Transcript Has the Moor*

More than Moore, how much more?
Fifty years of Moore’s Law
József Gyulai
Professor Emeritus
Chair Electronic Devices, BME-BUTE,
Inst. Tech. Phys.& Matl. Sci., MFA
Nobel prizes – near microelectronics
Nobel Committee considers usefulness
J. Bardeen, W.H. Brattain, W. Shockley, transistor
(1956)
L. Esaki, I. Giaever, B.D. Josephson, application of
tunneling (1973)
K. von Klitzing, quantum Hall-effect (1985)
E. Ruska electron microscope, G. Binnig, H. Rohrer
tunnel microscope (1986)
Z.I. Alferov, semiconductor laser, H. Kroemer, UHF
transistor, optics, J. S. Kilby, integrated circuit
(2000)
W.S. Boyle, G.E. Smith, CCD optics, Charles K. Kao,
optical fiber (2009)
A. Geim, K. Novoselov, graphene (2010)
N, Shuji, A. Hiroshi and A. Isamu, blue LED diode
Geim was awarded in 2000 with shared IgNobel
prize IgNobel: “first make people laugh, then
make them think”: diamagnetic levitation
1T- 10T field is
enough for
levitation of living
bodies
Miniaturization
• Key to success of microelectronics was that “scale down”:
worked: transistor with shrunk dimensions possess the
same characteristics except for heat dissipation...
• As fringe benefit, smaller sizes brought not only
portability, even wearebility, less power consumption, but
also increased reliability:
• amount of intelligence stacked into the device added to
reliability
• A good figure of merit is one mistake for 1010 steps, which
with added redundant organization can still be improved.
”Moore’s Law”
• “Doubling the number of elements on the chip
yearly”... “may work till end of the seventies…”
– says Gordon Moore (Electronics, 38(8), apr.19,1965)
• We may say that this is a generic law, which is
more a law of economy than that of technology
– production only satisfies market demands!
• International Technology Roadmap for
Semiconductors, ITRS: http://public.itrs.net/
a four-yearly study with biannual corrections
”Moore’s Law” in my past, develop
ionimplantation to technology
• Too early patent of W. Shockley became a fortune globally:
when the need for the technique became general, the
patent was worn out (17 years passed)...
• My first postdoc stay at Caltech (1969-70) was preceded
just by one year, when Intel (Moore, Noyce, Deal, Grove,
Vadasz...) left Fairchild – the company supplying the
forming Mayer group, including me
• Intel’s start-up success was a simple trick. They were first to
produce logic gates on a single chip
• The early success was the cause that the young Intel
refused to apply ion implantation as we argued at the 1st
Conf. on Ion Implantation (Thousand Oaks, CA,1970).
• In the coming decade, the Mayer-Gyulai (Caltech) group
contributed that ion implantation became a crucial
technology (pre-amorphization technique)
”Moore’s Law”
• First: National Technology Roadmap for
Semiconductors – determine developments
enabling fulfillment of ”Moore’s Law”
• Later changed to ”International Technology
Roadmap for Semiconductors, ITRS:
http://public.itrs.net/ – a four-yearly study with
biannual corrections
• Now, expanded, deals not only with memories
and processors, but with tools for telecommunication (high frequency, optical, etc.) and
portability
An example from ITRS: Difficult technology tasks, 2011:
black “known by industry”, Yellow “needs development” ,
white “no known solution”, “red brick wall”
– to date always found solution...
(PROCESS INTEGRATION, DEVICES, AND STRUCTURES
Success of ITRS on 2010 issue,
http://public.itrs.net/
How long will this work?
More than Moore, how much more?
Another, recent example from ITRS:
Difficult technology tasks, 2013:
(Lithography challenges)
Intel processor using 14 nm technology
- instead of 2022 news of the day
Moore’s Law today
• Moore’s Law is a ”law” more of
economics than of technology –
proof are present investments
into post-Moore solutions
• Mark Bohr, Intel's director of
process architecture and
integration: ”Moore’s Law in its
original form is dead, since a
decade”.
• In its original coverage, yes, but
• Scale down may work till 2020,
but question goes not only for
memories and processor, but for
telecommunication, etc., too.
Go to 3D, optics, spintronics,
etc. (Graphs from ENIAC
project)
Non-volatile memory forecast
• Going 3D, i.e., to stacked
structures, NAND and NOR
Flash memory, scaling down
to 12 nm half-pitch looks
straightforward till 2028
• Research is needed for
magnetic/spin torque and for
resistive devices
• Reliability issue is difficult
because of complex
structure: failure mechanisms
are very different for
transistors, for interconnects,
etc. May lead e.g., to need of
optical or carbon based
interconnects
Hot areas in our field
• Computer in telecommunication
–
–
–
–
Mobile devices
Wear-on devices
Ambience intelligence (intelligent car, intelligent “mote”)
Acoustic devices
• “Revolution” of sensors and coupled actuators
– Can be biomaterial, too…
– Automation of transport
• Micro- és nanotechnology
– „Energy harvesting”
• “Revolution” of lighting
– Light emitting diode (LED), Organic LED
• Priorities in EU:
– „Energy efficient buildings”,
– „Green car”,
– „Factory of the future”
ITRS Roadmap 2013 Conclusions
•
•
•
“First of all, the aggressive bi-annual introduction of new semiconductor
technologies allowed ICs, consisting of even hundreds of million of
transistors, to be produced cost effectively. This made it possible to
integrate extremely complex systems on a single die or in a single package
at very attractive prices. Furthermore, progress in packaging technology
enabled the placement of multiple dies within a single package. These
categories of devices are defined as system on chip (SOC) and system in
package (SIP).
Second, manufacturers of integrated circuits offering foundry services were
able to provide, once again, the “New ASICs” at very attractive costs. This
led to the emergence of a very profitable business for design “only” houses,
i.e., companies that do not manufacture ICs themselves, but produce the
designs that are manufactured elsewhere.
Third, development of sophisticated equipment for advanced integrated
circuits proliferated to adjacent technology fields and by so doing the
realization of flat panel displays (FPD), MEMS sensors, radios and
passives, etc., was made possible at reasonable costs. Under these
conditions system integrators were once again in the position to fully control
system design and product integration. “
Nanoelectronics
• Scale down cannot go ‘ad
infinitum’, new solutions are
needed (<14 nm node):
• Smart cut© - “peel” bulk
semiconductor materials, like
were mica, to make nearly
2D substrates, for SOI
wafers...
• New materials for channel,
higher mobility than that of
Si:
– Six Ge1-x
– InGaAs (L.Czornomaz:
Comp.
Semicond, 1/2014, 32)
– Graphene? Lithography
solutions by the Biró group
Some expectations
• The other day in a class on monolythic IC
technology, I was mumbling if the topic will
have importance in post-Moore times
• I think, it will, as it represents a basic
technical knowhow necessary even in
biotech applications
Hope for future of Moore’s law and potential
proof for its economic nature
• What I feel crucial is that new solutions in the
nanometer region will or will not be able to
satisfy reliability issues
• E.g., if biomolecules come in play, will the rather
exact electronic reliability issue merge with
redundancy-operated ‘reliability’ of biosystems?
• How close can we go to the kT-limit?
• I expect that once in the future, when we’ll have
whatever, e.g., biocomputers, then will come an
engineer, who for fun determines the gate level
equivalent circuit for that device
• and he’ll find that the exponential law remained
valid...
TRL: Technology readiness level
Future of nanoelectronics
• In my view, only solutions will
‘make it’ which fit into today’s
foundries, albeit with slight
modifications
• New solutions needed
because of scale down
limitations:
– Instead of electron
conduction, other binary
systems, e.g., shift register
using spin, ~tronics (D.
Jamieson, Melbourne)
– Optical data transport on
chip instead of metallization
– inevitable
– Optics: plasmonics?
– Analog vs digital systems
– So-called biomimetic
solutions
Insulator
28Si
Substrate
Quantumcomputer solutions
“Qubit” denotes entangled
particles,
e.g., seven qubit of 5 fluorine and
2 carbon can factorize 15:
3٠ׁ5=15
Radio waves trigger and NMR
reads out the result
2012: Superconductiong
stabilization (10 μs) of qubit using
silicon technology!
IBM Research
Dicarbonylcyclopenta
dienyl
(perfluorobutadien-2yl) iron
(C11H5F5O2Fe )
(ill.
pentafluorobutadienyl
cyclopentadienyldicar
bonyl-iron complex)
Vision
• 20th century almost erased border
between physics and chemistry,
• In the 21st, I’m expecting this to happen
towards biology
• Arsenal of mathematics is improving
simultaneously allowing ‘quasi-exact’
solutions to problems
„Ceterum censeo...”:
•
Science and technology of today can only have one
mission and maybe two main directions –
corresponding somehow to preservation of self and
of the race:
1. Search ways, modes whether, and if positive, on
what technical level can 8 – 10 billion people live
on Earth in kind of symbiosis with other forms of
life?…
2. Extension of life span of individuals attracts great
interest – causing unheard-of development of
biological science
If ‘recipes’ become known, will society absorb
them and put into action in time?
Buckminster Fuller, architect and one of Club of
Rome founders:
Operating Manual for Spaceship Earth - (1969)
“…One outstandingly
important fact regarding
Spaceship Earth, and
that is that no instruction
book came with it…„
Biosphere, Montreal, 1967
Recyling economy –
when?
Fullerenes
Thank you for your attention