computing is the defining feature of our
modern life and indeed the bedrock of
scientific discovery in almost every
field as layman users of computing
technology we’ve grown to be dependent
on computation to control almost every
aspect of our life from the running of
our trains to our airports to our
políticas of social media platforms
such as Facebook and Twitter however
there are only a few of us that has ever
wondered what makes this monumental
progress in technology possible it would
not be an exaggeration to say that
Moore’s law in the last twenty to thirty
years has been the most fundamental
innovation that has affected us of
course there are many innovations in the
20th century that has had profound
effects on our lives but Moore’s law
more than any deserves to be put on a
pedestal for has deeply and personally
touched every one of our lives the
realization that computing processing
power doubles roughly every two years is
what has enabled us to go from roughly
2,000 transistors on a single
microprocessor in the 1970s to now more
than 1 billion a processor that you can
purchase at any place right now so back
to that question what makes this
progress there’s seemingly unstoppable
progress in technology possible well I’m
here to tell you today that it is due to
a technology we called lithography our
ability to define at ever smaller
degrees and to pattern the dimensions of
the micro electronic circuitry that
makes every processor possible so let me
talk to you about evolution of
lithography just a brief history of it
we started roughly in the 1970s in
systems we call contact lithography
we’re inverting
higher system was in contact with the
substrate or a silicon that we were
trying to pattern today we’re using a
system called otherwise known as step
and scan where every field that you wish
to expose every microprocessor every
transistor that is design is exposed and
then repeated saw and so forth
we started at wavelengths in the order
of 460 nanometer in the 1970s in
lithographer x’ terms it was called G
line lithography today we’re at deep UV
or 193 nanometers on the right side you
can get a glimpse of these tools that
manufacturers such as Intel or Samsung
or many others use these are very heavy
equipments the lenses are designed very
precisely kept in a pristine environment
in a micro fabrication facility or
otherwise known as fab each tool costs
something like ten or eighty two hundred
million dollars and a particular
manufacturer might have ten of these
tools or twenty of these tools at their
disposal so they can make they can
realize the potential of manufacturing
in a high scale high-volume
manufacturing mode to make this a real
business you if you can look closely you
see that operators are gound or dawn in
this type of a protective clothing not
to protect them from the product but to
protect the product from the impurities
produced by by them so I’m just trying
to get give you an idea of how difficult
this business is and how expensive it is
getting to be able to make the progress
that we’re seeing today so what is
lithography you may be wondering at this
point optical lithography in very simple
terms is the process of patterning
patterning on a substrate in this trap
in this case silicon and to transfer
those geometric patterns onto wafers or
silicon through predetermined design the
design is on a photo mask or a reticle
and then this is all chance transfer to
a resist a light-sensitive material that
interacts with light and then we can
basically make devices so what I just
mentioned is realized in the picture you
see on the right you end up with lines
and spaces essentially the electronic
circuitry on a nanometer scale and you
have billions of these lines and spaces
enabling you to make a transistor let me
give you a basic processing sequence of
course there are many sequences in them
semiconductor manufacturing that are
critical but lithography is by far the
most difficult because it pushes the
limit of technology you start with the
soft straight silicon basically dirt
from Earth this whole industry is
created from dearth so that is what is
so amazing about it it’s purified into
crystalline form it is devoid of any
oxygen it is then coated with this type
of light sort of sensitive material
called resist that light travels to your
design your mask is exposed areas where
the resist interacts with lights are
stripped off in a developer chemical and
then the then you go through an edging
process and then finally you strip off
the resist again you end up with an
image like that on their electron
microscope so how do we do this this
imaging basics of this the equations
that describe this in very simple terms
are the angular angle of light refracted
light from from your mask related to the
wavelength of light and your slit
opening your mask a lithography a job of
a lithographer is to utilize multiple
orders of diffracted light to capture as
much information as possible the more
information you have the better your
resolution will be and the faster your
computer will be at the end so at the
heart of it all is Riley’s equation sort
of related to the first equation but
more condensed ly describing the
resolution or how precise your image is
and finally again how fast your device
will work to the wavelength of light and
the numerical aperture the numerical
aperture is the size of your lens you
can think about it in simple
the vagueness of light is simple it’s
the UV light that we use in 193 UV light
so you really have three choices three
fundamental choices in order to make
progress you can increase the numerical
aperture make bigger lenses more
expensive you can reduce the wavelength
of light or reduce this K factor this
fudge factor that you see a lot used in
physics or other you know electrical
engineering so so on if these factors
that have to do with the environment the
tool the lens everything you want to be
able to reduce it to improve the
resolution so this is the name of the
game so let me talk about the first mode
of progress increasing numerical
aperture numerical aperture is a maximum
angle of deflected light by the lens
when light goes to a slit its diffracted
because it’s a wave and then you have
these modes of diffraction these dark
spots and light spots if you all
captured only the ends n equal one order
of information your resolution is not
going to be good so again the the key
idea here is to design lenses that are
able to capture as much possible as many
diffracted orders of the diffracted
light as possible the spatial frequency
components of light so in recent years
at Intel and some other manufacturers we
introduced the concept of immersion
essentially putting ultra pure water
between your lens and your substrate the
silicon Intel was one of the first
manufacturers that was able to produce
this in 1268 technology 32 nanometers
very effectively at very high yields
thus being roughly 2 to 3 years ahead of
any other manufacturers such as Samsung
another leading manufacturer we use
water because it has a higher index of
refraction and the index of refraction
causes your reamer causes higher
numerical aperture as compared to air
which has the index of refraction of 1
so essentially ending up with a bigger
lens but your lens is not really
physically bigger it appears to be that
way there is a large there’s a trade-off
when you increase your numerical
aperture as everything in life nothing
is free so you can think about this in
simple terms simple terms like
photography when you zoom in on an image
you can capture a lot of information
about that area that you are trying to
take a picture of but you lose the
ability to resolve images at larger
scale so this is one of the challenges
of just relying on numerical aperture to
improve your resolution so we do other
things enhance k-factor that I talked
about this fudge factor you do this by
basically source engineering changing
your mask from a conventional circular
mode to these other modes quadruple
angular playing tricks with the optics
the most effective way the third way
that you can basically make faster
computers is is reducing the wavelength
of light we’re now working in a regime
where the wavelength of light is 13 and
1/2 nanometers to be introduced in the
next few years by Intel in many
computers that you will be able to
purchase and so this is getting very
difficult because the wavelength of
light is reducing so how do we do that
right now intel is working on this
technology called deep ultraviolet or
EUV extreme ultraviolet these are
sources that are plasma generated
sources so there is a lot of energy
highly energetic gases involved the
sources are the photons are energetic
when they hit your resist resist Hough
has to be able to withstand the flux or
energy of these photons so designing the
new types of resists is a challenge for
manufacturers your optics have to be
very precise your lenses or your masks
have to be very defect-free
your masks have to be películas there’s
nothing kind nothing that can touches
touch the mask so this is becoming very
difficult and challenging to do so today
I’m here to propose another way and this
is sort of hedging your bets because
we’ve spent so much money and resources
on EU
and it’s becoming more difficult to sort
of push the boundary of Moore’s law make
that happen for the next generation we
ought to explore other technologies one
such example is nano lithography or nano
imprint lithography where you’re
essentially designing these masks that
are on a nano scale and they’re a mold
there are a nano scale move you push
down this mold into a resist you go to a
series of steps such as soft lithography
thermal imprint little at room
temperature here you have the advantage
of not worrying about scattering you
don’t have to have interference you
don’t have the fraction because you’re
essentially making contact these molds
are making contact with the resist and
this is all done at room temperature so
I’m proposing that manufacturers pursue
this technology of course there’s many
challenges with this technology as well
but there is nothing that we cannot
overcome it’s just a matter of money and
sinking more resources into this to make
this truly happen so let me conclude by
saying a few words you’ve already got a
grasp of what Moore’s law means so as we
reduce feature sizes as we play with
optics as we reduce the original
wavelength of light to be able to make
devices faster we’re running into very
fundamental limits and these limits are
determined by the laws of physics
themselves they’re of course more
earthly laws and those laws are the laws
of diminishing returns
more expensive files may be files that
are roughly about ten billion dollars
every three years more expensive
equipment massive tools right more
expensive more cost you know more labors
in terms of in terms of research and
development and sustaining us so at this
point you might be asking yourself is
Moore’s law dead well I really believe
Moore’s law is not dead in fact we
barely begun to scratch the surface in
terms of the number of innovations we
will be able to bring about into the
marketplace in the next twenty to thirty
years
in the next twenty to thirty year time
horizon we will bear witness to some
more astounding number of innovations
made possible by artificial intelligence
quantum computing and optoelectronics
this hat has led many to speculate and
use this very forum the TED forum to
argue that we are on a verge of
singularity the singularity is
essentially a summation of our
scientific knowledge doubling roughly
every hour every day by 2050 rather than
every two years this is a bold idea but
if we believe we’re on an exponential
exponential curve of progress this will
indeed happen and in the 20th century
we’ve seen that we are exponentially
growing our technology so on a more
philosophical footing what does it mean
when you say Moore’s law is dead what
does it mean when you ask that when
you’re asking that what you’re really
questioning or asking is human ingenuity
debt is human curiosity that what I
firmly believe as Homan Homo sapiens
we’ve never lacked ingenuity or
curiosity ever since we’d set sail from
the northern plains of Africa and
expanded our species to other continents
we’ve always been explorers of our world
of our universe the world of the very
large today we’re exploring the world of
the very small the nano scale the
molecular scale the atomic layer finally
what does this mean for our society what
does it mean for us well can be
computing is pervasive and it’s here to
stay there is no going back
societies that are able to push the
emblem of progress are able to embrace
progress are able to not only even
embrace it but contributed contribute to
it or will be the societies that will be
successful our nation’s we have a duty
to contribute to a bar
of scientific knowledge and if we do
that if we expand our scientific
knowledge we will be able to not only
control the marketplace of ideas but
also control the marketplace itself
thank you very much
[Applause]
modern life and indeed the bedrock of
scientific discovery in almost every
field as layman users of computing
technology we’ve grown to be dependent
on computation to control almost every
aspect of our life from the running of
our trains to our airports to our
políticas of social media platforms
such as Facebook and Twitter however
there are only a few of us that has ever
wondered what makes this monumental
progress in technology possible it would
not be an exaggeration to say that
Moore’s law in the last twenty to thirty
years has been the most fundamental
innovation that has affected us of
course there are many innovations in the
20th century that has had profound
effects on our lives but Moore’s law
more than any deserves to be put on a
pedestal for has deeply and personally
touched every one of our lives the
realization that computing processing
power doubles roughly every two years is
what has enabled us to go from roughly
2,000 transistors on a single
microprocessor in the 1970s to now more
than 1 billion a processor that you can
purchase at any place right now so back
to that question what makes this
progress there’s seemingly unstoppable
progress in technology possible well I’m
here to tell you today that it is due to
a technology we called lithography our
ability to define at ever smaller
degrees and to pattern the dimensions of
the micro electronic circuitry that
makes every processor possible so let me
talk to you about evolution of
lithography just a brief history of it
we started roughly in the 1970s in
systems we call contact lithography
we’re inverting
higher system was in contact with the
substrate or a silicon that we were
trying to pattern today we’re using a
system called otherwise known as step
and scan where every field that you wish
to expose every microprocessor every
transistor that is design is exposed and
then repeated saw and so forth
we started at wavelengths in the order
of 460 nanometer in the 1970s in
lithographer x’ terms it was called G
line lithography today we’re at deep UV
or 193 nanometers on the right side you
can get a glimpse of these tools that
manufacturers such as Intel or Samsung
or many others use these are very heavy
equipments the lenses are designed very
precisely kept in a pristine environment
in a micro fabrication facility or
otherwise known as fab each tool costs
something like ten or eighty two hundred
million dollars and a particular
manufacturer might have ten of these
tools or twenty of these tools at their
disposal so they can make they can
realize the potential of manufacturing
in a high scale high-volume
manufacturing mode to make this a real
business you if you can look closely you
see that operators are gound or dawn in
this type of a protective clothing not
to protect them from the product but to
protect the product from the impurities
produced by by them so I’m just trying
to get give you an idea of how difficult
this business is and how expensive it is
getting to be able to make the progress
that we’re seeing today so what is
lithography you may be wondering at this
point optical lithography in very simple
terms is the process of patterning
patterning on a substrate in this trap
in this case silicon and to transfer
those geometric patterns onto wafers or
silicon through predetermined design the
design is on a photo mask or a reticle
and then this is all chance transfer to
a resist a light-sensitive material that
interacts with light and then we can
basically make devices so what I just
mentioned is realized in the picture you
see on the right you end up with lines
and spaces essentially the electronic
circuitry on a nanometer scale and you
have billions of these lines and spaces
enabling you to make a transistor let me
give you a basic processing sequence of
course there are many sequences in them
semiconductor manufacturing that are
critical but lithography is by far the
most difficult because it pushes the
limit of technology you start with the
soft straight silicon basically dirt
from Earth this whole industry is
created from dearth so that is what is
so amazing about it it’s purified into
crystalline form it is devoid of any
oxygen it is then coated with this type
of light sort of sensitive material
called resist that light travels to your
design your mask is exposed areas where
the resist interacts with lights are
stripped off in a developer chemical and
then the then you go through an edging
process and then finally you strip off
the resist again you end up with an
image like that on their electron
microscope so how do we do this this
imaging basics of this the equations
that describe this in very simple terms
are the angular angle of light refracted
light from from your mask related to the
wavelength of light and your slit
opening your mask a lithography a job of
a lithographer is to utilize multiple
orders of diffracted light to capture as
much information as possible the more
information you have the better your
resolution will be and the faster your
computer will be at the end so at the
heart of it all is Riley’s equation sort
of related to the first equation but
more condensed ly describing the
resolution or how precise your image is
and finally again how fast your device
will work to the wavelength of light and
the numerical aperture the numerical
aperture is the size of your lens you
can think about it in simple
the vagueness of light is simple it’s
the UV light that we use in 193 UV light
so you really have three choices three
fundamental choices in order to make
progress you can increase the numerical
aperture make bigger lenses more
expensive you can reduce the wavelength
of light or reduce this K factor this
fudge factor that you see a lot used in
physics or other you know electrical
engineering so so on if these factors
that have to do with the environment the
tool the lens everything you want to be
able to reduce it to improve the
resolution so this is the name of the
game so let me talk about the first mode
of progress increasing numerical
aperture numerical aperture is a maximum
angle of deflected light by the lens
when light goes to a slit its diffracted
because it’s a wave and then you have
these modes of diffraction these dark
spots and light spots if you all
captured only the ends n equal one order
of information your resolution is not
going to be good so again the the key
idea here is to design lenses that are
able to capture as much possible as many
diffracted orders of the diffracted
light as possible the spatial frequency
components of light so in recent years
at Intel and some other manufacturers we
introduced the concept of immersion
essentially putting ultra pure water
between your lens and your substrate the
silicon Intel was one of the first
manufacturers that was able to produce
this in 1268 technology 32 nanometers
very effectively at very high yields
thus being roughly 2 to 3 years ahead of
any other manufacturers such as Samsung
another leading manufacturer we use
water because it has a higher index of
refraction and the index of refraction
causes your reamer causes higher
numerical aperture as compared to air
which has the index of refraction of 1
so essentially ending up with a bigger
lens but your lens is not really
physically bigger it appears to be that
way there is a large there’s a trade-off
when you increase your numerical
aperture as everything in life nothing
is free so you can think about this in
simple terms simple terms like
photography when you zoom in on an image
you can capture a lot of information
about that area that you are trying to
take a picture of but you lose the
ability to resolve images at larger
scale so this is one of the challenges
of just relying on numerical aperture to
improve your resolution so we do other
things enhance k-factor that I talked
about this fudge factor you do this by
basically source engineering changing
your mask from a conventional circular
mode to these other modes quadruple
angular playing tricks with the optics
the most effective way the third way
that you can basically make faster
computers is is reducing the wavelength
of light we’re now working in a regime
where the wavelength of light is 13 and
1/2 nanometers to be introduced in the
next few years by Intel in many
computers that you will be able to
purchase and so this is getting very
difficult because the wavelength of
light is reducing so how do we do that
right now intel is working on this
technology called deep ultraviolet or
EUV extreme ultraviolet these are
sources that are plasma generated
sources so there is a lot of energy
highly energetic gases involved the
sources are the photons are energetic
when they hit your resist resist Hough
has to be able to withstand the flux or
energy of these photons so designing the
new types of resists is a challenge for
manufacturers your optics have to be
very precise your lenses or your masks
have to be very defect-free
your masks have to be películas there’s
nothing kind nothing that can touches
touch the mask so this is becoming very
difficult and challenging to do so today
I’m here to propose another way and this
is sort of hedging your bets because
we’ve spent so much money and resources
on EU
and it’s becoming more difficult to sort
of push the boundary of Moore’s law make
that happen for the next generation we
ought to explore other technologies one
such example is nano lithography or nano
imprint lithography where you’re
essentially designing these masks that
are on a nano scale and they’re a mold
there are a nano scale move you push
down this mold into a resist you go to a
series of steps such as soft lithography
thermal imprint little at room
temperature here you have the advantage
of not worrying about scattering you
don’t have to have interference you
don’t have the fraction because you’re
essentially making contact these molds
are making contact with the resist and
this is all done at room temperature so
I’m proposing that manufacturers pursue
this technology of course there’s many
challenges with this technology as well
but there is nothing that we cannot
overcome it’s just a matter of money and
sinking more resources into this to make
this truly happen so let me conclude by
saying a few words you’ve already got a
grasp of what Moore’s law means so as we
reduce feature sizes as we play with
optics as we reduce the original
wavelength of light to be able to make
devices faster we’re running into very
fundamental limits and these limits are
determined by the laws of physics
themselves they’re of course more
earthly laws and those laws are the laws
of diminishing returns
more expensive files may be files that
are roughly about ten billion dollars
every three years more expensive
equipment massive tools right more
expensive more cost you know more labors
in terms of in terms of research and
development and sustaining us so at this
point you might be asking yourself is
Moore’s law dead well I really believe
Moore’s law is not dead in fact we
barely begun to scratch the surface in
terms of the number of innovations we
will be able to bring about into the
marketplace in the next twenty to thirty
years
in the next twenty to thirty year time
horizon we will bear witness to some
more astounding number of innovations
made possible by artificial intelligence
quantum computing and optoelectronics
this hat has led many to speculate and
use this very forum the TED forum to
argue that we are on a verge of
singularity the singularity is
essentially a summation of our
scientific knowledge doubling roughly
every hour every day by 2050 rather than
every two years this is a bold idea but
if we believe we’re on an exponential
exponential curve of progress this will
indeed happen and in the 20th century
we’ve seen that we are exponentially
growing our technology so on a more
philosophical footing what does it mean
when you say Moore’s law is dead what
does it mean when you ask that when
you’re asking that what you’re really
questioning or asking is human ingenuity
debt is human curiosity that what I
firmly believe as Homan Homo sapiens
we’ve never lacked ingenuity or
curiosity ever since we’d set sail from
the northern plains of Africa and
expanded our species to other continents
we’ve always been explorers of our world
of our universe the world of the very
large today we’re exploring the world of
the very small the nano scale the
molecular scale the atomic layer finally
what does this mean for our society what
does it mean for us well can be
computing is pervasive and it’s here to
stay there is no going back
societies that are able to push the
emblem of progress are able to embrace
progress are able to not only even
embrace it but contributed contribute to
it or will be the societies that will be
successful our nation’s we have a duty
to contribute to a bar
of scientific knowledge and if we do
that if we expand our scientific
knowledge we will be able to not only
control the marketplace of ideas but
also control the marketplace itself
thank you very much
[Applause]
Please follow and like us: