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

Подождите немного. Документ загружается.

4 T.L. Lamers and B.L. Pruitt

1950

1960

1970

1980 1990 2000

1947:Transistors

Bardeen, Brattain, Shockley

(Bell Labs)

1952: Zone-refining process

Pfann (Bell Labs)

1952: Dopant diffusion

Fuller (Bell Labs)

1955: Ion Implant

Cussins

1956: Nobel Prize in Physics

Bardeen, Brattain, Shockley

1958: First IC

Jack Kilby

(Texas Instrument)

1959: Planar Transistor

Jean Hoerni (Fairchild)

1961: First silicon IC device

Noyce (Fairchild)

1967: Anisotropic

etching of silicon

Finne and Klein

1968: Intel was founded

Moore and Noyce

1968: Anodic Bonding

Pomerantz (US Patent)

1981: Electrochemical

P-N junction etch stop

Jackson, et al.

1974:oxidation enhanced

diffusion effect in silicon

1982: Silicon as a

Mechanical Material

Petersen, IEEE Proc.

1985:Wafer bonding

technology for SOI

Lasky, et al.

1986: Silicon-to-silicon

direct bonding method

Shimbo, et al.

1996: Bosch DRIE Process

Laermer and Schilp (US Patent)

1954: Piezoresistance

in Ge and Si

Smith (Bell Labs)

1958-60: Commercial

silicon strain-

gauge, metal

diaphragm pressure sensors

Kulite, Microsystem

s, and

Baldwin Lima Hamilton

1961:Transverse and

shear piezoresistance

Pfann and Thurston

(Bell Labs)

1962: Silicon diffused-

element piezoresistive

pressure sensor

Tufte, et al.

1963: Piezoresistive properties

of diffused silicon

Tufte and Stelzer (Hone

well)

1966: Integration of

piezoresistive strain gauge

into an accelerometer

Wall

1968: Diffused-element

piezoresistive cantilever

Wilfinger (IBM)

1977: First silicon-based

pressure sensor using ion

implantation technique

Marshall (USPatent)

1979: Integrated piezoresistive

accelerometer

Roylance and Angell

1982: Graphical piezoresistance

coefficient in silicon

Kanda

1991: Piezoresistive floating-

element shear stress sensor

Ng, et al.

1993: Non-linear piezoresistance

effect in Si

Matsuda

1993: Piezoresistive cantilever for

Atomic Force Microscopy (AFM)

Tortonese, et al.

1998: Dual-axis piezoresistive

cantilever using sidewall ion

implant technique

Chui, et al.

2000: 1/f noise consideration in

design of piezoresistive cantilevers

Harley and Kenny

1988: Pressure sensors

using Si fusion bonding

Novasensor

1954: Mesa Transistor

Lee (Bell Labs)

Fig. 1.2 Timeline of some technological advances in fabrication (above the horizontal line) and micromachining, particularly as pertains to piezoresistive

sensing (below the horizontal line)[1, 2, 5–33]. Modiﬁed ﬁgure, used with permission of IEEE. First appearing in Barlian et al. [34]

1 The MEMS Design Process 5

1.1.1 Design Process

The design process begins with deﬁning product requirements for the MEMS

device. These requirements are determined through interviews and surveys of cus-

tomers and users, as well as reviews of competitive products, and are deﬁned

in terms of customer speciﬁcations. Ultimately, quantitative metrics for the

speciﬁcations are needed to measure and compare product performance against the

requirements. The ﬁrst step of describing the customer and deﬁning the market can

be challenging for a MEMS designer. A ﬂowchart of the design process begins with

the product deﬁnition phase and requires customer and market knowledge (Fig. 1.3).

Very different requirements will be voiced depending upon whether the MEMS

device is targeted as a component in a consumer product (e.g., accelerometers or

microphones in Smartphones), or as an enabling technology integrated in a larger

system (e.g., Polychromix PhazIR

portable near-infrared spectrometer systems

leverage software, hardware, and custom chips) to deliver an integrated, portable

material inspection tool.

The product deﬁnition phase is often driven by marketing and sales teams who

interface directly with consumers. However, design, process, and test engineers

best know the in-house and supply chain capabilities, can readily deﬁne quanti-

tative process and performance metrics, and can evaluate how these trade off with

customer requirements. Thus, the product deﬁnition phase should involve all stake-

holders in this product and customer/market deﬁnition phase. Once participants

agree on the customer and market, the “voice of the customer” in terms of “soft”

customer wants and “hard” measurable performance metrics can be related and

ranked by importance. The design team then knows where to focus the most effort

to achieve a product that best meets the most customer requirements, usually trad-

ing off cost with complexity. Quality function deployment (QFD) is a tool that

formalizes this process, as explained in more detail later in this chapter.

A separate analysis of requirements and their relative weights should be com-

pleted for each customer and market. With these customer requirements, the design

team can brainstorm concepts that simultaneously consider materials, processing,

and packaging (Chapters 2–13 provide guidance on coupled material and process

selection as well as packaging options). Concepts with geometric and material prop-

erty detail are analyzed for predicted performance and the design can be reﬁned

based on results from analytical, numerical, or ﬁnite element models using data

from in-house processes or the literature. Models for the performance of common

classes of MEMS transducers are available elsewhere [35–37]. Concept screening

(CS), a tool to streamline the process of selecting a design idea to prototype, is

demonstrated later in this chapter.

Designers should take care to build in test structures and checkpoints in the

process to more accurately evaluate process-dimensional tolerances and result-

ing material properties of as-fabricated structures. The overall performance of

the device or system should be carefully considered with signal conditioning and

systems integration (see Chapter 14). MEMS commercial products use both comple-

mentary metal oxide semiconductor (CMOS) integrated signal conditioning on the

6 T.L. Lamers and B.L. Pruitt

Fig. 1.3 Design ﬂowchart. The product deﬁnition phase begins with deﬁning and understanding

the customer(s) and market. Structured design methodologies guide the translation of customer

requirements into design features and concepts and allow the designer to simultaneously con-

sider process, material, and performance in the design. The design process is not linear and short

loops on materials and process selection should feed back frequently through stakeholder design

reviews

1 The MEMS Design Process 7

same chip as the MEMS structures and also application-speciﬁc integrated circuits

(ASICs) in multichip modules for signal conditioning. Usually, the impact of reli-

ability and yield on overall size and cost requirements drive this decision. At each

stage of the process, short loops on critical design parameters (materials, dimen-

sions, interfaces, processes) should be investigated with appropriate test wafers and

test structures before entire lots of wafers are processed. This enables the design

team to revisit aspects of the design before an entire mask set or entire fabrication

run is committed.

The rest of this chapter provides more detail and case studies demonstrating the

ﬂowchart steps in Fig. 1.3.

1.2 Design Methods for MEMS

Many examples demonstrate the beneﬁts of using design methods [38–41], and

design methods are commonly applied in industries from automotive to aerospace

to semiconductors. Yet, design methodologies have less frequently been applied to

MEMS products [42, 43].

1.2.1 History of Design Methodologies

One of the key design methods used in product development is quality function

deployment (QFD). QFD is a tool that helps relate what is important to the cus-

tomer to decisions made in the design and manufacturing process. Mapping the

customer requirements to the metrics and speciﬁcations the design team uses helps

in the creation of a product that satisﬁes customer desires, resulting in more focused

development and higher sales volumes.

QFD was ﬁrst used in 1971 in the Kobe shipyard of Mitsubishi Heavy Industries

[38]. In 1972, Dr. Yoji Akao published an article on what he called Quality

Deployment and is now called QFD [44]. Akao had been developing and teaching

his method for several years prior to the publication of the article. The Kobe ship-

yard work was published by Nishimura shortly after Akao’s ﬁrst article [45]. Over

time, the demonstrated results of using QFD encouraged others to adopt the tool

and usage became widespread in Japan. Toyota and its suppliers continued devel-

opment of QFD to help with their focus on customer values. Through the 1970s

and early 1980s, Japanese companies used QFD to improve the effectiveness of

their product development [38, 46]. In the mid 1980s, Xerox and other U.S. com-

panies started to use QFD along with design for assembly (DFA) as part of their

drive to regain competitiveness in U.S. manufacturing. Primarily heavy industry

companies, such as automobile manufacturers, adopted QFD, with limited adoption

into high-technology ﬁelds such as semiconductors. QFD has typically been applied

to products as opposed to manufacturing processes. Professor Don Clausing at MIT

helped champion QFD in the United States [46], and some companies such as ITI

even created QFD software [38].

8 T.L. Lamers and B.L. Pruitt

QFD is meant to provide guidance in product development. As such, it should

be used with other management and technical tools, such as strategy planning, rapid

prototyping, design of experiments, and design for assembly, to produce an effective

development project. The output of QFD still needs to be subject to human judgment

and is not meant to automate the design process [47].

The effectiveness of QFD has been documented in many sources over the past 30

years. A few examples noted by ReVelle, Moran, and Cox [38] include:

• A Brazilian steel company, which used QFD on the components it produced for

automotive suspension springs, experienced a market s hare increase of 120%, a

fall in customer complaints by 90%, and a 23% reduction in production costs

between 1993 and 1998.

• Motorola America’s Parts Division, where customers reported being 60% more

satisﬁed with their product and pricing information after QFD principles were

used to improve the system.

• The Wiremold Company used QFD for its new product development starting in

1991. By 1994, new product development time had fallen from 24–30 months

to 6–9 months, enabling the introduction of six to eight times more products per

year.

• Toyota Auto Body Company in Japan reduced startup costs for the introduction

of four van models by 61% over a 7-year period, while simultaneously reducing

the product development cycle time by one-third.

QFD is a multiple-stage tool, although frequently only the ﬁrst one or two parts

are used on a given product development effort. The stages of QFD are commonly

referred to as houses. The houses are pictorially shown in the top part of Fig. 1.4.

House I, sometimes referred to as the House of Quality, translates the “soft” cus-

tomer wants into measurable engineering attributes or speciﬁcations. House II then

takes these engineering characteristics and maps them to parts characteristics, or

attributes of the product’s parts. House III matches up parts characteristics to process

operations, and House IV aligns process operations with production requirements.

1.2.2 Structured Design Methods for MEMS

Existing design methods are not always well suited to the MEMS product struc-

ture or development process. In the case of QFD, some of the difﬁculty of using

the tool for MEMS comes in QFD Phase II, where engineering metrics are linked

to parts. Most MEMS do not have physical “parts” that are assembled into a ﬁnal

device, but instead have product speciﬁcations and a manufacturing process used

to create the product. The physical form is determined through the fabrication pro-

cess, demonstrating a tight linkage between the MEMS product and process. MEMS

also have more integrally linked material functions with process and speciﬁcations

than many other types of products. For example, silicon comes in several crys-

talline forms such as amorphous, polysilicon, and single-crystal silicon, each with

distinctly different properties. The desired end product speciﬁcations and physical

1 The MEMS Design Process 9

Phase I

Engineering Metrics

Customer

Requirements

Phase II

Parts

Characteristics

sEngineering Metrics

Phase III

Key Process

Operations

Parts Characteristics

Phase IV

Production

Requirements

Key Process

Operations

QFD

Phase I

Engineering Metrics

Customer

Requirements

Concept

Screening

Engineering

Metrics

Design

Concepts

Fig. 1.4 (a) Houses of QFD, after Hauser and Clausing [46]and(b) QFD Phase I plus concept

screening to enable consideration of design of process and devices simultaneously for MEMS.

Modiﬁed ﬁgure, used with permission of K.L. Lamers. First appearing in Lamers [43]

form have large ramiﬁcations on which fabrication process and material variety are

chosen, and vice versa. The tight link between product and process was utilized

in creating concept screening, a tool that relates engineering metrics to design

concepts, including product conceptualization and manufacturing process. Concept

screening is a simple matrix tool that allows designers to evaluate design concepts

against important engineering metrics. The tool combines portions of QFD Phases

II and III into one phase to make the tool more readily applicable to MEMS. It also

contains aspects of Pugh concept selection [48], and differs from typical QFD pri-

marily in consideration of product idea and manufacturing process together in the

early phases of product deﬁnition. The design methods recommended for MEMS

are shown in the lower portion of Fig. 1.4.

QFD Phase I and concept screening are demonstrated in this chapter as design

methods well suited to MEMS. Additional methods, such as brainstorming and

materials selection, are utilized in concert with QFD and CS. Using methods to

be better aligned with the unique requirements of MEMS may lead to improved

MEMS product development. Improvements in MEMS product development may

be measured through a decrease in time to market, fewer development iterations,

reduced development or product costs, or increased product revenue.

1.3 Brainstorming

Brainstorming is an idea generation activity involving the key stakeholders in prod-

uct deﬁnition, development, and implementation [49]. It may be carried out in one

10 T.L. Lamers and B.L. Pruitt

session or it may be done over several sessions covering multiple topics, such as user

requirements, technical speciﬁcations, and design concepts. In a corporate setting,

the constituents participating in brainstorming may include engineers for devices,

packaging, testing, manufacturing, marketing, and sales. Brieﬂy, the brainstorming

session process might be comprised of: a moderator (project leader) who gathers

the key participants in a comfortable space with ample whiteboard space, mark-

ers, Post-its, or other means to record and move information around, keeping ideas

visible to participants. If the group does not regularly work together, the moder-

ator should lead icebreaker or warm-up activities to familiarize participants with

each other and encourage free-thinking and communication. The moderator then

describes the objective(s) of the exercise with the key requirements (high-level) for

the product; the moderator ensures all ideas are captured. Some brainstorming rules

the moderator must enforce include:

• No idea is a bad idea (try saying “yes, and...”, then morphing and building to

other ideas).

• Be creative and take risks (above all, keep ideas rolling).

• No criticism allowed (no “buts, excepts, or howevers”).

At the end of the brainstorming session, the group should review the ideas gen-

erated, consolidate like concepts, check these against product requirements and

restrictions, and then vote on the top candidates to trim the list to the top ﬁve to

ten ideas. These ideas are then ﬂeshed out with materials, processes, structures, and

rough performance models. The concepts are benchmarked against similar products

and processes, and then systematically evaluated against customer requirements.

Another brainstorming loop may be needed if none of the concepts looks promising.

1.4 Microphone Case Studies

1.4.1 Microphone Background

Microphones, in the class of acoustic sensors, provided motivation for Avago

Technology’s acoustic sensor project and Knowles SiSonic MEMS microphone.

The Avago case study demonstrates a technology-push-type of program that

employs formal design methods to evaluate the applicability of existing technol-

ogy to a perceived market opportunity, whereas the Knowles SiSonic case study

demonstrates a market-pull design process whereby a microphone company without

MEMS technology developed the tools and processes to meet market needs.

Microphone (mic) technology, which had been fairly stable for the better part

of 50 years, has seen large changes in the past few years with the introduc-

tion of MEMS microphones. Electret-condenser microphones (ECMs) comprise

the majority of microphones (mics) used in consumer electronic devices, such as

cellular handsets. They offer reasonable sound quality at low cost and typically

contain a polymer membrane within a metal case. ECMs could not withstand sol-

der reﬂow processes due to their low melting point materials. As a result, costly

1 The MEMS Design Process 11

hand insertion was required to mount ECMs on a board. This limitation provided

an opportunity for alternative mic technology to penetrate the cellular handset and

other large audio markets. Capacitive silicon mic technology is one such alternative,

currently growing over 100% per year and accounting for nearly all the growth in

new mic technology today.

Capacitive mics have limitations due to their dual plate design, such as failures

related to moisture exposure and the need for a charge pump to bias the mem-

brane. Piezoelectric mics comprise a single membrane, therefore alleviating these

challenges, although their sensitivity is lower for mics on the MEMS scale.

1.4.2 The Avago Story

Avago Technologies has produced more than one billion ﬁlm bulk acoustic resonator

piezoelectric bandpass ﬁlters, primarily used in cell phones. Inasmuch as Avago’s

FBAR technology dominates the cell phone bandpass ﬁlter market, the company

decided to look for ways to extend the use of their technology. Investigation began

to apply FBAR and other core technologies to new uses and markets such as acous-

tic sensor applications. The high-volume FBAR production process was leveraged

to enable fast prototyping of piezoelectric MEMS microphones. Using FBAR to

create an acoustic sensor involves many design analysis aspects, including basic

plate bending theory, more complex ﬁnite element analysis, device layout, and

fabrication. Characterization, acoustic modeling, packaging, and testing methods all

have to be developed and implemented. Design of each element must be cognizant

of the requirements of the other elements in order to optimize the product for per-

formance, cost, and ﬁt to market [50]. QFD Phase I and CS help focus development

efforts on the most important aspects of the design from a customer perspective, bal-

ancing efforts among the design elements and providing a means to assess tradeoffs.

Design for manufacturability methodologies can help project teams make their way

through the maze of product development issues and emerge with a technically and

ﬁnancially successful product. The application of QFD Phase I and CS to Avago’s

acoustic sensor program serve to demonstrate the use of these tools. Although devel-

opment of Avago’s acoustic sensor is on-going, activities described in this case study

exemplify a means to develop products rapidly and gain understanding of potential

market impact, which provide useful understanding for other product development

efforts.

1.4.2.1 Design Process and Methods

A standard FBAR fabrication process was used as the basis of the initial microphone

process to enable fast prototyping. A plate mechanics model and ﬁnite element

analysis guided initial design and process choices [50]. While the initial experiments

were being run, QFD Phase I and CS were applied to the acoustic sensor, with the

tool application repeated for several potential market applications. The results were

used in three ways. First, they were used to determine on which aspects of the design

to focus. Second, the results helped determine which design concepts were likely to

12 T.L. Lamers and B.L. Pruitt

be viable to carry forward through prototyping. Finally, CS was used to determine

which market or markets were a good ﬁt for the technology.

QFD Phase I

Figure 1.3 shows where QFD ﬁts in the overall product development effort. For the

Avago mic project, QFD Phase I was used in the Stanford University Manufacturing

Modeling Lab standard format, relating customer requirements to engineering

metrics [39]. In this format, a matrix is generated in which rows correspond to qual-

itative customer desires such as size or ease-of-use. The customer requirements are

listed in the leftmost column of the matrix, and deﬁning the customer requirements

is the ﬁrst step in QFD I. These customer needs are then rated 1, 3, or 9, where

“1” is considered somewhat important, “3” is important, and “9” represents very

important. Customer requirements and their weightings can be determined via cus-

tomer interviews, focus groups, competitive products, or sales and marketing inputs,

among other means. Brainstorming, described in the following section, can also be

used to generate aspects of QFD and CS.

Columns correspond to speciﬁc, quantitatively measured engineering metrics, for

example, linear dimension less than 1 mm. The engineering metrics are synonymous

with product or process speciﬁcations, and are listed in the top row of the matrix.

Where possible, quantitative targets that correspond to these metrics are located in

a row labeled “Technical Targets” near the bottom of the matrix.

Next, the matrix is ﬁlled out with the r elative r elationship between each customer

requirement and engineering metric in the intersecting matrix element. The weight-

ing scheme of 0, 1, 3, or 9 is used, where “0” denotes no relationship, “1” shows a

slight relationship, “3” means a signiﬁcant relationship, and “9” represents a very

important relationship. The products of the customer requirement importance and

the relationship weighting between customer requirement and engineering metric

are summed over each column to generate raw scores for each engineering metric.

Each column score is then translated to a percentage of the total score, with the

highest percentage items denoting the most critical engineering metrics. The rel-

ative importance percentages for the engineering metrics provide guidance on the

appropriate amount of product development effort to be spent achieving each engi-

neering metric. For background and more detailed information on performing QFD,

see [38, 39].

The results of QFD Phase I as applied to the use of the piezoelectric microphones

in cellular handsets, laptops, and automotive sensor applications are shown in

Figs. 1.5, 1.6, and 1.7, respectively. The customer requirements and their weightings

for each application were gathered through domain research and conversations with

industry insiders. For example, each application has a customer requirement related

to size. Cellular handsets have “ﬁts in cell phone,” laptops have a requirement of

“ﬁts in computer bezel or speaker phone,” and automotive has “ﬁts in engine com-

partment or body cavities.” Although size is important in all three applications, as

shown by t he size requirement having a weight of 9 in all three cases, the speciﬁcs

of the size required depend on the application. Auto engine compartments are much

1 The MEMS Design Process 13

Fig. 1.5 QFD Phase I application of piezoelectric acoustic sensor to cell phones (Modiﬁed ﬁgure,

used with permission of ASME. First appearing in Lamers and Fazzio [50])

larger than cell phones, providing more leeway for ﬁtting the automotive customer

space requirement.

The engineering metrics and their targets came from existing product data sheets

and the knowledge of the project technical team (Knowles; Akustica; Hosiden;

Panasonic). The relationship weights between customer requirements and engineer-

ing metrics also came from members of the technical team. Customer feedback

and comparisons to competitor products were used to validate the inputs and the

results of QFD Phase I , in an attempt to improve the robustness of the results. Many

engineering metrics appear in all three applications, although their target values

vary according to the requirements of each application. In Figs. 1.5, 1.6, and 1.7

the engineering metrics that garnered the highest relative weights are highlighted.

It is interesting to note that “Dynamic Range” and “Total packaged linear dimen-

sions” were in the top weighted engineering metrics in all three applications. At the

same time, the other top engineering metrics changed depending on the application.

For example, “sensitivity” had the highest weight in both the cellular handset and

laptop applications, but was just outside the top group for the automotive sensor

application.