Singapore, 23 May 2008
Advanced Energy Efficiency
for Process Industries and
Semiconductor Manufacturing
Amory B. Lovins 盧安武, For. Memb. Royal Swedish Acad. of Eng. Sciences
Chairman & Chief Scientist Rocky Mountain Institute
[email protected]
www.rmi.org
Copyright © 2008 Rocky Mountain Institute. All rights reserved. Unlimited free internal .PDF reproduction licensed to participants and sponsors.
Saving energy is cheaper than buying it, so
climate protection is profitable even if you
don’t think it’s necessary
IBM and STMicroelectronics
CO2 emissions –6%/y, fast paybacks
DuPont’s 2000–2010 worldwide goals
Energy use/$ –6%/y, add renewables, cut absolute
greenhouse gas emissions by 65% below 1990 level
By 2006: actually cut GHG 80% below 1990, $3b profit
Dow: cut E/lb 22% 1994–2005, $3.3b profit
BP’s 2010 CO2 goal met 8 y early, $1.6b profit
GE pledged 2005 to boost its eff. 30% by 2012
Interface: 1994–2006 GHG –62% (–9.2%/y)
TI new chip fab: –20% en., –35% water, –30% capex
So while the politicians debate theoretical “costs,”
smart firms are racing to pocket real profits!
2007 Vattenfall/McKinsey supply curve
for abating global greenhouse gases
(technologically very conservative, esp. for transport)
World emissions were 37 GTCO2e in 2000 and rising
27 GtCO2e in 2030 is 46% of base-case emissions
Average cost of whole curve ~ 2/TCO2e (Exec. Sum., p. 5)
www.vattenfall.com/www/ccc/ccc/577730downl/index.jsp
January 2007
Two Asian fab retrofit examples
(by one of RMI’s strategic engineering partners)
Big Asian back-end: 1997 retrofit, mainly HVAC
•
Cut energy use 56% (69%/chip) in 11 months with 14-month
average payback; further projects are saving more
STMicroelectronics’s world-class Singapore fab
’94–97 retrofits saved US$2.2M/y with 0.95-y av. payback
• ’91–97 improvements saved $30M; kWh/150mm std. wafer –
60%—providing 80% of energy capacity for 3.5 expansion;
80% paid back within 18 months
• All retrofits were performed during continuous operation via
cryogenic freeze-plugs and hot-taps (>20 each)
•
This is mainly just harvesting the low-hanging fruit
that already fell down and is mushing up around the
ankles (remember: the tree keeps producing more!)
If the fabs had been properly designed, none of this
would be possible—but they used infectious repetitis
Fab retrofits at STMicroelectronics
RMI analyzed eight fabs during 1998–2000
Found 30–50+% potential retrofit energy savings
Aftertax returns often 100%/y (59%/y in one)
Generally just HVAC—no changes to chipmaking process
Simple paybacks generally <2 y (0–3 y; a very few ~5–7 y)
Never worsened, usually improved, operational parameters
First two implementers cut HVAC energy up to 40% in year 1
Earlier, in 1994–97, STM had cut energy & water
use by 5%/y, or 10%/y per $ value added [“VA”]
End 1998 (1st y working with RMI): energy use
17% below 1994 at equal production value; #12
chipmaker, US$4.3b revenue, US$0.4b profit
1999: energy/$VA –6.5% in 1 y (= 1994 – 26%);
#8 chipmaker, US$5b revenue, US$0.5b profit
Retrofit results from
STMicroelectronics (continued)
2000: energy/$VA –22% in 1 y (= 1994 – 29%);
#6 chipmaker, US$7.8b revenue, US$1.5b profit, of
which saved energy & resources were US$77 million
2001: industry downturn began, but lower costs
helped STM become #3 at the time; VA –15%;
despite tough times, further kWh & water savings in
2001 added US$10 million profit
2002: energy/standard wafer –15% from 2001;
kWh/pin down by half over 6 y (= 1994 – 40%);
water use/wafer ~ 1994 – 50%, water/pin –40%
1994–2001: ~US$60 million total energy savings;
350 more projects identified for 2002–04 to add
>US$11 million/y extra savings; all paybacks <3 y,
av. 2 y
Petrochemical-complex retrofit
Cogen w/ (or don’t make/throttle) 1300# steam
› Turboexpander/cogen >10 MW, US$45M PV
› Replace a boiler with a second CO burner?
› Accept free GM HT fuel-cell offer; burns all HCs
› Vent no steam; absorption chilling (even 50#),
save condensate, save 130# steam 3:1
› Emphasize furnace optimization & innovation
Hi-temp heat distills H2O for boilers/CTs/finfans
Distributed and optimized distillation
Integrate with some neighbouring facilities
Compress air only to pressure required, no letdown
Sensors, graphical data presentation
Hi- CTs, overcool, ?wellwater summer sink
Some of the biggest retrofit
savings are the simplest
Turn off things you’re not using
Run existing cooling towers properly—run all
towers all the time at variable speed
Big slow fans use far less energy than small fast fans
Use free cooling: at STM’s Agrate fab, cost 80%
less to run than chillers, saved US$0.5M/y, –4
MW during >100 winter days/year, 1–3y payback
So all chillers should have variable-speed drive to exploit
seasonal differences in wetbulb temperature
No secondary pumping—primary only
1989 supply curve for saveable US
electricity (vs. 1986 frozen efficiency)
Best 1989 commercially available, retrofittable technologies
Similar S, DK, D, UK…
EPRI found 40–60%
saving 2000 potential
Now conservative:
savings keep getting
bigger and cheaper
faster than they’re
being depleted
Measured technical cost and performance data for
~1,000 technologies (RMI 1986–92, 6 vol, 2,509 pp, 5,135 notes)
–44 to +46˚C with no heating/cooling equipment, less construction cost
2200 m, frost any day, 39 days’
continuous midwinter cloud…yet
28 banana crops with no furnace
Lovins house / RMI HQ,
Snowmass, Colorado, ’84
Saves 99% of space & water
heating energy, 90% of home el.
(372 m2 use ~120 Wav costing
US$5/month @ US$0.07/kWh)
10-month payback in 1983
PG&E ACT2, Davis CA, ’94
Key: integrative
design—multiple
benefits from single
expenditures
Mature-market cost –US$1,800
Present-valued maint. –$1,600
82% design saving from 1992
California Title 24 code
Prof. Soontorn Boonyatikarn
house, Bangkok, Thailand, ’96
84% less a/c capacity, ~90%
less a/c energy, better comfort
No extra construction cost
Old design mentality:
always diminishing returns...
High efficiency doesn’t always raise
even components’ capital cost
Motor Master database shows no correlation
between efficiency and trade price for North
American motors (1,800-rpm TEFC Design B) up
to at least 220 kW
Buying this motor instead of this motor
can cost you >US$20,000 present value
E SOURCE (www.esource.com) Drivepower Technology Atlas, 1999, p. 143, by permission
Same for industrial pumps, most rooftop chillers,
refrigerators, televisions,…
“In God we trust”; all others bring data
= An oil major’s 1/99 min. (EP, which is
inherently ~0,4–1,6 % points below TEFC)
= best U.S. 2000 explosion-proof efficiency
Partial motor
survey in a
typical chip
fab found a
US$1.4M
potential PV
saving just
from using
premiumefficiency
motors to
replace 75
typically
inefficient
motors (2.5
MW)—1/3 of
the plant’s
total motors
New design mentality: expanding returns,
“tunneling through the cost barrier”
New design mentality: expanding returns,
“tunneling through the cost barrier”
“Tunnel” straight to the
superefficient lower-cost
destination rather than
taking the long way
around
Examples from RMI’s industrial
practice (>$30b of facilities)
Retrofit eight chip fabs, save 30–50+% of HVAC energy, ~2-y paybacks
Retrofit very efficient oil refinery, save 42%, ~3-y payback
Retrofit North Sea oil platform, save 50% el., get the rest from waste
Retrofit huge LNG plant, 40% energy savings; ~60% new, cost less
Retrofit giant platinum mine, 43% energy savings, 2–3-y payback
Redesign $5b gas-to-liquids plant, save >50% energy and 20% capex
Redesign next new chip fab, eliminate chillers, save 2/3 el. & 1/2 capex
Redesign new data center, save 89%, cut capex & time, improve uptime
Redesign new mine, save 100% of fossil fuel (it’s powered by gravity)
Redesign supermarket, save 70–90%, better sales, ?lower capex
Redesign new chemical plant, save ~3/4 of auxiliary el., –10% capex
Redesign cellulosic ethanol plant, –50% steam, –60% el, –30% capex
Retrofits save ~30–60% w/2–3-y payback; new ~40–90% w/less capex
“Tunneling through the cost barrier” now observed in 29 sectors
None of this would be possible if original designs had been good
Needs engineering pedadogy/practice reforms; see www.10xE.org
(RMI’s plot for the nonviolent overthrow of bad engineering)
Two ways to tunnel through
the cost barrier
1. Multiple benefits from single expenditures
Save energy and capital costs…10 benefits from
superwindows, 18 from efficient motors &
lighting ballasts,...
Throughout the design: e.g., RMI HQ’s arch has
12 functions, one cost
2. Piggyback on retrofits
A 19,000-m2 Chicago office could save 3/4 of
energy at same cost as normal 20-y renovation
— and greatly improve human performance
Cost can be negative even for
retrofits of big buildings
19,000-m2, 20-year-old curtainwall office near
Chicago (hot & humid summer, very cold winter)
Dark-glass window units’ edge-seals were failing
Replace not with similar but with superwindows
Let in nearly 6 more light, 0.9 as much unwanted heat, reduce
heat loss and noise by 3–4, cost US$8.4/m2glass more
Add deep daylighting, plus very efficient lights (3 W/
m2) and office eqt (2 W/m2); peak cooling –76%
Replace big old cooling system with a new one 4
smaller, 3.8 more efficient, US$0.2 million cheaper
That capital saving pays for all the extra costs
75% energy saving—cheaper than usual renovation
“People who seem
to have had a
new idea have
often just stopped
having an old
idea”
The Nine Dots Problem
The Nine Dots Problem
origami solution
geographer’s
solution
mechanical
engineer’s
solution
statistician's
solution
wide line
solution
New design mentality
• Redesigning a
standard
(supposedly
optimized)
industrial pumping
loop cut power from
70.8 to 5.3 kW (–
92%), cost less to
build, and worked
better
Just two changes
in design mentality
New design mentality,
an example
1. Big pipes, small pumps (not the opposite)
No new technologies, just two
design changes
2. Lay out the pipes first, then the
equipment (not the reverse)
No new technologies, just two
design changes
Fat, short, straight pipes —
not thin, long, crooked pipes!
Benefits counted
92% less pumping energy (12 reduction)
Lower capital cost
“Bonus” benefit also captured
70 kW lower heat loss from pipes
Additional benefits not counted
Less space, weight, and noise
Clean layout for easy maintenance access
But needs little maintenance—more reliable
Longer equipment life
Count these and save…~98%?
This case is archetypical
Most technical systems are designed to
optimize isolated components for single
benefits
Designing them instead to optimize the
whole system for multiple benefits typically
yields ~3–10x energy/ resource savings,
and usually costs less to build, yet improves
performance
We need a pedagogic casebook of diverse
examples…for the nonviolent overthrow of
bad engineering (RMI’s 10XE (“Factor Ten
Engineering” project—partners welcome)
Why focus on pumping examples?
Pumping is the world’s biggest use of motors
Motors use 3/5 of all electricity
A big motor running constantly uses its
capital cost in electricity every few weeks
RMI (1989) and EPRI (1990) found ~1/2 of
typical industrial motor-system energy could
be saved by retrofits costing <US$0.005
(1986 $) per saved kWh—a ~16-month
payback at a US$0.05/kWh tariff. Why so
cheap? Buy 7 savings, get 28 more for free!
Downstream savings are often bigger and
cheaper—so minimize flow and friction first
Typical areas for big savings
Thermal integration
Power systems
Designing friction out of fluid-handling systems
Water/energy integration
Superefficient and heat-driven refrigeration
Superefficient drivesystems
Advanced controls
Let’s look at one example: pumping systems
Designing for efficiency
Task elimination before task: why do it?
Eliminate muda: making defective or unwanted product,
anything not requested, mistakes requiring rectification,
unnecessary inputs or process steps, waiting for something to happen, moving things without purpose,…
Demand before supply
Downstream before upstream
Application before equipment
People before hardware
Passive before active
Quality before quantity
Designing for breakthrough industrial
energy efficiency: the eightfold way
1. Business vision, model, strategy,
& culture first: why do it?
2. Task elimination before task
•
Eliminate muda: making defective or
unwanted product, anything not requested, mistakes requiring rectification, unnecessary inputs or process steps, waiting for something to
happen, moving things without
purpose,…
3. Demand before supply
4. Downstream before upstream
5. Application before equipment
This approach makes
it possible to:
Capture multiple benefits
Make them compound
Free up the most capacity
Avoid the most capex
Eliminate the most waste & harm
Make the most profit
6. People before hardware
Do the most good
7. Passive before active
Have the most fun
8. Quality before quantity
But whole-system designers must think in
the opposite direction to the process flow
Save capex, not just opex,
by making equipment
unnecessary, smaller, or
simpler
Consider the whole system
all together
Optimize it for multiple
benefits
Reduce waste:
Design for whole-system
performance, not sub-system
performance!
Can wastes be reduced or eliminated
—designed out?
Can wastes be recycled as inputs?
Can wastes be made into other
products?
Capacity used to make
waste can now make value
instead—winning more
capacity at zero capex:
Debottlenecking
Throughput gains
So how do we do this magic?
“Like Chinese cooking. Use everything.
Eat the feet.”
— LEE Eng Lock, Singapore
efficiency engineer
Chinese food is world-famous for using every
part and wasting nothing. Why not do
everything else that way too?
Compounding losses…or savings…so start
saving at the downstream end to make
upstream equipment smaller and cheaper
So each unit of avoided flow or friction at the pipe saves ten
units of fuel at the thermal power station
First seek to eliminate part or all of the
flow: zero flow uses zero resources
LNG plant (–161˚C) in a +54˚C desert
Each 1 C˚ by which the site is cooled by raising albedo
(white sand, crushed shells, etc. instead of grey concrete
and black asphalt) saves A$106 million (in present value)
via lower chiller load & cooler air
Sun-rejecting pavings may save 10–20 C˚ = ~A$1–2b
Further potential with better pipe sheathing (what gets
hotter than black?)
Ice-cream plant, best-in-class equipment
Insulated box contains pipes to freeze the cream
The same box also contains the compressors and motors!
Taking them out of the box uses fewer kWh to freeze the
same flow of cream
Real-time flow optimization in cube-law
machines—no choked flow
Less cooling needed because less heat released
Right-sizing is critical
Fab designers typically assume that tools will use
~2–5 more energy than they actually use
Typical tool duty ~0.3–0.4; load diversity is ignored
Phantom loads mean hundreds of extra tons ($2k/t
capital cost) and incur HVAC part-load penalties
Inflated loads mean deep coils, big pressure drops,
oversized fans (heating air as much as tools do!)
Bigger fans, coils, silencers, chillers, towers, pipes,
valves, ducts, motors, electricals, land, foundations,
UPS & losses,…HENCE CAPITAL COST
More filters, resins, O&M, noise, insurance,....
Indirect benefits: retrofit tool’s CRT
display in cleanroom to LCD display
Cost << present-value energy saving ($1k)
LCD lasts longer, doesn’t drift, and is more reliable
LCD is easier to read (less fatigue, fewer errors)
Lightweight, small footprint, less UPS/HVAC sizing
Better laminar flow (no “thermal chimney”)
No static charge or outgas compromising cleanroom
Sealed, no slots with airflow to gather & stir up dust
No implosion, high-voltage, or electromagnetic
interference risks
Indirect benefits: Convert cleanroom
fluorescent lamps to light-pipe feed
Severalfold heat reduction, worth ~US$8–9/W
No disturbance to laminar flow, no EMI or static
No lamps to replace in cleanroom: less traffic, no
breakage risk, no particle shedding from contacts
No ballasts to fail or outgas
Easy to reconfigure tint or location
Indirect light: same/better visibility @ 5 fewer lux
Delivers attractive light with no flicker or hum
Less fatigue, better visibility and productivity
EXAMPLE
Then minimize friction
optional
vs.
99%
1%
Boolean pipe
layout
99%
hydraulic pipe
layout
High-efficiency pumping / piping retrofit
(Rumsey Engineers, Oakland Museum)
15 “negapumps”
Notice smooth piping design
– 45os and Ys
downsized CW pumps, ~4x energy saving, 15 negapumps
The bottom line: low operating
cost, high performance
Oakland Musem Chiller Retrofit
Annual Cost Savings
120,000
Yearly Electricity Cost [$/year]
COOLING TOWER
100,000
CW PUMPS
CHW PUMPS
80,000
62% ANNUAL COST
REDUCTION
60,000
COOLING TOWER
CHILLER
40,000
CHW PUMPS
CW PUMPS
20,000
CHILLER
-
Before
After
Develop your muda spectacles…
Which of these layouts has less capex & energy use?
Condenser water plant:
traditional design
return from tower
to
chiller
return from tower
to
chiller
return from tower
…or how about this?
return
from
tower
to
chiller
to
chiller
• Less space, weight, friction, energy
• Fewer parts, smaller pumps and
motors, less installation labor
return
from
tower
• Less O&M, higher uptime
Air handling: basic physics
Fan motor kW = cubic meter/s pressure drop (kPa)
fan efficiency motor efficiency
~2 opportunities: fan eff. (0.82, usually vaneaxial), motor system eff. (MotorMaster best, rightsized, high power factor,…—35 improvements), VFDs
Static or static+dynamic pressure yields static or total fanpower. To obtain fan motor hp from cfm (ft3/min) and inches w.g., divide by 6,354
~5–10 (or greater) opportunities:
• Reduce flow: air-change rates (base on actual
health goals and real-time sensors), displacement
• Reduce pressure drop: System design, wring out
friction (e.g. duct layout & sizing), low face velocity
• 60- vs. 50-cm duct saves 60% of fanpower (P d–5.1)
COMBINE ALL OF THESE, then downsize chillers
Comparision of a Typical Lab's Consumption to
an Efficient Lab's Consumption By Category
900,000
Typical Lab
800,000
700,000
Proposed
EPICenter
kWh/year
600,000
500,000
400,000
300,000
200,000
100,000
p.
Eq
b
Bo
ile
r,P
La
,M
ps
um
ui
isc
g
in
ol
Co
V
en
til
Li
at
gh
io
ts
n
0
…saving 62%, at lower capex, without improving lab equipment at all
Wet-chemistry exhaust hoods
Efficient hoods save 70–80%, safer, lower capex
Two different aerodynamic methods
Hoods often account for 50–75% of total wet-chem-lab energy
Use science-based indoor-air-quality standards
Use sensor-based real-time controls
Encourage aqueous systems, supercritical CO2, dry cleaning,…
If we don’t want to breathe it, why make our
neighbors breathe it?
Design out toxicity in the first place!
The right steps in the right
order: space cooling
0. Cool the people, not the building
1. Expand comfort envelope
2. Minimize unwanted heat gains
3. Passive cooling
•
Ventilative, radiative, ground- / H2O-coupling
4. Active nonrefrigerative cooling
•
Evap, desiccant, absorption, hybrids: COP >100
•
Direct/indirect evap + VFD recip in CA: COP 25
5. Superefficient refrigerative cooling: COP 6 (Singapore)
6. Coolth storage and controls
7. Cumulative energy saving: ~90–100%, better
comfort, lower capital cost, better uptime
Superefficient big HVAC
(105+ m2 water-cooled centrifugal, Singapore, turbulent induction air delivery — but
underfloor displacement could save even more energy)
E l ement Std kW /t Best kW /t H ow t o d o it
(CO P)
(COP)
Supply
0.60
0.061
Best vaneaxial , ~0.2–0.7
fan
kPa TSH (less w/UFDV),
VAV
ChW P
0.16
0.018
120–150 kPa head,
efficient pump/motor,
no pri/sec
Chiller
0.75
0.481
0.6–1 Cº approaches,
optimal impeller speed
CW P
0.14
0.018
90 kPa head, efficient
pump/motor
CT
0.10
0.012
Big fill area, big slow
fan at variable speed
TOTAL
1.75
0.59
Better comfort, lower
(COP 2.01) (COP 5.96, capital cost
3 better)
(Best Singapore practice w/dual ChW temp.: 0.52 total including 0.41 chiller kW/t, COP 6.8)
Low-face-velocity, highcoolant-velocity coils...
Correct a 1921
mistake about
how coils work
Flow is laminar
and condensation
is dropwise, so
turn the coil
around sideways,
run at <1 m/s;
29% better
dehumidification,
P –95%; smaller
chiller, fan, and
parasitic loads
Savings begin with measurement
Two 1996 Singapore hard-drive plants had a 54
range of kWh/drive (the high one went bankrupt in
1997)
One chipmaker’s rated chilled-water-plant COPs varied
so widely that the worst fab’s was 42% below the best
fab’s, despite having a less difficult climate
Only one fab’s chilled-water plant was measured; it
averaged 21% below its rated COP. Only actual
measured performance counts, not claims or guesses!
That best COP was >20% below the Singapore state of the art (6.8
or 0.52 kW/t), which costs less to build
The owner lost >US$1M/y by not adopting its own best practices
What is efficiency worth over 20 y?
(US$, $0.05/kWh, 5%/y real discount rate, zero HVAC capex
+ filter opex, nominal 1 kW/t HVAC + 10% parasitics)
1 watt of cleanroom power use and heat release =
US$7 opex…or US$8–9 including filters
1 L/s (2 cfm) cleanroom exhaust = US$132
Fan towers (humid climate): 25 Pa (0.1"w.g.) P =
US$230,000
"
250 Pa (1"w.g.) makeup/exh. P = US$2.6 per L/s
Each percentage point’s efficiency gain in an 8766 h/y
motor in conditioned space = US$95/kW
TI’s 2006 fab at Richardson, Texas
100,000 m2 (cleanroom 20,500 m2); all data courtesy of Paul Westbrook at TI
(see his Fabscape paper 26 Oct 2004)
Construction started 18 Nov 2004, after 3-day RMI design workshop Dec 2003;
completed spring 2006; awaiting tools. Why was it built in Texas, not in Asia?
Big HONKIN’ ideas (JD Bryant)
Holy Cow
Over the top
No Nonsense
Knock you out
I don't know why I didn’t do this before
Now because it will save me a *%$&^#+@
of money and time
3D - Data Driven Direction
Used Sematech
Data and TI Fab
Data
Wafer Fab Electrical Power Consumption
(Sematech Data - 14 fab average)
Process Tools Breakdown
UPS Controller
4%
RF Gen
6%
Non-Process
Pumps
9%
Remote
Plasma Clean
3%
Mini Environ
1%
Process
Pumps
52%
Misc
12%
Heaters
13%
Facilities Systems Breakdown
Process
Tools
40.7%
Facilities
Systems
59.3%
MU Air
4.9%
CDA
4.4%
PC Water Pump
2.9%
Exhaust
6.9%
Chillers
42.0%
UP Water
8.2%
N2 Plant
12.2%
Recirc Fans
18.5%
Design based on measurements
Vacuum pumps (21% of total electrical load)
TI/Sematech/vendors idle-signal protocol (tools use about as
much energy idling—nearly all saveable—as processing wafers!)
New vacuum pumps save much PC H2O (~30% higher efficiency
+ idle signal saves 300 tons of cooling), N2, ~7% of total el.
Exhaust
TI recovers some general-exhaust heat and works with tool
suppliers to optimize for key thermal constraints; better design
saved ~50 m3/s of exhaust (& makeup)
PC water
Design for small pressure drop and close-approach heat
exchangers reduced system flow by 20% (190 L/s)
Central utilities plant (21% of
fab load)
Split plant: 25% @ 4.4˚C for dehumidification (0.44–
0.51 kW/ton), 75% @ 12.2˚C for all other loads (0.32–
0.50 kW/ton)
12˚C chiller (steadier load) has heat recovery; build 1
boiler + 1 backup, not 6; mainly off; with high-P spray
(not steam) humidification, NOx emissions –60%
Variable-speed primary distribution; efficient pipes,
pumps, variable-speed motors (& fans)
One 4˚C spare chiller provides redundancy at both
temperatures via blending
Makeup air
Run-around coils for free reheat
Low-face-velocity coils (2 m/s) for small fans
High-pressure humidification, no steam boilers
(this + eliminated heating boilers cut NOx 60%)
Investigating enthalpy-wheel recovery (>70%
recapture of exhaust enthalpy)
Testing desiccant-wheel MUA option to eliminate
entire 4.4˚C chiller plant
Recirculated air (10% of fab load)
Take full credit for mini-environments: specify
Class 100 turbulent (ISO Class 5)
Reducing HEPA coverage from 50% to 25–30%
(FFU) eliminated 300 tons of cooling
Filter life rises as 1/velocity2, so 29% at 0.35 m/s
HEPA coverage, not 23% at 0.44 m/s, pays for
extra FFUs in 6 y—a 13% return on investment
TI is testing different smocks for cooler workers &
warmer rooms; less particle concern because
wafers are in front-opening unified pods (FOUPs)
Water efficiency too
DI water (>60% of total water use): using RO
reject & some recycling cut DI input 20%
CT evaporation/blowdown (20%) cut 50% by
using first-stage RO brine water
Scrubbers (10%) replace raw water purchase
with relatively pure “industrial waste”
Total reclamation saves nearly 4 ML/d of input
Waterless urinals (–2.3 ML/y), 8,000 m3
rainwater retention pond, native plants
Administrative building (and,
often, fab too)
Passive solar (E–W) orientation, exterior shading
Energy & daylight models: 30˚ rotation = –US
$30k/y
Lightshelf daylighting, dimming efficient el. lights
Optimized glazing (high visible light transmittance
and insulation, but rejects infrared rays)
Roof: high solar reflectivity and infrared emissivity
Demand-controlled ventilation (CO2 sensors)
Expected results vs. TI’s previous
best design
–20% energy, –35% water
–30% total capex/m2—cheaper than Chinese fab
Next fab can save even more and cost even less
(and it did so when TI recently designed it)
Better-optimized tool design—already drove half TI’s savings
Use heat-driven desiccant to eliminate low-temperature chiller
Onsite trigeneration (electricity, process & space heat, cooling)
Overall TI project economics
LEED (Leadership in Energy and Environmental
Design): Silver fab (a first), Silver/Gold admin. bldg.
LEED-related items cost US$2–3 million, mainly
efficiency that TI would have bought anyway
US$750,000 operating savings expected in first year
at old energy prices (which then doubled)
At full build-out, >US$3 million/y saved operating
costs (or twice that at today’s energy price)
Efficiency’s net extra capex: <1%, probably negative
An example of what’s next: fuel cells
Ultrareliable onsite power; no UPS capex or losses
Free process and space cooling and heating
Free ultrapure hot water (very valuable)
Onsite H2 production replaces tube-truck shipments
Even retrofitting today’s costly (2–3 US$/W) fuel
cells in a fab can be justified if properly sited &
used to capture “distributed
benefits” (www.smallisprofitable.org)
Four Times
Square, NYC
(Condé Nast
Building)
• 148,000 m2, 47 storeys
• non-toxic, low-energy materials
• 40% energy savings/m2 despite
doubled ventilation rates
• Gas absorption chillers
• Fuel cells
• Integral PV in spandrels on
S & W elevations
• Ultrareliable power helped recruit
premium tenants at premium rents
• Fiber-optic signage (signage
required at lower floor(s))
• Experiment in Performance Based
Fees rewarding savings, not costs
• Market average construction cost
Bundling PVs with end-use
efficiency: a recent example
Santa Rita Jail, Alameda
County, California
PowerLight 1.18 MWp project,
1.46 GWh/y, 1.25 ha of PVs
Integrated with Cool Roof and
ESCO efficiency retrofit (lighting, HVAC, controls, 1 GWh/y)
Energy management optimizes
use of PV output
Dramatic (~0.7 MWp) load cut
Gross project cost US$9M
State incentives US$5M
Gross savings US$15M, 25y PV
IRR >10%/y (Cty. hurdle rate)
Works for PVs, so should work
better for anything cheaper
Saving 1–2% of total costs matters
Saved energy costs, like any saved overhead, drop
straight to the bottom line
Basic energy efficiency retrofits can often add one
percentage point to total net profit
If new chip sales earn (say) 10% profit, then saving $1 worth of energy increases profits by the
same as $10 of new sales—harder and less certain (especially nowadays) than saving energy!
If you’re short of capital, don’t waste it on
oversized and overcomplex utility plant
STMicroelectronics’ CO2 goals
RMI showed how to cut CO2/chip by ~92%
profitably in 1999, ~98% profitably by 2010
STM adopted CO2–90% goal 1990–2000, 0 by 2010
(despite projected chip output 40 1990 level)
2010 supply goal: 65% fuel cells & cogeneration,
5% renewables, so CO2/$VA < 20% of 1990 level
STM expects 1994–2010 CO2 reduction >10 MT +
US$0.9b; also PFC reduction 10 1995–2008—
equivalent to >10 MT CO2 by 2010
Cut CO2/chip by 10–100
...at a profit!
0.44 from 200- to 300-mm shift if same yield
0.3 from state-of-the-art fab efficiency*
0.4 from onsite trigeneration (net of reformer loss)
0.94 from fuel-cell elimination of UPS losses
0.5 fueling with gas, not coal (less carbon/J)
0.5 switching energy supply to 50% renewables
These six steps cut CO2 per chip by ~99%**
So if output rises 30 (40%/y for 10 y or 18%/y for
20 y), and you fuel your growth this way, total CO2
drops by nearly 3, so you could sell carbon permits
Almost all steps are profitable now, the rest soon
All can also bring big operational benefits
*STMicroelectronics has published a path to 0.33, reducing a 1997 15-MW fab to 5 MW.
**STMicroelectronics published 5/98 a realistic path to a 92% reduction. They and I ignore
upstream options, e.g., 4–5 Czochralski savings.
Obstacles to resource-efficient fabs
Very risk-averse culture
Why innovate only inside the cleanroom, and not also in
how the utility functions are provided?
Is any fab in the world even optimized for its climate?
Organizational issues
Barriers and tribal behavior between process & utility staffs
Nobody owns losses or gets rewarded for savings
Schedule: there’s never a good time to design for efficiency
Key data are seldom measured or displayed
Only a few fabs in the world accurately measure ChW kW/t
Information is cheap, powerful,
but viscous
One factory saved US$30,000 the first year by…
labeling the light switches
A hard-drive factory saved a great deal of money
by properly labeling the red/green-zone “idiot
gauge” showing pressure drop in its big filter banks
“Cents per drive” and “Million $ profit per year” (nonlinear)
Innumerable facilities have saved untold energy and
maintenance costs by measuring
But many more use poor or uncalibrated sensors
Few plants are designed to measure what’s needed
And very few present key efficiency metrics to the
operator, real-time, in effective graphics
Benefits of monitoring
with good graphic display
Chiller Efficiency (kW/ton)
Finding 2 - The second and third chillers are running
before they are needed, due to a control problem.
Finding 3 - The maximum load is never
above 1500 tons. A fourth chiller called for
in the plant expansion is not required,
saving approximately $1,000,000
Finding 1 - Chillers are
always operating at less
efficient than manufacturer’s
specifications
Chiller Load (tons)
Courtesy Rumsey Engineers
Getting the value you want
Specify the physical performance you want
Reward the savings you get and measure
Reward designers for savings, not expenditures
Reward especially the toolmakers for system
value—if you don’t ask for high efficiency and tell
them what it’s worth, you won’t get it
Will the next fab save 50% of energy? 80%?
How much less will it cost? Let’s find out!
No limits to profitable industrial energy
efficiency for a very long time to come
Industry is a materials-processing activity, ~99.98% of the
materials are wasted, and most of this waste will ultimately be
turned into profit by dematerialization, virtualization, product
longevity, closed loops, industrial ecology, desktop mfg., etc.
Conventional technological innovation continues apace despite
appalling private and public underinvestment in energy RD&D
Important new classes of processes, like microfluidics
End-use efficiency keeps getting bigger and cheaper, esp. w/
integrative engineering to “tunnel through the cost barrier”
Next come two further design revolutions
Biomimicry: innovation inspired by nature (Janine Benyus)
Perhaps nanotechnology (in Eric Drexler’s original sense)
› Caution: nanomaterials look risky, and biomimicry is not biotechnology (often unwise): over time, Darwin always beats Descartes
Plus the options we haven’t yet thought of—but could live to
do so…if we quickly get the hang of responsibly combining a
large forebrain with opposable thumbs!
We are the people we have been waiting for
Companies that capture these opportunities for
elegant frugality will flourish
Those that don’t won’t be a problem—because
after a while, they won’t be around
“Only puny secrets require protection. Big
discoveries are protected by public incredulity.”
—Marshall McLuhan
www.rmi.org
Management recommendations (1)
Establish a serious corporate energy efficiency
program: site champions, coaches, accountability,
aligned incentives, continuous improvement
Promote necessary corporate cultural changes,
including curiosity and managed risk-taking
See facilities not as overhead to minimize but as a
profit center to optimize by mining valuable waste
Charge processes the shadow cost of services used
Review capital allocation rules top-to-bottom so the
financial and operating people share the same goal
Management recommendations (2)
Measure, visualize, and communicate the data
Convert efficiency metrics into money metrics
Require whole-system design
Set minimum performance standards; reward
better
For example, a new SE Asian plant should produce 5.5˚C chilled
water on the design day at not over 0.54 kW/t: 0.48 chiller* +
0.026 chilled water pump + 0.021 condenser water pump +
0.010 cooling towers. Why settle for worse and costlier?
*ST’s retrofitted AMK fab averages 0.44 chiller kW/t (half of typical, approaching
1/4 of some), producing 15˚C water at 0.38 kW/t and 5.5˚C water at 0.58 kW/t. A
new dual-temperature (15/6.7˚C) Singapore design can get 0.43 kW/t chiller, 0.52
whole-system
Management recommendations (3)
Technology and design are dynamic. Never stop
learning. If you’ve just retrofitted, retrofit again.
Remember the fecundity of the tree that keeps
growing more low-hanging fruit.
Traditional designers claim this approach doesn’t work,
or they already do it. Both can’t be true. If the latter
is, their designs’ technical efficiency should compare
favorably with the best in the world. Does it?
Demand and incentivize advanced efficiency from
vendors and contractors: reward measured savings,
not expenditures. Different outcomes require different
actions.