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

Подождите немного. Документ загружается.

694 D.R. Hines et al.

Fig. 9.22 Comparison between photoresist exposure proﬁle using standard and grayscale pho-

tomasks (Modiﬁed from [63], reprinted with permission)

been written for these photomasks. These masks can also be used in both projection

and contact lithography systems, which makes them more versatile when consider-

ing equipment availability and differing cleanroom capabilities. The cost of HEBS

masks can also be up to one order of magnitude more than standard chrome on

quartz plates, which has signiﬁcantly limited the wide adoption of this technique in

the MEMS community. However, when an extremely smooth 3-D proﬁle is required,

or contact lithography is the only option available, HEBS glass is an excellent choice

for grayscale lithography. HEBS glass is often used for lenses and high-density lens

arrays due to the smoothness achieved in the ﬁnal photoresist proﬁle. Figure 9.23

shows an example of patterned silicon structures using HEBS glass lithography.

As already mentioned, HEBS glass masks are fabricated using an electron beam

to initiate the chemical reduction of the silver-containing doped glass. The opacity

of the mask is determined by the electron beam parameters with the dose prac-

tically controlled by changing the dwell time and the energy of the beams. The

9 MEMS Lithography and Micromachining Techniques 695

Fig. 9.23 Silicon proﬁle

generated using a HEBS glass

mask and DRIE etching

(Reprinted with permission,

Lawrence Livermore National

Laboratory, https://www-eng.

llnl.gov/mic_nano/

mic_nano_MEMS.html)

ability to ﬁne-tune the dose allows for a very large number of gray levels, over

500 for most cases. Unlike a chrome mask, however, the dark areas of the mask

never become 100% opaque. Fabrication of a mask is performed exclusively through

Canyon Materials, Inc. (CMI) who provide detailed instructions of how to struc-

ture the mask design ﬁle for them to perform the electron beam exposure of the

mask. Inasmuch as many processing variations exist for the photoresists and opti-

cal lithography sources, CMI provides standardized calibration masks to establish

proper photoresist exposure times, thicknesses, and development procedures. There

are a number of publications available showcasing the capabilities of these masks

[64–66].

There is another commonly used method that has been more widely adopted,

utilizing a variable transmission mask just like the HEBS glass, but with standard

mask fabrication techniques. This is accomplished by pixelating the chrome pat-

terns on the masks so as to control the amount of light transmitted. This technique

necessitates the use of a stepper lithography system, which can be a problem if such

a system is not readily available. As discussed below, the limitations on pixelation

result in a ﬁnite number of transmission steps (called gray levels) between 0 and

100% transmission. For the photoresist, both sensitivity to UV light exposure and

696 D.R. Hines et al.

selectivity to the subsequent substrate etching will be important for utilizing all the

gray levels. For example, if the photoresist is too sensitive to UV light, light from

the ﬁrst grayscale level will fully expose the photoresist at the substrate surface

before light from the last grayscale level has a chance to expose the top surface

of the photoresist. If the photoresist is removed too fast during the substrate etch

then the photoresist will be consumed before the desired etch depth can be reached.

Likewise, if the photoresist is not consumed at all then none of the grayscale levels

can be transferred into the substrate etch.

9.3.1 Photomask Pixelation

To establish partial transmission of UV light during the exposure step in the pho-

tolithographic process, the opaque part of the photomask can be pixelated as shown

in Fig. 9.24. The pixels need to be designed so that they modulate the transmission of

UV light without being directly imaged onto the surface of the photoresist. For this

reason the pixel dimensions need to be below the diffraction limit of an optical pro-

jection s ystem. When making a particular mask design, the spacing (pitch) and the

pixel dimensions are varied to create different levels of transmission, generating the

gray levels in the photoresist pattern. In this way, the minimum dimensions available

in the design of a pixelated gray mask are limited by the projection lithography sys-

tem’s resolving power, and the resolution of the mask writing system. In addition,

less pattern control is available when making complex 3-D structures inasmuch as

pixel size and pixel spacing are all different based upon the particular transmission

or gray level being expressed. Minimum dimensions for singular gray levels are then

ﬁxed by this pixel size and pitch limitation. Pixelated grayscale masks are in gen-

eral best suited for larger patterns, or for situations where mask cost is a restrictive

parameter. A more detailed discussion of the mask design and pixelation is included

in the case study section of this chapter (Case Study 10).

Pixels

Pitch

Pixel ArrayMask

Fig. 9.24 Pixelation of a gray-scale photomask (Reprinted with permission from [67])

9 MEMS Lithography and Micromachining Techniques 697

9.3.2 Photoresist Properties for Grayscale Lithography

9.3.2.1 Contrast and Thickness

A photoresist contrast curve is shown in Fig. 9.25, with thickness plotted as a func-

tion of the logarithm of the UV light exposure. Two exposure points are noted: E

is the minimum exposure required for the photoresist to react to light and E

is the

minimum exposure required to completely remove (or clear) the photoresist from

the substrate surface [68]. The linear slope between these two exposure points is

deﬁned as the “contrast.” A larger photoresist contrast translates to a smaller dif-

ference between the two exposure endpoints and to a smaller processing window.

For grayscale lithography, larger processing windows are desirable so that good

differentiation between exposure levels can be realized. Therefore, low-contrast

photoresists are typically used for grayscale lithography.

Fig. 9.25 UV light exposure

proﬁle as a function of

photoresist thickness

(Reprinted with permission

American Institute of

Physics)

In addition to contrast, the thickness of the photoresist layer is also important.

The thickness should be chosen so that E

is the exposure produced by the largest

grayscale level and that it occurs at the top of the photoresist surface precisely when

is the exposure produced by the smallest grayscale level.

9.3.2.2 Exposure and Developing Times

The exposure time is determined as the time needed to deliver an exposure equal to

. Theoretically, determining the exposure time is straightforward but in practice

it can be problematic. Temperature, humidity, variations in photoresist thickness,

intensity ﬂuctuations of the stepper, and so on can all affect both the exposure and

the developing time. Therefore, it is typical to empirically determine these times by

performing test exposures using a calibrated mask. Overexposing or underexposing

will leave a thickness bias on the whole photoresist structure that will interfere with

698 D.R. Hines et al.

Fig. 9.26 UV exposure

levels for gray-scale

lithography (Modiﬁed from

[67], reprinted with

permission)

the proper transfer into the substrate beneath, creating a disparity in the ﬁnal device

thickness, as illustrated in Fig. 9.26.

9.3.2.3 Etch Selectivity

Photoresist etch selectivity is deﬁned as the ratio of the substrate etch rate to the

photoresist etch rate for a given set of etching conditions. A high etch selectivity

is usually desirable for standard photolithography applications where the original

ﬁdelity of the photoresist pattern is designed to survive the substrate etch, thus pro-

ducing a perfect replica of the photoresist pattern etched into the surface of the

substrate. For grayscale lithography, the photoresist layer is designed to be con-

sumed during the substrate etch. Often times varying levels of O

gas are introduced

during the etch to speed up the photoresist consumption rate. Using this method, it is

easier to ﬁne-tune the resulting selectivity rate. The resulting pattern etched into the

substrate reﬂects the thickness proﬁle of the original photoresist pattern. Each gray

level developed into the photoresist corresponds to a speciﬁc etch depth into the sub-

strate surface which is determined by the etch selectivity and the starting thickness

of the patterned photoresist layer. The difference in ﬁnal structure transfer is illus-

trated in Fig. 9.27 . Photoresist thicknesses of 6–10 µm, when combined with etch

selectivity between 40 and 100, translate into substrate etch depths ranging from

tens to hundreds of micrometers. An example of a fully transferred pattern using

HEBS glass has already been shown above (Fig. 9.24), and a typical transferred pat-

tern using a pixelated mask is shown in Fig. 9.28. More detailed discussion of the

silicon etching process is given in Chapter 7.

9.4 X-Ray Lithography

X-ray lithography was originally developed as an attempt to achieve smaller device

geometries by using the very short wavelength of X-rays (∼0.7–1 nm optimally), the

ability to have highly collimated X-ray sources, and the relatively s mall scattering

9 MEMS Lithography and Micromachining Techniques 699

Fig. 9.27 Etch selectivity

proﬁles for gray-scale

lithography (Modiﬁed from

[67], reprinted with

permission)

Fig. 9.28 Silicon Fresnel

lens patterned using a

pixelated grayscale mask and

etched using DRIE

(Reprinted with permission

from [69])

of X-rays through common materials. Theoretically, 20 nm resolution features can

be deﬁned, but in practice it is more common to see features in the 50 nm range [70].

This is due to a number of challenges encountered in the lithography process, such

as synchrotron X-ray wavelength spectral broadening, mask–substrate distance, and

mask distortions [71].

X-ray lithography follows a very similar process to that of UV lithography: a

pattern is transferred into an optically sensitive resist material (in this case X-ray

sensitive resist) by illumination through a mask onto the resist. Although there was

an initial interest in X-ray lithography for VLSI design, most X-ray lithography is

used today for the fabrication of high-aspect-ratio molds in a process commonly

referred to as LIGA (Lithographie, Galvanoformung, Abformung)[72]. LIGA was

initially used in the fabrication of nozzles for uranium enrichment (Fig. 9.29), with

700 D.R. Hines et al.

Fig. 9.29 LIGA-fabricated

uranium enrichment nozzle

(Reprinted with permission

from [73])

many other applications being developed after the initial demonstration of this tech-

nology [73]. LIGA takes advantage of the very high aspect ratios (100:1) achievable

through X-ray lithography and uses the resulting polymer resist structures as molds

for electroplating (metals), or other molding processes. There have also been exam-

ples of SLIGA performed which incorporates sacriﬁcial layers into the standard

LIGA process thus allowing for the creation of freestanding and moveable struc-

tures directly on the substrate [74]. X-ray lithography is used in MEMS processes

and devices primarily in this context [75, 76].

The primary challenge for X-ray lithography is in the required light sources for

highly collimated X-ray beams. These types of sources are generally much more

complex than the typical UV discharge lamps used in UV lithography. In addition,

due to their very high energy, X-rays cannot be focused and guided using tradi-

tional lenses and mirrors as in UV lithography. The synchrotron is one of the most

important enabling tools of high-aspect-ratio X-ray lithography, capable of emitting

high-power, highly collimated radiation. This high collimation permits relatively

large distances between the mask and the substrate without the penumbral blurring

that occurs from other X-ray sources, and also enables the high aspect ratios in

exposed photoresist, which is beneﬁcial to the development of LIGA.

In the electron storage ring or synchrotron, electrons follow a circular path. The

radial acceleration of the electrons causes electromagnetic radiation to be emitted

forward. The radiation is thus strongly collimated in the forward direction of the

moving electrons and is essentially parallel. These radiation sources also provide

higher ﬂux densities, which enable shorter exposure times. Photon energies for a

LIGA exposure are approximately distributed between 2.5 and 15 keV.

9.4.1 X-Ray Masks

X-ray masks are one of the bottlenecks in the commercial development of LIGA

and X-ray lithography [77], because there are a number of materials used for

this process, depending on the particular research group’s fabrication facilities and

9 MEMS Lithography and Micromachining Techniques 701

capabilities. The general concept employed in X-ray mask generation is to have

a highly transparent material for the ﬁeld of the mask (low-Z materials), and an

X-ray-absorbent material to deﬁne the features on the mask (high-Z materials). This

is commonly done with a substrate material of graphite or vitreous carbon, or some-

times silicon nitride, silicon carbide, or diamond membranes [78, 79]. The latter

are not as readily preferred as the membranes required for the masks in these cases

are very fragile. Beryllium substrates are very attractive mask materials due to their

very high X-ray transparency and the ability to make them much thicker and more

stable as compared to membrane approaches. The patterned absorbing features are

typically X-ray absorbing metals such as tin, copper, gold, or lead [76].

One of the challenges to making these large area masks is the ability to maintain

low distortion in the ﬁnalized masking image. Due to the 1:1 mask transfer, the accu-

racy of the mask features is paramount in importance. These distortions are caused

by a number of factors, with a major one being ﬁlm stresses in the membrane materi-

als. More masks are being made out of SiC due to this distortion issue. Research has

also been undertaken in an attempt to predict these distortions and counteract them

with the design of the mask itself, similar to what occurs in ultra-high-resolution

masks for UV lithography. Even the mounting of the mask contributes to distor-

tions in the mask pattern due to stresses; most masks are mounted in metal rings to

prevent some of these adverse effects.

The deﬁnition of the mask features can be done a number of ways and greatly

depends on the required resolution of the mask patterns. Direct-write methods such

as electron beam lithography are used to achieve the most accurate features for the

mask, but this carries the disadvantage of being very time consuming for the large-

area masks in X-ray lithography. Other methods (such as described in this chapter)

are also employed to pattern the X-ray masks with their own varying lithographic

tolerances.

Fabrication of these masks is often quite involved, in an attempt to reduce ﬁlm

stresses and achieve high-resolution features. Typically, the mask fabrication begins

with a silicon wafer that is coated with the membrane material (silicon nitride, sili-

con carbide). Two methods are commonly used to pattern the mask absorber metal

(tungsten, titanium, gold, etc.). The metal can be blanket deposited on the back-

side of this coated wafer, and then patterned using subtractive processes (dry or wet

etching). Other methods employ thicker photoresists and rely on additive processes

such as plating to deposit a thicker absorber coating which is required for lower

Z materials such as gold to achieve good exposure contrast. This can be done a

number of ways, one of which involves performing an X-ray lithography step on a

thinner coating of X-ray photoresist to f orm a plating mold for the ﬁnal mask, but

can be performed with normal thick photoresist plating mold technologies as well.

The thickness of this absorbing pattern can vary from a few microns to up to 50 µm,

and depends on the thickness of the photoresist to be exposed, both of which are

directly related to the required X-ray energy [76].

The completed mask is normally placed in a glass or metal ring at this point

in the process to provide the wafer with stability and strength. After this step, the

membrane is released by backside etching of the silicon wafer. A photograph of a

ﬁnalized X-ray mask is shown in Fig. 9.30.

702 D.R. Hines et al.

Fig. 9.30 An example of an

X-ray mask made from SiC

(Reprinted with permission

from [80])

9.4.2 X-Ray Photoresists

X-ray photoresists function very similarly to the resists used in e-beam lithography

(see Section 9.5.1), and thus most e-beam resists can be used for X-ray lithography;

poly(methyl methacrylate) (PMMA) is most commonly used for X-ray lithography,

however, SU-8 is also a possible choice [81]. PMMA is exposed through the gen-

eration of Auger and photo electrons by the X-rays entering the photoresist. These

interactions chemically alter the resist polymer chains through scission, reducing

their order from approximately 1,000,000 to 1000. This is what allows the polymer

to be dissolved by developers that attack the lower-order polymer chains [82].

Photoresists used in X-ray lithography for MEMS applications are generally

applied in thick layers to take advantage of the high aspect ratios achievable, par-

ticularly for LIGA. The typical choice PMMA is applied to the substrate for some

of the thinner applications by way of spinning, or more probably by a glue-down

process in which a precast sheet of PMMA is attached to the substrate material

[82]. This second option allows for much thicker molding layers and also prevents

the buildup of stress which can be caused by casting the photoresist directly on the

substrates. The PMMA is then milled down to the precise height before X-ray expo-

sure. In some cases, after exposure a second PMMA layer is glued to the surface for

a second exposure, facilitating the creation of even higher aspect ratios [83].

In the case of ultra-high- and very-high-resolution X-ray lithography, the thick-

ness of the photoresist is somewhat limited, with thicknesses of typically 5–50 µm.

These thicknesses have been found to be the threshold for adverse optical effects

such as diffraction and scattering which occur because of mask dimensions and

X-ray energy, respectively [84].

9.4.3 Exposure

Due to the fragility of X-ray masks fabricated on thin membranes, contact lithogra-

phy is not possible without damaging the masks. In addition, the high energy of the

9 MEMS Lithography and Micromachining Techniques 703

X-rays prevents the use of a typical projection lithography system where patterns

can be focused onto the surface of the substrate and reduced in scale. As neces-

sitated by these conditions, all X-ray masks must be 1:1. Instead of using these

more common lithographic techniques (projection and contact), proximity masking

is used. Divergence and near-ﬁeld diffraction commonly encountered in traditional

proximity UV lithography are not as large a problem for X-ray lithography due to

the highly collimated radiation and the very small wavelength of the X-rays. Mask-

wafer spacing of anywhere from 5 to 50 µm i s typical and varies depending on the

available equipment or the required resolution, as increasing this gap does have a

negative effect on the ﬁnal patterns [70, 82–84].

Exposure times and energies vary greatly based upon the thickness of the pho-

toresist used and the required critical dimension size. X-ray energies for exposure of

high-resolution patterns are typically around 1 keV, whereas a more typical MEMS

application will require X-ray energies of 5–15 keV. A typical dose for PMMA is

4–20 kJ/cm

. Anything higher than this could possibly cause the resist to burn [85].

Exposure times vary, again, based upon the thickness of the resist, but generally are

no more than a few hours. Care must be taken not to perform very long exposures

to prevent mask and substrate heating due to the X-ray exposure, which can cause

misalignment and mask distortions [86]. The time of the exposure plays a large part

in the choice of the masking material used.

Karl Suss has created a commercially oriented X-ray stepper tool that allows a

pattern to be exposed using X-ray lithography [71, 87, 88]. This particular device

operates with a proximity gap of 20–50 µm, and has a scanning speed of 0.5–

20 mm/s. In the literature, custom modiﬁcations to this tool have yielded very small

proximity gaps to approximately 10 µm[70]. The tool still requires a separate

synchrotron radiation source, however. The stepper functions similarly to that of

a standard stepper and can align masks to each other via CCD cameras to facilitate

multilevel X-ray exposures with high pattern alignment accuracy.

9.4.4 Development

For high-aspect-ratio structures the resist–developer system usually has a ratio of

dissolution rates in the exposed and unexposed areas of 1000:1. Developers vary:

GG developer (composed of diethothxy and morpholine) is sometimes used [82];

another standard developer used is a mixture of tetrahydro-1,4-oxazine (20%),

2-aminoethanol-1 (5%), 2-(2-butoxyethoxy)ethanol (60%), and water (15%) [89].

The former provides nearly 100% selectivity and does not induce any additional

stresses or cracks in the exposed resist. Developers vary on a case-by-case basis and

are usually empirically standardized and characterized.

After development, the substrate is rinsed with deionized water and dried either

in a vacuum or by spinning. At this stage, the PMMA structures can be released

as the ﬁnal product or can be used as molds for subsequent metal deposition. An

illustration of the complete process is shown in Fig. 9.31. Discussion of the plating