Industrial Energy Efficiency Accelerator - Guide to the microelectronics sector The UK microelectronics sector uses more than 1,300 GWh of energy each year representing nearly 710,000 tonnes in carbon dioxide emissions. The energy consumed is required to create the clean working environment (40%), process tools plant (40%) and cleanroom utilities, such as process gas and ultra pure water (20%). While energy reduction of over 69% has been achieved over the last 10 years, there are opportunities to reduce these further. Executive Summary The UK microelectronics sector uses more than 1,300 GWh of energy each year representing nearly 710,000 tonnes in carbon dioxide emissions. The three largest organisations account for approximately 25% of the total energy consumption. There are 26 significant regional manufacturers that make up the UK industry. According to a 2005 study by the International SEMATECH Manufacturing Initiative (ISMI) the global semiconductor industry could save $500 million/year (industry wide) in energy costs through modest improvements to tools and facilities 1 . The UK contributed approximately 15% to this global figure 2 . Between 2006 and 2008 the UK microelectronics sector achieved a reduction of 500,000MWh/year in energy use. This is equivalent to £2.65m reduction in the Climate Change Agreement levy 3 . The key productive component that is examined in this report is the wafer fabrication process used to create semiconductors. This process relies heavily on an electrical energy intensive controlled environment and the use of a complex manufacturing process. Both the fabrication and environmental components are supported by a vast array of utilities. The energy consumed in creating the clean working environment accounts for approximately 40% of the total consumption of a manufacturing plant. The “black box” of process tools plant also accounts for approximately 40% of the total consumption. Within the tools environment the largest energy consumers are those associated with pumps and furnaces; with equipment that often sitting in “idle” mode for large periods of time. The remaining 20% is associated with cleanroom utilities i.e. process gas and ultra pure water etc. 1 International SEMATECH (ISMI) Sematech News, 2005 2 “UK semiconductor design evolves and grows stronger”, NMI, August 2006 3 „Sector and process overview”, by NMI on behalf of Carbon Trust, sourced from „Technical and Project Management consultancy, Scope of Work”, 2010.
65
Embed
Microelectronics Industrial Energy Efficiency (CTG063) - Carbon Trust
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Industrial Energy Efficiency Accelerator - Guide to the microelectronics sector The UK microelectronics sector uses more than 1,300 GWh of energy each year representing nearly 710,000 tonnes in carbon dioxide emissions. The energy consumed is required to create the clean working environment (40%), process tools plant (40%) and cleanroom utilities, such as process gas and ultra pure water (20%). While energy reduction of over 69% has been achieved over the last 10 years, there are opportunities to reduce these further.
Executive Summary The UK microelectronics sector uses more than 1,300 GWh of energy each year representing nearly 710,000
tonnes in carbon dioxide emissions. The three largest organisations account for approximately 25% of the total
energy consumption. There are 26 significant regional manufacturers that make up the UK industry.
According to a 2005 study by the International SEMATECH Manufacturing Initiative (ISMI) the global
semiconductor industry could save $500 million/year (industry wide) in energy costs through modest
improvements to tools and facilities1 . The UK contributed approximately 15% to this global figure
2 . Between
2006 and 2008 the UK microelectronics sector achieved a reduction of 500,000MWh/year in energy use. This is
equivalent to £2.65m reduction in the Climate Change Agreement levy3 .
The key productive component that is examined in this report is the wafer fabrication process used to create
semiconductors. This process relies heavily on an electrical energy intensive controlled environment and the use
of a complex manufacturing process. Both the fabrication and environmental components are supported by a vast
array of utilities.
The energy consumed in creating the clean working environment accounts for approximately 40% of the total
consumption of a manufacturing plant.
The “black box” of process tools plant also accounts for approximately 40% of the total consumption. Within the
tools environment the largest energy consumers are those associated with pumps and furnaces; with equipment
that often sitting in “idle” mode for large periods of time.
The remaining 20% is associated with cleanroom utilities i.e. process gas and ultra pure water etc.
1 International SEMATECH (ISMI) Sematech News, 2005
2 “UK semiconductor design evolves and grows stronger”, NMI, August 2006
3 „Sector and process overview”, by NMI on behalf of Carbon Trust, sourced from „Technical and Project Management
consultancy, Scope of Work”, 2010.
Microelectronics Sector Guide 2
During the course of this report we have observed that the Industry demonstrates a high level of activity in the
implementation of energy saving measures. An energy reduction of 69.3% has been achieved in the sector over
the last 10 years4 . Such measures are keenly supported by the trade association, the National Microelectronics
Institute, who believes a further 15% can be found through innovative solutions. Good progression has already
been made within the cleanroom environment particularly around the HVAC systems and their associated plant
and equipment.
The report has considered the process in detail to identify what opportunities also might be available. The
opportunities have been divided between good practice measures and Innovations. Both face common business
barriers in the form of proven technology and capital expenditure. These opportunities are summarised as:
Annual Gas kWh 4,714,000 unknown unknown 15,500,000
These sites represent approximately 12% of the UK sector‟s energy use. The sites are typically components of a
larger, multi-facility, corporation and the data above represents the UK fabrication facility for each organisation.
The sites are typically linked to a sister site located somewhere else in the World, most notably Asia and US.
Distribution of the end product is either direct to industry clients or to the sister site for further fabrication.
These sites manufactured a variety of products and were involved in different stages of production from wafer
manufacture through to chip manufacturing and packaging.
The products included:
Wafer substrate Discrete Devices
Integrated Circuits Hard drives
Bipolar Transistors MOSFET
Diodes
Microelectronics Sector Guide 13
The sites also manufactured a variety of different wafer sizes including:
4” (declining) 6” (UK industry standard)
5” (declining) 8” (future trend in UK)
The majority of the sites were operating 24 hours a day, in shift patterns, and between 5- 7 days per week. The
operating periods are reflective of a high product demand that is currently prevalent in the industry. Most sites
noted that production has been increasing over the last few years through an increase in demand but that the
demand can be cyclic.
2.8 Semiconductor manufacturing process
Semiconductor device fabrication is the process used to manufacture integrated circuits. The process is a
complex sequence of photographic and chemical processing steps whereby electronic circuits are gradually
created on a wafer made of pure semiconducting material. From the raw material to completion can involve up to
500 steps and can take weeks to complete.
Figure 4: Process Overview
The overall process is illustrated within the diagram above and described in more detail within the sections and
diagrams below.
2.8.1 Wafer manufacture
The fabrication process starts with the preparation of wafers. Wafers are typically high purity silicon (99.9999%
pure) grown into mono crystalline cylindrical ingots, known as boules. An ingot is typically 100kg.
Integrated circuits are essentially linear, that is they are formed on the surface of the silicon so as to maximise
the surface area of silicon. The ingots/ boules are sliced into thin discs or wafers using a diamond saw or wire
saw.
The thickness of the wafers is a function of their diameter. The main criterion that determines their thickness is
the requirement to be sufficiently robust to ensure flatness across the surface, above all else, to minimise difficult
and expensive planarisation steps. As an example 300mm diameter wafers are typically 0.775 mm thick. The
preparation of wafers from an ingot involves a series of operations that typically take place in a light industrial
environment with the latter stages being carried out in a cleanroom.
The slicing of the ingot is typically carried out using a wire saw or diamond saw. Wire saws cut multiple slices
simultaneously. After slicing, the surfaces of the wafers are lapped using abrasive slurry until the wafers are flat
to within about 2μm. An etching process is carried out to remove crystal damage that may occur during the
Microelectronics Sector Guide 14
lapping process. Finally after etching the wafers are polished using a chemical / mechanical process that
smoothes the uneven surface left from previous processes and makes the wafer flat and smooth enough to
support optical photolithography.
2.8.2 Insulating
In order to protect the silicon substrate and to form transistor gates, a thin layer of silicon dioxide (SiO2) is formed
over the silicon wafer. This is typically done by exposing the wafer to high temperatures (900-1200°C) in a
furnace.
2.8.3 Placing
This is the first step in the processing of a raw wafer into semiconductor device. It involves the growth of a high
purity and defect free monocrystalline layer onto the surface of the substrate. This process is carried out in a
diffusion furnace heated to around 1,100°C and takes between 4 and 14 hours. The high temperature required
within the furnace and the duration of the process means that this is most energy intensive process of the various
activities carried out in the device fabrication process. This process is also known as epitaxy.
2.8.4 Patterning
Following this placing step the wafer is then ready for photolithography. The wafer is coated with a layer of photo
resist using a spin coating process. The photo resist-coated wafer is then baked for a short time (30 to 60
seconds) to drive off excess photo resist solvent, typically at 90 - 100 °C.
After baking, a mask is applied to the wafer and the photo resist is exposed to a pattern of intense light. Optical
lithography typically uses ultraviolet light. Positive photo resist, the most common type, becomes soluble in the
basic developer when exposed; exposed negative photo resist becomes insoluble in the (organic) developer.
This chemical change allows some of the photo resist to be removed by a special solution, called "developer" by
analogy with photographic developer.
The resulting wafer is then "hard-baked". The hard bake solidifies the remaining photo resist, to make a more
durable protecting layer in future ion implantation, wet chemical etching, or plasma etching.
2.8.5 Removing
In etching, a liquid ("wet") or plasma ("dry") chemical agent removes the uppermost layer of the substrate in the
areas that are not protected by photo resist. In semiconductor fabrication, dry etching techniques are generally
used, as they are more accurate and avoid significant undercutting of the photo resist pattern. This is essential
when the width of the features to be defined is similar to or less than the thickness of the material being etched
When a layer of photo resist is no longer needed, it is removed from the substrate. This usually requires a liquid
"resist stripper", which chemically alters the resist so that it no longer adheres to the substrate. Alternatively,
photo resist may be removed by a plasma containing oxygen, which oxidizes it. This process is called ashing,
and resembles dry etching.
2.8.6 Implanting
The electrical properties of selected areas of the developing device are changed by implanting energised ions
(dopants) in the form of specific impurities into the areas not protected by photo-resist or other layers. These
basic steps are repeated for additional layers of silicon, glass and aluminium
Implant is via diffusion or ion implantation. Diffusion is carried out in a furnace with a flow of gas running over the
wafers. The process is not selective so the photo resist and patterning need to be done before this step. Ion
implantation is different from diffusion in that ion implantation shoots the desired dopant ions into the wafer. The
disadvantage of ion implantation is that wafers have to be processed one at a time while a diffusion chamber can
handle many wafers simultaneously.
Microelectronics Sector Guide 15
Implant is followed by an annealing process which repairs the damage caused to the wafers. This process
involves heating to allow the crystal lattice structure to repair itself.
2.8.7 Interconnecting
The finished wafer is an intricate sandwich of n type and p type silicon with insulating layers of glass and silicon
nitride. All other circuit components are constructed during the first few masking steps and the following masking
steps connect the components together. An insulating layer of glass is deposited and a contact mask is used to
define the contact points of each of the circuit elements. After the contact windows are etched the entire wafer is
covered with a thin layer of aluminium. The metal mask is used to define an aluminium layer and therefore
leaving a network of metal connections. The entire wafer is then covered with an insulating layer of glass and
silicon nitride to protect it from contamination during assembly. The final mask and passivation etch removes the
passivation material from the bonding pads which are used to electrically complete the circuit.
2.8.8 Packaging
While still on the wafer, each device is tested and functional and non-functional devices are identified. Non-
functional wafers, i.e. defects, are then re-circulated back into the process to rectify the defects. The amount of
recirculation impacts on the product “moves” i.e. throughput, at the facility which is a key performance indicator of
a Fab plant.
Following a satisfactory test the wafer is then ready for cutting into individual chips and assembly / packaging.
Figure 5 below is a graphical description of the typical stages employed in manufacturing a semiconductor
device.
Microelectronics Sector Guide 16
Figure 5: Typical steps involved in manufacturing a device
Microelectronics Sector Guide 17
2.9 Sector energy use
This section introduces the site wide energy consumption breakdown within a Fab facility.
2.9.1 General
The primary energy consuming activities within the sector fall into three general categories:
Table 4: Energy use in the microelectronics sector
Category Energy Split
Facilities / Cleanrooms 40%
Utilities (compressed air etc.) 20%
Process Tools 40%
Irrespective of product group or energy use in process tools, approximately 60-70% of total energy is used in
facilitating the clean room i.e. maintaining temperature and humidity conditions and providing utilities within the
manufacturing zone. The visits to the host sites and the resulting acquisition of energy data has indicated that the
majority of the energy efficiency savings, which have been implemented to date, have been focussed within the
utilities/facilities areas.
Energy consumed by the process tools has remained relatively constant (proportional to production output).
Energy savings achieved in the Process Tools have not necessarily been recorded as these “shop-floor tweaks”
usually reflect quality and/or throughput improvements with energy savings being very much of secondary
interest. However, due to increasing utilities costs, energy savings are gaining significance in the implementation
of changes to the process.
The following charts illustrate this trend indicating the step increase in the proportion of electrical energy
consumed by the process tools as part of the total site energy consumption.
Figure 6 Chart of electrical load distribution trends in recent years
The final chart is based on the current trend continuing, with process tools energy consumption remaining static
and predicts a 50/50 ratio. The rate of change, in this trend, is likely to decrease as the lower cost/ease of
implementation opportunities are implemented.
The charts above are averaged across a number of enterprises and although it is indicative of the sector the
energy split within a particular site is clearly a function of the balance of activities and the site location. For the
sites visited the split varied from 40% process/60% facilities and utilities to 55% process/45% facilities and
utilities.
Microelectronics Sector Guide 18
The information presented is consistent with UK and international benchmarking. For example, according to the
World Semiconductor Council (WSC), the energy needed to run process equipment and tools accounts for up to
40-50% of the total energy consumption in a semiconductor facility.
It should be noted that this data will vary across the globe due to the climatic influences on Fab humidification
levels. For example, a Fab in the UK will only see a high humidification demand for a short period of time in
summer months whereas a Fab in Asia will have a high, all-year round humidification requirement. The figures
quoted are a generic average across worldwide semiconductor manufacturers.
2.9.2 UK energy use
The following table provides an indication of the estimated consumption associated with the UK Microelectronics
industry based upon 26 representative sites9:
Table 5: UK Fab energy consumption
Sites Annual Consumption
15 <50,000 MWh
5 50-100,000 MWh
3 100-150,000 MWh
3 >150,000 MWh
The host sites that were visited were all less than 100,000MWh. For purposes of indicative calculations it was
assumed a typical site consumes 50,000MWh/year.
These figures were converted into the following total consumption figures per fuel source (based upon data
recorded in 2008).
Table 6: Sector energy consumption
Sector Energy Consumption
Total Energy Use 1,916,695,322 kWh
Electricity 1,731,885,126 kWh (90%)
Gas 173,196,325 kWh (9%)
Gas Oil 11,613,871 kWh (1%)
The NMI has previously carried out a study of the electrical energy consumption across a range of sites and their
results are presented in Figure 7.
These results are consistent with the trends indicated in Figure 6 with 37% of the annual electrical energy
consumption coming from the process tools and the remaining 63% from utilities and facilities. The utilities and
facilities fraction is broken down into its main constituents and indicates that the Air Handling Units (AHU) and
Chillers typically consume the greatest proportion, approx. 66%, of the total utilities and facilities electrical energy
consumption.
9 “Technical and Project Management Consultancy: Scope of Work”, Annex D, Carbon Trust
Microelectronics Sector Guide 19
Figure 7 NMI Sector Average Breakdown of Annual Electricity Use
Based upon the breakdowns shown in Table 6 and Figure 6 and using the earlier consumption data as a
reference, we have estimated the typical consumption and emission rates for the primary services at an average
Fab site as shown in Table 7 using a grid electricity conversation factor of 0.545kgCO2e/kWh10
. This study
focuses on grid electricity usage in the sector only as this was considered the most influential in terms of typical
energy use. The proportions in values of the Climate Change Agreement levy reduction available for each fuel
type means electricity usage reduction will generate the highest overall savings.
Boilers for heating also represent a significant proportion of the total energy consumed by a typical site but gas
usage is not represented in Figure 7. Gas accounts for approximately 9% of the sector‟s energy use. Whilst
significant, the opportunities for the reduction in this specific fuel are reliant on known practices such as efficiency
in boilers and insulation which are thought to have already been implemented where possible. Gas is also has a
lower embodied carbon figure compared to grid electricity. Therefore this was not considered in this study.
Table 7: Plant equipment electricity use
Primary Services Proportion of Site
Load Distribution
MWh/annum tCO2/annum
Air Handling Plant and
Extraction
23% 12,000 6,000
Cooling Plant 20% 10,000 5,000
Process Tools 37% 19,000 10,000
Others 20% 9,000 5,000
Totals/average site: - 50,000 27,000
UK Industry Total: - 1,300,000 710,000
10
Carbon Trust
Microelectronics Sector Guide 20
The following charts, in Figure 7, provide a pictorial breakdown of the energy consumption recorded at four of the
host sites from their own metering data (data was not collected from the fifth site). The data is historical based on
energy consumption typically recorded within the last 5 years. The breakdown may not reflect current
consumption at each site but is generally indicative.
Figure 8: Host Site Energy Consumption Overview
It should be noted that the host sites operate dissimilar processes and in this regard the above charts should not
be used for comparative industry benchmarks. However, they are useful in verifying expected trends and
clarifying the main areas of energy consumption.
Microelectronics Sector Guide 21
3 Process energy use
The following section reviews energy consumption across the processes i.e. process tools, within a Fab facility.
3.1 General
The following graphic indicates the various processes, which were described in Section 2, and the external input
that is required to complete the process. The external input can be in the form of water (or other chemical),
electrical input power to energise equipment used in the process (i.e. pumps etc.) and gases (i.e. nitrogen, clean
dry air etc.), that maybe used as an agent in the completion of the process.
Figure 9: Process Energy Inputs
The energy used across the process spectrum can be approximately distributed as follows:
Pumps – 50-60%: a Fab facility can contain up to 600 pumps in the process tooling, depending on output
capacity.
Heaters – 20-30%: associated with heating elements within furnaces
Other - 10-25%: comprises of items such as cleaning machines, automated interconnection machines etc.
Microelectronics Sector Guide 22
Of the process energy used in a Fab, the consumptions have been distributed across the typical processes in
Figure 10 below.
Figure 10: Typical Process Energy Use
The processes implemented by the Fab can be reliant on some or all of the following variables:
Product output quantities;
Product specification;
Product variety;
Customer requirements/contractual agreements; and
Utilities available at the site.
Discussions with host sites have established that, to date, process tools (i.e. machines) energy savings have
been restricted or at least difficult to quantify. Reasons for this are noted as follows:
There is general resistance from production departments to changing tool operating parameters. The focus of
the production team is quality and output. Changes to tool operating parameters have the potential to impact
on both quality and output. They also involve production downtime together with the need to go through a
recertification/ qualification process to ensure that the process achieves the required output and quality.
Tool energy consumption is not a primary consideration when procurement of new tools is being considered.
Several of the sites visited identified that they procure partly used equipment. In this situation the choice of
equipment is limited and as such their ability to influence tool manufacturers is also limited.
Although progress in reducing process tool energy is advancing, albeit at a slower rate than the reduction in
utilities/facilities energy, it was observed some progress has achieved savings through the following:
Utilising energy monitoring software on their process tools to better understand their usage; and
Microelectronics Sector Guide 23
Switching off a cryo compressor associated with one specific process tool. The compressor was only needed
for a single step in the process but yet had been running 24/7.
These observations were site specific but are indicative of sector focus increasing in the process tools area.
From our preliminary observations and review of the process tools, and the way we understand how they are
operated, the following headlines are noted:
The process energy demand profile is relatively flat. Much of the equipment operates continuously with a
relatively small difference between demand whilst idling and the full load demand. Ancillary equipment
associated with process tools, such as cryogenic pumps are also required to operate continuously; and
For most sites visited a significant proportion of the energy consumed by process tools is used within the
furnaces i.e. heating elements, associated vacuum pumps and ancillaries.
As an example, for Site 3, a site highly reliant on the epitaxy process, the reactor furnaces accounted for
approximately 50% of process tool energy consumption. High energy use associated with reactors is
characterised by the requirement for the equipment to operate continuously at elevated temperatures even when
the furnace is not in use.
3.2 Process opportunities
From the process energy use breakdown and knowledge of the processes and their associated equipment it has
been possible to categorise areas of opportunities that offer potential to reduce energy. We have divided the
opportunities into the following primary consumer categories and presented in Table 8:
Water use;
Furnaces;
Pumps; and
Compressed Gases
Table 8: Process Categories
Process Element Water Use Furnaces Pumps Compressed Gas
Wafer Manufacture √ √
Insulating √ √
Placing √ √ √
Patterning/Lithography √ √ √ √
Removing/Etching √ √ √ √
Implanting/Diffusion √ √ √
Interconnecting √ √
Packaging √ √
Cleaning √ √
Microelectronics Sector Guide 24
As Table 8 demonstrates, lithography and etching are the highest energy processes in a site‟s overall process
energy use (40%). Table 9 below provides a brief overview of these categories with the subsequent sections
presenting the energy reduction opportunities in detail.
Table 9: Process Opportunities
Opportunity Comments
Water Use
Reverse Osmosis These processes demineralise feed water for
use in the Fab. Both are relevant to water
supply quality and are considered to be more
energy efficient than previous processes.
Depending on water quality, there are also
opportunities for recovery of waste water
available which reduces energy use further due
to reduced neutralisation demands.
Electrodeionised water
Water cooling Generally regarded as best practice. Reducing
cooling water requirements will largely optimise
available central systems while retaining
process tool and manufacturing requirements.
One site involved in this study has already
implemented improvements in this.
Rinse optimisation De-ionised water is the largest utilities cost for
any Fab due to energy in purification processes.
Water re-use can save considerable energy in
processing. One site involved in this study has
already implemented improvements in this.
Furnaces
Light Gauge Over Bend furnace elements Even temperature profile across furnace
element can be gained. Elements also found to
be much more energy efficient due to their
design and material makeup.
Vacuum insulation Vacuum insulation can be retrofitted to existing
furnaces to reduce radiation heat losses.
Current examples do not fit into the typical
operating temperatures of furnaces (<850°C or
>1650°C) but this technology may be extended
to this sector in the future.
Pumps – Standby operation
Load lock pump standby option (furnace
operation)
A vacuum is only required for a short period of
time in this part of the process. It was found
that most sites using this technology retain a
vacuum even if the chamber is empty. It is
thought that the pump can be set to standby in
these situations without any impact to product
quality.
Vacuum pump asset management Vacuum pump technology is such that efficiency
improvements are being achieved continually.
50% savings have been predicted in the near
future by leading vacuum pump manufacturer
Microelectronics Sector Guide 25
Edwards Limited.
Motor efficiency optimisation New standards in motor efficiencies have led to
high energy savings for any process reliant on
motors.
Gases
Clean Dry Air A lower cost alternative to Nitrogen than can be
used in some process applications.
On site nitrogen generation Can provide inherent advantages compared to
off-site generation subject to volumes
consumed.
3.2.1 Water use
Water is the most used „chemical‟ in the microelectronics industry. It is also the purist chemical involved in every
part of the manufacturing process. Historically comparisons between membrane processes and ion exchange,
and indeed developments in ion exchange systems themselves, have been addressed to capital costs rather
than a comparative energy cost. The operational costs that do embrace the energy element are usually part of
exercises that take in yield enhancement and increase production throughput. These are often from shopfloor
staff „tweaking‟ processes rather than a focussed improvement activity.
3.2.1.1 Reverse osmosis
Reverse Osmosis (RO) has replaced the Ion Exchange (IX) methods in recent years that were originally used to
generate pure water. RO is a form of water treatment that removes salt ions dissolved within the fluid. It is
particularly effective at removing heavy and trace metals, some Non Volatile Chlorinated Pesticides and
dissolved solids. Water molecules pass through a series of membranes which prevents other foreign ions
passing, to produce pure, contamination free water. This process typically rejects <25% of the feed water,
depending on its exact ion content, and achieves a purity of approximately 99.9%.
Figure 11 below shows the step changes in pure water generation that has occurred in recent times. Such has
been the pace of development in pure water generation that plant designed today has little relevance to 1st or
2nd generation facilities that were previously available (and still in used across established plants in the UK
today). Although an advanced technology industry, it is also a conservative one with regards to the central plant
and equipment. Today, pure water technology improvements are focused on improving membrane efficiencies.
Figure 11: Step changes in pure water production
Microelectronics Sector Guide 26
With the development of membranes many RO systems today, that have been retro-fitted with modern
membranes after the originals have expired, can now be operated at different parameters. For example a 2010
BW30-400HRLE (High Recovery, Low Energy) membrane is 15% more efficient than its 1990 BW30-400
counterpart although physically of the same appearance.
RO requires a higher driving pressure than the water‟s natural osmotic pressure. This pressure is exerted against
a semi-permeable membrane allowing the passage of the water molecules across the membrane and rejection of
the other molecules into a concentrate stream. To maintain this pressure the concentrate stream is “throttled” by
a valve.
Figure 12: Reverse Osmosis Membrane Energy Comparison11
Understanding exact energy savings from this technology is complex with salinity, temperature, pH, scaling
tendency, fouling index and ionic content all requiring consideration. Figure 12 shows how energy usage per unit
volume of water varies compared to feed water temperature and pressure.
Pre-heated water has process cycle time benefits that are also worth noting. Heating the water reduces its
osmotic pressure which increases the deionisation process efficiency. Whilst additional energy may be used to
pre-heat the water, significant energy savings can be made from the more efficient deionisation process. Heated
water is required for some sites as part of the process requirements such as controlling etchant rates. Pre-
heating of feed water for RO reduces the specific power from 0.78kWh/m3 at 10°C to 0.43kWh/m3 at 25°C,
realising an energy saving of 44%12
.
Lower temperatures, however, will give rise to higher quality water (i.e. number of total dissolved solids is lower).
This is due to the porous membrane contracting, relatively, compared to higher temperatures. A compromise
must be considered as smaller pores in the membranes require a higher pressure, and therefore greater energy
use, to pass the water though for the same flow rate. A rule of thumb used in RO design assumes flow rate
varies by 2.5% per 0.5°C change in feed water temperature.
Water purity has improved vastly as the RO process has been developed. Comparing today‟s purity level with
past RO or ion exchange processes is not viable due to material changes and advancing technology. RO is
11
Courtesy of Terry Cummings 12
Feedback from RO study at overseas facility
Microelectronics Sector Guide 27
generally reported to be approximately one third of the cost, compared to the ion exchange process it was
developed to replace, but it also achieves three times the water purity13
.
Recovery Reverse Osmosis (RRO) is a process enhancement that utilises the higher pressure to further treat the
water recovering a portion for re-use and absorbing the residual pressure energy for useful use. Thus, instead of
decaying pressure across a valve it is used to improve the water recovery of the overall RO system and makes
use of energy that would otherwise be wasted.
A design exercise for a UK wafer Fab (that was not progressed due to capex restrictions) provides an illustration:
Capital cost for RRO installation was £175k versus a payback possible in just over one year. The savings were
generated through recovered water, reduced electrical power and anti-scalant chemicals. This example takes
account of efficiencies and process realigning required to modify the existing system. This may not be
necessarily the case in all instances, further highlighting the potential of RRO.
The table below uses industry data showing the improvements in membrane technology over the last ten years.
These savings should not simply be considered in parallel as the energy savings achieved are only one of many
advantages that can be gained from the improved technology, see 6.2.1.2. The figures provide an order of
magnitude of the energy saving possibilities.
Table 10: Energy savings from RO Membrane Technology
2000 2010 Sample
Analysis (2011)
Future
kWh/m³ 1.13 0.64 various 0.5
Feed water
Temperature
10°C 25°C Pre-heat 75% heat
recovery
Process Energy Saving - 57% 63% Additional +22%
RO is fully automated, self-monitoring and diagnostic, thus significantly reducing labour costs. RO is also a
continuous process while IX is batch processed. Use of RO water eliminates blowdown and reduces
maintenance.
Some sites do not have RO as the feed water and/or some devices that are manufactured may preclude its use.
However, there is general application across the sector and hence taking advantage of the improvements that
have been made offer the sector good potential.
3.2.1.2 Electro deionised water production
Electrodeionisation (EDI) was first considered in the early 1950s with high hopes of cross-sector advantages in
water usage reduction and high quality outputs. In a 1952 newspaper article a figure of 20kWh/1000 gallons of
water was quoted. Originally used in desalination processes, the EDI process did not become more widely
accepted until the feed water could be conditioned to an acceptable level. However, 40 years on, this technology
is still not as widely seen in the UK microelectronics sectors today as perhaps could be potentially implemented.
EDI produces high purity water using an ion resin exchange process. EDI was developed to improve the ion
exchange process further. There are applications where ion exchange is suitable and EDI is not, such as
condensate polishing in power applications due to the presence of oil and grease; conversely EDI is superior to
13
Based on industry data with current electricity prices applied. See also reduction of total dissolve solids (TDS) in Figure 12
Microelectronics Sector Guide 28
ion exchange such as silica and boron reduction in ultra pure water. For a meaningful comparison the inputs and
expected outputs have to be comparable for the processes being considered.
EDI is typically used after RO/RRO processing to further demineralise the water. An electrical potential drives
the ions through a set of permeable membranes, separating ions and demineralising the water further. It is
considered advantageous for the following reasons:
Low energy consumption;
No chemical requirement – further reducing the associated hazard risk;
Regeneration of resins reducing maintenance requirements; and
Low operational costs.
EDI will typically recover 95% of the feed water. A continuous EDI machine with a capacity of 15,000L/hr
requires just 4.5KW of input energy. This is also supported by the fact that EDI is continuous but ion exchange
involves batch processing.
The amount of electrical power required to continuously regenerate the stack is directly proportional to the ions to
be exchanged (removed into the concentrate stream). It can be assumed that cost savings will directly relate to
energy reduction. It also follows that the performance of an RO process has a direct effect on the performance of
EDI. For this reason it is usual to deploy a two pass RO in EDI applications.
The use of EDI is a proven technology outside of the UK. Using first principles and industry data based upon an
installation in Singapore it has been possible to provide a comparison costs per regeneration14
as shown in the
table below.
Table 11: Comparison of EDI and RO technologies
RO/IX RO/EDI RRO/EDI
2009 Case Study
(£/regeneration)
£2,023 £594 £156
Estimated saving - 71% 90%
Another case study was completed recently for a system in Johor Bahru, Malaysia. Power, water and local cost
differences compared to the UK do not allow a similar cost comparison to be made as in Table 9 but the site has
confirmed a 17-19% cost saving per unit volume of pure water between RO/EDI and the RO/IX technologies.
There is a convincing argument for EDI to be followed up in more detail in the future as there are proven energy,
cost and resource usage benefits from this technology.
These savings are likely to be through energy reduction or other process changes but the additional benefits of
EDI, aside from the energy saving potential, should also be noted:
No use or bulk storage of chemicals;
No chemical waste and no wastewater to neutralise;
No interference of clean systems with dirty chemicals;
No invasive procedures;
14
EDI parameters as based on proven data from in Singapore for 50gpm stack latest development packed concentrate chamber units commissioned 18 months ago. Regeneration quantities depend on speed of process and number of processes. One regeneration is for a set volume of water passed through all phases in the deionisation process. A mixed bed regeneration process produces approximately 34cu.m of water, EDI over 47cu.m. Costs have been converted to reflect UK rates.
Microelectronics Sector Guide 29
No remixing of resins;
No rinsing;
No expensive Nitrogen Gas to remix;
Reduced Operator skills base;
No interruption to service flow – Continuous Regeneration;
Fully Automatic and Self Monitoring;
Reduced HPM Risk; and
Reduced Footprint, Height and Foundation Loads.
3.2.1.3 Water cooling opportunities
Reducing cooling water flow requirements has been typically received in this study as now being a best practice
industry standard to aim for. This is primarily due to recently escalating utilities costs in both water and electricity.
There are process limits to be considered when changing water temperatures which are invariably influenced by
other processes in the Fab and the individual machines and processes used. A cooling water ring main can be
installed, where possible, which has the benefit of removing a substantial number of independent pumps in the
system.
Typically there are three cooling systems - glycol, CHW (chilled water) and evaporative:
Glycol systems are used where low and very precise humidity control is required. Easing the humidity set
points and / or also control differentials has proven to obviate the need for glycol systems.
CHW is used for sensible cooling of air handling systems, dehumidification in less demanding areas and
Process Cooling Water (PCW).
Evaporative cooling makes use of the climatic conditions in parts of the UK utilising adiabatic cooling towers.
All of these systems have a cooling element and a rejection of heat element. This reject heat is usually in the
order of 20~22°C. This heat can be used for process use i.e. enhancing RO performance, frost coils, space
heating and pre-heat of domestic hot water though no data has been collected related to these options.
Reducing cooling water requirements is largely optimising available central systems while retaining process tool
and manufacturing requirements. Evolution of tools to increase efficiencies and hence heat losses is being
introduced but overall reduction in cooling demand from these activities may not be significant compared with the
other opportunities.
3.2.1.4 Rinse optimisation
Ultra pure water is the most expensive utility in a Fab when combining process costs and supply costs. An option
used by some Fabs to reduce this usage is the „quick dump rinse‟ after the etching process. The usage is
controlled by resistivity value of the water and is then purged when resistivity falls out of limits. It can also take
advantage of free cooling through heat recovery from other processes.
Microelectronics Sector Guide 30
Figure 13: Examples of Rinse Processes15
The overflow dump rinse options are used by two sites spoken to (who refer to it as a weir rinse). The quick dump
rinse was developed in the 1980s and uses much less water as the initial rinse is “dumped” and the subsequent
rinse is re-circulated. Further developments in this technology led to the recovery of the secondary rinses to the
DI water plant rather than discharge to drain.
The simple act of reducing water usage in a Fab has the benefit of reducing energy consumption due to the
volumes, processing and pumping energy required. Further work on this could involve a review of energy
specifically for water supplied into Fab and its circulation through the specific processes it is used in. The rinse
process is of particular interest as it is critical to the Fab process with an estimated 8%16
of process energy (3%
of Fab total energy use) dedicated to wafer cleaning. The potential of this saving is dependent on wafer moves
at each site.
There may be an additional benefit possible at some sites where wastewater from a Fab can be recovered. This
water almost always has a high acidity level (<pH2) and is predominantly very pure water. However, its ionic
content is very low which means less deionisation processing is required to enable the water to be used in the
processes again. It is believed that 50% of waste Fab water could be recovered with low capex and reused in
the manufacturing process. As much as 80+% can be achieved with a comparatively moderate additional
processing adding to energy savings per cubic meter of water used. This figure has been achieved at overseas
sites such as the Ang Mo Kio Technology Centre in Singapore though energy reduction has not been measured.
Where water quality is deemed not to be acceptable quality for processing use it can be used as a grey water
source for other parts of the site i.e. landscaping, flushing etc.
As a whole the consideration of water consumption will have an increase in focus. The sample projects we have
cited are located in Asia where water consumption is already a high priority. The water consumption opportunities
appear to offer a wide range of benefits to the sector that should make further consideration an attraction.
15
Ultrapure Water Use in Wafer Cleaning and CMP”, Chiarello, R, Stanford University, supported by International Sematech 16
Based on NMI sector data in 2008
Microelectronics Sector Guide 31
3.2.2 Furnaces
3.2.2.1 Furnace elements
Just as a corroded or old element in a household boiler is less efficient, old or damaged elements can be
inefficient in Fab furnaces. The patented Light Gauge Overbend (LGO™) element presents a step change in
furnace efficiency and capability. Industry literature claims 40% energy saving is possible (Koyo Thermo
Systems Ltd.17
) and that LGO elements can be retrofitted in some existing horizontal furnaces. They also
present a more consistent, uniform temperature profile in the cross section of the furnace with less temperature
variability at edges of the chamber (typically +/- 1°C). This has the advantage of a consistence heating process
across the wafer(s) which may assist in reducing wafer moves in a furnace.
Tetreon have confirmed that they use LGO elements in their furnaces with MRL Industries (furnace element
supplier to 2 host sites associated with this study) supplying „Black Max‟ elements that use the LGO technology.
However, it is understood that Black Max elements are only suited to low temperature furnaces (<850°C) – too
low an operating temperature in most semiconductor processes. It may however be suitable for low pressure
chemical vapour deposition which typically requires 700-850°C. Thermo Scientific do supply a 1200°C LGO Box
Furnace which suggests the technology can be used at higher temperatures in other types of furnace. The idea
of using this technology at higher temperatures may be beneficial in future furnace developments in the
semiconductor industry, including retrofittable elements. The development of this technology specific to the
furnace temperatures used in Fabs has not been developed as yet apart from in lower temperature processes
such as Chemical Vapour Deposition.
Potential energy savings have been estimated based on example site data provided by two of the sites as well as
sector wide data taken from the NMI. These are shown in the table below.
Table 12: Energy savings from LGO technology
Furnace Annual
Energy (kWh/year)*
Estimated saving
from element
replacement
(kWh/year)
Estimated cost
saving (OpEx only)
(£/year)**
Sector wide 113,960,000 25,400,000
(assumed 60%
uptake)
£2.03m
Site 1 (2004) 900,000 226,000 £18,000
Site 3 (2005) 2,409,000 602,250 £48,000
*assumes 22% of site energy is related to estimated furnace utilisation and matched sector wide figures of 20-30%. As
mentioned previously site 3 uses a greater proportion than this due to their specific process requirements.
** assumes 8p/kWh
The introduction of LGO technology to the microelectronic sector requires further clarification in the industry.
There does appear to be an opportunity for potential development for some semiconductor industry applications.
It is an established, patented furnace element design with high energy saving estimates that are significant
enough to warrant further examination.
17
www.crystec.com
Microelectronics Sector Guide 32
Figure 14: Light Gauge Overbend Elements18
3.2.2.2 Furnace insulation
In discussing with Tetreon and consulting Thermo Scientific literature it was understood that electrical
consumption of furnaces is highest during temperature ramp up. This removes the viability of the solution of
reducing furnace temperatures during standby times.
Most furnace heat losses are through air or water movements that are typically minimised in design rather than in
operation improvement. Furnace insulation materials are typically glass, ceramic or carbon fibre based. Vacuum
insulation, a potential retrofit solution for horizontal furnaces, works just as a vacuum flask does by reducing heat
radiation to a minimum. Tetreon have used carbon based vacuum insulation successfully with other customers
to those involved in this study but the solution was found to only be cost effective for furnaces operating above
1850°C. It is particularly successful for furnaces operating with a highly volatile atmosphere where the added
insulation of this type is a fundamental requirement rather than an advantageous add-on19
.
A form of vacuum insulation used by Thermo Scientific, Moldatherm®, is a high temperature fibre vacuum-formed
around the furnace chambers. According to literature from Thermo Scientific it provides efficient radiant energy
release, improved temperature uniformity across the furnace chamber and rapid heat-up and cool-down
properties. Moldatherm is used in furnaces operating at 100-1100°C.
Vacuum insulation is perhaps a solution that could be considered in principle with more research required into
cost effective solutions for semiconductor specific furnaces in the future. Current operating temperatures are
between 1200°C and 1500°C, out of the capabilities or value for the examples seen thus far.
3.2.3 Pump stand-by operation
3.2.3.1 Vacuum pumps
Pumps can be one of the largest energy users in a Fab, typically 50-60% of process energy. Microelectronic
Fabs have many numbers of pumps across the operations. Due to process control requirements it is not possible
to shut down some of these pumps even when there is temporarily zero through-put. However a structured asset
management plan to incorporate new pumps into an existing Fab would be beneficial to overall operating costs
and reduced energy usage at any semiconductor Fab.
18
Pictures from www.mrlind.com and Thermo Electron Corporation literature 19
Telephone conversation with Iain McGregor, Sales Manager, Tetreon
Microelectronics Sector Guide 33
Vacuum pump manufacturer Edwards Ltd, a leading global supplier in vacuum pumps, especially those for the
semiconductor industry, advocates a whole range of new, energy efficient pumps. Edwards have set corporate
aims to reduce energy usage of all their pumps including 10% energy reduction on heavy duty pumps‟ operations
specifically for the semiconductor industry.
There is, however, the barrier of substantial capital expenditure being required if a planned pump replacement
regime was to be undertaken in a single Fab. A phased replacement roll-out using new pumps when old ones
become uneconomical to repair would be a suitable option if long term investment planning was considered by
the sites.
3.2.3.2 Load lock vacuum pumps
A load lock allows a wafer to pass from the cleanroom into a process chamber for lithography via a compartment
having a door at each end. The outer door opens with the vacuum pump off, closes and the vacuum pump pulls
down the pressure in the chamber before the inner door opens and the wafer passes into the chamber where an
inert gas (usually nitrogen) is introduced to reduce surface oxidation. If the process chamber conditions do not
equal the vacuum chamber conditions, the doors will not open. Once the wafer is in the processing chamber
there is no process requirement to maintain the vacuum within the load lock. Understanding the throughput
impacts and energy use implications of constantly running load lock pumps compared to variable loading is
required to assess the energy usage requirements.
One such load lock pump that could be considered for this opportunity is the iXL120 by Edwards, launched in
2010, which advocates a „Green Mode‟. This product allows reduced energy usage during idle periods. This
pump is rated to consume circa. 83% less power for the same pumping capacity (see Figure 15 below).
water (i.e. higher dissolved solids). Therefore once
established in UK as well as further technology
improvements costs for this opportunity are likely to fall.
Recent projects in Asia estimate cost of setup to be
£175,000 for one site though water costs are considerably
higher.
Payback Site dependant, generic estimate of 1-3years
Overview of project The project would focus on existing water purification
practice. A suitable consortium would include a
microelectronics company, a consultant with previous
experience of implementing RO and EDI schemes
including the technology supplier
Demonstration project and
possible structure
Typical activities to be completed for a demonstration
project would include:
Assessment of existing purification process –
representative site example to establish equipment,
configuration and consumption
Scheme design and costing of proposed alternative
configuration c/w life cycle costing
Assess the roll out to the sector
Promote findings
Initiate a test case
Cost of carbon ~£9,000/site/year saving (based on 1st April 2013 carbon
floor price)
Barriers to adoption Providing a convincing illustration to the UK sector
Meeting the strict financial criteria that rules scheme
adoption
Microelectronics Sector Guide 59
Light gauge overbend (LGO) furnace elements
Technology maturity and need for
support
Support required to establish elements and suitable
temperature ranges for individual sites
Potential cost saving to site £130,240/site/year based on energy price of 8p/kWh
Potential carbon saving to site 887tCO2/site/year
Energy saving to site up to 1,628,000kWh/site/year (assuming 40% saving from
element and all furnace elements replaced)
Market penetration LGO elements are already use in LPCVD process. The
current technology does not reach the high temperatures
required to other processes as yet but elements are used
in other sectors that can meet these requirements.
Estimated 60% uptake for all furnace types if universally
proven.
Cost of technology (once mature) Unknown
Payback Unknown
Overview of project The project would focus on existing furnace elements, the
potential to retrofit and savings that could be achieved.
Demonstration project and
possible structure
Typical activities to be completed for a demonstration
project would include:
Assessment of application to conventional furnace
manufacturers
Application to retrofit
Assess the roll out to the sector
Promote findings
Cost of carbon £14,200/site/year saving (based on 1st April 2013 carbon
floor price)
Barriers to adoption Retrofit potential not possible with some furnace
models
Providing a convincing illustration to the sector
Meeting the strict financial criteria that rules scheme
adoption
Microelectronics Sector Guide 60
6 Next steps
The sector has a high level of awareness surrounding energy savings. The trade association is very active in
promoting and sharing advancement in reduction techniques through such initiatives as the “toolbox” project.
Through such projects, advancement in utility energy reduction has been significant over the last decade. Given
the various barriers the sector is faced with, these advancements are a real achievement and we believe this
sector would be one of the most educated in industry.
Conventional techniques are known to the sector, if not wholly implemented. These have been noted in this
report for reference.
Implement process specific good practice measure
We believe that some of the observations listed in the previous section would have merit for implementation but
that further development is required. Such opportunities hold much potential to reduce energy but may need
additional proving through structured pilot studies for the sector to adopt and integrate universally.
This would particularly apply to the following;
Stand-by options for load lock vacuum pumps; and
On-site gas generation
Therefore, further consultation with equipment suppliers is required to develop the business case in order to
assess the financial and process viability of these opportunities. Due to the complexity of the solutions presented
and the immediate impacts on the wider Fab sites pilot studies could be undertaken through the trade association
technical forums that the sector companies attend to accelerate this iterative process of implementation.
Implement process specific innovation measure
The development of RO with EDI and furnace element technology appears to be areas, associated with process
tools that the UK sector could take advantage of. Case studies of overseas Fabs that employ EDI technology are
available and LGO elements are already used in lower temperature furnaces. These could form the basis for the
commencement of UK feasibility studies to compare and illustrate the benefits.
Pilot studies of this nature could result in test case projects ultimately leading to a sector roll-out as its application
appears to be universal.
Microelectronics Sector Guide 61
Acknowledgements
This report has been prepared by the Carbon Trust in conjunction with Ove Arup and Partners Ltd and in
collaboration with the National Microelectronics Institute and Mr Terry Cummings. We are also grateful to the
support and data offered by 5 of the sector companies.
Microelectronics Sector Guide 62
Appendices
Appendix 1: Glossary and useful links
Microelectronics Sector Guide 63
Appendix 1: Glossary and useful links
Glossary
IC – Integrated Circuit
Discrete Device – Single circuit element e.g. transistor, diode
DRAM – Dynamic Random Access Memory: low cost and small size
SRAM – Static Random Access Memory: higher cost but faster than DRAM
EEPROM – Electrically Erasable Programmable Read-Only Memory: can be written on and read from
CPU – Central Processing Unit: microprocessor that comprises many different units
ASIC – Application Specific Integrated Circuit: performs specific task
DSP – Digital Signal Processor:
MOS – Metal Oxide Semiconductor
CMOS – Complementary Metal Oxide Semiconductor
NMOS – N channel Metal Oxide Semiconductor
PMOS – P channel Metal Oxide Semiconductor
NMOS devices are typically two to three times faster than PMOS devices
The main advantage of CMOS over NMOS and bipolar technology is the much smaller power dissipation.
BJT – Bipolar Junction Transistor: commonly used in analogue circuits, dissipates high amounts of power
MOSFET – Metal Oxide Semiconductor Field Effect Transistor: commonly used in digital circuits, dissipates
much less power than BJT
Epitaxy – depositing a monocrystalline film on a monocrystalline substrate
MBE – Molecular Beam Epitaxy
MOVPE – Metal Oxide Vapour Phase Epitaxy
EDI – Electronic De-Ionisation
Microelectronics Sector Guide 64
RO – Reverse Osmosis
UPW – Ultra Pure Water
LGO – Light Gauge Overbend
Industry Links
NMI – National Microelectronics Institute
WSTS – World Semi Conductor Trade Statistics
SIA – Semiconductor Industry Alliance
SEMI – Semiconductor Equipment and Materials International
iSuppli.com – Market Research Facility
Microelectronics Sector Guide 65
The Carbon Trust receives funding from Government including the Department of Energy and Climate Change, the Department for Transport, the Scottish Government, the Welsh Assembly Government and Invest Northern Ireland.
Whilst reasonable steps have been taken to ensure that the information contained within this publication is correct, the authors, the Carbon Trust, its agents, contractors and sub-contractors give no warranty and make no representation as to its accuracy and accept no liability for any errors or omissions.
Any trademarks, service marks or logos used in this publication, and copyright in it, are the property of the Carbon Trust or its licensors. Nothing in this publication shall be construed as granting any licence or right to use or reproduce any of the trademarks, service marks, logos, copyright or any proprietary information in any way without the Carbon Trust‟s prior written permission. The Carbon Trust enforces infringements of its intellectual property rights to the full extent permitted by law.
The Carbon Trust is a company limited by guarantee and registered in England and Wales under Company number 4190230 with its Registered Office at: 6th Floor, 5 New Street Square, London EC4A 3BF.