Rapid Rates of Improvement : Which Technologies and Why?
What does this tell us about the future?
Associate Prof Jeffrey Funk
National University of Singapore
More details on these ideas can be found in
1) What Drives Exponential Improvements? California Management Review, May 2013
2) Technology Change and the Rise of New Industries, Stanford University Press, January 2013
3) Exponential Change: What drives it? What does it tell us about the future? forthcoming 2014
4) http://www.slideshare.net/Funk98/presentations
How Objective are we?
Can we effectively think about the “future of
technology” without understanding rates of
improvement?
◦ Without improvements, how will the status quo
change
◦ Some technologies experience more rapid rates
of improvement than do other technologies
What happens when we don’t understand rates
of improvement?
Cognitive Biases
Nobel Laureate Daniel Kahneman
People assess relative importance of issues, including new technologies ◦ by ease of retrieving from memory
◦ largely determined by extent of coverage in media
◦ E.g., media talks about solar, wind, battery-powered vehicles, bio-fuels and thus many think they are have rapid rates of improvement - but only some are
Second, judgments and decisions are guided directly by feelings of liking and disliking ◦ One person invested in Ford because he “liked” their
products – but was Ford stock undervalued?
◦ Many people “like” some technologies and dislike others without considering rates of improvement
Source: Daniel Kahneman, Thinking Fast and Slow, 2011
Isn’t there a more deliberate and logical way?
Understanding rates of improvement can help firms,
universities, and governments better understand when
new technologies might become economically feasible
Technologies must have some level of performance and
price for specific applications before they begin to diffuse ◦ Technologies that experience faster rates of improvement are more likely to
become economically feasible….
◦ They are also more likely to become economically feasible for increasing
number of applications and thus diffuse…
◦ This has implications for R&D policy and solving global problems
But which technologies are experiencing rapid rates of
improvement and why?
Technology Dimensions of measure Time Period Rate/Year Integrated Circuits Number of transistors per chip 1971-2011 38%
MEMS Number of Electrodes/Eye 2002-2013 45.6%
Drops/second for printer 1985-2009 61%
Organic Transistors Mobility (cm2/ Volt-seconds) 1994-2007 101%
Carbon Nanotube
Transistors
1/Purity (% metallic) 1999-2011 32.1%
Density (per micrometer) 2006-2011 357%
Superconducting
Josephson Junctions
1/Clock period 1990-2010 20.3%
1/Bit energy 1990-2010 19.8%
Qubit Lifetimes 1999-2012 142%
Bits per Qubit lifetime 2005-2013 137%
Photonics Number of Optical Channels 2005-2011 91.9%
Computers Instructions per unit time 1979-2009 35.9%
Instructions per time and dollar 1979-2009 52.2%
Quantum Computers Number of Qubits 2002-2012 107%
Technologies Experiencing Rapid Rates of Improvements
(Information Transformation)
Technology
Domain
Sub-
Technology
Dimensions of
measure
Time
Period
Rate/
Year Information
Storage
Magnetic
Storage
Areal recording density of
disks
1991-2011 55.7%
Areal recording density of
tape
1993-2011 32.1%
Information
Transmission
Last Mile
Bandwidth
Bits per second 1982-2010 48.7%
Wireless Bits per unit time 1980-2008 104.0%
Materials
Transformation
Carbon
Nanotubes
1/Minimum Theoretical
Energy for Production
1999-2008 86.3%
Biological
Trans-
formation
DNA Sequencing per unit cost 2001-2013 146%
Synthesizing per unit cost 2002-2010 84.3%
Cellulosic
Ethanol
Output per cost 2001-2012 13.9%
Technologies Experiencing Rapid Rates of Improvements
Technologies Experiencing Rapid Rates of Improvements
Technology
Domain
Sub-
Technology
Dimensions of
measure
Time Period Rate Per
Year Energy Trans-
formation
Light Emitting
Diodes (LEDs)
Luminosity per Watt 1965-2008 31%
Lumens per Dollar 2000-2010 40.5%
Organic LEDs Luminosity per Watt 1987-2005 29%
GaAs Lasers Power/length-bar 1987-2007 30%
LCDs Square meters per
dollar
2001-2011 11.0%
Quantum Dot
Displays
External Efficiency 1994-2009 79.0%
Solar Cells Peak Watt Per Dollar 2004-2013 21.0%
Photo-sensors
(Camera chips)
Pixels per dollar 1983-2013 48.7%
Light sensitivity 1986-2008 18%
Energy
Transmission
Super-
conductors
Current-length per
dollar
2004-2010 115%
What Drives Rapid Improvements? Drivers of improvements
◦ 1) creating new materials (and often associated processes) to better exploit physical phenomena
◦ 2) geometric scaling: increases and reductions in scale
◦ Some technologies directly experience improvements while higher-level “systems” indirectly experience
Rapid improvements are driven by them when ◦ “Creating new materials” lead to rapid improvements
when new classes of materials are being created
◦ Technologies that benefit from reductions in scale (e.g., integrated circuits) have more rapid rates than those benefitting from increases (e.g., engines)
A summary of these ideas can also be found in
1) What Drives Exponential Improvements? California Management Review, May 2013
2) Technology Change and the Rise of New Industries, Stanford University Press, January 2013
3) Exponential Change: what drives it? what does it tell us about the future? forthcoming 2014
4) http://www.slideshare.net/Funk98/presentations
What do these Technologies tell us about the Future?
No end to Moore’s Law? ◦ Better Integrated Circuits and Computing
◦ “Big Data” Analysis enables better management of systems including “energy systems”
The Cyborg Era? ◦ Better bio-electronics and health care
◦ DNA sequencing
Cleaner transportation
Clean energy production
No End to Moore’s Law
Improvements in existing ICs will probably continue for at least 15 years ◦ Smaller wavelength light sources (13 nm or 1/10 as small as previous) will take us from 22 to 5 nm (nanometer) feature sizes
◦ 3D ICs enable multiple layers of transistors
New technologies will be available before then Photonics, carbon nanotubes, ultra-thin materials including graphene, Josephson (Quantum computers)
Smaller features, i.e., “reductions in scale,” lead to better ICs (more transistors per IC chip, faster speeds, greater functionality,
lower power consumption per transistor)
New Technologies are also becoming available for ICs
Photonics for faster connections between transistors
Carbon nanotubes (CNTs) for channel layer
Ultra-thin materials to replace silicon Superconducting Josephson junctions (for Quantum computers)
Most of these involve creation of new materials and processes for them ◦ And some reductions in scale (photonics)
Photonics (Optical Connections) is Another Layer on a 3D Chip
http://www.slashgear.com/ibm-silicon-nanophotonics-speeds-servers-with-25gbps-light-10260108/
IBM silicon nanophotonics speeds servers with 25Gbps light, Chris Davies, Dec 10th 2012
Improvements in CNTs for Transistor Channels
Electronics: The road to carbon nanotube transistors, Aaron D. Franklin, Nature 498, 443–444 (27 June 2013)
Ultra-Thin Materials to Replace Silicon
As of April 2013, >10 materials found that are one or a
few atoms thick
Boron nitride (insulator) has been fabricated in one-
atom sheet as has Molybdenum Sulfide
◦ Molybdenum Sulfide is semiconductor, Boron Nitride is
insulator, Graphene is for interconnect
◦ Together one atom thick flash memory devices have been
constructed (http://www.thessdreview.com/daily-news/latest-buzz/flash-
memory-to-be-based-on-2d-materials-a-single-atom-thick/)
◦ More complex devices can be constructed by doping one of
the layers
http://thessdreview.com/daily-news/latest-buzz/flash-memory-to-be-based-on-2d-materials-a-single-atom-thick/
April 29, 2013. http://edition.cnn.com/2013/04/29/tech/graphene-miracle-material/index.html?hpt=hp_c3
Bit Energy = power consumed per clock period x number of active devices
RSFQ: rapid single flux quantum, relies on quantum effects in superconducting devices Source: superconductivity web21, January 16, 2012. www.istec.or.jp/web21/pdf/12_Winter/E15.pdf
Improvements in Power Consumption and Speed of
Superconducting Josephson junctions
Improvements in Josephson
junctions also enable
increases in Qubit lifetime
and number of Qubits in a
Quantum Computer
Note: Performance increases faster than does
number of Qubits, Google bought a quantum
computer from D-Wave in 2013.
Science, Vol 339, 8 March 2013, pp. 1169-1174
http://nextbigfuture.com/2013/05/dwave-512-qubit-
quantum-computer-faster.html
Computing–“Big Data” Analysis
Improvements in ICs and computing enable more extensive data analysis of output from ◦ Particle accelerators, telescopes
◦ DNA sequencing equipment,
◦ other types of scientific and medical equipment
They also make it cheaper to create large mathematical models to make predictions, rather than pursue more efficient algorithms ◦ better translations
◦ better predictions of flu trends, inflation, health problems, loan defaults, rising food prices, and even social problems such as riots or terrorism
Big Data: A Revolution That Will Transform How We Live, Work, and Think, Viktor Mayer-Schonberger, Kenneth Cukier
Sensors Enable More Types of “Big Data” Analysis and Optimization
Higher resolution camera chips
Better MEMS (micro-electronic mechanical systems) ◦ Smaller feature size lead to higher performance
◦ Current feature sizes of 0.5 to 1.0 microns for MEMS and thus industry is like ICs were in 1980
◦ MEMS will probably have similar impact as ICs
In combination with conventional ICs, lasers, and Internet, better MEMS enables ◦ 3D scanners, printers, holographic displays
◦ eye-tracking devices, autonomous vehicles
◦ better health care and management of buildings, dams, bridges, power plants……..
Smaller feature
sizes lead to better
gas chromatographs,
ink jet printers
(drops/second and
resolution), and mobile
phone discrete
components (smaller)
Stasiak J, Richards S, and Angelos S 2009. Hewlett Packard's Inkjet MEMS Technology, Proc. of Society of SPIE:7318, Clark Ngyuen, Univ of CA, Berkeley
Improvements in ICs and Sensors are Also Improving Energy Usage
Better ICs and sensors enable better process control and better collection of data, extending the Internet to more devices
This data can improve simulation tools that are also coming from improvements in ICs
Traffic management being improved ◦ Traffic sensors, smart cards, better fare management
◦ Predictive analytics with better computers
◦ Navigation systems with better ICs and MEMS
◦ Goal should be to dramatically reduce public and private vehicle breakdowns and accidents
Consider Lighting Systems
Costs of LEDs are falling as better materials are created (higher flux/package) and size of wafers are increased (lower cost/lumen) ◦ their small size enables more
aesthetic designs
◦ by using sensors, we can create lighting systems that only illuminate those areas that are needed and when they are needed
◦ Motion, heat and other sensors track movements of humans, animals, and vehicles
http://www.globalsmtseasia.com/index.php?option=com_content&view=article&id
=4228:printed-electronics-for-flexible-solid-state-lighting&catid=109&Itemid=115
Improvements in Wireless (from improvements in ICs)
Enable Better Lighting Systems and Greater Access to Sensors
http://www.ruhr-uni-bochum.de/integriertesysteme/emuco/files/hipeac_trends_future.pdf
Wireless Enables Greater Access and Control of Sensors Many kinds of sensors and applications
◦ Environmental (temperature, pressure, gas content) ◦ Physiological (heart rate, brain wave, blood pressure) ◦ For vehicular and human traffic and many types of
infrastructure (factories, buildings, dams, bridges, power plants)
Data can be sent wirelessly to Internet for analysis and interpretation
The phone may become a major collection, analysis, and control point for data ◦ control and program the thermostat, lighting, and other appliances
◦ Rent bicycles, vehicles and other things to increase capacity utilization and reduce energy usage
Phones get better as Human-Computer Interfaces are Improved
Better and cheaper touch displays ◦ Create materials for “blind” typing or texture feedback
◦ Lower costs from larger substrates, new materials (e.g., OLEDs) and new processes (roll-to roll printing)
Better gesture displays ◦ Better cameras enable gestures
to be interpreted
Augmented Reality ◦ Google glasses will get better as
cameras, eye-tracking, and ICs get better
Neural interfaces ◦ Smaller features lead to better interfaces
Neural Interfaces: smaller MEMS/ electrodes lead to higher resolution
What do these Technologies tell us about the Future?
No end to Moore’s Law? ◦ Better Integrated Circuits and Computing
◦ “Big Data” Analysis enables better management of systems
The Cyborg Era? ◦ Better bio-electronics and health care
◦ Better DNA sequencers
Cleaner transportation
More clean energy
Bio-Electronic ICs
A special type of MEMS
Benefits from reductions in scale
◦ enable reductions in sample and reagent volume, faster
reactions and response time, higher throughput, and
analysis of smaller biological materials
◦ reductions in size of electrodes have enabled increases in
eyesight; further reductions can enable 20-20 vision
Result is better
◦ point-of care diagnostic equipment
◦ implants (e.g., bionic eyes)
Bio-electronic ICs will probably have similar
impact in future as conventional ICs in past
Source: Biomaterials 29(24–25): 3393–3399
MEMS-BasedElectrode
Electrode Implanted Into Retina
Smaller feature sizes for MEMS-
based electrodes lead to better
eyesight for people who suffer
from macula
In Combination with Skin Patches, Other Sensors and Wireless
More physiological data can be collected and sent wirelessly to phones or other devices
Faster detection of health problems
Enabled by flexible materials (e.g., higher mobility of organic materials) or island-bridge design with silicon
Huanli Dong , Chengliang Wang and Wenping Hu, High Performance Organic Semiconductors for Field-Effect Transistor, Chemical
Commununications, 2010,46, 5211-5222
Improvements in Mobility of “Flexible” Organic Materials Enable
Better Conformance of Electronics to Human bodies and Organs (performance of polycrystalline silicon is 100 and single crystal silicon is 1000)
Will Mobile Phones become Main Platform for Data Collection?
Phones have high-performance processors, memories, and displays
Easy to develop and download apps
Can create accessories/attachments ◦ test strips to analyze blood, skin, saliva; check for flu, insulin and other sicknesses
◦ microscope to analyze cells
◦ electrodes for electrocardigam
◦ others for ultrasound, MRI, etc.
Improvements in human-computer interface make mobile phones easier to use
http://www.economist.com/news/technology-quarterly/21567208-medical-technology-
hand-held-diagnostic-devices-seen-star-trek-are-inspiring
What do these Technologies tell us about the Future?
No end to Moore’s Law? ◦ Better Integrated Circuits and Computing
◦ “Big Data” Analysis enables better management of systems
The Cyborg Era? ◦ better bio-electronics and health care
◦ Better DNA sequencers
Human-computer Interface
Cleaner transportation
More clean energy
http://www.genome.gov/sequencingcosts/
What Drives Cost
Reductions?
New methods of sequencing
Improved lasers and cameras
to read fluorescent dyes
More parallel processing
Smaller feature sizes. Just like
ICs and MEMS, smaller
feature sizes lead to lower
costs and faster speeds
Implications of Falling Cost of Sequencing (and Synthesizing)
Enables better understanding for individuals of
risks for specific diseases
potential side effects from drugs
Faster and cheaper drug development ◦ drugs can be developed for smaller numbers of people
◦ personalized medicine
Develop better crops and materials from living organisms ◦ E.g., bio-mimicry
Green Machines for Better Crops?
Better sensors (cameras, infrared, fluorescence, lasers) and mechanical controls enable complete control and measurement over crop growth
DNA sequencing enables characterization and replication of high performing crops
What do these Technologies tell us about the Future?
No end to Moore’s Law? ◦ Better Integrated Circuits and Computing
◦ “Big Data” Analysis enables better management of systems
The Cyborg Era? ◦ better bio-electronics and health care
◦ Better DNA sequencers
Cleaner transportation ◦ But not how you think
More clean energy
Battery Performance Doubles about Once Every 15 years
But Gasoline has energy density 30 times higher!
Can we wait 75 years for adequate batteries?
High density system of charging stations ◦ Facilitated by smart grid, which is
enabled by continued improvements in Internet
◦ Users can quickly find charging stations with GPS, other sensors
◦ Reduces necessary energy density of batteries
Lower cost of power electronics enable move from mechanical to electronic controls ◦ Lower weight of vehicle reduces
required energy density of batteries
There are other options!
High density of charging stations facilitated by high temperature super-conducting (HTS) transmission lines
Autonomous vehicles enable faster moving and more densely packed vehicles ◦ Better fuel efficiency
through less congestion
◦ Their cost is falling as the cost of sensors fall
And other options!
Copper is 15-25$/kA-m
Looking Further to the Future
Carbon nanotube (CNT) or graphene-based flywheels have potential energy densities 100 times higher than do lithium-ion batteries
CNT or graphene-based automobiles would be much lighter than conventional ones and thus require much less energy storage densities in their batteries
But not 75 years in future!
Sources: Presentation by my students on April 11, 2013. Slides can be found on http://www.slideshare.net/Funk98/presentations. RIght figure:
Minimum Exergy Requirements for the Manufacturing of Carbon Nanotubes, Timothy G. Gutowski, John Y. H. Liow, Dusan P. Sekulic, IEEE,
International Symposium on Sustainable Systems and Technologies, Washington D.C., May 16-19, 2010
Falling Energy Cost for CNTs
What do these Technologies tell us about the Future?
No end to Moore’s Law? ◦ Better Integrated Circuits and Computing
◦ “Big Data” Analysis enables better management of systems
The Cyborg Era? ◦ better bio-electronics and health care
◦ Better DNA sequencers
Cleaner transportation ◦ But not how you think
Clean energy production
Only some have rapid rates of improvement
Like batteries, wind turbines have very slow rates of improvement (2%/year)
Cellulosic ethanol was experiencing rapid reductions in cost through increases in scale ◦ But they have recently
slowed and further reductions are needed..
Rapid reductions in cost of solar cells
plus rapid improvements in efficiency for some types of solar
cells (multi-junction III-V, multi-junction organic, organic single
junction, and quantum dot)
organic solar cells are much cheaper on a per area basis than
are other solar cells. Thus they are potentially much cheaper on
a peak-watt basis if their efficiencies are improved
Cost per peak watt of $0.25 are probably needed so we are
probably a factor of three from achieving grid parity
Cheaper
superconducting
transmission
lines facilitate
generation of
electricity in
North Africa
for Europe
(and in Mexico
for U.S.)
Source: Wikipedia
Desertec
Conclusions
We need to think more effectively about future of technology
This requires change in method of analysis ◦ from “instinctive and emotional”
◦ to “slower, more deliberative, more logical”
The future is too important to assess relative importance of technologies ◦ by ease of retrieving them from memory
◦ by letting judgments and decisions be guided by feelings of “liking” and “disliking”
It is easy to believe that certain technologies are important because ◦ the media regularly discusses them or
◦ we like them
Rates of Improvement are Important An important part of “slower, more deliberative,
more logical” method of analyzing technologies is ◦ better data on rates of improvement and a better
understanding of their drivers
Technologies must have some level of performance and price for specific applications before they begin to diffuse ◦ Technologies that experience faster rates of improvement
are more likely to become economically feasible….
◦ They are also more likely to become economically feasible for increasing number of applications and thus diffuse…
This has implications for R&D policy and solving global problems
Some Technologies Experience More Rapid Rates than do Others
Many technologies experience rapid rates of improvement ◦ And these improvements tell us something about
the future
But there are probably other technologies that are also experiencing rapid rates of improvements ◦ We need more data on these technologies
These slides also help us understand the drivers of the improvements and thus the reasons for them ◦ This can help us identify those technologies with
the potential for rapid improvements
Drivers of Rapid Improvements
Drivers of improvements ◦ 1) creating new materials (and often associated
processes) to better exploit physical phenomena
◦ 2) geometric scaling
◦ Some technologies directly experience improvements while higher-level “systems” indirectly experience them
Rapid improvements are driven by them when ◦ Creating new materials lead to rapid improvements
when new classes of materials are being created
◦ Technologies that benefit from reductions in scale (e.g., integrated circuits) have more rapid rates than those benefitting from increases (e.g., engines)
A summary of these ideas can also be found in
1) What Drives Exponential Improvements? California Management Review, May 2013
2) Technology Change and the Rise of New Industries, Stanford University Press, January 2013
3) Exponential Change: what drives it? what does it tell us about the future? forthcoming 2014
4) http://www.slideshare.net/Funk98/presentations
Creating Materials
Leads to orders of magnitude improvements when scientists and engineers create new forms of materials and do this with new processes
Sometimes these improvements involve new classes of materials (See next slide)
Without these new classes, the range of improvements might well be reduced below those achieved and documented earlier
Improvements done mostly in laboratories, not in factories
Technology
Domain
Sub-
Technology
Different Classes of Materials
Energy
Trans-
formation
LEDs Group III-V, IV-IV, and II-VI semiconductors
Organic LEDs Small molecules, polymers, phosphorescent materials
Solar Cells Silicon, Gallium Arsenide, Cadmium Telluride, Cadmium
Indium Gallium Selenide, Dye-Sensitized, Organic
Information
Trans-
formation
Organic
Transistors
Polythiophenes, thiophene oligomers, polymers,
hthalocyanines, heteroacenes, tetrathiafulvalenes, perylene
diimides naphthalene diimides, acenes, C60
Basic
Materials
Superconductors Simple elements (tin and aluminium), metallic alloys,
heavily-doped semiconductors, ceramic compounds
containing planes of copper and oxygen atoms (cuprates),
iron- and organic-based ones
Carbon
Nanotubes
Properties impacted by number of walls, diameter of walls,
axes of walls
Different Classes of Materials were found for Many Technologies
Geometric Scaling
Impacts on some technologies through both reductions and increases in scale
In both cases, large changes in both product and process design were implemented with each increment requiring non-trivial redesigns
Reductions in scale provide a mechanism for rapid rates of improvements in ICs, magnetic storage, MEMS, and DNA sequencing equipment ◦ involved better processes with completely new forms of
equipment and materials
◦ new equipment usually developed and implemented in labs
◦ Led to rapid improvements in many higher-level “systems”
Reductions in Scale
Lead to particularly rapid rates of improvement ◦ Most technologies become cheaper as they are made smaller
◦ But performance only rises for a few technologies
Performance rises for ICs, magnetic storage, MEMS, and DNA sequencing equipment as feature sizes are reduced
Finding these types of technologies is a major challenge
One technology that benefits from both reductions in scale and in creating new materials is nanotechnology ◦ Benefiting from nanotechnology is mostly about creating
materials that benefit from single nanometer feature sizes
◦ Such materials include carbon nanotubes, graphene, nano-particles, nanofibers and many others that can be made at single or nearly single-atom thicknesses
Implications for R&D Policy
One goal of R&D policy should be to fund
those technologies with
◦ rapid rates of improvement or
◦ with potential for rapid rates of improvement
◦ since these technologies will have larger impact
on world than will other technologies
These slides help us understand drivers of improvements and thus reasons for them ◦ This can help us identify those technologies with the potential for rapid improvements
Implications for Solving Global Problems
Rapidly improving technologies represent a kind of “tool chest” ◦ that can be used to solve global problems
It’s not just current performance and cost of them that provide us with useful tools ◦ their rapid rates of improvement mean that better
tools continue to emerge
Let’s think about how these better tools can help solve global problems ◦ For example, many of these technologies will have
bigger impact on solving “energy” problems than will the predominant view (batteries or wind turbines)
Appendix
Thank You