Nobel Laureate Robert Solow concluded that 85% of America’s productivity growth comes from innovation. But how can we characterize this innovation? One way we can characterize this innovation is through the improvements in cost and performance that technologies experience over time since many innovations are required for these improvements to occur. These slides investigate the rates of improvement for 33 different technologies and 52 dimensions of performance/cost and conclude that the drivers of these improvements can be placed in two categories: 1) creating materials (and their associated processes) that better exploit physical phenomena; and 2) geometrical scaling. For geometric scaling, some technologies experience improvements through increases in scale while a small number of technologies experience them through reductions in scale.

1. Exploring the Design Mechanisms
that Drive Improvements in
Performance and Cost
A/Prof Jeffrey Funk
National University of Singapore
Prof Christopher Magee
MIT
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?
http://www.slideshare.net/Funk98/exponential-change-what-drives-it-what-does-it-tell-us-about-the-future-14104827

2. Performance and Cost are Important (1)
Necessary (but insufficient condition) for
improvements in productivity (and value
propositions)
 Solow’s (1957) Nobel Prize winning research found
that most growth comes from innovation

◦ improvements in cost and performance for a technology is
one measure of innovation
◦ faster rates of improvement directly impact on output-to
input ratio of economic activities and thus on productivity
growth

3. Performance and Cost are Important (2)

Large impact on diffusion via effect on profitability of
users (Griliches, 1957; Mansfield, 1968)
◦ Greater profitability leads to faster rates of diffusion and the
first users tend to be those with the greatest profitability
◦ In summary, improvements in cost and performance of new
technologies impact on both the rate of diffusion and the level
of the impact of the technology on productivity

Helps us
◦ implement better R&D policies
◦ understand when new technologies become economically
feasible, which helps us solve global problems

4. But what drives improvements?

Predominant view is rather vague
◦ Changes in product design lead to improvements in
performance and changes in process design lead to
improvements in cost (Utterback, 1994; Adner and
Levinthal, 2001)
◦ Novel combinations of components (Basalla 1988; Iansiti
1995)
◦ Costs fall as cumulative production grows in learning or
experience curve (Wright 1936; Arrow 1962; Argote and
Epple 1990; Ayres 1992), some argue as automated
manufacturing equipment is introduced and organized
into flow lines (Utterback, 1994)

5. Another View: Geometric Scaling




Building from various engineering literatures, some argue:
changes in physical scale are important mechanisms for
improvements
Gold (1974, 1981) argued this phenomenon overlooked
when cumulative production and thus learning curves are
emphasized
Lipsey et al (2005) focus on theoretical reasons for
benefits from increases in scale, as does Winter (2008)
Winter also discuses technologies that benefit from
reductions in scale such as ICs and membranes. Winter
calls for better understanding of scaling, impact on
production functions, and thus drivers of cost and
performance improvements

6. Methodology
Looked for cost and performance data on wide variety of
technologies; called trajectories by Dosi (1982);
technologies are usually defined in terms of single
concept/principle (Uttterback, 1994; Henderson and
Clark, 1990)
 Began with already possessed data
 Found new data in

◦ Social science archival publications giving quantitative data
over time (Martino, 1971; Koh and Magee, 2006 and 2008)
◦ Scientific and engineering journals
◦ Google searches

7. Annual Rates of Improvement for Specific Technologies
Technology
Domain
Energy
Transformation
Sub-Technology
Dimensions of measure
Time Period
1 Lighting
2 LEDs
3 Organic LEDs
4 GaAs Lasers
5 Photosensors
6 Solar Cells
7 Aircraft engine
Light intensity per unit cost
Luminosity per Watt
Luminosity per Watt
Power/length-bar
Light sensitivity (mV/micrometer)
Power output per unit cost
Gas pressure ratio achieved
Thrust per weight-fuel consumed
Power of aircraft engine
Energy transformed per unit mass
Energy transformed per unit mass
Energy transformed per unit volume
Energy stored per unit volume
Energy stored per unit mass
Energy stored per unit cost
Energy stored per unit cost
Energy stored per unit mass
Energy stored per unit cost
Energy stored per unit mass
Energy transported times distance
Energy transported times distance per
unit cost
1840-1985
1965-2008
1987-2005
1987-2007
1986-2008
1957-2003
1943-1972
1943-1972
1927-1957
1896-1946
1880-1993
1890-1997
1882-2005
1882-2005
1950-2002
1945-2004
1962-2004
1983-2004
1975-2003
1890-2003
1890-1990
8 Piston engines
9 Electric Motors
Energy
storage
10 Batteries
11 Capacitors
12 Flywheels
13. Energy Transport
Improvement
Rate Per Year
4.5%
31%
29%
30%
18%
16%
7%
11%
5%
13%
3.5%
2.1%
4%
4%
3.6%
4%
17%
18%
10%
10%
2%

8. Information 14 ICs (Microprocessors)
Transfor15 MEMS Printing
mation
16 Computers
Number of transistors per chip/die
1971-2011
38%
Drops per second for ink jet printer
1985-2009
61%
Instructions per unit time
1945-2008
40%
Instructions per unit time and dollar
1945-2008
38%
17 Liquid Crystal Displays Square meters per dollar
2001-2011
11%
18 MRI
1/Resolution x scan time
1949-2006
32%
19 Computer Tomography
1/Resolution x unit time
1971-2006
29%
20 Organic Transistors
Mobility (cm2/ Volt-seconds)
1994-2007
101%
Information 21 Magnetic Tape
Storage
Bits per unit cost
1955-2004
40%
Bits per unit volume
1955-2004
10%
22 Magnetic Disk
Bits per unit cost
1957-2004
39%
Bits per unit volume
1957-2004
33%
Bits per unit cost
1996-2004
40%
Bits per unit volume
1996-2004
28%
Bits per unit time
1858-1927
35%
Bits x distance per unit cost
1858-2005
35%
Coverage density, bits per area
1901-2007
37%
Spectral efficiency, bits per unit
bandwidth
Bits per unit time
1901-2007
17%
1895-2008
19%
23 Optical Disk
Information
Transport
24 Wireline Transport
25 Wireless Transport

9. Living
Organisms
Materials/
Matter
Other
Biological
transformation
29 Transport of
humans/freight
30 Load Bearing
31 Magnetic
32 Machine
Tools
33 Laboratory
Cooling
26 Genome sequencing per unit cost
27 Harvest concentration of penicillin
28a U.S. wheat productivity (per input)
28b US wheat production per area
Ratio of GDP to transport sector
Aircraft passengers times speed
Strength to weight ratio
Magnetic strength
Magnetic coercivity
Accuracy
Machining speed
Lowest temperature achieved
1965-2005
1945-1980
1948-2009
1945-2005
1880-2005
1926-1975
1880-1980
1930-1980
1775-1970
1900-1975
1880-1950
35%
17%
1.3%
0.9%
0.46%
13%
1.6%
6.1%
8.1%
7.0%
6.3%
28%
Sources, from top to bottom: (Nordhaus,1997; Azevedo, 2009; Sheats et al, 1996; Lee, 2005; Martinson, 2007; Suzuki, 2010; Nemet, 2006;
Alexander and Nelson, 1973; Sahal, 1985; Koh and Magee, 2008; Wikipedia, 2013; Stasiak et al, 2009, Koh and Magee, 2006; Koomey, 2010;
Economist, 2012; Kurzweil, 2005; Kalender, 2006; Shaw and Seidler, 2001; Dong et al, 2010; Koh and Magee, 2006; Amaya and Magee, 2008;
NHGRI, 2012; Seth, Hossler and Hu, 2006; U.S. Department of Agriculture, 2012, Glaeser and Kohlhase, 2004: Martino, 1971; NAS/NRC, 1989;
Ayres and Weaver, 1998; American Machinist, 1977; Martino, 1971)

10. Methodology - continued

Our initial analysis of the technologies was aimed at
understanding the composition of a technology’s
system
◦ i.e., “nested hierarchy of subsystems” (Tushman and
Rosenkopf, 1992; Tushman and Murmann, 1998)

Then considered geometric scaling
◦ Examples of geometric scaling were searched for outside of
chemical plants, furnaces, and smelters (since these have
been empirically analyzed to some extent)
◦ For each instance of geometric scaling, type of geometrical
scaling was identified and data on changes in scale and on
cost/price for various levels of scale were gathered

This still left us with a large number of technologies
whose improvements were not well explained

11. Methodology - continued

Second mechanism is engineers and scientists
create (or improve existing) materials to better
exploit underlying physical phenomena
◦ This often involved simultaneously creating new processes
for producing them (Stobaugh 1988; Morris et al 1991;
Olsen, 2000; Linton and Walsh 2008, Magee 2012)


Word “create” is used because scientists and
engineers often create materials that do not
naturally exist (as opposed to finding them) and in
doing so must also create the processes
Improvements often involve new “classes” of
materials and not just modifications to existing
materials

12. Methodology - continued

Data on cost and performance
improvements was collected
◦ time series
◦ specific moments in time

Performance improvements from creating
materials were almost always in form of a
time series graph
◦ that included names of materials

For scaling, looked for data for a single
moment in time in order to isolate impact of
changes in scale, which was found for most
technologies

13. Methodology - continued

Each technology was assigned to one of two
mechanisms (and to identify important
component technology changes)
◦ even though many benefited from both
mechanisms

We also note that these two mechanisms are
attempt at categorizing complex set of
changes and that each mechanism is by
itself complex and in specific instances is
enabled or accompanied by other technical
knowledge

14. Outline of Results


Creating materials (and their associated
processes) that better exploit physical
phenomena
Geometrical scaling
◦ Reductions in scale: e.g., integrated circuits (ICs),
magnetic storage, MEMS, bio-electronic Ics
◦ Increases in scale: e.g., larger production equipment,
engines, oil tankers

Some technologies directly experience
improvements while others indirectly experience
them through improvements in “components”
◦ Computers and other electronic systems
◦ Telecommunication systems

15. Items 1, 2, 3: involve lighting

16. Other Evidence for Lighting


Full quote for LEDs from Azevedo et al, 2009: “In 1962,
Holonyak, while with General Electric’s Solid- State Device
Research Laboratory, made a red emitting GaAsP inorganic LED
[27]. The output was very low (about 0.1 lm/W), corresponding
to an efficiency of 0.05% [27]. Changing materials
(toAlGaAs/GaAs) and incorporating quantum wells, by 1980, the
efficacy of his red LED had grown to 2 lm/W, about the same as
the first filament light bulb invented by Thomas Edison in 1879.
An output of 10 lm/W was achieved in 1990, and a red emitting
light AllnGaP/GaP-based LED reached an output of 100 lm/W in
2000 [27]. In 1993, Nakamura demonstrated InGaN blue LEDs
[28]. By adding additional indium, he then produced green LEDs
and, by adding a layer of yellow phosphor on top of the blue LED,
he was able to produce the first white LED. By 1996, Nichia
developed the first white LED based on a blue monochromatic
light and a YAG down-converter.”
Quote for Organic LEDs: “The next few years should see major
advances in this area, and the availability of a much wider array
of durable materials and processes than currently exist for the
device designer.” (Sheats et al, 1996).

17. Item, 20, Organic Transistors
Note the different material classes and the improvements for each of them
Huanli Dong , Chengliang Wang and Wenping Hu, High Performance Organic Semiconductors for Field-Effect
Transistor, Chemical Commununications, 2010,46, 5211-5222

18. Different Classes of Materials were found for Many Technologies
Technology SubDomain
Technology
Dimensions of Different Classes of Materials
measure
Energy
Transformation
Light intensity
per unit cost
Lighting
LEDs
Luminosity per Group III-V, IV-IV, and II-VI semiconductors
Organic LEDs Watt
Small molecules, polymers, phosphorescent materials
Solar Cells
Energy
storage
Information
Transformation
Living
Organisms
Materials
Candle wax, gas, carbon and tungsten filaments,
semiconductor and organic materials for LEDs
Power output
per unit cost
Silicon, Gallium Arsenide, Cadmium Telluride, Cadmium
Indium Gallium Selenide, Dye-Sensitized, Organic
Batteries
Energy stored
Lead acid, Nickel Cadmium, Nickel Metal Hydride,
per unit volume, Lithium Polymer, Lithium-ion
mass or cost
Capacitors
Carbons, polymers, metal oxides, ruthenium oxide, ionic
liquids
Flywheels
Stone, steel, glass, carbon fibers
2/
Organic
Mobility (cm
Polythiophenes, thiophene oligomers, polymers,
Transistors
Volt-seconds)
hthalocyanines, heteroacenes, tetrathiafulvalenes, perylene
diimides naphthalene diimides, acenes, C60
Biological
U.S. corn output Open pollinated, double cross, single cross, biotech GMO
transformation per area
Load Bearing Strength to
Iron, Steel, Composites, Carbon Fibers
weight ratio
Magnetic
Strength
Steel/Alnico Alloys, Fine particles, Rare earths
Coercivity
Steel/Alnico Alloys, SmCo, PtCo, MaBi, Ferrites,
Couldn’t find different classes for GaAs lasers and for photosensors

19. Outline of Results


Creating materials (and their associated
processes) that better exploit physical
phenomena
Geometrical scaling
◦ Reductions in scale: e.g., integrated circuits (ICs),
magnetic storage, MEMS, bio-electronic ICs
◦ Increases in scale: e.g., larger production equipment,
engines, oil tankers

Some technologies directly experience
improvements while others indirectly experience
them through improvements in “components”
◦ Computers and other electronic systems
◦ Telecommunication systems

20. Geometric Scaling
Relationship between the technology’s
core concepts (Dosi, 1982), physical laws
and dimensions (scale), and effectiveness
 Or as others describe it: the “scale effects
are permanently embedded in the
geometry and the physical nature of the
world in which we live (Lipsey, Carlaw,
and Bekar, 2005)


21. Item 14:
“Intel, which has maintained this pace for decades, uses this golden rule as both a guiding principle
and a springboard for technological advancement, driving the expansion of functions on a chip at a
lower cost per function and lower power per transistor, by shrinking feature sizes while introducing
new materials and transistor structures.” www.intel.com/content/www/us/en/silicon-innovations/moores-law-technology.html)

22. (Item 26)
http://www.genome.gov/sequencingcosts/

23. Reductions in Scale: DNA Sequencing

Importance of scale can be seen by reading
highly cited papers such as “Genome
sequencing in micro-fabricated high-density
pico-liter reactors” (Margulies, 2005) and
“Toward nano-scale genome sequencing”
(Ryan et al, 2007)
◦ “The ability to construct nano-scale structures and
perform measurements using novel nano-scale
effects has provided new opportunities to identify
nucleotides directly using physical, and not
chemical, methods.”

In fact, just the titles of these papers are
fairly suggestive.

24. Outline of Results


Creating materials (and their associated
processes) that better exploit physical
phenomena
Geometrical scaling
◦ Reductions in scale: e.g., integrated circuits (ICs),
magnetic storage, MEMS, bio-electronic ICs
◦ Increases in scale: e.g., larger production equipment,
engines, oil tankers

Some technologies directly experience
improvements while others indirectly experience
them through improvements in “components”
◦ Computers and other electronic systems
◦ Telecommunication systems

25. Improvements from Increases in “Geometric” Scale (year in parentheses)
Technology
SubTechnology
Dimensions
of Scale
Production
Equipment
Liquid Crystal
Displays
Substrate
Size
Engines
Steam Engine
Horsepower
Marine Engine
Electricity
Increases in Scale
Small
0.17 m2 (1997)
1.4 m2 (2003)
10 (1800)
Large
2.7 m2 (2005)
5.3m2 (2008)
20 (1800)
2.3 (2010)
225 (2010)
Amount of Cost Reduction
Dimension
Equipment*
cost per area
Price per
horsepower
Amount
88%
36%
2/3
74%
Generation
1000s of
Watts
100,000
(1928)
600,000
(1958)
Capital cost
per Watt
59%
Transmission
Voltage
10,000 Volts
(1880)
790,000 Volts
(1965)
Price per
distance
2% per year
or >99.9%
Final cost of
electricity
1000s of
Watts
93
(1892)
1.4 million
(1969)
Price per
kilowatt hour
> 99.9%
38.5
(2010)
40
(2010)
132 (2012)
265
(2010)
170
(2010)
853 (2012)
Capital cost
per ton
59%
40 (2007)
220 (2007)
Transpor
Oil Tankers Capacity in
tation
1000s of
Equipment Freight Vessels
tons
Aircraft
Number of
Passengers
52%
Capital cost
per passenger
14%
Fuel usage per
passenger
48%
Sources (from top to bottom): (Keshner and Arya, 2004; DisplaySearch, 2010; von Tunzelman, 1978; Honda, 2010; Hirsh, 1989;
Koh and Magee, 2008; UNCTD, 2006; Airbus 2012 List Prices; Wikipedia, 2012; Morrel, 2007)

26. Items 7 and 8, Engines
Note scaling on left and pictures of steam engine,
modern day equivalent (steam turbine), and 90,000 HP marine engine
Cost of
cylinder
or piston is
function
of cylinder’s
surface
area (πDH)
Output of
engine
is function of
cylinder’s
volume
2
(πD H/4)
Result: output
rises
faster than
costs as
diameter is
increased

27. Outline of Results


Creating materials (and their associated
processes) that better exploit physical
phenomena
Geometrical scaling
◦ Reductions in scale: e.g., integrated circuits (ICs),
magnetic storage, MEMS, bio-electronic ICs
◦ Increases in scale: e.g., larger production equipment,
engines, oil tankers

Some technologies directly experience
improvements while others indirectly experience
them through improvements in “components”
◦ Computers and other electronic systems
◦ Telecommunication systems

28. Item 16, Computers
Note the similar levels of improvements between 1960 and 2000 (about 7 orders of magnitude)
Source: ICKnowledge, 2009; Koh and Magee, 2006)
As one computer designer argued, by the late
1940s computer designers had recognized that “architectural tricks could not lower the cost of a
basic computer; low cost computing had to wait for low cost logic” (Smith, 1988)

29. Items 18 and 19, MRI and CT

Improvements in MRI and CT were driven by
improvements in computers and they were
driven by improvements in ICs
Quote by Trajtenberg (1990)

Quotes from Kalendar, 2006

◦ “However, it was not until the advent of
microelectronics and powerful mini-computers in the
early seventies, the early seventies, coupled with
significant advances in electro-optics and nuclear
physics, that the revolution in imaging technologies
started in earnest. Computed Tomography scanners
came to epitomize this revolution and set the stage for
subsequent innovations, such as………..and the wonder
of the eighties, Magnetic Resonance Imaging”
◦ “Computed tomography became feasible with the development
of modern computer technology in the 1960s”

30. Item 25, Wireless Transport
Note reductions in feature sizes, which were needed for new cellular systems

31. Discussion/Conclusion

Most observed improvements can be
categorized into two mechanisms:
◦ 1) creating materials (and their processes) to better
exploit their physical phenomena
◦ 2) geometric scaling


Some technologies directly realize
improvements through these two mechanisms
while higher-level “systems” indirectly benefit
from improvements in “components”
Of 33 different technologies and 52 dimensions
of performance, these mechanisms explain
improvements for 31 technologies and 50
dimensions
◦ the exceptions are laboratory concentration of
penicillin and laboratory cooling

32. Summary Statistics
Mechanism
Specific Technologies in Table 1 by Item
Number
Number of
Technologies
Creating Materials
1, 2, 3, 4, 5, 6, 10, 11, 12, 20, 28, 30, 31
Scale Reduction
14, 15, 21, 22, 23, 26
6
Scale Increase
7, 8, 13, 17, 29
4
Component improvement
9, 16, 18, 19, 24, 25, 32
7
Components benefit from
14
9, 32
2
16, 18, 19, 24, 25
5
creating materials
Components benefit from
reductions in scale
Components benefit from
0
increases in scale
Other, Unknown
Total
27 (Penicillin), 33 (Laboratory Cooling)
2
33

33. Summary Statistics

Creating materials

Reductions in scale

Increases in scale
◦ Lighting (1,2,3), GaAs Lasers (4), Photosensors (5),
Solar Cells (6), battery (10), capacitor (11), flywheel
(12), organic transistors (20), crop yields (28b),
magnetic materials (30, 31)
◦ Through components: Electric Motors (9), machine
tools (32)
◦ ICs (14), MEMS (15), magnetic storage (21-22),
optical storage (23), DNA sequencing (26)
◦ Through components: Computers (16), MRI (18), CT
(19), wireline (24), wireless (25)
◦ engines (7, 8), LCDs (17), energy transmission (13),
transport (29)

34. 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
We identified new classes of materials for all
of the “material creation” technologies except
two of them (photosensors, lasers)
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

35. 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 that often involve completely new
forms of equipment and materials
◦ new equipment usually developed and implemented in labs
◦ rapid improvements in many higher-level “systems” were
achieved through improvements in ICs and other
components that benefit from reductions in scale

36. Relationship with Learning (1)

Results provide a deeper understanding of
learning in a technological context than do
current models
◦ they provide new insights into technological
diffusion (Griliches, 1957; Mansfield, 1968) and
productivity growth (Solow, 1956)

The technology diffusion and productivity
growth literatures pay little attention to
improvement rates
◦ but it seems apparent that rapid improvement
rates lead to earlier economic feasibility and
faster rates of diffusion and productivity growth

37. Relationship with Learning (2)
More attention to improvement rates is
required in research on technological
change
 The two mechanisms provide an initial
operational explanation for why some
technologies experience rapid rates of
improvement over long periods of time

◦ that is superior to any explanation that might
come from current theories such as the
learning curve (Wright 1936; Arrow 1962;
Argote and Epple 1990; Ayres 1992)

38. Relationship with Learning (3)
Incremental modification of equipment that
is emphasized by learning curve is one part
of both mechanisms but it is not the most
important part of the mechanisms
 It is in process side of both creating
materials and geometric scaling


39. Relationship with Learning (4)

Nevertheless, incremental modifications of
equipment cannot explain many orders of
magnitude improvements
◦ In fact, learning from production cannot explain
even small improvements in a per mass or
volume basis since such improvements clearly
involve something more basic about the artifact
than just small changes in processes
◦ Our work identifies the creation of new materials
and large reductions in scale as the changes
responsible for rapid improvements and such
learning requires R&D activities and not
necessarily cumulative production

40. 
Thank You
Appendix

41. Item 4 (GaAs Lasers)
Heat sink: heat must be removed
in order to prevent overheating of
laser
Mirror: contaminants in mirror
cause light to be focused on a
spot and thus burn up the mirror
Processes
1) Fewer defects can have large
impact on maximum power because
small reduction in defects can lead to
much higher power
2) Faster processes leads to lower
costs
Source: Martinson R 2007. Industrial markets beckon for high-power
diode lasers, Optics, October: 26-27. Personal Communication with Dr. Aaron Danner

42. Item 5, Photosensor
Note the names of the process and material changes
Source: T. Suzuki, “Challenges of Image-Sensor Development”, ISSCC, 2010

43. Item 6, Solar Cells
Note the different materials for each set of data points
More details on each set of data points can be found in various sources.
For crystalline silicon, see Green M, 2009. The Path to 25% Silicon Solar Cell
Efficiency: History of Silicon Cell Evolution, Progress in Photovoltaics 17: 183-189

44. Item10, battery
Note the names of different materials
Source: Koh and Magee, 2008;
Tarascon, 2009). For more details see Tarascon, J , 2010. Key Challenges in future Libattery research. Philosophical Transactions of the Royal Society 368: 3227-3241

45. Item11, Capacitors.
Note that energy density is a function of capacitance times voltage
squared and the names of different materials
Sources: Koh and Magee, 2008;
Naoi and Simon, 2008)

46. Item12, Flywheels.
Note that energy density is a function of mass times velocity squared and
stronger materials (carbon fiber) enable higher speeds
Sources: Koh and Magee, 2008; Renewable
and Sustainable Energy Reviews 11(2007):
235-258

47. Item 28b, Crop Yields for Corn
Note the different material classes and the improvements for each of them
Source: Troyer, 2006

48. Magnetic Materials (items 30 and 31)

49. Item 15, MEMS for Inkjet Printers
Note the reductions in scale that accompany increases in the number of nozzles
Source: Stasiak et al 2009
Quote for MEMS from (Stasiak et al, 2009): “The development of compact firing chamber architectures enabled smaller ejected drop volumes
and higher nozzle packing densities. The smaller drops required less firing energy per drop for increased frequency and higher throughput.
Furthermore, the smaller drops provided more colors per dot, lighter tones, and photo-quality printing on a wide variety of media.”

50. Magnetic (Items 21 and 22) and Optical (Item 23) Storage Density
Note that increases in density can only be achieved by making storage areas smaller
For more details, see (Daniel et al, 1999; Esener et al, 1999)

51. Item 13. Energy Transmission
Higher voltages lead to lower losses per mile because
losses are a function of surface area (function of radius)
and transmission is a function of volume (function of
radius squared) (AEP, 2008)

52. Item 17, LCDs.
Sources: Television Making: Cracking Up, Economist, Jan 21,
2012, p. 66. (Keshner and Arya 2004; Display Search 2010)

53. Item 29, Ratio of GDP to transport sector,
Aircraft Passenger Times Speed
Aircraft and aircraft engines benefit from
increases in scale
 Other transportation equipment (freight
vessels, oil tankers, trucks) benefit from
increases in scale
 Better computers also have an impact


54. From 1807 tons in 1878
To 500,000 tons in 2009
Oil Tankers

55. From 10 HP (horse power)
in 1817
To 1,300,000 HP today
(1000 MW)
Steam engine
Their modern day
equivalent: steam
turbine

56. Electricity Generating Plants
Edison’s Pearl Street Station
in NY City (1880)
More Recent Plant
From Kilowatts (125 HP engine) to Giga-Watts

57. From DC-1 in 1931
(12 passengers, 180 mph)
To A-380 in 2005
(900* passengers, 560 mph)
*economy only mode
*Economy only mode

58. Relative Price Per Output Falls as Scale Increases
10000
Oil Tanker:
1000s of tons
Smallest was
1807 tons
Steam Engine (in
HP) Maximum scale:
1.3 M HP
Relative
Price per Output
1000
Marine Engine
Largest is
90,000 HP
Commercial aircraft
Smallest one had
12 passengers
100
Aluminum
(1000s of
amps)
Electric Power
Plants (in MW); much
smaller ones built
10
LCD Mfg Equip:
Largest panel size is
16 square meters
1
0.1
1
Chemical Plant:
1000s of tons of ethylene
per year; much smaller plants
built
10
100
1000
Output (Scale)
10000

59. Improvements in Computations Per Second (Koomey et al, 2011)
Why do computers
experience
improvements in
processing
speed?
Are these large (or
small)
improvements in
processing
speed?
How many other
products
experience such
large
improvements?

60. Item 24, Wireline Transport
Figure 2.9 Reductions in Optical Loss of Optical Fiber
Based on personal communication with Dr. Aaron Danner
1000
Optical Loss (db/km)
100
10
1
0.1
0.01
1960
1965
1970
1975
1980
1985
Source: NAS/NRC, 1989.
Source: Koh and Magee, 2006
of lasers and fiber
Source: Fiber-Optic Communication Systems, Govind P.
Agrawal, Institute of Optics, University of Rochester

61. Item 9, Electric Motors: Better materials were needed for stronger magnets
Source: Koh and Magee, 2008

62. Item 32, Machine Tools
Improvements in material strengths led to faster cutting speeds. Note the
materials listed in the right hand figure.
Source: American Machinist, 1977