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

Подождите немного. Документ загружается.

984 P. Lin e t al.

reﬂectivity curve, especially the position of the SPR angle, is shown in Fig. 13.37a.

The SPR angle depends strongly on the change in refractive index on the metal sur-

face. As a result, the surface is often coated with a monolayer of active molecules

and is used to study the mass absorbed on the surface, such as protein and DNA.

The SPR angle shift as function of time can also be observed. Monitoring the angle

shift in real time allows the study of the kinetics of molecular interactions on the

surface at different conditions, for example during the injection of another chemi-

cal (Fig. 13.37b). This means that the association and dissociation of biomolecules

can be observed, and the rate constants and equilibrium constants can be

calculated.

Reflectivity

Angle

Time

(A)

(B)

Fig. 13.37 (a) The SPR angle changes as the surface absorption changes; (b) the time-dependent

SPR angle changes

As mentioned, the position of the reﬂectivity minimum is very sensitive to the

properties of the liquid-metal interface surface, such as refractive index (down to

–5

) and molecule absorption (up to a few hundred nanometers) [121, 123, 124].

No photons exit the reﬂecting surface, but the electric ﬁeld decreases exponentially

with distance from the interface, decaying over a distance of about a quarter of the

light wavelength beyond the surface. In the case of gold and glass with a light wave-

length of 633 nm, the SPP propagation decay length is calculated to be 9.98 μmin

the x-direction, and in the z-direction, 27.6 nm into gold and 213.5 nm into glass.

This illustrates the much shorter decay length in gold than in glass. The penetra-

tion depth of the evanescent ﬁeld depends on the wavelength of the incident light.

Factors such as metal ﬁlm thickness, surface adsorbed species, and liquid dielec-

tric properties can affect the SPR results. A background run is often performed as a

baseline prior to the study of the mass change on the surface.

SPR offers several advantages. First, it is a label-free detection technique.

Second, it is sensitive enough to distinguish surface interaction as opposed to bulk

interaction. It can be used to study kinetics by monitoring molecular interactions

in real time. Third, the detection is not affected by sample color. Lastly, many self-

assembly monolayers (SAMs) can be coated on gold or silver for various interfacial

interaction studies (see Section 13.4).

13 Surface Treatment and Planarization 985

13.6.2 Material Properties and Process Selection Guidelines

In many Optical MEMS or BioMEMS applications, surface coatings are effective

in creating new optical and chemical properties on the surface while retaining the

mechanical and optical properties in the bulk. These surface condition modiﬁcations

can be employed to manipulate light or to control bio and chemical/ and physical

interactions in the devices. Surface properties are often the determining factors for

the performance of a MEMS device. In terms of optical properties, many phenomena

can be manipulated on the surface of the MEMS devices, such as light reﬂection,

transmission, absorption, and polarization. These functional optical properties are

directly related to the fabrication methods and processing conditions. To achieve

the desired optical properties, many design and fabrication limitations must be con-

sidered thoroughly in advance, including the surface topography and roughness,

material selection and compatibility, fabrication process, and integration associated

interface issues. Therefore, this section will focus on methods for selecting and

arranging optical thin ﬁlms on a surface to provide the optical properties desired

for MEMS applications.

13.6.2.1 High Reﬂection Applications

The control of surface reﬂection of light is one the most common uses of surface

coatings. Two extremes of light reﬂection are often required, i.e. very high sur-

face reﬂection (high reﬂection) and very low surface reﬂection (antireﬂection). High

reﬂection is suitable for applications that require directional manipulation of light,

while antireﬂection is necessary for the maximum light intensity to pass through

the surface. Optical design considerations for these two extremes are relevant, and

considerations for each case will be discussed in the next two sections.

To achieve a very high surface reﬂection rate, one can either choose a thin metal

ﬁlm or multi-layers of coating. For metal thin ﬁlms, Equation (13.32)givesthe

reﬂectance R of a metal with good conductivity. Usually the values of n

and n

are

close, but both are much smaller than n

. As a result, an estimate of the reﬂectance

is given by

R =

− n

)

+ n

)

+ n

≈

(2n

)

+ 1

≈ 1–4(

)

(13.44)

If n

is 20% of n

, a reﬂectance of 0.84 can be easily achieved. For a metal

with very good conductance, i.e. a very low n

ratio, the reﬂectance approaches

100%. For example, Al (n

= 1.3/7.11 = 0.183 at 650 nm) and Ag (n

0.07/4.2 = 0.017 at 650 nm) have very high reﬂectances of 90.5 and 98.8%,

respectively, at 650 nm, and Au (n

= 0.142/3.74 = 0.042 at 650 nm) has a

reﬂectance of 99% in the infrared range (1500 nm). These metals have been coated

on the surfaces of MEMS devices to obtain high reﬂection.

986 P. Lin e t al.

On the other hand, to enhance the reﬂection efﬁciency, multiple layers can be

arranged together in a (glass)(HL)

H(air) system [125]. The multiple layers form a

high-reﬂectance central zone at ω

, and the width of the central zone increases with

the value of n

. The maximum reﬂectance can be close to 1.0 with an additional

H-layer on top of the periodical structures, as the shown in Fig. 13.38. As a result,

mirror-like surfaces can be achieved with high reﬂectance in the visible spectrum

via the proper arrangement of multilayer periodic systems.

(glass)(HL)

H(air)

(glass)(HL)

H(air)

(glass)(HL)

(air)

Wavelen

th (nm)

1.0

0.8

0.6

0.4

0.2

Reflectance

Fig. 13.38 Reﬂectance of multiple structures [125]. Reprinted with permission. Copyright

2008 Institute of Physics

Many MEMS mirror structures use Au on Si to reﬂect light. The reﬂectivity typ-

ically depends on the polarization. This is known as polarization dependent loss

(PDL) for optical switch applications. One example [126] showed that s-polarized

light (in transverse electromagnetic ﬁeld) reﬂected more strongly than p-polarized

light. To reduce the PDL, a multilayer dielectric coating on top of the gold mirror

was used. A gold ﬁlm, 60–100 nm thick, deposited over an underlying Ti or Cr

adhesion layer about 10 nm in thickness, was used as the reﬂective surface. Both

layers were deposited by thermal evaporation or sputtering. Finally, a multi-layer

dielectric overcoat was deposited using an e-beam evaporator. This multi-layer coat-

ing contained 4 pairs of SiO

/Ta

thin ﬁlm layers and had a total thickness of

2.037 μm. The PDL was reduced and the overall reﬂectivity was also improved

with this multi-layer coating [126].

13.6.2.2 Antireﬂection Applications

Antireﬂection coatings have been widely used in optical MEMS for reﬂection reduc-

tion and increased transmission of light into the materials [127–133]. AR coatings

13 Surface Treatment and Planarization 987

have a positive effect on energy efﬁciency (e.g. in solar cells), light transmission

(in micro display panels), light accumulation (in micro IR detectors), and photon

enrichment (in optical based bio-sensors) [127–133]. Many applications including

micro lenses [133], micro mirrors [ 125, 131, 134], micro prisms [124, 130], and

micro beam splitters [131, 134] have been reported. Antireﬂection coatings are also

widely employed to maximize light transmission into the active regions of solar

cell systems [135–147]; to lengthen polymer lifetime in outdoor applications such

as UV protection [148]; to lower energy consumption, for example by using ther-

mochromic glazings on building windows [149]; and to increase IR transmission in

sensitive infrared spectrometers and radiometers [150].

In MEMS related optical systems, dielectric and polymer materials are both com-

monly used in optical components. However, the surface treatment methods for each

type of material are distinct and must be considered carefully in the design phase.

Principles of antireﬂection for two different surface treatment methods (multi-layer

coatings and nano-structure fabrication) were described in Section 13.6.1.3.

Material Properties and Selection

For dielectric optical substrates, one of the most widely adopted methods for surface

thin ﬁlm coating is the vacuum coating process, including physical and chemical

vapor deposition. Physical vapor deposition, especially sputtering, is the method

employed most often f or thin ﬁlm coatings because it provides highly accurate thick-

ness control (in the nanometer range), good ﬁlm quality for both metal and dielectric

materials, and conformal coatings on structures [109]. Vacuum coating processes

for dielectric materials are mature processes and the process considerations are

described in detail in Chapter 4 (Section 4.2).

The properties of s everal dielectric substrate materials, such as commercially

available optical glasses and crystalline materials, are listed in Tables 13.5 and 13.6

as design references.

Case Study 6: Nanostructure on Surface for Antireﬂection (Moth Eye Effect)

In addition to thin ﬁlm coatings, nanostructures fabricated on the substrate surface

can also be used for antireﬂection. These s tructures are in the submicron range for

visible light (wavelength ranging from about 400 nm to about 700 nm) and can cre-

ate an effective index gradient from the substrate to the incident media [151]. The

phenomenon of antireﬂection by sub-wavelength structures was ﬁrst observed on

moths’ eyes for night vision [113], and has the advantage of low sensitivity to the

variation of light incidence angle, a characteristic that would usually require a com-

plex design for a multi-ﬁlm system. In the visible spectrum, the structure sizes must

be smaller than 300 nm and the total depth should be in the range of several hundred

nms. Several methods have been developed for the fabrication of nanostructures on

the substrate, including the interference of two coherent beams, holographic optical

process [152], stochastic dry etching processes [153], hot embossing/casting on a

nano mold [154, 155], rough PVD coating [156], nano phase separation [157–160],

988 P. Lin e t al.

Fig. 13.39 Nano structures fabricated by different technologies (a) nano casted PDMS structures

from Lotus leaf, (b) SU-8 nano ﬁbers made with the Self-Masked High-Aspect-Ratio (SMHAR)

process, and (c) self masked silicon oxide nano cones [155, 162]. Reprinted with permission.

the use of self assembled nanostructures [157, 161], and self nano masking pro-

cesses [162]. Some of the fabricated nanostructures are shown in Fig. 13.39 [155,

162].

The proposed mechanism of nanostructure formation in Fig. 13.39b is schemati-

cally depicted in Fig. 13.40 [162]. In conventional RIE etching (without a shielding

cover glass), short nanopillars were produced due to the nonhomogeneity of the

polymer material as prepared during the fabrication process. The aspect ratio of

the nanopillars remained low (usually smaller than 5) even with increased etching

time. This was because the etching rate at the top of the nanopillars was s imilar

to that on the bottom surface in RIE. However, in the Self-Masked High-Aspect-

Ratio (SMHAR) process [162], nanomasks were created from the cover glass and

deposited onto a Parylene C surface to help with the formation/passivation of the

HAR nanopillars. In the early stages of RIE etching in the SMHAR process, short

nanopillars (seed structures) were generated, and a small number of nanomasks were

produced simultaneously with no obvious passivation effect on the seed structures

(Fig. 13.40b). As the etching progressed, a sufﬁcient number of nanomasks were

produced to accumulate on the seed structures, thereby extending the length of the

seed structures and forming HAR nanopillars (Fig. 13.40c). This is preferred to the

substrate shown in Fig. 13.40d, which might result from the higher surface curvature

and thus, the higher charge density of the seed structures [162]. Nanomasks were

found to reduce the etching rate in RIE f or shielded nanopillars, as indicated by the

slope difference. In addition, ﬂuorocarbon was also generated in the RIE process,

and it provided sidewall passivation. The bottom surface has a stronger etching rate

than the sidewalls due to the direct physical ion bombardment during the RIE pro-

cess. Therefore, multiple effects were incorporated to induce both top and sidewall

surface passivation of the nanopillars. Consequently, the height of the nanopillars

continually increased as the RIE process proceeded, as shown in Fig. 13.40c, d.The

density distribution of nanopillars illustrated a gradient correlated with the distance

from the cover glass [162].

Most of the processes are cost effective and easy to use for batch fabrication,

and thus have high potential for commercial applications. However, the intrinsic

problems with nanostructure-based AR surfaces are the relatively weak mechanical

13 Surface Treatment and Planarization 989

(a)

Sidewall

passivation

Parylene C

Nanomasks

(b)

(c)

(d)

Plasma

Fluorocarbon

polymer

Nanomasks

etching

(a)

Glass substrate

Parylene C

Nanomasks

(b)

(c)

(d)

Plasma

Nanomasks

deposition

Fig. 13.40 The mechanism by which SMHAR nanopillars form [162]: (a) cover glass partially

shields the Parylene C; (b) short nanopillars (seed structures) form initially, and only a small num-

ber of nanomasks are deposited; ( c) more nanomasks are generated, and passivation masks form

during the RIE etching process; (d) HAR nanopillars form in RIE; the tops and sidewalls are both

Physics

properties of the ﬁlm and the difﬁculty of surface cleaning. These must be further

addressed before practical use is made.

The simplest one-dimensional AR nanostructure surface design is a binary

structure, similar to an array of square rods extruded from the substrate surface,

as illustrated in Fig. 13.41a. The analogous thin ﬁlm coating is illustrated in

Fig. 13.41b, with the effective index n

eff

equal to the average index of the nanorod

array in (a). The fraction factor of the rod array can be expressed as f = b/P, where

b is the width and P is the period of the rod structure. From the Fresnel equation, for

zero reﬂection from the ﬁlm, the depths of the binary structures must satisfy:

990 P. Lin e t al.

eff

(a)

(b)

eff

(a)

(b)

Fig. 13.41 (a) One dimensional binary AR surface (b) effective thin ﬁlm coating on substrate,

analogous to the binary surface

h =

√

(13.45)

where λ is the operation wavelength, and n

and n

are the indices of substrate and

medium, respectively. The effective index n

eff

also needs to satisfy the following

condition:

eff

√

= n





+ n



1 −



(13.46)

The nanostructures also need to satisfy the following conditions according to

Equations (13.38) and (13.39):

P ≤

and 0.4λ

max

≤ h ≤ λ

max

For example, considering the design of AR nanorod structures on a glass sub-

strate (index of 1.5) in air for green light at 530 nm, the required period P is:

P ≤

353 nm and 212 nm ≤ h ≤ 530 nm (13.47)

Assuming that the maximum P (353 nm) is to be used for easier fabrication, the

width of the square nanorod, b, is calculated to be about 159 nm and the height of

the nanostructures, h, is about 108 nm. These calculations are based on very simple

and perfect nanostructures. In a real process, the shape, size, roughness, aspect ratio,

and distribution of the nanostructures all affect antireﬂection results. The estimated

height may deviate from the range estimated by Equation (13.39) because of the

very simple two-layer design. For more complex designs, see references [116–118]

for more detailed calculations and simulations.

13.6.2.3 Considerations for Surface Smoothness and Roughness

When using microstructures for optical applications such as mirrors, prisms, or

lenses, the surface roughness must be controlled sufﬁciently to meet at least one

Rayleigh Limit (RL) for high optical performance. That is, the peak to valley optical

13 Surface Treatment and Planarization 991

path difference (PV-OPD) of the wavefront reﬂected or diffracted from the opti-

cal component must be smaller than λ/4 [109]. As a result, the corresponding root

mean square optical path difference (RMS-OPD) needs to be smaller than λ/14

[109]. For example, for a mirror in an integrated DVD pick up head in the blue

ray region (λ =405 nm), the RMS-OPD or surface roughness must be controlled to

less than 29 nm. Therefore, surface smoothness is a key parameter for micro optical

components.

Many methods for surface smoothing following microfabrication have been

reported for optical applications, including spin coating [163–169], surface pres-

surization [170], thermal annealing [171, 172], ion-beam/RIE etching [173–176],

and chemical mechanical polishing (CMP). The CMP process will be introduced in

detail in Section 13.7.

Among these surface planarization processes, spin coating is one of the most

commonly used methods due to its simplicity, low temperature, and effectiveness

in surface planarization. Spin coating is often used for planarization of polymers

[163, 164, 166–168] and spin on glasses [165]. For polymers, the control of spin

speed, number of layers, material properties (viscosity, volatility, and shrinkage),

and annealing conditions are important parameters. In general, low spin speed, low

viscosity, low shrinkage, and non-volatile polymer content contribute to improved

smoothness [163]. Annealing can induce reﬂow dominated by surface tension

effects for the planarization of thermoplastic type polymers. The application of

multiple layers can also improve the surface ﬂatness [164, 166, 167]. Many poly-

mer materials have been reported such as nonvolatile photopolymer Si-14 [163],

PC403 photopolymer [164], photoresist ORDYL PR125 [166], polyimide [166],

BCB polymeric resin [168], and polyimide-silica microcomposite ﬁlms [169]; in

those materials, the degree of planarization (DOP) has been shown to approach 95%

[163], 75% [164], 99% (for deep trench) [116], and 80% [118], respectively. Here

the DOP is deﬁned as:

DOP = 1 −

(13.48)

where S

and S

represent the step height of the surface before and after planariza-

tion [163, 168]. The degree of planarization (DOP) for spin on glass approaches

99% for very thin photonic crystal structures 160 nm thick [165].

In addition to polymers, metal thin ﬁlms can also be planarized by applying a

high pressure to the metal ﬁlm against a ﬂat surface [170]. For example, a pressure

of 600 MPa was applied to a 100 nm Ag ﬁlm against a (100)-oriented silicon wafer

[170] that had been polished on both sides to reduce the original 13 nm RMS rough-

ness on the Ag ﬁlm to less than 0.1 nm RMS. Bumps, asperities, rough grains and

spikes in vacuum-deposited metal ﬁlms can be thoroughly ﬂattened.

As for semiconductor/dielectric materials, such as silicon [171] and sol-gel glass

[172], thermal annealing is an effective surface smoothing technique. For example,

in one study [171], a hydrogen atmosphere was introduced during a high tempera-

ture (1000–1100

◦

C) annealing process, promoting silicon atom migration and thus

992 P. Lin e t al.

smoothing the surface [171]. The surface RMS roughness was reduced from 20 nm

after dry etching to 0.26 nm following the annealing process [171]. Annealing can

also be used to smooth the surface of a s ol-gel glass structure; annealing at 1260

◦

for 10 min was found to effectively smooth the surface without much structural

change [171]. As a result, very low scattering waveguide structures can be fabricated

using the annealing processes [171, 172].

Although adding material and rearranging surface atoms are effective methods

for surface planarization, removing material from the surface using etching is an

alternative technique for surface roughness reduction. Etching is usually performed

either by ion-beam etching [173, 174, 176] or by reactive ion etching (RIE) [ 175].

Small particles as large as 50 nm can be smoothed to 1 nm by in situ ion beam etch-

ing during thin ﬁlm deposition [173]. The direction of the beam is also important

for surface smoothness; smoothness can be attained either by using a grazing inci-

dent angle [176], or by rotating the sample to achieve a smaller incident angle [174].

Average surface roughness can be reduced from 40 to 10 nm [176] and from 2 into

0.2 nm [ 174], respectively. The latter is suitable for the smoothing of high precision

lenses, extreme-ultraviolet lithography, and X-ray optics [174]. RIE processes have

also been employed to smooth the surface of a CVD deposited diamond thin ﬁlm.

To increase the smoothing effectiveness, rapid thermal chemical vapor deposition

(RTCVD) was used to deposit a thin ﬁlm of SiO

onto the diamond ﬁlm [175].

Surface roughness was reduced from 40 to 14 nm by the RIE process; the treated

surface was suitable for X-ray lithography [175].

13.6.2.4 Polymer Materials for Optical Applications

Vacuum coatings on polymer materials are relatively new and have been devel-

oped for only about 20 years [177]. Since the use of polymer optical components

in MEMS is becoming increasingly popular, this ﬁeld will become more impor-

tant in the future. Due to signiﬁcant differences in physical and chemical properties

between polymers and dielectric materials, special efforts are required to determine

the proper substrate materials and process conditions for reliable surface coatings.

For precision optical parts, transparent polymers such as poly-methylmethacrylate

(PMMA) and biophenol-A polycarbonate (PC) have been the most commonly

used substrates for more than 50 years [177]. Other polymers such as polyamides,

polyether sulfones, and cyclooleﬁn polymers have been developed more recently

[177, 178]. These thermoplastic materials can be processed by injection molding,

hot embossing, or dry etching. Photo patternable polymers such as photoresists

and SU-8 have also been adopted for optical MEMS applications owing to their

ease of patterning and reshaping for optical component fabrication [131–134]. The

properties of some polymers used in optics are listed in Table 13.7.

13.6.2.5 Surface Coatings for Polymer Materials

Surface coatings can increase the performance of optical plastics for many applica-

tions such as antireﬂection, mechanical protection, chemical resistance, reduction

13 Surface Treatment and Planarization 993

of gas permeability, antistatic, and surface wetting angle control [110, 179, 180].

Usually coatings used on optical plastics consist of an optical interface, protective

layers, and outer surface layers [110]. The materials used for building up antireﬂec-

tion coating layers usually contain metal oxides. The most frequently used low index

material is silicon dioxide (n =1.5), while high index metal oxide materials range

from alumina (n =1.67) to titanium dioxide (n =2.4). Selection of the coating mate-

rial is critical to minimize unwanted effects such as heat or ultraviolet light emission

during the evaporation/sputtering processes, which can cause thermal stresses and

optical property changes.

Two technologies are often used in the preparation of thin ﬁlm coatings on plastic

materials: vacuum coatings and wet-chemical coatings. A low temperature process

(<120

◦

C) is required for vacuum coating of most thermoplastics [110]. However,

low temperature processes usually produce thin ﬁlm coatings with undensiﬁed

porous structures, which can cause adhesion problems on the substrate. As a result,

the optical properties of the thin ﬁlm are not stable when the coating is exposed to

moisture in the atmosphere [110]. To solve this problem, plasma ion-assisted depo-

sition (plasma-IAD, in Fig. 13.42) was developed to bombard the substrate with

ions during thin ﬁlm deposition [110, 181, 182]. An ion source used in one study

[177] was a low pressure plasma source with an ion energy level ranging from 70 to

150 eV. The material properties of the thin ﬁlm, including surface energy, adhesion,

refractive index, hardness, and surface topography, can be varied by adjusting the

surface ion bombardment [110, 183].

Other processing methods for vacuum coatings include plasma-impulsed chem-

ical vapor deposition (PICVD) and plasma enhanced chemical vapor deposition

Ion source

E-beam gun

Optical Society of America