Aalto- DD Alignment: High- Precision Robotic Microassembly...

9HSTFMG*aegeac+

ISBN 978-952-60-4640-2 ISBN 978-952-60-4641-9 (pdf) ISSN-L 1799-4934 ISSN 1799-4934 ISSN 1799-4942 (pdf) Aalto University School of Electrical Engineering Department of Automation and Systems Technology www.aalto.fi

BUSINESS + ECONOMY ART + DESIGN + ARCHITECTURE SCIENCE + TECHNOLOGY CROSSOVER DOCTORAL DISSERTATIONS

Aalto-D

D 6

9/2

012

Veikko Sariola

Droplet Self-A

lignment: H

igh-Precision R

obotic Microassem

bly and Self-Assem

bly A

alto U

nive

rsity

Department of Automation and Systems Technology

Droplet Self-Alignment: High-Precision Robotic Microassembly and Self-Assembly

Veikko Sariola

DOCTORAL DISSERTATIONS

Aalto University publication series DOCTORAL DISSERTATIONS 69/2012


Veikko Sariola

Doctoral dissertation for the degree of Doctor of Science in Technology to be presented with due permission of the School of Electrical Engineering for public examination and debate in Auditorium AS2 at the Aalto University School of Electrical Engineering (Espoo, Finland) on the 29th of June 2012 at 12 noon (at 12 o’clock).

Aalto University School of Electrical Engineering Department of Automation and Systems Technology

Supervisor Professor Heikki Koivo Instructor Adjunct Professor Quan Zhou Preliminary examiners Associate Professor Takafumi Fukushima, Tohoku University, Japan Assistant Professor Pierre Lambert, Université libre de Bruxelles, Belgium Opponents Research Professor Aarne Oja, VTT, Finland Assistant Professor Pierre Lambert, Université libre de Bruxelles, Belgium

Aalto University publication series DOCTORAL DISSERTATIONS 69/2012 © Veikko Sariola ISBN 978-952-60-4640-2 (printed) ISBN 978-952-60-4641-9 (pdf) ISSN-L 1799-4934 ISSN 1799-4934 (printed) ISSN 1799-4942 (pdf) Unigrafia Oy Helsinki 2012 Finland The dissertation can be read at http://lib.tkk.fi/Diss/

Abstract Aalto University, P.O. Box 11000, FI-00076 Aalto www.aalto.fi

Author Veikko Sariola Name of the doctoral dissertation Droplet Self-Alignment: High-Precision Robotic Microassembly and Self-Assembly Publisher School of Electrical Engineering Unit Department of Automation and Systems Technology

Series Aalto University publication series DOCTORAL DISSERTATIONS 69/2012

Field of research Control Engineering

Manuscript submitted 20 October 2011 Manuscript revised 19 April 2012

Date of the defence 29 June 2012 Language English

Monograph Article dissertation (summary + original articles)

Abstract Droplet self-alignment is a microassembly process where the surface tension of liquid aligns microparts to a substrate. Traditionally, solder has been used, but using unconventional liquids, such as water or adhesives in air, has several attractive properties. Water is compatible with most materials and processes, it is easy to achieve good droplet confinement and it evaporates quickly. Adhesives have the ability to make irreversible bonding. Nevertheless, achieving adhesive droplet self-alignment is difficult, because adhesives generally have a small surface tension. Both liquids can be adapted to low temperature processes.

This thesis describes water droplet self-alignment in high detail, by measuring yield, accuracy, capabilities to build complex structures, and speed of the process. Experiments have been done using an environment-controlled microassembly station and recorded using high-speed microscopy. The results show that droplet self-alignment can achieve industrially relevant performance, and the results may be used as a basis of future process design rules.

A new, microfabricated silicon capillary gripper has been developed, which picks microparts using the surface tension of water. Pick-and-place experiments showed that microparts are self-aligned to the tool by droplet self-alignment. The developed gripper enables handling microparts accurately and delicately.

Finally, new patterned oleophilic / oleophobic surfaces have been developed that enable self-alignment using oil-like liquids, including low-temperature curable adhesives. Self-alignment using an industrial adhesive on the patterns was demonstrated. While this surface was used for droplet self-alignment, a micropatterned oleophilic / oleophobic surface may well have other applications outside droplet self-alignment.

Keywords droplet self-alignment, microassembly, self-assembly, oleophobic, wetting, surface tension

ISBN (printed) 978-952-60-4640-2 ISBN (pdf) 978-952-60-4641-9

ISSN-L 1799-4934 ISSN (printed) 1799-4934 ISSN (pdf) 1799-4942

Location of publisher Espoo Location of printing Helsinki Year 2012

Pages 153 The dissertation can be read at http://lib.tkk.fi/Diss/

Tiivistelmä Aalto-yliopisto, PL 11000, 00076 Aalto www.aalto.fi

Tekijä Veikko Sariola Väitöskirjan nimi Pisaran itsekohdistus: Tarkkuusmikrokokoonpano hyödyntäen robotiikkaa ja itsekokoonpanoa Julkaisija Sähkötekniikan korkeakoulu Yksikkö Automaatio- ja systeemitekniikan laitos

Sarja Aalto University publication series DOCTORAL DISSERTATIONS 69/2012

Tutkimusala Systeemitekniikka

Käsikirjoituksen pvm 20.10.2011 Korjatun käsikirjoituksen pvm 19.04.2012

Väitöspäivä 29.06.2012 Kieli Englanti

Monografia Yhdistelmäväitöskirja (yhteenveto-osa + erillisartikkelit)

Tiivistelmä Pisaran itsekohdistus on mikrokokoonpanoprosessi, jossa nesteen pintajännitys kohdistaa mikrokokoisia kappaleita alustalle. Perinteisesti nesteenä on käytetty juotetta. Epätavanomaisten nesteiden, kuten veden tai liiman, käyttö ilmassa on useilla tavoilla houkuttelevaa. Vesi on yhteensopiva useimpien materiaalien ja prosessien kanssa. On helppo valmistaa pintoja, jotka rajaavat veden kostumista. Lisäksi vesi haihtuu nopeasti. Liimat taas pystyvät muodostamaan pysyvän liitoksen. Liimapisaraan perustuva itsekohdistus on kuitenkin hankalaa, koska liimoilla on pieni pintajännitys. Molempia nesteitä voidaan käyttää matalissa lämpötiloissa.

Tässä väitöskirjatyössä tutkittiin vesipisaraan perustuvaa itsekohdistusta mittaamalla sen saantoa, tarkkuutta, kykyä rakentaa monimutkaisia rakenteita ja nopeutta. Kokeita on tehty mikrokokoonpanoasemalla, joka sijaitsee olosuhteiltaan säädellyssä kammiossa. Kokeet nauhoitettiin suurnopeusmikroskoopilla. Tulokset näyttävät, että pisaraan perustuvalla itsekohdistuksella voidaan saavuttaa teollisesti merkittävä tehokkuus, ja tuloksia voidaan hyödyntää tulevaisuuden prosessisuunnittelun pohjana.

Lisäksi väitöskirjatyössä on kehitetty uudentyyppinen, piipohjainen kapillaaritarttuja, joka poimii mikrokappaleita veden pintajännitystä käyttäen. Poimintakokeissa osoitettiin, että kappaleet itsekohdistuvat työkaluun pisaran avulla. Tarttuja mahdollistaa mikrokappaleiden käsittelyn tarkasti ja niitä vahingoittamatta.

Väitöskirjassa näytetään myös, että kuvioituja oleofobisia / oleofiilisiä pintoja voidaan käyttää liimapisaraan perustuvassa itsekohdistuksessa. Itsekohdistus näytettiin kokeellisesti käyttäen kaupallista liimaa ja kehitettyä pintaa. Vaikka kyseistä pintaa käytettiin itsekohdistukseen, kuvioidulla oleofobisella / oleofiilisellä pinnalla voi hyvin olla muitakin sovelluksia.

Avainsanat pisaran itsekohdistus, mikrokokoonpano, itsekokoonpano, oleofobinen, kostuminen, pintajännitys

ISBN (painettu) 978-952-60-4640-2 ISBN (pdf) 978-952-60-4641-9

ISSN-L 1799-4934 ISSN (painettu) 1799-4934 ISSN (pdf) 1799-4942

Julkaisupaikka Espoo Painopaikka Helsinki Vuosi 2012

Sivumäärä 153 Luettavissa verkossa osoitteessa http://lib.tkk.fi/Diss/

6

List of publications

7

List of publications

This thesis consists of an overview and of the following seven publications

which are referred to in the text by their Roman numerals.

I. Veikko Sariola, Quan Zhou, and Heikki N. Koivo, “Hybrid

microhandling: A unified view of robotic handling and self-

assembly,” J. Micro-Nano Mechatronics, vol. 4, no. 1, pp. 5-16,

2008.

II. Veikko Sariola, Mirva Jääskeläinen, and Quan Zhou, “Hybrid

microassembly combining robotics and water droplet self-

alignment,” IEEE Transactions on Robotics, vol. 26, no.6, pp. 965-

977, 2010.

III. Veikko Sariola, Quan Zhou, and Heikki N. Koivo “Three

Dimensional Hybrid Microassembly Combining Robotic

Microhandling and Self-Assembly,” in Proc. IEEE Int. Conf. Robotics and Automation, ICRA’09, pp.2605-2610, 12-17 May

2009.

IV. Mirva Jääskeläinen, Veikko Sariola, and Quan Zhou,

“Environmental effects on droplet self-alignment assisted hybrid

microassembly,” in Proc. IEEE Int. Symp. Assembly and Manufacturing, ISAM’09, pp. 177-182, 2009.

V. Bo Chang, Veikko Sariola, Mirva Jääskeläinen, and Quan Zhou,

“Self-alignment in the stacking of microchips with mist-induced

water droplets,” J. Micromechanics and Microengineering, vol. 21,

no. 1, 2011.

VI. Veikko Sariola, Ville Liimatainen, Tatu Tolonen, Reidar Udd, and

Quan Zhou, “Silicon Capillary Gripper With Self-alignment

Capability,” in Proc. IEEE Int. Conf. Robotics and Automation,

ICRA’11, pp. 4098-4103, May 2011.

VII. Bo Chang, Veikko Sariola, Susanna Aura, Robin H. A. Ras, Maria

Klonner, Harri Lipsanen, and Quan Zhou, “Adhesive Self-assembly

of Microchips on Oleophilic/Oleophobic Patterned Surfaces in

Ambient Air,” Applied Physics Lett., vol. 99, no. 3, 2011.

8

Contributions of the Author

The author was the primary designer and conductor of the experiments in

[I - IV, VI], including setting up and automating the experimental platform.

The author was solely responsible for the mathematical analysis in [II] and

[VI], and for the machine vision algorithms in [V]. The author did all

microfabrication in [II - VII].

Adj. Prof. Quan Zhou was the instructor of all work. In [II] and [IV],

Mirva Jääskeläinen conducted part of the experiments (Section IVA, test

Sets 2 and 3 and Section IIIC, respectively). In [V] and [VII], Bo Chang

conducted the self-alignment experiments. In [VI], Ville Liimatainen, Tatu

Tolonen and Reidar Udd contributed to the mechanical setup of the

platform and pick-and-place experiments. Susanna Aura contributed the

porous ORMOCER® material in [VII]; the patterning was developed by the

author. Ph. D. Robin Ras developed concept of the surface functionalization

method used in [VII].

Publications [I - III, VI] were written primarily by the author. The author

contributed significantly to the writing of [IV], [V] and [VII].

Symbols

9

Symbols

Symbol Unit Definition

A m2 Surface area of the liquid interface

dtdx / m/s First derivative of x w.r.t. time t . The speed of a

part. 22 / dtxd m/s2 Second derivative of x w.r.t. time t . The

acceleration of a part.

E J Surface energy

f Fraction of solid surface area wet by the liquid,

10 �� f

F N Forceg m/s2 Gravitational acceleration

h m Liquid droplet film thickness

ph m Part thickness

l m Part length

CL m Capillary length of liquid

m kg Part mass

Cp Pa Characteristic pressure of liquid

r Surface roughness, ratio of the actual surface area

to the geometric surface area, 1�r

2,1R m Principal radii of curvature

*2,1R Non-dimensionalized principal radii of curvature

T Nm Torque

V m3 Droplet volume

w m Part widthx m x-bias, the difference between the initial position

and the equilibrium position of a part during self-

alignment

z m Height

10

Symbol Unit Definition

*z Non-dimensionalized height

Greek symbols� J/m2 Surface energy. For liquids equal to surface

tension.

SL� , LG� ,

SG�

J/m2 Surface energies of the solid-liquid, liquid-gas

and solid-gas interfaces, respectively

xE �� / N Partial derivative of E w.r.t. x�� /E Nm Partial derivative of E w.r.t. �

p� Pa Overpressure inside a meniscus

*p� Non-dimensionalized overpressure

� º Contact angle

0� º Contact angle of a perfectly smooth surface

1� , 2� º Contact angles of surface patterns 1 and 2,

respectively Pa·s Liquid viscosity kg/m3 Liquid density

p kg/m3 Part density

� º Pad edge angle

� º Part rotation around z-axis

Abbreviations

11

Abbreviations

ECA Electrically conductive adhesive

FSA Fluidic self-assembly

I/O Input / output

ICP-RIE Inductively coupled plasma reactive-ion etching

LMA Low melting point alloy

PCB Printed circuit board

PECVD Plasma-enhanced chemical vapor deposition

RFID Radio-frequency identification

RIE Reactive-ion etching

RMS Root mean square

SAM Self-assembled monolayer

SEM Scanning electron microscope

SU-8 A negative, high aspect-ratio photoresist

12

Contents

13

Contents

LIST OF PUBLICATIONS 7

Contributions of the Author 8

SYMBOLS 9

ABBREVIATIONS 11

CONTENTS 13

1. INTRODUCTION 15

1.1 Droplet self-alignment 161.1.1 Robotic microassembly and self-assembly 171.1.2 Terminology 18

1.2 Objectives of the research 19

1.3 Methods and tools 20

1.4 Thesis contributions 21

1.5 Related research 22

2. SURFACES WITH PATTERNED WETTING PROPERTIES 25

2.1 Surface energy 25

2.2 Contact angle 25

2.3 Manipulating contact angles 272.3.1 Surface chemistry 272.3.2 Surface roughness 28

2.4 Patterned wetting: surface patterns and protrusions 29

2.5 On the validity of the wetting models 30

3. DROPLET SELF-ALIGNMENT 33

3.1 Modeling of droplet self-alignment 343.1.1 Quasi-static modeling 34

14

3.1.2 Numerical modelling using Surface Evolver 363.1.3 Dynamic modeling and liquid properties 36

3.2 Handling strategies 383.2.1 Droplet self-alignment assisted robotic microhandling 383.2.2 Droplet self-alignment of mismatching shapes 393.2.3 Flipping parts 403.2.4 Capillary gripping 41

4. EXPERIMENTAL WORK 43

4.1 Microrobotic test platforms 434.1.1 Pick-and-place robots 434.1.2 Liquid droplet dispensing 444.1.3 High speed microscopy and machine vision 44

4.2 Microfabricated surfaces with patterned wetting properties 454.2.1 Hydrophilic surfaces using silicon dioxide 454.2.2 Hydrophobic surfaces using CHF3 plasma 454.2.3 Oleophilic / oleophobic patterns using porous ORMOCER® with perfluorinated trichlorosilanes and gold 464.2.4 Sharp edges using SU-8 464.2.5 Fabrication of a capillary gripper using inductively coupled plasma reactive-ion etching 47

5. RESULTS AND DISCUSSION 49

5.1 Water droplet assisted release in robotic microhandling [I – V] 495.1.1 Yield and statistical modelling [II] 505.1.2 Accuracy, speed and trajectories [II] 515.1.3 Mismatching and 3D structures [I] - [III] 515.1.4 Effects of environmental conditions on self-alignment [IV] 535.1.5 Parallel delivery of liquid using mist [V] 54

5.2 Silicon capillary gripper with self-alignment capability [VI] 55

5.3 Adhesive droplet self-alignment on oleophilic / oleophobic surfaces [VII] 56

6. CONCLUSIONS 59

ACKNOWLEDGEMENTS 63

REFERENCES 65

APPENDIX A: NEGLECTING GRAVITY IN SELF-ALIGNMENT 75

APPENDIX B: ENERGY OF A TWISTED MENISCUS 77

APPENDIX C: PUBLICATIONS 79

Introduction

15

1. Introduction

There are two prevailing trends in the semiconductor industry: “more

Moore”, transistors being cramped up in smaller and smaller space, and

“more than Moore”, the heterogeneous integration of different types of

component, such as logic, memory, radio frequency circuits, signal

conditioning, sensors, actuators, mechanics, fluidics, optoelectronics and

optics [1–3]. The fabrication processes and the material prerequisites for

the components may be incompatible, so that it is not possible to fabricate

them all on a common substrate [4]. One solution is to manufacture the

wafers separately, cut the wafers into dies, and assemble those dies into a

single package. Increasingly, the dies are stacked vertically in 3D, which

allows smaller packaging of the devices [3]. Currently, this is done using

high-performance robotic microassembly.

In the manufacturing of relatively simple semiconductor devices,

assembly has long been the major cost factor. For example with high

performance flip chip assembly of radio frequency identification devices

(RFID), assembly makes up 65% of the costs [5]. Due to the serial nature of

the robotic assembly, any throughput increase is transferred to the price

when manufacturing large volumes. Lowered manufacturing costs would

not only bring more profit, but open up new application fields by e.g.

bringing RFID tags to consumer products. Relatively high cost of the tags

has still hindered their acceptance [6].

Increasing the complexity of the components translates to an increasing

number of input / output (I/O) pads. Also, the total size of the packaged

device is still decreasing; consumers expect the same functionality in an

ever decreasing package size. Together these factors decrease the I/O pitch

(distance between the I/O pads), so the pads need to be aligned more

accurately. Especially optical interconnects need to be aligned very

accurately for good optical coupling [7]. Active alignment methods can

achieve this, but with a significant throughput penalty. High accuracy

passive alignment and high yield assembly methods would enable more

complex products [8][9].

16

1.1 Droplet self-alignment

When a droplet of liquid is placed between a part and a receptor site, the

droplet forms a meniscus and aligns the part to the substrate. This is called

droplet self-alignment (Fig. 1). The phenomenon is a consequence of the

surface energy of the droplet: the energy is minimized when the surface

area of the droplet is minimized, i.e. when the part is aligned to the receptor

site.

Figure 1 Principle of droplet self-alignment. a) A droplet of liquid forms a meniscus between a part and a receptor site. b) The surface tension of the droplet aligns the part to the receptor site.

The key to droplet self-alignment is to confine the wetting of the droplet

between the part and the receptor site. The wetting properties of the

surfaces depend on the surface materials, the droplet liquid, addition of

surfactants, medium (air, water, vacuum etc.), temperature, surface charge

etc. Two general approaches for inhibiting liquid spreading can be

identified: surface patterns [10], as done in the receptor site of Fig. 1,

and sharp edges [11], as done with the part in Fig. 1.

Solder self-alignment (where the droplet is solder) has been

extensively studied and is being used in the industry, for example in flip-

chip or surface-mount reflow soldering [9], [12–20]. In recent

developments, shaking has been used to help the self-alignment [21] and

solder self-alignment has been demonstrated in the presence of a viscous

fluid [19], which is important for flip chip assemblies using “no flow”

underfill. For solder, it is relatively easy to achieve high wetting contrast

using metallic pads. Nevertheless, it has its drawbacks; melting solder

requires relatively high temperatures, so that heat-sensitive materials and

parts cannot be used. The use of Electrically Conductive Adhesives (ECAs)

together with Low Melting point Alloys (LMAs) has been proposed [22],

[23]. The self-alignment was possible by the surface tension of the LMA,

but not with adhesives alone.

The use of unconventional liquids in air has several attractive properties.

These liquids include water [24] and adhesives (or resins, as they are

sometimes called) [25]. Adhesives can be adapted to low temperature

processes, and have the ability to make irreversible bonding. Nevertheless,

achieving adhesive droplet confinement for self-alignment on planar

Introduction

17

surfaces has not been demonstrated so far, partly because adhesives have

small surface tension and behave therefore like oils, not being easily

confined inside the patterns. Comparison of the typical viscosities and

surface tensions of adhesives, water and solders is shown in Fig. 2.

One of the differences between solder self-alignment and water droplet

self-alignment is that the solder balls are also used as electrical connectors.

Therefore, typically multiple solder joints are used, laid out in a matrix

pattern. In contrast, water droplet self-alignment usually uses a single

droplet, which can have a size comparable to the size of the part.

Water alone is not capable of making permanent bonding, but several

methods for combining water droplet self-alignment and bonding have

been proposed, e.g. using silicon oxide pads and a small amount of

hydrogen fluoride (HF) in the water [26], or using metallic bonding after

the water droplet self-alignment [27]. Water is an attractive alternative

because it is compatible with many materials and processes, it has relatively

high surface tension and therefore it is relatively easy to achieve good

droplet confinement. Furthermore, water evaporates rather quickly, leaving

the part in close contact with the substrate.

Figure 2 Comparison of surface tensions and viscosities of adhesives, water and solders in their typical operating temperatures. [25], [28–32]

1.1.1 Robotic microassembly and self-assembly

Microrobotic assembly methods can achieve either high throughput or

small part size, but not both at the same time. There is a tradeoff: the

smaller the part size, the slower the operation is [33]. Furthermore, there is

a fundamental limit in accuracy that can be achieved using optical

microscopes. Also, robotic assembly methods have long been known to

suffer from “sticking effects”, adhesive forces between the tool and the

object [34].

102

103

10-1

100

101

102

103

104

105

106

Surface tension [mN / m]

Surface tensions and viscosities of liquids

Adhesives

SoldersWater

18

Using droplet self-alignment for the final, fine alignment is an attractive

option. The accuracy of sensors or actuators does not affect the alignment

accuracy. Rather, the fabrication accuracy of the part and the substrate sets

the accuracy, so it can be very high [35], even when the robot is very fast.

Also, if the part is brought into contact with a liquid droplet (e.g. water), the

surface tension helps overcoming the adhesive force between the tool and

the object [36].

Since the beginning of the 1990s, there has been an increased interest in

self-assembly of mesoscopic and microscopic parts, in which the parts are

designed in such a way that they are automatically attracted to their

assembly sites. Self-assembly of microparts is typically based on surface

interactions [37]. The parts are typically delivered randomly to the

assembly sites, e.g. in fluid suspension [38].

Many self-assembly methods include droplet self-alignment as a part of

the process [37], [39], [40]. Self-assembly is inherently parallel, and droplet

self-alignment defines the accuracy. Therefore, such methods could yield a

higher assembly throughput rate and a higher alignment precision than

robotic placement [41]. In general, however, self-assembly methods have

not been used to create very complex devices. Furthermore, since the

assembly steps need to be considered during fabrication, reconfiguration of

the assembly is difficult compared to traditional methods.

Self-assembly has been used to build a range of interesting structures, for

example macroscopic crystalline structures [42], a microsized replica

keyboard [43] and a DNA smiley [44]. While such structures are impressive

scientific feats, their assembly methods are not directly applicable in the

assembly of microchips.

Surface tension is not the only force that has been used for self-alignment.

Self-alignment has been studied using several other physical forces,

including 3D shape matching [45], magnetic [46] and

electrophoretic forces [47]. Mastrangeli et al. [4] and the author [I]

have reviewed alternative methods in more detail.

Many of the ideas and techniques used in self-assembly can be adapted

and combined with traditional robotics. The author discusses this concept

of hybrid assembly in [I], but in this summary only one combination, the

combination of droplet self-alignment with robotics, is discussed.

1.1.2 Terminology

Droplet self-alignment has been called by several other names, including

solder self-alignment (when the liquid is solder) [48], resin self-alignment (when the liquid is resin, actually a type of adhesive) [25], surface tension

Introduction

19

driven self-assembly [49], capillary self-assembly [50] and fluidic self-assembly [51]. The terminology is yet to be fully established across research

fields, and the author himself has been unable to adapt a single standard

when publishing in journals of different fields. To avoid confusion, this

summary reserves the term self-assembly only for stochastic, parallel

assembly methods. Self-alignment refers to a process where forces bring

the part accurately to its desired position and orientation. A self-assembly

process may or may not include a self-alignment step.

Sometimes it makes sense to distinguish between self-centering and self-alignment. In this thesis, self-centering refers to position correction, while

self-alignment includes orientation correction. Achieving self-centering, but

not self-alignment, is typical for axially symmetric droplet joints [52]. In

Section 3.2.1 and [VI], a special case, which achieves self-alignment, but not

self-centering, is detailed. However, in general self-alignment includes self-

centering, which is assumed hereon unless otherwise noted.

The word droplet underlines that the liquid used is not necessarily

solder. Fluidic and liquid self-alignment could be easily confused with

fluidic self-assembly methods, where the whole process is done in a liquid

medium. Capillary self-alignment is confusing, as there is no capillary

(small narrow tube), or capillary action (liquid rising spontaneously in a

narrow tube). Therefore, this summary uses the term droplet self-alignment, but the reader should be aware of the possibly synonymous

terms in the literature.

The word microassembly highlights that distinct microparts combined,

as opposed to monolithic microfabrication methods where all the required

functionalities are built on the device by sequential material deposition,

removal and modification steps. Microparts should have at least one

dimension less than 1 mm. In this summary, assembly is understood as the

placement of only a single part on a substrate or another part. The assembly

of multiple parts can often be broken down into several single-part

assembly operations, which is why multipart assembly is not considered

separately. Subsequent packaging operations, such as sealing or wire

bonding, are out of the scope of this thesis and are therefore excluded from

the definition.

1.2 Objectives of the research

The objective of this research is to increase the accuracy of the robotic

microassembly process by combining it with droplet self-alignment using

unconventional liquids, such as water and adhesives. The goal is to have

20

benefits of both technologies, without the drawbacks. The specific research

objectives include:

� Characterizing the water droplet self-alignment process in high

detail, including yield, accuracy, its capabilities to build complex structures, and studying the speed of the process

� Developing a new miniaturized capillary gripper which picks up

a micropart using the surface tension of water. The droplet self-aligns the micropart to the tool

� Developing new patterned surfaces that enable droplet self-

alignment using water and oil-like liquids, including low-

temperature curable adhesives

1.3 Methods and tools

This research is multidisciplinary, combining elements of microrobotics,

microfabrication, surface chemistry and automation, which are

cross-scientific in their own right. The approach taken is predominantly

experimental.

The research builds upon previous microrobotic research conducted by

the research group during the last 20 years [53]. For the experiments [I -

V], [VII], a microhandling station was setup. The station comprises of

microrobotic pick-and-place tweezers that can handle parts down to about

50 μm, several microscopes, positioning stages and liquid dispensing

systems.

In most experiments, the robot picked a micropart, liquid was delivered to

the receptor site and the part was placed in contact with the droplet. The

part was released, and it self-aligned to the receptor site.

The self-alignment is a relatively fast process. To observe the process in

more detail, high-speed optical microscopes were used [II], [III], [V].

Furthermore, the ambient environment (temperature and humidity) affect

microhandling and droplet self-alignment, so the ambient environment was

carefully controlled using an environmental control system [54], [IV].

Accurate dispensing of the water was achieved using non-contact [I - IV]

microfluidic dispensing systems. As an alternative, a parallel dispensing

method based on water mist condensation was studied [V]. Adhesives were

dispensed using contact dispensing methods [VII].

For making the test parts and surfaces with patterned wetting properties,

several microfabrication methods were used, including standard

lithography, etching, plasma processing and surface chemistry. Some

special techniques that were used include making dummy parts and

patterns using SU-8 [55], [56], [I - V], [VII]. SU-8 is a spin-coated, thick

Introduction

21

photoresist which can be used to make structures as thick as 450 μm with

aspect ratios around 1:10 or even 1:20. Also, Inductively Coupled Plasma

Reactive Ion Etching (ICP-RIE) was used for making protruded wetting

patterns on silicon and through-silicon microchannels [VI]. Commercial

chips were used in [IV], [VI].

For hydrophilic surfaces, silicon dioxide was grown using thermal growth

process [VI] or deposited using plasma-enhanced chemical vapor

deposition (PECVD) from silane (SiH4) precursor [IV], [V]. Hydrophobic

fluorocarbon surfaces were deposited using trifluoromethane (CHF3)

precursor in a plasma process [IV], [V]. For oleophobic surfaces, porous

ORMOCER® was functionalized using fluorinated trichlorosilanes [VII].

ORMOCER® is an inorganic-organic polymer with a silicon skeleton.

1.4 Thesis contributions

This thesis brings new knowledge in using unconventional liquids in

droplet self-alignment, including water and adhesives, instead of the more

traditionally used solder.

This thesis brings new, experimental data from the droplet self-alignment

process itself, by studying the phenomenon using high-speed microscopy

and an environment-controlled microassembly station. Yields, accuracies,

capabilities and the speed of the process have been measured. This

information can be used as a basis of future process design rules.

A new, miniaturized microgripper that integrates water droplet self-

alignment in a capillary microgripper has been developed. The tool may be

used in future microassembly platforms which need to handle parts

accurately and delicately. Current flip-chip machines actively measure the

misalignment between the picked part and the tool; the newly developed

tool may remove the need for this and simplify the picking operations

considerably. The developed tool is suitable for smaller parts than the ones

previously used in capillary gripping.

Finally, a new fabrication method has been developed for patterned

oleophobic / oleophilic surface for adhesive droplet self-alignment. This is,

as far as the author knows, the first time adhesive droplet self-alignment

has been achieved using oleophobic surfaces for the droplet confinement.

While this surface was used for droplet self-alignment, producing a

micropatterned oleophilic / oleophobic surface itself is no trivial task, and

may well have other applications outside droplet self-alignment.

22

1.5 Related research

Several authors have reported water droplet self-alignment [24], [27], [35],

[57–60]. Parts with lateral dimensions of several millimetres [27], [35],

[57], [60] and around one millimetre [24], [59] have been used. Compared

to those works, the parts used in this thesis work were significantly smaller

(from 50 μm to around 300 μm). Consequently, also smaller amount of

water was used.

Fukushima et al. [61] reports water droplet self-alignment assisted release

from tweezers. Compared to the work done in this thesis, the initial

misalignment was not accurately controlled and the chips were larger than

the parts used in this thesis. Refs. [61] and [35] also report an assembly

approach using water droplet self-alignment for positioning chips on an

intermediate carrier wafer. The self-alignment accuracy was 0.43 μm.

Furthermore, hydrofluoric acid solution together with silicon dioxide

surfaces has been used for achieving final bonding [57]. The bonding

strength was over 10 MPa.

Sato et al. [24] reports water droplet self-alignment using surface

patterns. The results show that the alignment accuracy is decreased by the

friction force between the microparts, the existence of areas which are not

wet on the high wettability area and the overflow to the low wettability area.

Theoretical analysis showed that hexagonal pattern created the largest

restoring force of the studied patterns.

Noda et al. [58] reports droplet self-alignment using a significantly larger

water droplet and a smaller chip (75 μm × 75 μm × 5 μm) than the ones

used in this thesis work. The droplet was either water or methanol solution.

The chip floats on an excessively large (diameter around 1 mm) droplet,

which slowly evaporates. After the droplet evaporation, the chip is located

within the hydrophilic region of the surface, but not positioned in a definite

location or orientation inside the region.

Tsai et al. [59] reports self-alignment using a solid edge for inhibiting

water droplet spreading. The study shows that large droplets decrease self-

alignment accuracy. Furthermore, the chips stay aligned when transporting

the substrate.

Kaneda et al. [60] reports modelling and experimental study on

oscillations of a tilting circular pad on a glycerine or a water droplet. While

those results are important for understanding the dynamics of the tilt

oscillations, it is hard to predict lateral dynamics or random variations in

the self-alignment process based on that work. The tilting motion was

captured on video using high speed microscopy. In this thesis work, high

speed microscopy was also used.

Introduction

23

Chapuis et al. [27] reports a combination of a water droplet and a solder

bump self-alignment. First a large water droplet self-aligns the microchip to

the substrate, after which smaller solder bumps are reflowed and perform

the final self-alignment. That study focused on fabrication methods of the

chips and substrate, but the self-alignment process itself was only shortly

studied. The alignment error was estimated to be below 20 �m on both

axes.

Bark et al. [62] reports the use of surface tension as a gripping force. The

gripper in that work was significantly larger (4 mm × 4 mm square) than

the gripper presented in this work. Both grippers had a similar rectangular

geometry. The self-alignment effect was reported [62], but the phenomenon

was not modelled.

Aoyama et al. [63], Lambert and Delchambre [64], Lambert [65] and

Biganzoli et al. [66] report different capillary gripper designs. In all of the

aforementioned work, the grippers were axially symmetric and could not

achieve self-alignment; only self-centering. Furthermore, the parts handled

were significantly larger (> 1 mm) than the ones used this thesis. One

important difference between the previous works and this work is that in

this work, the capillary gripper was microfabricated from silicon, allowing

greater control of the gripper head shape and definition. Vasudev and Zhe

[67] report manipulating the gripping force using electrowetting. The

electrowetting gripper did not have a liquid inlet, so that the

microfabrication of the electrowetting pads would still have to be integrated

into the fabrication process of the liquid inlet.

Moon et al. [22] and Yasuda et al. [23] report self-alignment using LMA

filled ECAs. Adhesive droplet self-alignment was reported in Kim et al. [25].

The wetting contrast was achieved with sharp edges, instead of surface

patterns, and the chips were larger than the ones used in this thesis.

Ethylene glycol droplet self-alignment using only sharp edges and similar

parts as the ones in this thesis has been reported by Corral et al. [68], who

was a member of the research group of the author.

24

Surfaces with patterned wetting properties

25

2. Surfaces with patterned wetting properties

2.1 Surface energy

When bulk material is cut, work must be done to break the intermolecular

bonds of the material. The work done is proportional to the area of the cut.

The work is stored as energy in the interface, and one can associate a

material interface with energy defined as

AE � (1)

where E is the energy stored in the interface, � is the surface energy

(constant, units 2/ mJ ), and A is the area of the interface. The surface

energy depends on several factors including materials, chemical

composition, surface roughness, charge and temperature at the interface.

For liquids, surface energy is equal to the surface tension of the liquid.

Surface tension makes liquid interfaces act like a rubber sheet and is the

reason that liquid droplets tend towards spherical shape. With the energy

definition it is fairly easy to see why airborne liquid droplets assume a

spherical shape: spheres have the minimal surface area for a fixed volume,

minimizing the surface energy.

2.2 Contact angle

The Young equation establishes the contact angle at which a liquid droplet

comes at rest when placed on surface. Young equation states that

SGLGSL �� cos (2)

where SL� , LG� and SG� are the surface energies of the solid-liquid, liquid-

gas and solid-gas interfaces, respectively (See Fig. 3 for definition) and � is

the contact angle of the liquid.

26

Figure 3 When a droplet placed on a surface, the contact line settles at the contact angle, which is a characteristic of the solid-liquid-gas combination used.

The Young equation (2) is a consequence of the force balance at the

contact line. Cohesive forces try to curl the liquid droplet into a sphere,

while adhesive forces try to spread the liquid on the surface.

In practice, the Young equation is not always useful for predicting the

contact angle of a new material, partly because neither SL� or SG� is known

accurately. Instead, one just resorts to measuring the contact angle, since it

can be done fairly easily.

For a particular liquid, surfaces can be divided into two categories:

� High wettability surfaces, which have contact angles less than 90º.

When the liquid used is water, such surfaces are called hydrophilic, in

the case of oil they are called oleophilic (or more generally: lyophilic)

and when both types of liquids wet a surface it is called omniphilic.

� Low wettability surfaces, which have contact angles over 90º. Water

repellent surfaces are called hydrophobic, oil repellent oleophobic or

lyophobic and surfaces repelling all types of liquids omniphobic.

Water is a special case. Being an extremely polar solvent, there are

comparatively large cohesive forces associated with it and thus it is

relatively easy to find surfaces that are hydrophobic. For less polar liquids,

such as oils, this is not the case.

The Young equation (2) does not cover several nonidealities observed in

real surfaces. When a liquid contact line moves along the surface, the

contact angle can assume a range of angles between advancing and

receding contact angle denoted as A� and R� , respectively, depending on

the direction of the movement of the contact line [69]. This is called contact angle hysteresis and is a friction phenomenon.

When a droplet is dispensed on a surface, it is possible that the contact

line is advancing, i.e. the observed contact angle is A� on every side of the

droplet. On the other hand, when for example a droplet is evaporating, it is

possible that the contact line is receding everywhere, i.e. the observed

contact angle is R� on every side of the droplet. The observed contact angle

can also be anything in between A� and R� , if no care is taken to ensure the

direction of the movement of the contact line [70]. Thus, without specific

knowledge on how a contact angle measurement was done, a single contact


27

angle measurement should only be understood as an upper bound for R�

and as a lower bound for A� . For some materials, the angle hysteresis is

very low ( RA �� ) so a single contact angle fully describes the wetting on

the surface.

Surface chemistry and surface roughness also affect the observed contact

angle, which is discussed in the following Subsection. For simplicity, it is

assumed that the contact angle hysteresis is negligible, but the reader

should be aware that this is not necessarily the case. Both roughness,

especially local surface curvature [71], and chemistry [72] also affect the

contact angle hysteresis.

2.3 Manipulating contact angles

For the purposes of this thesis, it is of special interest to explore how the

wetting properties of a surface can be manipulated. Since it is the outermost

layer on the surface that defines the wetting properties, chemical methods

can be used to alter the adhesive properties of a surface. Surface roughness

also affects the contact angle and can have radical effects on the contact

angle, even when the materials stay the same.

In the following two subsections, these two methods are detailed.

2.3.1 Surface chemistry

Silicon dioxide is a naturally hydrophilic material (Fig. 4a). This is due to

having a lot of hydroxyl (–OH) groups at the surface [73], which are created

when the oxide comes into contact with water. Those hydroxyl groups have

strong interactions with water molecules, resulting in large adhesive forces.

Fluorocarbons (Fig. 4b) have been shown to belong to ultra-low

wettability materials [74]. This phenomenon likely arises from the low

polarizability and high ionization potential of the carbon-fluorine bond

[75]. Consequently, fluorocarbon coatings exhibit excellent repellency

towards water and in some cases even oil.

A breakthrough in tailoring surface properties was the development of

self-assembled monolayers (SAMs) [76]. SAMs are formed by amphiphilic

molecules that have two ends, head and tail. The head shows specific

affinity to the substrate, while the tail group can be chosen for its interfacial

properties, in effect allowing greater freedom in the choice of surface

material and the choice of the interfacial energy (Fig. 4c).

Common SAMs include thiols (head group –SH) that can bond with gold

and trichlorosilanes (head group –SiCl3) that can bond with hydroxyl

28

groups. For achieving low interfacial energies, SAMs with fluoroalkane tail

groups are used.

Figure 4 Three different ways to use surface chemistry and materials to manipulate interfacial energy: a) Silicon oxide is a naturally hydrophilic material due to the large number of hydroxyl groups on its surface; b) Fluorocarbons are naturally hydrophobic materials; and c) SAMs can be used to tailor interfacial energies, by the choice of the tail group R.

2.3.2 Surface roughness

The Wenzel model [77] of surface roughness is based on the following

observation: when the roughness of a surface increases, the actual surface area of the interface is increased (Fig. 5a). Wenzel model

states that

0coscos �� r (3)

where � is the apparent contact angle, 0� is the contact angle for a perfectly

smooth surface and 1�r is the roughness factor (ratio of the surface area

to the projected surface area). Two cases can be identified:

a) When º900 �� , 0�� i.e. the contact angle is decreased or stays

equal

b) Conversely, when º900 �� , 0�� i.e. the contact angle is

increased or stays equal

The important observation is that the surface cannot go from high

wettability to low wettability, in spite of the roughness, if the assumptions

of the Wenzel model are valid.

However, in porous surfaces, pockets of gas may be trapped in pores of

the solid (Fig. 5b). This reduces the contact area between the liquid and the

solid, leading to Cassie-Baxter model [69]:


29

1coscos 0 �� ffr �� (4)

where 1,0�f is the fraction of the droplet area that is in contact with the

surface.

When 1 f , (4) is reduced to (3). When 0�f , º180�� regardless of

the original contact angle. This explains how roughness can make highly

wettable materials non-wettable.

In practice, neither of the models is very good for predicting contact

angles: factors r and f are often unknown for a new material beforehand.

Again, it is often easiest to measure contact angles experimentally. The

models serve as an explanation on what kind of mechanisms contribute to

the observed contact angle and how to manipulate it. The Cassie-Baxter

model suggests that, in order to make surfaces with extremely low

wettability, one should have significant porosity, with pockets of gas trapped under the liquid.

Figure 5 Effects of surface roughness on contact angle: a) in Wenzel model, the wetting of the rough surface is complete; and b) in the Cassie-Baxter model, pockets of gas are trapped under the droplet, reducing the area in contact with the droplet.

2.4 Patterned wetting: surface patterns and protrusions

When a droplet is placed on an area with high wettability, framed by edges

with low wettability, the liquid spreading stops at the edges of the area and

the droplet assumes the shape of the area (Fig. 6a). In this summary, any

such shape inhibiting liquid spreading is called a pad.

To completely wet the pad, there is a minimum volume required for the

droplet. Below this volume, the contact line does not reach all the way to

the pad edge and the position of the droplet inside the pad is ambiguous1.

At the edge, the droplet can have any contact angle between the contact

angle of the pad and the contact angle of the background (Fig. 6b);

consequently, there is a range of volumes that can fully wet the pattern.

1 Methods for spontaneously positioning droplets into the center of a pad have been developed [100]; they are based on surface energy gradient. Such methods might be helpful for droplet self-alignment, but remain yet to be tested.

30

When the droplet volume goes past the maximum volume, the liquid

overflows the pad edge and wets also the background.

Figure 6 a) When a droplet is placed on a pad with high wettability (�1) on a low wettability background (�2), the droplet assumes the shape of the pad. b) When there is too little liquid, the pad is not fully wetted. If there is too much liquid, the liquid overflows and wets the background. c) Liquid spreading can also be inhibited by a sharp edge.

Pads with sharp edges can also inhibit liquid spreading [11]. A geometric

condition for the liquid confinement was already derived by J. W. Gibbs:

00 )º180( �� (5)

where � is the angle of the pad corner (see Fig. 6c).

The ability of a sharp angle to inhibit liquid spreading is similar to having

high wettability surface patterns on low wettability background, with a

theoretical difference in contact angles.

One of the attractive properties of using sharp edges instead of surface

patterns is that sharp edges can naturally confine even low surface tension

liquids [25]. However, in many cases it is undesirable or even impossible to

introduce a significant profile to the pads. Therefore, whichever approach is

appropriate depends on the particular task at hand.

Throughout this summary, many figures are drawn so that sharp edges

inhibit liquid spreading. This is for the reason that large part of the research

in this thesis concentrated especially on confining wetting using sharp

angles. In most cases the wetting confinement could have also been

possible using surface patterns.

2.5 On the validity of the wetting models

There has been considerable discussion on the validity of the Wenzel and

Cassie-Baxter models [78–80]. The discussion stems from an

experimentally proven observation: The contact angle behaviour is only determined by interactions of the liquid and the solid at the contact line alone [80]; not the average properties of the contact area.

��º180


31

Any surface properties (roughness, surface energy) inside the contact area

of the droplet have no contribution to the observed contact angle.

The author fully agrees with the contact line view; and, in fact, takes

identical stance with [79] in Fig. 6: After overflowing the pattern in Fig. 6b,

the contact angle is exactly 2� , the surface energy of the pad has no

contribution to the observed contact angle.

It seems that the controversy rises partly from misinterpretation of the

Cassie-Baxter model. The author believes that Cassie and Baxter fully

understood that it is the surface properties at the contact line that

determine the contact angle, and that their analysis of the surface energy

should be only applied to the infinitesimal area covered by the moving

contact line. They observed contact angle instability when a water droplet

moved on a coarse wire mesh, when the contact line moved discontinuously

from wire to wire [69]. Had they thought that the contact angle is set by the

average properties of the whole contact area such instability would be

unexpected.

Cassie and Baxter considered homogenous surfaces that have roughly

similar wetting properties throughout and with microscopic surface

roughness. It was for this reason that the surface properties at the contact

line could be replaced by the average properties of the surface.

Regardless of the controversy, Wenzel and Cassie-Baxter modelswere originally formulated to explain why rough or porous surfaces have low wettability. The fact that they have is already enough for this

thesis.

32

Droplet self-alignment

33

3. Droplet self-alignment

When a droplet of liquid is placed between a part and a substrate with

matching wetting patterns, the droplet forms a meniscus between the

patterns and aligns the part to the substrate. This is called droplet self-

alignment (Fig. 7). The phenomenon is a consequence of the surface energy

of the droplet: the energy is minimized when the surface area of the droplet

is minimized i.e. when the patterns are perfectly centered and aligned to

each other. The confinement of the droplet inside the patterns can be

achieved by any of the methods discussed in Subsection 2.4.

Figure 7 Illustration of the droplet self-alignment principle. a) A droplet forms a meniscus between a part and a receptor site. b) The surface tension of the droplet self-aligns the part to the receptor site. c) Perspective view of the phenomenon. d) View from the top: droplet self-alignment can correct both position and orientation of the part.

In this summary, the freely moving object is called the part (sometimes also

called the chip), while the substrate is the supporting structure that the part

is assembled on. In practice, the substrate can be e.g. a PCB, a wafer or

another part; in this thesis, mostly dummy parts and substrates were used

in order to demonstrate the principle. A receptor site is the shape on the

substrate which inhibits the liquid spreading and to which the part self-

aligns.

34

The difference between the initial location of the part and the final

location of the part, after the successful self-alignment, is called the bias. In

some other contexts [24], this is called the initial error. Term error was not

adopted in this work, because the bias could be accurately controlled and

thus is considered a process input.

3.1 Modeling of droplet self-alignment

3.1.1 Quasi-static modeling

Quasi-static models of droplet self-alignment only consider the energies

associated with the liquid droplet interface [15], and possibly the

gravitational potential energy associated with the liquid or microscopic

parts. The scaling effect [81] explains why the gravity is negligible

compared to capillary forces when the dimensions are small. This holds

true when the characteristic dimensions are much smaller than the

capillary length of the liquid; in practice this is so when the dimensions

approach sub-millimeter sizes (the capillary length CL of water is 2.7 mm).

This is generally true for the liquid joints used in droplet self-alignment.

Also the parts used in this thesis are of submillimeter size, so their gravity

can be neglected. This can be shown using non-dimensionalization of the

Young-Laplace equation, full derivation given in Appendix A.

Using quasi-static models, one can calculate the energy associated with

the interface as a function of the part location and orientation. As a result,

one gets the shape of the energy well, which can be used to study

problematic self-alignment configurations e.g. local minima or positions

where the restoring force is small [82].

The energy well depends on the shape of the part and the receptor site. In

this thesis, only square-shaped or rectangular parts and receptor sites have

been studied. The reason for this is that many dies are rectangular, because

they are cut from a wafer along horizontal and vertical lines. A single

rectangular receptor site can achieve both self-centering and self-

alignment. Rectangular parts have 180º symmetry positions, while square

parts have 90º symmetry. This means that there are multiple

configurations with identical energy.

A simplistic geometric model of the surface energy is presented in Fig. 8.

The droplet shape is approximated with a right parallelepiped.


35

Figure 8 Approximating droplet shape with a right parallelepiped.

By increasing the bias x in the geometric model, the volume of the liquid

stays constant and the surface areas of the front, back, top and bottom faces

stay constant. Only the surface areas of left and right faces (indicated in Fig.

8) increase, and their total energy is given by

2222 hxwAE � �� (6)

where w is the width of part, h is the height of the liquid film and x is the

bias in x-direction (see Fig. 8). It is clear that the energy minimum with

respect to x is found when 0 x i.e. the part achieves self-centering.

Even if such a geometric model is grossly simplified, it already brings

useful information about self-centering. By taking the derivative of (6) w.r.t

x-bias, the force F is

22/2/ hxwxxEF �� (7)

The force starts to drop as the part approaches the equilibrium, being

exactly zero when the part is in equilibrium. When x is much smaller than

h , (7) can be approximated with

hwxFhx

/2��

(8)

In practice, due to necking effects, contact angle hysteresis and other

wetting related phenomena, the droplet height cannot be controlled

accurately. However, it is more feasible to control the droplet volume Vand h is a monotonic function of V .

Decreasing h , by decreasing V , increases the force near the equilibrium

point, i.e. smaller droplets create larger forces. This suggests that smaller

droplets could be beneficial to the droplet self-alignment. On the other

hand, intuitively, using too small droplets, no droplet self-alignment can be

36

possible. Therefore, the droplet volume was selected as one of the key

parameter of the study, and its effects on the process yield was measured in

multiple experiments. This will be discussed in Subsection 5.1.1.

To show self-alignment, one should consider the energy of a meniscus as a

function of part rotation. This derivation is left to Appendix B.

3.1.2 Numerical modelling using Surface Evolver

The simplistic geometric model already captures the basic features of self-

alignment, but it is still a very crude approximation. Real surface shape is

smooth, more like shown in Fig. 9a and 9b.

The Surface Evolver software [83] can be used to numerically study

minimal surfaces more accurately. The software breaks the surface into

smaller elements, and tries to minimize the energy, by optimizing the

location of each vertex. During the optimization constraints are respected,

such as liquid volume being constant and that the droplet wets only inside

the pads. The actual optimization is based on well-known methods, such as

the gradient, Hessian or conjugate gradient method. As a result, the shape

of the energy well can be calculated more accurately.

The difference between the Surface Evolver simulation and the geometric

approximation of (6) can be very small so that (6) may be a good enough

approximation for some applications. Equation (6) only allows calculation

of the force in lateral direction. When moving the part in vertically with a

constant droplet volume, the liquid will bulge, neck or retract from edge of

the pad and the situation is more complex. More general cases are usually

handled with Surface Evolver, when analytical and geometrical

approximations would be complicated.

3.1.3 Dynamic modeling and liquid properties

The static models do not explain damping, which is a result of

hydrodynamic forces of the liquid. Assuming Newtonian fluid and linear

velocity profile inside the liquid [15], the viscous force can be approximated

with

dtdx

hwlF

� (9)


37

Figure 9 Example of droplet self-alignment simulation using Surface Evolver. Contact angles on the receptor site and the part are assumed to be 30º. The boundary conditions of the pads were assumed to be one-sided i.e. the contact line is forced to be inside the edges of the pads, but not necessarily at the pad edge. The liquid is assumed to be water (� = 72.8 mJ / m2), the size of the receptor site and the part is 300 μm × 300 μm and the liquid amount is 3 nL. a) Shape of the surface with bias between the receptor site and the part; b) Shape of the surface with perfect alignment. Notice the small necking of the liquid; and c) comparison of the energies calculated using (6) and Surface Evolver. In Surface Evolver, the x-bias was initially 0 m and the liquid thickness h was optimized so that the meniscus was in equilibrium. The h was then fixed for the rest of the simulation. In eq. (6), the liquid

thickness was approximated with wlVh .

where is the viscosity of the liquid, l is the length of the part (see Fig. 8)

and dtdx / is the velocity of the part. This approximation is valid only for

high viscosity liquids or small h . Eq. (9) underestimates the viscous force

for low viscosity liquids or large h ; in such cases, the liquid flows have to be

taken into account by solving Navier-Stokes equations [84] or some

simplification of thereof [85], [86].

Quasi-static modeling assumes that at each instant the surface is at

equilibrium and the force can be calculated by taking the derivative of the

energy. The force calculation can be done numerically using either Surface

Evolver or by some analytical approximation. For this approximation to be

valid, the dynamics of the liquid surface should be much faster than the

dynamics of the part. This is generally true for droplet self-alignment

applications [41].

When the bias is small, one can use (8), (9) and Newton’s law to get a

linear second-order dynamic equation of the part:

022

2

��hwx

dtdx

hwl

dtxdm �

(10)

38

where m is the mass of the part and 22 / dtxd is the acceleration of the

part. This explains some of the basic dynamic properties of the self-

alignment. The dynamics can be over- or under-damped, depending on the

balance between the mass, surface tension, viscosity and dimensions. When

the system is under-damped, the part overshoots during the self-alignment

and oscillates. It is clear that the choice of liquid has an effect on self-

alignment.

3.2 Handling strategies

3.2.1 Droplet self-alignment assisted robotic microhandling

As discussed in Subsection 1.1.1, droplet self-alignment can be used in many

applications in robotic microassembly and self-assembly. One particular

use is to help positioning parts after coarse positioning by a robot.

In this thesis work, the following handling strategy was extensively

studied: first, a microgripper picks a micropart from a surface and brings it

close to a receptor site. A droplet is dispensed on the receptor site, and the

gripper brings the part in contact with the droplet. When the part is

released, the surface tension of the droplet self-aligns the part to the

receptor site. Finally, at least in the case of water, the droplet evaporates;

leaving the part adhered to the receptor site. This handling strategy is

illustrated in Fig. 10.

In the simplest case, the shape of the receptor site and the part match

exactly. But the following Subsection discusses that this does not need to be

the case and good alignment can still be achieved.

Figure 10 Droplet self-alignment experiments using a tweezer-type robot. a) A pad with sharp edges is the receptor site. Receptor site could also be surface patterns; b) Liquid (in most cases water) is dispensed on the receptor site; c) Microgripper approaches the liquid droplet with a part in its grasp. In this thesis, the whole backside of the part is used and the liquid spreading is inhibited by the sharp edges. Surface patterns could also have been used; d) The part is placed in contact with the droplet and released; e) The part is self-aligned to the receptor site. In the case of water, the water evaporates and leaves the part aligned and adhered to the receptor site.


39

3.2.2 Droplet self-alignment of mismatching shapes

If the shapes of the part and the receptor site do not match exactly, there

exist multiple energy minima for the part (Fig. 11a). If, however, the part is

released outside the receptor site, the part can still align to the receptor site,

as shown in Fig. 11b. This happens when the part is sufficiently damped.

With too little damping (small liquid viscosity or large h ), the part can

overshoot past the edge and pull the contact line with it, away from the left

pad edge. In such a case, the part could end up to one of the positions in

Fig. 11a.

Figure 11 Self-alignment of mismatching shapes. a) There exist multiple energy minima, so the final position after the self-alignment may be ambiguous; b) If, however, the self-alignment is started outside the border of the receptor site and the system is sufficiently damped, the part aligns to the edge of the receptor site.

It is interesting to note that in all cases in Fig. 11, the rotation is still

corrected by the self-alignment. So, in this particular case, it is possible to

have self-alignment, without self-centering.

When only one edge confines the wetting changes, the shape of the energy

well is not the same as one side of the energy well of a perfectly matching

part and a receptor site. This happens because the liquid is able to wet

further on the receptor site. As a consequence, the alignment is not perfect,

but the part may align a bit inside the edge of the receptor site. When the

liquid film is thin, this misalignment is negligible.

When the liquid film is very thick, the ambiguity in the energy well

disappears completely. This is illustrated in Fig. 12. There is enough liquid

to wet the receptor site completely, and a small part can center on a larger

receptor site.

40

Figure 12 Alignment of a small part on a larger receptor site with a large amount of liquid. The wetting of the receptor site is complete, and the part aligns to the middle.

When there is only a small mismatch in the size of the part and the receptor

site, small droplet is already enough for making the position unique and

align the part to the middle of the receptor site. Thus droplet self-alignment

is robust: the part aligns in the middle of the receptor site even if there are

small tolerances in the part or receptor site dimensions, due to

manufacturing.

Finally, the alignment is even possible in two dimensions (Fig. 13). By

starting the alignment over both edges of a corner, a part self-aligns to the

corner of a large receptor site.

Figure 13 Alignment of a small part on a corner of a large receptor site. If the alignment is started with edges of the part outside both edges of the large receptor, the part aligns to the corner of the receptor site.

3.2.3 Flipping parts

In addition to realizing lateral part movement and part rotation around z-

axis, the surface tension of a droplet can be used to power rotations around

x- or y-axis. The principle is illustrated in Fig. 14. This can be used to flip

parts in place, which would otherwise require either dexterous

manipulations [87], complicated kinematics capable of object-centric

rotations [88] or rotating the part by pushing it against an external object

[89].

The phenomenon is known in solder assembly as part tomb-stoning [90],

where a passive component spontaneously rises up due to the surface

tension of the droplet. In electronics assembly, it is usually considered a

manufacturing defect. However, surface tension powered rotations have

also been used to create self-assembled out-of-plane structures in a

controlled manner [91].

When using surface tension powered rotations with a microrobotic

handling platform and a water droplet dispenser, interesting handling

possibility emerges: since the robot can approach the receptor from a

controlled direction, the rotation direction and final orientation of the part


41

can be chosen during manipulations. The water droplet dispenser allows

dispensing droplets on the fly and performing multiple manipulations with

the same part and receptor site, with the liquid evaporating after each self-

alignment. Combining multiple such manipulations, the part can rotated

with almost any desired face facing the receptor site.

Figure 14 Using surface tension of a droplet to flip parts. A side of the part wets and when the part is released, the part rotates so that the side faces the receptor site.

3.2.4 Capillary gripping

So far, droplet self-alignment has been discussed in the context of parts

self-aligning on a receptor site after the part is placed on the receptor site.

In a capillary gripper, the surface tension of a droplet is used to pick a part

from a surface. The surface tension lifts the part from the surface and self-

aligns the part to the tool. The principle is shown in Fig. 15.

The working principle of a capillary gripper is as follows: first, a water

droplet is formed in the head of the gripper. For this purpose, the gripper

usually contains a small water inlet (capillary) through which the water is

pumped (Fig. 15a). The gripper approaches the part on the surface, and the

droplet forms a meniscus between the gripper and the part (Fig. 15b). If the

surface tension is able overcome the adhesion of the part to the surface, the

part is picked (Fig. 15c) and self-aligns to the gripper. The gripper is usually

axially symmetric, so that only self-centering is possible. If, however, the

gripper is e.g. square, also self-alignment is possible (Fig. 15d-e).

Capillary gripping has several attractive features. First, due to the scaling

of the capillary force, the force becomes comparatively larger than inertia

when the parts are minimized. Second, the self-alignment mechanism may

simplify manipulations considerably due to eliminating the need for

measuring the misalignment between the tool and the part after picking.

Finally, the liquid joint provides compliance during manipulations and

applies force evenly on the surface, protecting the parts from damage.

One problem with capillary gripping is the release of the parts: if the

picking force was large enough to overcome the part-surface adhesion, the

part-surface adhesion is not large enough to release the part from the tools

42

grasp. Several solutions to this problem have been proposed in [92],

including pumping excessive amount of liquid through the gripper head.

Figure 15 Picking parts using a capillary gripper. a) A square-shaped gripper head approaches the part. Water droplet is inserted through the capillary. b) The gripper touches the part. The water droplet forms a meniscus between the gripper and the part. c) The gripper retracts and picks the micropart using the surface tension of the water. d-e) The part self-aligns to the gripper head, using surface tension.

Experimental work

43

4. Experimental work

4.1 Microrobotic test platforms

To capture the phenomenon of droplet self-alignment in detail, test

platforms were needed, which could a) position the microparts near the

receptor sites with known accuracy; b) dispense accurate amounts of liquid;

and c) record the phenomenon in real time. To these ends, different kinds

of custom microrobotic test platforms were developed, which included

accurate positioning stages, water dispensers and high speed microscopes.

An example of a test platform is shown in Fig. 16.

Figure 16 A microrobotic test platform used to study droplet self-alignment in [I – V].

4.1.1 Pick-and-place robots

In [I – V, VII], a tweezer microgripper was used to pick-and-place the parts

for self-alignment. The tweezer microgripper could handle relatively small

parts, down to 50 μm.

The gripper also suffers from the “sticky finger” phenomenon when

handling very small parts. Nevertheless, the surface tension of the liquid

can overcome the adhesion of the part to the tip and releases the part from

the tip.

To achieve accurate timing, positioning and repeatability in the

experiments, high level of automation was required. The test platforms

44

were computer controlled, and the experiments were automated using the

LUA scripting language [93].

4.1.2 Liquid droplet dispensing

Several different concepts for delivering the liquid to the assembly site were

tested. For water, a non-contact dispenser (GeSIM / PicPIP), capable of

shooting a water droplet to the assembly site, was used in [I – IV]. Each

droplet could be tuned to have a volume of about 60 to 250 pL. Larger

droplets could be formed by dispensing multiple droplets.

As an alternative, the parallel delivery of water droplets using

condensation from humid air was tested in [V]. Water mist was created

using an air humidifier (Bionaire / Ultrasonic Compact BU1300W-I), and

guided near the assembly site using tubing. The humidifier could be turned

on and off by the computer through an I/O card and a relay.

In [VI], water was delivered to a capillary microgripper using through-

silicon microchannel. On the backside of the silicon, 1/16” tubing was

connected using microfluidic connectors (Upchurch / Nanoport™). The

liquid amount was controlled using a volume controlled pump (Cavro /

XCalibur, Syringe size 25�l).

In [VII], adhesive liquid was used instead of water. The adhesive had a

much higher viscosity, so that a non-contact dispenser could not be used.

Instead, a contact dispenser (EFD Inc. / Mikros TM dispense pen) was used

for the liquid delivery. The contact dispenser makes a small droplet at its

head and touches the receptor site, delivering controlled amount of the

liquid on the receptor site.

4.1.3 High speed microscopy and machine vision

To observe the self-alignment phenomenon in detail, the process was

recorded using a high speed camera (Imperx / IPX-VGA210-G), capable of

recording up to 3000 frames per second. The camera was mounted on an

optical microscope. This enabled recording of the time it takes for the self-

alignment to complete [II], observing the trajectories of the part during

self-alignment [II] and recording videos of the self-alignment in high detail

[III].

In [V], machine vision algorithms were used to estimate the amount of

condensed water on the patterns. The algorithms were able to locate and

measure the size of each droplet from individual images. Using the

algorithms, the water condensation process was shown to be linear with

respect to time, and the amount of liquid could be easily controlled by

controlling the time of the condensation.

Experimental work

45

4.2 Microfabricated surfaces with patterned wetting properties

In this thesis, confined wetting has been achieved using both sharp edges

and surface patterns (see Subsection 2.4). All studies used sharp edges in

one way or another, while in part of the studies, surface patterns were also

used [IV, V, VII]. Table 1 summarizes the materials and liquids used in this

thesis, and the corresponding contact angles.

Table 1 Contact angles of different materials and liquids used in this thesis.

Material Liquid Contact angle (º)

Silicon dioxide Water < 10 [94]

SU-8 Water 65 [VII], 81 [V]

Delo 18507 adhesive 47 [VII]

Gold Delo 18507 adhesive 53 [VII]

CHF3-plasma

deposited fluorocarbon

Water 105 [95]

Functionalized porous

ORMOCER®

Olive oil 133 [VII]

Delo 18507 adhesive 119 [VII]

4.2.1 Hydrophilic surfaces using silicon dioxide

Hydrophilic surfaces were made from silicon dioxide, and two methods for

creating silicon dioxide surfaces were used: thermal oxidization [VI] and

depositing silicon dioxide in PECVD process from silane (SiH4) source gas

[IV], [V].

Silicon wafers were wet oxidized in oxidation furnaces. Patterning of the

oxide was achieved by using lithography and by dry etching the oxide in

RIE. A practical problem with the wet oxidation furnaces used is that any

metallized wafers are not allowed. Therefore, only new and clean wafers can

be thermally oxidized.

For wafers that had already patterns on them, silicon dioxide was

deposited using the PECVD process. Silicon dioxide patterns alone were

nearly invisible under an optical microscope. To see the edges of the

patterns clearly, a thin layer of aluminum was sputtered under the patterns

before SiO2 deposition.

4.2.2 Hydrophobic surfaces using CHF3 plasma

In [IV], [V], deposition of fluorocarbon surfaces was achieved using

trifluoromethane (CHF3) plasma [95]. Five minutes in pure CHF3 plasma

was enough to make the surfaces hydrophobic. The surface treatment was

stable enough that a photoresist could be spincoated, exposed, developed

46

and removed from top of it without significantly altering the hydrophobic

qualities of the surface.

4.2.3 Oleophilic / oleophobic patterns using porous ORMOCER® with perfluorinated trichlorosilanes and gold

In [VII], patterned oleophilic / oleophobic surfaces were developed for

testing the self-alignment using industrial adhesives. The oleophobic areas

were made out of ORMOCER® [96]. ORMOCER® is an inorganic-organic

polymer with a silicon skeleton. When exposed to oxygen plasma, the

organic material is removed, leaving only porous skeleton behind [97].

Furthermore, the silicon skeleton is oxidized.

After the oxygen exposure, gold patterns were defined on the

ORMOCER®, using lift-off lithography. Finally, perfluorinated

trichlorosilanes were used to functionalize the surface. The trichlorosilanes

are able to bind hydroxyl (–OH) groups of the porous ORMOCER®, but

not on gold. As a result, the gold remains oleophilic but the porous

ORMOCER® becomes oleophobic. The full fabrication sequence in shown

in Fig. 17.

Figure 17 Fabrication process of oleophobic/oleophilic patterned surface. Not drawn to scale. (a) ORMOCER® resist is spin coated on a silicon wafer; (b) oxygen plasma treatment to make porous ORMOCER®; (c) lithography; (d) gold evaporation; (e) lifting-off gold; (f) functionalization with trichlorosilane.

4.2.4 Sharp edges using SU-8

A large part of the self-alignment experiments were done using rectangular

microparts, made out of SU-8 (Fig. 18). SU-8 is a spin-coated, thick

photoresist which can be used to make structures as thick as 450 μm with

aspect ratios around 1:10 or even 1:20. SU-8 was very suitable for making

thick parts with small (around 100 μm) lateral dimension in various shapes

and sizes very quickly. For the self-alignment experiments, SU-8 has the

nice property of being naturally relatively omniphilic, and the edge of the

patterns can be used as a sharp edge for confined wetting. Furthermore,

Experimental work

47

SU-8 is optically transparent, allowing the observation of the self-alignment

process from above.

Figure 18 A scanning electron microscope (SEM) micrograph of an SU-8 part. The part size is approximately 100 μm × 100 μm × 40 μm.

The SU-8 parts used did not have any kind of electrical functionality. For

real world applications, silicon chips would have been more practical. Some

experiments were done using silicon chips [IV, VI], but SU-8 allowed much

faster prototyping. A silicon surface can be easily oxidised, which makes it

omniphilic so many of the techniques tried with SU-8 are expected to carry

over to silicon chips also. For very fine control of the wetting, any of the

methods discussed in Subsection 2.3 could have been used.

4.2.5 Fabrication of a capillary gripper using inductively coupled plasma reactive-ion etching

In [VI], inductively coupled plasma reactive-ion etching (ICP-RIE) was

used to create near vertical walls and through-silicon microchannels for

fabricating a silicon capillary gripper. Since polymer based photoresists are

not selective enough for ICP-RIE, the photoresist pattern was first

transferred to a hard mask from thermally grown silicon dioxide and

sputtered aluminum.

Two different etch depths (through-silicon channels and gripper head

shape) were achieved using nested masks: outer mask was made of

aluminum and masked the through-silicon etching, and inner mask was

made of silicon dioxide and masked the gripper head etching. The

fabrication method was adapted from [98] and is illustrated in Fig. 19.

48

Figure 19 Etching two different depths using nested masks. The figure is not drawn to scale. a) Silicon oxide patterns are defined on the silicon wafer. On top of the oxide, aluminium patterns are defined; b) Through silicon channels are etched, using aluminium as the mask; c) The aluminium is removed; d) Second silicon etching uses the oxide as a mask. e) Perspective view of the final structure.

Liquid was inserted through the through-wafer channel. One of the nice

side benefits of using silicon dioxide as a mask material was that it could be

left on the surface. The naturally omniphilic silicon dioxide, bordered by

sharp edges, was good for achieving confined wetting of water [VI].

Results and discussion

49

5. Results and discussion

Papers [I] – [V] show how water droplet self-alignment can be used to help

part placement in robotic microassembly. Specifically, a tweezer

microgripper was used (see Section 4.1.1) to pick-and-place microparts on

water droplets, which self-aligned the microparts to the assembly sites after

releasing. In those studies, a microdispenser was used for local water

droplet delivery.

Paper [I] surveys the merits of robotic microassembly and self-assembly.

Self-assembly has high efficiency, as it can have very high throughput. On

the other hand, robotic handling has good capabilities, in the sense that it

can build complex structures and can be adapted by reprogramming to the

assembly task. Based on the analysis, the paper proposes the combination

of droplet self-alignment with robotics, demonstrates the handling strategy

in experiments and analyses the proposed handling method using the

results of the survey. In paper [II], the yield, accuracy and speed of the

handling method are experimentally measured. Paper [III] reports the

building of 3D structures using the aforementioned handling method.

Paper [IV] reports the effects of the ambient environment conditions

(temperature, humidity) on the handling method. In paper [V], parallel

delivery of water by condensation from humidified air was reported.

Paper [VI] shows how water droplet self-alignment can be used to help

part picking in robotic microassembly. Specifically, a new microfabricated

capillary microgripper is presented, where the picked micropart is self-

aligned to the tool.

Finally, paper [VII] reports how oleophilic / oleophobic patterns can be

fabricated and used for adhesive droplet self-alignment.

The results are now explained more thoroughly.

5.1 Water droplet assisted release in robotic microhandling [I – V]

In paper [I], the merits of the droplet self-alignment assisted robotic

microhandling were qualitatively evaluated, by performing the following

50

experiments: a) aligning two, identically shaped 300 μm × 300 μm square-

shaped SU-8 parts on each other, and b) aligning a smaller part on the

corner of a larger part (see Subsection 3.2.1). The following major points

were identified for further study:

1) The handling strategy is robust, in the sense that the self-alignment

can be successful, even when the robot has large positioning errors.

Good alignment accuracy can still be achieved. This was quantitatively

measured in paper [II], by determining the yield as a function of the

bias and droplet volume.

2) The handling strategy is accurate, since it is determined by the

manufacturing precision of the parts and the receptor sites. The

droplet self-alignment assisted release overcomes the sticking effects

[34], where the part is adhered to the tool. This is because the

capillary force from surface tension is larger than the adhesion of the

part to the releasing tool [61]. The alignment accuracy is mostly

limited by the fabrication accuracy of the microparts. The alignment

accuracy was quantitatively measured in paper [II].

3) Self-alignment can happen even when the size of the parts do not

match exactly. This allows great flexibility in the handling, and

different possibilities were explored in papers [II] and [III].

4) The ambient environment has an effect on the handling strategy, due

to affecting the evaporation rate of the droplets. This was

quantitatively measured in paper [IV]. Furthermore, if the humidity of

the environment can be made high, no water dispenser is necessary, as

the water can be delivered to the receptor site through condensation.

This possibility was further discussed in paper [V].

5.1.1 Yield and statistical modelling [II]

In paper [II], the yield of the handling strategy, presented in [I], was

reported as a function of the process parameters. A large number (> 500) of

experiments with varying x-, y-, z-bias and water amount were conducted,

and the success of each self-alignment was recorded.

Using logistic regression [99], the yield was statistically modeled. The

model was used to calculate the optimal droplet volume and z-bias. For

300 μm × 300 μm parts, these were 1.8 nL and 47 μm, respectively. For the

range of acceptable parameters, the model was used to show that 98% yield

could be achieved when a) x-bias was inside ±82 �m; b) y-bias was inside

±88 �m; c) z-bias was inside 25 and 69 μm; or d) droplet volume was

between 0.97 and 3.07 nL; while other parameters were kept optimal. Of

the positioning dimensions, z-bias had the tightest bounds.


51

Furthermore, the model predicts correctly the intuitive result that the

yield is maximum when there is no x- or y-bias.

5.1.2 Accuracy, speed and trajectories [II]

In paper [II], the final alignment accuracy of the self-alignment process for

matching and mismatching shapes was measured using a scanning electron

microscope. Water droplets and SU-8 parts were used. In all cases, the

root-mean-square accuracy of the self-alignment was close to 2 μm and

0.4º.

SU-8 side walls have a slight taper, which is a result of the lithography

process. This alone can create a size difference of ~ 1-2 μm between the top

and the bottom side of the part, and was not taken into account when

conducting the experiments. As discussed in Section 3.2.2, the self-

alignment process is expected to be robust, in the sense that the part should

center to the receptor site even when there are small size differences.

Nevertheless, it is possible that these size differences made the alignment

accuracy worse, and parts with better manufacturing precision should be

used to measure the ultimate limits of droplet self-alignment. The

alignment accuracy was still reasonably high.

The alignment accuracy was measured using high-vacuum SEM. First, the

water droplet self-aligned SU-8 parts on top of SU-8 parts were left to air

dry. This drying process created a fairly strong bonding between the two

SU-8 parts, as discussed in [III]. While the bonding strength was not

measured quantitatively, qualitatively it was so strong, that detaching these

two parts using tweezers created significant damage to the parts before the

bonding failed. The chemical origin of this bond is not known. Therefore, it

was expected that a combination friction and adhesion would keep the

parts firmly in place, even when the samples were introduced to high

vacuum.

Using high-speed microscopy, the duration of the water droplet self-

alignment was measured to be around 100 ms, and to have large variations

from one test to another. In some cases, the standard deviation was as large

as 146 ms. Nevertheless, the alignment was successful in all cases,

regardless of the variations in the trajectories.

5.1.3 Mismatching and 3D structures [I] - [III]

As discussed in Section 3.2.1, self-alignment is even possible when the part

sizes are not matching exactly. In paper [I], first results of aligning parts to

corners of larger parts were shown (Fig. 20). The liquid amount was

identified as a critical parameter for the success of this handling strategy.

52

In paper [II], further testing of the alignment of mismatching shapes was

carried out. In the tests, 95% yield was achieved when aligning a part on the

corner of a larger part. In edge alignment experiments, the RMS-accuracy

in y-direction was measured as 3.4 μm and 0.3º. Fig. 21 shows a SEM

micrograph after the alignment of a square SU-8 part on the end of a

rectangular receptor site.

In paper [III], the edge alignment was shown in a reversed context: the

top part was large and rectangular, while the receptor site was significantly

smaller (Fig. 22). Using this method, overhanging structures could be

created.

Finally, in paper [III], flipping microparts 90º by surface tension was

shown (Fig. 23), as discussed in Section 3.2.3. By approaching the receptor

site from opposite direction, the direction of the flip could be reversed. The

handling strategy is an easy way to realize object-centric rotations, which

usually would otherwise require complicated kinematic structure.

Figure 20 Aligning a small part on the corner of a larger part. a) The gripper approaches with the part in its grasp. b) A water droplet is dispensed on the corner of the receptor site. c) The part contacts with the droplet, and the droplet forms a meniscus between the part and the corner of the receptor site. d-f) The part self-aligns to the corner of the receptor site. Originally published in [I]. With kind permission from Springer Science and Business Media.

Figure 21 Aligning a small, square-shaped, SU-8 part onto the end of a larger, rectangular part. The square part is approximately 300 μm × 300 μm × 40 μm. The picture is from one of the self-alignment experiments from [II]. Image was taken using a SEM.


53

Figure 22 Aligning a larger part on a smaller part, creating an overhanging structure. The part is a 600 �m × 300 �m × 40 �m and the receptor site is 150 �m × 300 �m × 40 �m. Originally published in [III]. © 2009 IEEE

Figure 23 Flipping parts using droplet self-alignment. This particular manoeuvre is very difficult for conventional microrobotics. a) The gripper approaches the receptor site with a part in its grasp. b) A water droplet is dispensed on the receptor site. c) The droplet wets the side of the part, forming a meniscus between the receptor site and the side of the part. d) The part is released and rotates 90º, the side now facing the receptor site. e-f) Water evaporates, leaving the part aligned and adhered to the receptor site. Originally published in [III]. © 2009 IEEE

5.1.4 Effects of environmental conditions on self-alignment [IV]

The water droplet self-alignment is affected by the ambient environment,

namely its temperature and humidity. This is partly explained by the fact

that the water droplet evaporation rate is affected by the liquid volume,

temperature and the humidity of the environment.

In paper [IV], firstly the droplet evaporation time was measured as a

function of temperature, humidity and liquid volume (Fig. 24). Naturally,

smaller droplets evaporated faster in lower humidity and higher

temperature. In self-alignment experiments, decreased relative humidity

was observed to increase self-alignment duration when aligning silicon

parts on hydrophilic patterns. Also, in very low humidity (5%), small water

54

droplets evaporated too quickly, before the self-alignment was completed,

resulting in failed alignment.

The results suggest that very low humidity and small droplet is

detrimental to water droplet self-alignment, due to droplet evaporating too

quickly. However, this can be compensated by increasing the water droplet

size.

Figure 24 Droplet evaporation time as a function of humidity and liquid volume in a) 25°C and b) 40°C. Smaller droplets evaporated faster in lower humidity and higher temperature. Size of each droplet was approximately 300 pL. Data from [IV].

5.1.5 Parallel delivery of liquid using mist [V]

In paper [V], a study on a method for delivering water to the receptor sites

by condensation from water mist was reported. An ultrasonic humidifier

created water mist, which was delivered to the vicinity of the receptor sites

(Fig. 25). The mist condenses on the surface, forming the droplet for self-

alignment.

In the paper, it was shown that the water condensation process is

approximately linear with respect to time and thus the amount of water can

be easily controlled by adjusting the duration of the humidifying. Part of the

difficulty in developing the model was to estimate the amount of water on

the receptor sites. A machine vision technique was developed that identified

droplets and estimated their volume from the microscope images.

The fact that there are also some droplets on the background, outside the

receptor site, complicates the wetting behaviour. The results showed that

with too large a bias, the yield of the self-alignment drops, partly because

the droplet is not fully confined inside the receptor site. Nevertheless, with

a suitably small bias (< 60% of part size in one direction), a 100% yield was

achieved.

The paper shows that using a humidifier is a viable strategy for the water

delivery. This eliminates the need for the water droplet dispenser, which is


55

a very delicate component in the system. The parallel nature of the water

mist condensation makes this strategy very appealing, as compared to the

serial nature of the single droplet dispenser.

Figure 25 Water mist is created using a humidifier, and directed near the parts. The mist condenses on the parts, forming the droplet for self-alignment. [V]

5.2 Silicon capillary gripper with self-alignment capability [VI]

In paper [VI], a novel capillary microgripper was reported. The basic

working principle of capillary gripping was outlined in Subsection 3.2.4.

The gripper was fabricated out of silicon using the method detailed in

Subsection 4.2.5. The final, fabricated gripper is shown in Fig. 26.

Figure 26 a) Two different gripper head shapes on a wafer, before singulation. The through-silicon holes are visible in the middle of the gripper heads. The die singulation trenches are also etched during the through-silicon etching. The singulation is done by cutting the small silicon bridges. b) Singulated gripper. A microfluidic connector is attached on the backside of the gripper. [VI]

Using Surface Evolver and geometric models, the gripper was shown to

have self-alignment capability, even when the parts do not have exactly the

same size as the gripper (Fig. 27).

Furthermore, using Surface Evolver models, the gripping force was

estimated for various gripper / part configurations. The gripping force was

estimated to be in the range of 100 μN to 1 mN for parts with lateral sizes

ranging from 250 μm to 500 μm.

The self-alignment capability of the gripper was demonstrated

experimentally in pick-and-place experiments. An image sequence of a part

aligning to the tool is shown in Fig. 28.

56

Figure 27 Shape of the energy well of the gripper. a) All models predict that the gripper is able to self-center the part when both the gripper and the part are matching squares; b) If, however, the part is rectangular, but the gripper is square, there are multiple minima for small amount of liquid, as predicted by the Surface Evolver model. With large enough amount of liquid, the minimum becomes unique and the part centers to the gripper. Data from [VI]

Figure 28 a-d) An image sequence of an experiment showing the self-alignment of a silicon chip to the capillary microgripper. Originally published in [VI]. © 2011 IEEE

5.3 Adhesive droplet self-alignment on oleophilic / oleophobic surfaces [VII]

In paper [VII], adhesive (i.e. glue) was used as the self-alignment droplet.

The components were made out of SU-8 and the alignment patterns were

oleophilic / oleophobic patterns made on ORMOCER®, fabricated using

the method described in Subsection 4.2.3. Electron micrograph of porous

ORMOCER® surface is shown in Fig. 29.


57

Figure 29 ORMOCER® surface after exposure to oxygen plasma. The porous silicon skeleton is left behind and it is highly oxidized.

Fig. 30 shows the patterns and the contact angle of the adhesive on the

patterns. The achieved contrast was large enough to confine the liquid on

the patterns without the adhesive wetting on the background.

Fig. 31 shows a self-alignment experiment using a droplet of adhesive on

the patterns. The successful self-alignment was shown with bias up to 90

μm for 200 μm × 200 μm microparts. The test was repeated 8 times with

100% yield when the amount of adhesive was 0.5 nL – 1.5 nL. The self-

alignment failed for various reasons when the amount of adhesive was

significantly outside this range.

The accuracy was estimated to be less than 1 μm in x-axis and 3.5 μm in y-

axis, where the measurement of this accuracy was limited by the accuracy of

the optical microscopes.

The duration of the self-alignment was about 500 ms, which is about 10

times longer than with water droplets of similar size and parts of similar

dimensions. The reason for this difference is the increased viscosity of the

adhesive.

58

Figure 30 Contact angle contrast on oleophilic / oleophobic patterns. a) Oleophilic gold patterns on fluorinated porous ORMOCER® background, fabricated using the method described in Section 4.2.3. b) Contact angle of an adhesive droplet on the oleophobic surface: 119º. c) Contact angle of adhesive droplet on the oleophilic patterns: 53º. Reprinted with permission from [VII]. Copyright 2011, American Institute of Physics.

Figure 31 a-c) An image sequence of an adhesive droplet self-alignment. The part is made of SU-8. The self-alignment takes about 500 ms. x- and y-bias were 35μm and 90μm, respectively. Reprinted with permission from [VII]. Copyright 2011, American Institute of Physics.

Conclusions

59

6. Conclusions

Droplet self-alignment using unconventional liquids, such as water and

adhesives, was studied experimentally. Several experiments were

conducted to measure properties that are relevant for practical applications

of the technique: yield, accuracy, duration and adaptability to handle parts

of varying size.

The alignment accuracy that was achieved is comparable to the

manufacturing precision of the microparts and patterns used in the

experiments. The accuracy is close or even better than the accuracies of the

best currently available industrial robots. Parts with more accurate

definition should be used to see the limits of the droplet self-alignment

process itself. The main benefit of using droplet self-alignment is that it can

achieve high accuracy without using high precision robotics.

In [VII], the accuracy of the adhesive droplet self-alignment (~ 3.5 μm)

was measured using optical microscopes, which had a comparable

resolution to the accuracies measured. Therefore, it is possible that the real

alignment accuracy is better than reported. As a future work, the alignment

accuracy should be validated using SEM or some other method e.g. vernier

patterns [35].

To achieve high yield, z-bias and droplet volume should be optimized and

controlled more accurately than x- or y-bias. When other parameters were

optimal, water droplet self-alignment could correct positioning errors as

large as half the lateral dimension of the part, but the positioning tolerances

in z-bias were much tighter. In some extreme cases, water droplet self-

alignment was observed to be successful even when the lateral bias was

almost as large as the size of the part i.e. when only a small edge of the part

was overlapping with the receptor site.

The estimated yield of the alignment process (around 99%) using water

droplet self-alignment is comparable or better than the yield of many self-

assembly processes [4]. However, even higher yields can be expected from

industrial chip assembly processes and to even measure accurately yields

that high, a full-scale industrial test plant would be required. Nevertheless,

for many applications, the achieved yield could already be high enough.

60

Droplet self-alignment is a very fast process, occurring in the time range

of 100 ms in the assembly cases presented in this summary. Whether

droplet self-alignment time should be counted into the cycle time of the

whole assembly process of a device depends on the process: if the

subsequent assembly steps cannot be started before the self-alignment step

is complete, the droplet self-alignment time has to be included in the cycle

time. In cases where the part can be left to self-align spontaneously, the

self-alignment time is not critical.

As a future work, droplet self-alignment methods could be integrated with

electrical connections and bonding steps. Bonding could already be

achieved using adhesives as discussed in Section 5.3, suggestions for

integrating droplet self-alignment with electrical connections were given in

Subsection 1.1.

There is ongoing work on the commercialization of the droplet self-

alignment process. Existing pick-and-place machines can be adapted for

the process, as long as they are able to house the liquid dispensing systems.

The required accuracy of the pick-and-place machine is very low, because

the droplet self-alignment can correct even large placement errors.

Some limitations of the results obtained in this thesis should be explained.

First of all, in [I – VI], water was used for droplet self-alignment, and the

results can be, to some extent, liquid specific. Some of the problems were

identified in [VII] and solutions were offered for adhesives and other oil-

like liquids. However, the work done in this thesis cannot be used to infer

what liquid properties are ideal for droplet self-alignment, because different

liquids were not exhaustively tested. On theoretical considerations alone,

high surface tension is expected to help liquid confinement, and low

viscosity and high surface tension speed up the self-alignment process.

These considerations cannot be used to conclude that high surface tension

and/or low viscosity are always beneficial for droplet self-alignment. In

future research, the experiments done in this work should be expanded by

using various liquids and the pros and cons of each liquid should be

carefully evaluated. Once the critical liquid properties are identified,

methods should be developed to tailor and enhance the desired liquid

properties. To close the circle, the handling method should be

benchmarked against solder self-alignment in comparable experiments.

Droplet self-alignment has been modeled, but the models often include

several simplifications, such as perfect contact line pinning to the pad edge,

no contact angle hysteresis, quasi-static meniscus etc. The typical output of

the models is the shape of the meniscus in equilibrium and self-alignment

forces. However, for wider adoption of droplet self-alignment, the models

should ultimately be able to answer a simple question: using a particular

Conclusions

61

combination of materials, liquids, and process parameters, will droplet self-

alignment work (as judged by practical criteria, such as yield, speed or

alignment accuracy)? While such model may still be quite far away, one

prerequisite for it is the prediction of wetting behavior (contact angle) of a

liquid on a new material. Development of such models in near future seems

more likely.

Furthermore, the microrobotic platform and the microgripper in

particular can have some effect on the results of the droplet self-alignment

experiments in [I – V]. When releasing the microparts, the part can adhere

unevenly to the tips of the microgripper. With a different type of a gripper,

e.g. a vacuum gripper, the adhesion force would be towards a different

direction and would have a different magnitude, which could affect the

success of the droplet self-alignment. As a future research topic, different

grippers should be tested.

Finally, the results, as grouped under the research objectives (Subsection

1.2), can be summarized as

� Several experiments were done to study water droplet self-alignment

assisted robotic placement. Yield, accuracy, capabilities to build

complex structures, and speed of the process were evaluated.

� A new capillary gripper was fabricated from silicon and experiments

showed that the gripper can pick and self-align parts

� Oleophilic / oleophobic patterns were developed for adhesive droplet

self-alignment. Self-alignment of parts on the patterns was shown

experimentally.

62

Acknowledgements

63

Acknowledgements

I am grateful to Adj. Prof. Quan Zhou for his active guidance throughout my

research. Indeed, he hired me as a young student and introduced me to the

whole field of micro- and nanotechnology. Also, this work would have not

been possible without the encouraging leadership of Prof. Heikki Koivo. He

helped in the preparation of this summary, but most importantly, visits to

his office were crucial for keeping the spirits high.

It has been an honor to work with the current and past members of the

Micro- and Nanorobotics research group, including Ville Liimatainen, Bo

Chang, Iiris Routa, Antti Virta, Reidar Udd, Mirva Jääskeläinen, Petri

Hänninen and Petteri Korhonen. They managed creating a comfortable and

a casual atmosphere, which made work days much more enjoyable. I thank

Prof. Yrjö Konttinen for introducing me to the field of cell biology in the

past year, during which I completed writing the summary of this thesis. I

am indebted to Dr. Tech. Antti Pohjoranta, Matias Lahti and my father

Prof. Hannu Sariola for proofreading and invaluable comments on the draft

of the summary. This work was made possible by the financial support I

have received from the Graduate School in Electronics,

Telecommunications and Automation, the Academy of Finland, the

European Commission and the Emil Aaltonen foundation. Assoc. Prof.

Takafumi Fukushima and Asst. Prof. Pierre Lambert did outstanding work

pre-examining this thesis.

I am grateful for the enthusiastic support from my family, mother Anna-

Paula, father Hannu and sisters, who primed me with a deep interest in the

natural world. Finally, I thank my wife Laura, for her loving support and

crazy projects which keep the life worth living for. Had she not wanted to

make Carmen dance with robots, we would not have met.

Any factual errors and omissions remain my sole responsibility.

Espoo, April 2012

Veikko Sariola

64

References

65

References

[1] J. Lu, “3-D Hyperintegration and Packaging Technologies for Micro-Nano Systems,” Proc. IEEE, vol. 97, no. 1, pp. 18-30, 2009.

[2] G. Q. Zhang, M. Graef, and F. van Roosmalen, “Strategic Research Agenda of ‘More than Moore’,” in 7th. Int. Conf. on Thermal, Mechanical and Multiphysics Simulation and Experiments in Micro-Electronics and Micro-Systems, 2006, pp. 1-6.

[3] C. Giovanni, G. Ripamonti, and R. Micheloni, “3-D Integration Technologies,” Proc. IEEE, vol. 97, no. 1, pp. 5-8, 2009.

[4] M. Mastrangeli, S. Abbasi, C. Varel, C. Van Hoof, J.-P. Celis, and K. F. Böhringer, “Self-assembly from milli- to nanoscales: methods and applications,” J. Micromechanics & Microengineering, vol. 19, no. 8, pp. 1-37, 2009.

[5] G. Swamy and S. Sarma, “Manufacturing Cost Simulations for Low Cost RFID systems.” [Online]. Available: http://www.autoidlabs.org/uploads/media/MIT-AUTOID-WH017.pdf. [Accessed: 26-Jul-2011].

[6] B. S. Vijayaraman and B. A. Osyk, “An empirical study of RFID implementation in the warehousing industry,” Int. J. Logistics Management, vol. 17, no. 1, pp. 6-20, 2006.

[7] D. V. Plant and A. G. Kirk, “Optical interconnects at the chip and board level: challenges and solutions,” Proc. IEEE, vol. 88, no. 6, pp. 806-818, 2000.

[8] M. F. Dautartas, G. E. Blonder, Y.-H. Wong, and Y. C. Chen, “A self-aligned optical subassembly for multi-mode devices,” IEEE Trans. Components, Packaging & Manufacturing Technology: Part B, vol. 18, no. 3, pp. 552-557, 1995.

[9] M. Wale and C. Edge, “Self-aligned flip-chip assembly of photonic devices with electrical and optical connections,” IEEE Trans. Components, Hybrids & Manufacturing Technology, vol. 13, no. 4, pp. 780-786, 1990.

[10] P. Lenz, “Wetting Phenomena on Structured Surfaces,” Advanced Materials, vol. 11, no. 18, pp. 1531-1534, 1999.

66

[11] J. F. Oliver, C. Huh, and S. G. Mason, “Resistance to Spreading of Liquids by Sharp Edges,” J. Colloid & Interface Science, vol. 59, no. 3, pp. 568-581, 1977.

[12] C. Kallmayer, H. Oppermann, G. Engelmann, E. Zakel, and H. Reichl, “Self-aligning flip-chip assembly using eutectic gold/tin solder in different atmospheres,” in 19th IEEE/CPMT Int. Electronics Manufacturing Technology Symp., 1996, pp. 18-25.

[13] R. Chen, B. Fennell, and D. F. Baldwin, “Flip chip self-alignment mechanism and modeling,” in Proc. Int. Symp. on Advanced Packaging Materials, 2000, pp. 158-164.

[14] W. Pittroff et al., “Flip chip mounting of laser diodes with Au/Sn solder bumps: bumping, self-alignment and laser behavior,” in Proc. 47th Electronic Components and Technology Conf., 1997, pp. 1235-1241.

[15] S. K. Patra and Y. C. Lee, “Modeling of Self-Alignment Mechanism in Flip-Chip Soldering - Part II: Multichip Solder Joints,” in Proc. 41st Electronic Components and Technology Conference, 1991, pp. 783-788.

[16] R. Zhao, Y. Zhang, R. W. Johnson, and D. K. Harris, “A Study of Normal, Restoring, and Fillet Forces and Solder Bump Geometry during Reflow in Concurrent Underfill/Reflow Flip Chip Assembly,” in Proc. 51st Electronic Components and Technology Conf., 2001, pp. 699-703.

[17] L. S. Goldman, “A heuristic force-height equation for molten axisymmetric solder joints,” in Proc. IEEE 43rd Electronic Components and Technology Conf., 1993, pp. 1120-1124.

[18] B. Yost, J. McGroarty, P. Børgesen, and C.-Y. Li, “Shape of a nonaxisymmetric liquid solder drop constrained by parallel plates,” IEEE Trans. Components, Hybrids, and Manufacturing Technology, vol. 16, no. 5, pp. 523-526, 1993.

[19] B. Fennell, S. Lee, and D. F. Baldwin, “Translation Solder Self-Alignment Mechanics Modeling for a Flip Chip in the Presence of a Viscous Fluid,” J. Electronic Packaging, vol. 132, no. 4, pp. 041013-1-5, 2010.

[20] M. Kong, S. Jeon, H. Au, C. Hwang, and Y.-cheng Lee, “Development and Experimental Validation of a 3-D Solder Self-Alignment Model for Alignment Accuracy Prediction of Flip-Chip Assembly,” IEEE Trans. Components Packaging & Manufacturing Technology, vol. 1, no. 10, pp. 1523-1532, 2011.

[21] D.-A. Wang and B. Liao, “Shaking assisted self-assembly of rectangular-shaped parts,” J. Materials Processing Technology, vol. 210, no. 2, pp. 343-350, 2010.

References

67

[22] K.-S. Moon, J. Wu, and C. P. Wong, “Study on Self-Alignment Capability of Electrically Conductive Adhesives (ECAs) for Flip-Chip Application,” in Proc. Int. Symp. Advanced Packaging Materials: Processes, Properties and Interfaces, 2001, pp. 341–346.

[23] K. Yasuda, J.-M. Kim, and K. Fujimoto, “Adhesive Joining Process and Joint Property With Low Melting Point Filler,” J. Electronic Packaging, vol. 127, no. 1, p. 12, 2005.

[24] K. Sato, K. Ito, S. Hata, and A. Shimokohbe, “Self-alignment of microparts using liquid surface tension—behavior of micropart and alignment characteristics,” Precision Engineering, vol. 27, no. 1, pp. 42-50, 2003.

[25] J.-M. Kim, K. Yasuda, and K. Fujimoto, “Resin self-alignment processes for self-assembly systems,” J. Electronic Packaging, vol. 127, no. 1, p. 18, 2005.

[26] T. Fukushima et al., “Surface tension-driven chip self-assembly with load-free hydrogen fluoride-assisted direct bonding at room temperature for three-dimensional integrated circuits,” Applied Physics Lett., vol. 96, no. 15, p. 154105, 2010.

[27] Y. A. Chapuis, A. Debray, L. Jalabert, and H. Fujita, “Alternative approach in 3D MEMS-IC integration using fluidic self-assembly techniques,” J. Micromechanics & Microengineering, vol. 19, no. 10, pp. 1-9, 2009.

[28] F. H. Howie and E. D. Hondros, “The surface tension of tin-lead alloys in contact with fluxes,” J. Materials Science, vol. 17, pp. 1434-1440, 1982.

[29] J. H. O. Lee and D. N. Lee, “Use of Thermodynamic Data to Calculate Surface Tension and Viscosity of Sn-based Soldering Alloy Systems,” J. Electronic Materials, vol. 30, no. 9, pp. 1112-1119, 2001.

[30] W. Gasior, Z. Moser, and J. Pstrus, “Density and Surface Tension of the Pb-Sn Liquid Alloys,” J. Phase Equilibria, vol. 22, no. 1, pp. 20-25, 2001.

[31] Z. Moser, W. Gasior, and J. Pstrus, “Surface Tension of Liquid Ag-Sn Alloys: Experiment versus Modeling,” J. Phase Equilibria, vol. 22, no. 3, pp. 254-258, 2001.

[32] J. Kim, Y. Shin, and K. Fujimoto, “Dynamic modeling for resin self-alignment mechanism,” Microelectronics Reliability, vol. 44, no. 6, pp. 983-992, 2004.

[33] C. J. Morris, S. A. Stauth, and B. A. Parviz, “Self-assembly for microscale and nanoscale packaging: steps toward self-packaging,” IEEE Trans. Advanced Packaging, vol. 28, no. 4, pp. 600-611, 2005.

68

[34] R. S. Fearing, “Survey of sticking effects for micro parts handling,” in Proc. IEEE Int. Conf. Intelligent Robots and Systems, 1995, pp. 212-217.

[35] T. Fukushima et al., “Multichip self-assembly technology for advanced die-to-wafer 3-D integration to precisely align known good dies in batch processing,” IEEE Trans. Components, Packaging & Manufacturing Technology, vol. 1, no. 10, 2011.

[36] Q. Zhou and B. Chang, “Microhandling using robotic manipulation and capillary self-alignment,” in Proc. IEEE/RSJ Int. Conf. Intelligent Robots & Systems, 2006, pp. 5883-5888.

[37] U. Srinivasan, D. Liepmann, and R. T. Howe, “Microstructure to substrate self-assembly using capillary forces,” J. Microelectromechanical Systems, vol. 10, no. 1, pp. 17-24, 2001.

[38] G. M. Whitesides and B. Grzybowski, “Self-assembly at all scales,” Science, vol. 295, no. 5564, pp. 2418-2421, 2002.

[39] R. Sharma, “Thermally controlled fluidic self-assembly,” Langmuir, vol. 23, no. 12, pp. 6843-6849, 2007.

[40] E. Saeedi, S. Abbasi, K. F. Böhringer, and B. A. Parviz, “Molten-alloy driven self-Assembly for nano and micro scale system integration,” Fluid Dynamics & Materials Processing, vol. 2, no. 4, pp. 221-246, 2006.

[41] J. Berthier, K. Brakke, F. Grossi, L. Sanchez, and L. Di Cioccio, “Self-alignment of silicon chips on wafers: A capillary approach,” J. Applied Physics, vol. 108, no. 5, p. 54905, 2010.

[42] T. L. Breen, J. Tien, S. R. J. Oliver, T. Hadzic, and G. M. Whitesides, “Design and self-assembly of open, regular, 3D mesostructures,” Science, vol. 284, no. 5416, pp. 948-951, 1999.

[43] S. E. Chung, W. Park, S. Shin, S. A. Lee, and S. Kwon, “Guided and fluidic self-assembly of microstructures using railed microfluidic channels.,” Nature materials, vol. 7, no. 7, pp. 581-587, 2008.

[44] P. W. K. Rothemund, “Folding DNA to create nanoscale shapes and patterns,” Nature, vol. 440, no. 7082, pp. 297-302, 2006.

[45] H.-J. J. Yeh and J. S. Smith, “Fluidic self-assembly for the integration of GaAs light-emitting diodes on Si substrates,” IEEE Photonics Technology Lett., vol. 6, no. 6, pp. 706-708, 1994.

[46] C. J. Morris, B. Isaacson, M. D. Grapes, and M. Dubey, “Self-assembly of microscale parts through magnetic and capillary interactions,” Micromachines, vol. 2, no. 1, pp. 69-81, 2011.

References

69

[47] K.-F. Böhringer, K. Goldberg, M. Cohn, R. Howe, and A. Pisano, “Parallel microassembly with electrostatic force fields,” in Proc. IEEE Int. Conf. Robotics and Automation, 1998, vol. 2, pp. 1204-1211.

[48] M. Gershovich and Y. C. Lee, “Gas flow effects on precision solder self-alignment,” IEEE Trans. Components, Packaging, & Manufacturing Technology: Part C, vol. 20, no. 4, pp. 305-311, 1997.

[49] X. Xiong, S.-H. Liang, and K. F. Böhringer, “Geometric binding site design for surface-tension driven self-assembly,” in Proc. IEEE Int. Conf. Robotics & Automation, 2004, pp. 1141-1148.

[50] M. Mastrangeli, W. Ruythooren, J.-P. Celis, and C. Van Hoof, “Challenges for Capillary Self-Assembly of Microsystems,” IEEE Trans. Components Packaging & Manufacturing Technology, vol. 1, no. 1, pp. 133-149, 2011.

[51] A. A. O. Tay, H. Li, X. Gao, J. Chen, and V. Kripesh, “Simulation of forces acting on tiny chips during fluidic self-assembly,” in Proc. 8th Electronics Packaging Technology Conf., 2006, no. 1, pp. 243-248.

[52] M. A. Verges, M. C. Larson, and R. Bacou, “Force and shapes of liquid bridges between circular pads,” Experimental Mechanics, vol. 41, no. 4, pp. 351-357, 2001.

[53] Q. Zhou et al., “A microassembly station with controlled environment,” in Proc. SPIE, 2001, vol. 4568, pp. 252-260.

[54] Q. Zhou, C. del Corral, P. J. Esteban, A. Aurelian, and H. N. Koivo, “Environmental influences on microassembly,” in Proc. IEEE/RSJ Int. Conf. Intelligent Robots & Systems, 2002, pp. 1760-1765.

[55] H. Lorenz, M. Despont, N. Fahrni, N. LaBianca, P. Renaud, and P. Vettiger, “SU-8: a low-cost negative resist for MEMS,” J. Micromechanics & Microengineering, vol. 7, no. 3, pp. 121-124, 1997.

[56] H. Lorenz, M. Despont, P. Vettiger, and P. Renaud, “Fabrication of photoplastic high-aspect ratio microparts and micromolds using SU-8 UV resist,” Microsystem Technologies, vol. 4, no. 3, pp. 143-146, 1998.

[57] T. Fukushima et al., “Self-assembly of chip-size components with cavity structures: high-precision alignment and direct bonding without thermal compression for hetero integration,” Micromachines, vol. 2, no. 1, pp. 49-68, 2011.

[58] H. Noda, M. Usami, A. Sato, S. Terasaki, and H. Ishizaka, “Self-aligned Positioning Technology to Connect Ultra-small RFID

70

Powder-chip to an Antenna,” in Proc. Electronic Components and Technology Conf., 2011, pp. 1009-1014.

[59] C. G. Tsai, C. M. Hsieh, and J. A. Yeh, “Self-alignment of microchips using surface tension and solid edge,” Sensors & Actuators A: Physical, vol. 139, no. 1–2, pp. 343-349, 2007.

[60] M. Kaneda, M. Yamamoto, K. Nakaso, T. Yamamoto, and J. Fukai, “Oscillation of a tilted circular pad on a droplet for the self-alignment process,” Precision Engineering, vol. 31, no. 2, pp. 177-184, 2007.

[61] T. Fukushima et al., “Three-dimensional integration technology based on reconfigured wafer-to-wafer and multichip-to-wafer stacking using self-assembly method,” in IEEE Int. Electron Devices Meeting, 2009, pp. 1-4.

[62] C. Bark, T. Binnenböse, G. Vögele, T. Weisener, and M. Widmann, “Gripping with low viscosity fluids,” in Proc. IEEE 11th Int. Workshop Micro Electro Mechanical Systems, 1998, pp. 301-305.

[63] H. Aoyama, S. Hiraiwa, F. Iwata, J. Fukaya, and A. Sasaki, “Miniature robot with micro capillary capturing probe,” in Proc. 6th Int. Symp. Micro Machine & Human Science, 1995, pp. 173–178.

[64] P. Lambert and A. Delchambre, “A study of capillary forces as a gripping principle,” Assembly Automation, vol. 25, no. 4, pp. 275-283, 2005.

[65] P. Lambert, “A contribution to microassembly: a study of capillary forces as a gripping principle,” Ph.D. dissertation, Faculté des sciences appliquées, Université libre de Bruxelles, Brussels, Belgium, 2005.

[66] F. Biganzoli, I. Fassi, and C. Pagano, “Development of a gripping system based on capillary force,” in 6th IEEE Int. Symp. Assembly and Task Planning: From Nano to Macro Assembly and Manufacturing, 2005, pp. 36-40.

[67] A. Vasudev and J. Zhe, “A capillary microgripper based on electrowetting,” Applied Physics Lett., vol. 93, no. 10, p. 103503, 2008.

[68] C. del Corral et al., “Droplet based self-assembly of SU-8 microparts,” in Proc. 2nd VDE World Microtechnology Congr., 2003, pp. 293–298.

[69] A. B. D. Cassie and S. Baxter, “Wettability of Porous Surfaces,” Trans. Faraday Society, no. 40, pp. 546-551, 1944.

[70] L. Gao and T. J. McCarthy, “Contact angle hysteresis explained,” Langmuir, vol. 22, no. 14, pp. 6234-7, 2006.

References

71

[71] A. Tuteja et al., “Designing superoleophobic surfaces,” Science, vol. 318, no. 5856, pp. 1618-22, 2007.

[72] Y. L. Chen, C. A. Helm, and J. N. Israelachvili, “Molecular Mechanisms Associated with Adhesion and Contact Angle Hysteresis of Monolayer Surfaces,” J. Physical Chemistry, vol. 95, pp. 10736-10747, 1991.

[73] G. D. Parfitt, “Surface Chemistry of Oxides,” Pure & Applied Chemistry, vol. 48, no. 4, pp. 415-418, 1976.

[74] C. J. Drummond, G. Georgaklis, and D. Y. C. Chan, “Fluorocarbons: Surface Free Energies and van der Waals Interaction,” Langmuir, vol. 12, no. 11, pp. 2617-2621, 1996.

[75] D. Barriet and T. R. Lee, “Fluorinated self-assembled monolayers: composition , structure and interfacial properties,” Current Opinion in Colloid & Interface Science, vol. 8, pp. 236-242, 2003.

[76] E. Sabatani and I. Rubinstein, “Organized self-assembling monolayers on electrodes. 2. Monolayer-based ultramicroelectrodes for the study of very rapid electrode kinetics,” J. Physical Chemistry, vol. 91, no. 27, pp. 6663-6669, 1987.

[77] R. N. Wenzel, “Resistance of solid surfaces to wetting by water,” Industrial & Engineering Chemistry, vol. 28, no. 8, pp. 988-994, 1936.

[78] R. Lipowsky, P. Lenz, and P. S. Swain, “Wetting and dewetting of structured and imprinted surfaces,” Colloids and Surfaces A: Physicochemical & Engineering Aspects, vol. 161, pp. 3-22, 2000.

[79] C. W. Extrand, “Contact angles and hysteresis on surfaces with chemically heterogeneous islands,” Langmuir, no. 2, pp. 3793-3796, 2003.

[80] L. Gao and T. J. McCarthy, “How Wenzel and Cassie were wrong,” Langmuir, vol. 23, no. 7, pp. 3762-3765, 2007.

[81] W. S. N. Trimmer, “Microrobots and micromechanical systems,” Sensors & Actuators, vol. 19, no. 3, pp. 267-287, 1989.

[82] K. F. Bohringer, U. Srinivasan, and R. T. Howe, “Modeling of capillary forces and binding sites for fluidic self-assembly,” in 14th IEEE Int. Conf. Micro Electro Mechanical Systems, 2001, pp. 369-374.

[83] K. A. Brakke, “Surface Evolver,” Experimental Mathematics, vol. 1, no. 2, pp. 141-165, 1992.

72

[84] C. Lin, F. Tseng, H.-C. Kan, and C.-C. Chieng, “Numerical studies on micropart self-alignment using surface tension forces,” Microfluidics and Nanofluidics, vol. 6, no. 1, pp. 63-75, 2008.

[85] H. Lua and C. Bailey, “Dynamic analysis of flip-chip self-alignment,” IEEE Trans. Advanced Packaging, vol. 28, no. 3, pp. 475-480, 2005.

[86] P. Lambert, M. Mastrangeli, J.-B. Valsamis, and G. Degrez, “Spectral analysis and experimental study of lateral capillary dynamics for flip-chip applications,” Microfluidics & Nanofluidics, vol. 9, no. 4–5, pp. 797-807, 2010.

[87] T. Tanikawa and T. Arai, “Development of a micro-manipulation system having a two-fingered micro-hand,” IEEE Trans. Robotics & Automation, vol. 15, no. 1, pp. 152-162, 1999.

[88] N. Dechev, W. L. Cleghorn, and J. K. Mills, “Microassembly of 3-D Microstructures Using a Compliant, Passive Microgripper,” J. Microelectromechanical Systems, vol. 13, no. 2, pp. 176-189, 2004.

[89] Q. Zhou, P. Korhonen, J. Laitinen, and S. Sjovall, “Automatic dextrous microhandling based on a 6 DOF microgripper,” J. Micromechatronics, vol. 3, no. 3–4, pp. 359-387, 2006.

[90] W. J. Plumbridge, “Defects in electronics and their significance for structural integrity,” Fatigue & Fracture of Engineering Materials & Structures, vol. 30, no. 5, pp. 432-442, 2007.

[91] R. R. A. Syms, E. M. Yeatman, V. M. Bright, and G. M. Whitesides, “Surface tension-powered self-assembly of microstructures - The state-of-the-art,” J. Microelectromechanical Systems, vol. 12, no. 4, pp. 387-417, 2003.

[92] P. Lambert and A. Delchambre, “Design rules for a capillary gripper in microassembly,” in 6th IEEE Int. Symp. Assembly and Task Planning: From Nano to Macro Assembly and Manufacturing, 2005, pp. 67-73.

[93] R. Ierusalimschy, L. H. de Figueiredo, and W. Celes, Lua 5.1 Reference Manual. Lua.org, 2006.

[94] D. Janssen, R. De Palma, S. Verlaak, P. Heremans, and W. Dehaen, “Static solvent contact angle measurements, surface free energy and wettability determination of various self-assembled monolayers on silicon dioxide,” Thin Solid Films, vol. 515, no. 4, pp. 1433-1438, 2006.

[95] V. Jokinen, L. Sainiemi, and S. Franssila, “Complex droplets on chemically modified silicon nanograss,” Advanced Materials, vol. 20, no. 18, pp. 3453-3456, 2008.

References

73

[96] K.-H. Haas and H. Wolter, “Synthesis, properties and applications of inorganic–organic copolymers,” Current Opinion in Solid State & Material Science, vol. 4, no. 6, pp. 571-580, 1999.

[97] S. Aura, V. Jokinen, M. Baumann, and S. Franssila, “Pore formation in inorganic-organic hybrid material by oxygen plasma treatment,” in Proc. Transducers, 2009, pp. 461-464.

[98] L. Sainiemi et al., “Fabrication and fluidic characterization of silicon micropillar array electrospray ionization chip,” Sensors & Actuators B: Chemical, vol. 132, no. 2, pp. 380-387, 2008.

[99] D. W. Hosmer and S. Lemeshow, Applied logistic regression, 2nd editio. John Wiley and Sons, 1989.

[100] H. S. Khoo and F.-G. Tseng, “Spontaneous high-speed transport of subnanoliter water droplet on gradient nanotextured surfaces,” Applied Physics Lett., vol. 95, no. 6, p. 063108, 2009.

74

Appendix A: Neglecting gravity in self-alignment

75

Appendix A: Neglecting gravity in self-alignment

The shape of a liquid meniscus is governed by the Young-Laplace equation

under gravity:

21 /1/1 RRgzp �� (11)

where p� is the overpressure inside the meniscus, is the density of the

liquid, z is the height, g is the gravitational acceleration and 2,1R are the

principal radii of curvature. Eq. 11 can be non-dimensionalized by using the

capillary length g

LC �

and the characteristic pressure CC Lp /� , and

setting CLzz /* , Cppp /* � � , CLRR /2,1*

2,1 to arrive at

*2

*1

** /1/1 RRpz � �� (12)

Hydrostatic pressure (gravity acting on the droplet) can be neglected when

CLzz �� 0* .

For a rectangular part where the liquid fills the bottom of the part fully (as

in Fig. 8), the overpressure under the part due to the part weight is

ghwl

gwlhwlmgp pp

pp

� (13)

where m is the mass of the part, p is the density of the part and ph is the

thickness of the part. Thus

� /*Cpp gLhp � (14)

76

Assuming that �p (for example, the density of SU-8 is

3kg/m1,190 p ), 2/1/ Cp Lg �� and we get Cp Lhp /* �� . The gravity of

the part can be neglected when Cp Lhp �� 0* .

To summarize: Assuming that a) the part bottom is fully wetted; b) the

part density is close to the density of the liquid; and c) the height of the

liquid film and the height of the chip are much smaller than the capillary

length of the liquid, the gravity can be neglected. Similar conclusions

cannot be drawn for parts with large lateral dimensions, where the droplet

wets only a small portion of the part bottom, since the overpressure due to

chip weight can be much larger.

Appendix B: Energy of a twisted meniscus

77

Appendix B: Energy of a twisted meniscus

To show self-alignment, the energy of the meniscus should be calculated as

a function of part angle � . In Appendix B of ref. [41], equation for the

surface area of a rectangle upon a twist was developed. When a rectangle

with the length l and height h is twisted by angle � along the axis of h ,

the surface area A is given by [41]

22

2 1ln12

aahahlA ��

(15)

whereh

la2�

. This can be rewritten as

)(arcsinh1 22

aaahA ��

(16)

If the chip is assumed to have same length and width ( wl in Fig. 8), the

total energy from twisting all the four faces is given by

)(arcsinh144 22

aaahAE ��

�� (17)

The torque T is then given by

)(arcsinh144 22

2

aaahAET ��

� ��

� �

��

��

(18)

In ref. [41], the equation for torque was given in a slightly different form,

but it can be checked that the both forms are equivalent. The direction of

torque was chosen as negative, to highlight that the torque is self-aligning.

Fig. 32 shows energy and torque plotted as a function of part angle.

78

Figure 32 Calculating surface energy and torque of a twisted meniscus using equations 17 and 18. Following values were used: 2mJ/m8.72 � , �m10 h and �m300 l . a)

Energy as a function of chip angle. The energy minimum is exactly when º0 � . b) Torque as a function of chip angle. Nm0 T exactly when º0 � .

The method can be extended to rectangular chips ( wl � ), by considering

the energy and torque of front and back faces separately from the energy

and torque of left and right faces.

Appendix C: Publications

79

Appendix C: Publications

�

9HSTFMG*aegeac+

ISBN 978-952-60-4640-2 ISBN 978-952-60-4641-9 (pdf) ISSN-L 1799-4934 ISSN 1799-4934 ISSN 1799-4942 (pdf) Aalto University School of Electrical Engineering Department of Automation and Systems Technology www.aalto.fi

BUSINESS + ECONOMY ART + DESIGN + ARCHITECTURE SCIENCE + TECHNOLOGY CROSSOVER DOCTORAL DISSERTATIONS

Aalto-D

D 6

9/2

012

Veikko Sariola

Droplet Self-A

lignment: H

igh-Precision R

obotic Microassem

bly and Self-Assem

bly A

alto U

nive

rsity

Department of Automation and Systems Technology


Veikko Sariola

DOCTORAL DISSERTATIONS

Date post:	19-Oct-2020
Category:	Documents
Upload:	others
View:	2 times
Download:	0 times

Aalto- DD Alignment: High- Precision Robotic Microassembly...

Documents