M E T H O D S I N M O L E C U L A R B I O L O G YTM
John M. Walker
School of Life Sciences
University of Hertfordshire
Hatfield, Hertfordshire, AL10 9AB, UK
For other titles published in this series, go to
Microengineering in Biotechnology
Michael P. Hughes and Kai F. Hoettges
University of Surrey, Guildford, UK
Michael P. Hughes
University of Surrey
Centre for Biomedical
Duke of Kent Building
United Kingdom GU2 7TE
Kai F. Hoettges
University of Surrey
Centre for Biomedical
Duke of Kent Building
United Kingdom GU2 7TE
Library of Congress Control Number: 2009933982
# Humana Press, a part of Springer ScienceþBusiness Media, LLC 2010
All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or
hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as
such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights.
While the advice and information in this book are believed to be true and accurate at the date of going to press, neither
the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be
made. The publisher makes no warranty, express or implied, with respect to the material contained herein.
Printed on acid-free paper
Just as the twentieth century witnessed developments both in electronic engineering
and molecular biology which have revolutionized the way we live, so the twenty-first
Since so many biochemical procedures occur at the molecular level, microelectronic
engineering offers the opportunity to reduce the way in which such procedures are
performed to the same level. The advantages of miniaturized analysis are manifold;
reducing the sample volume increases reaction speed and detector sensitivity whilst
reducing sample and reagent requirements and device cost. Some of the world’s largest
having potential to provide, for example, bench-top versions of large and expensive
equipment that could make analyses like flow cytometry as commonly available as gel
electrophoresis is now.
The aim of this book is to provide biochemists, molecular biologists, pharmacolo-
gists and others with a working understanding of the methods underlying microengi-
neering and the means by which such methods can be used for a range of analytical
techniques. It describes the methods by which microengineered devices can be built to
perform a number of applications and considers how the field may progress by examin-
ing some more complex lab on a chip devices which have great potential in the
advancement of the way in which molecular biology is performed. We also hope that
this book will be of use to microengineers, both as a reference guide for practical
microengineering techniques and as a route into the development of new devices for
biological applications. The union of molecular biology and microelectronics offers
huge promise but one which will be all the stronger wherein both sides understand the
needs of the other.
27 April 2009
Michael P. Hughes
Kai F. Hoettges
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1. Microfabrication Techniques for Biologists: A Primer on
Building Micromachines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2. The Application of Microfluidics in Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
David Holmes and Shady Gawad
3. Rapid Prototyping of Microstructures by Soft Lithography
for Biotechnology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Daniel B. Wolfe, Dong Qin, and George M. Whitesides
4. Chemical Synthesis in Microreactors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Paul Watts and Stephen J. Haswell
5. The Electroosmotic Flow (EOF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Gary W. Slater, Fre ´de ´ric Tessier, and Katerina Kopecka
6. Microengineered Neural Probes for In Vivo Recording . . . . . . . . . . . . . . . . . . . . .
Karla D. Bustamante Valles
7. Impedance Spectroscopy and Optical Analysis of Single Biological Cells
and Organisms in Microsystems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Shady Gawad, David Holmes, Giuseppe Benazzi, Philippe Renaud,
and Hywel Morgan
8. Dielectrophoresis as a Cell Characterisation Tool. . . . . . . . . . . . . . . . . . . . . . . . . .
Kai F. Hoettges
9. AC-Electrokinetic Applications in a Biological Setting . . . . . . . . . . . . . . . . . . . . . .
Fatima H. Labeed
10. Wireless Endoscopy: Technology and Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
David R. S. Cumming, Paul A. Hammond, and Lei Wang
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
GIUSEPPE BENAZZI • School of Electronics and Computer Science, University of
Southampton, Highfield, Southampton, UK
KARLA D. BUSTAMANTE VALLES • Orthopaedic & Rehabilitation Engineering Center,
The Medical College of Wisconsin & Marquette University, Milwaukee, WI, USA
DOUGLAS CHINN • Sandia National Laboratories, Albuquerque, NM, USA
DAVID R. S. CUMMING • Department of Electronic and Electrical Engineering,
University of Glasgow, Glasgow, UK
SHADY GAWAD • Swiss Federal Institute of Technology, Lausanne, Switzerland
PAUL A. HAMMOND • Department of Electronic and Electrical Engineering, University
of Glasgow, Glasgow, UK
STEPHEN J. HASWELL • Department of Chemistry, University of Hull, Hull, UK
KAI F. HOETTGES • Centre for Biomedical Engineering, University of Surrey, Guildford,
DAVID HOLMES • School of Electronics and Computer Science, University of Southampton,
Highfield, Southampton, UK
KATERINA KOPECKA • Department of Physics, University of Ottawa, Ottawa, ON,
FATIMA H. LABEED • University of Surrey, Guildford, Surrey, UK
HYWEL MORGAN • School of Electronics and Computer Science, University of Southamp-
ton, Highfield, Southampton, UK
DONG QIN • University of Washington, Seattle, WA, USA
PHILIPPE RENAUD • Swiss Federal Institute of Technology, Lausanne, Switzerland
GARY W. SLATER • Department of Physics, University of Ottawa, Ottawa, ON, Canada
FRE´DE´RIC TESSIER • DepartmentofPhysics,UniversityofOttawa,Ottawa,ON,Canada
LEI WANG • Department of Electronic and Electrical Engineering, University of Glas-
gow, Glasgow, UK
PAUL WATTS • Department of Chemistry, University of Hull, Hull, UK
GEORGE M. WHITESIDES • Department of Chemistry and Chemical Biology, Harvard
University, Cambridge, MA, USA
DANIEL B. WOLFE • Department of Chemistry and Chemical Biology, Harvard
University, Cambridge, MA, USA
Microfabrication Techniques for Biologists: A Primer on
In this chapter we review the fundamental techniques and processes underlying the fabrication of devices
on the micron scale (referred to as ‘‘microfabrication’’). Principles laid down in the 1950s now form the
basis of the semiconductor manufacturing industry; these principles are easily adaptable to the production
of devices for biotechnological processing and analysis.
Key words: Fabrication, photolithography, photomask, etching, thin films.
Fabrication of electronic devices on a micron scale was developed
by the integrated circuit industry, beginning with the invention of
the integrated circuit (IC) simultaneously by Jack Kilby of Texas
Instruments and Robert Noyce of Fairchild in 1958. Microma-
chining began with the challenge issued by Professor Feynman in
1959 to build a tiny motor. In 1965, Gordon Moore of Intel
postulated what is now known as Moore’s Law, where the data
density of integrated circuits doubles every 18 months. Today,
2006, the largest companies are fabricating complex state-of-the-
art chips with over 40 photomask layers, including 10 metal layers,
and are approaching 109transistors in an area the size of a postage
stamp. Lateral dimensions are below 100 nm and shrinking all the
time. Vertical dimensions for the thinnest oxides are a few mono-
layers, with films of metal and dielectrics on the order of a few
nanometers (nm)to amicrometer (micron,mm).A modernsilicon
integrated circuit fabrication facility costs over $2 billion to build
M.P. Hughes, K.F. Hoettges (eds.), Microengineering in Biotechnology, Methods in Molecular Biology 583,
DOI 10.1007/978-1-60327-106-6_1, ª Humana Press, a part of Springer Science+Business Media, LLC 2010
and equip. The new MESA micromachine laboratory at Sandia
National Laboratories costs almost $500 million. Only large cor-
porations and governments can make this kind of capital invest-
ment. Chips must be made in huge quantities with exceptional
quality control to achieve an adequate return on investment.
Because chip-making processes advance so quickly, process equip-
ment is typically depreciated after 2 years.
An integrated circuit is fabricated using only four steps: film
growth and deposition, patterning, etching, and annealing. A
typical process begins by depositing a film, spinning photoresist,
patterning and developing the resist, etching the thin film, and
finally stripping the resist. By repeating these steps over and over,
the most complex devices can be made. This is a gross oversimpli-
fication but it demonstrates how a number of simple steps can be
combined to make very complex devices. In industry, complex
processing tools such as etchers and thin film deposition systems
are dedicated to a single process. Engineers in semiconductor and
micromachine factories devote entire careers to reducing variabil-
ity by characterizing processes and equipment and thus improving
device yield. Even so, scrap rates run up to 50% of all material
started in the line, depending on the maturity of the process.
Mature lines with well-characterized processes will skip testing in
wafer form and package every device made, rejecting the defective
parts at that stage, indicating that well over 99% of all wafers were
Most modern wafer processing machine tools are ‘‘cassette to
designedwaferholdersknownas cassettesor boats.Humansnever
handle wafers. Instead, cassettes are placed into ‘‘indexers’’ that
returning them to a cassette when the process is finished. Since
humans make mistakes and add defects, they must be removed as
far as possible from the wafers.
Here we must distinguish between the terms ‘‘fab’’ and ‘‘lab.’’
Fab refers to a large, commercial factory set up to turn a profit,
which produces a small variety of integrated circuit devices that
have similar processes. Lab refers to a smaller facility, usually at a
university or government laboratory, set up to produce a wide
variety of electronic, optical, and mechanical devices with many
is designed for high throughput (low cost) and must be designed
Uptime is a major consideration, and fabs will schedule routine
maintenance into the process. Although programmers who use
chips to create amazing programs get all the glory, the people who
design and build the machines that make the chips are the real
heroes because the semiconductor industry arguably has the high-
est technology machine tools in existence. Creating a 25-mm
square chip with dozens of layers, thousands of process steps,
geometries below 0.25 mm, and one billion transistors with no
defects is not an easy task!
The emerging micromachine industry has benefited from all
the advances made by the semiconductor industry, but is several
Moore’s law generations behind it. In many micromachining
applications, lateral dimensions are above 1 mm, and vertical
dimensions are above 100 nm. Universities and other small labs
do not need the expensive dedicated cassette to cassette machines
designed for high throughput and quality control. This chapter is
written for the small lab, with contrasts to large production facil-
ities. We hope to give rules of thumb that help get devices fabri-
cated without the reader having to ‘‘re-invent the wheel’’ while
learning to build microdevices.
Why is this chapter called ‘‘Microfabrication for Biologists’’
when it has absolutely no reference to biology in it? Because most
commercially successful micromachines have biological applica-
tions and most of those are microfluidic devices that sense some
driving micromachining away from silicon-based devices that
often have moving parts to devices that analyze fluids built on
glass and polymer substrates.
Many good books on the subject of microfabrication have
been written (see Suggested Reading List). These books list
some exotic and complex processes and the reader is referred to
these books and others for rigid analysis and mathematical detail.
first device needs practical information to get started on a project.
Here we attempt to supply the reader with a lot of practical knowl-
edge, gained from the author’s experience in building devices in
industry, university, and government laboratories. The approach
used in this chapter is to teach a novice engineer, scientist
or biologist how to design a device pattern using a CAD
(computer-aided design) program and how to develop a process
run sheet to get it fabricated with the help and assistance of
professionals who work in microdevice process labs every day.
We have avoided the use of trade names where possible. This
chapter is intended to be a practical guide to getting started in
the field, making no attempt to define the underlying physics and
chemistry. Since most biologically oriented micromachining
involves substrates that are not silicon or period II–VI or III–V
semiconductor compounds, this chapter is written with that in
Any device built by micromachine techniques has hundreds of
potential variables to control. Small changes in any one of these
can have a large impact on the final product. The keys to
successful microdevice fabrication are good process control in
Microfabrication Techniques for Biologists3
well-characterized machines, a well-defined fabrication process,
well-designed photomasks, and careful fabrication. To maximize
the possibility of success, it is highly recommended that the novice
microfabricator devote a great deal of time to designing the device
and process before beginning production. Properly designed
photomasks and a complete process run sheet, or traveler, can be
assembled by talking with people familiar with the process tools
their many variables in the proper order, and a simple example is
given below. In the world of microfabrication, the tiniest of details
of all the processing is necessary to repeat a successful process.
engineers building their first microdevice typically expect the first
attempt at fabrication to work. This is rarely the case, so one
should expect several scrap wafers before having success. The
reason is that in a research setting, well-characterized, reproduci-
ble, standardized processes are not often available, and as Mur-
phy’s law states, anything that can go wrong will. With hundreds
of variables to control, and only limited knowledge of any given
process step, fabricating a device correctly during the first attempt
is not highly probable. An engineer or biologist must fabricate the
device under the limitations of the laboratory equipment available
and the time available to characterize processes, but the resources
available to the researcher limit processing capabilities.
The first consideration for microfabrication is the substrate, or
wafer, the device will be built upon. Integrated circuits are built
on wafers, where micromachines can be built on a variety of sub-
strate materials, so the term substrate is used interchangeably with
wafer. Micromachining is often divided into two categories, bulk
and surface. In bulk micromachining, the substrate is formed into
the finished device, whereas in surface micromachining layers are
added on top (or on the bottom) of the substrate, and the added
layers are formed into the desired device. Many technologies
combine surface and bulk micromachining. It is highly recom-
mended that one obtain many more substrates than are required
for the project to use as dummy wafers, test wafers and to take into
account scrapped wafers caused by misprocessing. Plan on drop-
ping and breaking some substrates too. It is also recommended
that substrates be scribed at the beginning of the process to make
traceability easier. The variables affecting the specification of
substrates can be seen in Fig.1.1.
The selection of a substrate determines much of the
subsequent processing. Most microfabrication tools are designed
for silicon wafers with a flat. Theflat issimply an area ground off of
the edge of the round wafer. It is there so that the crystal orienta-
tion can be determined and is used by automated processing
equipment to orient the wafer. Square or other shaped substrates
can be used, but non-circular substrates are much harder to work
sizes that fit into most machines, so it is advantageous to use a
standard size. Two-inch (50 mm), three-inch (75 mm), four-inch
(100 mm), and six-inch (150 mm) substrates are common stan-
dards. However, the metric and English sizes are not necessarily
compatible in some machine wafer holders (1 inch=25.4 mm).
When ordering substrates,several variablesmustbe accounted
for. After diameter, and thickness, one must determine the flat-
ness, surface roughness, edge grind, the finish on the back of the
substrate, temperature limitations, coefficient of thermal expan-
sion, and transparency to a given wavelength of light and
mechanical toughness. Poorly ground edges can be a source of
particles when they chip, and can initiate cracks which will cause
the substrate to break. In a plasma etch process, note that the
conductivity of the substrate can have an impact on the way the
Silicon has the advantage that it is widely available, relatively
inexpensive and very well characterized. It is also the most perfect
material obtainable, being a slice of a single crystal. The crystal-
linity can be taken advantage of in directional etches and in
that can determine how successful the device fabrication process will be.
Microfabrication Techniques for Biologists5
applications where it is desired to etch through the wafer. The
most common types are (100), (111) and (110). These are the
Miller indices, indicating which crystal plane makes up the surface
of the wafer. Silicon can be ordered with either p (boron) or n
doping can be done with an ion implanter or high-temperature
furnace. If thin membranes are required in the device, silicon may
be the optimum material because some wet etchants, such as
KOH, can be highly selective between different crystal planes
and levels of doping. Custom silicon can be ordered from many
manufacturers, but off-the-shelf material is usually much cheaper.
Glass is being used more often in micromachining applica-
tions. It is amorphous, rather than crystalline like silicon. Dozens
of types of glass are available, and it can be ordered in custom
shapes and sizes. It is inert, transparent, and mechanically tough.
Some types of glass can be very inexpensive, such as microscope
slides, and some, such as fused silica and quartz, can be very
expensive. Many glasses are chemically impure, and because it is
amorphous, it can have a non-uniform microstructure. This can
affect etching, optical, and electrical properties.
Polymer substrates are now being used. It is very difficult to
generalize about polymers, since so many are available. Plastics are
inert to acids and can be formed by molding and embossing, but
plastics tend to have a non-uniform microstructure. Polymer
substrates can absorb water and solvents, changing their
Metal substrates have been used, but due to the reactivity of
metals they generally hold up very poorly during chemical treat-
ments. It is also difficult to get surfaces with low roughness in
metals. Ceramics are also used, but the polycrystalline nature of
ceramics can present severe processing difficulties
Wafer supply companies often have film deposition equip-
ment. If all of your substrates need a particular film, it may be
cheaper, faster, and cleaner to buy wafers with a film already
deposited on them by the supplier.
used to expose some kind of photoresist. The most common kind
is made of a thin film of chrome on a square of glass that is
transparent into the UV range. Many different kinds of glass are
on stray reflectance from the front surface. Most align tools use a
mask that is 1 inch/25 mm larger than the wafer size.
Contact the commercial mask fabrication shop at the begin-
ning of the design cycle so that your design is compatible with the
shop’s tooling. Many drawing programs are available, but line
types and shapes used by generic drawing programs cannot be
recognized by mask shop tools. Mask shops require the data to
be in specific formats, so an important factor is being able to
convert the finished CAD drawing into a format that the mask
capable of saving the data in an acceptable format.
Inexpensive photomasks can be designed using any drawing
program and printed out using standard office printers on Mylar
transparencies. Such masks are only effective for larger geometries
with loose tolerances, but can be made quickly and cheaply.
A reticle is a photomask that is scaled, typically used in various
types of steppers. The advantage of using a reticle over a photo-
mask is that small geometries are achieved optically by the end
user, rather than by trying to write small features directly on the
photomask. Each type of stepper requires its own reticle, where
photomasks can be used in most machines, if sized right. If your
device is small and several copies can fit on a single wafer, a shop
may generate a reticle and step it over and over to make a mask,
rather than write the same pattern over and over.
Sometimes a photomask has a pellicle attached to it, a thin
membrane that is set away from the chrome geometry by a spacer.
Many designers have been frustrated because an expensive mask
does not do what it was designed to do. As illustrated below, the
mask design is not independent of the process. Several decisions
must be made during the design stage.
1. Wafer or substrate and die size
2. Type and size of mask
3. Dark or light field
4. Spot size
5. Minimum dimension/pitch
6. Defect density
8. Process bias
9. Positive or negative resist
10. Alignment keys
12. Dicing the wafer into individual chips
13. Howthe device will bepackaged orhow the inputs andoutputs
will travel from the macroworld to the chip scale microworld.
Microfabrication Techniques for Biologists7
One of the firstdecisions to make isthe die, or chip, size. Theword
die comes from the IC industry, where a typical wafer will have
compound semiconductor) chip inside that could be anywhere
from a fraction of a mm2to ?30 mm2in size. These individual
dies are cut from a larger wafer. Many micromachine devices can
take up an entire wafer. If your device is smaller than the wafer it
may be beneficial to have many dies on the same wafer.
The type of lithography tool to be used and the substrate size
determine the size and type of the mask to be made. Contact
printers, steppers, and projection printers all have many specific
mask requirements, so decide on the alignment tool before doing
the CAD design. Masks can be ordered as masters or copies.
Copies are cheaper than masters, but can have lower resolution.
A master must be made to make a copy. If contact printing is to be
used, a mask can be damaged, so it may be economical to use
copies rather than the original master.
Dark field masks are mostly chrome with holes in the film;
light field masks appear to be mostly glass with chrome geometries
on them. The type of resist used and the desired pattern determine
whether dark or light field is called for. Masks are usually drawn
light field, and dark field is specified on the mask shop’s run sheet.
The way a mask exposes photoresist is a function of how much
geometry is present and how it is distributed. During lithography,
dense geometries in one area may expose differently than areas
are a function of geometry density and size.
A mask shop starts with a photoresist covered chrome film and
the writing is a major determinate of the price, with 0.25 mm being
write. Both line and space (pitch) minimum dimensions on the mask
can determine what type of equipment is used to make the mask:
larger dimensions are sometimes cheaper to make. The smallest geo-
the sum of the line space. Mask shops may define a minimum pitch.
size of defects allowed. If your device has small, dense geometries,
very few defects in the form of missing chrome or extra spots of
chrome can cause fatal defects in the device. If you have large, low
generalrule, a defect 10% ofthegeometrysize may causeproblems.
The tolerances in a given layer and the tolerances between layers
can also impact the price of a photomask. If all lines and spaces are
above 1 or 2 mm the mask can be fairly inexpensive. Mask shops
cannot alter their quality control for each mask. If a mask has tight
tolerances and low defect levels, a shop may have to make it several
times to get it right, increasing the cost. Masks often have special
geometries specifically for critical dimension (CD) measurements.
Understand that any metrology tool will have error in its measure-
ment and that measurements can vary across a wafer. Ideally measure
understand how much error is in a process.
A very critical idea that must be incorporated into any mask
design is process bias (Fig. 1.2). The geometry dimensions laid out
in the CAD work of the photomask are very unlikely to be
reproduced exactly on the finished product. Every time a pattern
is transferred it will change dimensions, growing or shrinking, and
corners will be rounded (Fig. 1.3)
With positive tone resists, the part that is exposed becomes
tone resists, the exposed part becomes cross linked, and the unex-
posed parts are removed during develop (Fig. 1.4). Typically with
negative resist, openings in the chrome translate to larger geome-
tries after exposure. Positive resists re-create the photomask, but
chrome patterns will shrink after exposure and develop.
Most processing today is done with positive photoresists.
Positive resists give finer geometries than negative resists. Because
positive resist developer chemistry uses aqueous bases it presents
less waste disposal issues than negative resists which use solvent-
based developers and rinses.
Fig. 1.2. Process bias during exposure and etching. Upon exposure, the photoresist will
change from the mask dimension by some factor 2a. Depending upon processing
conditions, a can be positive or negative relative to the mask dimension. After etching,
another bias is introduced, 2b. Etching conditions and the resist profile determine b. The
pitch is constant throughout all processing. The designer must account for these biases
when laying out the photomask.
Microfabrication Techniques for Biologists9
Well-designed alignment keys are critical mask features for
devices that have more than one layer, since the second layer
must be aligned to the first layer.
Many different alignment keys have been used, and each
designer has a favorite. Understand what works well in the litho-
graphy tool before doing the design. In the figure, both light and
Fig. 1.3. Photoresist profiles. The ideal profile is shown. Its transfer characteristic will be
very good and its dimension is easy to measure, since its edges are well defined and
distinct from the substrate. It is not seen in real processing. The second drawing
demonstrates a more realistic profile, where the edges are not distinct at all. Measure-
ment tools use various algorithms to determine where the edge actually is. Note the
rounding of the corners. The third drawing illustrates what is possible in resist proces-
sing. Such a profile is used in liftoff processes.
Fig. 1.4. Positive and negative photoresists.
dark field keys are shown as crosses. This author likes to use
crosses about 100 mm long and 20 mm wide for the first layer
for contact printing. The layer to be aligned to it has a cross
98 mm long and 18 mm wide, giving a tolerance of about –
1 mm. Thus, a smaller cross fits on top of a larger cross that is
etched into the wafer. Recall that the mask design dimensions
and the geometry dimensions on the first layer will vary based
on the process biases introduced, so those biases need to be
understood before laying out the align keys. Most often all
layers are aligned to the first layer, so the first layer needs keys
for each layer to be aligned to it. Align keys must be posi-
tioned on the mask and die so that the microscopes on the
align tool can easily locate them.
A very common mistake in mask layout occurs in dark field
masks. Trying to find align keys using a microscope on top of a
dark field mask can be very difficult. Since most of the mask is
chrome, the underlying layer can only be seen through the align
key itself. Draw a large box of clear geometry and put a small
chrome align key inside the clear box. The large box makes it
easy to find the align key and makes it much easier to find the
align key on the underlying layer (Fig. 1.5).
Fig. 1.5. Alignment keys for dark- and light-field masks. On the left, align keys in a dark
field mask must be surrounded with a large clear area to make it possible to see the
geometry underneath that a layer is being aligned to.
Microfabrication Techniques for Biologists11
Another common mistake is to draw closed shapes by putting
one square or circle on top of another (Fig. 1.6). Human eyes can
circles on topof eachother and writesboth. Whatwassupposed to
be a donut or outline becomes a solid shape. Therefore, closed
shapes must be drawn as two parts, as seen in Fig. 1.6.
Besides the tolerances of specific geometries, another impor-
tant tolerance indicates how each layer lines up to the previous
layer. Masks and wafers can have run-in and run-out errors. These
are where a design may call for a dimension across an entire wafer,
say 100.000 mm, but the actual dimension is, for example,
100.005 mm. The layer placed on top of that layer may be
99.009 mm, giving a 0.006 mm or 6 mm run-out error. Such
errors can be introduced at the mask shop by bowing of substrates
and masks in the align tool and by thermal mismatch (Fig. 1.7).
Fig. 1.6. Drawing closed shapes. A computer recognizes only solid shapes and cannot
subtract one pattern from another.
Fig. 1.7. Alignment and runout errors seen upon alignment of the top layer (light gray
crosses) to the layer etched into the wafer (dark crosses).
Taking handedness or chirality into account is important when
designing a photomask. Masks are usually used chrome side down
or viewed through the back of the mask relative to the design
artwork. One can buy boxes of unexposed photomasks covered
Do not open them except in yellow lights. By placing an existing
mask in contact with a blank mask, a copy can be made. Typically
some pattern fidelity is lost, and the copied mask is a left-hand
version of the original. Copying a mask a second time restores the
chirality to its original. To reverse a mask, i.e., to use it with
negative photoresist, for example, one can make a copy, strip the
resist, sputter or evaporate a different metal onto it (such as tita-
nium), etch the chrome in a chrome specific etchant, and have a
titanium reversal copy. Edges may be somewhat rougher than the
Sometimes test patterns that can define resolution, or measure
overlay alignment (such as a vernier pattern), are added to the
photomask. Many chip designers have various test patterns avail-
able to cut and paste into designs. Electrical devices often have
special test devices such as resistors or individual transistors built
into a device in the test pattern section of the chip.
If multiple devices, or dies, are on a single wafer, it is usually
diced up as a final processing step. The blades used in dicing saws
least 500 mm makes the dicing operation much easier.
Packaging is often the most expensive part of an IC and is
often overlooked during the device design stage. Final packaging
can be a very difficult part of building a device. In ICs, only
electrons go in, and only electrons and heat leave the chip. In
micromachines, electrons, heat, fluids, mechanical energy,
photons, and perhaps even magnetic energy all may go in and
out; consequently, the package can be very complex and probably
cannot be purchased off the shelf. It is difficult to get information
from a micro level to the human who will make decisions based on
what happens at the micro level. Thus the designer of microma-
chines must take the package into account early in the design
Consider when designing a mask how the finished device will
world is usually done with bond pads. Bond pads are squares of
metal that are the terminus of electrical wires patterned on the
chip. Special micromanipulators are available to put tiny probes
onto bond pads, but if space allows, bond pads 2 mm on a side can
be contacted with the crudest probe tips. We recommend making
them as large as space permits.
Another type of mask is a shadow mask. This is usually a thin
sheet of metal with holes cut in it. The shadow mask is placed in
contact with the substrate and put into an evaporator or sputterer
Microfabrication Techniques for Biologists13
(below). Metalisdepositedthroughthemask,eliminating the need
for lithography. It is cheap and quick, but edge definition is poor
and it isdifficult to align shadowmasks to existingpatterns. Resolu-
tion is limited to above 50 mm. Shadow masks can be made in a
Shadow masks can be ordered commercially from companies that
do lithography on stainless steel sheets and etch patterns into them.
To build a micromachine or microelectronic device, many vari-
ables must be controlled. Below we outline the issues.
The adhesion of spin-cast films to the substrate is an important
variable, and must be dealt with before spinning a film. Modern
resists adhere well to many materials, and dry (plasma etch) pro-
cesses usually do not affect the adhesion. However, when subse-
quent processing involves mechanical stress or liquid chemicals,
the interface between the spun film and the substrate may be
attacked. The most common and well-proven adhesion promoter
(CH3)3SiNHSi(CH3)3). This material is designed for silicon-
based materials and may not work on metal films and polymers
because it is designed to work with hydrated surfaces. Its reaction,
R-OH + (CH3)6Si2NH= 2 R-O-Si(CH3)3+ NH3. R is a substrate
molecule such as silicon. Ammonia is liberated from the reaction.
Primedwafers do notneed to bebaked, but a124?C bake can help
eliminate the ammonia. HMDS makes a previously hydrophilic
organic surface. It is often used as a diluted solution (?20% in an
organic solvent) spun on at 3,000 rpm prior to spinning the
photoresist. More modern labs will use a vapor prime system,
where the adhesion promoter is released in vapor form into an
evacuated chamber, usually well above room temperature. Vapor
prime systemsuseless material and coatvery uniformly.Thekeyto
using adhesion promoters is that they must be a monolayer,
forming a bridge between two incompatible surfaces, such as an
oxide and an organic. Adhesion promoters designed to work on
many classes of material are available, so research an adhesion
promoter that forms a good interface between the two materials
you are using if HMDS is not effective in your application.
By dispensing a small amount of a liquid-phase material onto a
for spinning may have several solvents in them to control
evaporation rates and film uniformity. Often in a research lab new
and novel materials need to be spun, so the engineer must first
characterizethe material’sspinning characteristics beforedevoting
the material to a device. A very simple spinner can be made using a
small motor and a chuck appropriate to hold the substrate. How-
ever, most labs have commercially available spinners with vacuum
chucks. When working with odd-sized substrates, a sheet of red
silicon rubber with a hole in the middle can be placed on most
spinner chucks, allowing the chuck to successfully hold the sub-
strate. Always close the lid of a spinner to control air flow around
from objects thrown off the chuck.
The variables of interest when spinning are as follows:
1. solution viscosity and solids content
2. solvent system vapor pressure
3. acceleration rate
4. final spin speed
Other factors can determine the thickness and quality of the
film, such as the atmosphere inside the spinner and the air flow
around the substrate while spinning, but these variables are sec-
ondary. With commercially available materials, an amount of solu-
tion is poured into the center of the substrate roughly one-fourth
the diameter of the substrate. With lab-made solutions, the quan-
content is important because films will tend to crack if the volume
changes too much when the solvent evaporates. Spin speeds below
500 rpm rarely work well. A 3,000 rpm spin with at least
1,000 rpm/s up to 5,000 rpm/s acceleration rate is a common
place to start when characterizing a material on a spinner. Thirty
seconds to one minute is usually an adequate spin time. High
vapor pressure/low boiling solvents can evaporate so quickly that
a film forms on top of the liquid creating a poor spin. Low vapor
pressure/high boiling solvents can take so long to evaporate that
film dries like a ‘‘coffee stain’’ leaving a poor-quality non-uniform
film, so selection of solvents is an important factor when designing
materials for spin casting. Particles on the surface of the wafer or in
the resist can create ‘‘comet tails,’’ a common spin-cast film defect.
These are summarized in Fig. 1.8.
Photoresists have a short shelf life, but often work well a
few years beyond that date if stored in a refrigerator. Produc-
tion fabs never use any chemical past its expiration date. To
prevent contamination, always pour resist, or any laboratory
chemical, from the factory bottle into a dry, cleaned brown
bottle for use. Never dip any pipette or other items into the
clean bottle of resist. Keep bottles closed and never expose
resists to room light. It is a good idea to wrap photoresist
bottles in aluminum foil to keep light out.
Microfabrication Techniques for Biologists15
After casting, photoresists need a soft bake to dry out the resi-
develop poorly. Baking can be done on a hot plate or in an oven,
near 100?C, depending on the resist type. Bake times on hot plates
are typically 1–2 min, oven times are usually 10–30 min. More
uniform temperatures are available in ovens. Some more sophisti-
catedlabshave thebake plateinlinewiththe spinner, andindustrial
fabs have the spinner, bake plates, and develop systems in line with
the stepper for a fully automated process. In such highly automated
process equipment it can be difficult to change parameters for
experiments or to use non-standard substrates.
Clean the spinner and replace the liner before spinning your
wafer. Vacuum out lint left over from the wipers used in the
cleaning procedure. Always spin a dummy wafer with resist before
committing the actual device to gather particles left over from
cleaning and verify that the spinner is working properly. These
dummies can be used to characterize the lithography process, so
store them under yellow lights.
Spin on glasses (SOG) are solutions containing chemicals that
can be thermally degraded into oxides and can be spun onto
wafers. Any spun on material has an advantage over vapor depos-
ited materials because they planarize, or level, the surface, analo-
gous to snow falling on a field or water filling a shallow pool.
Some materials designed to become part of themicrostructure
can be made photosensitive, such as SU-8 (a trademark) negative
resist and polyimides. Both families of materials are very tough in
their cured form and can be very simple to process. These kinds of
polymeric materials often require thermal treatments to cure
them, so heat cycles must be accounted for in the overall process.
4.2. Lithography Tools
Most lithographic patterning is done using a mercury or mercury
argon light source, which has peaks at 365 nm (i-line), 408 nm,
and 436 nm (g-line). Some machines filter out specific
Fig. 1.8. Photoresist spinning defects. The edge bead is normal and usually causes no
problems. It is sometimes stripped off by carefully spraying a stripper on the edge of the
wafer while it is spinning.
photoresist depends on the absolute wavelength, the relative
intensities of the peaks, resist sensitivity, resist thickness, desired
profile, and the exposure tool. The reflectivity of the substrate can
sources age, thereforetotal intensity and relative peakintensity can
change over time in any given exposure tool. Most labs have a
meter for measuring the light output to aid in setting the exposure
Several types of tools exist to transfer a pattern from a photo-
mask to a substrate. Contact printers place the chrome side of a
photomask in contact with the photoresist on the substrate. These
machines are fast, reproduce the mask well, and have high relia-
bility. The disadvantage is that they are typically limited to 1 mm
and above in geometry size and they damage the photoresist and
mask through physical contact. Modern versions of these tools
often have backside alignment capability, where photographic or
infrared methods can be used to align patterns on both sides of a
substrates that have smooth surfaces.
Proximity printing is similar to contact printing, but the mask
never comes into contact with the photoresist. Resolution is lost,
but neither the mask nor the resist is damaged.
Projection printers separate the mask and the substrate
through complex optics and give resolution similar to that of
contact printers. These were used extensively in the IC industry
for many years due to speed and the quality of the image they
project, but have fallen out of favor in recent years.
Steppers are the workhorse of industry and are found in
larger universities and corporate labs. They are expensive to
buy and maintain, require specialized photomasks, and are
usually operated by specialists. Steppers can expose large
300-mm wafers with no defects added. A 10? reduction
stepper can create submicron geometries by reducing a 5-mm
line on a reticle to a 0.5-mm line on a wafer. Because of the
size reduction, the image must be placed on a wafer over and
over until the entire surface is exposed, known as step and
repeat. These tools do not work well with large die sizes as are
common in micromachines. If large high-resolution patterns
are required, different pieces must be stitched together.
Steppers require special align keys to be put onto the wafer
known as globals.
The smallest geometries are still made using an electron beam
writer, often a modified electron microscope. Geometries as small
as the spot size of the e-beam can be written, but only very tiny
areas can be exposed. E-beam writing is a very time-consuming
Microfabrication Techniques for Biologists 17
Over the last few years, many researchers have had excel-
lent results using polydimethoxysilane, or PDMS, to make
‘‘rubber stamps.’’ A conventional photomask is designed, and
the image is etched into an appropriate substrate. A PDMS
film is cast onto the substrate. After peeling the polymer off of
the substrate, it can be coated with a variety of materials,
which are then transferred to other substrates. The reader is
directed to the original literature, much of it from Professor
George Whitesides’ laboratory at Harvard for details on this
Another technique is to use synchrotron X-rays to image
polymethyl methacrylate substrates (PMMA) creating very high
aspect ratios up to a millimeter thick. This technique was
(lithographie, galvanoformung, and abformung). Very high
resolution can be obtained in high-aspect structures, but since
time on a synchrotron is difficult to get, the technique is limited
to a few laboratories worldwide.
After being exposed through a photomask with UV light, the
exposed regions (positive resist) or the unexposed regions
(negative resist) of photoresist must be removed. Developing can
be done in a beaker or in a sophisticated spray develop spinner.
Most small labs use a beaker for immersion develop. Sufficient
agitation is necessary to get the dissolved photoresist away from
the surface so that a fresh developer can get to the surface. Spray
developers eliminate this problem, but require far more chemicals
depends on resist thickness and type, developer concentration,
agitation rate, and temperature. For a 1-mm-thick positive resist
with a 3–2% concentrated basic developer it is roughly 1 minute.
Do not re-use or save positive developers because they can absorb
CO2from the atmosphere and become less effective. Common
bases are tetramethyl ammonium hydroxide (TMAH) and sodium
TMAH is sometimes known as an MIF, or metal ion free, devel-
oper. Developers may have surfactants in them to assist in wetting.
Develop time can determine the resist profile. Overdevelop
narrows CDs, underdevelop can leave residual films on the surface
or leave a thin film around the edge of the pattern. Plan on doing
several exposures on dummy wafers to determine the proper
exposure for your combination of substrate and resist. Typically,
dummy wafers are both over- and underexposed to bracket the
ideal exposure time/develop time around the optimum. Use
optical microscopes to inspect the resist pattern, but microscope
light can expose resist. Underdeveloped resist can be re-developed
to improve its appearance.
After developing, a rinse step must be done. Negative
resists use organic solvents to remove the residual developer,
positive resists use DI water. A minute or so of immersion is
usually adequate, but flowing fluids are best. In small labs, a
nitrogen blow gun is often used to dry the wafer after rinse
(see below about blow guns). Hold the wafer vertically using
clean tweezers with the bottom edge of the wafer on a special
clean room wiper. Blow from top to bottom, doing the front
and back alternatively. The cloth will wick off the last drops,
which are nearly impossible to blow off into the air.
Automated develop tracks spin the wafer dry, a much better
Resist patterns that are unacceptable can be stripped off and
re-spunevenafter exposure,a processknown as re-working.Often
a ghost pattern can be seen on the substrate where the original
This can be alleviated by doing a short oxygen plasma, often called
a descum. Typically 30 s at 400 mt/100 watts in pure oxygen is
affective. Such a process should remove 100–300 nm of resist. As
with all other processes, characterize it before committing wafers
to be good practice. Sometimes a descum is done before
depositing the resist to improve adhesion by roughening the
substrate or etching away organic films.
4.5. Hard Bake
Many resists require a ‘‘hard bake’’ after develop to harden the
resist for further processing. This step is important in wet etching,
but may make the resist difficult to remove. Hard baking is done
on the same equipment as the soft bake, but is usually done about
10?C hotter, with typical hot plate times for a 1-mm -thick resist of
about 2 min.
4.6. Resist Stripping
Photoresists are designed to be stripped off after transferring a
pattern to a material beneath the resist layer. Positive resists
are often easier to strip off than negative resists. (Photoresist
stripping steps are often left out of process run sheets.)
Piranha clean (detailed below) works very well to remove
both positive and negative tone photoresists, but can attack
underlying layers. Acetone strips positive resists that have not
had much post-processing that hardens resists, such as baking,
ion implant, or plasma etching. Acetone should always be
followed by an isopropyl alcohol (IPA) rinse. Positive resists
can also be stripped by exposing the developed pattern with a
blanket exposure in the alignment tool or another UV source,
Microfabrication Techniques for Biologists19
and stripped with the same developer used to delineate the
pattern in the first place. Acetone/IPA and re-expose/re-develop
are effective in stripping resists for rework. Many commercially
made photoresist strippers are made, but check with substrate
compatibility before using these products.
Another method to strip photoresists is to use an oxygen
plasma, sometimes known as ashing, which is just a long descum
process. Ashers can be much simpler tools than the plasma etchers
discussed below. Barrel etchers or small parallel plate etchers work
hydrogen, 85% nitrogen) to achieve a pressure of 100–500 mt is
used, with a power setting between 100 and 300 watts. Time
depends on resist type, thickness, processing, and the etcher con-
figuration and settings. Times can be from a few minutes to hours.
Many substrates are not affected by oxygen plasmas, other than
some minor sputtering, but some substrates can be etched in
oxygen, particularly polymers. Even mild ashing can damage sub-
strates and leave undesirable defects. Often a wet cleaning is done
after ashing to remove residual particles. Sometimesa wet cleaning
is done first to remove the majority of the resist, followed by a
short plasma clean to remove residue left by the wet stripper.
It is important to do long pump downs prior to introducing
the etch gas because substrates or films can outgas, changing the
chemistry of the plasma. Water and other contaminants on the
surface of the plasma chamber are slowly pumped away. Always
note the color of the plasma as a quality-control check. Oxygen
plasmas look bluish. If the system has a leak and nitrogen is pre-
sent, the plasma will have a pink tint to it.
All surfaces in a typical lab will have a thin film of organic
residue on them from the environment or non-volatile organic
for 30 s to 1 min or so can sputter away residues, and oxygen or
CF4plasmas can etch them away. Such activated surfaces will
return to their original condition within a few minutes of leaving
the vacuum chamber. Such a treatment can be very effective for all
types of substrates to improve adhesion and may be used just prior
to deposition of an adhesion promoter.
4.7. Substrate Cleaning
Cleaning wafers or substrates is often overlooked or viewed as an
afterthought, but doing it well can make a lot of difference in the
performance and yield of the final product. Wafers may need to be
cleaned several times before they are completed. New wafers
should be cleaned prior to any processing.
A common and very effective method to clean silicon or glass
substrates is using piranha clean. It is so named because it eats
almost anything organic. This is a mixture of 98% concentrated
two chemicals is exothermic. It destroys almost any carbon-based
compound and many metals, especially aluminum. The ratio of
acid to peroxide is from 2:1 to 10:1. Many labs have a
permanently set up bath for this process. The peroxide breaks
down into water and oxygen and must be periodically
replaced. To work well, the solution should look like a clear
bubbly soda pop solution (lemonade in Europe, Sprite in the
US). Needless to say, it is very dangerous to humans. Wear an
apron, acid-resistant gloves, and a face shield when working
with this material.
containers. All wafer holders placed into the bath must be made of
Teflon. Do not put plastic coated metal objects into a piranha bath.
Typically 5 min cleans any silicon or glass surface. Metal and plastic
substrates cannot be cleaned in piranha. It is highly recommended
thatcommercially available waferhandlingboatsandtoolsbeusedin
piranha clean baths.
Waste piranha bath can be stored in polypropylene or
polyethylene containers when cool. Do not re-use solutions after
removal from the original bath. Disposal is problematic because
the peroxide can remain active and create gas for many weeks after
the solution is cooled. Do not put waste piranha in a tightly sealed
A 5-min rinse in flowing deionized water removes sulfate from
the surface after cleaning. This is best done in a cascade or dump
rinser. Cascade rinsers flow copious amounts of pure deionized
water over a weir from back to front. Place the clean wafers in the
front (relatively dirtier) tank, and after 5 min transfer them to the
back (cleaner) tank. Dump rinsers fill the tank, let the water out
through the bottom and re-fill the tank. If no special rinse tank is
available, rinse in flowing water. Particles tend to float on the
surface of liquids, and any time you pull a wafer through a surface
between air and liquid it can pick up the particles on the surface,
analogous to a Langmuir Blodgett trough.
a boiling mixture of 25% ammonium hydroxide (NH4OH), water
and hydrogen peroxide (H2O2) can be used instead. Use a ratio of
roughly 1:5:1 for about 10 min. This is sometimes known as an
RCA1 or SC-1 clean and is often done in conjunction with an
ultrasonic or megasonic agitation. Ultrasonic energy can damage
some substrates. This clean removes particles and organics.
Some laboratories use hydrochloric acid based cleaning solu-
tions, such as HCl:H2O:H2O2in a 1:6:1 ratio, sometimes called
an RCA2 or SC-2 clean. (The process was developed by the RCA
Corporation, but is now known as a semiconductor clean.) This
clean was designed to reduce metal contaminants on the surface,
not typically applicable in micromachining applications. This
solution rinses off of surfaces much better than do sulfuric acid
Microfabrication Techniques for Biologists 21
Another way to clean a surface, especially silicon, is to etch
the native surface oxide off in very dilute hydrofluoric acid
(1%). Most metals and all silicon have an oxide on the surface.
Silicon oxide or silicon dioxide is hydrophilic or water loving.
After a few seconds in HF, the surface becomes hydrophobic or
water hating. This is easily seen because silicon wafers ‘‘dewet,’’
where the acid sheets off the wafer leaving an almost dry
substrate when pulled out of the acid. By etching away a thin
layer, particles or sulfate-laden layers left over from piranha
cleaning can be removed from a surface.
Small particles (below about 5 mm) CANNOT be blown off of
a surface with an air stream or flowing liquid. Fluid streams have a
velocity of zero near the surface. Below about 1 mm, the adhesion
force of a particle is far greater than any force reasonably possible
with a moving fluid stream.
4.8. Rinsing and Drying
Rinsing and drying are perhaps the single most important step in
any wet process. Dedicated wafer processing labs have clean water
delivered in special piping, detailed below. Clean tanks, glassware,
and other hardware are also very important for getting clean
surfaces because a fingerprint inside a rinse tank can lead to an
let water drops dry on a wafer surface. DI water is very corrosive
and will leave water stains.
A very good tool for rinsing a wafer after a cleaning or wet
etching process is the spin rinse dryer. After an immersion rinse
for 5 min in flowing DI water, substrates in a boat are spun
while being rinsed with more DI water. A final high-speed spin
with flowing filtered nitrogen finishes the process and leaves
clean, dry wafers. Since entropy always increases, a freshly
cleaned wafer surface is a magnet for particles. Immediately
place the clean wafers in a clean container. As with all tools,
verify that the machine is clean by running clean dummies in it
before processing your wafers.
More modern facilities use variations on alcohol dryers.
These tools are based on the Marangoni effect, which is
defined in the IUPAC Compendium of Chemical Technology,
2nd ed. 1997, as ‘‘Motions of the surface of a liquid are
coupled with those of the subsurface fluid or fluids, so that
movements of the liquid normally produce stresses in the
surface and vice versa. The movement of the surface and of
the entrained fluid(s) caused by the surface tension gradients is
called the Marangoni effect.’’ These dryers can leave very clean
dry surfaces and lower a lab’s water usage. However, they are
larger tools and not often found in smaller laboratories, but
are the ideal tools for drying wafers. Effectively, alcohol dis-
places the water on the surface of a wafer and surface tension
effects remove particles.
5. Thin Film
At the heart of most types of microdevice fabrication are vacuum
systems. Thin films of metals and oxides and silicides can be
deposited using vacuum systems. Many materials can be etched
in vacuums, thus it is instructive to review some basics of vacuum
systems. The somewhat antiquated unit of torr is still in use in
vacuum systems. One atmosphere=760 torr=101 kPa, one
torr=133.322 Pa, 1 Pa=1 newton/m2.
Pumps can be categorized by the basic pressure they reach,
which roughly corresponds to the three regions of pressure: low
vacuum (atmosphere to 10–3torr=1 mt), high vacuum (10–3to
10–7torr), and ultrahigh vacuum, (10–7torr and below). Most
wafer process tools operate at base pressures in the high-vacuum
Leaks into vacuum systems from cooling systems, bad fittings
and feedthroughs are common problems. Another common issue
is contamination in the chamber. Fingerprints and other surface
contaminants can be ‘‘virtual leaks.’’ A fingerprint on the inside of
a clean vacuum system can outgas and coat the entire chamber and
your wafer, with a thin film of oil, making it very difficult to
reproduce some results. Another type of virtual leak is from air or
gas trapped underneath a screw in a blind hole, or gas trapped
inside a microstructure. Films, particularly polymer films, can also
outgas and slow down pumping times. Every time a vacuum
chamber is opened the walls absorb water and other atmospheric
contaminants which must be pumped away when the system is
closed up. All systems leak, due to outgassing and poor seals. The
simplest way to leak check a system is to pump it down, close all
valves to all pumps and watch how fast the pressure rises. If it rises
faster than what was historically normal for the system, a leak may
be present. Systems need to be opened and manually cleaned on
occasion, depending on their use.
regime, where a gas behaves as a fluid and the mean free path of a
molecule is much much smaller than the chamber size. In the
molecular flow regime, molecules do not interact with each
other and the mean free path of a molecule bouncing around
inside the chamber is much greater than the size of the chamber.
There are several types of vacuum pumps, each with its advan-
tages and disadvantages.
The most basic and common type of vacuum pump is the
rough pump. These pumps are designed to move a lot of gas
quickly from atmospheric pressure to about 50 mt, roughly
where viscous flow ends. Some simple plasma etch systems have
only a roughing pump. These pumps have a rotating shaft and
Microfabrication Techniques for Biologists23
chamber immersed in oil that moves gas molecules from the
low-pressure side to the high-pressure side. When used for
pressures below about 100 mt, any pump with oil in it will have
some backstreaming of oil into the vacuum chamber. This can be
minimized with the use of a trap that absorbs the oil. The trap
requires routine cleaning and maintenance. Two types of
lubricants are used in roughing pumps – hydrocarbon and fluori-
nated oils. Hydrocarbon oils are cheap, but cannot be used in
systems that pump oxygen or other reactive gases. Fluorocarbon
oils are very expensive, but are also very inert. It is our experience
that pump oils are rarely changed in university laboratories.
Changing the oil can improve pumping capabilities, reduce
chamber contamination, and increase pump life. Treat used
pump oil like toxic waste. Modern systems use more expensive
oil-free dry roughing pumps.
High-vacuum pumps come in three common types, the oil
diffusion pump, the cryogenic pump and the turbomolecular
pump. All high-vacuum pumps operate in the molecular flow
region, so these pumps all rely on gas molecules randomly mov-
ing to the pump. All high-vacuum pumps take a system from
?50 mt to the high-vacuum regime, so the majority of the gas
in the chamber must be removed with a roughing pump prior to
opening the valve that exposes the chamber to the high-vacuum
pump. Oil diffusion pumps have been around for many years, are
cheap and work very well. They take a long time to heat up
when cold and have limited pumping capacity. Sometimes these
pumps burn the oil and must be cleaned. Oil diffusion pumps
can reach pressures less than 5?10–8torr. Cryogenic pumps or
cryos are very expensive and remove gas from a chamber by
freezing it out on a material that is held near the temperature
of liquid helium, around 20 K above absolute zero. They are
very ‘‘clean’’ pumps since they add no contamination whatsoever,
but some cannot pump out helium very well. On a routine basis
cryogenic pumps must be regenerated, where the cold head is
warmed up so the frozen gases can evaporate. This takes several
hours and can result in considerable system downtime. Cryo
pumps can achieve pressures as low as 5?10–9torr. Turbomo-
lecular or turbo pumps have increased in popularity recently
because of improved reliability, cleanliness, and pumping speed.
They are effectively a small jet engine, driven by a motor. Some
modern systems use a magnetically levitated turbine, limiting
friction and particle generation from the bearings. Turbos can
achieve pressures of less than 5?10–9torr. An additional type of
pump is the titanium sputter pump. Titanium is a very reactive
metal, so during a sputter operation it reacts with many gas
molecules, depositing the gas on the side of the chamber. In
any sputterer with a titanium target, a short titanium sputter can
dramatically decrease the vacuum pressure.
Vacuum system design is a complex subject because the
way a chamber and its pumping lines are laid out can drama-
tically affect the way a system pumps. The foreline (the pump
line leaving the large process chamber) and all its associated
components (valves, traps, gages) are critical design parameters
that can affect ultimate vacuum, cleanliness, pumping speed,
reliability, and cost. All system dimensions depend on the
pumps used and the desired operating characteristics of the
5.2. Thin Films
To deposita filmofmetal, oxide,or semiconductor ontoa surface,
the material must be made into a gaseous or plasma state. This can
be done by heating the metal above its boiling point or by knock-
ing atoms off of a surface using an ionized gas. Most physical vapor
deposition (PVD) systems are designed to deposit metals. A second
way to deposit a film is known as chemical vapor deposition (CVD).
CVD is done by flowing a reactive gas across the surface, and
supplying thermal or plasma energy to break down the gas, thus
reducing it to a desired film material. Some metals, such as tung-
sten, can be deposited by CVD, but the most common films
deposited by CVD are silicon oxides, silicon nitrides, polysilicon,
and amorphous silicon.
Electroplating deposits metal films by reducing ions of a salt
in a solution. Electrophoresis and dielectrophoresis use an electric
field to pull polar or non-polar molecules out of a solution and
make them stick to a surface. These techniques are not dealt
5.3. Physical Vapor
Evaporators can be very simple systems, as shown schematically in
Fig. 1.9. All that is needed is a chamber with a good vacuum, at
least 10–6torr, and a method to melt a metal. The two major types
of evaporators are thermal and electron beam. In a thermal
evaporator, a small quantity of the film forming metal is put
into a boat, typically a refractory metal such as tungsten or
it up. The metal melts and evaporates since the chamber is free of
gas. The better the vacuum, the purer the metal film deposited, so
it is advantageous to pump for a long time before beginning the
The second type of commonly found evaporator is the
electron-beam evaporator. It uses a refractory metal filament to
boil off electrons, which are then directed by a magnetic field
into the boat containing the source metal. E-beam evaporators
are more complex than thermal evaporators, but almost any metal
can be evaporated, regardless of melting temperature. Commonly,
carbon crucibles are used to hold the metal charge. E-beam
systems may have large, moving substrate holders and are used in
high-volume production environments.
Microfabrication Techniques for Biologists 25
The only variables to control are the current through the
heating element, the distance from the source to the substrate,
and the pressure. One disadvantage of evaporators is that they
always require that the substrate be above the source, thus
fixturing is required to hold the substrate. Evaporation is a line
of site technique, so sidewall coverage can be poor. Evaporators
can have very high deposition rates and can deposit very pure and
very thick metals. Heating of the substrates is rarely a problem,
unless the source–substrate distance is too close.
Another way to deposit films is using a sputterer, such as that
shown in Fig. 1.10. These systems are more complex than
evaporators, but allow better control over the film properties. A
metal target is attached to a ‘‘gun,’’ usually a magnetron type. The
magnets in the gun increase the plasma density near the target
surface, increasing the bombardment of the target with positive
gas ions, increasing the sputter rate. Targets are typically discs 2 or
3 inches in diameter and can be fairly inexpensive for common
metals such as aluminum and titanium. For rare metals such as
gold, sputterers are much more efficient in the use of metals
than evaporators are. (Note that iron, nickel, gadolinium, and
cobalt – the ferromagnetic metals – may be difficult to sputter
Theatmosphere isevacuatedand an inertgas at lowpressure is
allowed to backfill the chamber. The most common sputter gas is
argon. A dc or RF plasma is generated between the gun and the
substrate. Some machines sputter up, requiring fixturing for the
a drawing of the hearth of an electron beam gun, which replaces the resistively heated
source in the thermal evaporator. E-beams can even melt refractory metals. Carbon
crucibles or refractory metal crucibles are often used to hold the source metal.
substrates, some sputter down. The plasma is often struck at a
high pressure, 15–20 mt. After stabilizing, the pressure can be
dropped to 2–3 mt for the actual deposition operation, since
sputter rates go up as the pressure is dropped. Ions in the
plasma knock atoms off of the sputter target by momentum
transfer, creating a ‘‘gas’’ or ‘‘cloud’’ of atoms. These atoms
then condense out on all surfaces inside the sputter chamber.
Most machines also rotate the substrate during deposition so
that the material coats the side walls of microstructures, thus
sputtering has better step coverage than evaporation does.
evaporation, such as pressure, gas type, substrate temperature,
dc bias, power, and source–substrate distance, which allow for
finer control of film properties. Metals and conductors are
usually sputtered with a dc power supply, from 50 to 500
watts. Dielectric materials such as oxides can be sputtered
using a radio frequency (RF) power supply. Many materials
can be sputtered that cannot be evaporated. Compound
materials can be sputtered, but the composition and crystal
structure of the resulting film may vary from that of the original
target material. By introducing a reactive gas, such as nitrogen,
materials such as titanium nitride can be deposited from a pure
titanium target. Another way to make compounds in a sputterer
is to run two targets simultaneously. To adjust the stress in
the deposited film, some sputterers have heating capabilities in
the substrate holder. Another advantage of sputterers is that the
substrate can be cleaned by sputtering away some of the
substrate material before depositing the desired film.
Fig. 1.10. A magnetron sputtering system. A vacuum load lock keeps the inside of the
chamber at high vacuum while the substrates are being loaded. This improves vacuum
quality and speeds up processing.
Microfabrication Techniques for Biologists27
micromachining because of their chemical inertness, high conduc-
tivity, and high work function. However, the chemical inertness of
these metals means that they do not adhere well to many sub-
strates. Very thin chrome, tungsten or titanium films (50 nm) are
often deposited before depositing these metals to act as adhesion
layers. However, these metals are more reactive than noble metals
and may cause processing problems.
Films deposited by any technique may have stress in them,
which can bow the substrate or cause films to crack (Fig. 1.11).
Stress is controlled by deposition parameters, temperature, and
film chemistry. Annealing a film may change its stress level.
5.4. Chemical Vapor
The advantage of oxides, nitrides and polysilicon films for use in
microdevices is that they are not reactive, are easy to deposit, and
are widely available. The disadvantage of these films is that they
generally require high temperatures, vacuums, and expensive
equipment to deposit. Many chemical vapor deposition (CVD)
processes rely on SiH4, silane, as the source of silicon. This gas is
uses this dangerous gas diluted in a carrier gas such as nitrogen or
CVD is subdivided into many different technologies. Atmo-
spheric pressure (APCVD), is good for oxides and requires
?350?C heat to crack the feed molecules. Low pressure
(LPCVD) can deposit silicon oxide, silicon nitride, silicon, and
tungsten, as well as silicides, and requires ?550?C heat to provide
reaction energy, but provides high quality films. Plasma enhanced
(PECVD) usesa plasma to provide theenergy to crack the feedgas
and can be done at temperatures as low as 100?C, but film quality
may be poor at lower temperatures. It is best at depositing oxides
and nitrides. Metal organic (MOCVD) is used for depositing
compoundsemiconductorfilms epitaxiallyon crystalline
Fig. 1.11. Stresses that can develop in deposited films.
substrates with an approximate lattice match to the film being
deposited. CVD reactors need frequent manual cleaning to mini-
mize particle deposition on substrates.
Plasma deposition equipment is very similar to plasma etching
equipment, and both processescan bedone in thesame chamber, if
the substrate is sufficiently heated and the right gases are plumbed
in. Machinery dedicated to one or the other tends to work better.
In silicon processing, the best dielectric thin films are oxides
created by growing a film on a clean silicon wafer at temperatures
over 1,000?C by introducing oxygen or water. Dry oxygen gives
the best quality films, but ‘‘wet’’ oxides grown with water can be
made thicker because of the higher diffusion of water through the
existing oxide.These filmsareoftenused as gatedielectricsand are
not often needed in micromachining. These thermal oxides grow
on the back of the wafer as well as the front.
As with all processes, you may be limited in what films you can
deposit by the equipment available and substrate temperature
limitations. Keep in mind thermal expansion issues as substrates
and films heat and cool.
Electroplating can be thought of as the opposite of wet etching. If
metallic salt, metal ions can be reduced to metals on the conduct-
ing film as shown schematically in Fig. 1.12. It is usually done in
Fig. 1.12. A simple electroplating bath. Due to poor film uniformity, the geometries must
be overplated. Final thickness is determined by a process that grinds off the top of the
wafer until the correct thickness is achieved, known as chemical mechanical polishing
Microfabrication Techniques for Biologists29
organic solvents. The metals that can be deposited from any sol-
vent must have a reduction potential less than that of the solvent.
In other words, the water will break down into oxygen at the
anode and hydrogen at the cathode before the metal ions will be
reduced to metal at the cathode. Aqueous electroplating is usually
limited to iron, nickel, copper, chrome, silver, and gold. It can be
very cheap to do, requiring little more than a beaker, electrodes,
solution, and power supply. Electroplating can deposit into very
deep trenches, unlike sputtering and evaporation. Control of
metallurgical properties is difficult because the plater has little
control over crystal structure. Plating solutions contain many
additives to control film properties. Very thick films can be
grown this way, but uniformity is very poor, controlled by the
non-uniform current density.
The IC industry has replaced sputtered aluminum with
electroplated copper in what is known as the Damascene process,
named after the famed Damascus swords with their swirled
patterns in the metal. Copper conducts better than aluminum,
and will not suffer electromigration (breaking of lines in high
current applications due to atomic drift) but cannot be plasma
etched. Holes are etched into an oxide down to a plating base and
copper is plated up into the holes. The plating base is often a
silicide, used to prevent copper from diffusing into silicon. After
plating, the wafer is flattened, or planarized, using a process called
lapping or chemical mechanical polishing, CMP.
Etching is the removal of materials from a substrate and can be
divided into two categories, wet and dry. Wet etching can be done
under the most simple laboratory conditions in a beaker, whereas
dry etching can involve considerable capital and maintenance
6.1. Wet Etching
In wet etching a few variables can be controlled, such as the
chemical composition and concentration, as well as temperature.
Agitation can be an important factor in some wet etching situa-
tions. Liquid etchants can have very high etch rates compared to
plasmas. Most useful etchants are concentrated mineral acids, such
as hydrochloric (HCl) and hydrofluoric (HF) and nitric (HNO3)
and sulfuric (H2SO4) acids. We recommend that these chemicals
be purchased in semiconductor, electronic, or clean room grade.
Etchants for some metals such as chrome can be purchased ready
to use, sometimes etchants must be mixed in the laboratory. An
excellent reference book is the CRC Handbook of Metal Etchants
(1), which contains recipes for etching semiconductors as well as
metals. Remember the AAA safety rule of always adding acid to
water: never add water to an acid because the exothermic reaction
can splash hot acids on you. Also, be very careful when mixing
acids and organic solvents, as sometimes these mixtures can be
explosive. Always dispose of acids and solvents in an environmen-
tally acceptable manner. One driving force in industry to convert
from wet etching to dry etching is the elimination of liquid wastes
from wet etching operations, since environmentally sound acid
disposal can be very expensive. Some factories have on-site acid
recycling facilities. However, dry etching uses materials that are
often discharged directly into the atmosphere, which can
contribute to atmospheric pollution and potentially ozone
depletion. With proper abatement equipment, such as burn
boxes and scrubbers, plasma etching can easily meet environmen-
tal regulations, albeit at a rather high cost.
Hydrofluoric acid, HF, deserves special mention for safety.
Although not one of the strongest acids, it is commonly used in
49% concentration, it is almost always diluted prior to use as an
etchant. HF does not immediately cause skin burns and it has the
appearance of water. However, HF soaks through the skin and
attacks bone tissue. It may leave a red rash on the skin, but little
sign that it is doing damage deep within the body. It also report-
edly attacks eye tissue, so when working with HF extra special care
must be taken to keep this dangerous acid away from human
the glass. HF can be safely used in Teflon or other fluoropolymer
labware, polypropylene, or polyethylene. Since the cost of
fluoropolymer labware is so high, we have had excellent results
using inexpensive polyethylene plastic food storage containers
with snap fit lids for storing and using acids and solvents in the
departmentstores. Although we havenever specificallytested such
cheap containers for contamination issues, we have never noted
any problems caused by them.
Perchloric acid (HClO4) finds some use in micromachining,
a special hood, as its crystals are explosive.
The water present in HF oxidizes metals to their oxide, then
the F–ion reacts with the oxide to form a soluble byproduct.
Sometimes HNO3(nitric acid) is added to acid etchants to oxidize
metals, such as in aqua regia (royal water), a mixture of hydro-
chloric and nitric acids used to etch noble metals. HF/HNO3
mixtures will attack silicon. Pure HF etches silicon oxides and is
sometimes sold in a buffered mixture with NH4F and surfactants
in a 2:13 ratio known as buffered oxide etch (BOE) or buffered
Microfabrication Techniques for Biologists 31
hydrofluoric (BHF). If you mix your own acids, be aware of the
concentration. HF, for instance, can be supplied in many different
concentrations in water, so take the diluent into consideration
when mixing. Whenever any two acids are mixed, the reaction
may be exothermic, so carefully consider the type of container
used in mixing. Always wear heavy acid resistant gloves, eye
protection and an acid-resistant lab apron and work in a fume
In all etching, selectivity is an important parameter, the
difference in etch rates between the material that is desired to be
etched and other materials present that should not be etched, such
as the etch mask and underlying layers. In wet etching, selectivity
between the etch mask, usually photoresist, and the substrate can
be almost infinite. In plasma etching, most materials present in the
plasma will be etched to some degree. Wet chemical etching is
usually an isotropic technique, with the material being etched
horizontally as well as vertically at the same etch rate. This
horizontal etching is responsible for much of the dimension
change discussed above. The thicker the layer being etched, the
larger the dimensional change (Fig. 1.13). Plasma etching can be
way the plasma is set up.
Fig. 1.13. Wet and dry etching. Wet etching is isotropic for most materials, etching in all
directions at the same rate. Dry etching can be anisotropic, with the vertical etch rate
many times higher than the horizontal etch rate. The etch mask is usually eroded during
In silicon and other crystalline materials, many liquid etchants
have been developed that are highly specific to certain crystal
planes or to the dopant level in the silicon. The classic is the
warm KOH etch (Fig. 1.14). When used with (100) silicon, the
(111) planes are etched very slowly. Details of KOH etching and
some decorative etches can be found elsewhere (2, 3).
6.2. Metal Etchants
The metals most often encountered in device processing are
aluminum, gold, platinum, and copper. Titanium, chrome,
tungsten, and TiW (10%Ti, 90%W) are often used as adhesion
metals underneath a noble metal. Nickel is used because it is easy
to electroplate, but hard to sputter because it is a ferromagnet.
Aluminum (usually as an aluminum ?1% silicon alloy) is used in
integrated circuits as a conductor because it is easy to deposit, has
low resistance, does not tend to diffuse into silicon, and can be
plasma etched. Copper was not traditionally used, as it tends to
diffuse easily in silicon, putting states in the band gap. Copper
cannot be plasma etched because there is no volatile copper com-
pound. Copper has come into use recently using the Damascene
process, which avoids etching (see above). Tin doped indium oxide
(indium tin oxide (ITO)) 90% In2O3+ 10% SnO2is generally
sputtered and is used as a transparent electrode in electrolumines-
centdevices. Cobalt silicideis listed becauseit andother silicidesare
often put down under copper as a barrier to diffusion. Other metals
are used, but they tend to be specific to an application.
pure, but forms hydrochloric acid when water is added). The recipes
that follow are generally based on the most concentrated chemicals
Fig. 1.14. Directional etching of silicon using KOH, (potassium hydroxide). KOH prefer-
entially etches the (100) plane of silicon relative to the (111) plane. KOH also etches n
(phosphorous) doped silicon much faster than p (boron) doped silicon, so a thin
membrane can be made by etching through the back of the wafer to the p++ (heavily
doped p-type) silicon, which is called an etch stop.
Microfabrication Techniques for Biologists 33
available. For HF, this is 49%, H2O2(hydrogen peroxide) 30%, HCl
38%, HNO385%, H3PO4(phosphoric acid) 85%, H2SO498%,
NH4OH (ammonium hydroxide) 28.5%, CH3COOH (acetic acid)
98%. NaOH (sodium hydroxide) and KOH (potassium hydroxide)
are sold as solids. They will attract moisture from the air, so keep the
containers tightly closed. Recipes are specified by volume:volume
ratios. These recipes are starting points. Individual recipes may be
modified for specific effects. Dilution typically slows etch rate. Some
be very very careful when boiling an acid. As a general rule, one does
in thickness, and very high etch rates tend to overetch one part of a
on how the film was deposited, purity, and heat treatments. A list of
treatments is shown in Table 1.1.
Basic etching recipes for common materials
1. GoldAqua regia, tri-iodide
2. Platinumaqua regia, chloroplatinic acid/lead acetate to oxidize platinum
3. Nickel1:1 Nitric/acetic acid, HF/nitric mixes, CAN1
4. Coppernitric/sodium chlorite, nitric/hydrochloric acids etch copper in various
concentrations, as do FeCl3solutions.
5. Chromium-Conc. HCl, commercial etchants, CAN2
6. Titanium HF in any concentration, 20% H3PO4, 25% formic acid, 20% sulfuric acid
7. TungstenH2O2, 1:1 H2O2:HF, conc. H2SO4, 1:4 HNO3:HF
8. TiWH2O2, aqua regia, 1:2 NH4OH:H2O2
9. Aluminum10% NaOH or KOH, HCl, PNA
10. Indium Tin Oxide
Conc. H2SO4, piranha etch, conc. HCl, 1 M oxalic acid, 55% HI (hydroiodic
acid); see note below
11. Cobalt silicideSee note below. Any of the concentrated mineral acids mixed with H2O2
should etch the material.
12. Silicon1:1 to 1:10 HF:HN03, KOH (selective), HNA
6.3. Dry Etching
Plasma, or reactive ion etching (RIE), etching is a complex
subject, and the results obtained depend on many factors,
including gas type, pressure, power, dc bias, electrode spacing,
substrate type, and chamber configuration, as well as the
ability of the machine to control pressure and gas flow. To
the engineer designing a process, the etch equipment available
and the gases plumbed into it ultimately determine what kind
of etching can be done.
Table 1.1 (continued)
13. Silicon oxides –
Dilute HF, BOE
14. Silicon nitrides –
H3PO4+few %H20 at 160?–180?C. HFalso etches this nitride, but etchrates
vary depending on how the film was formed.
1. Aqua Regia – 3 parts HCl, 1 part nitric acid. Mixture may be explosive.
2. Tri-iodide – 400 g KI, potassium iodide, 200 g I2, iodine solid, 1,000 ml water
platinum in biology is as an electrode, usually coated with platinum black, which increases the current
available through the electrode. Platinum electrodes are typically oxidized electrolytically in a 3% solution
of chloroplatinic acid, H2PtCl6.6H2O. The addition of a small amount of lead, copper, or mercury salt
increases the available current, for example, lead acetate at 0.06% in solution.
4. CAN1 – ceric ammonium nitrate (NH4)2Ce(NO3)650 g, 10 ml HNO3, 150 ml water. Note that the
Handbook of Metal Etchants gives the formula for CAN as 2NH4NO3.Ce(NO3)3.4H2O, showing one
less NO3–group than the material that can be purchased from standard chemical catalogs. The four H2O
groups attached aid in dissolution, but the material is not generally sold as a hydrate.
5. Nitric sodium chlorite – 375 ml HNO3+ 150 g solid NaO2Cl plus water to make 1 l.
6. Chromium etchants are usually purchased mixed from a vendor that specializes in these etches. They are
based on ceric ammonium nitrate and are designed for minimal undercut and high selectivity. Chrome is
used for photomasks and as an adhesion layer, hence the need for good selectivity to gold and platinum.-
CAN2 – ceric ammonium nitrate 10 g, nitric acid 100 ml, 1,000 ml water.
7. BOE or BHF, buffered oxide etch or buffered HF – typically contains 13:2 NH4F:HF or a similar ratio.
8. HNA – Hydrofluoric nitric acetic is the classic silicon etch. Nitric oxidizes the silicon, HF etches the
oxide, and acetic acid is a pH buffer. Etching of silicon is so common that the reaction is given here: Si +
HNO3+ 6HF= H2SiF6+ HNO2+ H2O + H2(gas). The ratios are HF 8%, nitric 75%, acetic acid 17%.
9. PNA – Phosphoric nitric acetic – 80 parts phosphoric, 5 parts nitric, 5 parts acetic, and 10 parts water.
Commercially available as a mixed acid, the most common aluminum etchant.
10. KOH – 7–8 Molar at 80?C with stirring. 450 g KOH/liter of water. Many concentrations will work,
but 6–8 M have the best uniformity.
11. # ITO is generally plasma etched, often in CH4/H2and argon mixtures, generally thought to be
primarily a physical etch, rather than chemical. ITO is almost always deposited on glass, so any etchant for
the doped oxide will also etch the glass substrate. Piranha etch is the same as the piranha clean, 4:1 to 10:1
12. +Cobalt silicide and other metal silicides are now used in silicon device processing and are usually
plasma etched due to the small geometries generally used. Cobalt silicide has the highest conductivity of
the silicides. Due to lack of volatile cobalt compounds, plasma etching is difficult, although chlorine and
other halogen plasmas have been used with a heated substrate, as CoCl2is volatile at 200?C. CF4/O2
plasmas are also reported to work. We would also expect HF/HNO3mixtures to etch the film, but have
not tested this mixture.
Microfabrication Techniques for Biologists35
As with all microdevice processing, repeatable results depend
on having machinery that operates consistently. During plasma
etching, the byproducts of the etch are removed in the flowing gas
and ideally exit the system through the pump exhaust. However,
many etch byproducts are not volatile and deposit inside the
chamber, in the pump lines, and in the pump. Most laboratories
specify routine plasma cleaning of chambers using an oxygen/CF4
mixture. This cleaning can remove only some deposits, so it is
prudent to open the chamber on occasion and check for deposits
on the walls and electrode, which can seriously degrade the etcher
performance. Removing these deposits by scrubbing and wiping
can have a large effect on the characteristics of the plasma. Some
etchers introduce gas through a showerhead-type electrode. The
holes can become plugged and must be cleaned out occasionally.
Plasma etchers come in many different styles and configurations.
Simple machines such as barrel etchers create a plasma by wrapping a
large cylindrical chamber with a radio frequency (RF) coil, usually
though the glass chamber. These machines have low plasma density
modifications. Most control pressure by adjusting the flow rate of gas
One advantage of the inductively coupled barrel etcher is that no
in modern etchers, where a plasma is generated in an ICP (inductively
coupled plasma) head and driven into the wafer by an additional
potential applied to the wafer holder below the ICP head (Fig. 1.15).
Much modern plasma etch equipment operates in this manner.
Fig. 1.15. A simple plasma etcher. Such systems often have only one pump and cannot
reach pressures below 50 mt. These are best used for descum, stripping photoresist or
activating a surface.
Another tabletop plasma reactor is a parallel plate etcher,
which has a chamber large enough for a single wafer, and thus a
higher density of plasma (Fig. 1.15). Most of these types of table-
top systems only have a single pump, and thus are limited to
pressures above 50 mt.
Another configuration of modern plasma reactors uses an
electron cylclotron resonance (ECR) chamber on top of the actual
etch chamber to create a plasma, which is accelerated to the
substrate by a powered substrate chuck, similar to the ICP reactor
of Fig. 1.16.
Whenever a plasma is created with an RF power supply, a
matching network must be in the system. This variable capacitor
matches the feed from the power supply to the power reflected
back to the power supply by the chamber. The reflected power
should be less than 10% of the supply power. Many machines tune
automatically, but some machines require manual tuning.
Running a dummy wafer can get the tuning adjusted to an
approximate setting before an important wafer is placed into the
system. Do not operate a system that cannot be tuned properly.
Most ICP or ECR etchers have a load lock which loads wafers
into the chamber without breaking vacuum. These robots only
work with standard wafer sizes, so if you are etching a small or odd
sized substrate, it may be necessary to put it on a larger wafer.
Special pastes are made that contain a mixture of high vacuum
Fig. 1.16. A modern ‘‘deep reactive ion etcher’’ DRIE. The plasma is created using an RF
chuck. Such systems can etch trenches into wafers with vertical walls and high aspect
ratios (up to 20:1).
Microfabrication Techniques for Biologists 37
grease and metal particles. These work well for gluing small pieces
to a large substrate, since they provide both good thermal and
electrical contact between the carrier wafer and the small substrate
6.4. Plasma Processes
During a plasma etch, three different processes can take place.
Pressure is the primary determinate of which process dominates,
with gas chemistry being a secondary variable. The relative con-
tribution of each process depends on pressure, power, and
temperature, as well as gas flow rate and type, and chamber
When etching a substrate, the parameters to control are selec-
tivity, etch rate, uniformity, and the profile of the sidewall of the
remaining material. Uniformity is determined by how uniform the
power and gas distributions are across the wafer surface. From a
a much lower etch rate (high selectivity) than the material being
etched (Fig. 1.13).
Plasma etching usually refers to higher pressure processes in
simpler machines, typically between 200 and 2,000 mt. Reactive
ion etching is the dominant mechanism in the 50–100 mt range,
and sputter etching takes place below ?50 mt. Ion beam etching
involves three electrodes: the plasma is generated in a separate
chamber and accelerated toward the substrate by a grid or series
At very low pressures, sputter etching removes material by
bombarding the surface with neutral atoms, knocking out atoms
of the substrate. It is a not a very selective process, depending
primarily on the binding energy of the atoms in the substrate.
Higher electrode potentials increasesputter etching at the expense
of selectivity and possibly damaging the surface. Sputter etching
can be very directional or anisotropic. Atoms sputtered away leave
the vicinity of the surface due to their high energy and condense
out of the plasma on chamber walls and in the pumping system.
Almost any material can be sputter etched, but may contaminate
the etch chamber. A common problem is that sputtered atoms can
re-deposit on the wafer creating micro masks. The result is known
as grass, a very rough surface in etched-out areas.
Chemical etching is exactly what it says: ions and radicals react
removed in the flowing gas. Chemical etching is limited bythe fact
can be pumped away. If the reaction leaves a solid material, it
cannot be chemically etched using a plasma. Chemical etching
tends to be isotropic, dependent primarily on gas composition
and how effectively the plasma is ionized. The primary purpose
of the plasma is to create ions and radicals, thus chemical etching is
a weak function of power, but the etch rate tends to increase with
pressure. The ratio between neutral and ionized gas species in a
glow discharge plasma is something like 104–106:1.
Ion assisted or ion enhanced etching is found during simulta-
neous reactive etching and etching enhanced by ion bombard-
ment. The etching of the substrate is chemical in nature and the
reaction rate is determined by bombardment of energetic ions.
Thus lower pressures and higher powers tend to increase the etch
rate. The etch rate can be enhanced by the application of a bias
across the plasma, but such capability is found only on advanced
Highly directional etching, in silicon known as Bosch etching,
takes place when an etch-inhibiting polymer is deposited on the
substrate (2, 3). In one step, a polymer is deposited uniformly
across the surface; subsequently, the chemistry of the plasma is
changed to a reactive mode which attacks the polymer and the
substrate. Since the polymer deposits uniformly, but the etching is
primarily vertical, deep trenches with nearly vertical walls are pos-
sible with this technique.
Many different gases can be used in plasma etching. Oxygen
is the most common, and reacts well with all polymeric materials.
CF4, sometimes known as Freon 14, is also very common, as are
C2F6, C3F8, C4F8, and SF6. The commonality is that all these
gases are a good source of fluorine. Fluorine radicals are one of
the most reactive chemicals known, and the only thing that can
effectively react with oxides. Chlorine and its compounds are also
frequently found in plasma etch systems, since many chlorides,
particularly aluminum trichloride (AlCl3), are volatile. Chlorine
tor etching. Chlorine and fluorine chemistries are not typically
compatible in a single chamber. Chlorine etched metals need to
be handled carefully, as residual chlorine may remain on the
surface once the substrate is removed from the vacuum. The
chlorine can react with atmospheric moisture creating HCl
which will corrode the metal pattern and substrate. Rinsing in
water can remove the HCl. Treating the substrate with an oxygen
plasma before removal from the vacuum will also alleviate the
In ion beam etching in an ion mill, a plasma is created in a
chamber above the substrate and accelerated toward the substrate
with a grid or series of grids. These machines can have high etch
rates and etch difficult materials.
One way to avoid etching metals is to do a liftoff process. With
careful resist processing, the third profile shown in Fig. 1.3 can
be obtained. By putting the photoresist pattern where the metal
does not go, and then evaporating a film, a metal pattern is
defined. Sputtering does not work as well in liftoff process
because it is less directional than evaporation and will coat the
Microfabrication Techniques for Biologists39
sidewalls of the resist pattern. The resist is then dissolved away in
acetone, floating off the metal on top of the resist, leaving a well-
defined metal pattern. The edges of the metal pattern may be
An area where there is much misunderstanding and myth is in
contamination control. Simply by doing a process in a clean
room in no way assures contamination free processing. Workers
who understand contamination control can do much better work
in anon-clean roomthanpoorly trained workers can doin thebest
clean rooms. Industrially, billions of dollars are spent to eliminate
every source of contamination, and extreme practices must be
utilized to obtain and maintain tools, wafer handling equipment,
people, and rooms at levels of cleanliness acceptable for the pro-
cess. In smaller labs, such practices become burdensome and
expensive. The level of cleanliness depends on the processes
being carried out. One must decide what is ‘‘clean enough.’’ The
10% rule applies in thin films as well as in x-y plane geometries – a
defect 10% of the thickness of a film may cause problems with it.
There are several types of contamination. The most common
is particles. Particles are ubiquitous in the environment, and the
smaller the size, the larger the number of particles. They come
from bacteria, people, abraded surfaces, aerosols, and especially
in half. There are two large pieces, a few small pieces, and thou-
sands of tiny crumbs. The smallest and most numerous particles
cause the most problems and are hardest to eliminate. Preventing
particles from getting onto a surface is much better than trying to
remove them later. Thermodynamics requires that a clean surface
become dirty, so a great deal of effort in any clean room is
involved in removing the entropy increase that comes with
The second type of contamination is ionic. This source of
contamination is not a large problem in micromachining, but is a
significant problem for the integrated circuit industry. Ionic
contamination comes from people, processes, and chemicals.
High-grade chemicals made for the semiconductor industry are
purified for trace elements at the factory, are filtered, and shipped
in specially engineered bottles. These come with many different
monikers, but all have specification sheets that tell what levels of
ionic and particulate contaminants are present. ‘‘Off the shelf’’
chemicals, even spectrophotometric grade, are not as pure as
chemicals made specifically for microelectronic processing.
The third type of contamination is known as non-volatile
residue (NVR). All solvents have residues in them, even various
microelectronic grades. It is easy to see how much is present by
evaporating a drop of any solvent on a clean silicon surface and
observing it under intense light. A single fingerprint on the inside
of a chemical bottle, wafer boat, tweezers, or process chamber can
add NVR to wafer.
Before spending a lot of money and introducing complex
machines and procedures to reduce contamination, do a Pareto
analysis. This is simply for identifying the major source(s) of
contaminants and eliminating them first. The 80/20 rule often
applies here – 20% of the effort will remove 80% of the
contaminants. In industrial processes, the machine tools are
probably responsible for most of the contamination, but in a
small lab the people and all the surfaces that touch wafers are
more likely to be the major source of contamination.
Since most substrate materials are dielectrics, they can easily
pick up static electricity charges of 20,000 V or higher. This highly
charged surface attracts oppositely charged particles. The only way
to effectively discharge a polymer or glass surface is to spray it with
an ionizing air or nitrogen gun. Spraying a surface to remove
particles with an air stream that is not ion controlled is likely to
charge the surface, actually increasing the number of particles on
surfaces making them airborne so that they land on wafers.
7.1. Counting Particles
There are now many machines made to count particles on surfaces
and in fluids. Airborne particle counters can be obtained for a few
somewhat more. Both machines are reliable and helpful in
monitoring the room itself and the fluids that touch wafers. If
you have access to an airborne particle counter, turn it on and
watch the counts as you move near it inside a clean room.
Particle counters for wafer surfaces are much more expensive
and complex, especially those designed for patterned surfaces.
Universities rarely have these tools. An inexpensive alternative is
excellent one is made by Spectroline and is sold as a ‘‘BlakRay,’’
although any very bright light in a darkened room will do to look
at wafers and other surfaces. It may expose photoresist.
7.2. Wafer Handling
In small research clean rooms, tweezers are commonly used to
handlesubstrates.Fingers,nomatterhow wellgloved, willputoils
fingers. Vacuum wands are the optimal wafer handling tool since
they only touch the back of the wafer, but are only effective in
production environments. Hundreds of varieties of tweezers are
available, so select stainless steel tweezers or tweezers with plastic
Microfabrication Techniques for Biologists41
jaws that are appropriate for the size of substrates you are using.
Clean tweezers frequently with acetone and alcohol because any
dirt on the jaws will be immediately transferred to your wafers.
Store tweezers in a clean container, and do not use tweezers for
levers,screwdrivers,and such,asnicksand damagetotweezers will
Before a product wafer is put into any machine for proces-
sing, one or several dummy wafers should be run in the process to
make sure it is running in a repeatable fashion. The previous user
affect your process. For example, if you are depositing a metal
film, the chamber could be coated with a metal that is incompa-
tible with your process. In a sputterer, material on the walls may
be sputtered away, depositing an unwanted contaminant in your
film. By coating the chamber with the metal film you desire
during a dummy run, your wafers will not be affected by the
previous user’s run. This is also true of plasma etch chambers
and CVD deposition systems. Always do a dummy run before
committing your product so that the chamber conditions are the
same. Plan on having several wafers scrapped while you develop
7.3. Clean Rooms
since few particles on wafers actually come from the air in a clean
room. Most particles come from process solutions, process
chambers, people, and dirty surfaces in contact with wafers.
Clean rooms are measured by their ‘‘class.’’ The older
classification is measured by the number of particles greater
than 0.5 mm per cubic foot of air. A class 100 room has, for
example, less than 100 0.5 mm particles per cubic foot of air in
a room that has no people in it and has had time to come to a
steady state. The modern classification is based on a metric
standard, so a class 100 clean room in the older system is
now a class 3.5 (3.5 particles greater than 0.5 mm per liter of
air). Some universities have class 10 clean rooms, but most are
class 1,000 or class 10,000. Clean rooms are always measured
with no people in them.
The idea of a clean room is that air flows in a laminar flow
regime, sweeping any particles out of the air and away from wafers.
People and objects in the laminar air flow stream cause turbulence,
which picks up particles which can then deposit them on nearby
Regardless of the classification of the clean room, the HEPA
(high efficiency particulate air) filters used to purify the air are
common to all types of clean rooms and clean benches. If an
airborne particle counter is used within a few cm of a HEPA
filter, frequently no particles will be measured. After the laminar
flow air passes equipment and people it picks up particles. Better
clean rooms are built by controlling the way the air flows out of Download full-text
the room. For example, class 100 rooms often have sidewall
returns, where class 10 or better rooms have air that returns
through holes in the floor. Obviously, the better the clean
room the higher the initial cost and the higher the maintenance
Housekeeping is critical for clean rooms. The floor should be
mopped with special mops and cleaners and should be vacuumed
regularly with a HEPA filtered vacuum cleaner. Remove clutter.
Position tables and machine tools so that air can flow around them
optimize cleanliness. Particles build up in dead air spaces and
where the air rolls due to turbulence. Regularly clean tabletops,
surfaces and doorknobs with clean room wipers and isopropyl
7.4. Human Behavior in
How people behave in a university type clean room may be the
most important factor in how clean the wafers are, regardless of
how much effort is put into engineering clean rooms and pro-
cesses. Simply by dressing in clean room garments does not
assure that people will not shed particles. In fact, making people
dress in special garments is primarily a barrier to people entering
a clean room. Even though gowned in a smock or jumpsuit
(bunny suit), head covering, face covering, gloves, and shoe
covers, you are still a source of particulate contamination. Hav-
ing less people in a clean room is the best way to reduce particle
counts, not by going to more exotic and expensive garments.
Garments should have static control fibers woven into the polye-
ster fabric, and the garments should be designed to prevent air
puffs coming out of the sleeves and neck openings. As people
move aboutin a clean room, the laminar airpassesaround people
and objects picking up particles, which can then be deposited on
surfaces. It is important that your head and hands never get
above clean wafers. Move slowly in clean rooms. Never store
wafers at floor level, even in closed boxes. Never put your feet up
on a chair or table.
clean rooms have no powder on them to assist in donning and are
frequently pre-washed. Always wash your hands before entering a
clean room, even if you wear gloves, since finger oils can soak
through most glove materials. Every surface your fingers touch
will gain particles and oily films, and these contaminants can be
moved around by diffusion (concentration gradients), by process
fluids, and simply by putting wafers in contact with a surface that
has been touched. For example, never place fingers inside a wafer
boat, rather always hold boats by the outside.
Microfabrication Techniques for Biologists43