lower speed means higher power of the marking per distance unit. A higher scan speed increases the distance between laser pulses, as explained in Section 5.1. This can result in scan lines that are discontinuous.

Fill type - Refers to the type of pattern used to fill a structure with scan lines. The three fill types available are displayed in Figure 10. These are: single (Figure 10 (a)), cross (Figure 10 (b)) and triple lines (Figure 10 (c)). As the fill type determines the number of lines per pass, it will affect the etch rate where triple lines will have a higher etch rate than single lines. Furthermore, the surface structure is expected to differ between the fill types due to the orientation of the overlapping scan lines.

Passes - Refers to the number of repetitions of the pattern that is being marked. This will determine the depth of the structure.

Fill option - Specifies if the laser follows the scan lines unidirectional or bidirectional. This parameter was not optimized and was set to bidirectional for all structures.

Fill space - Determines the spacing between the scan lines. This parameter was not optimized and it was set to the minimum distance of 0.01 mm for all structures.

(a) Single (b) Cross (c) Triple

(d) Single (e) Cross (f) Triple

Figure 10: Three fill types used for etching a filled area. (a) Single, (b) cross and (c) triple lines. Optical microscope image of a 1 cm square filled with (d) single, (e) cross and (f) triple lines using a fill space of 2 mm.

4.1.2 Restrictions

For safe operation of the laser marker, several restrictions should be followed. Structures should be placed outside of a 20 mm square directly under the lens when processing highly reflective materials such as silicon. This is done to avoid any reflections that could damage the lens. A way to mitigate the risk and allow structures under the lens is to do a primary sweep of the structure using a low power setting to reduce the reflectivity of the surface.

(16)

The range of power and frequencies in this study were limited to avoid damage to the laser system. The upper limit for power was at 95% and frequencies below 10 kHz must not be combined with powers higher than 70%. Low-power processing should not be done below 50% power as the pulse energy will no longer be stable.

4.2 Materials

In this study, the LMM process was optimized for silicon and borosilicate glass. These materials are fre- quently used in microfluidics since they are inert materials, making them suitable for applications in life science. Furthermore, both materials have a high acoustic impedance which is required for acoustophoresis applications.

The silicon wafers used in this study are highly reflective for the wavelength of the laser which intro- duces restrictions to the LMM process.

Borosilicate glass (80% SiO₂, 5-15% B₂O₃with additives Na₂O, K₂O, Al₂O₃, CaO and MgO [22]) is used to minimize the risk of crack formation during machining due to the thermal stresses induced in the processed material. The reason for this is that borosilicate glass has a low thermal expansion coefficient compared to other transparent materials such as soda lime glass or quartz [22]. Processing borosilicate glass with a 532 nm laser could be challenging since the glass has a spectral transmission range of 0.35 to 2 µm [22] and is therefore transparent to the wavelength of the laser.

For the Silicon chip, the structures were etched in 4", single crystalline <100> wafers with a thickness of 525 µm. The wafers were polished on one side. The Silicon chips were sealed using a 725 µm thick borosilicate glass wafer. Similar silicon wafers were used for the Sandwich chip, but both sides were polished and the wafers had a thickness of 300 µm. Sealing of the Sandwich chips was attempted using 175 µm borosilicate glass.

For the Glass chip, double polished 4" borosilicate glass wafers with a thickness of 1000 µm were used.

Bonding of the second batch of Silicon chips was done using Ormocomp (Micro Resist Technology), which is a biocompatible and UV curable hybrid polymer that has glass-like properties when fully cured [23].

4.3 Evaluation methods

The LMM process was optimized to give a uniform rectangular cut-section profile with vertical walls and flat bottom. Process time was not optimized. The structures etched for the optimization were 200 by 2000 µm rectangles. Each laser parameter was tested by doing a parametric sweep i.e. testing a range of values for the parameter to determine the impact and optimal values.

One challenge was to establish a method to consistently evaluate the etched structures. The preliminary testing was evaluated using scanning electron microscopy (SEM) assisted by visual inspection of profilom- etry maps. This method did not yield any quantifiable data.

Therefore, the evaluation of the remaining tests was complimented by an analysis of cut-section profiles acquired from the structures, as seen in Figure 11. (a). For the analysis, a ratio between the etched area and the total area of a rectangle was calculated. To ensure a consistent comparison, the top of the structure was defined at 5% of the maximum depth of the structure. To determine the uniformity, two additional ratios were calculated for each side of the structures. The method is illustrated in Figure 11. (b). The method did not account for the depth of the structure and required to be complemented by a visual inspection of the cut-section profile as well as SEM imaging.

4.3.1 Optical surface profiler

An optical surface profiler is an instrument that utilizes interferometry for non-contact surface measure- ments. A measurement yields a 3D map of the surface with the ability to extract profiles of the desired structures. The instrument used in this study was a Nexview™NX2 (ZYGO). Using the profiler, high resolution profile maps were acquired and cut-section profiles of the etched structures were extracted.

4.3.2 Scanning electron microscope

A scanning electron microscope (SEM) utilizes a beam of electrons instead of light. This allows a resolution that is sub-wavelength of the visual spectra. Samples analysed with SEM are required to be conductive.

(17)

(a) (b)

Figure 11: (a) The acquired 100 by 300 µm slice, vertical red rectangle with yellow border, across an etched structure. (b) Visualization of the calculated ratio between the etched area (green) and the total area of the rectangle (green and blue). Two sub-boxes limited by the two dotted lines indicates the evaluation of each side.

Therefore, the glass samples in this study were sputter coated with a few nm thick layer of gold and palla- dium.

4.4 Etch processes and assembly of microfluidic chips

4.4.1 Test for positioning of structures

The impact of structure placement, in regards to structure quality, was tested by etching six lines positioned as displayed in Figure 12. The lines were 1000 µm long and the acquired profiles were a mean profile over the whole line. The etching was done at 80% power, a frequency of 20 kHz with 10 passes and a scan speed of 200 mm/s.

Figure 12: Positioning of lines at 12, 20 and 26 mm from the centre of the work table. Gray box is the restricted zone referred to in Section 4.1.2.

4.4.2 Optimization of silicon processing

Preliminary testing For the preliminary test, structures were etched using a wide range of powers, frequencies and scan speeds. The settings are displayed in Table 2. The intervals were selected to test both the upper and lower limits.

Powers and frequencies From the preliminary test it was concluded that well defined structures were etched at higher powers such as 70 to 80% and lower frequencies of 10 to 20 kHz. Based on these results a narrowed range of powers and frequencies were selected for a more thorough evaluation. The settings for the test are displayed in Table 3.

(18)

Table 2: Preliminary test settings for silicon.

Power [%] 40, 50, 60, 70 , 80 Frequency [kHz] 10, 20, 50, 60, 90, 100 Scan speed [mm/s] 100, 200, 300

Passes 50

Fill type Single lines

Number of structures 90

Table 3: Power and frequency settings for silicon.

Power [%] 50, 60, 70 Frequency [kHz] 10, 30, 50 Scan speed [mm/s] 100

Passes 100

Fill type Single lines

Scan speeds In the test for powers and frequencies, it was concluded that the best structures were etched at 50% power and 10 kHz frequency. Additionally, it was concluded from the preliminary tests that a lower scan speed appeared to give a better profile. Therefore, these settings were used to test the impact of a range of lower scan speeds. In addition to the lower scan speeds tested, three structures were etched at higher scan speeds. The settings for the processes can be found in Table 4. For these tests, the number of passes was not adjusted between the different scan speeds. This made the comparison more difficult since a lower scan speed gives deeper structures.

Fill types The impact of fill type was tested by etching three rectangles using single (S), cross (C) and triple lines (T). The settings used were based on the results of the previous tests and can be found in Table 5.

As cross and triple lines have multiple sweeps per pass, the number of passes were adjusted to acquire a similar depth. Single lines was used for 30 passes, cross lines was used for 20 passes and triple lines was used for 10 passes.

Table 4: Scan speed for silicon.

Low High

Power [%] 50 70

Frequency [kHz] 10 20

Scan speed [mm/s] 20, 60, 100 100, 200, 300

Passes 100 50

Fill type Single lines Single lines

Number of structures 3 3

Table 5: Fill type for silicon.

Power [%] 50

Frequency [kHz] 10

Scan speed [mm/s] 100

Passes S:30, C:20, T:10 Fill type Single, Cross, Triple

Post-process cleaning To investigate the effects of post-process cleaning, three structures were treated with buffered hydrofluoric acid (buffered HF) for 25 minutes. The treatment duration was based on an etch rate of 0.08 µm/min for SiO2 which would theoretically remove 2 µm oxide [24]. The settings for the structures in this test can be found in the column for high scan speed tests in Table 4.

Etching through a thin wafer Two methods for etching structures through a silicon wafer were tested.

The first method etched filled structures while the second method etched the outline of the structures. The structures in this test were an intersection between a resonance channel with a width of 200 µm and three outlet channels with a width of 150 µm. The settings displayed in Table 6 were not optimized and were selected based on the preliminary test.

Table 6: Performed test to etch structures through a silicon wafer.

Power [%] 70, 75

Frequency [kHz] 25

Passes 280

Fill type Single lines Fill option Bidirectional, outline

Structure broadening During optimization, it was observed that the etched structures were larger than originally designed. This was the case even when the incident angle of the laser was taken in consideration.

Structure broadening is to be expected as the laser beam spot size has a minimum diameter of 16 µm.

(19)

To compensate for the broadening, two sets of structures with widths ranging from 355 to 380 µm and 92 to 127 µm, with a step size of 5 µm, were etched. The ranges of widths were selected to match the targeted widths of 380 µm for the resonance channel, and 127 µm for the outlet channels. The optimized settings displayed in Table 10 were used for the test.

Etch rate To establish the etch rate for the optimized settings, Table 10, a set of squares were etched using 25 to 35 passes with a step size of 1 pass. A profile was then acquired across the etched structures.

The lowest point in each structure was measured and a line was fitted to the measured depth and the number of passes.

4.4.3 Optimization of glass processing

Preliminary testing The optimization of glass was done using the same method as for silicon. Prelim- inary testing consisted of finding the power threshold where etching was initiated and then do multiple parametric sweeps based on the result.

Powers and scan speeds A more in-depth analysis for the impact of scan speed and power was done based on the results from the preliminary testing. High power settings were combined with a wide range of scan speeds. Based on the first test, a second test using higher scan speeds was carried out. Both tests are displayed in Table 7.

Fill types To investigate the impact of fill type, a structure was etched using each of the three fill types. A high power and scan speed were used based on the results from the power and scan speed tests. The settings can be found in Table 8. Triple and cross lines were etched for 50 passes while single lines were etched for 100 passes since it was part of a previous test batch.

Table 7: Tests of power and scan speed for glass.

Test 1 Test 2

Power [%] 75, 80, 85 75, 80

Scan speed [mm/s] 300, 350, 400 500, 700, 900

Passes: 100 50

Table 8: Fill type for glass.

Power [%] 80

Frequency [kHz] 10

Passes 50 (Single 100) Fill type Single, Cross, Triple

Post-process cleaning To observe the effects of potassium hydroxide (KOH) treatment, four structures were etched using the optimized settings displayed in Table 14. The structures were treated in KOH for 10 minutes and were then examined using SEM. The treatment duration was selected for a gentle removal of surface debris without enhancing cracks that formed during processing.

Etching inlets Both the Sandwich and the Glass chips require in- and outlets to be etched through the thin glass wafer after bonding. To optimize this process, in- and outlets were etched through 150 µm thick glass wafers. Different sizes of inlets were etched using a range of scan speeds and powers which can be found in Table 9.

Table 9: Tests for etching in- and outlets in glass.

Power Scan speed

Diameter [mm] 0.5, 1, 2 0.5, 2

Power [%] 55, 60, 65 55

Scan speed [mm/s] 50 100, 200, 300, 400

Passes 30 -

(20)

Structure broadening To determine the structure broadening of structures etched in glass, two sets of structures with widths ranging from 355 to 380 µm and 92 to 117 µm with a step size of 5 µm were etched.

The ranges of widths were selected to match the targeted widths of 380 µm for the resonance channel, and 127 µm for the outlet channels. The test was done using the optimized settings, Table 14, for 160 passes.

Etch rate To establish the etch rate for the optimized settings, Table 14, a set of squares was etched using 140 to 160 passes with a step size of 2. A profile was then acquired across the etched structures. The lowest point in each structure was measured and a line was attempted to be fitted to the measured depth and the number of passes.

4.4.4 Microfluidic systems in silicon

First batch of Silicon chips Six structures were etched for the first batch of Silicon chips using the optimized settings displayed in Table 10. The process was done in three steps, displayed in Table 11, with a total process time of of five hours.

Table 10: Final settings used to etch structures in silicon.

Power [%] 50

Frequency [kHz] 10

Fill type Triple lines

Table 11: Table listing the number of passes during the processing of the first Silicon wafer.

Step 1 Step 2 Step 3 Total

Channels 35 - - 35

In- & outlets 50 50 50 150

Markings & label 15 - - 15

As discovered during the optimization, the positioning will impact the quality of the etched structure.

As the function of the system depends on the dimension and profile of the resonance channel, the resonance channel was aligned with an axis radiating out from the centre of the work table. This is explained in detail in Section 5.2. The outlet channels were placed as close to the centre as possible, to minimize the incident angle of the laser and thereby reduce the low slope on the angled outlet channels.

Dicing lines were placed to maximize the distance from the structures, while allowing cuts to be made across the whole wafer. Chip labels were placed close to the inlet. The placements of the structures can be seen in Figure 13. (a).

(a) (b)

Figure 13: (a) Positioning of the structures when etching the first wafer for Silicon chips. (b) Schematic representation of the stacked wafers during the anodic bonding for the first batch of Silicon chips. From bottom to top: The structured silicon wafer, the glass wafer and lastly the chisel-shaped anode.

For the bonding process, the structured silicon wafer and a 725 µm thick glass wafer were cleaned in a Piranha-bath (H₂SO₄ + H₂O₂, 1:1) for 20 minutes. The wafers were stacked in the anodic bonder as displayed in Figure 13. (b). The stack was heated to 370 °C and the bonding was done with an applied potential of 700 V for 6 hours.

Dicing of the bonded stack was done with the blade aligned to the dicing marks. However, due to complications during bonding the dicing had to be done with an offset, destroying all but two structures.

Silicone sleeves (228-0700, VWR), with an inner diameter of 1.09 mm, were glued to the in- and outlets of the two chips, C1 and F1, using silicone rubber (A-07, Wacker). The chips were left over night to cure before being tested for leakage. A piezoelectric transducer (APC-840, APC international) with an optimal

(21)

resonance frequency of 2 MHz was attached to the back of chip F1 using LOCTITE 420 glue and applying pressure for 60 seconds.

Second batch of Silicon chips The second batch of Silicon chips was etched using the same settings and positioning as for the first batch, Table 10. However, the dicing marks were replaced with shorter line segments and the chip labels were reduced in size similarly to the Sandwich chip in Figure 15. (a). The etching process for the second batch can be found in Table 12.

Table 12: Table listing the number of passes for the second Silicon wafer.

Channels 35 - - 35

In- & outlets 35 50 50 135

For the bonding process, the structured silicon wafer and a 200 µm thick glass wafer were cleaned in a ultrasonic bath of acetone for 60 seconds and were then dipped in two new acetone baths and one bath of isopropyl alcohol (IPA). The glass wafer was cleaned in a plasma asher for 60 seconds at 1000 W and was then spin coated with 1 ml of Ormocomp (Micro resist technology). The Ormocomp was partially hardened by UV exposure for 40 seconds at 12 mJ/(cm²s). The structured silicon wafer was combined with the coated glass and the stack was fully hardened for 90 seconds as illustrated in Figure 14. UV exposure was done in a mask aligner in hard-contact mode for the full hardening. The bonded stack was then baked at 130° for 30 minutes.

Figure 14: Schematic representation of the stacked wafers during the Ormocomp bonding for the second batch of Silicon chips. From bottom to top: The structured silicon wafer, the partially hardened Ormocomp adhered to the glass wafer and UV radiation.

The bonded stack was diced into individual chips following the dicing marks and one of the chips was then assembled similarly to the first batch of Silicon chips.

Table 13: Table listing the number of passes for each feature for the Sandwich chips.

Channels 35 35 10 80

In- & outlets 35 35 - 70

Sandwich chips For the Sandwich chips, six structures were etched through a thin silicon wafer using the optimized settings in Table 10. The process, displayed in Table 13, was done in four steps for a total process time of roughly 8 hours. The placement of the structures, dicing marks and chip labels can be seen in Figure 15. (a). Additionally, an alignment mark was added to the bottom left of the design.

For the bonding process, the structured silicon wafer was cleaned in a plasma asher for 20 minutes at 1000 W. It was then dipped in buffered HF for 60 seconds, rinsed in deionized water (DI water) and dried thoroughly. The 175 µm glass wafer was cleaned in sulphuric acid (H2SO₄) for 10 minutes, rinsed in DI

(22)

(a) (b)

Figure 15: (a) Positioning of the structures when etching the first wafer for Sandwich chips. (b) Schematic representation of the stacked wafers during the anodic bonding for the Sandwich chips. From bottom to top: the structured silicon wafer, spacers, the thin glass wafer to be bonded, a cover glass wafer and lastly a cubic anode.

water, dried and cleaned in a plasma asher for 10 minutes at 1000 W. The wafers were stacked on the bonding plate according to Figure 15. (b). A thicker glass wafer was inserted on top of the stack in order to prevent the anode from damaging the thin glass. The separation of the two wafers was done in order to heat them to bonding temperature, 380 °C, before putting the surfaces in contact. When bonding temperature was reached, the spacers were removed and a voltage of 800 V was applied. After 4 hours, the heating and voltage were turned off for the night. The process was resumed the following day for an additional 7 hours.

4.4.5 Microfluidic systems in glass

Glass chips Six structures were etched for the Glass chips using the optimized settings displayed in Ta- ble 14. The process took a total of 1 hour to complete and was done in one step with 140 passes per feature.

The structures were positioned similarly to the Silicon and Sandwich chips, as displayed in Figure 16.

Table 14: Optimized settings for etching glass.

Power [%] 60

Frequency [kHz] 10 Scan speed [mm/s] 460 Passes 140 Fill type Single

Figure 16: Positioning of the structures when etching the Glass chips.

(23)

5 Results and discussion

In this chapter, all the results of the previously described tests are presented and discussed. The chapter is divided in four sections which covers striation, optimization and chip manufacturing for silicon, as well as glass.

5.1 Striation

During the optimization process striations, i.e. a series of ridges, were observed at the bottom surface of multiple etched structures. This can be seen in Figure 19. (a). The number of ridges and their orientation was found to correlate with the scan speed, frequency and fill type as well as the placement of the etched structure on the work table.

When the laser is swept along a scan line, the laser pulses are separated by a distance that is determined by the scan speed and frequency of the pulsed laser. In turn, the separation of scan lines are controlled by the fill space. This is illustrated in Figure 17.

Figure 17: The impact of different laser parameters on the separation of pulses, scan lines and size of laser pulses. The blue arrow indicates a scan line.

Ridges are formed when the pulse separation and/or fill space is too large to ensure overlap of laser pulses. A mismatch between the pulse separation and the fill space can give rise to the striped pattern observed in Figure 19. (a). The pulse separation can be calculated using equation 3.

Pulse separation=Scan speed

Frequency (3)

An example of ridge formation is illustrated in Figure 18. were the ridges are formed due to excessive pulse separation. The laser has a fixed scan direction in relation to the work table i.e. independent of structure orientation. As a result of this, the orientation of a structure will impact the orientation of the ridges. Figure 18. (a). shows a vertical structure (I) and a tilted structure (II).

(a) (b)

Figure 18: Formation of striations for channels filled with single lines. (a) (I) A channel placed perpendicu- lar to the scan direction and (II) a channel placed tilted against the scan direction. (b) Placement of the two channels on the work table. The scan direction is indicated by the blue dashed arrow. The red dots are laser pulses and the grey lines are formed ridges.

For example, a laser with a pulse frequency of 10 kHz and a scan speed of 300 mm/s will result in a pulse separation of 0.03 mm along the scan lines. As expected, this correlates to the width of the ridges observed in Figure 19. (a). Consecutively, at a lower scan speed of 200 mm/s (Figure 19. (b)) the ridges

Micromachining of microfluidic systems using a nanosecond laser

Examensarbete 30 hp November 2019

Micromachining of microfluidic systems using a nanosecond laser

Process optimization and application

Per Söderbäck

Abstract

Micromachining of microfluidic systems using a nanosecond laser

Tillverkning av mikrofluidiksystem med laserskärare

Abbreviations

Contents

1 Introduction

2 Laser micromachining

3 Microfluidic system for acoustic manipulation of particles

4 Materials and methods

5 Results and discussion