di strutture MEMS e microsistemi con elevato aspect-ratio utilizzando il microma-

Sommario

In questo lavoro ` e stato fatto un passo avanti significativo verso la fabbricazione

di strutture MEMS e microsistemi con elevato aspect-ratio utilizzando il microma-

chining elettrochimico del silicio in soluzioni a base di HF (ECM). L’ECM ` e una

tecnologia di microstrutturazione recentemente proposta che combina i vantaggi di

entrambe le tecniche di micromachining tradizionalmente utilizzate, l’alta flessibilit` a

delle tecniche dry e il basso costo delle tecniche wet. Tra le caratteristiche principali

dell’ECM sono degne d’essere menzionate la possibilit` a di cambiare l’anisotropia

dell’attacco (da uno a zero) e l’alto aspect-ratio delle strutture ottenibili. Queste

caratteristiche permettono di avere una maggiore flessibilit` a nella microfabricazione,

anche rispetto alla tecnologie dry, e rendono possibile la fabbricazione di strutture

MEMS ad alta densit` a in un singolo passo di attacco elettrochimico. Sono state

fabbricate strutture con elevato aspect-ratio, di forma e dimensioni differenti, cos-

tituite da masse inerziali free-standing equipaggiate di comb-fingers e sospese per

mezzo di molle ripiegate e non-ripiegate dal substrato. La tecnica ECM di fabbri-

cazione di strutture MEMS (processo, caratteristiche e svantaggi) ` e qui discussa nei

dettagli con riferimento ai risultati sperimentali. Le regole di design identificate

nella fase di sviluppo del processo sono state quindi applicate alla progettazione

e fabbricazione di tre prototipi di microgripper. I microgripper, infatti, sono dei

microsistemi complessi, di grande interesse attuale per la movimentazione di micro

oggetti, che integrano al loro interno strutture MEMS di varia complessit` a necessari

all’attuazione del gripper stesso. I microgripper sono quindi ideali per testare le

potenzialit` a della tecncica di micromachining ECM.

Abstract

In this work, a significant step towards the fabrication of very high aspect-ratio

MEMS structures and complex microsistems by silicon electrochemical microma-

chining in HF based solution (ECM) is given. ECM is a recently proposed bulk

micromachining technology combining advantages of both dry (high flexibility) and

wet (low cost) traditional micromachining tools. Among the main features of ECM

are worthy to be mentioned the possibility of changing the etching anisotropy (from

one to zero) as well as the high aspect-ratio of viable structures. Such features allow

an enhanced flexibility in silicon microfabrication to be achieved, even with respect

to dry etching technology, and enable high density MEMS fabrication by one-single-

etching step. Very high aspect-ratio structures, with different shape and dimension,

consisting of inertial free-standing masses equipped with comb-fingers and suspended

by folded or unfolded springs from the substrate were fabricated. ECM fabrication

technique (process, features and drawbacks) of MEMS structures is here detailed and

discussed with reference to experimental results. The mask-design rules identified

for MEMS fabrication in the development phase of the process were then applied to

the design and fabrication of three microgrippers prototypes. Microgrippers repre-

sent today an ideal benchmark for testing all the challenging tasks involving complex

and large area MEMS structures.

Contents

Introduction 1

1 MEMS: Applications and Fabrication 5

1.1 Definition of MEMS . . . . 5

1.2 Materials in microsystems . . . . 7

1.3 Benefits and Drawback of miniaturization . . . . 10

1.4 MEMS Applications . . . . 12

1.4.1 Microsensors . . . . 12

1.4.2 Optical MEMS devices . . . . 15

1.4.3 Microactuators . . . . 16

1.5 MEMS fabrication techniques . . . . 17

1.5.1 Thin-film deposition . . . . 18

1.5.2 Lithography . . . . 19

1.5.3 Etching . . . . 21

1.5.4 Substrate bonding . . . . 22

1.6 Surface Micromachining . . . . 23

1.7 Bulk micromachining . . . . 28

1.7.1 Wet etch . . . . 28

1.7.2 Dry Etch . . . . 31

1.7.3 High Aspect-Ratio Dry-Etch . . . . 34

1.7.4 SCREAM process . . . . 45

1.7.5 Performance of a comb drive actuator fabricated with DRIE . 47

2.1.1 The silicon/HF electrolyte contact . . . . 55

2.1.2 The J-V characteristic of Silicon/HF electrolyte . . . . 58

2.1.3 Dissolution chemistries . . . . 61

2.1.4 Electropolishing region . . . . 66

2.1.5 Pore formation . . . . 68

2.2 Macroporous silicon . . . . 70

2.2.1 Macropores in p-type silicon . . . . 70

2.2.2 Macropores in n-type silicon . . . . 72

2.2.3 Pore initiation . . . . 75

2.2.4 Stable pore growth . . . . 77

2.2.5 Porosity, growth rate and transport . . . . 80

2.2.6 Models for the macropore growth in n-type silicon . . . . 83

2.3 Porous silicon for micromachining . . . . 87

2.3.1 Micromachining with micro and mesoporous silicon . . . . 87

2.3.2 Micromachining with macroporous silicon . . . . 89

2.3.3 Capacitors . . . . 90

2.3.4 Silicon Oxide Microneedles . . . . 92

2.3.5 Trenching of silicon for oxidation . . . . 93

2.3.6 Photonic Crystals . . . . 94

2.3.7 Membranes for Micropumps . . . . 97

2.3.8 Free-standing compliant structures . . . . 98

3 Fabrication of free-standing MEMS structures by means of ECM 101 3.1 Experimental setup . . . . 101

3.2 Process Flow . . . . 103

3.3 Electrochemical Etching . . . . 104

3.3.1 Anisotropic phase . . . . 104

3.3.2 Isotropic Phase . . . . 109

3.4 Mask design rules . . . . 112

3.5 Oxidation and Metallization . . . . 119

4 Micro-manipulators: A branch of MEMS 125

4.1 Gripping methods . . . . 127

4.2 Non-contact Manipulation . . . . 128

4.2.1 Aerostatic gripper . . . . 128

4.2.2 Magnetic gripper . . . . 128

4.2.3 Laser Tweezer . . . . 129

4.3 Manipulation by physical contact . . . . 130

4.3.1 Sticking effects for microparts handling . . . . 130

4.3.2 Adhesive grippers . . . . 134

4.3.3 Vacuum Gripper . . . . 136

4.3.4 Mechanical grippers . . . . 137

4.3.5 Releasing . . . . 141

5 Design and fabrication of microgrippers by ECM technology 143 5.1 Open-Gripper . . . . 144

5.1.1 Operating Principle . . . . 145

5.1.2 Structure simulation . . . . 148

5.1.3 Mask design and fabrication . . . . 150

5.2 Closed-Gripper 1 . . . . 155

5.2.1 Operating principle . . . . 156

5.2.2 Structure simulation . . . . 157

5.2.3 Mask design and fabrication . . . . 158

5.3 Closed-Gripper 2 . . . . 162

5.3.1 Operating principle . . . . 162

5.3.2 Structure simulation . . . . 165

5.3.3 Mask design and fabrication . . . . 165

Conclusions 168

List of Figures

1.1 Geometry and circuitry connection of a Resonant Gate Transistor with a C-F resonant beam [2]. . . . . 6 1.2 Comparison of Stress-Strain relations of Silicon and Stainless Steel . . . . 8 1.3 SEM image of the prototype of the Draper Lab comb drive tuning fork

gyro. Due to the superior mechanical properties of monocrystalline silicon, a better performance was achieved using monocrystalline silicon with bulk micromachining process. . . . . 13 1.4 Basic micromirror structure for precision alignment of optical components [21]. 14 1.5 Cross-connected optical switch developed at University Neuchatel, Switzer-

land [24]. . . . . 15 1.6 MEMS approaches for optical cross-connect switching. (a) digital or 2D

technology. (b) Analog, scanning mirror, or 3D technology [25]. . . . . 16 1.7 Schematic drawing of the photolithographic steps with a positive photoresist. 20 1.8 Profile for isotropic (a) and anisotropic (b) etch through a photoresist mask. 21 1.9 Glass to silicon anodic bonding setup . . . . 22 1.10 SEM image of a surface-micromachined RF MEMS switch realized by poliymide

sacrificial etching [57]. . . . . 24 1.11 Process flow for single layer polysilicon micromachining. . . . . 25 1.12 Bulk micromachining. Anisotropically wet etched pit in <100> Si wafer (a).

Cross-section of pit (b). Membrane formation using backside wet etch (c). 30

1.13 Schematic representation of a generic system for plasma etching. . . . . . 32

gas (c). . . . . 33 1.15 Arrays of vertical silicon (100) nanopillars . Silicon nanopillars as narrow

as 40 nm diameter and 1.5 µm tall with vertical sidewalls arrayed on an ordered 235 nm pitch grid (a). Higher magnication view showing sidewall roughness less than 10 nm peak-to-peak (b). Wide eld view showing a large area array with long range order (c) [93]. . . . . 36 1.16 Schematic of the DRIE etched slanted trench (a). Deep trench profile,

plasma etched in silicon through silicon dioxide mask (b) [98]. . . . . 37 1.17 Scallop formed due to cyclical nature of BOSCH process is shown in SEM mi-

crographs. Deep trenches (c), magnified image of the sidewall with nanoscal- lops (d) [86]. . . . . 38 1.18 Micrograss in the etched features. (a) [95]. (b) [86]. . . . . 40 1.19 Aspect-ratio dependent etch ARDE. A series trenches with widths range

from 2 to 50 m (a). The final etch depths in wider trenches are larger (b).

The etch rate decrease as the aspect-ratio increase (c) [106] . . . . 41 1.20 Evolution of the depth of the silicon trench as a function of time (a), and

the apparent etch rate as a function of the aspect-ratio (b) [107]. . . . . . 42 1.21 The effect of ARDE under different process conditions on features with

widths from 2.5-10 µm, normal ARDE lag (∼10%) (a), optimized ARDE lag (<2%) (b), and inverse ARDE lag (-5%) (c) [108]. . . . . 42 1.22 Problems of the notching phenomenon in the conventional SOI process.

Cross-section of the conventional SOI process with DRIE showing the footing- induced notching (a); roughened and distorted bottom surface by the footing (b); loose silicon fragments resulting from the footing phenomenon (c) [109]. 44 1.23 SCREAM MEMS. Deposition and patterning mask on top of Silicon (a).

Silicon DRIE (b). Conformal oxide deposition (c). Timed, directional oxide

etch (d). Isotropic etching (e). Deposition of metal contacts (f). . . . . 46

1.24 Schematic of a lateral comb-drive actuator. . . . . 47

1.25 Relative electrostatic force vs. angle of comb finger (a). Normalized lateral

stiffness vs. angle of folded beam (b) [119]. . . . . 48

1.26 Uneven deep comb fingers: SEM image of the damaged comb finger (a) and schematic of uneven deep comb finger (b) [119]. . . . . 48

2.1 Cross-sectional view of a lateral anodization cell. . . . . 53

2.2 Cross sectional view of a single-tank anodization cell. . . . . 54

2.3 Cross-sectional view of a double-tank anodization cell. . . . . 55

2.4 Schematic model of a Metal/Electrolite interface: Formation of two charged layers with in the middle an high electric field. . . . . 56

2.5 Band diagram for a n-type silicon immersed into a HF-based electrolyte [132] 57 2.6 Typical p-type or illuminated n-type silicon J-V curve [132]. . . . . 59

2.7 Reaction regions under anodic polarization: Porous Silicon Region (a); Tran- sition region (b); Electropolishing Region (c). . . . . 60

2.8 Reaction scheme for the anodic, divalent dissolution of silicon electrodes in HF [132]. . . . . 63

2.9 Reaction scheme for the anodic tetravalent dissolution of silicon electrodes in HF [132]. . . . . 63

2.10 Dissolution valence n

as a function of anodic current density for low doped p-type and strongly illuminated, low doped n-type samples (<10

cm

). Etching performed with an aqueous solution with 2.5% HF at room temper- ature [132]. . . . . 65

2.11 Critical current density J

of <100> silicon electrodes for different HF concentrations as a function of the inverse absolute temperature 1/T. An Arrhenius-type behavior, with an activation energy of 0.345 eV, is onserved [132]. 66 2.12 Cyclic voltammograms of silicon anodes of different crystal orientation. The critical current density J

has been found to be largest for <100> oriented samples indipendently of the Hf concentration [132]. . . . . 67

2.13 Mechanism of pore formation. Random pore initiation on the Si surface

(a), formation of depletion layers and directional growth of pores (b), and

dissolution advance only at the pore tips (c). . . . . 68

2.15 Equilibrium (V=0) field currents and diffusion currents across the space- charge region for the macropore tip and wall region in p-type Si (a). Field currents and diffusion currents under forward bias (V>0). The tip currents are always larger than the pore wall currents. . . . . 72 2.16 Sketch of the equilibrium charge distribution and the electric eld around

pores in a n-type semiconductor electrode. A concentration gradient of holes is possible, because of compensation by electrons [132]. . . . . 73 2.17 SEM micrographs of spontaneous pore initiation on polished surfaces of

n-type Si electrodes anodized under white light illumination of the front side [128]. . . . . 75 2.18 SEM micrograph of an n-type silicon electrode with an array of inverted

pyramids, produced by standard lithography and subsequent alkaline etching (a), and an etched macropore array (2.5 Ω·cm, <100>, 2 V, 75 min, 5% HF) (b). . . . . 76 2.19 Complete etch pyramids lead to formation of single pore (a), while incom-

plete lead to formation of four pores (b). . . . . 77 2.20 The dependence of pore growth on doping density is shown for samples

anodized under indentical conditions. The center figure shows a sample with a well adjusted doping density in respect to the lithographic pattern.

High doping density leads to branching (a), while low doping density results in a dying of pores (c). A abrupt end of the pattern leads to branching even for a well selected substrate doping density (b) [152]. . . . . 78 2.21 Schematic illustration of current variation and coverage of sili- con oxide on

the surface of a pore bottom (a) [153]. SEM micrograph of the tip of two macropores (b). . . . . 79 2.22 Preferred macropore growth direction on [100](a) and [322](b) [154] substrates. 80 2.23 By an increase in bias or doping density the round (a) or slightly faceted (b)

cross-section of macropores becomes star shaped by branching (c), (d) or

spiking (e) along the <100> directions orthogonal to the growth direction [132]. 80

2.24 Cross-section of macropore arrays parallel to the electrode surface for an array square (a), for a triangular array of squares (b), and for array of squares and trenches (c). The pores and trenches (black) collect holes from the area indicated by the broken lines. . . . . 81 2.25 Normalized pores tip concentration as a function of the normalized pores

depth for: (1) the analytical reduced model, eq (2.25) (solid line); (2) the asymptotic model, eq. (2.28) (dotted line); (3) the numerical solution of eq. (2.21) (dashed line); (4) Lehmann model presented in [138] (dash-dot line) [155]. . . . . 86 2.26 Basic process sequence for sacrificial porous silicon. Growth of n-type epi

on p-type substrate (a). Using an oxide, or other mask, the epi is dened to reveal the substrate (b). Using a back contact, the exposed regions of the substrate are made porous (c). Finally, the porous silicon is removed to create free-standing structures (d). . . . . 88 2.27 Structure of different shape fabricated by means of electrochemical etching

in HF on on n-type silicon substrate. Cross section of a wall array (a), cross- section of microtubes array (b), 45

view of a microtips array (c), 45

view of a micropillar array (d), cross-section of a square shaped spiral array (e), and cross section of a round shaped spiral array (f). The latter is obtained using an isotropic etchant to define the initial defects (seeds), starting points of the electrochemical etching [151, 176]. . . . . 90 2.28 SEM micrograph of the cross-section of a regular array of macropore with

variable diameter (a) and (b) [177]. SEM top view (c) and tilted cross- section view of the SS’ section (d) of a meander-shaped trench obtained by means of electrochemical micromachining [178]. . . . . 91 2.29 Sketch of a MOS capacitor (a) and SEM image of a complete MOS capacitor

structure with 120×120 µm

top electrode surface [179]. . . . . 92

capacitor: the highly doped substrate and poly silicon layer which constitute the electrodes and the thin oxide-nitride-oxide (ONO) dielectric in between (b) [180]. . . . . 93

2.31 Fabrication process steps of a silicon dioxide microneedles for transdermal drug delivery system: silicon surface patterning (a); macropores growth (b); wet oxidation (c); microneedles release (d); reservoir definition. SEM micrographs of the front-side of a silicon chip for transdermal drug delivery which contains a 0.5 cm×0.5 cm array of silicon dioxide hollow microneedles with inner size of 4 µm, outer size of 6 µm, period of 10 µm, and protruding length of 100 µm (a) and (b) [184]. . . . . 94

2.32 SEM cross-section of a silicon wall array (2P4) fabricated by means of the electrochemical etching in HF electrolyte (a). SEM cross-section (b) and magnication (c) of a thick silicon dioxide layer produced by thermal oxida- tion of the silicon wall array [185]. . . . . 95

2.33 Different defects structures realized in macroporous silicon with 1.5 µm in- terpore distance [186]. . . . . 95

2.34 SEM images of three dimensionally shaped macroporous silicon. A con-

trolled defect plane between 2 Bragg mirrors with 10 modulation periods

(a). A cross-section of 3 pores with 3 modulations fabricated by the devel-

oped etching process (b). Single modulation (c). Structure with symmetry

close to cubic resulting from the widening of the single modulation (d). SEM

image of the prepared simple cubic photonic crystal (100)planes, with the

upper part showing the lithographically defined square lattice and lower one

the etched square lattice (e). SEM image of the (100) and (110) planes of

the crystal [187]. . . . . 96

2.35 SEM cross-section image for a sample with ten periods obtained by ten corresponding periods in the etching current profile (a). Schematic cross- section of the pump. The carrier liquid, including the dispersed particles of different size, is pumped periodically through the membrane with the asymmetrically shaped pores (b) [199]. . . . . 98 2.36 The basic steps of the SEEMS process [200]. . . . . 98 2.37 The beam-mass structure fabricated by SEEMS, proposed as accelerometer

(a). The clamping point of the cantilever (b) [200]. . . . . 99

3.1 Sketch of the electrochemical experimental set up. The main component is the electrochemical cell. . . . . 102 3.2 Process flow for the fabrication of free-standing MEMS structures

by means of the ECM technology. Definition of an array of holes and trenches (pattern) on the silicon surface (a) and (b); anisotropic electrochemical etching of the pattern into the substrates (c); release of the structures by isotropic etching (d); dry thermal oxidation (e);

metallization by sputtering (f). . . . 103 3.3 Experimantal I-V characteristic of the system electrolyte/silicon electrode

at a temperature of 22

C and a HF concentration of 5%(vol). . . . . 105 3.4 SEM micrographs showing cross-section and wall (a) and a zoom of the

cross-section (b) of the marked structure. Fit of the experimental data (c).

The markers obtained by enhancing the etching current for time necessary to affect the trench morphology without altering the HF concentration. Due to the constance of the applied etching current the porosity increase with the etching depth. . . . . 106 3.5 Critical current density J

and etching current density J

calculated for

a porosity P =50%, as a function of the etching time. . . . . 108

of pores with round shaped cross-section, but the coalescence of pores inside the trenches is loss after few micrometers (a). For bias voltages of +2.2 V (b) and +2.5 V (c) the coalescence of pore is maintained, but the cross-sesction degenerate form round to star shaped. . . . . 109 3.7 Etching current density and bias voltage as a function of the etching time

(a). SEM images of the resulting etched structures: cross section of an array of pores (P =50%) (b), and array of pores and wall of a trench (d). . . . . 110 3.8 Etching current density versus etching time. At 7700 s the current density

is switched to the releasing value J

(a). ∆J

versus growth depth. The amount of the increase of the etching current decrease with the depth of the released structures (b). SEM image of an array of macropore partially released (c). SEM image of a mass suspended over the substrate (b). It is clearly visible a gap between the mass and the substrate. . . . . 111 3.9 Sketch of the release phenomenon. . . . . 112 3.10 SEM images showing the dependence of the release behavior on the applied

release current density J

: ∆J < ∆J

(a); ∆J = ∆J

(b); ∆J > ∆J

(c). 112 3.11 Mask layout (a) and SEM images of serpentine spring, 2 µm thick and 30

µm in depth, after the anisotropic phase (b) and the isotropic phase (c). . 113 3.12 Mask layout (a) and SEM image (b) of a beam-mass structure with comb-

fingers anchored to the substrate by the anchoring patterned structure. The inertial mass is composed by a 1.5 µm side, and 6.8 µm pitch triangular array of holes. Beam and comb-fingers are 2 µm thick walls. SEM image of the cross section of a serpentine springs mass structure in which is clearly visible that the structure is suspended on and anchored to the substrate through the anchor walls (c). . . . . 114 3.13 Mask layout (a) and SEM image of a 30 µm in depth serpentine spring-

mass structure (b). The clamping point to the anchor wall is over-etched,

as sketched in (c). . . . . 116

3.14 Sketch of the KOH overetch at the edge of two walls disposed with the heads facing each other (a). SEM image of a beam-mass structure etched for 30 min in a 25% wt KOH solution at 50

C (b). SEM image of two arrays of dummy layers, with a 2 µm initial distance, etched for 1.5 h in a 25% wt KOH solution at 50

C (c). . . . . 116 3.15 SEM view at two different magnitudes of mass-spring structures of different

depth: 60 µm (a), 98 µm (b), and 130 µm (c). . . . . 118 3.16 SEM images of a serpentine spring-mass structure (a) and (b) and oa beam-

mass structure (c) and (d) after 1 h dry oxidation. The beam-mass structure is moved to respect its rest position because of the bend of the springs due to the mechanical stresses induced by the enhanced volume. . . . . 119 3.17 SEM images of the cross section of a 1.5 µm thick wall (a) and of the bottom

side of a anchoring walls (a) of a 98 µm in depth etched structure after 1 h dry oxidation. SEM image of the bottom side of the anchor walls after Au sputtering (c) and (d). In (c) it should be noted the semi-conformal nature of the metal sputtering process, which guarantees the electrical isolation of the free-standing structures. . . . . 120 3.18 SEM view of MEMS structures fabricated by ECM, after 6h wet oxidation.

Oxidation gives a preliminary proof-of-concept about structure actuation. . 122 4.1 The different phases in the operation of an adhesive gripper. . . . . 136 4.2 SEM micrographs showing the electrostatic microgripper [247]: break lines

on PSG membrane and etch channel formed under the PSG membrane, in which the microgripper is embedded (a); the flexible comb-drive sructures, extension arms, and the gripper jaws (b); clos-up micrograph of microgripper jaws (c); one-celled protozoa, being held by microgripper (d). . . . . 138 4.3 Electrostatic microgripper, combination of a linear motor, an amplication

mechanism and a grasping stage (a) Detail of a ground-link (b) Four polysil- icon layers required to fabricate the mechanical structure (c) [248]. . . . . 139 4.4 SEM view of silicon gripper actuated by SMA emlements (a). Parallel grip-

ping method (b) [249]. . . . . 140

5.1 SEM micrographs of the microgripper fabricated by Volland et al. using DRIE on SOI wafer (a) [277], and of the microgripper fabricated using ECM. 144 5.2 Comb actuator (a) and suspension beam (b) schematic view. . . . . 145 5.3 Operating principle of the gripper system (a) and realized system (b) [277]. 148 5.4 Sketch of electrode cross-section after oxidation (a). Equivalent beam cross-

section, which is assumed to be made of metal (b). . . . . 149 5.5 Horizontal displacements ux (a) and vertical displacement uy (b) with a load

of 0.225 mN. . . . . 149 5.6 Mask layout (a) and SEM image showing a view of the fabricated structure. 151 5.7 Mask layout (a) and SEM image view of the anchoring point of the of the

spring system on the substrate. . . . . 152 5.8 Mask layout (a) and SEM images (b) and (c) of the tip of the gripper fingers. 153 5.9 SEM image view of a 30 µm (a),(c), and (e) and 60 µm (b),(d), and (f) in

depth grippers. . . . . 154 5.10 Schematic view of the left side of the designed structure (a). SEM image

of a view of the device fabricated by means of ECM (b). Systems of beam springs 2 µm thick with length L=275 µm and L

=500 µm, L

=184 µm, sustain the movable parts. The blue areas are anchored to the substrate. . 155 5.11 Comb actuator (a) and suspension beam system (b) schematic view. . . . 156 5.12 Horizontal displacement ux (a) and vertical displacement uy (b) with a load

F

=0.5 mN. . . . . 158 5.13 Mask layout (a) and SEM image of a view of a beam springs system of a

gripper 60 µm in depth. . . . . 159 5.14 SEM image of the finger tips after the anisotropic phase of the etching (a)

and after the isotropic phase (b) of a gripper 60 µm in depth. SEM image view of a broken structure, in which the gripper tip is moved (c). . . . . . 160 5.15 SEM image of the fingers base after the anisotropic phase (a) and after the

isotropic phase of the etching (b). . . . . 161

5.16 SEM images views of a 90 µm deep gripper. . . . . 162

5.17 Schematic view of the grasping force sensor (a). SEM image of a view of the device fabricated by means of ECM (b). Two beam springs 2 µm thick with length L

=500 µm, and L

=184 µm, and four open-folded beams 2 µm thick with characteristic lengths L=240 µm, and l=10 µm sustain the movable parts. The blue areas are anchored to the substrate. . . . . 163 5.18 SEM image of the tri-plate capacitive sensor (a). Capacitive sensor schematic

view (b). Sensor suspension open-folded spring schematic view (c). . . . . 164 5.19 Horizontal displacement ux (a) and vertical displacement uy (b) with a load

of 20 µN applied in the x direction on the tip of the gripper finger. . . . . 165 5.20 SEM images views of the sensorized gripper: parallel plate sensor (a) and

(b); gripper finger base (c) and (d); open-folded spring (e); finger tips (f). . 167

List of Tables

1.1 Comparison of silicon properties and other material properties . . . 9

1.2 Combination of materials and etchants for surface micromachining. . . . . 27

Introduction

Micro Electrochemical Systems (MEMS) is a technology that uses the processes de- veloped for the integrated circuit fabrication to build microscopic electromechanical devices that combine electrical and mechanical components. The field of MEMS has merged as a technology with significant impact on every day life. MEMS pro- vide inexpensive means to sense and, in a limited way, control physical, chemical and biological interactions with nature. Today MEMS technology is still far from producing complex structures, especially three-dimensional, using low-cost, straight- forward processes. Suspended, deformable micro-mechanical structures have been to date fabricated by mainly using two different methods: anisotropic wet etching tools, such as KOH or TMAH; dry etching tools, such as Deep Reactive Ion Etching (DRIE). Wet anisotropic etching entails the use of a simple set-up at low cost. How- ever, dry etching tools are usually preferred to standard wet techniques, because they offer an enhanced flexibility in obtaining high aspect-ratio structures, which enable the fabrication of a wider range of mechanical elements. On the other hand, such tools are somewhat expensive, and anyway more expensive than wet methods. Am alternative method is the electrochemical micromachining of silicon in HF based so- lution (ECM) that solves many problems of standard MEMS fabrication techniques.

It allows a higher flexibility of shape, size and depth of the etched structures with respect to standard wet etching. Furthermore, it overcomes some limitations of dry etching tools, such as the reduced value of viable aspect-ratio.

In this work, the ECM technology has been pushed to the fabrication of very

high aspect-ratio free-standing structures in order to obtain MEMS structures as

well as complex microsystems. The technique consists of two distinct phases: 1) an initial anisotropic phase, and 2) a final isotropic phase. The anisotropic phase is exploited to deep etch the layout into the silicon substrate and ends as soon as a given etching depth is reached. At this stage the etched structures are still anchored to the bulk substrate through the bottom of the structure itself. The next isotropic phase is exploited to etch the fabricated structures at their bottom and, in turn, release them from the substrate. In order to apply this fabrication technique to produce MEMS structures it is necessary to electrically isolate the dynamic section and from the static section of microstructures for electrical actuation and sensing.

To achieve this goal, we can exploit the idea that underlies the SCREAM (Single Crystal Reactive Etch and Metallization) process for MEMS fabrication. The bulk silicon serves as both the structural and sacrificial material. A silicon oxide thermal growth and a subsequent sputtering deposition of metal that coats the top and the side walls of the structures are then exploited to form highly conductive electrodes for electrostatic actuation or capacitive sensing.

The ECM technology has been tailored, in a first phase, to MEMS fabrication

by using test structures, in which inertial masses were anchored to the substrate by

high aspect-ratio folded and un-folded springs. These structures can be envisioned as

basic micro-mechanical blocks of every MEMS structure. A few experiments about

oxidation and sputtering metallization of these structures have been also prelimi-

nary performed, in order to evaluate the fabrication of electrically actuated MEMS

structures. After MEMS-like structure fabrication by ECM has been thoroughly

investigated, the know-how on such etching technology has been applied to the

fabrication of three different microgrippers. The fabrication of microgrippers for

micro-object handling and assembling is today a very challenging task, subject of

study of many research groups in the world. Actually, microgrippers would greatly

reduce assembly costs of hybrid micro-systems, which are a great portion of fab-

rication costs and are mainly governed by degree of automation, flexibility of the

assembling tools. Moreover microgrippers encompass all the challenging tasks of a

complex and large area MEMS structure, thus making these structures ideal to test

the potential of the ECM technique. Every designed microgripper is electrostatically

actuable, one of them for closing and two of them for opening their fingers under voltage application. The opened type microgripper consists of an electrostatically driven micro-actuator which provides the force to operate the gripper tweezers. The microactuator generates a linear motion, which is converted into a rotational “grip- ping” motion by a hinge system. The two closed type microgrippers consist of an electrostatically driven micro-actuator which provides the force to operate a single gripper tweezer, and a capacitive force sensor, for gripping force measurement, con- nected to the second tweezer. The gripping force sensor permits a secure grasping without applying excessive forces. Both microgrippers are equipped with the same electrostatically driven actuator, designed to work also as force/displacement sen- sor. An alternative capacitive sensor has been designed for the second microgripper.

To obtain informations about grippers structural stiffness and their behavior during actuation and sensing, mechanics of the structures has been simulated with a finite elements method (FEM) using the simulation tool ANSYS.

The Thesis is organized as follows:

• In Chapter 1 an overview to the MEMS environment is given, including appli- cations and standard manufacturing techniques;

• In Chapter 2 an introduction of the mechanisms of silicon dissolution in HF based solutions, with particular attention to the macroporous formation in n-type silicon and its application as microstructuring technique is given;

• In Chaper 3 the ECM fabrication technique for high aspect-ratio free-standing structures is presented and discussed;

• In Chapter 4 an overview about micromanipulators and microgrippers is given;

• In Chapter 5 the operating principle, the simulation results, and the results of the fabrication process relating to the mask layout are discussed;

• Conclusions take stock of the results and identify possible lines of future re-

search.

Chapter 1

MEMS: Applications and Fabrication

Small high performance machines has always been a dream of scientist and engi- neers. MEMS technology has transformed this dream into a reality. The beginning of MEMS history can be identified in 1959 when Richard Feynmann, in “There’s Plenty of Room at Bottom” [1] tried to stimulate the invention of new micro-fabrication techniques, arguing that the entire Encyclopedia Britannica could be printed on a pinhead. Only eight years later comes the first “MEMS” device: in 1967, Nathalson et al. [2] uses for the first time the “surface micromachining” for fabricating the “res- onant gate transistor” which consisted in a gold oscillating cantilever (Figure 1.1).

However, only in 1989 in a meeting held in Salt Lake City was used for the first time the acronym MEMS - (Micro Electro Mechanical Systems) [3].

1.1 Definition of MEMS

Microsystems, are very small systems or systems made of very small components.

The name implies no specific way of building them and no particular type of func-

tionality. MEMS is an acronym for Micro-Electro-Mechanical Systems, where: Mi-

cro establishes a scale, Electro suggests electricity and electronics, and Mechanical

Figure 1.1: Geometry and circuitry connection of a Resonant Gate Transistor with a C-F resonant beam [2].

suggests moving parts. Nonetheless, the MEMS concept encompasses many others fields, including thermal, magnetic, fluidic, and optical devices and systems, with or without moving parts. In practice, MEMS shares several common features:

• MEMS involve both electronic and non-electronic elements, and perform func- tions that can include signal acquisition (sensing), signal processing, actuation, display, and control. They also serve as vehicles for performing chemical and biochemical reactions and assays;

• MEMS are systems in the true sense, which means that important issues, such as packaging, partitioning into components, calibration, signal-to-noise ratio, stability, and reliability.

Most MEMS devices and systems are fabricated using the standard Integrated

Circuit (IC) processes. The micro-mechanical elements are fabricated with special-

ized techniques called micromachining. Three main assets of MEMS are their size,

easy batch fabrication, and seamless integration with microelectronics. These assets

of MEMS have produced fully self-contained systems and have helped to expand the distribution of MEMS and reduced the production costs.

The field of MEMS has, over the past 20 years, emerged as a technology that promises to have significant impact on every day living in the near future. MEMS provide inexpensive means to sense and, in a limited way, control physical, chemical and biological interactions with nature. Nowadays the impact on every day living of MEMS is bigger than we can account. A modern car can be equipped with several pressure sensors (to control the pneumatics pressure or the air in the combustion chamber), several accelerometers (to activate airbags or the electronic stability sys- tem), gyroscopes (for the acceleration slip regulation), and so on. In addition to automotive, which is very important for the development of MEMS applications, the areas of application are numerous and ever-widening: environmental monitor- ing, biomedicine, telecommunication, industrial automation, optics, etc. This is the main reason why it is important develop new silicon micromachining techniques, in order to increase the wide spread of the micro-systems and reduce the fabrication costs.

1.2 Materials in microsystems

The benefits that bring MEMS technology suggest that there isn’t reason to limit

the choice of the material. The choice of materials in microsystems is determined

by microfabrication constraints. Integrated circuits are formed using various semi-

conductors, metals and insulators that can be deposited and patterned with high

precision. Most of these are inorganic materials (silicon, silicon dioxide, silicon

nitride, aluminum, and tungsten), although certain polymers are used as well. Mi-

crofabrication that extends beyond conventional microelectronics open the way to a

much broader range of materials and to a corresponding set of additional techniques

such as electroplating of metals, and molding of plastics. Although this, to the

acronym MEMS it is associated immediately silicon. In fact, silicon is the mostly

utilized material in microsystems industry. The use of silicon has its origin in the

industry of integrated circuits, but the main reasons to choice silicon are:

Figure 1.2: Comparison of Stress-Strain relations of Silicon and Stainless Steel

• Optimal knowledge of electrical properties;

• Monocrystalline substrate, cheap to produce;

• Availability of a large number of processing techniques easily accessible;

• Good knowledge of mechanical properties;

• Mechanical properties ideal for most application.

Knowledge of electrical properties and cost of substrate have certainly been two important starting points for the use of silicon in microelectronics industry. Me- chanical properties of silicon have been also widely studied and documented. This has extended the silicon applications beyond the manufacture of integrated circuits.

Silicon has proven to be a material ideal for the integration of electrical, mechan-

ical, optical and thermic function. In particular, monocrystalline silicon deforms

elastically up to the fracture point, without any mechanical hysteresis, is lighter

than aluminum, and with an elasticity modulus similar to stainless steel as shown in

Figure 1.2. Stainless steel is normally used as reference material, given its wide use

in the manufacture conventional transducers. The mechanical properties of silicon

Properties Si<111> Stainless Al Al ₂ O ₃ SiO ₂ Quartz

Steel (96%)

Young’s

Modulus (GPa) 179 200 70 303 73 107

Poisson’s

Ratio 0.27 0.3 0.33 0.21 0.17 0.16

Tensile

Strength (GPa) 7 3 0.17 9 8.4 9

Thermal Expansion

Coefficient (10/K) 2.3 16 24 6 0.55 0.55

Thermal Conductivity

at 300 K (W/cm K) 1.48 0.2 2.37 0.25 0.014 0.015 Melting

Temperature ( ^◦ C) 1414 1500 660 2000 1700 1600 Table 1.1: Comparison of silicon properties and other material properties

are anisotropic, i.e. dependent on the orientation with respect to crystalline axis. In

Table 1.1 is shown a comparison between the properties of silicon and other materi-

als commonly used in MEMS manufacturing. It is easy to understand why silicon is

the mostly used material in micromachining. A direct consequence of the monocrys-

tallinity of silicon is the uniformity of the mechanical properties on an entire batch

of wafer. Until the dopant concentration don’t reach high values (∼ 10 ²⁰ cm ⁻ 3),

there is no residual stress. One of the techniques commonly used in MEMS manu-

facturing is the surface micromachining. In surface micromachining polysilicon and

amorphous silicon are used as structural materials. Actually the mechanical prop-

erties of polysilicon and amorphous silicon, although dependent on the deposition

conditions, are similar to that of monocrystalline silicon. The substantial difference

lies in the higher intrinsic stress of both polysilicon and amorphous silicon.

The good thermal conductivity of silicon, slightly less than the aluminum con- ductivity, allow to use silicon as heat sink in complex integrated circuits, or to use silicon in temperature sensors. Moreover, up to 500 ^◦ C the silicon properties remains constants, allowing the use of silicon for application in harsh environments.

Despite the interaction with gases, fluids and enzymes is still under study, silicon can be considered resistant to most of chemicals used in the main MEMS applica- tions. Moreover, has been demonstrate the bio-compatibility of silicon, i.e. when a silicon sensor is placed inside the human body, any toxic substance is released. For all this reason silicon remain the mostly used material in MEMS manufacture.

1.3 Benefits and Drawback of miniaturization

In the development of MEMS technology, it is important to understand that the miniaturization of macro machines is not the most effective way to create MEMS applications and explore their usability. The same physical laws govern both the macro and the micro domain [4]. Miniaturizing a device can enhance its properties as well as introduce new problems to solve.

The feasibility of millimeter-sized devices with sub-micron precision offered by MEMS has enabled the introduction of sensors into systems where conventional macroscopic devices could not be used. For example, these dimensions are partic- ularly suitable for use in medicine and surgery. Moreover, the scaling of the size corresponds to a change in the physical properties of the devices, which leads to several advantages including:

• An increase of the sensitivity: very small signals can be detected;

• Lower power consumption;

• Better performance;

– The lower mass allow to some systems to be faster, and reduce vibration problems due to the enhanced resonant frequency;

– Using micro-fluidic devices is possible to analyze small quantity of fluid,

and in turn small quantity of reagents and analytes;

– An increase of resolution and contrast in optic devices.

The technical realization of MEMS differs significantly from those used for tradi- tional devices. This gives rise, in most cases, to a greater reliability for the following reasons:

• The structures are monolithic, seamless, without welding or joint subject to breakage;

• The final device is usually sealed in a controlled environment to prevent cor- rosion and the presence of dust inside. This avoids the use of lubricants and maintenance.

• MEMS devices can be self-calibrating;

• Redundant copies of certain parts of the object can be put on the same device, thereby ensuring operation even in case of failure of one or more structures.

The smaller mass of microsystems is reflected in a reduced amount of material and therefore lower weight. This leads to particular benefits in mobile devices, but especially in space applications, where costs can be greatly reduced by reducing the weight of the equipment necessary for navigation. Moreover, the small size of MEMS allows batch processing to be performed, i.e. the simultaneous production of thousands or million of pieces, with a consequent reduction of production costs.

Despite the many positive aspects of miniaturization, some problems in device

design also arise from it. For example, the physical properties of thin films of ma-

terial, can be dramatically different from those of bulk materials. Hence, certain

properties that are critical in devices performance, for example, the elastic modulus

or residual stress of a suspended beam, must be monitored to ensure repeatability

from device to device. This demands new methods for material properties measure-

ment. Moreover, the MEMS technology, using processes typical of the integrated

circuit industry, is partly limited (or complicated), in the fabrication of structures

that develop in the direction orthogonal to the silicon wafer surface. We still con-

sider compatibility issues between surface micromachining and bulk micromachining,

which are often used together in the manufacturing of some devices.

1.4 MEMS Applications

The field of MEMS is so rich and varied, that it is a challenge to select a small set of devices to illustrate the field. Although, a little description of the principal application of MEMS, can give an idea of the influence of MEMS on the modern society.

1.4.1 Microsensors

MEMS are now widely popular sensors for measuring motion, acceleration, incli- nation, and vibration. MEMS sensors are System-in-Package solutions, delivering high resolution and low power consumption, in an extremely small size. The sensors can add an intuitive man-machine interface to a mobile phone, MP3/MP4 player, PDA, or game controller, creating interaction by linking the user’s movements to applications, navigation, gaming, and much more. MEMS accelerometers are also commonly integrated as vibration detectors in today’s electronic home appliances, such as washing machines or dryers, to alert users to unbalanced loads and to pro- tect against the excessive wear of parts, before a failure occurs. One and two axis accelerometers are widely used in the automotive market for passive safety systems, like frontal and lateral air-bags; pressure sensors are used to control of the pneumat- ics or the air in the combustion chamber. Accelerometers and gyroscopes are also used in navigation systems and active safety systems, like ABS and Vehicle Dynamic Control.

An accelerometer measures the rate at which the velocity of an object is chang-

ing. A typical accelerometer utilizes a proof mass, a spring (or its equivalent) joining

the proof mass to a stationary housing or substrate, and a sensor to measure dis-

placement of the proof mass. The accelerometer is attached to the moving object,

and as the object accelerates, inertia causes the proof mass to long behind as its

housing accelerates with the object. The force exerted on the proof mass (given

by Newton’s second law) is balanced by the spring, and as the displacement of the

spring is it self proportional to the applied force, the acceleration of the object is

proportional to the displacement of the proof mass.

Figure 1.3: SEM image of the prototype of the Draper Lab comb drive tuning fork gyro.

Due to the superior mechanical properties of monocrystalline silicon, a better performance was achieved using monocrystalline silicon with bulk micromachining process.

The displacement sensor is the key component in determining overall accelerome- ter performance in terms of sensitivity, stability, and packaging constraints. Presently available devices utilize any of several sensing techniques, including capacitive [5], piezoresistive [6], piezoelectric [7], optical interference [8, 9], and tunneling ap- proaches [10].

A pressure sensor measure pressure, typically of gases or liquid. A typical pres- sure sensor utilizes a force collector as a diaphragm placed to separate two chambers, to measure strain or deflection due to the different pressure in the two chambers.

Also in this case the deflection sensor is the key component. Presently, available devices make use of several sensing techniques, such as piezo-resistive [11, 12], ca- pacitive [13], electromagnetic [14], piezoelectric [15] and optical [16].

A gyroscope is a device for measuring orientation, based on the principles of

conservation of angular momentum. Many types of MEMS gyroscopes have been

Figure 1.4: Basic micromirror structure for precision alignment of optical components [21].

reported in the literature, most of which falling in the categories of tuning-fork gyros, oscillating wheels, Foucault pendulums, and wine glass resonators.

Tuning fork gyroscopes contain a pair of masses that are driven to oscillate with equal amplitude but in opposite directions. When related, the Coriolis force creates an orthogonal vibration that can be sensed by a variety of mechanisms. The Draper Lab Gyro [17, 18, 19] shown in Figure 1.3 uses comb-type driving structures to induce the tuning fork into resonance. Rotation causes the proof masses to vibrate out of plane, and this motion is sensed capacitively with a custom CMOS ASIC.

Many reports of vibrating-wheel gyroscopes also have been published [20]. In

this type of gyroscope, the wheel is driven to vibrate about its axis of symmetry,

and rotation about either in-plane axis results in the wheel’s tilting, a charge that

can be detected with capacitance electrodes placed under the wheel.

Figure 1.5: Cross-connected optical switch developed at University Neuchatel, Switzer- land [24].

1.4.2 Optical MEMS devices

Optical MEMS (MOEMS) devices are products ranging from a few microns to a few centimeters in size that combine mechanical, electrical, and optical components.

Numerous types of MOEMS devices have been used successfully in a wide range of applications, including optical communications, display systems, biomedical in- strumentation, and adaptive optics [22, 21]. Because MOEMS devices have the advantages of small size, high robustness, and low power consumption, they have recently attracted particular attention for space optical systems, such as the James Webb Space Telescope [23].

The most commonly used MOEMS device is the micromirror, which has rapid

tilting speed and high reliability compared with large conventional mirrors (Fig-

ure 1.4). For space optical systems, research into optical MEMS-based micromir-

ror arrays has been reported for multi-object spectroscopy [26] and adaptive op-

tics [27, 28]. The development of free-space MOEMS technology has also enabled

Figure 1.6: MEMS approaches for optical cross-connect switching. (a) digital or 2D tech- nology. (b) Analog, scanning mirror, or 3D technology [25].

the implementation of optical switching, in particular optical cross-connect switch- ing (Figure 1.5). To date, several types of MEMS-based optical switches that have been reported include 1×2 switches [29, 30], 2×2 switches [24], 1×4 switches [31, 32], N×N 2D or 3D switches (N= 4, 8, 16, 32, ...) (Figure 1.6) [25, 33, 34]. The superior performance of cross-connect switching includes all-optical mode, low insertion loss, low cross talk, low wavelength and polarization dependence, higher port count, and the integration of micro-optics and micro-actuators on the same substrate.

1.4.3 Microactuators

The development of microactuators has been a key research area in MEMS technol- ogy. A simple microactuator can consist of a beam or a membrane as the moving element, actuated by a driving potential. This potential can be piezoelectric [35], electrostatic [36], electromagnetic [37], thermal bi-morph, thermopneumatic [38], shape memory alloy, or hydrogel [39].

Microactuator are largely used in microfluidics. Microfluidics is the science of manipulate extremely little quantity of liquids. Microfluidic systems have enabled familiar systems such as inkjet printers and no so familiar systems such as mi- crochips used for analysis of biological samples (often called “lab-on-a-chip”) [40].

Critical components in microfluidic device are micropump and microvalves whose

design, functionality and suitability have to be carefully considered for specific ap- plications [41].

Nowadays, microvalves can be roughly classified in two groups: active microvalves using mechanical and non-mechanical moving parts, and passive microvalves using mechanical and non-mechanical moving parts. Micropumps can be also roughly categorized into two groups: mechanical micropumps with moving parts and non- mechanical micropumps without moving parts. Two movement mechanisms have been employed in mechanical micropumps: reciprocating and peristaltic movements.

According to various actuation principles, the mechanical micropumps can be di- vided into several categories: piezoelectric [42], pneumatic [43, 44], thermopneu- matic [45], electrostatic [46, 47], electromagnetic [48, 49, 50], etc.

An other important field of application of the microactuators is the manipulation of micro-objects. Nowadays the design and the fabrication of microrobots is a very difficult task, which not only requires substantial progresses in many areas of mi- crofabrication techniques, but also the ability to integrate these factors into robotic microsystems. Actuators are recognized as critical components of microrobots. The need for better actuators is widely recognized in many fields of application. In robotics, the lack of actuators with high power to mass and torque to speed ratios is at the origin of many design limitations. In the field of microrobots, suitable microactuators are required even more urgently, since the ordinary electromagnetic motors exhibit serious limitations when miniaturized.

1.5 MEMS fabrication techniques

To fabricate all possible MEMS no single process flow can be used. Most MEMS

fabrication techniques have their roots in the standard manufacturing methods de-

veloped for semiconductor industry. In addition, several MEMS-specific techniques

developed for fabricating micromechanical structures have been added to the mi-

cromachininst’s toolbox. Here we will present the key microfabrication techniques

that cover the basic MEMS fabrication concepts and form a basis for many other

derivatives. Subsequently, we will present three major “hard” MEMS technologies

(surface, bulk, and high aspect-ratio micromachining).

As previously anticipated, basic fabrication methods currently used in MEMS and micromachining are the same as those employed by microchip manufacturers over the last four decades. These techniques are:

1. thin-film deposition;

2. lithography;

3. etching;

4. substrate bonding;

Thin films are deposited using various chemical or physical techniques and are used for masking, isolation, and structural purposes. Subsequently, the lithography step is performed in order to transfer the designed pattern onto the substrate. The pat- terned substrate is etched using various chemicals in liquid and gas phase. Finally, substrate bonding is used either to integrate multiple functionalities or for packaging purposes. These steps can be repeated numerous times depending on the complexity of the design and process.

1.5.1 Thin-film deposition

To deposit thin films of different materials on a substrate a wide variety of tech- niques is used . The selection of a suitable material should be accompanied by determination of the appropriate deposition technique, since the properties of the deposited film, as internal stress (tensile or compressive), adhesion to the substrate material, and conformality to the substrate topography, will be strongly process- dependent. The four main thin-film deposition techniques are oxidation, electrode- position, chemical vapor deposition (CVD), and physical vapor deposition (PVD).

Oxidation is typically performed on semiconductor substrates by heating them up to temperatures ranging from 800 to 1200 ^◦ C in an atmosphere containing O 2

or H ₂ O vapor. The result is high quality thin film of oxide (SiO ₂ if the substrate is

silicon).

Electrodeposition (or electroplating) is typically used to obtain thick (tens of micrometers) metal structures. This is an electrochemical process, the substrate is immersed in a solution containing a reducible form of the ion of the desired metal and it is maintained at a negative potential (anode). The ions are reduced at the sample surface and the non-soluble metal atoms are deposited [51].

Chemical Vapor Deposition uses the reaction of chemicals in a gas phase to form the deposited thin film [52]. The most common CVD techniques used in microfabrication are Low Pressure CVD (LPCVD) and Plasma Enhanced CVD (PECVD). These methods are typically used to deposit inorganic materials such as silicon dioxide, silicon nitride, and polycrystalline silicon (polysilicon). LPCVD is performed at pressure ranging from 0.1 to 1 Torr and temperatures in the range of 550-900 ^◦ C. The deposited material can be highly stoichiometric and the films obtained are usually very conformal to the surface of the substrate. PECVD is performed at lower temperature, which is often the reason for its choice. However, the obtained films are often less conformal and of poor chemical quality.

Physical vapor deposition systems are based on two different principles: evap- oration and sputtering. In evaporation systems, a piece of material is heated in a vacuum chamber containing the substrate. The evaporated material is spread all over the chamber and a thin film is deposited on top of the substrate. In sputter- ing systems, a target of material is bombarded with high-energy inert ions (usually argon). The individual atoms or clusters that are removed from the surface of the target material are ejected toward the substrate. Metals are easily deposited using these techniques, although many other materials can also be evaporated or sput- tered [53]. Due the nature of the sputter deposition, it is semi-conformal and the metal will laterally deposit on the side walls, but will not deposit underneath the beams.

1.5.2 Lithography

Lithography is the technique used to transfer a specific pattern onto a substrate.

This pattern is subsequently used to etch an underlying thin-film for various pur-

poses. The starting point, subsequent to the creation of the computer layout for

Figure 1.7: Schematic drawing of the photolithographic steps with a positive photoresist.

a specific fabrication sequence, is the generation of a photomask. This involves a

sequence of photographic processes which results in a glass plate having the de-

sired pattern in the form of a thin (obout 100 nm) chromium layer. Lithographic

techniques generally require the use of flat substrates. Silicon is often used even

where there are no electronic components in the device because the tools and in-

struments needed for microfabrication are designed to match the characteristics of

silicon wafers. The lithography proceed as shown in Figure 1.7. After depositing or

growing the desired material on the substrate, the photolithography process starts

with spin-coating of the substrate with a photosensitive resist material. The sample

is prepared for the photoresist with a primer which promotes the adhesion of the

photoresist to the surface. The thickness of the resist material depends on viscosity

and spin speed. Following spinning, the substrate is baked in order to remove sol-

vents from the resist and improve adhesion, time and bake temperature depends on

Figure 1.8: Profile for isotropic (a) and anisotropic (b) etch through a photoresist mask.

the resist type. Subsequently, the mask is aligned to the wafer and the photoresist is exposed to a UV source.

After exposure the photoresist is developed by washing off the UV-exposed or the unexposed regions, depending on whether the resist material is “positive” or

“negative”, respectively. The resist is subsequently baked in order to further improve the adhesion. The post-baking concludes the photolithography sequence that creates the desired pattern on the wafer. Next, the underlying thin film is etched away and photoresist is stripped in acetone or other organic removal solvents.

1.5.3 Etching

As mentioned above, a lithography step is usually followed by an etching step in order to obtain a patterned film or selective material removal from the substrate. Two important figures of merit for any etching process are selectivity and directionality.

Selectivity is the degree to which the etchant can differentiate between the masking material and the material to be etched. Directionality has to do with the etch profile under the mask. In an isotropic etch, the etchant attacks the material in all directions at the same rate, hence, creating a semi-circular profile under the mask as shown in Figure 1.8(a). In an anisotropic etch, the dissolution rate depends on specific directions, and one can obtain straight sidewalls, as shown in Figure 1.8(b), or other non-circular profiles. The etching techniques can be also divided into “wet”

and “dry” categories. Wet etchants are largely isotropic in reactivity and show

Figure 1.9: Glass to silicon anodic bonding setup

superior material selectivity as compared to various dry techniques. Due to the lateral undercut, the minimum feature achievable with wet isotropic etchants is limited. An exception to this rule is provided by the anisotropic wet etching of monocrystalline subtrates. In the 1960s, silicon etch was found to be dependent and concentration-dependent in some chemical solution [54, 55]. The development of anisotropic etching of crystalline silicon is considered to have marked the beginning of micromachining and the MEMS discipline.

1.5.4 Substrate bonding

Wafer bonding is one of the most important fabrication techniques in microsystems technology [56]. It is frequently used to fabricate complex 3-D structures, both as a functional unit and as a par of the final microsystem packaging and encapsulation.

The two most important bonding techniques are silicon-silicon fusion and silicon- glass electrostatic (or anodic) bonding. In addition to these techniques, several other alternative methods which utilize an intermediate layer, have also been investigated.

Direct silicon or fusion bonding is used in the fabrication of micromechanical

devices and silicon-on-insulator (SOI) substrates. The technique is used primarily

to bond two silicon wafers with or without an oxide layer. The bonding procedure is as follows: the silicon or oxide-coated silicon wafers are first thoroughly cleaned.

Subsequently, the surfaces are rendered hydrophilic by hydroxylation in HF or boil- ing nitric acid. The substrates are then brought together in close proximity, starting from the center to prevent void formation. As a result, attractive van der Waals forces bring the surfaces into intimate contact on an atomic scale, and hydrogen bonds form between the two hydroxylated surfaces, joining the substrates together.

These steps can be performed at room temperature; however, in order to increase the bond strength, a high temperature anneal is usually required (800-1200 ^◦ C). A major advantage of silicon fusion bonding is the thermal matching of the substrates.

Silicon-glass anodic bonding is another substrate joining technique that has been extensively used for microsensors packaging and device fabrication. The main ad- vantage of this technique is its lower bonding temperature. Figure 1.9 shows a schematic diagram of the bonding setup. A glass wafer (usually Pyrex 7740 due to its thermal expansion match to silicon) is placed on top of a silicon wafer and the sandwich is heated up to 300-400 ^◦ C. Subsequently, a voltage of ∼1000 V is applied to the cathode. Bonding starts immediately upon application of the voltage and spreads outwards from the cathode contact points. During the heating period, glass sodium ions move toward the cathode and create a depletion layer at the silicon- glass interface. A strong electrostatic force is therefore created at the interface, which pulls the substrates into intimate contact, apparently creating conditions for covalent silicon-oxygen bonds to form at the interface.

The techniques outlined so far are common to the semiconductor industry and to MEMS technology. We now turn to MEMS-specific methodologies.

1.6 Surface Micromachining

The most commonly used surface micromachining process is sacrificial-layer etch-

ing [58]. This technique relies on the deposition of structural thin films on a sacri-

ficial layer wich is subsequently etched away resulting in movable micro-mechanical

structures (beams, membranes, plate, etc.). As an example, Figure 1.10 shows a RF-

Figure 1.10: SEM image of a surface-micromachined RF MEMS switch realized by poliymide sacrificial etching [57].

MEMS switch consisting of electroplated nickel and serpentine folded suspensions, obtained using PI2545 Polyimide as sacrificial layer [57]. The main advantage of sur- face micromachining is that extremely small sizes can be obtained. In addition, it is relatively easy in principle, to integrate the micromachined structures with on-chip electronics for increased functionality. The primary disadvantage is the fragility of surface microstructures, particulates and condensation during manufacturing.

Polysilicon micromachining is the most common form of surface micromachin-

ing [58, 13]. Figure 1.11 shows a basic process flow. The process starts with the

passivation of the wafer with a layer of 0.15 µm of LPCVD (low-pressure chemi-

cal vapor deposition) silicon nitride deposited on top of the silicon substrate (Fig-

ure 1.11(a)). This layer, often, is silicon-rich nitride, which can have a tensile stress

much less than stochiometric Si 3 N 4 and thus adheres better to the substrate. The

next process step (Figure 1.11(b)) is the deposition of a LPCVD phosphosilicate

glass (PSG), which is patterned to form what will be the structural anchor to the

Figure 1.11: Process flow for single layer polysilicon micromachining.

substrate of the movable structures (Figure 1.11(c)). A subsequent deposition of LPCVD phosphorous-doped polysilicon forms a 2 µm thick structural layer Fig- ure 1.11(d). The polysilicon is dry etched to define the structures (Figure 1.11(e)).

In the last step, the wafer is immersed in aqueous hydrofluoric acid (typically 10

:1 diluted hydrofluoric acid (HF) or buffered HF) to dissolve the sacrificial PSG

layer (Figure 1.11(f)). The required etch time for releasing the structure must be

limited, because the HF has a small but finite etch rate for silicon nitride. Large

microstructures must have spaced-apart holes designed into them so as to allow the

release etch to remove the underlying spacer. The wafer is then rinsed in deionized

water and dried. An important consideration in the design and processing of sur-

face micromachined structures is the issue of stiction, which can occur either during

fabrication when wet etchant is used to remove the sacrificial layer, or during the

device lifetime. The source of stiction during fabrication is surface tension of the