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

Подождите немного. Документ загружается.

1044 P. Lin e t al.

263. C. Gui, G.J. Veldhuis, T.M. Koster, P.V. Lambeck, J.W. Berenschot, J.G.E. Gardeniers, and

M. Elwenspoek: Nanomechanical optical devices fabricated with aligned wafer bonding,

Eleventh Annual International Workshop on Micro Electro Mechanical Systems, MEMS,

Heidelberg, Germany, 482–487 (1998)

264. C. Kourouklis, T. Kohlmeier, H.H. Gatzen: The application of chemical–mechanical polish-

ing for planarizing a SU-8/permalloy combination used in MEMS devices, Sens. Actuators

A 106, 263–266 (2003)

Chapter 14

MEMS Process Integration

Michael A. Huff, Stephen F. Bart, and Pinyen Lin

Abstract In this chapter we provide a concise understanding and appreciation of

MEMS process integration – the technique by which processing steps are com-

bined into an ordered sequence with the goal of creating a methodology that

allows the production of functional MEMS devices with acceptable yields and cost

levels. Process integration is one of the most important topics in MEMS technol-

ogy because the choices made about how to produce a MEMS device have a large

impact on the success or failure of any MEMS-related product development activity

or business venture. This chapter describes why process integration is important in

MEMS, what it entails, and how it is done. A review of the similarities and dif-

ferences between MEMS and the related integrated circuit technology is provided,

along with some guidance on how MEMS process sequences are successfully devel-

oped. We also review many of the most noteworthy and important MEMS process

sequences and technologies that have been developed to date, showing examples

both with and without microelectronics integrated with MEMS. In addition to being

a general reference guide, these wide ranging examples can bolster a designer’s

experience base by illustrating MEMS process integration strategies that have been

successfully demonstrated. A general strategy for MEMS cost analysis is also pro-

vided. We hope that we can give the r eader at least a sufﬁcient understanding to

appreciate the importance of this topic with the “real-life” examples of process

integration to the extent possible. The important process integration issues such as

design for manufacturability is also highlighted in this chapter.

14.1 Introduction

This chapter is the culmination of the preceding chapters of this book and is

directed at the extremely important topic of process integration. Process integration

is the technique by which processing steps are combined into an ordered sequence

M.A. Huff (B)

MEMS and Nanotechnology Exchange, Reston, VA, USA

e-mail: mhuff@mems-exchange.org

1045

R. Ghodssi, P. Lin (eds.), MEMS Materials and Processes Handbook,

MEMS Reference Shelf, DOI 10.1007/978-0-387-47318-5_14,



Springer Science+Business Media, LLC 2011

1046 M.A. Huff et al.

with the goal of creating a methodology that allows the production of functional

MEMS devices with acceptable yields and cost levels. In general, process integra-

tion involves the selection of processing steps to be used in device fabrication. But

it also entails the understanding, characterizing, and optimizing of these steps and

their interrelationships with the goal of having them effectively work together as a

standardized production method to meet a speciﬁc acceptable performance, yield,

and cost level. Process integration is one of the most important topics in MEMS

technology because the choices made about how to produce a MEMS device have

a large impact on the success or failure of any MEMS-related product develop-

ment activity or business venture. This chapter describes why process integration

is important in MEMS, what it entails, and how it is done. It also reviews a num-

ber of notable examples of successfully developed MEMS process t echnologies and

presents some of the economic issues involved. Realistically, it is nearly impossible

to teach a reader how to become a “process integration expert” in only one chap-

ter of a book. However, our hope is that we can give the reader at least a sufﬁcient

understanding to appreciate the importance of this topic as well as highlight some

of the important process integration issues that occur in any MEMS development

endeavor. Our approach is to use “real-life” examples of process integration to the

extent possible.

14.2 What Is Process Integration?

It is best to begin by establishing some deﬁnitions in order to help clarify the ter-

minology. We have already deﬁned what the term process integration means in the

introduction above; let’s now expand our vocabulary.

A processing step is deﬁned as a single process that is performed on a substrate or

a set of substrates in order to advance the fabrication of some device. This book has

provided the reader with a review of a number of examples of processing types and

capabilities in the previous chapters including depositions, etches, photolithogra-

phy, electrochemical plating, and so on. An example of a processing step under this

deﬁnition would be the deposition of polycrystalline silicon (polysilicon) on a batch

of wafers using LPCVD. More exactly, the processing step would be limited to only

the actual deposition of the polysilicon thin-ﬁlm material on the wafers and not to

the cleaning and preparation steps performed on the wafers prior to the deposition

or the metrology that would be performed after the deposition was completed. The

wafer cleaning and preparation steps as well as the metrology steps would rightly

be considered processing steps on their own. Having s aid that, one must be careful

because these distinctions are often made somewhat arbitrarily by many fabricators

and it is common to see several steps grouped together and labeled as a process s tep.

Using the example from above, it would be common practice to see the individual

process steps of cleaning, deposition, and metrology lumped into a group and then

called a process step.

14 MEMS Process Integration 1047

A process module is deﬁned as a grouping of processing steps that have

been assembled into an ordered sequence and which have been characterized and

standardized, and are reused on a periodic basis in the production of devices, but

which is not sufﬁcient in and of itself to result in functional devices. Therefore,

under this deﬁnition a process module could be applied to anything ranging from

two or more process steps up to any number of processing steps just shy of the

complete fabrication of a device. Building on the example of polysilicon depo-

sition illustrated above, the term “process module” could be used to describe

the collection of the processing steps including cleaning and preparation of the

wafers, followed by the deposition of polysilicon on the wafers, followed by the

performance of metrology on the wafers to measure the thickness and index of

refraction of the thin-ﬁlm polysilicon, if this collection of processing steps or pro-

cess module were used on a repeated basis for the production of devices. In practice

most process modules in foundries involve the grouping of several major process-

ing steps (e.g., deposition of polysilicon, or photolithography, etc.) and therefore

would be much more complicated than merely a clean, deposition, and inspection

cycle.

A process sequence is deﬁned as a grouping of processing steps that have been

assembled into an ordered sequence and which have been characterized and stan-

dardized, and are reused on a periodic basis in the production of devices, and which

is sufﬁcient to result in functional devices. Obviously, the distinction between a pro-

cess module and process sequence is a matter of degree. A process module does not

result in a functional MEMS device, whereas a process sequence does.

The term “process sequence” is frequently used interchangeably with process

technology. For our purposes, the distinction between a process sequence and a

process technology relates to its maturity, particularly in the commercial sense.

A process technology is deﬁned as a well-developed and standardized process

sequence, which is used in the commercial production of devices. The term is typ-

ically applied to a process sequence developed by a research and development or

academic fabrication laboratory, whereas a process technology would be used to

describe the manufacturing method used to produce speciﬁc MEMS devices in a

commercial foundry. Additionally, it is frequently the case that a process technology

has an associated set of design rules and technology ﬁles that will aid in developing

a suitable design and mask set.

Figure 14.1 is an example of a very simple process sequence for fabricating

a polysilicon cantilever that is held to the substrate at one end and free on the

other. This would be classiﬁed as a “polysilicon surface micromachining process

sequence.” This illustration shows a cross-section of the wafer at various impor-

tant points in the process sequence. These types of visual aids are extremely

useful for designing a process for any MEMS device and are commonly used by

MEMS designers, particularly in the early stages of a development effort. These

types of illustrations are also frequently used to describe many of the example

process sequences that we review later in this chapter. As can be seen, the ﬁrst

step (Fig. 14.1a) is describing the starting wafer as a single crystal silicon wafer.

Subsequently, a thermal oxide is grown on the surface of the wafer (Fig. 14.1b).

1048 M.A. Huff et al.

(a) Starting wafer is single crystal silicon wafer.

(b) A silicon dioxide layer is deposited or grown

on the surface of the wafer.

layer is etched to expose areas on the surface

of the wafer. Photoresist is removed.

(d) A layer of polycrystalline silicon

(polysilicon) is deposited on the

surface of the wafer. Photolithography

is performed and the polysilicon

is etched. Photoresist is removed.

(e) The oxide layer is removed by etching.

(f) A micromachined capping wafer of silicon

is wafer bonded to the device wafer to

hermetically seal the micro-structure in a

protective cavity.

PolysiliconSilicon Silicon Dioxide

Fig. 14.1 Example of a very simple surface micromachined process sequence surface

Next, this thermal oxide layer is patterned and etched using photolithography

(Fig. 14.1c) to open up areas in the oxide to expose the underlying silicon wafer

surface. A layer of polycrystalline silicon (polysilicon) is then deposited on the

surface of the wafer and this layer is patterned and etched in the shape of the can-

tilever (Fig. 14.1d). The oxide is then removed from the wafer (Fig. 14.1e) and

a micromachined capping wafer i s wafer bonded to the device wafer. The capping

wafer has cavities cut into its surface that will encapsulate the polysilicon cantilevers

(Fig. 14.1f).

The description of the above process is highly simpliﬁed and leaves out many

of the important details, including: thickness of the deposited layers; depth of the

etches; type of photolithography process performed to pattern the layers; what type

of mask was used, or whether a mask was even used at all; as well as the criti-

cally important process step parameters. Process step parameters include the exact

details of the equipment that is used to perform the process step, the temperature

and chemistries used in the process step, and so on. Obviously, the fabrication of a

MEMS device has many levels of detail that must be speciﬁed before implementa-

tion can begin. Process sequences are commonly deﬁned in what are called process

runcards or process travelers. These are documents that deﬁne the processing steps

and modules for the fabrication of a device. The level of detail on runcards is usually

sufﬁcient to capture a high level of detail, such as process steps employed (including

the identiﬁer of the process recipe), thickness of layers, equipment that the process

step is performed on, speciﬁcation of the mask and wafer to be used, and the like,

14 MEMS Process Integration 1049

but would not provide a complete and exhaustive level of detail. The higher level of

detail would be implied by the speciﬁcation of the process recipes and the tools to be

used and these would usually be documented elsewhere. In a production-level facil-

ity, this hierarchy of documentation would be kept on a computerized manufacturing

support system that would be integrated with the production equipment.

Figure 14.2 shows an example of a process runcard for the polysilicon surface

micromachining process sequence described above. As can be seen, it starts with

a speciﬁcation of the masks to be used in the process sequence, details the wafer

descriptions to be used, and then lists each of the processing steps in the process

sequence, numbered Step 1 through Step 11. Some of the processing steps, such as

Steps numbered 1, 2, 5, 6, and 10, are actually metaprocess steps and are composed

of several individual processing steps, such as 2.1 through 2.10. A document such

as this would be conveyed from the MEMS designer to the foundry to specify the

fabrication of any MEMS device.

Fig. 14.2 Example of process runcard or process traveler for the simple example of the polysili-

con surface micromachined cantilever surface (Reprinted with permission, copyright MEMS and

Nanotechnology Exchange)

1050 M.A. Huff et al.

14.3 What Is an Integrated MEMS Process?

Process integration, as deﬁned in the previous section, does not have the same

meaning as an integrated process.Anintegrated MEMS process is one wherein

the MEMS device(s) is (are) fabricated on the same substrate as microelectronics.

Many MEMS devices require some type of electronics interface in order to be func-

tionally useful. So the natural question is: when does it make sense to produce a

MEMS device and microelectronics on the s ame wafer? This is a difﬁcult question

to answer, but always depends on whether such integration provides the best sys-

tem solution in terms of performance and cost. There are other issues as well, such

as the magnitude of the development cost and the length of time associated with

developing a manufacturable integrated MEMS process technology.

In cases where the system-level tradeoffs do point to the need for merging MEMS

and electronics on the same substrate, an integrated process must be designed and

developed. At ﬁrst blush this may seem relatively straightforward, however, the

complexities of such integration are quite signiﬁcant and often drive the devel-

opment cost up very signiﬁcantly. Some of the key difﬁculties in developing an

integrated process technology are topography and photolithography accuracy, layer-

to-layer alignment, materials incompatibility, and thermal budget problems (see

discussion below). In any case, this is a fundamentally important system-level

question of MEMS design. And although there is no single answer because each

individual case depends on performance, size, cost, and packaging requirements,

among others, of that particular product and market, we examine some of these

issues and tradeoffs to give an appreciation of the important decisions related to the

partitioning of the system design.

We show several examples of both nonintegrated and integrated MEMS process

sequences later in this chapter, as well as examine how to develop a production-cost

estimation for these different scenarios. With the matter of deﬁnitions established,

let’s now discuss the importance of process integration.

14.4 Differences Between IC and MEMS Fabrication

MEMS technology has its roots in and heavily leverages the fabrication technolo-

gies developed by the integrated circuits industry and therefore is often associated

with IC technology. This association is valid because MEMS and IC share common

attributes related to some of the fabrication methods and materials used. However,

there are several very important differences between IC and MEMS fabrication

and reviewing these differences will help to better understand some of the process

integration issues related to MEMS.

The ﬁrst difference is that the microelectronics domain has converged to a rel-

atively small number of “standardized” or “ﬁxed” process technologies, such as

CMOS (complementary metal–oxide semiconductor), bipolar, BiCMOS (bipolar

and CMOS), and so on, for making integrated circuits. Although each manufacturer

usually has a proprietary process technology for their ICs, in reality there are very

14 MEMS Process Integration 1051

few and they have very subtle differences between them. In comparison, the MEMS

domain tends to use a totally different process sequence for each device type. That

is, the sequence of processing steps to implement an accelerometer may have little

in common with a process technology to implement a lab-on-a-chip. This is partly

due to the fact that microelectronics primarily has only three basic device types,

namely transistors, resistors, and capacitors, which are wired together into circuits

using metal interconnect layers. In comparison, MEMS technology has an extremely

large number of different types of devices. Some examples include: pressure sen-

sors, accelerometers, gyroscopes, vibration sensors, inkjet heads, magnetic sensors,

microvalves, micropumps, optical switches, modulators, and so on. It is important to

note that each MEMS device usually requires its own specialized process sequence

in order to meet the performance requirements for a given application. As a result,

there is a tremendous amount of customization involved in the fabrication of MEMS

devices and, therefore, process integration is an extremely important issue in the

implementation of MEMS.

Another major difference between MEMS and IC fabrication is the vast array of

processing capabilities and materials used to make MEMS devices. MEMS fabrica-

tion leverages conventional process capabilities borrowed from the IC world such as

oxidation, LPCVD, and photolithography and combines these processes with highly

specialized “micromachining” techniques. These micromachining techniques have

been reviewed in previous chapters of this book and include bulk micromachining,

surface micromachining, wafer bonding, and LIGA, among others. In fact, each of

these micromachining techniques actually includes a very broad spectrum of differ-

ent processing technologies depending on the materials used and exact ordering of

the processing steps.

Consequently, MEMS developers have a much richer range of choices regard-

ing materials and fabrication techniques than is commonly available in the IC

domain. This has many advantages, but also poses some challenges as well. A

major advantage is that MEMS developers have considerably more design and pro-

cessing freedom to obtain outstanding device performance. On the other hand, the

MEMS developer has comparatively little in the way of standard process technolo-

gies from which to leverage. This means that the MEMS designer must choose from

an enormous range of process options without a priori knowledge of exactly which

processing steps will work best for the fabrication of a given device, not to mention

how to integrate these processing steps into a viable process sequence. Moreover,

MEMS designers very often need to develop from scratch one or more of the indi-

vidual processing steps used in the process sequence. In short, there is an enormous

amount of customization that routinely takes place in MEMS fabrication, which

results in a tremendous number of variables that must be understood and controlled

in order to develop a process sequence that meets the performance requirements,

provides an acceptable yield, and meets the product cost goals. Most MEMS devel-

opment efforts are necessarily focused on the development of working devices in the

early stages of a venture, which means that the efforts are really directed at devel-

oping a viable process sequence that allows devices to be made that meet the design

performance and cost objectives.

1052 M.A. Huff et al.

Another signiﬁcant difference between MEMS and IC fabrication (and one that is

extremely important to the success of a business venture) relates to the scale of pro-

duction. Speciﬁcally, the production volumes in the t wo technologies are often quite

different. The volume of IC wafers typically produced in a commercial foundry each

year is truly enormous. ICs are produced using batch-fabrication and continuous

ﬂow manufacturing techniques whereby large numbers of wafers are processed in

an identical fashion (i.e., a ﬁxed process technology such as CMOS) and each wafer

may have hundreds or even thousands of individual, but identical, integrated circuits.

The advantage of batch fabrication is that the cost is spread over millions of die,

and therefore large economies of scale can be obtained thereby allowing a com-

paratively low die cost. High volumes are essential for IC manufacturing because

the investments that must be made in order to enable state-of-the-art IC manufac-

turing are now in excess of several billion dollars. Consequently, there is a strong

incentive in IC production to move toward larger diameter substrates as well as

smaller devices in order to lower t he cost per die. The industry is now at the 300 mm

diameter level with linewidths of tens nanometers. High volumes are typical in IC

manufacturing because microelectronic circuit process technologies are sufﬁciently

generic to cover a wide number of applications (e.g., TSMC’s foundry model) or the

production volumes (and margins) are sufﬁciently high (e.g., Intel’s microprocessor

production) to recoup the investment and still make a proﬁt.

For MEMS the situation is most often quite different. Typically, the volumes in

MEMS are inherently lower due to the specialized nature of the speciﬁc devices.

For example, a MEMS inertial sensor designed for automobiles as a crash air bag

deployment sensor is very useful for that speciﬁc application, but the process tech-

nology used to make this device may not be suitable for most other applications. As

a consequence of the specialization of MEMS devices the overall market size for

a given device type, as measured by the number of devices manufactured annually

to satisfy the market need, is frequently quite small in comparison to microelec-

tronic production volumes. Indeed, many MEMS devices have market sizes that

are very small (e.g., hundreds or thousands of devices per year). Therefore, the

production volumes in MEMS vary over a huge range. As a general rule, MEMS

devices with higher production volumes tend to be manufactured using essentially

the same methodology used in IC production, namely batch fabrication and contin-

uous ﬂow manufacturing techniques, whereas MEMS devices with low production

volumes are usually manufactured using job-shop production methods. Clearly, a

good understanding of the production volumes required to satisfy the expected mar-

ket demand as well as the approach taken for the development and production of

MEMS devices (e.g., outsource or insource) are critically important.

14.5 Challenges of MEMS Process Integration

There are substantial challenges associated with MEMS process integration.

Perhaps most important, the properties of materials used in MEMS fabrication,

particularly their mechanical properties, can be highly dependent on the process

14 MEMS Process Integration 1053

sequence as well as the speciﬁc processing conditions. It is well known in

microelectronics fabrication that the electrical properties of materials used must

be very well controlled to achieve good transistor device performance. However,

in order to achieve good performance in MEMS devices, mechanical as well as

electrical, and potentially other material properties (e.g., optical, thermal, etc.) must

also be tightly controlled. This can be particularly challenging for thin-ﬁlm materi-

als used in MEMS fabrication because t he properties of these materials can have a

very large impact on the resultant performance of MEMS devices. Moreover, these

materials will typically possess a range of values in the “as-deposited state” depend-

ing on the exact processing conditions (i.e., deposition rate, temperature, etc.) used

during the deposition of the thin-ﬁlm material.

Consequently, if a thin-ﬁlm layer used in a MEMS device is in the as-deposited

state, then knowledge of the properties of that material layer would be required,

but may not be available until they are measured. It is also important that these

material properties be reproducible. Models for the properties of these materials

are extremely useful in this regard. Additionally, it is often the case that more

processing will be performed after a thin-ﬁlm layer has been deposited. Any subse-

quent processing steps can signiﬁcantly modify the material properties of thin-ﬁlm

materials. That is, the processing steps performed after the thin-ﬁlm was deposited

can substantially change the material properties from their as-deposited values.

Moreover, the amount of change in these material properties can vary depending

on the nature and type of the subsequent processing steps performed. Obviously,

this makes it extremely difﬁcult to predict the properties of these materials with any

reasonable level of certainty unless the exact same process sequence has been previ-

ously characterized (i.e., the material properties are known for the process sequence)

and is reproducible. Consistent control of these effects is often critical and therefore

must be part of the process integration method used from the outset.

Another challenge that must be addressed is to determine how and where to

develop and manufacture a MEMS device. Because of the enormous number of

MEMS fabrication methods (including processing technologies and materials), it is

virtually impossible for any single MEMS foundry to have all of the needed methods

within their capabilities. Most microelectronics foundries provide one-stop shop-

ping for all of the frontend processing (i.e., fabrication up to packaging and test)

such that everything needed to implement an IC according to a standardized process

technology resides within a single foundry. On the other hand, the fabrication core

competencies in the MEMS ﬁeld tend to vary, and sometimes quite signiﬁcantly,

from foundry to foundry.

Given the enormous diversity in materials and processes, it is prohibitively

expensive for individual MEMS foundries to have all the techniques, and therefore

the design and process freedom at any one MEMS foundry can be constrained. To

overcome this problem, distributed fabrication, whereby the wafers are processed by

multiple foundries, is becoming more widespread in the MEMS industry. For exam-

ple, it is becoming more commonplace for the production of integrated MEMS to

form a split-manufacturing scenario between an IC and MEMS foundry wherein

the individual foundry capabilities complement each other. A CMOS foundry