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

Подождите немного. Документ загружается.

684 D.R. Hines et al.

material will be dissolved away at the top than at the bottom of the exposed area.

For a positive photoresist, such an exposure proﬁle versus depth results in an “over-

cut” sidewall proﬁle (a sidewall angle less than 90

◦

). This is acceptable for etching

applications but not for “liftoff” applications. For a negative photoresist, such an

exposure proﬁle versus depth results in an “undercut” sidewall proﬁle (a sidewall

angle greater than 90

◦

). This is desirable for liftoff applications but can result in

loss of pattern ﬁdelity for etching applications. Ideal, overcut, and undercut sidewall

proﬁles are illustrated in Fig. 9.17.

Fig. 9.17 Edge proﬁle of photoresist layer resulting from absorption of light during exposure

9.2.3.1 Positive Photoresist

Upon exposure to light, the solubility of a positive photoresist increases due to poly-

mer chain scission. Common formulations of positive photoresists have included

diazoquinone ester (DQ) as a sensitizer in a novalac resin (phenol formaldehyde

resin) (N). These molecules are shown in Fig. 9.18. Such formulations are referred

to as DQN resists. A typical DQN resist formulation is listed in Table 9.2.

9 MEMS Lithography and Micromachining Techniques 685

Fig. 9.18 Chemical formulas of (a) diazoquinone ester and (b) novolak resin components of

standard positive photoresists (Reprinted with permission from [7])

Table 9.2 Photoresist

components for positive DQN

resist

Resist component Weight, g

DQ 2

Novolak 8

Cellosolve acetate 20

Butyl acetate 2

Xylene 2

Water (trace) 0.05

Reprinted with permission from [7]

DQN-based resists are insensitive to oxygen and work very well at near ultravio-

let (UV) wavelengths provided by the 365 (i-line), 405 (h-line), and 435 (g-line) nm

lines of an Hg lamp typically found in research and development cleanroom facili-

ties. For deep UV (DUV) (248 nm) and extreme UV (EUV) (193 nm) applications,

positive photoresist formulations that do not contain aromatic molecules, such as

PMMA-based materials, need to be utilized. A 15% loading of the photoresponsive

DQ is enough to photochemically transform the two-component resist into a polar

(base) soluble product upon exposure to UV light [7].

Positive resists are generally more expensive than negative resists; however,

they exhibit higher pattern resolution causing them to be more widely used for

photolithographic processing applications. Positive resists do not tend to swell

when submerged in developer and can be developed in aqueous-based solvents

that decrease the amount of generated hazardous waste material. Photochemical

686 D.R. Hines et al.

reactions in a positive resist are not sensitive to oxygen and the phenolic resin com-

ponent provides plasma etch resistance to the resist. DQN-based positive resists

can be thermally stabilized to 200

◦

C and are suitable for use for image reversal.

However, they tend to have a narrower processing window and exhibit lower wet

chemical etch resistance and lower adhesion to substrate surfaces. For this reason an

adhesion promoter such as hexamethyldisilazane (HMDS) is applied to the substrate

surface prior to coating the substrate with a photoresist layer [7, 36].

Commonly used positive resists include Shipley 1813 and Fujiﬁlm OiR-906-10.

Processing case studies are included below for both of these resists. Other common

formulations of positive photoresist are shown in Table 9.3.

Table 9.3 Commonly used positive photoresists

Polymer References from Moreau

Polyhydroxystyrenes [23, 185]

Polyvinylphenol [24]

Styrene-methacrylic acid [25, 186]

PMMA-methacrylic acid [26]

Styrene-acrylic acid [27]

Esteriﬁed polyvinyl alcohol [28]

Polyimide [29]

Silylated novolak [30, 211]

Reprinted with permission from [7]

9.2.3.2 Negative Photoresist

Upon exposure to light, the solubility of negative photoresists decreases due either

to cross-linking of the polymer chains or a photochemical reaction. Many polymers

will cross-link when exposed to light. Various types and classiﬁcations of negative

photoresists are listed in Table 9.4.

An early negative photoresist formulation referred to as KTFR combined a bis-

aryldiazide photosensitive cross-linking agent with a polyisoprene cyclized polymer

as shown in Fig. 9.19 [7, 37, 38].

Such resists suffer from sensitivity to oxygen and solvent swelling. To over-

come such deﬁciencies, onium salts [39, 40] were identiﬁed as photosensitive

Table 9.4 Classiﬁcations and types of negative photoresists

Thermoset Elastomers Thermoplastics

Diallylphthalates Acrylates Acetates

Epoxies Polysulﬁdes Acrylates

Phenolics Neoprenes Cellulosics

Urethanes Polycarbonates

Silicones Polystryenes

Polyimides

Reprinted with permission from [7]

9 MEMS Lithography and Micromachining Techniques 687

Fig. 9.19 Negative photoresist components before and after exposure to UV light (Reprinted with

permission from [38])

additives that, when exposed to light, catalyze (chemically amplify) cross-linking

or decreased solubility.

Negative resists are not widely used for photolithographic processing applica-

tions because they exhibit a lower pattern resolution in comparison to positive

resists. They also tend to swell when submerged in developer. Negative resists are

cheaper and have higher chemical resistance than positive resists. They are also

more adhesive to substrate surfaces and harder to remove after processing. In addi-

tion, negative resists exhibit larger processing windows and can be formulated from

a wide variety of polymers.

Common types of negative resist include chemically ampliﬁed epoxy or vinyl

derivatives. A case study of the common Futurrex NR7-1500PY is presented below

[30, 38].

For MEMS processing, SU-8 has recently become a commonly used negative

resist material. SU-8 was ﬁrst developed by IBM in 1989 and is now sold by

MicroChem Corporation. SU-8 is highly transparent above 360 nm. This means

that UV light can penetrate through a thick layer (up to about 2 mm). The lack of

absorption results in nearly vertical sidewalls, thus allowing the fabrication of SU-8

structures with very high aspect ratios (>20). Because of this unique feature SU-8 is

not usually used as a sacriﬁcial pattern transfer layer but rather as a patterned poly-

mer layer that is permanently incorporated into device fabrication. SU-8 is presented

in more detail in Chapter 5.

9.2.3.3 Image Reversal for Positive Resist (Converting Positive Resist

into a Negative Resist)

Because of the typical overcut sidewall edge proﬁle in a positive photoresist, it is not

suitable for liftoff processes. Also, a negative resist is not as attractive as a positive

688 D.R. Hines et al.

resist in many applications because of lower resolution and the generation of larger

amounts of hazardous waste. Therefore, image reversal processes have been devel-

oped that enable a positive resist to be used to produce a negative tone image. The

image reversal process for a positive resist is accomplished by postbaking the resist

layer in an ammonia vapor oven [31, 41–43]. The interaction of the ammonia with

the UV exposed regions of the photoresist (i.e., soluble regions) renders the pho-

toresist insoluble and also insensitive to further UV exposure. A subsequent ﬂood

exposure converts the previously unexposed regions of the photoresist to a soluble

form. Thus, upon developing, the initially unexposed regions of the photoresist are

dissolved away leaving a negative image of the initial photomask pattern.

9.2.4 Substrate

Historically, MEMS and IC fabrication have been performed using Si, Ge, or GaAs

substrates. Flat panel displays are typically fabricated on glass substrates. More

recently, microﬂuidic devices, lab-on-a-chip, and ﬂexible electronics have incor-

porated polymeric substrates. Often transparent substrates can be more challenging

to perform lithography on, as the reﬂection or lack thereof can broaden exposure

regions, or reduce the dose directed to the photoresist. In all cases, photolithogra-

phy demands that the substrate be ﬂat and well cleaned. Inorganic s ubstrates can

be cleaned in several ways: for example: RCA clean, plasma (or UVO) clean, or

solvent clean.

9.2.5 Processing Steps for UV Lithography

9.2.5.1 Deposit Photoresist

Spin Coating

Photoresist is typically applied t o substrates via spray or spin coating. Although

spin coating is a rather simple process, the tolerances for ﬁlm thickness, unifor-

mity, and repeatability raise a number of important issues that must be understood

and controlled. As illustrated in Fig. 9.20, there are typically four stages of ﬁlm

evolution during the spin-coating process. The puddle formation stage results from

the dispensing of the photoresist onto the substrate surface prior to spinning. The

photoresist at this point is a thick puddle that covers some or all of the substrate.

The ﬁlm formation stage develops during acceleration of the spin coater as spinning

commences.

The photoresist at this point covers the entire substrate, is starting to thin out at

the center of rotation and is getting thicker toward the edges of the substrate. The

thick-ﬁlm stage develops after acceleration when the spin coater has reached the

desired spin speed. The photoresist at this point is a relatively uniform but thick

ﬁlm (tens to hundreds of microns thick) that may contain thickness variations near

9 MEMS Lithography and Micromachining Techniques 689

Fig. 9.20 Stages of ﬁlm evolution during the spin coating process (Modiﬁed from [44] and used

with permission)

the edge of the substrate. The thin-ﬁlm stage develops after tens of seconds at the

desired spin speed. The photoresist at this point is approaching its target thickness

over the entire area of the substrate and may still have thickness variations only at

the very edge of the substrate. As the ﬁlm progresses from the thick-ﬁlm stage to

the thin-ﬁlm stage, optical interference fringes resulting from reﬂection off both the

top surface of the resist layer and the interface between the resist and substrate can

be seen to change color as it gets thinner. Viewed at an appropriate angle, these

thickness fringes can be seen to appear at the center of the wafer and rapidly travel

radially out to the edge of the wafer while the wafer is spinning.

Several artifacts can appear in the photoresist as a result of spincoating: (1) stri-

ations that are nonuniformities in thickness typically on the order of 30 nm caused

by nonuniform drying of the resist after spin coating; (2) edge effects that result

from excess resist buildup at the wafer edge that can be 20–30 times thicker than

the resist layer; and (3) streaks caused by particulate contamination on the surface

of the substrate or in the resist itself.

As IC feature sizes decrease, resist thickness must also decrease. For MEMS pro-

cessing however, such state-of-the-art feature sizes are typically not encountered.

Rather, MEMS processing may require thicker resist layers that cover surfaces con-

taining intricate topology, thus deﬁning a different set of r esist constraints t hat affect

690 D.R. Hines et al.

viscosity, solvent evaporation rates, and so on. Such constraints must be factored

into the production of ﬂat, uniform, stress-free ﬁlms.

After spin coating, the remaining solvents must be removed from the resist ﬁlm.

This is typically performed with a prebake (or soft-bake) on a hotplate or in an oven.

On a hotplate the temperature increase of the photoresist starts at the bottom of the

ﬁlm and progresses upward. This prevents solvent vapors from getting trapped in the

ﬁlm during the prebake. In contrast, in an oven the temperature rise is more uniform

throughout the ﬁlm which can decrease the rate at which solvents are driven out

of the ﬁlm. In an oven, the top surface of the photoresist can form a skin that may

trap solvent vapors and subsequently form bubbles or other imperfections in the

resist layer. For this reason, prebaking in an oven is typically a slower process than

prebaking on a hotplate.

Spray Coating

For substrates that have very nonuniform surfaces or deep trenches and tall features,

as is common with MEMS processes, spin coating will not be able to produce a uni-

form photoresist coating. In cases such as these, the more commonly used method is

spray coating [45, 46]. In a spray coating system, photoresists are mixed in different

ratios with solvents to facilitate the spraying onto the surfaces of the wafer [47]. In

the spray-coating process, there is direct perpendicular impingement of the coating

solution that promotes coverage into deep trenches and along tall features. This is

usually achieved by a reciprocating spray nozzle that coats the wafer surface to the

desired photoresist thickness. For thicker ﬁlms, the solutions used in spray coating

are often diluted as compared to that used to achieve a similar spin-coated ﬁlm thick-

ness. These systems are critical to the fabrication of many high-aspect-ratio MEMS

structures. Unfortunately, very few commercialized spray coating photoresists exist

and most need to be mixed and characterized on a case-by-case basis [47].

9.2.5.2 Expose Photoresist

For a positive photoresist layer, the solubility of certain well-deﬁned areas can be

increased by the exposure to UV light. In contrast, for a negative photoresist layer,

the solubility of certain well-deﬁned areas can be decreased by the exposure to UV

light.

This is performed by mounting a photoresist coated wafer and a photomask con-

taining the desired mask pattern into an optical projection system (i.e., a contact

aligner or stepper). The mask pattern and wafer are aligned in X, Y, Z, and Rotation

as needed. Once properly aligned, UV light is transmitted through the mask to

expose the desired area of the photoresist layer as described above. Depending on

the resist properties, a postexposure bake may be necessary to achieve the desired

solubility change in the exposed region of the resist layer.

9 MEMS Lithography and Micromachining Techniques 691

9.2.5.3 Develop Photoresist

After exposing selected areas of the photoresist layer to UV light, the soluble regions

can be removed by submerging the photoresist coated wafer in a solution of the

appropriate solvent. A 30–90 s soak is typically sufﬁcient to wash away the soluble

regions of the exposed photoresist layer. To halt the developing process, the wafer is

removed from the developer solution and then typically submerged into a deionized

(DI) water bath for approximately 30 s and dried in a stream of clean/dry N

gas.

At this point the photoresist layer should contain a pattern of open areas identical

to the mask pattern. A careful inspection of the open areas and sidewalls with a

high-powered optical microscope is recommended.

If there is any photoresist visible in the open areas then either the photoresist

was not exposed to enough UV light or it was developed for too short a time. If the

photoresist was underexposed then the lithography process will need to be repeated

with a fresh layer of photoresist and a longer exposure time. If the photoresist was

underdeveloped, the photoresist layer can be returned to the developing bath for a

short time. It is noted that there exist a large number of developer solutions that

can be used to an equal effect. In most cases, however, manufacturers of photore-

sist create various developers that are speciﬁcally optimized to be used with each

photoresist and should be consulted when developing a more detailed or sensitive

process ﬂow.

9.2.5.4 Transfer Pattern

The patterned photoresist layer covering the surface of a substrate contains areas that

are “open” (i.e., where the photoresist has been removed). At this point material can

either be added to or removed from the substrate surface within these open areas.

Removing material is called etching. Wet etching techniques are presented in

Chapter 8. Dry (plasma) etching techniques are presented in Chapter 7. Etching

processes rely on the photoresist having a much slower etch rate than the substrate

surface. The ﬁdelity of the pattern transfer as a result of the etching process depends

both on how directional the etch process is and on the proﬁle of the photoresist side-

wall at the edge of the open areas. If the photoresist sidewall is perfectly vertical

and the etch rate in the horizontal direction is zero (i.e., perfectly anisotropic) then

the pattern developed into the photoresist layer will be perfectly transferred into the

substrate surface due to the etching process. Any deviations in the perfection of the

photoresist sidewall will correspondingly result in removing more substrate mate-

rial than desired. Etch rates that are not perfectly anisotropic will either produce less

than vertical sidewall proﬁles in the substrate (sidewall angle less than 90

◦

), under-

cut beyond the edge of the photoresist opening (sidewall angle greater than 90

◦

)

or over etching (features in the substrate surface that are larger than the openings

deﬁning the photoresist pattern).

Adding material is referred to as deposition. Deposition of semiconductor

and dielectric materials is presented in Chapter 3 and deposition of metals in

692 D.R. Hines et al.

Chapter 4. Typically material is added over the entire substrate thus covering the

photoresist surface and the substrate surface that is accessible in the open areas of

the patterned photoresist layer. The process of photoresist depositing and develop-

ing can leave a thin invisible residue on the substrate surface in the open areas of

the photoresist pattern. This residue can cause a decrease in t he adhesion of the

deposited layer to the substrate surface and should be removed prior to deposition.

This residue layer is typically removed by performing a low-power oxygen plasma

etch (sometimes referred to as a descum) just prior to materials deposition.

9.2.5.5 Remove Photoresist

Once deposition is completed, the unwanted material has to be removed from the

surface of the photoresist and also the photoresist removed. For a method that

includes an additive process, this is accomplished all in one step called “liftoff”.

Liftoff is performed by submerging the substrate (covered with the photoresist and

deposited layers) into a solvent bath. The solvent dissolves the photoresist l ayer

thus causing the deposited layer on the photoresist surface to ﬂoat away. Lift-off

processes can be slow because the photoresist has very little surface area exposed

to the solvent. The processing time can be decreased by increasing the temperature

of the solvent bath. Also applying the solvent in the form of an aerosol spray can

help the liftoff process. For this to work properly the deposited layer must be dis-

continuous at the edges of the opening in the patterned photoresist. Therefore the

photoresist sidewall proﬁle should be undercut. For this reason, negative photore-

sists or positive photoresists with image reversal work best for liftoff applications

inasmuch as they inherently have undercut proﬁles. This is illustrated in Fig. 9.21.

For processes that involve etching rather than deposition, only the photoresist

layer needs to be removed after substrate etching. Photoresist is typically removed

in one of two ways: either using the appropriate solvent or using plasma etching

Fig. 9.21 Illustration of liftoff with positive and negative resists

9 MEMS Lithography and Micromachining Techniques 693

(sometimes referred to as ashing). Acetone is the most common solvent for photore-

sist in the cleanroom and is usually sufﬁcient to clean the surface. In some cases,

the photoresist can be more difﬁcult to remove, such as after successive baking steps

or etching steps, and for these cases the photoresist manufacturers provide speciﬁc

photoresist solvents, usually called “strippers,” or “resist removers.” These solvents

tend to be much more aggressive than acetone, and may damage other materials on

the surface of the wafer. Care must be taken when using these solutions to remove

photoresist.

Plasma etching to remove photoresist is more s traightforward. “Ashing” con-

sists of an O

plasma that chemically reacts with and consumes organic matter on

the wafer, and thus the photoresist. In some cases an argon plasma can be used to

remove photoresist, but in this case the argon plasma merely physically sputters the

photoresist material away and can damage other materials on the wafer. In many

situations ashing is preferred over some of the more aggressive solvent options, but

it is noted that the plasma techniques do not always remove photoresist, such as is

the case with some negative resist formulations that are less reactive.

9.3 Grayscale Lithography

UV lithography processing methods presented in the previous section strictly per-

tain to high-resolution, 2-D patterning on ﬂat substrates. Great effort has gone

into producing near-vertical sidewalls for etching applications and slightly under-

cut sidewalls for liftoff applications. This is ﬁne for the semiconductor industry,

however, there are MEMS-based applications where precise control over the side-

wall proﬁle of the photoresist and subsequent etch processing is highly desirable.

For example, applications in the area of optics [48–50], microﬂuidics [51], and pho-

tonic crystals [52] require the ability to apply 3-D fabrication methods to MEMS

substrates.

Two-dimensional lithography photomasks are designed to either block or trans-

mit light. The resolutions of these masks are diffraction limited. The primary goal

is to completely expose and subsequently remove all the photoresist from spe-

ciﬁc area(s) on the coated substrate. In 3-D lithography applications, the goal is

to pattern the photoresist such that the patterned photoresist layer is comprised of

areas containing variable thickness proﬁles. To accomplish this, both the photomask

and the photoresist properties can be manipulated. For the photomask [53–62], the

opaque areas that deﬁne the pattern can be modiﬁed to be partially transparent. An

illustration of this is shown in Fig. 9.22.

There are a number of methods that can be used to achieve this. One method

uses commercially available high energy beam sensitive (HEBS) glass that uses the

beam-induced reduction of silver ions in a silver–alkali–halide material to produce

the darkening of the glass via coloring specks of silver atoms. This directly modu-

lates the opacity of the photomask with no graininess. HEBS glass mask resolution

is limited only by the beam writers, and examples of sub-0.25 µm features have