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

Подождите немного. Документ загружается.

424 S. Tadigadapa and F. Lärmer

Figure 7.9 shows the polymer ﬁlm thickness and etch rates of silicon and silicon

dioxide upon 0–50% H

addition to CF

plasma [43]. An increasing thickness of the

ﬂuorocarbon with increasing hydrogen concentration and a proportional decrease in

the etch rate of silicon is observed, whereas the SiO

shows neither an enhanced

formation of the ﬂuoropolymer ﬁlm nor any appreciable dependence of the resulting

etch rate. Using this technique SiO

: Si selectivity of up to 20:1 has been reported.

− H

plasma provides high selectivity between photoresist and SiO

ﬁlms as

shown in Fig. 7.10.

Percentage of H

in CF

Fluorocarbon Film Thickness (nm)

0 5 10 15 20 25 30 35 40 45 50

0.5

1.5

2.5

3.5

For Si Surface

For SiO

Surface

Percentage of H

in CF

Etch Rate (nm/min)

SiO

/Si Etch Rate Ratio

0 5 10 15 20 25 30 35 40 45 50

Silicon Etch Rate

Silicon Dioxide Etch Rate

SiO

/Si Etch Rate Ratio

Fig. 7.9 Fluorocarbon ﬁlm thickness on Si and SiO

measured after 5 min RIE under identical

condition (bottom). Formation of the relatively thick layer of ﬂuorocarbon effectively suppresses

the etch rate of silicon in CF

(top) plasma providing high selectivity conditions suitable for

etching SiO

7 Dry Etching for Micromachining Applications 425

% H

in CF

Etch Rate (nm/min)

0 5 10 15 20 25 30 35 40 45

RF Power = 0.26 W/cm

Pressure = 0.47 Pa

Flow Rate = 28 sccm

Silicon

Silicon Dioxide

AZ1350B Resist

PMMA

Fig. 7.10 Etch rate of photoresists, silicon, and silicon dioxide in H

/CF

plasma. At high hydro-

gen percentage, a large value of selectivity between SiO

and these materials is obtained [42](Used

Silicon to Silicon Dioxide Etch Selectivity: Fluorine plasma naturally provides an

inherent selectivity for etching silicon with respect to silicon dioxide. Flamm et al.

measured the luminescence during silicon etching in ﬂuorine plasma and concluded

that ﬂuorine atoms etch silicon at a rate (R

F(Si)

)given by [44]

F(Si)

= 2.91 ± 0.20 × 10

−12

√

−0.108 eV/k

, (7.4)

where n

is the atom concentration of ﬂuorine given in #/cm

−3

. The etch rate for

silicon dioxide (R

F(SiO

)

) in ﬂuorine plasma is given by the expression

F(SiO

)

) = 6.14 ±0.49 × 10

−13

√

−0.163 eV/k

(7.5)

resulting in a relative Si to SiO

etch ratio of

F(Si)

F(SiO

)

= (4.74 ± 0.49)e

−0.055 eV/k

(7.6)

and ranges from a value of 41 at room temperature to 26.2 at 100

◦

C. Etching of

silicon in chlorine plasma has been presented by Schwartz et al. [7]. In general

silicon etching in chlorine plasma is an ion-beam induced process and therefore

results in poorer selectivity.

Addition of Inert Gases: In addition to oxidants and radical scavengers, inert

gases are often added to the plasma. This is done for several reasons: (i) inert gases

such as Ar stabilize the plasma because they readily donate electrons to the plasma

426 S. Tadigadapa and F. Lärmer

in contrast to the chemical species which, owing to their high electronegativity, cap-

ture available electrons needed to maintain the glow discharge; (ii) the high sputter

yield of Ar

ions improves the ion-assisted etch rate; (iii) the directional nature

of ion bombardment through the sheath improves the anisotropy of the etch; and

(iv) the high thermal conductivity of inert gases such as He when added to the

plasma can improve the heat transfer from the substrate wafer to the supporting

chuck and plasma chamber. Of course large quantities of inert gas addition will

dilute the reactive gas component and the etch process will shift towards physical

sputter etching.

Recently, Yamakawa et al. [45] have demonstrated very high speed etching of

silicon dioxide (BPSG) ﬁlm using microwave excited nonequilibrium atmospheric

pressure plasma. The authors were able to demonstrate an ultrahigh etch rate of

SiO

(14 μm/min) and an unprecedented selectivity of 200 with respect to silicon

using NF

/He along with addition of H

O as the etching gas. The fast etch rate can

be attributed to the presence of NF

radicals arising from the breakup of NF

into

∗

−radicals. NF

−radicals (with x > 0) are extremely aggressive towards

SiO

and other dielectric (and polymers) materials and etch them at extremely high

speed. Addition of water vapor (H

O) consumes the ﬂuorine radicals by HF forma-

tion (2F

∗

O → 2HF+ < O >) whereas NF

radicals are much less scavenged.

Water vapor therefore acts as a selective ﬂuorine radical scavenger and consequently

reduces any undesirable Si-attack. However, if NF

is completely broken down, for

example, as in high-density, high-power plasma, then the plasma primarily consists

only of ﬂuorine and nitrogen radicals. Under these conditions, a complete quenching

of the fast etch rate of SiO

and other dielectrics has been observed. The complete

breakup of NF

molecules under appropriate plasma conditions is also conﬁrmed

by the plasma color which changes from a deep red to a bluish color, indicating a

change in preponderance from NF

to ﬂuorine radicals in the plasma.

Silicon Nitride Etching: Silicon nitride is readily etched in ﬂuorine plasma and

has intrinsic etch characteristics between silicon and silicon dioxide. Because sili-

con nitride is typically used in the LOCOS (local oxidation of silicon) process, there

is often a need to etch silicon nitride over silicon dioxide and silicon with high selec-

tivity. The intrinsic rate of etching for Si

is about 5–10 times higher than that

of SiO

depending upon temperature. In a CF

plasma the obtained Si

:SiO

selectivity is ∼2[46]. In general adding nitrogen to CF

plasma has been found

to improve the selectivity of silicon nitride over silicon dioxide to 10:1.

Sanders et al. reported achieving a high selectivity of >10:1 for etching silicon

nitride over silicon dioxide by the addition of CF

Br to CF

plasma [46]. It

is considered that this chemistry converts some of the F atoms into BrF and BrF

which selectively etch silicon nitride whereas silicon dioxide is not affected by the

interhalogen gas species. Furthermore, Br and Cl radicals in the plasma are thought

to convert the atomic oxygen (radicals) into more inert oxygen molecules. This

quenching of oxygen radicals, which would otherwise oxidize the Si

surface

explains the enhancement of Si

etching in comparison to SiO

in interhalogen

gases. The highest selectivity in etching silicon nitride over silicon dioxide has been

reported when using interhalogen ﬂuoride gases such as ClF

and BrF

[47].

7 Dry Etching for Micromachining Applications 427

In summary, due to the spontaneous formation of SiF

, silicon is more easily

etched than SiO

which requires the etching process to be induced via ion bom-

bardment. Addition of oxygen in RIE etch processes increases the etch rate of both

Si and SiO

. Addition of hydrogen can be used to selectively etch silicon dioxide

over silicon through the formation of ﬂuorocarbon ﬁlms. Silicon nitride etches at a

rate that is in between that of silicon and silicon dioxide. Highly selective etching

of silicon nitride over silicon dioxide layer can be performed in 1:2 ratio NF

/Cl

downstream etchers.

7.5 Case Study: High-Aspect-Ratio Silicon Etch Process

Fluorine-based plasma etching processes for silicon yield high etch rate, etch depth,

and mask selectivity, and in particular enable photoresist masking with good selec-

tivity between silicon and photomask. This is mainly due to the high chemical

reactivity and spontaneous etching nature of the ﬂuorine radicals towards silicon,

and the high volatility of the silicon ﬂuorides as reaction products. As a conse-

quence, the etch is intrinsically isotropic. Fluorine radicals do not need ion activation

to initiate or enhance their reaction with silicon, or to remove the reaction products

from the silicon surface; anisotropy can only be achieved by the inclusion of side-

wall passivation schemes to the process. The existing approaches to deep reactive

ion etching (DRIE) of silicon are distinguished by the way sidewall passivation

is achieved, the key to anisotropy and overall performance of the etch process.

Cryogenic etching and the so-called “Bosch process” with alternating etch and pas-

sivation cycles are the two most well-known high-aspect-ratio silicon etch processes

and are discussed in this section.

A high-density plasma tool with so-called remote or decoupled plasma is ben-

eﬁcial or even a must for all silicon DRIE approaches, be it cryogenic or Bosch

processing. Silicon DRIE is based on the chemical activity of the plasma. Chemical

activity requires both a sufﬁciently high concentration of passivating r adicals, and a

sufﬁciently high concentration of ﬂuorine radicals to obtain the high etch rates. For

a high density of chemically active species, a process pressure regime of typically

between 1 and 10 Pa is most appropriate. Energy of the ion ﬂux bombarding the

wafer surface must be controlled independently from the plasma excitation, through

substrate electrode biasing by a radio frequency or low frequency bias generator.

For a good control of ion acceleration independent of plasma excitation, the plasma

potential should be kept as low as possible near ground potential. The latter involves

decoupling of the plasma excitation zone from the substrate region.

Traditional plasma sources for RIE fail in one or more of these require-

ments. Diode and triode reactors show insufﬁcient plasma excitation density, below

−3

. Electron cyclotron resonance (ECR) sources are inappropriate for the

very low pressure regime (<0.1 Pa) tolerated by the cyclotron resonance mode.

ECR is a typical ion-current type of source and lacking sufﬁcient chemical activ-

ity. Microwave surfatron sources (a tube or horn made from dielectric material,

to guide the microwave ﬁeld along the interface between the dielectrics and the

428 S. Tadigadapa and F. Lärmer

plasma burning inside the tube or horn, down to the process chamber, (e.g., see

[48]), helicon sources, and inductively coupled plasma (ICP) sources open the pres-

sure window up to 10 Pa and to even higher pressures. Amongst these alternatives,

the ICP has emerged as the most appropriate and widespread source technology for

DRIE and as the industry standard.

7.5.1 Cryogenic Dry Etching

As explained, sidewall passivation is necessary to obtain anisotropic silicon etches.

One approach is to use ﬂuorine radicals created from sulfur hexaﬂuoride gas (SF

)in

the plasma discharge as the main etching species which would readily and isotropi-

cally remove silicon wherever it is unprotected and accessible. Oxygen radicals are

created in the plasma from oxygen gas (O

) and serve for silicon surface oxidation.

Oxidation saturates dangling bonds on the silicon surface and builds up a passivat-

ing silicon oxide ﬁlm, which inhibits ﬂuorine attack of the silicon. After ﬁnding the

appropriate SF

to O

ﬂow ratio, an oxide scavenging gas such as triﬂuoromethane

CHF

may be added to improve results and widen the process window to some

degree (so-called “Black Silicon Method”, see [49]). At room temperature, the per-

centage of oxygen needed to achieve anisotropic proﬁles is larger than 50%, which

reduces both etching speed and mask selectivities considerably.

Cryogenic dry etching is a variation of the sidewall passivation technique based

on surface oxidation. Lowering the wafer temperature into the cryogenic regime

typically around 173 K (–100

◦

C) by liquid nitrogen (LN

) cooling of the substrate

electrode, brings down the required oxygen amount for anisotropic etching to a few

percent of the total gas ﬂow, thus pushing up etching speed and mask selectivi-

ties drastically. In addition, the process window for achieving anisotropic etching is

widened strongly.

This results from an enhancement of the chemical stability of the inhibiting sil-

icon oxide ﬁlm in a ﬂuorine-based plasma, mainly because desorption of reaction

products of silicon oxyﬂuoride-type from unbombarded surfaces is frozen by the

low temperature. Ion bombardment of the etch ﬂoor removes the passivating ﬁlm

there by sputtering off most of the chemical compounds. Perfectly vertical etch-

ing in silicon can be achieved with a strong overweight of etching ﬂuorine radicals,

compared to the passivating oxygen radicals, yielding high etch rates and a relatively

wide process window, in comparison to the room temperature situation.

Impressive results are achieved with cryogenic dry etching in inductively cou-

pled plasma tools such as the Alcatel reactor, with silicon etch rates reaching

4–6 μm/min, and mask selectivities of several 100:1 for SiO

-hardmasks [50, 51].

Use of photomasks is difﬁcult due to cracking and delamination reasons at the

low substrate temperatures, however, not impossible if special resist treatment for

stress adjustment is performed before the etch. The main drawback of the cryogenic

etching method is related to the low substrate temperatures needed and the criti-

cal dependence of process properties on the substrate temperature, given the strong

heat impact from the plasma to the substrate and the exothermic etching reaction

7 Dry Etching for Micromachining Applications 429

between silicon and the ﬂuorine radicals. In addition, the substrate as the coldest

part in the whole reactor conﬁguration acts as a cryopump, freezing out compounds

from the plasma which act as hard-to-remove micromasks on the wafer surface.

Micromasking is responsible for the formation of microneedles and micrograss on

the silicon which is often observed in deep cryogenic etches as a disturbing and

often unacceptable factor. Nevertheless, cryogenic dry etching is a very important

silicon microstructuring technique, for example, in microoptical applications, where

smooth sidewalls on a nanometer scale are of key importance.

7.5.2 Bosch Process

The etching technique described before makes use of relatively hard to remove com-

pounds as passivating layers, in the form of silicon oxides or oxyﬂuorides resulting

from silicon surface oxidation. Their complete removal from the etch ﬂoor without

leaving micromasking residues requires energetic ion impact, eventually in com-

bination with added scavenging agents. Excessive ion sputtering and the use of

scavengers reduce the selectivity towards the masking materials. This is especially

true in the case of photoresist masking. There is a tradeoff between a clean trench

bottom and high mask selectivity.

A technique that avoids the formation of such hard to remove compounds is the

deposition of smooth polytetraﬂuoroethylene (PTFE) or Teﬂon



-like ﬁlms on the

silicon surface during etching. Early published work describes the importance of

the balance between etching and polymerizing species, and effects of changing this

balance for both Si and SiO

etching [52].

Plasma polymerization can be achieved by the generation of (CF

)

-type radi-

cals in the plasma, starting from precursor gases such as hexaﬂuoropropene (C

)

or octaﬂuorocyclobutane (RC318



), the latter being a nontoxic and stable

dimer of tetraﬂuoroethylene (TFE, C

). The deposited smooth passivating ﬁlm

consists of a network of long linear (CF

)

-chains with only little cross-linking,

which can be easily removed from the etch ﬂoor by low-energetic ion bombard-

ment without leaving behind any residues. A mixture of sulfur hexaﬂuoride gas for

the ﬂuorine radical supply and octaﬂuorocyclobutane gas for the supply of polymer-

forming radicals can be used in the plasma to achieve both passivation at the sidewall

and etching at the etch ﬂoor, yielding anisotropic proﬁles in silicon. In addition, hav-

ing ﬂuorocarbons in the process to some extent scavenges undesired oxygen in the

background gas or oxygen precipitates in Si and thus prevents the formation of hard

passivation (silicon oxides) from the etched surface (ﬂoor).

However, the presence of ﬂuorine radicals as etching species and of Teﬂon



ﬁlm-forming monomer species together in the plasma, at the same time, leads

to recombination and mutual extinction of both kinds of species. This makes the

“mixed process” difﬁcult to control for deeper etches and lowers process perfor-

mance mainly with respect to etch-rate. Although, these drawbacks are somewhat

mitigated using high-density ICP plasma, still the potential of the “mixed process

approach” remains limited to shallow trenches with a depth on the order of 10 μm.

430 S. Tadigadapa and F. Lärmer

The recombination problem was overcome by the patented Bosch-process tech-

nology [48], which is a variation of the Teﬂon



-based sidewall passivation

technique. Figure 7.11 illustrates the process mechanism: passivation and etching

gases are introduced separately and alternately into the process chamber and the

substrate to be etched is exposed to a high-density plasma, during the so-called pas-

sivation and etch cycles, respectively. In each passivation cycle, thin Teﬂon



-like

ﬁlm is deposited onto the walls of etched structures from C

-precursor species.

Some scavenging of oxides from the silicon etch ﬂoor may also happen during

or after the Teﬂon



deposition step. During the subsequent etching cycle, part of

this ﬁlm is removed from the coated sidewall by off-vertical ion i mpact and driven

deeper into the trench. At the same time the trench bottom is cleared of ﬂuorocarbon

polymer and etched by ﬂuorine radicals released from SF

in the plasma. Switching

times between steps normally range from a few seconds up to 1 min, depending on

sidewall roughness that can be tolerated. Because the passivating polymer ﬁlm can

be easily removed by small ion impact sputtering, mask selectivity reaches very high

values, for example, 150:1 for photoresist and »300:1 for SiO

hard masks. If harder

passivating polymers were used, more aggressive ion impact would be required and

as a consequence, the selectivity towards the masking material would be reduced.

CF2-species

Passivation

F-species

Etching

positive ions

Fig. 7.11 Illustration of the basic mechanism of the Bosch process. The protecting Teﬂon



-like

ﬁlm deposited onto the structures during the passivation step is cleared from the trench ﬂoor and

driven deeper into the trenches during the (intrinsically isotropic) ﬂuorine-based etching step

Transport of the sidewall ﬁlm (i.e., removal and redeposition of Teﬂon



material) yields some local anisotropy of the etching steps that would otherwise be

completely isotropic, in the vicinity of the sidewall, but does not adversely affect the

main etching reaction away from the sidewall. Sidewall roughness is thus reduced

despite the discontinuous process. Plasma polymerization is nonconformal i n nar-

row trenches and ﬁlm formation preferentially takes place at the trench opening,

leaving deeper sidewall regions with less deposited passivating ﬁlm, therefore the

ﬂow of polymer material along the sidewalls towards the trench ﬂoor also provides a

7 Dry Etching for Micromachining Applications 431

Fig. 7.12 Trench proﬁles and sidewall structure resulting from a typical Bosch DRIE process

recipe

more uniform sidewall passivation over the trench depth. Figure 7.12 shows typical

structures, proﬁle forms, and characteristic sidewall scalloping arising from standard

Bosch DRIE process recipes.

The discontinuous nature of the process overcomes a number of general problems

involved with anisotropic etching.

Firstly, recombination loss of reactive species, by gas phase reactions causing

their pairwise extinction, is excluded by separating them in the time domain. This

is most important for the volume of the plasma source itself, where the species are

generated at high concentration levels and where recombination would be strong.

Outside the high-density source area, in a more diluted state, coexistence of species

is easier to achieve and some mixing there, in the space between the source and

the substrate, can be tolerated. For a high-density plasma source, sufﬁciently small

in volume, switching times to below 1 s, for example, down to 100 ms are pos-

sible. Passivating and etching species remain separate in the limited volume of

high-density plasma excitation, but may overlap on their way from the source to

the substrate. This so-called “ultrafast gas-switching condition” yields very smooth

sidewalls with hardly any remaining sidewall scalloping at all, as is depicted in the

SEM pictures of Fig. 7.13.

Secondly, a vertical etching process having high mask selectivity bears an inher-

ent risk for micromasking the trench ﬂoor. Even minor contamination on the etch

surface may lead to formation of residues, microroughness, and even microneedles.

Having a sidewall with a periodic nanostructure or ripple on the order of 10 nm

in size is an advantage in this respect: micromasks smaller than the characteristic

dimension of the scallops are undercut and extinguished from the surface, leav-

ing back no etch residues. Only if the size of a micromask exceeds the critical

dimension limit that the process can tolerate by undercutting will it cause a potential

residue problem. This has to be considered in connection with the before-mentioned

approach where the sidewall scalloping is practically gone: the latter process is more

critical with respect to formation of microroughness, resulting from micromasking

effects starting from a random distribution of incidental contaminants on the etched

surface.

432 S. Tadigadapa and F. Lärmer

Fig. 7.13 SEM pictures of the sidewall resulting from a Bosch process using an ultrafast switch-

ing recipe: hardly any horizontal scalloping is visible at the sidewall any more. Some vertical

striations are visible in the lower part of the left-hand side SEM picture, indicating the beginning

of microroughness formation at the trench bottom, which also affects the sidewall

7.5.3 Understanding Trends for DRIE Recipe Development

High-Aspect-Ratio Etches: As a general observation in plasma etching processes,

net etch rate decreases with rising aspect ratio of the trenches. In addition, reduced

ion impact reaching the trench ﬂoor shifts the etching process towards more effective

passivation with higher aspect ratios. As a consequence, proﬁles tend to develop

positive slopes and typically trenches end up tipping when their aspect ratios begin

to approach values of greater than 20:1.

Given the condition of having all trenches within a similar range of widths, adap-

tation of the process recipe stepwise or continuously [53, 54] helps to reduce the

amount of passivation in the process and keeps the proﬁles straight. For example,

passivation cycle time or passivation gas ﬂow can be ramped down steadily. As

an alternative, etching cycle time or etching gas ﬂow can be ramped up steadily.

Although other options including source power, bias power, or pressure ramping do

exist; changing cycle times is the most obvious parameter ramping s trategy, to affect

the etch : deposition balance in a predictable and straightforward manner. With the

use of parameter ramping, trenches of >50:1 aspect ratio can be fabricated, while

keeping the sidewalls straight and vertical.

Aspect Ratio Dependent Etching – RIE Lag: Etching speed and proﬁle evolu-

tion depend strongly on the depth : width ratio of the trench, the so-called aspect

ratio. In general, the etch in narrower trenches progresses at a lower speed than in

wider trenches; that is, higher-aspect-ratio trenches lag behind the lower-aspect-ratio

trenches. This is partly due to the fact that the amount of ions reaching the trench

ﬂoor progressively diminishes for reasons of aperture effects at the trench opening,

in combination with the angular distribution of the ions. Another important aspect

is transport limitations of gaseous species – radicals – in high-aspect-ratio trenches.

7 Dry Etching for Micromachining Applications 433

The amount of the lag effect is determined mainly by the mean free path of the

gaseous species, that is, gas pressure.

Control of RIE Lag: RIE lag is a common and well-understood transport phe-

nomenon in plasma etching [55]. The amount of the lag effect depends mainly

on the mean free path of the gaseous species which determines their transport in

etched trenches. The dependence of etch rate on the aspect ratio of the feature has

been elucidated based upon a simple vacuum conductance model [56]. In addition,

interaction and change of balance between etching and passivating species during

transport in etched trenches is important, which is again related to mean free path

and process pressure. Changes in the balance of etching and passivating species with

increasing aspect ratio also affect the anisotropy and the proﬁle. Figure 7.14 shows

a characteristic example of RIE lag in high-aspect-ratio Bosch-process etches.

Fig. 7.14 Illustration of RIE

lag effect: etch depth

decreases with narrower gap

sizes corresponding to higher

aspect ratios

In a discontinuous process with separate etching and passivating cycles that can

be controlled independently of each other, there is an individual lag effect both for

the etching and for the passivating part of the overall process. For the etch step, the

higher the aspect ratio of the trench the fewer ﬂuorine radicals per time unit can

reach the trench ﬂoor and etch the silicon, the size of the effect depending mainly

on the etch step pressure. For the deposition step, the higher the aspect ratio of the

trench the fewer polymer formers per time unit can reach the trench ﬂoor and build

up an inhibiting ﬁlm on the ﬂoor, the size of the effect depending again mainly

on the deposition step pressure. Because a thinner inhibiting ﬁlm on the etch ﬂoor

boosts the rate for the subsequent etch step, this involves some compensation of

the rate decrease with the aspect ratio resulting from the reduced ﬂuorine radical

supply. Deposition step lag and etching step lag can compensate each other to reach

a lagfree overall process over a wide range of aspect ratios. This is achievable by

individual control of the etch step and deposition step parameters; mainly by the

pressure values set.

Setting the pressure higher for the individual step increases its lag effect; setting

the pressure lower decreases its lag effect. Increasing deposition step pressure with