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

Подождите немного. Документ загружается.

704 D.R. Hines et al.

X-ray radiation

Proximity gap

Fig. 9.31 X-ray lithography process. Steps 4–6 are unique to the LIGA micromachining method.

(1) Mask is fabricated with transparent membrane and metal absorbing pattern and brought within a

proximity to a conductive substrate coated with PMMA or other thick X-ray photoresist. (2) X-ray

radiation exposes PMMA resist. (3) Development of PMMA. (4) Metal is plated into the PMMA

mold (nickel, gold, copper). (5) Wafer is lapped to remove the metal cap; for mold fabrication, this

step may be omitted. (6) Substrate and PMMA are removed and the metal mold/part is ready for

additional processing or assembly

and injection techniques that can be used for the ﬁnal LIGA fabricated structures

are explored in more detail in Chapters 4 and 5.

9.5 Direct-Write Lithography

The lithographic processes discussed above include the use of a mask for patterning

a layer of resist. It is also possible to pattern the resist layer directly by forming

the source into a narrow beam and scanning it across the surface of the resist. By

blocking the beam or scanning it in a desired fashion, the resist layer can be exposed

pixel-by-pixel. After exposure, the r esist layer can then be further processed just as

it would have been had it been exposed through a mask in the photolithographic

method. Electron beams, ion beams, and laser beams can all be used in such direct-

write techniques.

9.5.1 E-Beam Lithography

Electron-beam (e-beam) lithography uses electrons to expose a resist layer. Like

photons in photolithography, the solubility of a resist layer can be changed due to

interactions with electrons. By scanning the electron beam across the surface of the

9 MEMS Lithography and Micromachining Techniques 705

resist layer in a fashion almost identical to scanning electron microscopy (SEM),

the resist layer can be patterned. In an SEM, magnetic or electrostatic deﬂectors are

used to scan or raster the beam across the surface of the resist layer. Added to this

is the means to blank the beam by inserting electrostatic plates into the column that

can deﬂect the beam so that desired areas (or pixels) are left unexposed. Dedicated

systems that have been designed solely for e-beam lithography can provide state-

of-the-art capabilities. However, in many research and development labs, e-beam

lithography is performed using an SEM ﬁtted with a beam blanker controlled by

a separate computer running commercially available SEM lithography software. A

diagram of the scanning and beam blanking of an SEM-based e-beam lithography

system is shown in Fig. 9.32.

Fig. 9.32 Illustration of an e-beam lithography system (Reprinted by permission from [90])

For a Gaussian beam focused using magnetic lenses, a resolution of 10 nm

or less is possible with e-beam lithography. The exposure speed is too slow,

however, for large-scale production and is therefore typically used for mask fab-

rication, prototyping, or research and development applications. For example,

photomasks and nanoimprint lithography templates are routinely fabricated using

e-beam lithography due to the high resolution available.

Depending on the magniﬁcation setting of the microscope during writing, the

electron beam can only be scanned over a ﬁnite area (e.g., a 100 × 100 µm area).

To pattern larger areas, exposed areas can be “stitched” together by moving the

substrate to a new area and continuing the exposure. This is similar to the action of

a stepper but with e-beam, the subsequent exposures do not have to be identical and

can be connected.

706 D.R. Hines et al.

E-beam lithography is capable of producing high-resolution patterns, thus it is

also well suited for NEMS applications. Many NEMS devices incorporate mechani-

cal resonators, cantilevers, and doubly clamped beams. Applications involving mass

sensing with cantilevers have a mass in the ictogram and mass detection capabil-

ity in the attogram [91] and zeptogram [92] range. Ekinci and Brugger [93]give

an account of fabrication of NEMS cantilevers from Si, GaAs, AlGaAs, AlN, SiN,

SiC, and nanocrystalline diamond all fabricated using e-beam lithography. The res-

olution achievable with e-beam lithography is limited by machine stability, beam

diameter, and proximity effects. Yamazaki [90] gives an account of several factors

associated with machine stability that can affect pattern resolution.

As the electron beam interacts with the resist layer and the underlying substrate,

electrons can be scattered or absorbed. The scattered electrons can contribute to

unwanted exposure in nearby areas of the photoresist layer. This type of exposure

error is referred to as the “proximity effect.” A Monte Carlo simulation of electron

scattering for PMMA resist on a silicon substrate is shown in Fig. 9.33 [94]. Three

main types of electron scattering are important in e-beam lithography: forward

scattering, backscattering, and secondary electron emission. Forward scattered elec-

trons can be minimized by using thin resist layers and high acceleration voltages.

Backscattered electrons are largely responsible for proximity effects. Secondary

electrons are a cascade of low voltage electrons that are created from the interaction

of primary beam electrons with the resist or substrate materials. Secondary electrons

produced within the resist layer account for the bulk of the resist exposure dose and

can cause beam broadening on the order of 10 nm. Because secondary electrons

have low energy, they travel only a short distance within the resist and thus do not

contribute much to proximity effects.

The proximity effect is highlighted in Fig. 9.34a for an uncorrected exposure.

Note that larger and more closely spaced features are more fully developed (less

Fig. 9.33 Simulation trajectories for electrons in PMMA ﬁlm on Si substrate (Reprinted with

9 MEMS Lithography and Micromachining Techniques 707

(c)

(a)

(b)

Fig. 9.34 Examples of proximity effect corrections (a) uncorrected exposure (noted by residual

resist in patterned areas), (b) corrected exposure using dose modulation and (c) schematic illustrat-

American Vacuum Society)

residual resist). To some extent, proximity effects can be minimized by assigning

different dose rates to different sized features. Also ﬁlm thickness and beam energy

can be adjusted to minimize beam scattering under certain controlled settings. The

most common way to correct proximity effects is to modulate the dose rate. There

are several dose modulation techniques [95, 96] based on calculations that solve

the shape-to-shape interactions. If the dose for a large-area feature is normalized to

unity, then smaller and isolated features are assigned a somewhat larger dose. The

result of the dose modulation proximity effect corrections is shown in Fig. 9.34b.

An alternative approach to proximity effect corrections is to bias the patterning [95,

96]. As illustrated in Fig. 9.34c, pattern biasing compensates for the overdosing of

closely spaced features by slightly reducing the size of the patterns. This is more

easily implemented within an e-beam writing system but does not have the dynamic

range of dose modulation.

Polymethylmethacrylate (PMMA) is a standard positive e-beam resist. Other

resist materials include poly(2,2,2-triﬂuoroethyl--chloroacrylate) (EBR-9) which

is an acrylate-based resist that is commonly used in mask writing applications

and is ten times faster but also has ten times lower resolution t han PMMA,

and poly(butene-1-sulfone) (PBS) which is a high-speed positive e-beam resist

708 D.R. Hines et al.

commonly used for photo mask patterning. See [98] for a more detailed discus-

sion about different e-beam resists. A case study of PMMA e-beam resist is given

below. For liftoff applications, a bilayer resist is typically applied to the substrate.

For example, the 200-nm layer of methylmethacrylate (MMA) is spin coated onto

the substrate and baked. A subsequent 200-nm layer of PMMA is then spin coated

onto the substrate and baked. The MMA is more sensitive to the e-beam exposure

than PMMA so it dissolves away at a faster rate during development producing an

undercut below the edges of the PMMA features.

Several companies provide design and control software for performing e-beam

lithography with an SEM. One of the most popular systems is the Nanometer Pattern

Generation System (NPGS). Within the NPGS software application, parameters for

controlling the SEM are stored in a run ﬁle and the desired patterns are drawn using

DesignCAD software. Within a DesignCAD ﬁle, color coding is used to designate

different dose exposures for different features. When a speciﬁc DesignCAD ﬁle is

selected within a run ﬁle, separate beam control lists (referred to as layers) appear

for each color. Within a given layer, magniﬁcation, distance between pixels, beam

current, and doses can be speciﬁed. Some of these details are discussed in more

detail in the case study below.

9.5.2 Ion Beam Lithography and Focused Ion Beam (FIB)

Beams of ions can be used to create patterns directly on the surface of a sub-

strate (i.e., no photoresist). A focused beam of ions can be used to selectively

remove or deposit material at a desired location. Whereas photons and electrons

are effective for patterning soft materials, ions are sufﬁciently heavy to allow

patterning/removal of hard materials such as metals, ceramics, and inorganic semi-

conductors. Therefore, a sacriﬁcial resist layer is not needed as the ion beam

can directly pattern (remove material from) the substrate. The removal process is

referred to as sputtering, which is the removal of material by ion bombardment.

This is a purely mechanical process. Popular ion sources include As, Be, Ga, and

Si (for the related process of ion implantation in the semiconductor industry, Ar,

B, and P are commonly used). Various ion-target interactions produce competing

processes such as swelling, deposition, milling implantation, backscattering, and

nuclear reactions. Ion species, incident angle, ion energy, and ion interaction with

the target material are all important interactions. For appropriate ion energies, ion–

target atom collisions can transfer sufﬁcient energy to the atom to overcome binding

energy (3.8 eV for Au and 4.7 eV for Si) resulting in the removal of the atom from

the substrate surface (i.e., sputtering). Most typically, Ga

ions are used for focused

ion beam lithography applications. A common type of source is a liquid metal reser-

voir that feeds metal ions to a sharpened (tungsten) needle. Such a source has a

typical brightness of 10

A/cm

sr. A summary of ion sources and related properties

can be found in Utke, Hoffmann, and Melngailis [99].

A FIB system consists of an ion source, ion optics, beam deﬂectors, and a sub-

strate stage [100]. A review of these components and each of their functions and

9 MEMS Lithography and Micromachining Techniques 709

designs is given by Tseng et al. [101]. The machine is similar to an SEM in both

function and size. Typically the system is housed in a high-vacuum chamber (10

−7

torr) to minimize scattering of ions from the beam. The vacuum chamber column is

typically 30–50 cm tall and 15–20 cm in diameter [99]. The r ange of beam energies

is 1–50 KeV. Typical spot size for these is about 50 nm. As in e-beam lithogra-

phy, the ion beam is scanned across the substrate surface in the desired pattern and

blanked off at points where no materials removal is desired. The amount of mate-

rial removed for a given beam energy and diameter is determined by the time the

beam is held at a speciﬁc point (dwell time) and the size of the beam step to an

adjacent position (pixel spacing (PS)). The roughness of the milled structure can be

minimized by using a smaller PS. For Gaussian beams with a standard deviation of

the Gaussian distribution σ , smooth f eatures have been produced with PS/σ = 1.5

[102].

For a single pass (scan) of the ion beam, a V-shaped channel (line) is produced.

The depth proﬁle of such a line is shown in Fig. 9.35 from Tseng [102].

The proﬁle is notably wider than the beam diameter indicating that the energy of

scattered ions is large enough to cause sputtering. Also redeposition is evident at the

side of the line-depth proﬁle. For precision pattern formation both sputtering and

Fig. 9.35 AFM image of FIB single-line scan and associated height proﬁle scan (Reprinted with

710 D.R. Hines et al.

Fig. 9.36 FIB proﬁle after single and multiple scans with the same total dose (Reprinted with

permission from [103])

redeposition must be controlled. For many features in nano- and microfabrication,

it is desirable to have vertical sidewall proﬁles and ﬂat-bottomed features with high

aspect ratios. To obtain most types of desired patterns with FIB, multiple passes

must be performed. As shown in Fig. 9.36, multiple scans can lead to better feature

proﬁles. Figure 9.36a, b were fabricated using the same total exposure time: for (a)

a single pass was performed at 5 µm/s whereas for (b) 200 passes were performed

at 1 mm/s. The total dose rate was the same for both cases (1.9 × 10

ions/cm

)but

the single-pass feature shows an inclined bottom due to redeposition (keep in mind

that for some applications such a feature proﬁle may have advantages).

Because of the dynamics of redeposition, different scanning histories can pro-

duce dramatically different features. For example, Fig. 9.37 shows the difference

between scanning from edge-to-center as opposed to scanning from center-to-edge.

Review papers covering FIB and associated uses of these tools can be found in [100,

104–117].

9.5.3 Gas-Assisted Electron and Ion Beam Lithography

With the introduction of a precursor gas, both electron and ion beams can be used

to perform either material removal or deposition. The removal process here is beam

induced etching rather than sputtering as was presented above. For electron beam

9 MEMS Lithography and Micromachining Techniques 711

Fig. 9.37 FIB feature resulting in scanning from (a) center-to-edge and from (b) edge-to-center

(Reprinted with permission from [118])

induced etching, the etching process is initiated by an excitation/dissociation of

molecules in the precursor gas. The resulting gas molecules can then chemically

react with atoms on the surface of the substrate forming volatile molecules contain-

ing substrate atoms that are easily removed in vacuum [119]. Precursor gases can

also be added to an ion beam to combine the affect of etching and sputtering.

In addition to gas-assisted milling, electron and ion beams can also be used to

perform gas-assisted deposition of materials [119, 120]. An example of a similar

process is often observed in SEMs with the formation of thin-ﬁlms resulting from

contamination after the electron beam has scanned over a speciﬁc area on the surface

of a sample. Silicone vacuum pump oil was deliberately used to form a polymer

ﬁlm on a substrate in an SEM as one of the ﬁrst demonstrations of this technique

[121, 122]. More recently the process of electron-beam-induced deposition has been

exploited to deposit other desired materials [123–125]. van Dorp and Hagen [126]

list more than a dozen materials and precursor gases that have been used for gas-

assisted e-beam deposition. A compelling example of e-beam-induced deposition is

illustrated using a tungsten precursor gas to fabricate a nanoscale map of the world

onaSi

substrate. The map is 230 nm across and contains height variations that

represent mountain ranges around the world [127].

9.5.4 Dip-Pen Lithography (DPN)

Dip-pen lithography is an extension of scanning probe microscopy and atomic force

microscopy (AFM). Instead of imaging a surface with a probe tip, the probe tip is

used to deliver an “ink” to the substrate surface. An AFM tip is coated with an ink,

and then when brought into proximity to a surface, a meniscus is formed between the

712 D.R. Hines et al.

tip and surface. As the tip is moved with respect to the surface, ink is pulled off the

tip (due to surface tension) and remains on the surface. In this way, sub-50 nm lines

of material can be deposited either sequentially or in parallel (using multiple tips).

Possible inks include small organic molecules, polymers, DNA, proteins, peptides,

colloidal nanoparticles, metal ions, and sols. Both hard and soft substrate materials

can be patterned, including insulating, semiconducting, and metallic substrates. The

process of writing dots and lines is dependent on parameters such as temperature,

humidity, writing speed, and contact force between tip and substrate, in addition to

the physical and chemical properties of both the ink and substrate materials.

Biological samples are also possible with these schemes, rhodamine 6G (R6G)

being a good example [128]. An ultrasharp Si

tip was dipped into a 2 × 10

−5

solution of R6G for 30 s and then dried in N

gas. The pattern was then written on a

mica substrate in a 23

◦

C and 36% humidity air environment. The feature was written

at both 0.01 and 0.04 µm/s. In this case, the 0.01 µm/s drawn feature generated a

smooth and continuous line, whereas the 0.04 µm/s case produced a discontinuous

line.

For the writing of chemical materials, DPN can be used to deposit nanometer-size

features comprised of self-assembled monolayers (SAM). For example, high-

resolution patterns of octadecanethiol (OTS) molecules on an Au ﬁlm [129] and

hexamethyldisilazane (HMDS) molecules on either a Si or a GaAs substrate [130]

can be fabricated. Such SAMs can be used to fabricate high-resolution hydropho-

bic/hydrophilic patterns or chemical etch barriers for further processing. Zhang et al.

[131] demonstrated DPN by writing 16-mercaptohexadecanoic acid (MHA) features

onto a Au-coated Si substrate. The MHA pattern was used as a wet chemistry etch

mask for the Au ﬁlm and then the patterned Au layer was subsequently used as a

reactive ion etch (RIE) mask for the Si.

9.5.5 Direct-Write Laser

Since their advent, the interaction of laser beams with materials has been of prime

interest. Any laser process that modiﬁes, adds, or removes material from a substrate

can be thought of as a form of direct-write laser lithography. This is a maskless

technique that scans the laser beam from point to point. This can allow for a one-

step process that does not need a resist layer, backing, or wet chemistry. It is also not

restricted to 2-D processing on ﬂat surfaces. The laser beam and/or substrate can be

moved in x, y,orz in a manner that can produce a 3-D feature of arbitrary shape. This

can also be performed on arbitrary-shaped substrate surfaces not only ﬂat wafers.

Inasmuch as the laser beam can be programmed to raster in any desirable way,

the need to design and fabricate new or multiple mask sets in eliminated. Several

main components are needed: (1) laser source, (2) optics to shape and deliver the

laser bean to a desired location, (3) substrate mount, and (4) r elative translation

between laser and substrate. Three main classes of laser direct-write lithography

can be enumerated: subtractive, additive, and modiﬁcation.

9 MEMS Lithography and Micromachining Techniques 713

The subtractive mode typically involves photochemical, photothermal, or photo-

physical processes [132] of which ablation, surface cleaning, milling, and drilling

are examples. The modiﬁcation mode typically involves thermal methods causing a

chemical or structural change. For example, thermally induced phase change from

crystalline to amorphous, sintering, and changes in solubility (similar to UV expo-

sure in photolithography but on a point-by-point processing). The addition mode

typically involves the transfer of material from one substrate to a second substrate.

For example, a bilayer ﬁlm on a transparent substrate where the bottom layer is

vaporized by the laser beam thus tearing off the top layer which is then propelled

across the gap to be deposited onto the second substrate. The additive process is

illustrated in Fig. 9.38 and has been used to fabricate a thin-ﬁlm transistor (TFT) by

Blanchet et al. [133].

Fig. 9.38 Illustration of the direct-writing laser additive process to transfer one material onto a

Physics)

Laser-induced chemical-vapor deposition (LCVD) is an example of the modi-

ﬁcation method [134] that can be used to form 3-D structures on a substrate. The

substrate is surrounded by material in the gas phase and the laser beam is focused to

the desired location to induce a chemical reaction t o change the gas into a solid form.

The laser beam is moved around to create the desired structure [135]. Two types of

reactions are generally used, either decomposition reactions or combination reac-

tions. These reactions can be either photolytic or pyrolytic. Duty et al. [135] list

materials that have been deposited using pyrolytic methods. The list includes metals

such as Al, Cu, and Au; semiconductors such as Si and Ge; and ceramics such as

, SiC, and TiN. The list also includes starting reagents along with various pro-

cessing parameters. Figure 9.39 illustrates the LCVD process where a laser beam is

depicted to induce a chemical reaction to convert a gas or liquid phase into a solid as