Imagine a manufacturing technology capable of making
trillions of tiny machines — each the size of a bacteria.
Each machine could contain an onboard device programmed to control
a set of molecular scale tools and manipulators. An individual
machine could be designed to manufacture superior materials, convert
solar energy to electricity, or even, ultimately, enter the body
to fight disease and aging at the cellular and molecular level.
Materials hundreds of times better than today’s best materials,
vastly more powerful computers, precise machinery that doesn’t
wear out, and a revolution in clean manufacturing are but a few
of the predicted benefits of applying these new machines.
Nanotechnology is the enabling technology for
this vision. Nanotechnology is almost certain to spring upon
the world in the next twenty years and, with directed research
and development, may happen much sooner. The benefits of developing
nanotechnology are so great, and the cost to take the first step
so comparatively modest, that those people who understand the
issues are racing to be first to harness this technology. Twentieth-century
manufacturing has progressed from the blacksmith, to the production
line, to the wafer fab. Along the way, we have learned effective
ways of making small numbers of highly customized objects (model
shop), large numbers of simple 3D objects (mass production via
special purpose stamping, casting, and molding machines), and
massive numbers of complex 2D objects, such as photolithography.
This pathway has led to much lower costs and a level of complexity
unthinkable one hundred years ago. Humans are close to having
the technology to take the next step to true molecular nanotechnology — atomically
precise manufacturing using arrays of billions of molecular machines.
It is an understatement to say this technology will transform
the world more than the semiconductor revolution. It is just
as important to say that it won’t happen during the next
week — but rather in years.
Nobel Laureate Dr. Richard Smalley, testifying
before Congress during a hearing on nanotechnology1 on
June 22, 1999, said: “The impact of nanotechnology on health,
wealth, and lives of people will be at least the equivalent of
the combined influences of microelectronics, medical imaging,
computer-aided engineering, and man-made polymers developed in
Dr. Ralph Merkle, then with Xerox PARC, now with
Singularity University, also testified, stating: “Nanotechnology
will replace our entire manufacturing base with a new, radically
more precise, radicallyess expensive, and radically more flexible
way of making products.”
Dr. Eugene Wong, Assistant Director of the
Engineering Directorate at National Science Foundation, told
the committee: “Recent discoveries
at this scale are promising to revolutionize biology, electronics,
materials, and all their applications. We’re seeing inventions
and discoveries that were unimaginable only a short time ago.”
of this subcommittee, Congressman Nick Smith (R-Michigan) stated: “Nanotechnology
holds great promise for breakthroughs in health, manufacturing, agriculture,
energy use, and national security.”
It should be noted that many firms are now embracing
the word ‘nanotechnology’ in their corporate mission statements,
product literature, and stated capabilities, due to the new distinction
of nanotechnology as “the next thing” in the technical economy.
Our purpose in this paper is to put nanotechnology
in its proper perspective and seriously avoid much of the “über-hype” that
surrounds the field. As the first molecular nanotechnology company
in the United States, Zyvex feels a genuine responsibility to clarify
what’s required to tap into the significant promise of nanotechnology.
What is molecular nanotechnology?
The renowned physicist Dr. Richard Feynman
gave a talk in 1959 entitled “There’s Plenty of Room
at the Bottom,”2 describing
the possibility of constructing objects atom-by-atom. Feynman
concluded that we haven’t done this yet only because “our
hands, and our available tools are too large.” Eric Drexler
analyzed potential approaches and in his 1992 book, “Nanosystems,”3 he
concluded that there is no known physical reason molecular nanotechnology
cannot be mastered. Dr. Ralph Merkle has an extensive web site
Nature provides empirical proof of functioning
molecular machines in the form of life. Starting with the relatively
general “molecular machinery” common to all cells and
the specific program for an individual contained in that cell’s
DNA, one cell can build a complete organism of immense complexity.
Our challenge is to clearly understand and fabricate a broader variety of materials
and products than nature provides.
The primary technological goal of molecular nanotechnology
is to build one of the key enablers: the assembler.5 An
assembler is a system capable of manufacturing materials or complex structures
with atomic precision, positioning nearly every atom in the desired location.
Starting with a generic feedstock (such as methanol), this general purpose
machine would, in theory, manufacture any precisely defined object that could
be built from stable arrangements of the feedstock atoms. Potential fabricated
products would include most solid objects made today — from cars, to
chairs, to computers. At the very least, new materials could be made from
nanostructured ultra-strong, light, and low cost materials. It is quite likely
that these new products would contain embedded computers, actuators, and
power sources, so they could be programmed by their designers to have desired
The popular picture of a molecular assembler
is that of a small, robotically-controlled arm. However, a real
assembler must be built as a complete system, not just a robotic
arm, because the massive parallelism required for a practical
manufacturing system requires at least as much development as
the individual arm performing the assembly operation. Macroscopic
objects are made of trillions of atoms; therefore, building something
of that size requires billions of assembler components coordinating
a hugely parallel manufacturing process.
Although a semiconductor manufacturing plant manufactures
extremely complex 2 or 2 1/2-dimensional structures (with a severely
restricted set of chemical compounds), it lacks atomic precision.
On the other hand, bacteria are atomically precise, self-replicating “manufacturing
plants” that manufacture substances with atomic precision.
Genetic engineering can enable them to manufacture certain novel
compounds with atomic precision, but bacteria cannot make arbitrary
3D structures out of arbitrary materials.
The assembler differs from both semiconductor
manufacturing and bacteria because it operates on atomically-precise
molecular building blocks to build precise structures of arbitrary
complexity, as specified by a CAD/CAM program. Very simplistically,
an assembler could be a bank of molecular-scale robotic arms
with chemical binding sites on some arms and grippers to hold
components being built on another set of arms — all under
the control of an outboard computer instructing it how to snap
together the building blocks for a desired product. The assembler’s
control computer totally controls the product being built and
drives the manipulators to execute the sequence of motions specified
by the manufacturing software. In the morning, this assembler
might make computer memory modules; in the afternoon, it might
make medical manipulators; later in the evening, it might be
programmed to build power storage devices.
This first assembler will most likely be a crude
device — its purpose to demonstrate that molecular nanotechnology
is feasible and to help build a better device. This can start the
field on a Moore’s law type of improvement pathway.6 A
related concept is the semiconductor “learning curve” where
increasing volume allows manufacturing costs to decrease as the manufacturing
process matures. This occurs because process variables are more carefully
controlled and other economies of scale reduce the unit cost. To
reach the desired type of learning curve, we must make a product
of commercial value that cannot be economically made in another way.
For rapid progress, the manufacturing system
must be capable of being improved by the same products it manufactures.
A semiconductor manufacturing plant cannot manufacture itself;
hence, the cost of a semiconductor production line increases
with each generation, becoming more and more unaffordable. A
well-designed nanotechnology manufacturing plant should not suffer
from this problem, since it will be built using the same technology
and techniques it uses to manufacture other goods.
A practical design for an assembler requires that
the assembler be made out of materials it can handle. Assemblers
can be manufactured as inexpensively as the products they fabricate.
This capability is often called self-replication, although we prefer
the term “exponential assembly,” which is less likely
to be confused with living entities. Living systems carry their own
instructions in DNA, while exponential assembly systems do not. Exponential
assembly devices must receive instructions from a conventional computer
control system. This system design is both simpler and safer than
Many people use the term “nanotechnology” to
describe anything with characteristic dimensions at the nanometer
scale (one billionth of a meter). For clarity, we will frequently
use the phrase “molecular nanotechnology” (MNT) to
more carefully describe the goal of adaptable, affordable, and
molecularly precise manufacturing.
Molecular nanotechnology is defined as the use of a controlled sequence of
nanopositioning to perform mechanochemistry at exactly the reaction sites desired,
flexibly manufacturing atomically-precise products under software control.
Nanoparticle manufacturing is not MNT because it fails all aspects of the above
definition. In a broad sense, it is nanotechnology, because it deals with nanometer-sized
objects, but in that broad a sense, so is a chemical manufacturing plant, petroleum
refinery, or drug manufacturing facility.
Biochemistry is arguably MNT, but it has limits
in the product being made. One could argue that DNA can specify
a virtually infinite number of proteins, but the materials properties
of the resulting product cannot be tailored sufficiently to achieve
the desired goals of MNT. Creating a good conductor or insulator
capable of working in a vacuum at 200 degrees Celsius is easy
for true MNT, but extremely unlikely with biochemistry.
Self-assembly is frequently postulated as
a more likely way to achieve molecular nanotechnology than Drexler’s
mechanical constructor arms. We believe self-assembly is likely
to be of value with small building blocks, but is insufficient
for building larger assemblies. Self-assembled materials almost
always have grain defects, which cause weaknesses and unpredictable
With templating to guide grain boundary growth,
self-assembly might be able to create materials having desirable
properties, but self-assembly is unlikely to create truly complex
structures. Note that life processes are not pure self-assembly;
there is a huge amount of templating going on, as well as uncountable
numbers of existing cellular molecular machines actively following
their own instructions. To appreciate the difference between
pure self-assembly and directed molecular manufacturing, consider
the likelihood of a dog spontaneously self-assembling from raw
Nanoelectronics is probably the best funded
nanoscale research at this time. While this is a good application
area for an assembler, nanoelectronics cannot flexibly manufacture
products. Indeed, it appears likely that even more expensive
semiconductor fabrication plants will be necessary for nanoelectronics
than what we have today. Molecular electronics promises to break
this price spiral if it succeeds, and therefore seems a promising
avenue for achieving nanoelectronics. However, if we had an assembler
system, assembling arbitrarily complex nanoelectronics (based
on today’s devices) would be a straightforward extension
of the basic chemistry used in the first assembler. We believe
that developing a mechanical assembly system is at least as important
as trying to build nanoelectronics today.
The mechanical approach to molecular nanotechnology
can be pursued either from the top-down or the bottom-up. Dr.
Richard Feynman suggested a top-down approach (i.e., building
successively smaller generations of machines until we get to
a level at which there is no room for inaccuracy, with every
atom precisely placed and accounted for). Dr. Eric Drexler proposes
a bottom-up approach (i.e., using a tool such as an atomic force
microscope) to precisely place molecular building blocks to build
larger structures with precision. Feynman’s speculations
predate the invention of the AFM by decades. Drexler made his
original conjecture several years before the invention of the
atomic force microscope.
Both approaches are likely to contribute valuable insights. Therefore, it is
important to work on top-down projects that will have more near-term payoffs,
and bottom-up projects that are essential for the full fruition of molecular
One promising top-down approach involves using
microelectromechanical systems (MEMS) components to create a
micron scale analog of the assembler.
MEMS components are made via the semiconductor
processing technology used for integrated circuits (IC), and
can take advantage of the large investment made in IC fabrication
equipment and techniques.
MEMS offers several advantages over alternative,
more conventional-sized components or pure simulation:
• MEMS components are light enough that
gravity and inertia are relatively unimportant, so surface forces
dominate, just as they do at the nanometer scale.
• MEMS components can be made inexpensively — a typical fabrication
run delivers structured parts costing pennies each7 — compared
to dollars or hundreds of dollars each for similar prototype parts at the centimeter
(or larger) scale.
• Many of the mechanisms useful in MNT also work in MEMS, such as electrostatic
actuators, snap connectors, flex joints, and manipulator geometries. Conversely,
macroscopic parts such as magnetic stepper motors or solenoids are not effective
below the micron scale, so many changes are necessary for such a design as it
• Objects must be assembled at the micron and larger
scale even with MNT, so system designs for parts assembly are required at this
scale. MEMS offers an excellent test bed for quickly prototyping the system
• The components are visible under a microscope, so designs can be easily
debugged by watching them operate.
• As the MEMS market develops, other companies will need to assemble systems
at the micron to millimeter scale, creating an outside market for MEMS assembly
devices. Today’s MEMS designers try to integrate entire systems on a monolithic
chip, but that strategy is limiting if assembly is to be inexpensive and reliable.
• Complex systems can be designed and built more easily.
By integrating common subcomponents and assembling 3D structures, one can postulate
a cubic centimeter device with millions of moving parts,8 perhaps
performing sensing functions, doing chemical separations or reactions, or even
functioning as a catheter-guided surgical device. Building MEMS-based artificial
ears, eyes, or nerve stimulators that could partially restore lost functions
are exciting longer-term applications under development by others.
Of course, MEMS have some disadvantages when
compared to molecular nanotechnology:
• Parts are not atomically precise, so
bearing surfaces exhibit increased wear and system design is
limited to poorer tolerances.
• The inflexibility of having to pre-manufacture parts means it is slower
and more difficult to build prototypes.
• Chemistry is severely restricted, mostly to materials
compatible with a semiconductor wafer fab line. Those materials are excellent
by our current standards but poor compared to true MNT materials.
• Part attachment is not done by forming covalent bonds, so all forms of
interconnect need to be designed carefully.
• While inexpensive, the parts are still higher-priced when compared with
This real-world, mechanistic approach to the development
of molecular nanotechnology requires technological advances in
three areas: nanomanipulation, mechanochemistry, and system design.
All three areas must progress to deliver MNT. Mechanochemistry
is the most fundamental technology needed, but probably the easiest
to scale into production, given the other two capabilities. Nanopositioning
is vital to hold reactants precisely at the right location so that
a chemical bond can be formed at a desired reaction site and not
a chemically identical site nearby. System design is the key to
successfully implementing economically-viable MNT. A single nanopositioner,
performing a single mechanochemistry operation per instruction,
will take “geological time” to build a visible object.
It is a fundamental requirement that we have immense numbers of
manipulators and reactions operating simultaneously to achieve
The fundamental operation that differentiates
molecular nanotechnology from other precision manufacturing is
mechanochemistry. With mechanochemistry, one can literally fabricate
an object one molecule or atom at a time, placing new building
blocks precisely where desired. “Machine-phase chemistry” is
another term often used to describe this activity. Rather than
performing chemical reactions in a solution exposed to heat and
pressure using mechanochemistry, one might hold the two reacting
molecules in precise orientations relative to one another and
push them together, forcing them over their reaction barrier
by mechanical energy. The chemistry of this is the equivalent
to a solution heated to a temperature promoting that reaction;
the difference is that, in solution, the molecules come together
randomly, joining in a random manner. With mechanochemistry,
the same chemistry is occurring, but due to the positional accuracy
and ability to exert great forces, one can provide reaction conditions
comparable to extreme temperatures and pressures, and build objects
with exact structure. Mechanochemistry can also be performed
in an ultra-high vacuum (UHV) where extremely reactive molecules
or atoms with unterminated bonds can be brought together precisely
to allow a reaction with no reaction barriers to occur at a desired
Research is necessary to develop suitable molecules
to act as building blocks. The 2-4 Diels-Alder reaction has been
proposed as a model (i.e., pushing together a diene and a dienophile
with precise positional control joins the carbon atoms into a six-member
ring, building a new composite molecule in the process). A useful
building block must have enough “linker groups” to be
joined to others at multiple sites, so the resulting “molecule” is
suitably cross-linked internally. A molecule with tetrahedral or
octahedral symmetry would be a good choice for this fundamental shape.
Depending on the molecule chosen, mechanochemistry
can be performed in fluids (the way biochemical processes work),
inert gases, clean air, or in a UHV. When using individual atoms
as building blocks, part of the preparation process may result
in unterminated bonds, which are highly reactive. This need not
be a deterent: by controlling the environment where the chemistry
occurs, one can ensure that there are no other reactive atoms around
to react with the “dangling bond” before attaching
it to the desired spot.
System design deals with controlling immense
numbers of nanomanipulators performing mechanochemistry. A successful
system controls and provides power to the manipulators, delivers
raw materials to the manipulators, and moves finished subassemblies
to the next stage. With a good modular system design, advances
in nanomanipulation and mechanochemistry can be rapidly incorporated
into the entire manufacturing process.
One promising technique is what Dr. Ralph Merkle describes as “convergent
assembly.”9 The entire assembler system is
composed of a series of stages at different scales. The largest, outermost
system may have an output assembly compartment of 10 cm on a side. The back
side of this compartment may be fed from four smaller boxes, each 2.5 cm on
a side. Each of those, in turn, is fed from four, smaller assembly compartments.
A series of such stages ultimately leads to a very large number of molecular
scale manipulators (number of smallest assembly stations = 4x4x4…). Each
of those manipulators might perform mechanochemistry to make a small component,
then pass that component up the hierarchy to the larger stages, assembling
components from those subcomponents. With convergent assembly, thirty stages
can span nine orders of magnitude —from the nanometer to the meter size
Assemblers may do their mechanochemistry work in fluids,
inert gases, a UHV, or some combination of these. A design might use fluids
for material transport, passing materials through a manufactured membrane
under an UHV where reactive chemistry is performed. Such a design would mimic
nature’s own design for living organisms (e.g., passing materials through
a lipid bilayer into or out of a controlled environment).
In addition to the geometry and engineering
of the physical assembler, one must have software tools to design
parts made with that assembler. Postulating a convergent assembly
approach, one can envision two major components of such a computer-aided
design (CAD) system: parts design and assembly sequencing.
Individual parts would be made in large numbers
deep inside the molecular scale devices of the assembler system.
The parts design software would need to assist the designer in performing
molecular design work for that part, characterizing its properties,
and generating assembly sequence commands to drive the nanomanipulator
performing its mechanochemistry. Designing such parts would require
knowledge of chemistry and engineering expertise. CAD software for
this would encompass molecular design and simulation, mechanical
3D part design, and knowledge of the mechanochemistry being performed.
Since computational chemistry requires computing resources that scale
exponentially with the number of atoms, one would like to minimize
the number of atoms in a part needing detailed chemical simulation.
Once a basic library of usable parts was developed,
most designers could simply use that library to design larger assemblies.
In an ideal case, there would be enough existing parts so the system
designer would not have to make custom parts at all. CAD software
would assist the designer in combining parts from a library to
create the desired object. A graphics display designer might choose
display modules from a catalog, build power and data distribution
busses from snap-together pieces, design a custom package using
existing outer finish components, and integrate communications
modules to complete the design. A designer of drug delivery devices
might integrate sensors, drug storage tanks, injectors, power supplies,
and logic. Each of those integrated components might, in turn,
be built from other subassemblies. The required CAD software should
be able to use the characterization data for the parts at a high
level to avoid the need for chemistry simulation of such a large
system. Simulation would be at the system level, predicting performance,
and verifying behavior.
A prototype of such an assembler system could be
built today, using MEMS components, rather than molecularly precise
components, such as low-level building blocks.
Furthermore, we expect to find commercial use in
using such systems to create MEMS scale systems that cannot be
assembled today. The inability to assemble MEMS-scale parts means
those parts don’t get designed or they get designed as an
integrated whole. This means that a large portion of the surface
area is taken up with actuators that may get used once during the
initial configuration. For example, some have designed stand-up
mirrors (erected by actuators on the chip) that take up a considerable
amount of the chip area simply to erect the mirror one time.
Given a very basic MNT capability, the first thing
a company would likely make would be a better MNT capability. Such
a process should quickly lead to a commercially viable nanomanufacturing
capability, which could be sold or licensed to any company engaged
Clearly, MNT is a technology that will dramatically
change most industries. Companies successfully making products
today have product know-how, customer relationships, and networks
of support that we cannot, and should not, try to take away.
For example, one of Zyvex’s goals is to provide the MNT
assembler, partnering with our customers (as appropriate) to
share our expertise, so our customers can change their own businesses
with this technology. While nimble companies will adapt and new
ones will form — slow-paced companies will go out of business.
This process is described by economist Joseph Schumpeter as “creative
destruction:” the dynamic renewal process that keeps free
markets healthy and active. MNT applications extends across a
broad spectrum of industries.
We expect business opportunities to follow
the development of MNT in five major phases.
Business opportunities include selling nanomanipulation
devices, building small simple structures (fewer than one hundred
atoms), and developing the required CAD/CAM software. Systems
may be manufactured conventionally in a machine shop or MEMS
This phase requires progress in nanomanipulation, mechanochemistry, control
software, and component design.
With the Phase One capability working, one should
be able to fabricate nanostructured and nanocomposite materials
in volume. Nanocomposite materials are attractive because they
can have performance exceeding any of the individual component
materials. For example, the abalone shell, a composite of calcium
carbonate plates sandwiched between organic material, is 3,000
times more fracture resistant than a single crystal of the pure
Phase Two requires progress in system design (to
make a cost-effective nanomanufacturing plant capable of manufacturing
bulk quantities of materials) and materials design (to make interesting
materials with the chemistries available at this stage).
Limiting oneself to simple materials means one can
quickly build a large variety of materials with relatively little
design time. It is much easier to specify the unit cell of a repeating
material than to specify a complex object, such as a computer or
Once Phase One capabilities are demonstrated, complex
designs can reasonably be initiated. It will take some time for
most designers to become aware of the possibilities, commit to
learning the new capabilities, and develop new design paradigms.
Designing new approaches to sensing, actuation, and control will
require more time to specify than the approaches to the simple
materials of Phase Two. Phase Three requires advances in CAD/CAM
software, object design, simulation, testing, and packaging.
By making products such as bio-implants for restoring lost hearing or vision,
the potential for MNT will rapidly become apparent. Thousands of designers
will invest in learning the new skills and macroscopic-sized objects (built
with atomic precision) will show up in the marketplace.
We envision two major paths to the development of
Path One occurs if mechanochemistry develops rapidly
and the assembler becomes capable of building structures from all
of the atoms used in current silicon electronics (such as silicon,
germanium, nitrogen, oxygen, phosphorous, boron, arsenic, copper,
and aluminum). In this optimistic scenario, existing semiconductor
computer designs can simply be manufactured atom-by-atom by banks
of inexpensive assemblers. Nanocomputers might then precede the
Phase Three objects mentioned above, due to the huge economic incentives
of inexpensive semiconductor manufacturing.
Path Two occurs if mechanochemistry cannot yet build
with the variety of atoms needed; therefore, electronic computers
can’t be built atom-by-atom. One could still build computers,
but they would need to be designed more along the lines of Drexler’s
mechanical computers (i.e., physically moving small molecules around).
Due to the thousands of years of intellectual capital invested
in computer designs and the requirement to develop new design rules
and techniques, it is unlikely that any company would invest in
designing a mechanical nanocomputer before Phase Three is demonstrated.
To be competitive with electronic computers, a nanocomputer
company would have to invest considerably more effort than its
electronic computer competitors. The need for such an investment
would slow the advent of the nanocomputer.
Therefore, to obtain a nanocomputer, one must either improve the mechanochemistry
required to handle all of the needed semiconductor materials, or spend a huge
amount of time designing computers with a brand new principle of operation.
The step that excites nanotechnology enthusiasts
the most is that of making autonomous devices which carry their
own onboard computers, power sources, actuators, and sensors. These
tiny “robots” might travel the body fighting disease,
or patrol our food crops recognizing and destroying pests. They
might act as road surfaces, automatically repairing damage, while
joining together the nation’s highway system in a vast grid
of solar power generators (each device could be a solar cell and
a road-building robot).
While expected to lead to a huge market, this technology is extremely advanced
and, as a prerequisite, requires the first four phases. The design effort to
build such autonomous devices is unprecedented and, like the nanocomputer,
should not be seriously attempted until feasibility of MNT assemblers is proven.
The power source for such a device is currently a serious technological hurdle,
but the biggest unknown is how one would program it. Our current techniques
for programming large systems are limited. Rough estimates of the complexity
of such a device, programmed with today’s techniques, are in the tens
of millions of lines of code. Better programming methods are required to achieve
Phase Five. Autonomous learning systems are promising, but testing the resulting
device is problematic, and deterministic performance is unlikely.
Molecular nanotechnology is predicted to be the most
powerful technology yet developed by humankind. It will lead to
major changes in our civilization. However, three key breakthroughs
are necessary: nanomanipulation, mechanochemistry, and system design.
There is an immense amount of work to be done, but the number of
people working in this field grows larger every day. With this
increased effort, there is no question that we will see major steps
toward the goal of creating molecular nanotechnology in the next
1. Prepared Statements for
House Science Committee, Subcommittee on Basic Research, June
22, 1999, http://www.house.gov/science/106_hearing.htm
2. Feynman, Richard: “There’s Plenty
of Room at the Bottom,” 1959 speech, http://www.zyvex.com/nanotech/
3. Drexler, K. Eric: “Nanosystems: molecular
machinery, manufacturing, and computation,” Wiley Inter-science,
1992, http://www.zyvex. com/nanotech/nanosystems.html.
4. Merkle, Ralph: Nanotechnology web site: http://www.zyvex.com/nano/index.html.
5. Assembler description from Zyvex: http://www.zyvex.com/CorpInfo/New
6. Moore’s “law” is a thirty year
old observation by Gordon Moore of Intel that the number of transistors
in an integrated circuit doubles every eighteen months as the manufacturing
process improves in accuracy.
7. An advanced MEMS prototyping process run costs
approximately $40,000–$90,000 depending on number of process
steps and wafers run. Typical runs would cost $70,000 for two wafers,
$90,000 for ten. Each four inch wafer may hold 500,000 parts of dimension
10 microns by 100 microns, resulting in a per-part cost ranging from
$0.07/part for a small run, to $0.018/part for a large run. In a
production setting, larger wafers would be used, and the silicon
mold and masks would be reused, driving the per-wafer cost to below
$1,000, producing 1,000,000 parts per wafer, for a per-part cost
of $0.001. Regular arrays of transistors, packaged and delivered,
currently cost less than $0.25 per million transistors.
8. Part counts, even with multimicron sized parts,
becomes quite high if three dimensions are used. An early MEMS system
might have sub-components that occupy an average volume of 50x50x200
microns. In a 3D cube, this would result in 2,000,000 components
per cubic centimeter, where a component might be an actuator bank,
electronic switch, or simple mechanical structure. More advanced
MEMS parts might be 2x4x50 microns, leading to 1,250,000,000 parts
per cubic cm.
9. Merkle, Ralph: “Convergent Assembly,” http://www.zyvex.
10. Smith, Bettye, et al: “Molecular Mechanistic
Origin of the Toughness of Natural Adhesives, Fibres and Composites,” Nature
399, 761–763 (1999).
Zyvex Labs, LLC.