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

Подождите немного. Документ загружается.

464 D.W. Burns

adjustments to the overall process architecture may be helpful. Alternatively, device

designs can often be adjusted to accommodate undue constraints on etching or other

process sequences.

8.2.1 Surface Reactions and Reactant/Product Transport

Etching systems generally comprise an etch beaker or tank sufﬁcient in size to

hold one or more wafers comfortably in an often vertical orientation, surrounded

completely with liquid etchant during an etch cycle, as illustrated in Fig. 8.3.A

heater, thermocouple, and agitator may be included. The tanks often have a drain

facilitized into an acid/base neutralization system, a dedicated hydroﬂuoric acid

collection system, or a solvent collection tank. Etchant may be poured into the

tank from bottles of etchant or from permanently installed delivery systems with

external chemical storage facilities. The etch tanks and liquid distribution systems

are formed from inert materials with respect to the etchant used, such as perﬂu-

oroalkoxy (PFA), ﬂame-retardant polypropylene (PP), high-density polyethylene

(HDPE), polytetraﬂuoroethylene (PTFE or Teﬂon



), molded polyvinylidene ﬂu-

oride (PVDF), or quartz [27, 33]. Stainless steel tanks are occasionally used for

solvents and sometimes for KOH etchants. PFA tanks and holders may be used for

KOH etching. The etch tank may be placed inside another tank or sink for safety and

spillage control. To protect the operator, the tank and sink are typically exhausted

Etchant Rinse WaterEtchant

Wafers

Etch

Bath

Drain

Heater

Agitator

Etchant

Rinse Water

Rinse

Bath

Drain

Rinse Water

Supply

Fig. 8.3 Etch baths (a) for wet etchants often have a heater, a thermocouple, an agitator or stirrer,

and valving to drain the etchant. A separate rinse bath (b) allows copious amounts of deionized

water to rinse the wafers adequately prior to spin or blow-drying. A dedicated holder or cassette

(not shown) of polyaryletheretherketone (PEEK) with embedded carbon ﬁbers, polypropylene,

PFA, PTFE or Teﬂon



, PVDF, or quartz may be used for wafer transport during etching and

rinsing

8 MEMS Wet-Etch Processes and Procedures 465

with an overhead exhaust hood implemented with splashguards as part of a wet

bench.

The wet bench may have timers, light and fan s witches, heater controllers, alarm

systems, and one or more dedicated etch baths along with rinse tanks. The rinse

tanks are similarly constructed and allow for copious amounts of deionized water to

spray on or bubble up past the wafers, diluting and removing the etchant from the

wafers. The rinse baths may have drop doors at the bottom of the tank to rapidly

remove the rinse water in multiple dump–rinse cycles. Some etch sequences require

that the wafers be etched and rinsed in the same bath (i.e., dilution rinses), however,

wafers are generally transported from the etch bath to the rinse bath by manual or

automatic lifting. The wafer holders or cassettes must also be inert to the etchant and

are usually dedicated to a particular wet bench or etch process to avoid possibilities

of cross-contamination.

A generalized model for liquid-etch processes is illustrated in Fig. 8.4. Etching

occurs when reactants in a liquid etchant chemically react at an exposed surface of

the material being etched, with by-products dissolving back into the liquid etchant

or released as a gas. The etch rate is affected by the concentration of reactants in the

liquid etchant, the local transport of these reactants to the exposed surface, the reac-

tion rate at the interface between the liquid etchant and the material being etched,

the removal rate of the reaction products from the surface, and the concentration of

reaction products in the etch bath.

The reactions at the liquid–solid interface can be surprisingly complex. Etching

may involve a composite reduction–oxidation reaction that produces an oxide of the

material being etched that is then etched to expose fresh material in a continuous

manner. Secondary effects can occur and sometimes dominate, such as vapor bar-

riers at the interface from a high-aspect-ratio feature or from nonwetting masking

material. Unwanted by-products, such as a salt residue or other solid formation at the

etched surface, can generate an etch barrier that slows or even stops an etch. Local

evolution of gas can create a micromasking effect that slows the etch rate while

bubbles stay adhered to the surface, resulting in a roughened surface and possible

hillock formation. Inadvertent, localized etch rate reductions can occur with viscous

or low-solubility etchants, as when a wafer is partially rinsed and then reinserted

into the etch bath with insufﬁcient agitation.

Enhancements to the etch rate are often achieved by: increasing the concentration

of the reactants, performing the etch at an elevated temperature, using an agitator to

assist etchant ﬂow, changing the etch bath to get fresh etchant and expel accumulated

by-products, and minimizing obstructions to the etchant that may occur with direct

contact to a wafer holder or close proximity to another wafer. The rate-limiting step

of the generalized model usually determines which enhancements have the most

impact. Increases in the etchant temperature usually result in a faster etch, although

room-temperature etching is often the most convenient and safest with reasonable

etch rates obtained by careful selection of the etchant and etchant concentration.

Nearly all wet etches, including anisotropic etches, encompass some degree of

undercutting when using an overlying photoresist or hard mask layer having pat-

terned features. An ideal etch, best realized with a dry etch rather than a wet etch,

466 D.W. Burns

Mask Opening

Mask Defect

Liquid Etchant

Masking Layer

Etched Layer

Etch Stop

Substrate

Normal Undercut

Accelerated Undercut

stcudorPstnatcaeR

Backside

Mask Opening

Mask Defect

Liquid Etchant

Masking Layer

Etched Layer

Etch Stop

Layer

Substrate

Normal Undercut

Accelerated Undercut

stcudorPstnatcaeR

Backside

Protection

Fig. 8.4 Illustration of transport mechanisms for liquid etching of masked layers with an underly-

ing etch stop layer. A supply of reactants in the etchant react heterogeneously with exposed portions

of the etched layer resulting in one or more by-products that dissolve back into the etch liquid or

egress as a gas. Insufﬁcient reactants or an excessive concentration of products can slow the etch

inordinately, particularly in tight geometries. With most thin ﬁlms and amorphous materials, the

reaction is isotropic, resulting in circular corners viewed from above and a generally rounded etch

proﬁle at the sidewalls from the side. As the etch front strikes an underlying etch stop layer, the

reactants are less exhausted and the etch can speed up laterally, resulting in a more angular and

less rounded sidewall. Masking layers with high internal stress or etched layers with a deviated

morphology at the upper surface can result in appreciably accelerated etch fronts at the interface

between the masking layer and the etched layer, effectively accelerating the undercutting. Defects

in the masking layer can result in unwanted local etching. A dark-ﬁeld masking layer is illustrated;

similar concerns and phenomena occur for light-ﬁeld masks though the amount of material etched

is usually much greater

translates the overlying patterned mask features into the etched layer with high

ﬁdelity and minimal feature distortion. Wet etches, with their inherent property of

etching any exposed surface, generally etch the sidewalls at a similar rate as the

bottom. Isotropic etches produce a lateral undercut distance on the same order of

magnitude as the etch depth, resulting in a curved proﬁle having a radius approxi-

mately equal to the etch depth as shown on the left side of Fig. 8.4. Undercutting bias

may be compensated with adjustments to the widths of corresponding features on

the photomask or beam-writing tool. Undercutting concerns may preclude the use

of wet etchants for ﬁne features in favor of dry etching that is more vertical and has

a diminished lateral etch rate. With anisotropic materials or anisotropic etchants, the

sidewall proﬁle becomes more angular. For the isotropic case, as the etchant breaks

through to an underlying etch s top layer, the etchant can become less exhausted

8 MEMS Wet-Etch Processes and Procedures 467

locally and etch laterally more quickly, which tends to linearize and straighten the

sidewall.

Other phenomena such as a highly tensile masking layer can cause a slight lift-

ing of the masking layer near a feature opening, which causes the mask to peel

back slightly during etching. The peeling results in accelerated undercutting at the

interface and a shallow angular proﬁle as illustrated near the center of Fig. 8.4.A

compressive mask may also tend to lift via mechanical buckling as the aspect ratio

(undercut distance to mask thickness) increases, which further increases the etch

rate.

The lateral etch rate can be accelerated by modifying characteristics of the etched

layer such as decreasing the density at the top or bottom of t he ﬁlm, generating a

high defect density near the top surface, or locally injecting high levels of impuri-

ties. In cases where the etched feature is deep or where extensive lateral etching is

desired, the etchant can become locally constrained and exhausted, decreasing the

local etch rate. Defects in the masking layer, such as small particulates, can allow

the etchant to creep along the particulate or move blatantly through an open mask

defect to inadvertently etch the underlying material, as shown on the right side of

Fig. 8.4.

8.2.2 Etchant Selectivity and Masking Considerations

A major consideration in the selection and evaluation of a wet-etch process sequence

is the etch selectivity, deﬁned as the etch rate of the etched material relative to the

etch rate of the masking or underlying material. Etch selectivities of 100 or more

are preferable although selectivities as low as one or less can prove adequate for

thin etched layers. Etch selectivity with respect to the etch stop layer or to other

layers exposed during the etch sequence may also be of consequence, although it is

often of lesser concern because the underlying layer is generally exposed for only a

short time during the overetch phase of the sequence. As depicted in the graphic of

Fig. 8.5, a higher etch selectivity usually allows a thinner mask layer, a deeper etch,

a thinner etch stop layer, and improved overetch capability.

The choice of etchant may be determined by the masking layer, and vice versa.

The masking layer prevents the etchant from reaching covered surfaces and must

hold up during the time for the etch. A thicker masking layer can increase the achiev-

able etch depth. The masking layer may also etch laterally to some degree as the

etch proceeds. This additional source of bias may be compensated for in the pho-

tomask design by automatically adding or subtracting a suitable amount from light

or dark ﬁeld mask features, respectively. Over time, the wet etchant may absorb into

the masking layer, causing it to swell, crack, or possibly lift, thereby exposing the

etched layer in unintended areas. Attention must be paid to the status of the ﬁlms

on the backside of the substrate, inasmuch as this portion as well as the sides of the

substrate are generally exposed directly to the etchant. Additional protection may be

468 D.W. Burns

Etch Selectivity

Lower

Higher

Mask Thickness

Thicker

Thinner

Etch Depth

Shallower

Deeper

Etch Stop Layer

Thicker

Thinner

Overetch

Shorter

Longer

Fig. 8.5 As the selectivity of an etchant compared to an etch mask or an etch stop layer increases,

the mask thickness can be thinner, the achievable etch can be deeper, the etch stop layer can be

thinner, and the overetch can be longer

provided with an application of photoresist to the backside, the deposition of back-

side etch-resistant protection layers, or use of special ﬁxturing that limits backside

exposure during the etch step.

Many factors should be considered for the appropriate selection of an etch-

resistant masking layer, as shown in Fig. 8.6. The predominant consideration is

whether an existing or standard process is adequate for the device. Because of

the cost and time required to set up a new etch sequence, the best approach is

usually a standard photoresist-based etch sequence including steps such as spin,

Standard Process

Easy Mask Removal

Improved Etch Resistance

Higher Selectivity

Lower Development Cost

Lower Unit Cost

Fewer Process Steps

Shorter Process Time

Deeper Etch

Self-Aligned Structures

Shorter Spin/Deposition Time

Better Step Coverage

Backside Protection

IC Compatible

Less Contamination

Higher Resolution

Better Feature Size Control

Hard

Mask

Fig. 8.6 Considerations and consequences of using a hard mask or a photoresist mask shows that

a PR mask is generally the mask of choice, except for situations where higher etch resistance, a

deeper etch, or substrate backside protection is needed

8 MEMS Wet-Etch Processes and Procedures 469

prebake, expose, develop, postbake, etch, rinse, dry, and strip. Basic photoresist

processes are generally IC compatible, offer less risk of contamination, provide bet-

ter resolution, and offer improved feature size control over a hard-mask sequence

that likely uses an initial photoresist patterning and etching sequence to form

the hard-mask features. Photoresist masking sequences, once installed in a facil-

ity, reduce development costs for a new etchant. Photoresist processing generally

provides lower unit costs, contains fewer process steps, has a shorter overall pro-

cess time, provides better step coverage in most situations, and spins onto the

wafer faster than coating the wafer with a thin-ﬁlm deposition for use as a hard

mask.

When needed, photoresist processes can often be adapted for MEMS device

requirements. For example, step coverage can be improved with a thicker photore-

sist layer simply by slowing the spin speed or selecting a different photoresist with

higher viscosity and solids content. Improved etch resistance with photoresist masks

can sometimes be achieved with an elevated postbake cycle or heavy UV expo-

sure for negative-acting resist. With careful handling and processing considerations,

photoresist can be spun on the backside of a wafer to provide any needed masking

or backside protection. Similarly, backside photolithography sequences can beneﬁt

from a frontside photoresist coating to provide etch resistance and scratch protection

during the etch sequence.

8.2.3 Direct Etching and Liftoff Techniques

Liftoff techniques may be invoked when a difﬁcult-to-etch material such as a noble

metal or a stack of several metal layers requires patterning, and direct etching is

unsuitable. Rather than spinning and patterning photoresist on top of the patterned

layer and etching the layer directly, the liftoff technique ﬁrst forms a patterned pho-

toresist layer on the substrate and then the patterned layer is deposited directly on

the resist such as by e-beam evaporation (see Fig. 8.7). After stripping away the

photoresist using, for example, a solvent in an ultrasonic bath, selected material in

contact with the substrate remains and the rest ﬂoats away or is rinsed, brushed,

or blown from the substrate. The liftoff process may result in jagged edges due to

thin deposited material on the photoresist s idewalls prior to liftoff. Enhanced liftoff

techniques use a two-layer resist system with a relatively thin, somewhat developer-

resistant upper layer above a lower layer that is less resistant. When the photoresist

system is exposed and developed, small overhangs emerge over the lower layer,

which subsequently shadow a sputtered or e-beam deposited ﬁlm. Note that the tone

of the photomask is reversed for the liftoff process (i.e., a dark ﬁeld mask is needed

where a light-ﬁeld mask would be correct for direct etching) or negative-acting

resist may be used. Alternatively, the enhanced liftoff technique can be implemented

effectively with a carefully selected thin-ﬁlm stack.

470 D.W. Burns

a) Direct Etching

b) Lift-Off

c) Enhanced Liftoff

Fig. 8.7 Liftoff techniques may be used in lieu of direct etching for materials that are difﬁcult

to etch or the unavailability of etchants with adequate selectivity to other exposed materials.

Conventional etching, as in (a), directly etches the exposed material, leaving the covered mate-

rial unetched. The general lift-off process places the to-be-removed material on top of patterned

photoresist and on the substrate as shown in (b), where the underlying photoresist is dissolved

away, requiring no direct etching of the patterned material. Material codeposited on the sidewalls

of the photoresist or liftoff layer can prove difﬁcult to remove cleanly and may result in unwanted

stringers and particulates. An enhanced liftoff process as in (c) has a composite photoresist layer

wherein the thin upper resist layer is more resistant to the developer than the underlying resist

layer; in this way, an overhanging region is produced to minimize sidewall deposition and improve

the robustness of the etch

8.2.4 Sacriﬁcial Layer Removal

Sacriﬁcial layer etching presents speciﬁc challenges that are normally secondary

considerations for direct wet or dry etching. Sacriﬁcial layers are positioned under-

neath structural layers, where features in the sacriﬁcial layer such as anchor

openings, dimples, sidewall proﬁles, ﬁlm thickness, and ﬁlm steps shape the

overlying structural layer. Selective removal of the sacriﬁcial layer is generally

accomplished with extensive lateral etching using a liquid etchant to undercut the

structural layer and leave portions or more freestanding. The etchant must have

extremely high selectivity to the sacriﬁcial layer relative to the structural layer and

other exposed materials for lengthy sacriﬁcial etches. Lack of an etchant with suit-

able selectivity restricts the attainable undercut distance and can limit the choices of

materials usable for the structural layer.

Successful sacriﬁcial etch processes generally have inherently high overetch

capability that allows the operator to ensure complete sacriﬁcial layer removal and

the designer to place features with short and long undercut times on the same device.

Removal of sacriﬁcial layers generally proceeds with relatively short downward

etching followed by extensive lateral etching. Lateral etching is generally isotropic

in the plane, etching in all directions equally. As described in Section 8.1, the etchant

can become exhausted with insufﬁcient reactants or become poisoned with too many

etch products, either of which can reduce the local etch rate. Design features should

be placed so that the undercutting frees up all portions of a structure nearly at the

same time; however, a robust etchant, sacriﬁcial layer and structural layer combina-

tion allows ﬁdelity in the formation of structures even with vastly diff ering undercut

8 MEMS Wet-Etch Processes and Procedures 471

time. During the etch s equence, anchored features need to stay anchored, with the

sacriﬁcial etchant unable to separate the structural layer from the substrate. Upon

completion of the etch, test structures or device features that can be inspected readily

validate the etch procedure.

Enhancements can be made to the s acriﬁcial etch process for improved perfor-

mance. For example, careful selection and abidance with design rules for etch holes

in the structural layer allow local access of the etchant to the sacriﬁcial layer and

reduces the time required to completely clear the sacriﬁcial layer. The sacriﬁcial

layer itself may be selected to increase the etch rate, such as a bilayer of fast-etching,

heavily doped deposited oxide over a thermally grown oxide. Wet etching in general

and sacriﬁcial layer etching in particular require consideration of the wafer backside

so that any deposited ﬁlms stay intact and do not ﬂake off during the sacriﬁcial etch

process.

8.2.5 Substrate Thinning and Removal

Wet etchants are often used in the thinning or selective removal of substrate mate-

rial (see Fig. 8.8), largely because of the relatively high volume of the etched

material to be removed and the high density of reactants in the liquid etchant com-

pared to reactants in a reduced-pressure plasma etching system. Uniform substrate

removal or thinning combines physical grinding mechanisms with chemical etching

in chemical–mechanical processes to rapidly remove material on either the backside

or the frontside of a wafer. Masked features, such as holes, vias, trenches, chan-

nels, and anisotropically etched diaphragms conﬁne the etch choice to selective

wet etchants or to high etch-rate reactive ion etchers. Wafer thinning or substrate

Fig. 8.8 Wet chemical etching can be used to (a) thin, (b) selectively etch from the backside,

or (c) selectively etch substrates from the frontside. Chemical–mechanical polishing or planariza-

tion (CMP) combines mechanical grinding with chemical etching to uniformly thin the substrate

backside or planarize the topside. Masked features for anisotropically etched diaphragms, through-

wafer holes and vias, or topside trenches and channels require a suitable etch mask and may be wet

etched or dry etched. The wafers become increasingly fragile and special care must be placed

during subsequent handling and dicing

472 D.W. Burns

etching is generally placed near the end of the process ﬂow, as the wafers become

increasingly fragile with thinning and have higher embedded costs.

8.2.6 Impact on Process Architecture

The choice of the type of wet etch has a direct impact on the process architec-

ture. As illustrated in Fig. 8.9, the process ﬂow can be generalized into a series of

masking sequences, each including a deposition step, a patterning step, and an etch

step. Although deposition steps clearly encompass sputtered, e-beam evaporated,

chemical vapor, or plasma-deposited ﬁlms, also falling into this category are ther-

mal oxidation, ion implantation, or acquisition of starting material. Patterning of

photoresist with photomasks or direct-writing equipment provides a suitable mask

for most etch processes, whether wet or dry. Hard masks provide increased etch

resistance where needed, although they themselves are generally patterned with a

photoresist sequence. During the deliberations for choosing the order and type of

etching processes, each mask sequence must be considered to ensure that nonstan-

dard materials are not included early in the process; that possible contamination

from etch tanks, holders, or etchants themselves is avoided; that restricted fume

hoods or other wet-etch facilities are used only after critical processing sequences

have been completed; that masking materials are limited where possible to pho-

toresist masks or to standard oxide, polysilicon, and metal layers for use as hard

masks; that structural layers be limited to standard materials when contamination is

a concern; that the masking material be readily stripped after etch completion; that

selected ﬁlm thicknesses be compatible with subsequent processing including pho-

toresist coverage; that step coverage after a ﬁlm has been etched does not generate

stringers or other sources of particles; and that wafer handling not be compromised

Mask Sequence

Deposit

Pattern

Etch

Mask Sequence

Deposit

Pattern

Etch

Mask Sequence

Deposit

Pattern

Etch

Fig. 8.9 Process ﬂows for ICs, MEMS/NEMS, or integrated MEMS/NEMS devices can be simpli-

ﬁed, with exceptions, into a series of masking sequences, each with a deposition step, a patterning

step, and an etch step. Complex CMOS and BiCMOS processes may have two dozen or more mask

sequences whereas many MEMS-only process ﬂows have fewer than ﬁve. Each masking sequence

is affected by the prior sequences and has an impact on the subsequent sequences. The order and

choice for each sequence is often a part of the device design. To achieve the desired result, the pro-

cess steps or the design may be varied. Decisions on the order of execution for wet-etch sequences

and their associated masking and material layers must consider a number of issues including con-

tamination, ﬁlm t hickness limitations, front- and backside processing requirements, and substrate

handling

8 MEMS Wet-Etch Processes and Procedures 473

by creating fragile structures too early into the overall ﬂow. This all being said, the

clever device designer and process engineer can often work together and ﬁnd com-

promises that achieve the design and production goals i n a cost-effective, technically

successful manner.

8.2.7 Process Development for Wet Etches

When considering development of a new wet-etching procedure: avoid it if at all

possible. A new etch process can involve setting up new facilities, installing new

equipment, setting up new chemical handling and safety procedures, obtaining

purchasing approvals, determining chemical storage and disposal, establishing train-

ing procedures, performing operator training, providing system maintenance, and

spending a large amount of characterization time with sometimes expensive wafers.

Adapting an existing etch procedure, or better yet using an existing etch procedure,

is a highly recommended alternative. In many established laboratories or fabrication

facilities, a wide array of processes and materials for photoresist and hard masks are

available for use or adaptation. Hard masks are sometimes necessary, however, it is

best to consider and use a photoresist mask whenever possible.

In cases where a new etch is needed or an existing etch needs adaptation, guide-

lines for etching procedures and safety considerations are available from a variety

of sources including handbooks, user guides, equipment manuals, manufacturer’s

recommendations, etchant suppliers, photoresist manufacturers, the Web, university

websites, publications, and expert knowledge from an experienced user or process-

ing expert. Appreciable effort is required to set up an etch facility with suitable

services, adequate ventilation, appropriate disposal means, and controlled access.

Several etch stations may be needed in a facility to accommodate the needs of dif-

ferent users and process ﬂows. Facilities should provide ready access to goggles,

facemasks, aprons, and spill carts with established safety procedures and chem-

ical handling in case of spills. Wet benches are outﬁtted with suitable beakers,

tanks, holders, and cassettes, with dedicated labware for contamination control.

PFA, PTFE, or quartz labware are generally the best choice, although other mate-

rials may serve for room-temperature etchants. Etch sequences may be executed in

heated tanks or in beakers on hot plates, although etch processes that run at room

temperature are highly favored. Etch-process development generally requires plenty

of wafers to establish etch rates of the etch layer and potential masking layers, to

determine selectivity to other materials in the wafer stack, and to perform sensitiv-

ity studies with changes in etchant concentration, solution temperature, and other

variables. The amount of rinsing may need to be qualiﬁed, with aqueous etchants

preferred over solvents. A run-traveler template and operating procedures should

be ﬁnalized and documented. Light sensitivity for the etch should be checked, and

operation in a tank with a lid may be needed. Agitation capability may be desired for

improved etch uniformity. Adequate removal of photoresist and hard masks must be

veriﬁed.