Ghodssi R., Lin P., MEMS Materials and Processes Handbook

Подождите немного. Документ загружается.

934 P. Lin e t al.

be completely hydrophilic. Less strongly hydrophilic solids exhibit water contact

angles up to 90

◦

. A hydrophobic surface is one on which the water contact angle

is larger than 90

◦

. On very hydrophobic surfaces, water contact angles as high as

150

◦

can be observed [41], reaching almost 180

◦

in extreme cases. These surfaces

are termed “superhydrophobic,” and can be obtained in situations where surfaces

have been appropriately micropatterned to achieve multiscale roughness. The con-

tact angle provides information about the interaction energy between surface and

the liquid. Unlike XPS, this technique probes large (mm-sized) surface areas, typi-

cally larger than MEMS devices; therefore, it needs to be implemented on suitably

prepared surfaces that are representative of the real devices.

13.3 Adhesion and Friction of MEMS

13.3.1 Measurements of Adhesion and Friction

A number of microinstruments have been speciﬁcally developed to examine the

surface-surface interaction involved in silicon microstructures [42–53]. These are

described below.

13.3.1.1 Cantilever Beam Array Technique

To quantify the adhesion characteristics of microstructures, Mastrangelo and Hsu

developed a simple microinstrument to measure the work of adhesion between

two contacting surfaces [42]. The instrument involves an array of single-clamped

beams with varied lengths parallel to the surface. This is commonly referred to as

a cantilever beam array (CBA) test structure, a SEM picture of which is shown in

Fig. 13.3b. A schematic side-view of a single cantilever beam is shown in Fig. 13.3a.

To measure the work of adhesion, the beams are brought into contact with the

underlying substrate. Once the applied actuation force is removed, the beams begin

to peel themselves off the surface. In the original analysis [42], beams shorter than

a characteristic length (so-called detachment length, 

) were stiff enough to free

themselves completely from the underlying surface, while beams longer than this

characteristic length remained adhered to the surface. Based on this deﬁnition, the

beams in the transition region were adhered to the substrate only at their tips. By

balancing the elastic energy stored within the beam and the beam-substrate interfa-

cial energy, the work of adhesion, W, between the two surfaces can be calculated

using the following equation:

W =



(13.1)

where E is Young’s modulus, h is the spacing between the beam and the substrate,

and t is the thickness of the polysilicon beams. Since then, a number of different

schemes have been developed to relate the measurements to the work of adhesion

13 Surface Treatment and Planarization 935

Si Substrate

anchor actuation pad beam landing pad

Ground-plane polysilicon

Increasing

beam length

(a)

(b)

Fig. 13.3 Cantilever beam array: (a) schematic side-view of a single cantilever beam, and

[43–45]. If the two contacting surfaces are perfectly ﬂat and indistinguishable, then

the measured work of adhesion is twice the surface energy of the individual surfaces

[40, 43]. However, the polysilicon surfaces used in micromachining are most often

rough, and thus the CBA method provides information about the apparent work

of adhesion since the actual area of contact is different from the projected area of

contact.

13.3.1.2 Double-Clamped Beam Array Technique

Another test structure used to probe adhesion in MEMS consists of an array of

double-clamped cantilever beams with varying lengths [46]. By bringing the beams

in contact with the underlying substrate and removing the external actuation, a

transition is observed between adhered and freestanding beams. Since the stored

elastic energy in the double-clamped beam contains a stretching term in addition to

the bending term, the equation relating detachment length to work of adhesion is

modiﬁed accordingly:

W =



εEh



(13.2)

where ε is the residual strain in the beams. By measuring the detachment length,

the work of adhesion can be found, assuming the other parameters are known. In

contrast to the cantilever beam array microinstrument, this test structure requires an

independent measure of the residual strain in the beams. As described in [49], this

936 P. Lin e t al.

microinstrument can also be used to determine the adhesion force, which from the

design point of view is sometimes more desirable than adhesion energy.

13.3.1.3 Friction Test Structures

A number of microinstruments have been developed to study friction in MEMS

[50–53]. As an example, here we describe a device that allows the measurement

of the static coefﬁcient of friction between in-plane surfaces. The instrument con-

sists of a shuttle that is suspended above an underlying electrode by a folded beam

suspension [50]. The application of a voltage, V

, to the comb drives causes the

shuttle to be displaced by a speciﬁed amount. Applying a voltage between the shuttle

and the underlying substrate, V

, pulls the shuttle downwards until the dimples con-

tact the landing pads. By reducing the V

bias, a threshold voltage, V

, is reached,

at which point the shuttle snaps back to equilibrium. The normal applied force is

calculated from the threshold voltage. The tangential force is calculated using the

initial displacement of the shuttle and its spring constant. The friction coefﬁcient

can then be determined from the normal and tangential forces.

13.3.2 Effects of Surface Roughness

One method used to reduce adhesion is to reduce contact area, as done by Fan

et al. by fabricating a micromechanical bushing, or dimple [54]. Another approach

is to intentionally roughen (texture) one or both of the contacting surfaces. In this

way, the average separation at contact is reduced and hence adhesion is reduced.

Examples of this approach include texture etch-back techniques [55] and pref-

erential oxidation of silicon grains at their boundaries [56] to obtain very rough

polysilicon surfaces. Yet another scheme is to grow silicon carbide (SiC) nanoparti-

cles (in the tens of nm range) using SiO

ﬁlms deposited from tetraethylorthosilicate

[57, 58].

13.4 Chemical Modiﬁcation of MEMS Surfaces

13.4.1 Treatments for Low Surface Energy

The release solutions described in Section 13.2 do not help alleviate the problem of

in-use adhesion. This problem can be addressed by adapting one or a combination

of the following strategies:

(i) conservative designs to prevent surfaces from coming into contact whenever

possible;

(ii) physical modiﬁcation of surfaces, as discussed in Section 13.3.2;

(iii) chemical modiﬁcation of surfaces to reduce surface energy.

13 Surface Treatment and Planarization 937

The third strategy, referred to as “boundary lubrication,” is the most versatile and

probably the most actively pursued approach, and will be discussed next.

13.4.2 Siloxane and Silane Treatments

Self assembled monolayers are thin organic ﬁlms that spontaneously adsorb or bond

on solid surfaces [59]. The precursor molecule has three parts. One is a surface-

active head group; because of this group, the molecule has the propensity to adsorb

on surfaces through covalent or electrostatic interactions. Second is an alkyl chain,

which helps in the formation of a densely packed array, due to the resulting van der

Waals interaction between the adjacent chains. Third is the tail group that can be var-

ied to impart a variety of functions, such as hydrophobicity or hydrophilicity. For

modiﬁcation of the silicon surface, several monolayer systems have been developed,

including alkyl- and perﬂuoroalkyl-trichlorosilane SAMs (R-SiCl

), dichlorosilane

monolayer ﬁlms (R

-SiCl

), and alkene-based molecular ﬁlms. Schemes for apply-

ing self-assembled monolayers to microstructures have been developed. These

monolayers have been shown to eliminate release adhesion, and to signiﬁcantly

reduce in-use adhesion. This approach has been recently reviewed [60, 61]. Speciﬁc

examples of these ﬁlms and their resulting properties are summarized in Table 13.1.

In addition, a number of two-step monolayer processes have been investigated for

MEMS applications [62, 63].

Early schemes for SAM coating of microstructures involved deposition from the

liquid phase. However, due to challenges associated with scale-up, vapor-based

coating processes have been pursued more recently, although these approaches

require a successful release as the starting point. Indeed, SAM deposition from

vapor phase has been demonstrated following a separate release process using super-

critical CO

drying [64–66]. Table 13.2 lists the properties achieved via the vapor

phase deposition process. Tools that implement vapor phase antistiction coatings are

commercially available [67, 68]. One such tool is referred to as MVD, Molecular

Vapor Deposition [67]. It utilizes plasma surface cleaning and proprietary seed layer

Table 13.1 Physical property data for several surface treatments [taken from Ref. 53]

Surface

treatment

Water contact

angle (

◦

)

Work of

adhesion

(mJ/m

)

Static

friction

coefﬁcient

Thermal

stability in air

(

◦

Particulate

formation

OTS 110 0.012 0.07 225 High

FDTS 115 0.005 0.10 400 Very high

DDMS 103 0.045 0.28 400 Low

1-octadecene 104 0.009 0.05 200 Negligible

Oxide 0–30 20 1.1 – –

1H,1H,2H,2H-perﬂuorodecyltrichlorosilane (FDTS); octadecyltricholorosilane (OTS); dimethy-

dichlorosilane (DDMS); 1-octadecene; untreated oxidized silicon

938 P. Lin e t al.

Table 13.2 Selected properties of the DDMS and FDTS monolayers deposited from vapor phase

versus liquid phase

Surface

treatment

Water contact

angle (

◦

)

Work of

adhesion

(mJ/m

)

Static

friction

coefﬁcient

Thermal

stability in air

(

◦

Particulate

formation

V-FDTS 111 0.003 0.12 400 Low/none

L-FDTS 115 0.005 0.10 400 Ve ry high

V-DDMS 102 0.062 0.35 400

L-DDMS 103 0.045 0.28 400 Low

Oxide 0–30 20 1.1 – –

Oxide values are given for comparison. V- (L-) denotes deposition from vapor (liquid) phase

deposition to increase the number of active sites on the surface. This is then fol-

lowed by the vapor deposition of the precursor of interest, based on the i ntended

surface property. In addition to FDTS and DDMS listed in Table 13.2, the tool

can deposit coatings based on a wealth of monolayers. Some examples include

1H,1H,2H,2H-perﬂuorooctyltrichlorosilane (FOTS), which has properties that are

rather similar to those of FDTS. Another monolayer commonly used for surface

functionalization is 3-aminopropyltrimethoxysilane (APTMS), which is an exam-

ple of a hydrophilic surface treatment. Lastly, 3-mercaptopropyltrimethoxysilane

(MPTMS) is a mercapto-terminated silane that is useful in thiol based chemistries

or as an adhesion layer for deposition of smooth gold.

Another vapor coating approach with enhanced thermal stability was reported

by Analog Devices, Inc. [69, 70]. It is based on phenyl-methyl silicone or diphenyl

siloxane that has good resistance to temperature and oxidation. A small amount of

liquid is placed onto each package before the die is sealed. At elevated temperatures,

the liquid evaporates, and the molecules contacting the surface chemically react

and bond to it. This process results in a monolayer coating with a high organic

content.

13.4.3 Weakly Chemisorbed Surfactant Films

Up to this point, we have emphasized the use of silicon as the structural mate-

rial, due to its continued dominance. However, many metals and alloys have long

been used in applications requiring high reﬂectivity or low contact resistance.

Moreover, metal deposition methods such as electroforming are gaining favor as

inexpensive and versatile fabrication methods, and recent research [71] has validated

their use in applications in which semiconductor or ceramic materials have been

traditionally believed to yield superior performance (such as high Q resonators).

Functionalization schemes are readily available for many of these metals, and they

can be used to render the surfaces hydrophobic in order to avoid microstructure

13 Surface Treatment and Planarization 939

adhesion caused by surface forces [72]. Unlike silane coupling, these schemes usu-

ally rely on weaker chemisorbed headgroups, such as thiol (suitable for binding to

most noble metals, including copper and its oxide) and carboxyl (suitable for bind-

ing to aluminum and its oxides, and notably employed in the Texas Instruments

digital micromirror device, also known as DMD). Weakly chemisorbed species

such as alkanethiols or alkanoic acids possess many advantages with respect to the

strongly chemisorbed siloxanes, including the inability to crosslink (which leads to

clean, particle-free surfaces after treatment) and the ability to self-heal in response

to mild wear. For the latter property to take effect, the presence of a reservoir of pre-

cursor molecules is required. This can be accomplished by hermetically packaging

the device under the pressure of the precursor. In the case of the TI DMD device,

a standard 5-year warranty on a DLP

(Digital Light Processing) chip implies

hundreds of billions of stiction-free impacts, a feat made possible by vapor phase

lubrication of the kind just discussed. Additional thiol and silane related surface

treatments are discussed in Section 13.5.

13.4.4 Materials Properties and Process Selection Guidance

Despite the large number of coatings now available, it is difﬁcult to identify one

treatment as the best surface coating. Tables 13.1 and 13.2 provide a variety of

properties for some anti-stiction coatings. As evident from the table, one can choose

from a wide range of molecular coatings yielding different degrees of thermal sta-

bility, particulate susceptibility and tribological performance. Other considerations

that affect the choice of coating are selectivity, lubricity and the very presence

of an oxide layer. Because of the wide ranges of ﬁlm properties and deposition

methodologies, and due to the greatly varying end-use applications and processing

requirements, no one speciﬁc coating technology can be claimed to be “universal”.

Ultimately, it is the performance of the micromechanical device that will dictate

which anti-stiction coating should be utilized, if any. Because it is performance that

drives design, this underscores the need to plan for the coating at the microstructure

design level.

13.5 Surface Considerations for Biological Applications

BioMEMS is a specialized branch of MEMS dedicated to the design, fabrication

and application of MEMS devices that contain or handle broadly-deﬁned biological

materials, ranging from natural or synthetic biomolecules, biochemical complexes,

and cellular entities to tissue masses. Unlike most conventional MEMS devices

that primarily operate in liquid-free environments, BioMEMS devices often operate

under conditions that require parts of the devices’ solid surfaces to contact liquid

media and interact with biological materials. Therefore, speciﬁc challenges related

940 P. Lin e t al.

to the interactions between BioMEMS device surfaces and biological materials at

solid-liquid interfaces are of special interest in BioMEMS.

Although BioMEMS is recognized as a specialized ﬁeld, and can be distin-

guished from conventional MEMS, BioMEMS still utilizes many fundamental con-

cepts, raw materials and processing techniques employed in conventional MEMS.

Indeed, in the past several decades, the principles of conventional MEMS have con-

tributed to the rapid development of BioMEMS. However, it should be noted that

to ensure proper functioning of BioMEMS devices, some restrictions must be con-

sidered when applying principles of conventional MEMS to BioMEMS. First of all,

the raw structural materials chosen must be inert and stable when exposed to the

liquid media (mainly buffered aqueous media). Additionally, the selected materi-

als must be compatible with biological materials, and non-toxic to living biological

entities. The raw materials commonly used to fabricate BioMEMS devices include

noble metals (Au, Ag, Pt, etc.), polymers (PMMA, SU-8, PDMS, etc.), silicon, silica

(glass, quartz, etc.), and silicon nitride. As these materials are also commonly used

in conventional MEMS, it is advantageous that most of the processing techniques

well-established in conventional MEMS can be readily applied for the fabrica-

tion of BioMEMS devices. Essentially, the diversiﬁed and sophisticated physical

architectures for BioMEMS devices can be fabricated by the processing techniques

developed in conventional MEMS. However, for BioMEMS applications, beyond

elaborate architectural designs, the performance and functions of BioMEMS devices

are also highly dependent on the interactions of biological materials with device sur-

faces. Therefore, the surface modiﬁcation of the pristine surfaces of conventional

MEMS materials is a major research topic in the BioMEMS area.

The two main purposes of surface modiﬁcation in BioMEMS are (i) to reduce

undesired non-speciﬁc adsorption of biological materials which can cause device

degradation or interfere with device transduction and (ii) to generate desired sur-

face properties which can promote device performance or expand device functions.

To reduce non-speciﬁc adsorption, surface coatings based on bovine serum albumin

(BSA) or poly(ethylene glycol) (PEG) are commonly used. To prepare biologically

active surfaces, immobilization of biological molecules, such as oligonucleotides,

peptides, carbohydrates or proteins (including enzymes and antibodies), is required.

Modiﬁcation of MEMS surfaces for BioMEMS applications is mainly based on

bottom-up molecular engineering. The surface modiﬁcation procedure can be

a single-step process. However, for the construction of robust and/or complex

structures, multi-step processes are commonly applied. Surface modiﬁcation by

molecular engineering can cause changes in surface wettability and/or surface

functionality. Altering surface wettability can generate new surfaces with desired

adsorption behaviors, while altering surface functionality can create new surface

functional groups for desired chemical reactions.

Surface modiﬁcation by molecular engineering utilizes the interactions between

molecules. The interactions can be covalent or non-covalent. Non-covalent inter-

actions include van der Waals interactions, electrostatic interactions, hydrogen-

bonding and certain biological interactions (such as antigen/antibody conjugation,

DNA hybridization, etc.). Covalent interactions involve the formation of covalent

13 Surface Treatment and Planarization 941

bonds. Covalent interactions frequently used for surface modiﬁcation in BioMEMS

are primarily based on certain chemical reactions involving the following ﬁve major

functional groups: −OH, −NH

−COOH, −SH, and −CHO. These functional

groups are commonly found in natural biomolecules. Moreover, those functional

groups can be derivatized to chemical compounds of non-biological origin, and thus

can also be found in synthetic biomolecules, cross-linkers, and on modiﬁed substrate

surfaces. The cross-linkers are deﬁned as the chemical compounds used to generate

the linkage between the pristine surface (i.e. the original MEMS device surface) and

the outermost materials. The cross-linkers can be heterobifunctional (two different

functional end groups) and homobifunctional (same functional groups at the both

ends).

13.5.1 Surface Modiﬁcation Techniques

A generalized surface modiﬁcation procedure is illustrated in Fig. 13.4. The pris-

tine substrate is the original, clean substrate prior to the surface treatment. After

the surface treatment, the chemical compounds containing desired terminal groups

are placed at the outermost layer. In general, the outermost chemical species on

the device surfaces provide the functionalities of BioMEMS devices. The outermost

chemical species can be biological materials, biocompatible materials or mixtures

of both. However, the performance of BioMEMS devices can also be affected by

other factors. For example, the layers underneath the outermost layer could affect

the robustness, durability and reliability of the outermost layer. In addition, the

untreated pristine surface areas should be properly blocked or passivated to prevent

non-speciﬁc binding (see Case studies 1 and 2).

Chemical compounds that can react with the surfaces of chosen substrate mate-

rials, such as organosilanes (for silica-based surfaces), are often used for the ﬁrst

Pristine

1 step

2 steps

3 processing steps

4 processing steps

Pristine

Substrate

Fig. 13.4 Schematic illustration showing a single-step or multi-step surface modiﬁcation proce-

dure. The letters correspond to interactions or reactions used to create linkage between the substrate

surface and a chemical species or between chemical species represented by blocks

942 P. Lin e t al.

reaction to get a ﬁrm anchoring on the pristine surface. As shown in Fig. 13.4, there

are several different ways to create the linkage in a given modiﬁcation procedure. In

a one-step procedure, the interaction occurs between the pristine substrate surface

and the outermost material (Reaction a). The two-step procedure involves the pris-

tine substrate surface reacting with a cross-linker (Reaction b), and then with the

outermost material (Reaction c). Multiple-step procedures (n ≥ 2) include multiple

reactions with cross-linkers (Reactions b, d, and/or e), and then with the outermost

material (Reaction c). The modiﬁcation techniques can be categorized as (i) tech-

niques for modifying pristine substrate surfaces (i.e. general MEMS device surfaces

created with typical clean room fabrication tools), such as Reactions a and b, and

(ii) techniques for modifying pre-treated substrate surfaces (i.e. treated pristine sur-

faces that require additional steps to complete the entire process), such as Reactions

c, d and e. Following the categorization, the techniques commonly used for surface

modiﬁcation in BioMEMS are summarized and discussed in the following sections.

Many examples are accompanied by the reaction principles, to provide the materials

and processing details.

13.5.2 Modiﬁcation of Pristine Substrate Surfaces

13.5.2.1 Plasma Treatment

Plasma treatment can be applied to a variety of substrate materials and surfaces

that can withstand plasma bombardment. Plasma treatment can be used to change

the surface wettability or to generate reactive surface-bound functional groups for

chemical reactions. For example, NH

(ammonia) plasma can be used to gener-

ate surface-bound amino groups, while H

OorO

plasma can be used to generate

surface-bound peroxide, hydroxyl, carbonyl, and/or carboxyl groups.

EXAMPLE 1:

As shown in Fig. 13.5,NH

plasma was applied by Salber et al. to introduce

amino groups on polydimethylsiloxane (PDMS) surfaces [73]. After the treatment,

the contact angle with water was slightly decreased (from 107

◦

to 93

◦

). X-ray pho-

toelectron spectroscopy (XPS) on the treated surface showed about 2 atom % of N,

indicating that amino group formation was concurrent with the NH

plasma treat-

ment on the surface. NH

functionalization enables the subsequent immobilization

of isocyanate-terminated star-shaped polyethylene glycol (PEG). This PEG coating

system can prevent non-speciﬁc cell adhesion.

PDMS PDMS

Plasma

Fig. 13.5 Schematic illustration showing surface modiﬁcation on a PDMS surface by NH

plasma

treatment

13 Surface Treatment and Planarization 943

EXAMPLE 2:

A PMMA-based capillary electrophoresis microchip was fabricated using the Si

wafer imprint techniques [74]. Prior to the plasma treatment, the PMMA substrate

was annealed at 100

◦

C for 1 h and rinsed with methanol and water, and then dried

with nitrogen. Plasma was generated at 13.56 MHz/80 W and at a pressure of 100

mTorr for a 3 min treatment. The ionized oxygen or oxygen radicals reacted with the

PMMA chains, creating oxygen-containing groups such as hydroxyl and carboxyl

groups on the microchannel surface (Fig. 13.6). These functional groups were used

for subsequent surface polymerization of PEG polymer to improve the efﬁciency of

protein and peptide separation [74].

COOH

PMMA

Plasma

PMMA

COOH

Fig. 13.6 Schematic illustration showing surface modiﬁcation on a PMMA surface by O

plasma

treatment

13.5.2.2 Physical Adsorption

Physical adsorption is a method of surface modiﬁcation via non-covalent interac-

tions (hydrogen-bonding, van der Waals interaction, electrostatic interaction, etc.)

between the substrate surface and the applied material. Physical adsorption can be

applied to various substrate materials or surfaces, as long as the applied material is

compatible with the substrate material or underlying surface. Many proteins, includ-

ing antibodies and some enzymes, BSA, etc. can be coated onto substrate surfaces

via the weak physical interactions between protein molecules and substrate surfaces.

EXAMPLE 3:

A PDMS microﬂuidic system with trypsin membrane reactor for protein diges-

tion was fabricated by Gao et al. [75]. A porous poly(vinylidene ﬂuoride) (PVDF)

membrane (0.45 μm pore size, obtained from Millipore) was used to form an

enzyme reactor via non-covalent adsorption of trypsin molecules, a serine protease

that hydrolyzes proteins. The high surface-to-volume ratio of the PVDF membrane

allows high trypsin concentration in the membrane, resulting in rapid protein diges-

tion. The adsorption was carried out by pumping a trypsin solution (2 mg/ml, in pH

8.0 Tris buffer) through the membrane at a ﬂow rate of 0.2 μL/minfor2h.The

membrane was then ﬂushed with a Tris buffer for 20 min to remove any unbound

trypsin. This miniaturized membrane reactor combined with microﬂuidic channels

on a PDMS substrate enables the rapid digestion of trace amounts of protein samples

(1 ng or less) in minutes for subsequent mass spectrometry analysis [75].

13.5.2.3 Covalent Linkage

Surface modiﬁcation of pristine MEMS substrates via covalent linkage has been

applied for many BioMEMS applications. Depending on the substrate materials,