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

Подождите немного. Документ загружается.

This is Blank Page Integra xxxii

Chapter 13

Surface Treatment and Planarization

Pinyen Lin, Roya Maboudian, Carlo Carraro, Fan-Gang Tseng, Pen-Cheng

Wang, and Yongqing Lan

Abstract Released structures, structures that move, free surfaces, and surfaces that

contact one another in operation are all common in MEMS devices. Free surfaces

on MEMS elements are crucial to device performance; hence, surface characteris-

tics must be well controlled during release processes. In this chapter, materials and

processes pertinent to device surfaces are discussed in seven main sections. Case

studies and examples illustrate the processing details. The ﬁrst section centers on

release processes that create the free device surface, including surface treatments to

prevent stiction. Tools for surface analysis are discussed in the second section, and

the third section focuses on (undesired) adhesion of MEMS structures, i.e. stiction

(Sections “Release Processes and Surface Treatments to Prevent Stiction”, “Surface

Analysis”, and “Adhesion and Friction of MEMS”, in this chapter).

As MEMS diversify into chemical, analytical, and biomedical applications,

MEMS devices interface with chemical and biological environments, making the

free surface between MEMS elements and the environment crucial to performance.

The chemistry of surface treatments and the variety of substrate materials constitute

a distinct ﬁeld for surface design and processing of MEMS devices. The func-

tionalized surface must respond to changes in the environment and translate those

changes into measurable signals. Accordingly, the fourth and ﬁfth sections describe

chemical and biological treatments to create appropriately functionalized surfaces

(Sections “Chemical Modiﬁcation of MEMS Surfaces” and “Surface Considerations

for Biological Applications”, in this chapter).

As light interacts with MEMS devices, three optical parameters are relevant to

performance: reﬂection, transmission, and absorption. The design of optical coat-

ings to meet the optical requirements of MEMS applications is discussed in Section

“Surface Coating for Optical Applications”, in this chapter. The materials and

processes for optical coating and surface requirements are also discussed.

The ﬁnal section, Section “Chemical Mechanical Planarization”, in this chap-

ter, concerns chemical mechanical planarization (CMP) for MEMS applications.

MEMS structures with high aspect ratios are often created to achieve speciﬁc

P. Lin (B)

Touch Micro-system Technology Corp., Taoyuan; Walsin Lihwa Corporation, Taipei, Taiwan

e-mail: pinyen_lin@alum.mit.edu

925

R. Ghodssi, P. Lin (eds.), MEMS Materials and Processes Handbook,

MEMS Reference Shelf, DOI 10.1007/978-0-387-47318-5_13,



Springer Science+Business Media, LLC 2011

926 P. Lin e t al.

performance objectives, creating raised topography. CMP is required to level the

surface for subsequent microlithography. Details of the CMP process for various

materials and design criteria are also described.

13.1 Release Processes and Surface Treatments to Prevent

Stiction

As described in earlier chapters, a number of fabrication techniques are employed

in microsystems technology. Among them are bulk micromachining, surface micro-

machining, and LIGA [1–3]. Although these methods employ lithographic, etching

and deposition techniques that are substantially similar to those employed in IC

device fabrication, the mechanical nature of MEMS devices introduces new pro-

duction and reliability issues. Foremost among these is s tiction-related failure. The

inherently large surface area-to-volume ratio of surface microstructures causes these

devices to be vulnerable to adhesion, either to the underlying substrate or to adjacent

microstructures [4–13]. The term stiction, borrowed from the disk-drive industry, is

used to describe this phenomenon, because the restoring forces in micromechanical

devices generally have both tangential and vertical components.

Stiction can occur both in the production phase and during device operation.

Consider, for example, the process of surface micromachining, which is probably

the most ﬂexible and widely used fabrication method. The basic steps in the pro-

cess are illustrated in Fig. 13.1. The ﬁnal step of sacriﬁcial layer removal is called

the “release step.” If the release process occurs from solution, the surface tension

Crystalline Si

Isolation layer

(silicon nitride)

(a)

Sacrificial layer

(silicon oxide)

Crystalline Si

(c)

Crystalline Si

Microstructural layer

(polysilicon)

(d)

(b)

Fig. 13.1 Schematics of basic steps employed in a surface micromachining fabrication process.

First, the substrate is typically coated with an isolation layer (a) that protects it during s ubsequent

etching steps. A sacriﬁcial spacer layer is then deposited on the substrate and patterned (b). The

microstructural ﬁlm is then deposited and etched (c). Finally, selective etching of the sacriﬁcial

layer releases the microstructures, creating the freestanding micromechanical structures such as

Elsevier

13 Surface Treatment and Planarization 927

of the draining rinse liquid draws the microstructure into contact with the underly-

ing substrate, leading to “release stiction.” Stiction occurring later during operation

is generally called “in-use stiction” and is due to intentional or accidental contact

between microscopic parts. As implied by the term stiction, MEMS surfaces gener-

ally undergo both normal and sliding contact, so friction and wear are also important

issues [14–17], limiting both the production yield and the useful lifetime of many

microdevices.

Strong adhesion is generally caused by capillary, electrostatic and van der Waals

forces, and in some cases by “chemical” forces such as hydrogen bonding and solid

bridging. Treatments that render the surfaces hydrophobic can be used, eliminating

capillary forces. Such treatments offer varying degrees of thermal and chemical

stability, surface hardness, wear resistance, and electrical conductivity. Electrical

properties and chemical stability are factors to consider when trying to minimize

electrostatic forces (e.g., charge trapping). Lastly, van der Waals forces are results

of the polarizability of a medium and as such cannot be eliminated. Their effect can,

however, be mitigated by appropriate surface texturing. We mention only in passing

that retardation effects transform the van der Waals force into the so-called Casimir

force at large body separation (several tens of nm). Therefore, while the Casimir

force, by deﬁnition, does not act at contact, and thus cannot strictly be considered

a player in stiction failure, it can contribute signiﬁcantly to the contact pressure in

those cases when the microstructure contact occurs at high aspect ratio asperities

(e.g., for very rough surfaces) [18].

Fabrication processes severely constrain surface treatments [19]. This implies

that the ﬁrst and most important factor in stiction prevention is good design. A

good design considers the mechanical response of a device, in order to avoid or

minimize surface contacts. A secondary factor is a good material choice, but this

is often constrained by various other considerations (including cost and available

infrastructure). The approach of last resort is to treat surfaces during, or right after,

the microstructure release. It is the easiest approach, since it has the least impact

on the front-end fabrication steps, but it is also the approach that carries the highest

risk of compromising reliability and production yield. A back-end coating process

must be conformal to ensure uniform deposition in areas that are not in the line of

sight. It must also account for the thermal budget of eventual packaging and bond-

ing processes (which often require thermal stability in excess of 300

◦

C). Finally,

one cannot overemphasize that all too often, stiction failure is the unintended (but

foreseeable) consequence of poor design. The surface treatments discussed in this

chapter are not intended as a substitute for good design practice.

Release processes and their speciﬁc recipes must be developed and optimized

based on the combinations of structural and sacriﬁcial materials involved. Examples

include metal and polymer (e.g., aluminum and hardened photoresist), metal and

metal (e.g., nickel and copper), semiconductor and oxide (e.g., silicon and silica),

or semiconductor and semiconductor (e.g., silicon carbide and silicon) [1–3, 13, 20,

21]. Polycrystalline silicon, deposited by low-pressure chemical vapor deposition

(LPCVD), is used frequently in surface micromachining, with an oxide ﬁlm as the

sacriﬁcial layer and an oxide or nitride ﬁlm as the isolation layer. We will discuss

928 P. Lin e t al.

this combination of materials in detail to illustrate the subtleties involved in the

choice of a release process.

13.1.1 Wet Chemical Release Techniques

Silicon oxide sacriﬁcial ﬁlms are etched by liquid hydroﬂuoric acid (HF). HF also

etches silicon nitride, although relatively more slowly, and usually has a negligible

effect on polysilicon. If long etch times are needed in order to etch hard to reach,

recessed areas, hydrochloric acid is added to the etching solution in a 1:1 ratio with

HF to prevent chemical attack of the silicon nitride insulation layer. This is followed

by one or more rinses in water to displace the etch solution and etch residue, and

drying. The oxidized polysilicon surface that results from the rinse and dry process

is hydrophilic, with a water contact angle in the range from zero to 30

◦

. Capillary

forces are thus generated by the meniscus formed between the suspended structure

and substrate. These forces may be sufﬁciently strong to collapse the microstructure

onto the substrate. The oft-ensuing permanent adhesion is called release stiction.

Release stiction can be further magniﬁed by solid bridging of dissolved sacriﬁcial

residues from the rinse solution [5–7].

Several release approaches, such as freeze-sublimation drying and supercrit-

ical drying, can help to minimize or eliminate release-related stiction [22–31].

These approaches share the common feature of avoiding meniscus formation. The

most commonly adopted approach is supercritical drying using CO

[24], with

commercial tools now available for realizing this process at the wafer level [30].

However, it is preferable to use freeze-sublimation drying [23] because it is a sim-

pler process. An alternative approach to wet release is to make the microstructure

surfaces hydrophobic or nonwetting, e.g., by formation of an organic self-assembled

monolayer (SAM) ﬁlm [29]. These coatings eliminate the capillary attraction that

collapses the microstructures upon drainage of the wetting liquid phase, and have

the additional appeal of providing in-use stiction relief.

13.1.2 Dry Release Techniques

The difﬁculties inherent in the drying of microstructures can be avoided altogether

if a dry release process is available. This is generally the case when the sacriﬁcial

material is a polymer, which can be removed by an oxygen plasma etch. A wet-dry

hybrid release scheme is sometimes also employed, whereby a partial undercutting

of the surface microstructure is performed and the voids are backﬁlled with a poly-

mer layer. The polymer holds the microstructures in place after the wet etch of the

sacriﬁcial layer and can be removed with a dry etch [25, 26].

Increasingly popular is the use of vapor HF to etch away sacriﬁcial SiO

[27, 28].

This process is an attractive alternative to liquid phase HF etch, not only because

it eliminates the need for drying the microstructures, but also because it is more

13 Surface Treatment and Planarization 929

cost effective and generates much less chemical waste. As a consequence, a variety

of commercial tools are now available to carry out this process. Note that vapor

HF is particularly aggressive toward nitride, so with this approach, exposed silicon

nitride must be absolutely avoided. Despite this limitation, vapor HF release [31]

is successfully employed by large-volume manufacturers like Bosch in their inertial

sensors technology.

Other dry release chemistries are enabled by suitable choices of sacriﬁcial mate-

rials. For example, polysilicon can be used as a sacriﬁcial, rather than structural,

layer in the fabrication of ceramic devices, such as silicon carbide MEMS. ( SiC is

used as structural material in high value applications, such as MEMS for harsh envi-

ronment.) In this case, xenon diﬂuoride can be used very effectively as a dry etchant

for the sacriﬁcial polysilicon [32].

In summary, dry release processes are recommended because they are econom-

ical and environmentally friendly. Consideration of the choice of release process

must take place at the design level, given the materials selection limitations imposed

by the relatively small number of dry release chemistries that are commercially

available for large scale, high volume production.

13.2 Surface Analysis

A number of analytical techniques have been utilized to monitor surfaces and thin

ﬁlms at various stages of the microfabrication process (Fig. 13.1), including the

release process. Here we brieﬂy summarize several techniques that are most com-

monly employed, starting with those used for chemical characterization, followed

by those used for physical characterization of surfaces.

13.2.1 Surface Chemical Composition

13.2.1.1 X-Ray Photoelectron Spectroscopy (XPS or ESCA)

X-ray photoelectron spectroscopy (XPS, or sometimes called ESCA for electron

spectroscopy for chemical analysis) is a powerful technique used to determine the

chemical composition and the bonding state of atoms in the near surface region

of solids [33]. In an ESCA experiment, X-rays are directed onto a surface, and

electrons in the core levels and in the conduction or valence bands are excited. If the

electrons have momentum towards the surface and sufﬁcient energy to overcome

the surface potential barrier, they escape into the vacuum (photoelectric process),

where they are monitored by an energy analyzer to determine their kinetic energy.

For a given photon energy and work function of the spectrometer, a measurement of

the photoelectron kinetic energy immediately yields its binding energy. The binding

energy of core-level photoelectrons is speciﬁc to individual elements, and varies

slightly but often measurably according to bonding conﬁguration. ESCA data thus

provides information about chemical composition bonding for the surface being

930 P. Lin e t al.

analyzed. Since X-ray photons can penetrate solids to a depth of microns, the surface

sensitivity of XPS is limited by the mean free path of the photoelectrons exiting the

solid. The mean free paths of the photoelectrons are typically in the range of a f ew

nm, and consequently, the sensitivity of XPS is also of this order. Lower take-off

angles can be chosen to further enhance the surface sensitivity; maximum sensitivity

is achieved for grazing angles.

A sample XPS spectrum for as-is silicon is shown in Fig. 13.2, in which the

intensity recorded by the electron multiplier is plotted as a function of the secondary

electron binding energy [34]. The main peaks visible in this plot indicate the pres-

ence of silicon, carbon, and oxygen. The peak labels 1s, 2 s, and 2p indicate the

core level from which the photoelectron was ejected for that particular element. The

results indicate that the silicon surface is hydrogen terminated (i.e. Si-H bonds on

surface) and no Si-O bonds exist on the surface.

Fig. 13.2 X-ray photoelectron spectrum obtained on Si(100) sample, identifying the presence of

carbon and oxygen in addition to Si. The inset shows a close-up of the Si 2p region of the spectrum.

The small peak at ~103.5 eV signiﬁes Si in a silicon-oxide matrix, highlighting that the sample has

a thin layer of oxide on it

13.2.1.2 Scanning Auger Electron Spectroscopy (AES)

Scanning Auger electron spectroscopy (AES) allows one to obtain a chemical image

of a surface. The Auger effect is at the heart of this technique [33, 35]. When a

material is irradiated by a beam of photons or a beam of electrons with energies

in the range of 2–50 keV, a core state electron may be removed, leaving behind

a charged vacancy or “core hole”. The core hole may be ﬁlled by an outer shell

electron. The energy lost by the electron transitioning to the lower energy level,

which equals t he difference in the energies of the two orbitals, may be coupled

to a second outer shell electron. If the energy transfer is greater than the electron

13 Surface Treatment and Planarization 931

binding energy, the electron can be emitted from the solid. Since orbital energies

are unique to an atom of a speciﬁc element, analysis of the ejected electrons can

yield information about the chemical composition of a surface.

Scanning Auger electron microscopy (SAEM) involves rastering a focused elec-

tron beam across a sample surface and measuring the intensity of the Auger peak

corresponding to a particular element. In this way, high resolution, spatially resolved

chemical images of the surface are obtained. In addition, sputtering can be cou-

pled with Auger spectroscopy to perform depth-proﬁling experiments. Sputtering

is typically performed by bombardment with Ar ions of energies up to a few kV.

In this case, sputtering removes thin outer layers of the surface so that AES can

be employed to evaluate the underlying composition. AES is often used as an eval-

uation tool on and off fabrication lines in the microelectronics industry, while the

versatility and sensitivity of the Auger process makes it a standard analytical tool in

research laboratories.

While Auger electron and X-ray photoelectron spectroscopies provide similar

information, there are distinct differences between the two techniques. Since the

incident electron beam in Auger can be tightly focused, the spot size (or sampling

size) in Auger is much smaller than that in XPS, and thus has the capability of

identifying ﬁne features on the surface. On the other hand, XPS has the capability

to provide detailed information regarding surface chemical structure and bonding

through the use of chemical shifts. Although Auger lines also exhibit chemical

shifts, these are not generally as large or as well documented as those obtained by

XPS. In addition, X-ray radiation used in XPS imparts less damage to the sample

surface than does the electron beam used in SAEM. This feature becomes par-

ticularly important when using these techniques to analyze soft samples (such as

self-assembled monolayers). As mentioned above, the spatial analysis and imaging

capabilities of the scanning Auger microprobe make it a very useful complementary

technique to XPS.

13.2.1.3 Energy Dispersive X-Ray Spectroscopy (EDS or EDX)

Energy dispersive X-ray spectroscopy is another analytical technique used for the

elemental analysis or chemical characterization of a sample [36]. Most commonly,

a high-energy electron beam (e.g., in a scanning electron microscope, where EDS is

often offered as an accessory) is focused onto the sample. The incident beam may

eject an electron from an inner shell, creating a hole. An electron from an outer,

higher-energy shell then ﬁlls the hole, and the difference in energy between the two

shells may be released in the form of an X-ray. The number and energy of the X-rays

emitted from a specimen can be measured by an energy dispersive spectrometer. As

the energy of the X-rays is characteristic of the difference in energy between the

two shells, and of the atomic structure of the element from which they are emitted,

this technique allows the elemental composition of the specimen to be measured.

A chemical map of the surface is obtained by scanning the electron beam across the

sample and measuring X-ray intensity.

932 P. Lin e t al.

13.2.1.4 Secondary Ion Mass Spectroscopy (SIMS)

Secondary Ion Mass Spectroscopy can be used to determine the chemical composi-

tion of the near surface region [37]. The technique involves the removal of surface

species by the process of sputtering, collecting the ejected (so-called secondary) ions

and analyzing them with a mass spectrometer. Sputtering is performed by bombard-

ing the sample with energetic ions. The incident ions set off a series of collisions

within and between target atoms (the so-called “collision cascade”). If atoms reach

the target surface with an energy greater than the surface binding energy, t hey can

be ejected. SIMS is arguably the most sensitive surface analysis technique, capable

of detecting elements present in the parts per billion or less range. It can also detect

all elements, including hydrogen. Since it involves the removal of surface species,

one can also obtain a depth proﬁle of the sample (i.e., composition as a function of

depth). For the same reason, however, it is a destructive technique.

13.2.2 Surface Structure and Morphology

13.2.2.1 Atomic Force Microscopy (AFM)

Atomic force microscopy is a powerful technique for surface analysis [38]. It

employs a sharp probe tip mounted at the end of a compliant cantilever. With the

cantilever held ﬁxed, the surface of a sample is brought close to the tip by a piezo-

driven motorized stage. (The opposite situation is also possible, where the sample is

held ﬁxed and the cantilever brought close to it by the piezo stage.) At short enough

surface-tip separation, surface forces (such as dispersion forces, electrostatic forces,

capillary forces, hydrogen bonding and other chemical forces, magnetic forces and

forces of entropic origin, like solvation forces) become effective and can be mea-

sured. Indeed, a surface-tip force leads to a deﬂection of the cantilever which is

measurable, e.g., by optical means, whereby a laser beam is reﬂected from the

top surface of the cantilever and detected by a photodiode array. Other detection

methods include optical interferometry, capacitive sensing, or piezoresistive AFM

cantilevers.

Constant force imaging is preferred to constant height imaging, in order to avoid

the tip crashing into a sample with abrupt topographic features. This imaging mode

is achieved by a feedback loop. With proper calibration, the feedback response

is converted to a topographic image of the sample, normally with sub-nm reso-

lution. The AFM can be operated in several modes, which may be divided into

static and dynamic modes. The static mode measurements are usually done in “con-

tact” regime (where the tip senses the repulsive core of the surface potential). In

the dynamic “noncontact” mode, the cantilever is made to vibrate at or near one

of its resonance frequencies. The oscillation amplitude, phase, and resonance fre-

quency are affected by the tip-sample forces; these changes provide information

about the surface force, and thus, about the chemical and topographic characteristics

of the sample. Atomic force microscopy can also be used to measure the adhesion or

13 Surface Treatment and Planarization 933

pull-off force for a given tip-surface combination. The data obtained using this tech-

nique is in the form of an AFM force curve, in which the deﬂection of the cantilever

beam holding the AFM tip is measured as the tip approaches, makes contact with,

and withdraws from a surface. Additionally, AFM can be used to measure frictional

properties of the sample by tracking the lateral deﬂections of the cantilever as the

probe tip is scanned perpendicular to the longitudinal axis of the cantilever at various

applied loads. A hardened AFM tip (e.g., diamond or diamond-like carbon coated)

can also be used to estimate the hardness of the sample through nanoindentation

experiments.

13.2.2.2 Scanning Electron Microscopy (SEM)

Scanning electron microscopy (SEM) employs an electron beam, which typically

has an energy ranging from a few hundred eV to 40 keV [36]. The beam is focused

onto the sample to a spot size in the range of a few to less than 1 nm. The energy

exchange between the electron beam and the sample results in the reﬂection of high-

energy electrons by elastic scattering, emission of secondary electrons by inelastic

scattering and the emission of electromagnetic radiation, each of which can be

detected by specialized detectors. The beam is scanned across the sample while

the secondary electrons are detected synchronously.

When compared to optical microscopy, this technique offers several advantages,

including a larger depth of focus and higher magniﬁcation, which are of great value

for MEMS inspection. The electrostatic interaction between the electron beam and

MEMS structures is sometimes employed to actuate the structures, thus allowing

researchers to probe MEMS dynamics while directly observing the structures under

high magniﬁcation.

13.2.3 Surface Energy Measurements

Contact angle measurements can be used to estimate surface energy of solids by

measuring the contact angle, θ , of a liquid drop on a solid surface [39, 40]. This is

the angle formed by the liquid-vapor surface and the solid-vapor surface, measured

in the liquid such that a zero value of the contact angle denotes complete wetting of

the solid surface by the liquid. A common method for contact angle measurements

is the “sessile drop” method, in which a liquid drop is placed onto a solid surface

and the resulting t angent angle at the liquid/solid/vapor intersection point deter-

mines the contact angle. The interaction energies between solid and vapor, solid

and liquid and liquid and vapor are related through Young’s equation [39]. Contact

angle measurements are typically made with pure liquids in saturated vapor, and

often advancing and receding angles (the angles measured when the liquid drop is

advancing or receding over the surface), or the average of the two, are reported.

If the liquid interacts strongly with the solid surface (e.g., water on oxidized sili-

con, a hydrophilic surface), the droplets spread out on the solid surface, yielding a

water contact angle close to 0

◦

. If the liquid is water, the wetted surface is said to