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

Подождите немного. Документ загружается.

434 S. Tadigadapa and F. Lärmer

respect to etch step pressure, even to values higher than the etch step pressure,

reduces the overall RIE lag of the total process and even inverts it at some stage,

with higher-aspect-ratio trenches (i.e., narrower gaps), etching faster than smaller

aspect ratio trenches (i.e., wider gaps). However, compensation of RIE-lag is at the

expense of a reduced etch rate in the wider trenches, which approaches the low

etch-rate value in the narrow trenches. On the other hand, running a deposition step

pressure lower than the etch step pressure will cause a signiﬁcant RIE-lag of the

total process, however, at the beneﬁt of higher overall etching rates in the wider

trenches. This is the “normal” process regime which is used in most cases, in the

interest of higher rates.

The stability of the polymer deposited onto the etch ﬂoor is an important

weight factor in balancing etch and deposition lags, therefore the amount of

achievable compensation depends also on the substrate temperature. Lowering the

substrate temperature makes the ﬂoor polymer more robust and increases the weight

of the deposition step in overall lag balancing. For this reason, a substrate temper-

ature of 0

◦

C or even lower, instead of the normal 40

◦

C was found beneﬁcial to

achieve lagfree etching over a wide range of aspect ratios [57].

Suppression of Notching at Dielectric Interfaces: As explained at the beginning

of this chapter, sidewall attach at the bottom of the etched features is observed when

the etch terminates on a dielectric layer. This effect arises due to the redirection of

the ions with charged dielectric at the bottom of the etch feature during the overetch

phase [14, 58]. Pulsing the substrate–electrode bias power provides a mechanism for

the discharging of the dielectrics during bias off-periods. Laermer et al. compared

the pulsed biasing for low frequency (380 kHz) and high-frequency (13.56 MHz)

carriers [59]. Notch suppression capabilities were demonstrated on high-aspect-

ratio trenches (1.5 μm wide, 11 μm deep, 16% overetch on SiO

). These results

showed that pulsed LF (40 Hz, 50% duty cycle) and “double-pulsed” RF (100 kHz

at 10% duty cycle + 140 Hz at 50% duty cycle) are equivalent with respect to notch

suppression [59].

7.6 High-Aspect-Ratio Etching of Piezoelectric Materials

Dry etching techniques based upon the use of reactive plasma offer very attractive

alternatives to the wet patterning techniques for piezoelectrics. Using a Langmuir

probe, Steinbruchel made pioneering measurements for identifying the role of ions

in reactive plasma etching [60]. Since then, several gas species have been used for

the etching of various piezoelectric materials. In this section we present etching

processes suitable for several piezoelectric materials.

7.6.1 Case Study: High-Aspect-Ratio Etching of Glass

(Pyrex



) and Quartz

Silicon dioxide in its crystalline form (quartz) as well as its amorphous form (glass)

is ﬁnding increasing applications in microsystems, as active resonator structure as

7 Dry Etching for Micromachining Applications 435

well as passive support and packaging components. Recently Pyrex



and quartz

substrates have been etched with very high aspect ratios and very high surface

smoothness using SF

and Ar/Xe gases [61–63]. The main difference between the

etch processes developed for quartz micromachining processes and SiO

etch pro-

cesses described in the earlier section pertain to the desired high etch rates and

high-aspect-ratio etching of quartz. In this context the process relies upon ion bom-

bardment to accelerate the etching process and ﬂuorine-based gases are used to

provide the reactive component for etching. The use of heavier Xe helps reduce

the redeposition and more effectively removes any nonvolatile residues resulting in

smoother surfaces with an average surface roughness of ∼2nm.

An inductively coupled plasma system is once again well suited for this applica-

tion where the source generator driving the inductor coils creates the high density

plasma and ion bombardment is independently controlled using a separate sub-

strate RF generator. This enables excellent control over plasma density and kinetic

energy of etchant ions. Low processing pressure and high plasma density, essen-

tially resulting in high ionic current and greater radical ﬂux density, improve the

mass transfer rates of the r eactant gases and the etch products in addition to being

instrumental in the removal of nonvolatile residues. Nonvolatile residues are typi-

cally generated from the masking materials, the substrate holder, r eaction chamber

walls, or as reaction by-products. These result in micromasking causing high surface

roughness (often referred to as grass), microtrenching, and formation of plateaulike

structures. Additionally, the increased mean free path at low pressures improves the

anisotropy of the etched features by minimizing the randomizing collisions between

the radicals, ions, and other plasma species.

In the case of deep reactive ion etching of silicon dioxide (quartz or Pyrex



high Ar:SF

ratio is required to maintain l ow RMS surface roughness. Figure 7.15

shows the dependence of the etch rate and RMS surface roughness as a function

of substrate RF power, chamber pressure, Ar, and SF

ﬂow rate. In all cases the

pressure in the chamber was maintained at 0.26 Pa throughout the ﬂow ranges. The

ICP source power was 2000 W and a substrate bias power of 475 W (Bias voltage

of 80 V) was used in generating these results. From the graphs it can be seen that

the best surface roughness of ∼2 nm is obtained at high Ar ﬂow rates, low chamber

pressure, and high substrate power, corresponding to conditions dominated by phys-

ical sputtering of the material. The etch rate can be increased by increasing the SF

ﬂow rate from 5 to 50 sccm from 0.54 to 0.74 μm/min, however, the surface rough-

ness was found to degrade under these conditions to >100 nm. Pulse electroplated

nickel was used as the etch mask layer and a selectivity of ∼25:1 was obtained for

SiO

etching under these conditions.

Figure 7.16 shows an SEM of a high-aspect-ratio feature etched in quartz using

these conditions. Similar results were obtained by Li et al. while etching SiO

using

Xe instead of Ar. The higher sputter yield of Xe gave a lower RMS surface rough-

ness value as compared to Ar for the same mole fraction of the inert gas in SF

Although silicone grease or a small drop of Fomblin



oil can be used for mounting

the quartz/glass substrates onto a 4 in. silicon carrier wafer, these materials cannot

withstand the long process times and can leave the backside of the sample with

436 S. Tadigadapa and F. Lärmer

hard to remove, stubborn residues. Furthermore, these mounting materials do not

provide a reliable and uniform thermal contact between the carrier wafer and the

sample throughout the entire etch process. In order to avoid these problems, indium

solder can be used for mounting the sample directly onto a silicon wafer. However,

the mounting side of the SiO

sample needs to be coated with 20/80 nm of Cr/Au

to provide a surface to which the solder can adhere. Of course if the sample is large

enough it can be directly mechanically clamped or an electrostatic chuck can be

used for the mounting of the sample. In all cases the backside of the chuck/substrate

is cooled using helium gas maintained at the desired temperature.

Pressure (Pa)

Etch Rate (m/min)

rms Surface Roughness (nm)

0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5

2.75

0.375

0.4

160

0.425

240

0.45

320

0.475

400

0.5

480

0.525

560

0.55

Etch Rate

rms Roughness

Substrate Power (W)

Etch Rate (m/min)

rms Surface Roughness (nm)

100 150 200 250 300 350 400 450 500

–20

0.1

0.15

0.2

0.52

0.3

0.53

100

0.4

210

0.54

140

0.5

160

0.55

Etch Rate

rms Roughness

Fig. 7.15 Etch rate and surface roughness dependence of Pyrex



7740 glass dependence as a

function of: (a) substrate power in watts, (b) chamber pressure in Pa, (c) argon ﬂow rate (sccm),

and (d) SF6 ﬂow rate (sccm). In the above graphs, all other etch parameters except the vari-

able parameter are held constant at the following values: ICP power = 2 kW, substrate power =

475 W, chamber pressure = 0.26 Pa, argon ﬂow rate = 50 sccm, SF6ﬂow rate = 5 sccm, and

substrate temperature − 20

◦

7 Dry Etching for Micromachining Applications 437

Argon Flow Rate (sccm)

Etch Rate (m/min)

rms Surface Roughness (nm)

0 5 10 15 20 25 30 35 40 45 50

0.15

0.3

0.45

0.6

0.75

Etch rate

rms Roughness

Flow Rate (sccm)

Etch Rate (m/min)

rms Surface Roughness (nm)

0 6 12 18 24 30 36 42 48 54 60

0.52

0.56

0.6

0.64

0.68

100

0.72

1200.76

Etch Rate

rms Roughness

Fig. 7.15 (continued)

Fig. 7.16 SEM picture of a

60 μm deep etched feature in

quartz. The sidewalls have

been found to be roughened

due to redeposition of

nonvolatile etch products and

mask erosion. Trenching is

also observed at sharp corners

in the feature, clearly

indicating the dominant role

of the ion-induced physical

sputtering in the process [64]

(Used with permission,

438 S. Tadigadapa and F. Lärmer

7.6.2 High-Aspect-Ratio Etching of Piezoelectric Materials

As seen in the case of SiO

, most piezoelectric materials are primarily etched via

physical sputtering with some assistance from chemical species in the process. The

role of the chemical species is clearly observed in the comparative etching exper-

iments of SiO

(quartz) and lithium niobate/tantalate [65]. In these experiments it

is clearly observed that increasing the ratio of CF

to CHF

has no effect on the

etch rate of lithium niobate/tantalate whereas it signiﬁcantly increased the etch rate

of SiO

where a regime of ion-enhanced chemical etching is obtained. Because the

ﬂuorides of lithium and tantalum oxide and niobium oxide are all nonvolatile, the

reactive component of etching is an insigniﬁcant part of the etching process, which

simply relies on the physical sputtering of the surface atoms due to the bombard-

ment of energetic ions. Most piezoelectric etching recipes use inert gases as part of

the etch gas composition [66]. This is mainly because inert gases provide plasmas

with higher ion densities and ionization in comparison to the more electronegative

chemical compounds (etch gases) and the higher sputter yield of inert gases is also

much higher than other elements/molecules.

Figure 7.17 shows the SEM of lithium tantalate etched to a depth of ∼8 μmatan

etch rate of 0.2 μm/min. Surface roughness of <5 nm was obtained for this sample.

The etching was performed in an inductively coupled plasma etcher with an SF

/Ar

ﬂow rate ratio of 10/50 sccm and a substrate bias voltage of 75 V. The sample surface

roughness is critically dependent on chamber pressure with low pressures offering

a lower RMS surface roughness.

Aluminum nitride is typically etched using chlorine chemistries inasmuch as alu-

minum ﬂuoride is an extremely stable and nonvolatile compound [67, 68]. Shul et al.

have studied the comparative etching of GaN, InN, and AlN in chlorine and BCl

plasmas and were able to achieve etch rates of 0.23 μm/min for AlN in BCl

plasma

with a small percentage additions of Ar/N

[69]. The etching characteristics of ZnO

and etch selectivities of ZnO to SiO

in CF

/Ar, Cl

/Ar, and BCl

/Ar plasma are

reported by Woo et al. [70]. High etch rates of 0.12 μm/min have been reported. In

several of the optimization reports of the etch processes, it was important to ensure

that the photoresist mask was able to survive the piezoelectric material patterning.

Fig. 7.17 SEM picture

showing a high aspect ratio,

smooth etching of lithium

tantalate in 10:50 SF

:Ar

plasma. The chamber

pressure was <0.66 Pa, with

an ICP power of 2 kW and a

substrate power of 400 W. A

nickel hard mask was used to

pattern the substrate

7 Dry Etching for Micromachining Applications 439

Often such optimization results in a compromise between the etch rate of the

piezoelectric and mask selectivity because such etches are performed under high

pressure and low substrate bias conditions. In cases where it is necessary to use

a hard mask, typically electroplated nickel is used. Nickel provides a very high

selectivity in ﬂuorine plasmas where the selectivities of 20–30:1 for PZT etching

have been reported [71]. PZT is typically etched in ﬂuorine or chlorine plasma.

Using SF

and Ar several groups have reported high-aspect-ratio etching of PZT.

Using ﬂuorine-based plasma chemistry, a maximum etch rate of 19 μm/hr for PZT-4

and 25 μm/hr for PZT-5A compositions have been reported [72]. This work also

demonstrated a high-aspect-ratio etch (>5:1) on a 3 μm feature size.

Figure 7.18 shows the SEM image of the etched features for PZT. The square

pillars created have a lateral dimension of 15 μm with a gap of 3 μm in between.

The etched depth was ∼15 μm. Almost vertical sidewalls with relatively smooth

surfaces were obtained. 2–5 μm thick nickel on a Cr/Au was used as the hard

mask. Selectivity of ∼25:1 between the nickel hard mask and PZT was obtained.

The availability of reliable, high-throughput, high-aspect-ratio micromachining pro-

cesses has now created new opportunities for realizing novel MEMS devices from

bulk piezoelectric materials such as quartz, lithium tantalate, aluminum nitride,

PZT, and single-crystal PMN-PT. Table 7.4 summarizes the dry etching of various

piezoelectric materials.

Fig. 7.18 SEM image of the etched feature in PZT ceramic substrate with a minimum feature size

Institute of Physics)

440 S. Tadigadapa and F. Lärmer

Table 7.4 Summary of the dry etching characteristics of various piezoelectric materials [73]

Material Etch gas(es)

Pressure

(Pa)

RF frequency

(MHz)/power

(W)

Etch rate

(μm/min) Comments

Quartz [74]CF

355 27 MHz 4 Nonlithographic plasma

conﬁnement method

was used

Quartz [61]SF

/Xe 0.59 13.56/90 0.4 ICP source with 150 W

power was used to

obtain highly smooth

surface

Pyrex 7740/

quartz

[63]

/Ar 0.26 13.56/475 0.54 ICP source with 2 kW

power was used. Highly

smooth surface with

= 1.97 nm. Similar

rates were obtained for

quartz etching as well

LiNbO

LiTaO

[65]

CHF

/CF

6.58 13.56/350 ∼0.01 Etching proceeds mainly

via physical sputtering

AlN [75]Cl

/Ar 0.66 13.56 MHz 0.75 ICP Source with 500 W

power was used.

Etching proceeds by

physical bombardment.

No signiﬁcant etching

was observed below a

threshold substrate

voltage of −50 V. The

paper also reports

etching GaN and

AlGaN. Another good

reference is [76]

AlN [67] BCl

/Cl

/Ar 0.66 13.56 MHz 0.4

ZnO [77]SiCl

/Ar 23 13.56/0.56 W/m

0.027 Additional references on

ZnO etching are

available [70, 78].

Typically etching is

found to proceed via

physical bombardment

ZnO [79]C

/Ar 0.66 13.56/300 0.05

PZT (Bulk)

[71]

PZT [80]

PZT [72]

/80%Ar

:Ar::1:10

0.66

1.97

0.66

13.56/200

13.56/700

13.56/475

0.12

0.143

0.42

In [71] pure SF

gave the

best etch rate but the

angle was shallow

which could be

improvedbyAr

addition but at the cost

of etch rate. The recipe

used in [72]mainly

uses physical

sputtering, however,

aspect ratios of >5:1

were obtained

7 Dry Etching for Micromachining Applications 441

7.7 Etching of Compound Semiconductors

Almost all III-V compound semiconductors and their heterostructures can be etched

in chlorine and bromine plasmas. More recently CH

-based dry etching pro-

cesses have also been explored. The three main considerations in the etching of

III-V semiconductor structures include (i) smoothness of etched features inasmuch

as many of these structures are used in optical devices, (ii) low damage to the

semiconductor layer, and (iii) conﬂicting requirement of high selectivity as well

as equirate etching of the heterostructure layers depending on the particular pro-

cess step. Typically group V halides have very high vapor pressure and therefore

are readily removed in the dry etching process. However, group III halides are not

so volatile. For example, the boiling points of AlCl

is 262

◦

C, GaCl

is 535

◦

GaCl

is 201

◦

C, InCl is 608

◦

C, and InCl

∼600

◦

C. Furthermore, group III ﬂuo-

rides are extremely stable compounds and do not volatilize readily; for example, the

boiling point of AlF

is 1291

◦

C, GaF

is ∼1000

◦

C, and InF

is >1200

◦

C. Thus,

purely chemical etching of GaAs can be done in chlorine plasma but the same pro-

cess does not achieve good etching results for InP due to the involatility of InCl

Furthermore, the addition of ﬂuorine to chlorine plasma can be used to obtain high

selectivity during the etching of GaAs on AlGaAs. Fluorine-containing plasmas also

provide a means for selectively etching silicon nitride and silicon dioxide masks on

III-V compound semiconductors.

The most commonly used masking materials for etching of III–V compounds

include metals such as Cr, Ni, Ti, Al, chemical vapor deposited dielectrics such

as Si

,SiO

, and novolac resin-based photoresists and electron beam photore-

sists. Metals exhibit the highest selectivity for these etching applications; however,

the grain size of the deposited ﬁlm determines the sidewall roughness and lim-

its the achievable smoothness. The smoothest etch results have been achieved

by using hardbaked, multilayer, novolac-resin-based photoresists. In general poly-

methyl methacrylate (PMMA) e-beam resist is found to exhibit poor etch resistance

in comparison to the novolac resists. However, positive tone e-beam resists such as

ZEP-520A and negative tone SAL-601 have been found to offer good etch resis-

tance. Care has to be taken to prevent any erosion of mask which might result in

roughening of the top part of the etched features.

7.7.1 Case Study: Etching of GaAs and AlGaAs

In this section, various GaAs gas chemistries are brieﬂy reviewed. Speciﬁc advan-

tages for each process are brieﬂy indicated. Table 7.5 lists the etch rate of some of

the commonly used masking materials in GaAs etching [81].

Reactive Etching in SiCl

Plasma: Etching at moderate pressures of 2.63–

13.15 Pa is chemical in nature. Typically at moderate pressure, etching of GaAs

in chlorine and bromine plasma is rapid, spontaneous, and crystallographic as seen

in Fig. 7.19 [37]. The relative chemical etch rates of the various etch planes in GaAs

are: (111)

As - rich

> (100) > (110) > (111)

Ga - Rich

. Silicon tetrachloride is not as

442 S. Tadigadapa and F. Lärmer

Table 7.5 Commonly used masking materials compatible with etching III–V semiconductors in

chlorine plasma

Material Etch rate (nm/min) GaAs: material etch rate ratio (selectivity)

GaAs 28.33 1

Au-Pd 2.5 11.3333

Au 1.75 16.19048

Glass (SiO

) 0.708333 40

Cr 1.25 22.66667

Ni 0.5 56.66667

Ni-Cr 0.833333 34

AZ-1350 J (Photoresist) 2.916667 9.714286

These etch rates were obtained in 90% Ar/10%BCl

plasma chemistry obtained in diode RIE

set-up operated at 50 W RF power (300 V bias), 1.97 Pa pressure [81]. Used with permission,

Fig. 7.19 Etch

characteristics of GaAs in

SiCl

reactive ion plasma

(2.63 Pa, 150 W,

200 mW/cm

). The high

pressure and low power etch

results in the emergence of

the various crystallographic

facets [37] (Used with

Springer)

corrosive as pure Cl

gas and the plasma leaves no nonvolatile residues and is a

clean process well suited for crystallographic dry etching.

Reactive Ion Etching in Cl

Plasma: Smooth and vertical trenches are obtained

due to the high DC-bias ion bombardment. Aspect-ratio-dependent etch rate is

clearly observed. In order to avoid any roughness or grass formation due to

microloading, this etch must be performed at very low pressures in the range of

0.066–0.66 Pa. The addition of BCl

and Ar can result in very smooth trench for-

mation [82]. Figure 7.20 shows an SEM picture of a high-aspect-ratio, smooth etch

proﬁle obtained in GaAs using chlorine plasma [37].

Reactive Ion Etching in BCl

+ Ar Plasma: Using this etching chemistry at low

pressures and high Ar%, equirate etching of GaAs and AlGaAs can be obtained. An

equi-etch rate of 280 nm/min for 50 W, 1.97 Pa, and 90% Ar has been reported [81].

Addition of 10% oxygen to this mixture provided a selectivity of ∼5:1, however, at

a remarkably slow etch rate of 10 nm/min for GaAs.

7 Dry Etching for Micromachining Applications 443

Fig. 7.20 Etch characteristics of GaAs in Cl

in an inductively coupled plasma reactive ion etcher

(0.2 Pa, 150 W, ∼500 V dc bias). Highly anisotropic etch showing width dependent etch rate is

clearly observed. It is possible to obtain smooth etch morphologies using low pressures (below

Reactive Ion Etching in SiCl

and SiF

Plasma: This etch chemistry provides a

high selectivity GaAs etch against AlGaAs. The use of separate chlorine and ﬂuorine

containing gases as opposed to CCl

(CCl

is an ozone depleting chloroﬂuo-

rocarbon the use of which currently conﬂicts with “Montreal Protocol”.) affords

greater ﬂexibility in controlling the Cl/F ratio. In addition, no polymer formation is

observed in this gas mixture. Formation of AlF

results in the etch stopping mech-

anism for AlGaAs in ﬂuorine containing plasma. Figure 7.21 shows the selectivity

of GaAs over AlGaAs as a function of the percentage of SiCl

[83]. Clearly pref-

erential high selectivity of GaAs over AlGaAs is obtained for aluminum-rich GaAs

compositions and at low SiCl

concentrations.

Percentage of SiCl

Selectivity (Etch Ratio of GaAs: AlGaAs)

0 10 20 30 40 50

100

200

300

400

500

600

700

800

0.07

0.93

0.35

0.65

Fig. 7.21 Selectivity of

GaAs over AlGaAs as a

function of %SiCl

SiCl

+ SiF

plasma. The

data were obtained for a RIE

plasma at a pressure 7.89 Pa

and a dc bias of −60 V and

offer a selectivity in excess of

500:1 for Al

0.35

0.65

As [83]

(Used with permission,

Vacuum Society)

Reactive Ion Etching in CH

and H

Plasma: Hydrogen–alkane mixtures can

also be used to etch GaAs [84, 85]. However, the etch rates in this case are