<^  Office of Air and Radiation
g  Office of Atmospheric Programs, Climate
^  Change Division
*   EPA-430-R-06-901

This report presents the results of initial research on the uses of liquid perfluorinated
compounds (PFCs) as heat transfer fluids (HTFs)—materials to manage heat—
throughout the electronic sector. The information was obtained from searches of the
Internet using the search engine provided by Google.  The objective of the searches was
    •   Identify major uses and suppliers of liquid PFCs employed as heat transfer fluids;
    •   Identify leading users— suppliers/manufacturers of equipment that employ liquid
       PFCs and manufacturers of devices, device assemblies, and electronic products,
       which included the consumer, industrial and military sectors;
    •   Identify published estimates of air emissions and trends of liquid PFCs employed
       as HTFs in the electronic sector; and
    •   Identify options for estimating emissions from the top emitting sources.

During the manufacture of semiconductor devices, HTFs serve as coolants in chillers,
removing excess heat during many manufacturing processes. During semiconductor
device testing, containers of HTFs  are cooled or heated to a desired temperature into
which the devices are immersed to test their integrity.  In addition, when testing the
function of devices, HTFs are used to remove the heat the devices generate while being
tested. HTFs are also used to attach semiconductor devices to circuit boards via solder,
which may be melted by the vapor of an  HTF heated to its boiling point. HTFs may also
serve to cool semiconductor devices and other devices or systems that generate high
heat during operation.

The remainder of this report consists of seven additional sections. The next section,
Section 2, describes the physical and chemical characteristics of liquid PFC HTFs for
electronic manufacturing. Section 3 presents the history of liquid PFC uses in the
electronics sector, which is followed,  in Section 4, by detailed descriptions of the current
uses,  focusing on the quantities used and the factors that govern selection of HTFs. In
Section 5, annual emissions are estimated for each use. Section 6 presents the
distribution of sources (users of HTFs) in light of the current and evolving structure of the
electronics sector. Section 7 describes three  options for estimating emissions, based  on
the information obtained from this initial research. The last section proposes additional

work aimed at clarifying the distribution of sources, the relative proportions of HTF use
among competing technologies as well as the modes and amounts of emissions from
liquid RFC HTFs.


Tuma reported that for managing heat loads in the electronic sector, liquids are sought
that have high dielectric strengths, acceptable thermal conductivities and are also inert
and safe. Perfluorocarbon (PFCs) liquids possess these properties (Tuma, 2003).
Among these liquids are fully fluorinated linear, branched or cyclic alkanes,  ethers,
tertiary amines and aminoethers, and mixtures thereof. The chemicals can be straight
chains, branched or cyclic, or a combination thereof (such as alkylcycloaliphatic and
cycloethers) and are saturated (i.e., free of ethylenic, acetylinic or aromatic  unsaturated
parts). The skeleton can include catenary oxygen, trivalent nitrogen heteroatoms, or
both, providing stable links between fluorocarbon groups that preserve the inert
character of the compound (3M, 1991). Also among the RFC liquids with the desired
properties for the electronics sector are perfluoropolyethers and mixtures thereof (Solvay
Solexis, 2004).1

The searches revealed two global manufacturers of liquid RFC HTFs. Solvay Solexis
offers the perfluoropolyethers (often as mixtures) under the trade  name Galden™ (e.g.,
Galden™ HT-135) with a practical temperature range of -40°C to +120°C for use in
chillers; other Galden fluids are useful at  higher temperatures (e.g., DO2-TS, up to
160°C in thermal shock testing (TST) and LS/215 up to 215°C) (Solvay Solexis, 2004).
3M also offers, perfluoroalkanes, perfluoroethers and tertiary perfluoroamines under the
trade names Fluorinert™ (e.g., Fluorinert™ FC-3283), with a practical temperature range
of -40°C to +120°C for use in chillers; other Fluorinert fluids are useful at higher
temperatures (e.g., FC-5320 is used in TST up to 160°C and FC-70 up to 215°C (3M,
2004). Galden and Fluorinert fluids have  molecular weights between 350 to 820 g/mole.
Occasionally Krytox, a product of DuPont, appeared in the searches as a heat transfer
1 The molecular formulas for the simpler classes of fully-fluorinated compounds are:
perfluoropolyethers, CF3-(O-CF(CF3)-CF2)n-(O-CF2)m-O-CF3; perfluoroalkanes, CnF2n+2;
perfluoroethers, CnF2n+i(O)CmF2m+i; perfluorocyclic ethers, CnF2nO; tertiary perfluoroamines,
(CnF2n+i)3N. In these compounds, n>6 and m>2-3. FC-72, for example, consists mostly of C6F14
with a boiling  point of 56°C, while FC-75 is mixture of C8F18 and c-C8F16O with a boiling point of

fluid. However, according to DuPont product literature, Krytox fluids, while being a
perfluoropolyether (like Galden), serve only lubricant markets (DuPont, 2004). The
molecular weights of Krytox fluids are high— between 1177 and 6034 g/mole, consistent
with Krytox being lubricants (oils and greases) not heat transfer fluids (DuPont, 2004).

The dielectric properties and chemical inertness of Galden and Fluorinert liquids make
them particularly attractive as heat transfer media in electronic fabrication, testing,
assembly and operation.  Their very high dielectric strengths and stable resistivity
provide a dielectric barrier and reduce energy losses (i.e., stray currents) from the high
fluctuating voltages and magnetic fields present in etching and CVD chamber
environments.  In addition, despite their relatively low thermal conductivities (compared
to water), they possess other properties that make them useful HTFs,  including, a wide
range of  boiling points, high densities,  low viscosities, low pour points, low surface
tension, and high thermal and chemical stability. Their compatibility with metals, plastics
and elastomers makes these fluids  attractive in thermal testing, reflow soldering as well
as temperature control during device or system operation. They are odorless,
nonflammable, nonexplosive, evaporate cleanly and are practically non-toxic, which
results in easier safe usage and storage (Tuma, 2003).

Typical HTFs such as Galden and Fluorinert are encumbered by their relatively high
vapor pressures, estimated long atmospheric lifetimes, and relatively strong absorbance
of infra-red radiation, which indicates that, if not contained, their potential contribution to
global warming is a concern. The few published reports indicate that the global warming
potentials (GWPs) of liquid PFCs are similar to  the gaseous PFCs used in
semiconductor device and TFT-FPD (thin film transistor flat panel display)
manufacture—emissions that are currently being controlled globally (see Table 1).

Table 1: PFC Gases and Liquids Used in the Manufacture, Testing and Operation
of Electronic Components and Systems
HT-1 1 0
HT-1 35
HT-1 70
point, °C)
Molecular weight
(Avg. for liquids)
Potential (100
yrtime horizon)
Molina et al, 1995
T&T, 2001
T&T, 2001
T&T, 2001
3M Data, 2004
SS, 2004
3M Data, 2004
3M Data, 2004
3M Data, 2004
SS, 2004
T&T, 2001
SS, 2004
3M, 2004
3M, 2004
SS, 2004
3M, 2004
SS, 2004
3M, 2004
3M, 2004
GWP values

TAR, C6F14
TAR, C6F14
^' Denotes properties at standard temperature and pressure. (2) It appears T&T have adopted methodology
from Third Assessment Report to estimate certain GWP. (3) T&K denotes Tuma and Knoll, 2003, see
reference 8 where T&K cite lifetimes of liquid PFC HTFs >2000 years to give GWP >6000 for 100-year
integrated time horizon.

Recently, 3M and Solvay Solexis announced the availability of non-PFC heat transfer
fluids with lower, although still significant, GWP values (T&T, 2001; Solvay Solexis,
2003).  Examining these HTFs in detail, which have reported GWPs below approximately
400 and 3000 for the 3M and Solvay Solexis materials, respectively, is outside the scope
of this report (for GWP estimates see Table 7.7 at tar/wg1/pdf/TAR-06.PDF ). Nevertheless, both 3M and
Solvay Solexis claim to have lowered the atmospheric lifetimes and, therefore, GWPs
without compromising the properties that make them attractive HTFs to the electronic

sector.2 By virtue of judiciously substituting certain F-atoms with H-atoms within the fully
fluorinated carbon compound, the lifetime is shortened via chemical reaction of the H-
atoms with the ubiquitous atmospheric OH-radical.

PFCs have been and are applied to cooling electronics in evolving and expanding ways
(Tuma, 2003). Heat is removed either by direct or indirect contact of the RFC liquid with
the heat source, relying on either single or two phase modes of cooling, with the latter
taking advantage of the latent heat of evaporation to remove more heat. In the early
1950s, liquid PFCs served as  heat transfer media to cool sensitive military electronics,
principally devices used in electronic countermeasure (ECM), radar, sonar, and
guidance systems (Tuma, 2003). In the 1970s and 1980s, RFC fluids moved into
commercial applications, initially for thermal testing and reflow soldering operations,  and
then to cool supercomputers, lasers, and x-ray targets (Tuma, 2003).

In the early 1970s, 65 percent of RFC fluids were used in thermal testing (which,
replaced glycol/water mixtures) and 25 percent were used for cooling electronics in
military systems. The remainder was presumably used to cool large transformers and
power transmitter tubes, in competition with CFC-113 (Tuma, 2003).

In the 1990s,  increasing semiconductor processing temperatures as well as the growing
adoption of complementary-symmetry/metal-oxide semiconductor (CMOS) technology
(which is adversely affected by higher temperatures) forced manufacturers to begin
replacing deionized water/glycol fluids with liquid PFCs. Until then the Dl/glycol fluids
had been used exclusively in temperature control units (TCUs), units for regulating
temperatures during etching, plasma enhanced chemical vapor deposition (PECVD), ion
implantation, and photolithographic processes. In the latter 1990s, in anticipation of
increasing semiconductor heat generation (from exponentially increasing transistor
speeds and densities), semiconductor manufacturers initiated research on device spray
2 3M calls its substitutes segregated hydrofluoroethers, referring to the inclusion of a -OCnHm (the
H-containing) segment that is separated ("segregated") from the fully fluorinated segment. 3M
trade names use HFE in the name. Solvay Solexis calls its substitutes hydrofluoropolyethers,
referring to the substitution of H-atoms for the F-atoms on the terminating ends of the
fluoropolyether. Further information for both supplies is available at and s.155/97132.

cooling with liquid PFCs in closed systems (Tuma, 2003). Hewlett Packard, for example,
began exploring the utility of combining Fluorinert FC-72 and the nozzle-control offered
by its inkjet technology (in a closed system) to spray-cool devices when and where
cooling is needed (Hewlett Packard, 2002). During this period, Defense Advanced
Research Projects Agency (DARPA) of the Department of Defense (DoD) and other
DoD funding sources initiated programs to exploit the application of liquid PFCs to cool
high-power electronics in military systems (SDA, 2000).

The initial transition from liquid water/glycol to liquid PFC cooling led to inadvertent
increases in PFC emissions. Initially, device manufacturers typically replaced the
water/glycol coolant with the liquid PFC, without any changes to the TCU pump, seals
and connectors. The lower surface tension of liquid PFCs compared to water resulted in
PFC loss via leakage, which equated to PFC atmospheric emissions. Current practice
among semiconductor manufacturers is, when switching from water/glycol to liquid PFC
TCU units, to replace pumps, seals and connectors to prevent leakages or to replace
water/glycol TCUs with suitably designed PFC-using TCUs, which may or may not use
lower-GWP HTF alternatives.2

In 2000, reports of new uses of liquid PFCs emerged. Exploration was initiated of liquid
PFCs to cool rectifiers (which consist of high power isolated gate bipolar transistors for
converting DC to AC current for electric motors) in hybrid vehicles.

By 2002, the predominant use of liquid PFCs had changed. Sales of PFC liquids for just
TCUs for semiconductor manufacturing and testing approached 50 percent of all PFC
heat transfer sales (Tuma, 2003).

By 2005, penetration of liquid PFC-usage  reached applications in electronics
manufacturing: (1) TCUs, (2) thermal shock testing (TST), (3), vapor phase reflow (VPR)
soldering, (4) automatic test equipment (ATE), and (5) device cooling. The volume
usages (%) corresponding to these five categories in 2005 were. 50 (TCU), 20 (TST), 10
(VPR soldering), 15 (ATE) and 5  (device cooling).3 (Werner, 2005).
3 Werner (2005) also reported that, in 2005, approximately 10% of the TCU usage employed low-
GWP HTF fluids while the corresponding usage for ATE was 65%. Challenges are slowing the
use of low-GWP alternatives TST and VPS applications.

This initial research provided information about commercial TCUs, thermal shock testing
(1ST), reflow soldering products, and cooling individual devices, clusters of devices, and
commercial-off-the-shelf (COTS) assemblies.  Consequently, the discussion of RFC-
liquid use in these applications is the main focus for the remainder of this report.

Temperature control is of increasing importance during the fabrication, testing and
operation of electronic devices, products and systems.4 As electronic components (or
integrated circuits, ICs) and systems have advanced in speed, transistor density, and
function, managing thermal budgets during fabrication has also increased. Testing
device performance under expected temperature environments and providing for the
dissipation  of heat produced during device, product and system operation have also
increased. Temperature regulation also determines the performance and reliability of
surface-mounted devices (SMDs), where the electrical leads of numerous packaged 1C
devices and passive components (e.g., resistors) are interconnected by soldering,
specifically a process called reflow soldering.5

Temperature control—the removal of excess heat by cooling—is achieved by air cooling
and liquid cooling. Tuma & Tousignant (T&T) (2001) review the uses of liquid  PFCs as
heat transfer fluids. T&T as well as others indicate that liquid cooling is becoming the
preferred method for dissipating the increasing amounts of heat associated with both
device fabrication and shrinking electronics (cf. Simons, 1996). The reasoning is
straightforward: the highest heat transfer coefficient for air (achieved through forced
convection) is roughly 0.005W/cm2/°C compared to the corresponding figure for a
4 In this report the term electronic devices, product and system includes power and optoelectronic
applications, such as metal oxide semiconductor-controlled thryristors and telecommunication
cabinets. Power and optoelectronic applications are sources of growing revenues to the
electronics industry.
5 The common method of soldering onto a substrate is by mounting devices on a preprinted
substrate using a solder paste. The entire substrate is then heated, remelting the solder to
interconnect the devices. This method of soldering is called "reflow soldering." In 1999, reflow
soldering equipment accounted for 78 percent of electronic soldering equipment sales and was
projected to exceed 85 percent by 2006 (as reported in Fabtech at A less common method is
wave soldering, where a machine generated wave of molten solder alloy makes controlled
contact with the devices being mounted. The surface mounting process may employ both reflow
and waver soldering.

dielectric liquid (achieved via two-phase coolant spray) is 1.5 W/cm2/°C, 300 times
greater (Incopera and DeWitt, 1996).

RFC liquids, however, have long atmospheric lifetimes and high global warming
potentials (T&T, 2001). Thus, to the extent they evaporate to the atmosphere, their
contribution to global warming is a concern.

The remainder of this section summarizes four uses of RFC liquids: temperature control
units used during the manufacturing process; integrated circuit (1C) testing; soldering to
connect surface-mount devices; and cooling individual devices, clusters of devices, and
commercial-off-the-shelf (COTS) assemblies.6'7 Each of these uses is described below.

4.1 Temperature Control Units (TCUs) For Manufacturing
Temperature control during  1C fabrication is achieved through temperature control units
(TCUs), part of the support equipment (such as vacuum pumps)  attached to wafer
cluster tools. TCUs use coolants to remove unwanted heat from wafers during film
etching and deposition processes as well as ion implantation, and lithographic
processes.  Coolant is circulated through the wafer mounts during the process. TCUs are
also used to remove heat from quartz windows and reactor domes. The semiconductor
industry's continuous pursuit to reduce wafer processing costs (i.e., to lower the life-
cycle cost of its equipment by considering equipment throughput, uptime, and effects on
1C yield) is, among other things, motivating TCU design innovations, which facilitates
switching from deionized water/glycol coolants to fluorocarbon liquids (Sheppard, 1999;
BOC Edwards, 2004; Solid State Systems Cooling, 2003).

The available literature suggests that each process tool that requires temperature control
has its own TCU (Sheppard, 1999; BOC Edwards, 2004). A modern 200 or 300 mm
wafer fab will operate and maintain, respectively, approximately 1600 and 1400 distinct
pieces of manufacturing equipment (Deutsche Bank, 2002; Smith Barney, 2004). In the
 COTS electronics refers to military electronics and have lower maximum temperature ratings
than the usual military specification. Conduction cooling, using dielectric liquids, is common in
military electronics. (Wilson, 2003)
7 Because these uses of liquid PFCs occur in closed systems that are still limited to commercial
and military systems (as opposed to the more ubiquitous consumer systems), the potential for
significant amounts of direct air emissions appears less, although maintenance and disposal of
systems cooled in this manner remains a potential concern.

300 mm wafer fab, approximately 10 percent of the equipment requires cooling; in a 200
mm wafer fab, approximately 8 percent of the equipment requires cooling and 10
percent for testing (Smith Barney, 2004).  Product literature and other reports indicate
that each TCU requires 1.5 to 50 gallons to fill, which presumably depends on heat loads
(Solid State Cooling Systems, 2003; BOC Edwards, 2003; and Tuma and Knoll, 2003).
Tuma and Knoll (2003) report that "by far the majority" of TCUs require approximately 8
to 50 gallons/system, with 5 gallons per system being typical.8

Table 2 summarizes the coolant used in the high-volume production of 200 and 300 mm
wafer fabs, excluding "top-off. The range for fab-coolant usage is large due to the
limited knowledge of TCU coolant volume requirements (a factor of approximately 5)
rather than limitations in knowledge about the number of tools that require TCUs.
Coolant cost, which ranges from $200 to $500/gallon, is  not insignificant—ranging from
approximately $38,000 to $3,500,000 per fab, depending on assumed fab size, coolant
capacity and coolant unit cost. However,  coolant life is essentially infinite and never
needs replacing if properly used, although some topping-off seems required every 1 to 2
years for a well-designed system (Tuma and Knoll, 2003).9

Table 2: Estimates of Liquid RFC Use for a 200 and 300 mm Production Fab That
Uses Only Perfluorocarbons as Coolants in Temperature Control Units (TCUs)
      Wafer size, mm            Number of TCUs3      TCU Coolant Capacity13, gal
           200                      125 units                   188-6250
           300                      140 units                   210-7000
aAssumes one TCU for each PECVD, etch (dry), ion implant and stepper tool In fab.
bAssumes a coolant capacity range of 1.5 to 50 gj
Source:  Smith Barney, 2004 for number of tools.
bAssumes a coolant capacity range of 1.5 to 50 gal/TCU; excludes "topping off (Tuma and Knoll, 2003).
4.2 Automated Test Equipment and Thermal Shock Testing
 Later in this report there is discussion of the additional fluid volumes required to "top-off TCUs
due to leaks and evaporation in systems with "free-breathing" expansion reservoirs. These
expansion reservoirs are designed to permit air saturated with the fluid vapor to leave the
expansion reservoir as the fluid expands and contracts with each thermal cycle.
9 A brief, qualitative telephone survey of TCU equipment and HTF vendors by the author during
July 2005 indicated that semiconductor manufacturers top-off TCUs at monthly intervals and that
the top-off volumes appear to be decreasing as TCU designs are improving.
10 For convenience, automated test equipment and thermal shock testing are presented together
in this report. However, the reader should be aware that HTF-usage in automated test equipment
is rapidly transitioning from high-GWP to low-GWP fluids, which is more difficult for thermal shock

1C testing is undergoing transformation. Automated test equipment (ATE) accounted for
approximately 10 percent of total capital equipment market in 2003 and 10 percent of the
capital equipment cost of a new 300 mm wafer fab (Smith Barney, 2004). Recently, the
Semiconductor Test Consortium revised the specification for the Open Semiconductor
Test Architecture (OPENSTAR) to include air-cooling and liquid HTF cooling
( HTF cooling improves thermal management during testing,
providing for improved test stability and greater throughput when testing, for example,
high-speed, large capacity memory devices (see see also that also produces similar ATE). This initial research, however, did not
produce information about the quantities of HTFs needed to fill or top-off these cooling
systems. It appears, however, that the amounts would be similar to those for TCUs.

As ICs have continued to add functionality and speed  in ever shrinking form (i.e., shape)
factors, testing has grown more vital for ensuring design and manufacturing integrity.
Thermal shock testing  (TST) is an IC-screening test to discover defects that are
attributed to differing coefficients of thermal expansion among materials (silicon, metals
and dielectrics), which can  result in 1C or packaged-IC delamination during assembly or

Component manufacturers or their subcontractors perform TST in one of two ways: (1)
alternating hot and  cold air; or (2) alternating hot and cold liquid. Manufacturers have
considered liquid-to-liquid TST too expensive because, despite its proven effectiveness,
replacing lost liquid—fluorocarbons—through evaporation is too high. However, recently
TST equipment manufacturers have designed new chambers with fluid recovery features
that lower fluid loss by up to 65 percent (ESPEC, 2004; Cincinnati Sub-Zero, 2003; ACS,

Liquid PFC use is more uncertain in TST than in TCUs. This initial research produced no
information about (1) the proportion of fabs using air-to-air TST compared to liquid-liquid
TST, (2) the  proportion of fabs that outsource TST, and (3) testing protocols, which
specifies the fraction of packaged die that require TST. Without this information, only
testing because of problems of material incompatibility between available low-GWP test-fluids
and device packaging.

very rough estimates of liquid RFC use for testing at a typical fab can be made.
Moreover, information was not found that either described or permitted estimates of the
proportion of HTFs used in 1ST that are high-GWP fluids.3
A typical high-volume 300 mm wafer fab (30,000 wafer starts per month) requires
approximately 150 tools to perform automatic testing of packaged components (Smith
Barney, 2004). However, these tools are used for both electrical and thermal tests (i.e.,
thermal tests include both liquid-liquid 1ST and air-air equipment for 1ST).

The cost of RFC liquids for liquid-liquid 1ST appears to discourage this form of shock
testing. Recent changes in test equipment designs, however, may favor the use of
liquid-liquid over air-air 1ST because the duration of liquid-liquid test can be shortened.
(More time is needed to bring the same mass of components to a fixed temperature
(without overshoot) with air than with a liquid.  Temperature uniformity throughout the
packaged die also appears to be easier with liquids than with air.)

Consider liquid use at a 30,000 wafer start per month 300 mm wafer fab. At full
production and commercial yields, this fab would produce approximately 13,000,000 to
15,000,000 die per month.11 Not all of these packaged die would  be tested. A sampling
plan that specifies testing 10 die per wafer lot is roughly equivalent to sampling 0.1
percent of all die produced, or 13,000 - 15,000 packaged die per month in this example.
If each packaged die weighs 5 g, then each month 65 - 75 kg of packaged die require
testing. Commercially available liquid-liquid TST systems handle  between approximately
1.5- 3.5 kg  of packaged die per TST test, with each test requiring between
approximately 1 to 17 hours to complete (which includes loading and unloading the
testing system), depending on the specific test (ACS, 2004; MEFAS, 2004).12  If,  for this
example, only  17 hour tests are assumed to be performed, a fab could choose to have
one or two 2kg or 3.5 kg TST systems, depending on its practices for tool utilization and
system availability. If the lower (2 kg) capacity unit is chosen for this  example,  then
11 In arriving at this range, a die of 110 m is chosen together with reasonable assumptions about
area for edge exclusions, defect densities, defect model and dies lost to test sites and alignment
marks. The size of an MPU (microprocessor unit) is approximately 110m2 (Intel, 2004).
12 Many protocols exist for liquid-liquid thermal shock testing, which are characterized by the
number of cold-hot cycles per test and the dwell time at each temperature. The test protocol used
in this example specifies 100 cycles, with hot and cold dwell times of 5 minutes each and
transition period between each cycle less than 10 seconds. Other protocols may specify as few
as 15 cycles and as many as 1000 cycles with longer or shorter dwell times (ESPEC, 1997;
ESPEC Technology Report: Evaluation Reliability, Report No. 3).

approximately 650 hours (=38 tests x 17 hrs/test, or, 90 percent of available hours) of
testing per month is required to shock test 75 kg of packaged die. A unit with 2 kg
capacity requires 18 gallons of liquid RFC (for both baths that hold the packaged die) at
a cost of $3,700 to $9,000, plus the cost to replace liquid lost via evaporation (Cincinnati
Sub-Zero, 2004).13  ACS specifies evaporative loss rates of approximately 3 g/hour of
testing for its 2 kg unit, so that over the course of a month, 1,950 grams of fluid is lost, or
approximately 1 liter (the density of RFC liquids ranges between 1.6 -1.9 kg/liter) for an
additional cost of $50 to 130/month ($600 - 1600/yr).  This estimate compares favorably
to recent published reports by ESPEC for its newest design, which cite expected annual
fluid replacements costs less  than $1,500/yr for its 2 kg system (ESPEC, 2004). The
new systems from ACS, ESPEC, and CSZ use new designs aimed at significantly
reducing evaporative losses.  ESPEC reports that it has heard of annual fluid
replacement costs as high as $20,000 for older TST system designs, approximately 40
to 100 gal/yr (assuming a liquid PFC cost of $200 to 500/gal). Depending on the capital
cost of liquid-liquid TST units  and the extent the claimed reduction in fluid loss is
achieved, increased use of the liquid-liquid TST method may occur. However, a recent
report touted Intel's purchase of a new kind of fast air-air shock testing  system  claiming
significant increases in throughput (IRTS, 2003; Neves, 2003).

4.3  Vapor-Phase Reflow (VPR) Soldering
Reflow soldering is the principal soldering method to connect surface-mount devices
(SMDs), multi-chip modules and hybrid components to circuit boards (CBs) in
commercial, industrial, and military electronic products. Among the reflow soldering
methods—simple air convection, forced-air convection, infrared, laser and vapor-phase
reflow (VPR)—only VPR uses a heat transfer fluid. However, among the alternatives, the
literature contains conflicting claims about the merits of each alternative. The principal
encumbrance of vapor-phase reflow seems to be operating cost—machine
maintenance,  chemical cost and waste management requirements (Apell and Howell,

In vapor-phase reflow soldering, the melting (reflow) of the solder is achieved by placing
the  printed circuit board (PCB) with SMDs into a chamber with the heat transfer fluid,
13 Throughout this report reference is made to the cost of liquid PFC fluids. For convenience and
consistency, the range $200 to 500/gal is adopted (Tuma and Knoll, 2003).


heating the fluid to its boiling point (the melting point of the solder, 215°C for lead-
containing solders and 20 to 35°C higher for lead-free solders), and then removing the
SMD-PCB after gradual cool-down. The total time in the chamber is 3 to 5 minutes,
depending on the mass and density of the SMD-PCB. Liquid PFCs serve as the HTFs.
Both in-line (conveyor-belt handling) and batch-type systems are used, with the latter
favored for mass-production (Intel, 1998; Concoat Limited, 2004). Device manufacturers
provide instructions for surface mounting the devices they manufacture. For example,
both Micron (memory products) and Altera (programmable logic devices) do not
recommend vapor-phase reflow for surface mounting their devices, while Intel and
Intersil do support vapor-phase reflow for surface mounting its products (MICRON, 2004;
Altera, 2002; Intel, 2004; Intersil, 2004).

Quantification of the industry's use of VPR soldering compared to reflow alternatives that
do not use liquid PFCs was not found during this initial research. Nor was information
obtained  regarding the share of HTFs represented by high-GWP fluids compared to the
low-GWP substitutes.14

Robinson (2004) reported increasing interest in and opportunities for VPR because of
the advent of lead-free soldering. He noted the virtues of VPR—less risk of overheating
or device damage—but also cited its major hurdle was relatively low throughput
(because of longer reflow time) and high liquid cost (Robinson, 2004). Nevertheless,
some 1C manufacturers, system assemblers, and manufacturers of soldering equipment
promote the virtues of VPR soldering—for both lead and  lead-free surface mount
technologies (Intel, 2004; Intersil, 2004; Gregory and Liecht2004; 3M, 1990 and 2004;
Concoat, 2004).

An earlier study report provided by 3M (1990) seems to support the more recent claims
of Gregory and Liqht (2004) and Robinison (2004). The 3M report, using information
from a study conducted by Centech of Singapore in 1990, compared the device
mounting cost per circuit board of four conveyorized (in-line) soldering machines: one
VPR soldering, and three infrared-heated reflow soldering machines (3M, 1990). In the
3M report, the cost-performance of VPR soldering was superior.  The total cost for
surface mounting devices was $0.0207/board, approximately 20 percent lower than the
14 See footnote 3, wherein it appears low-GWP fluids are not in use in VPR.

cost of the closest alternative. This and the reports of Gregory and Liqht (2004) and
Robinson (2004) suggest that VPR is popular today, despite the emergence of other
methods, which do not use expensive RFC liquids, such as forced convection heaters.
Apell and Howell observe, however, that VPR is an all but forgotten technology—
perhaps not as popular as it was in the 1970's and 1980's (Apell and Howell, 2003).

A substantial part of the cost of VPR soldering is fluid loss. The cost-performance for
VPR reported by 3M (1990),  for in-line machines included fluid loss. The FC-70
(Fluorinert, liquid PFC) loss rate was 0.12 Ib/hr (54 g/hr) of use (3M, 1990). At an annual
utilization rate of 80 percent (7,000 hrs per year) for the in-line soldering system, the
annual liquid loss is 55 gallons, an annual cost of $11,000 to $28,000. The cost of this
liquid loss contributed $0.005 to $0.01 to the total cost of surface device mounting (or,
23 to 48 percent of the total cost for surface-mount devices) (3M, 1990). Gregory and
Leicht (2004) give a lower range for fluid loss rates for in-line systems, between 10 to15
g/hr. For an assumed  utilization of 7,000 hr/yr, the annual loss rate, using Gregory and
Leicht's range, is between 10 and  15 gallons/yr, an annual cost of $200 to $7,500/yr.
Gregory and Leicht also note that VPR is particularly well suited when using lead-free
solder, currently being introduced by the electronics industry.  In this instance, according
to Gregory and Leicht (2004), it is simply a matter of using a different fluorocarbon heat
transfer fluid, one that has a  higher boiling point to  reach the higher melting point of most
lead-free solders.

Gregory and Leicht (2004) also provided a range for fluid loss from the use of batch VPR
SMD systems. For batch systems, Gregory and Leicht give a  loss rate of 5 to 10 g/hr,
which,  for 7,000 hr/year utilization, corresponds to 5 to  10 gal/yr of fluid loss, Gregory
and Leicht do not indicate whether the lower range for fluid  loss reflects the use of a
secondary fluid above the primary liquid, which, according to 3M product application
literature, reduces aerosol formation and vapor loss of the primary working fluid (3M,
2004).15 Regardless, Gregory and Leicht's is the only estimate of fluid loss from batch
systems identified from this initial research.
15 For example, 3M offers a secondary perfluorocarbon fluid, SF-2, for use above the primary
fluid. The thin vapor blanket formed by heating this fluid above the primary fluid helps to prevent
aerosol formation and vapor loss of the primary liquid to the atmosphere.

As noted at the beginning of this section, this initial research neither resolved the extent
of nor the prospects for VPR-soldering in device assembly.

4.4 Temperature Cooling Of Devices. Clusters Of Devices And Commercial-Off-
The-Shelf (COTS) Assemblies
This section summarizes the apparently growing commercial and continuing military use
of liquid PFCs for cooling high-power-density electronic systems. The literature in this
area is vast, esoteric, and dense. However, while many techniques exist for cooling, in
almost every situation, just two  fundamental objectives drives cooling these high power
systems: (1) prevent exceeding the critical device junction temperature (which, if
exceeded either degrades performance16 or produces device failure), and (2) remain
below the solder melting point (which, if exceeded produces system failure). The
exception to these two objectives for cooling electronic systems applies to high-powered
electric motors: their performance improves with liquid cooling compared to air cooling.
For example,  large electric motors vibrate less and operate more efficiently with liquid
cooling compared to air cooling. Less vibration means quieter operation, an important
result when the electric motors  power submarines (Harris and Hildebrandt, 2003).

No information was obtained about the current or expected annual consumption of RFC
liquids  for high-power electronic-cooling applications. Neither was their information
obtained that  provided a means for estimating that consumption during operation. Thus,
unlike the discussion of liquid RFC use in device fabrication, testing and assembly, the
discussion that follows is more  qualitative,  providing  little insight about the quantities of
liquid PFCs used in cooling high-power electronic systems. However, for all of these
applications, the  liquid PFCs are used in closed (often hermetically sealed) modules,
where  losses  would  seem to  be negligible except during maintenance and

For the purposes of  this report,  liquid-PFC cooling applications of high-power electronics
are separated into consumer, commercial (non-consumer), and military applications.
Furthermore,  the only consumer application for liquid-PFC cooling of potential
significance is assumed to be high-performance desktop computers.  However, liquid
cooling is not currently in use in mainstream consumer applications because, despite the
16 As junction temperature increases in semiconductor devices, switch times decrease; but at
some upper temperature the devices fails.


ever increasing heat dissipation from advanced microprocessor chips and wireless RF
(radio frequency) modules, new chip designs and innovations in thermal packaging
continue to extend the life of conventional heat-sink-assisted air cooling (Bar-Cohen,
Avram, 1999; Mukherjee, S. and Mudawar, S., 2002; CIS, 2004). Nevertheless, liquid
cooling kits are sold, apparently, to the few users who operate microprocessors at higher
voltages, which increase the processing speed (called "overclocking") and heat
generation of the PC, exceeding the off-the-shelf cooling capability. These kits use
approximately 300 ml of either water or fluorocarbon coolant to remove the excess heat
produced (see for a description of the product cooling method
offered by Asetek called Waterchill). Liquid RFC use in this application is understandably
preferable to water inside a PC. These after-market PC products do not provide
guidance to prevent the release of the liquid-PFC to the atmosphere.17

The commercial liquid PFC applications category, as described earlier in this report,
includes cooling of supercomputer, telecommunication and airport radar systems, which
as described by Tuma (2003) and others (Simons, 1996; Bar-Cohen, 1999), is 1970-
vintage technology. In these applications, bare packaged chips or modules of packaged
chips are cooled by direct and indirect contact with liquid fluorocarbons. Boiling of the
coolant at the exposed packaged surfaces provides high heat transfer, sufficient to meet
chip cooling requirements. The PFC-containing enclosure possesses internal  metal fins
to provide the means to condense the vapors and remove heat from the liquid. Air-
cooling of a cold-plate attached to the enclosure lowers the coolant temperature via
attachment to a cold-plate. Other well known cooling concepts that  use liquid  PFCs and
may be on the cusp of mainstream commercial applications are spray cooling, jet
impingement, or pumpless pool boiling of fluorocarbon liquids directly onto the chip;
these concepts provide the required heat transfer by combining a heat exchanger to
condense the coolant for reuse (Bar-Cohen, 1999; Mukherjee and Mudawar, 2002; CTC,
2004). The designs of these cooling systems envisage a very long coolant life without
replacement,  and also provide environmentally sensitive instructions for
decommissioning the system.
17 Other, more expensive approaches to cool overclocked PCs use elaborate refrigeration


Cooling of high-power military systems with liquid PFCs will likely increase. Keltec (a
Crane Aerospace company) recently announced a system that uses liquid PFCs for
continuously cooling airborne and mobile high powered microwave assemblies and
transmitters that is mounted on a metal plate through which coolant passes (see Evidence also exists of several military R&D
programs and the prospect for increased use of liquid cooling. Spray cooling, jet
impingement cooling and pool boiling appear favored for future military systems.
Isothermal Research Systems (ISR ) seems to be the dominant if not the only enterprise
doing commercial development work and shipping operational spray cooling systems to
the military ( SDA, 2000). In all of these cooling applications, the
liquid RFC appears to be contained in a reservoir from which it passes (pumped through
a cold plate in the case of simple change of phase cooling, or nozzles in the case of
spray and jet impingement cooling) to the chip-containing module. The vaporized RFC
then passes through a heat exchanger where it condenses and returns to a reservoir
after being purified by a filtration system. The life of the coolant matches the life of the
electronic system, so no maintenance or coolant replacement appears required.

This initial  research identified four estimates of air emissions resulting from using liquid
PFCs at semiconductor manufacturers. Additional information was obtained that
permitted estimating emissions from cooling, testing, and reflow soldering processes that
use liquid PFCs. The searches provided no emissions information for emissions from
liquid PFCs associated with the operation and maintenance of electronic equipment.
However, the information obtained during this initial research suggests, as previously
noted, that emissions from cooling electronic equipment are smaller than other sources,
albeit unquantifiable at this time.

5.1 PFC Emissions from HTF Use in Manufacturing ICs—Cooling and Testing
Of the four estimates for manufacturing, T&T provided an estimate for a typical but
uncharacterized high-volume fab, vintage 1999-2000. Semiconductor manufacturers
provided the other three estimates for their global operations (Seif and Hermanns, 2002;
Seiko Epson, 2002; Mitsubishi, 2002).  The estimates reported by semiconductor
manufacturers do not explicitly identify the uses—cooling, testing and VPF soldering—
included in their estimates.

T&T provide estimates of air emissions from "a typical high volume fab". T&T do not
specify the capacity (i.e., level of wafer production) for this "typical high volume fab". The
fab probably uses 200 mm wafers in light of the 1999 vintage. Neither does T&T specify
the number of pieces and  the nature of the equipment that comprise their estimate, i.e.,
T&T neither provide the proportion of equipment that performs cooling for etching,
deposition, etc. processes nor the corresponding proportion that performs testing. T&T
evidently combine the two functions and  report that "during a year when no
manufacturing capacity is  being added",  the consumption of "virgin RFC liquid" is, on
average, 500 gallons. Accepting the validity of this interpretation as  well as T&T's
assumption to use of the properties of C6F14 (specifically the density and GWP of C6F14),
these 500 gallons equate to 0.0079 million metric tons of carbon equivalent (MMTCE).18

T&T compare this estimate to the annual uncontrolled gaseous RFC emissions for a
hypothetical 200 mm high  volume fab (Beu and Brown, 2000), which was estimated by
Beu and Brown using the  Intergovernmental Panel on Climate Change (IPCC) Tier2a,
2b and 2c emission estimation methods.  T&T's comparison shows that the air emissions
from liquid RFC losses range between 17 to 19 percent of the total emissions  during 1C
manufacture (i.e.,  use of gaseous and liquid PFCs), with the smaller figure
corresponding to the Tier 2a method (Beu and Brown, 2000).

Seif and Hermanns (2002) presented the share of total global warming emissions from
"topping off liquid RFC in  2001  across AMD's worldwide operations. While
acknowledging several uncertainties in each component of the total  (which is comprised
of emissions from using energy, N2O and gaseous and liquid PFCs), the authors
estimate that 5 percent is the contribution from using liquid PFCs. Sief and Hermanns
did not indicate what figure or figures were used for the GWPs of the liquid PFCs they
use. AMD's estimate can be restated in terms of the fraction of PFCs emitted from its
facilities in 2001. The result is 15 percent for the share of AMD's global emissions from
all PFCs, i.e., from use of  gaseous and liquid  PFCs. This 15 percent figure is roughly
18 T&T use the properties of C6F14 to convert the 500 gallons/yr lost to 0.0079 MMTCE (= 500 gal
x 3.7854 I/gal x 1.7 kg/I x 9,000 (GWP) x (12 C/44 CO2)/109 kg/MMT). The GWP for C6F14 from
the Second Assessment Report is 7,400; the more recent estimate from the Third Assessment
Report is 9,000, Table 10-8, p. 10-28 for 100-year time horizon. Substituting 7,400 for 9,000
would lower T&T's estimate from 0.0079 to 0.0065 MMTCE. This adjustment is not made in this

comparable to the corresponding figures reported by T&T (2001) (Seif, and Hermanns,

Two other reports of liquid RFC emissions from semiconductor manufacturers were
obtained from this initial research. Japanese semiconductor manufacturers publish
annual environmental reports, which may include estimates of both gaseous and liquid
RFC emissions. Not stated is whether the reported emissions for liquid RFC use include
those for cooling and testing. However, they likely include both uses. Table 3
summarizes the available figures for Seiko Epson and Mitsubishi for the period  1997 -
2001. There are no  references in the report as to how RFC emissions are estimated. It
appears, however, that emissions of the gaseous PFCs used during 1C manufacture rely
on IPCC procedures and that emissions of liquid PFCs rely on estimates of annual liquid

In reporting these figures, the companies  appear to use the GWP of C6F14, as did T&T.
Table 3 shows that over the 5-year period, emissions from both gaseous and liquid
PFCs decreased. In the early years of 1997 and 1998, when emissions were practically
unchanged and uncontrolled at Seiko Epson, the share of total emissions from the use
of liquid PFCs was approximately 18 percent, close to the corresponding share
estimated by T&T. For Mitsubishi, the corresponding fraction is 14 percent, closer to the
corresponding figure reported by Seif and Hermann for AMD. In the later years, when
manufacturers were reducing emissions associated with both gaseous and liquid PFCs,
the shares of total PFC emissions from liquid PFCs decreased to as little as 5 percent
for Seiko Epson.
Table 3: Reported Air Emissions from Use of Gaseous and Liquid PFC at Seiko
Epson (A) and Mitsubishi (B) for 1997 - 2001 (10,000 Mtons, CO2)
a. Seiko Epson

Type of
Liquid PFCs
All PFCs
Liquid PFCs, %
of total





b. Mitsubishi
Liquid PFCs
All PFCs
Liquid PFCs, %
of total





Sources: Seiko Epson, 2002 Environmental Report, p. 54; Mitsubishi, 2002 Environmental Sustainability
Report, p. 35.

The estimates in Table 3 for uncontrolled emissions from using liquid RFC may also be
compared to the absolute emissions estimated by T&T for the hypothetical fab.
Converting the Seiko Epson emissions from liquid PFCs of 4.8 x104 Mtons CO2 to
MMTCE units, gives 0.013 MMTCE or 1.6 times the 0.0079 MMTCE figure forT&T's
single hypothetical high production fab. The corresponding 1999 figure for all of
Mitsubishi operations is 0.016 MMTCE, or 2.1 times the T&T single-fab figure.  These
comparisons are not only consistent with the T&T's single fab (0.0079 MMTCE) estimate
but are close to T&T's per fab estimate in light of the number of production fabs operated
by Seiko Epson in 1999 and 2000 (see footnote 22).

These published estimates  of uncontrolled emissions may also be compared to
uncontrolled values calculated for a new high-volume 200 or 300 mm wafer fab, each
with a capacity of 30,000 wafer starts per month. A 200 fab would need approximately
290 tools, of which 165 would be devoted to 1C testing; the corresponding 300 mm wafer
fab would need  approximately the total same number of tools but fewer for testing—150
instead of 165 (cf. Table 2). Using the 290 estimate as the maximum  number of tools
that might use liquid PFCs together with the usage figures given in Table 2 and Tuma
and Knoll's (2003) estimate that,  on average, the annual loss rate is <10 percent, an
estimated range of annual emissions from liquid PFC use can be developed. The
equation use to estimate the annual  emissions,  Q, is:

Q(A, T) = [290 tools (12 C/44 CO2) /109 kg/MMT] Ij pi GWR Ai Ti = 79 x 10'9 p GWP A T,

where for the purposes of these estimates, single (average) values replace the
subscripted values with p, GWP, A and T denoting, respectively, the density (at 1.7 kg/I),
the global warming potential of C6F14 (9,000), the average leak rate across all liquids
(%/yr) and T is the average quantity of liquid required to initially fill the tools (liters). The
quantities A and T are
treated as variables,
allowed to vary over a
reasonable range for each
(Tuma and  Knoll, 2003). In
simplifying the formula, it is
reasonably assumed that
the quantities are

The results are given in
Figure 1, where lines of
constant emissions
(MMTCE) are shown. The
             Contours of constant
             emissions, M MICE '
6%  |
5%  I.
4%  o
3%  a
2%  -•»
1%  ^
           0 5 10 15 20 25 30 35 40 45 50
              Liquid PFC per tool, gal
      D 0-0.005 • 0.005-0.01 D 0.01-0.015 D 0.015-0.02 • 0.02-0.025
Figure 1. PFC air emissions for high volume 200 or 300
mm wafer fab with 290 liquid-PFC-using tools, MMTCE.
GWP for C6F14 of 9,000 (SAR) is assumed for these calculations. If the emissions
estimate from T&T (0.0079 MMTCE) is accepted as roughly correct (which may or may
not include emissions from losses during testing), then, for an average liquid PFC per
tool of 25 gal/tool, the results in Figure 1 show that an annual average loss rate over all
tools that use liquid PFC falls between approximately 7 and  8 percent.

As previously mentioned in Section 4.2, the estimated annual liquid losses ranges
between 3 to 100 gal/yr for semiconductor manufacturing. The lower amount is
associated with newly designed low-loss TST equipment (ESPEC, 2004); the larger
amount was estimated by dividing ESPEC's customer reports of the annual cost of liquid
losses ($20,000) by $200/gal, the lowest reported cost of liquid PFCs for testing. This
range in lost liquid becomes 0.00005 to 0.0016 MMTCE of air emissions, assuming the
fluid density is 1.7 kg/I and GWP is 9,000 (C6F14).19 This range is consistent with T&T's
estimate for liquid PFC loss of 0.0079 MMTCE.
19 The figure 0.0016 MMTCE = 100 gal/yr x 3.785 I/gal x 1.7 kg/I x 9,000 x (12 C/44 CO2)/109
kg/MMT. The figure 0.00005 MMTCE = (3 gal/100 gal) x 0.0016 MMTCE.

In Section 5.1, it was noted thatT&T's estimate of 0.0079 MMICE was likely to include
emissions from both the losses from temperature control and testing equipment,
although the portions were not provided. Using the range 0.0005 to 0.0016 MMTCE from
testing, suggests that as much as 20 percent of T&T's estimate could include losses
from testing (= 0.0016/0.0079). However, if T&T's estimates excludes emissions from
testing, then T&T's estimate could be increased to 0.0095 MMTCE (=0.0079 +  0.0016).
It should also be noted that the estimates for TST emissions, by definition, exclude HTF
losses from ATE.

5.2 PFC Emissions from Vapor-Phase Reflow Soldering
This section considers  emissions during VPR soldering, the arguably popular and only
surface mounting technology among several that use liquid PFCs. Because no
information about emissions from either assembly or product manufacturing facilities
(which use VPR soldering) was obtained from the initial research, this section presents
calculated estimates.

As shown next, such an estimate shows that if all packaged ICs from a high-volume
wafer fab were assembled using VPR soldering, those emissions could equal 55 - 66
percent of emissions from using liquid PFCs in equipment cooling and device testing, or
8-11 percent of the total emissions associated with electronics manufacturing. Those
emissions from VPR soldering, however, most likely occur at locations other than where
the packaged ICs were manufactured and tested. The basis for this range is described

In the section on VPR soldering, a large range for liquid loss during soldering was
presented—5 to 55 gallons per 7,000 hours of soldering. For the purposes of estimating
annual loss from a facility that might use vapor-phase reflow, the results from the
comparative study published by 3M are consistent and useful.

According to the 3M-published study, 55 gallons would be lost over the course  of 7,000
hours. The published study also reported that 620 (7" x 6") boards were completed per
hour. Over the course of the 7,000 hours, assuming approximately 7 mounted devices
per board, 30,380,000  tested, packaged ICs would be soldered while loosing, through

evaporation, 55 gallons of liquid PFCs.20 Thus, approximately 5 - 6 similar soldering lines
would be needed to surface-mount the annual quantity of packaged ICs produced at one
modern high-volume wafer fab (e.g., 5.1 soldering lines = [13,000,000 ICs
mounted/month x 12 month/yr]/[30,380,000 ICs mounted/year/soldering line]. Together
these 5 - 6 lines would emit (each at 55 lost gallons/yr) the equivalent of 275 to 330
gallons of liquid PFCs. This volume equates to 55 - 66 percent of the emissions from
liquid RFC loss that T&T estimated for a typical high volume fab (e.g., 330 gallons
evaporated = 0.0052 MMTCE compared to T&T's estimate of 0.0079 MMTCE)21.

Table 4 summarizes estimates of RFC emissions associated with electronic
manufacturing that were identified in or developed using  information from this initial
Table 4: Estimates of Uncontrolled RFC Emissions per High-Volume Electronics
Manufacturing Facility


Liquid (loss)
liquid shock



of Total



Estimate 2



% of Total




Bue & Brown, 2000

Tuma &
Tousignant, 2001
This work

This work

Hypothetical, 1999
vintage hi-volume fab,
IPCC Tier2a method

Typical 1999-2000 hi-
volume fab, 500 gallons
annual loss
Only liquid-liquid testing
done; 0.1 % of all
packaged die tested
All packaged products
surface mounted using
VPR soldering

 Totals may not sum to 100 because of rounding.

The estimates of emissions in Table 4 are initial estimates of uncontrolled emissions
associated with electronics manufacturing. The estimates conceal many assumptions
although those assumptions have been identified explicitly throughout this report. Those
assumptions fall into two broad categories: (1) fab characteristics, such as technology
vintage, capacity, utilization, wafer size, tool mix, heat management practices, die size,
yield,  sampling protocol, etc., and (2) HTF usage, such as the proportion and magnitude
20 In this estimate, the number of mounted-devices per board is important. The figure 10 devices
per board is similar to figures from Concoat, which reports that, for its vapor-phase reflow Delta-5
system, 48 devices are mounted on a 50 cm by 40 cm board or roughly 2.4 devices per 100 cm2
area of substrate (cf. 7 70 cm2 or 2.6 per 100 cm2 in this estimate).
21 0.0052 MMTCE = 330 gal x 3.7854 I/gal x 1.7 kg/I x 9,000 x (12 C/44 CO2)/109 kg/MMT.

of HTF use in cooling, testing and soldering; the quantities of HTFs used and lost via
evaporation from each use; whether HTFs have high or low GWPs (and what that value
is22); and what the values for fluid densities of the HTFs should be. The estimates in
Table 4 are those associated with a specific single fab—a large product fab—a proxy for
facilitating comparisons of fab-infrastructure emissions  from high-GWP materials rather
than inter-fab emissions of high-GWP emissions. The estimates of emissions in Table 4
represent the emissions during 1C manufacture (e.g. from etching, chamber cleaning,
chillers, and testing provided testing is done at the fab) as well as the emissions
associated with 1C and product manufacture (e.g.,  testing and VPR soldering when both
are outsourced). Finally, the estimates were developed using a single density (1.7 kg/I)
and single GWP (9000) to convert volumes of liquids to MMTCE units. The 1.7 kg/I value
could be  approximately 10 percent lower or 6 percent higher. In addition, liquid PFCs
can have higher or lower GWPs, from as low as 300 to as high as 11,000 (T&T, 2001).

The column labeled Estimate 1 denotes the low-end of the range for VPR soldering,
while the column labeled Estimate 2 reflects the higher end of the range. However, both
estimates assume all ICs are mounted using VPR  soldering.

As indicated in Table 4, approximately 90 percent (87 - 91 percent) of all emissions
associated with electronics manufacturing occur during 1C manufacture—before
packaging and surface mounting in products and, perhaps, before testing. This share of
all emissions is the sum of emissions from both gaseous and liquid PFC use, during 1C
manufacture, tool cooling, and possibly testing. In addition, it should be recalled that
some testing is outsourced, so that, even if liquid-liquid shock testing is employed and
included in T&T's estimate, those emissions may not always occur at the 1C fab.

The second observation about the estimates in Table 4 is that emissions from liquid-
liquid shock testing (0.0016 MMTCE) appear relatively  small (<5 percent of the total),
regardless of where they occur (i.e., whether they occur during 1C manufacture or
elsewhere during testing). Furthermore, if only air-air shock testing is performed,
emissions of high-GWP from testing would be nil. Equally important, however, is the
effect on  this estimate of emissions of using alternative sampling rates. For liquid-liquid
22 In choosing a specific value for a GWP, for example, is it taken from the IPCC Second
Assessment or Third Assessment Report?


TST, emissions are directly proportional to sampling rate, which was taken as 0.1
percent for the estimate (0.0016 MMTCE) in Table 4. The only support for this sampling
rate was its consistency with expected annual economic value of fluid losses specified
by a TST equipment manufacturer. If there were no other changes when making this
estimate, except that the sampling rate were assumed to be 0.2 percent (twice the 0.1
percent estimate), the emissions would be 0.0032 MMTCE, twice the estimate in Table
4. However, as that sampling rate increases, fluid  replacement costs also increase,
which at some sampling rate would limit the use of liquid-liquid TST in favor of air-air

The final observation about the entries in Table 4 applies to VPR soldering. At
approximately 10 percent of total emissions, it is notable. However, the  evidence for the
extent of VPR soldering for surface mounting devices is conflicting and  perhaps
changing.  More than one report suggested that VPR soldering technology was passe.
Reports promoted the virtues of convection ovens, which avoid the high cost of PFC
liquid  replacement, provide high throughput and have much less waste  disposal (except
for lost heat associated with convection ovens). However, other reports promoted the
virtues of VPR soldering (especially for lead-free soldering), which is emerging rapidly as
the required method for surface-mounting devices. Thus, it could be argued that  PFC
emissions from VPR soldering might increase.

The initial  research suggests a fragmented, dynamic and global distribution of sources
(i.e., users) of liquid PFCs in the electronic sector. Three segments of this sector—
semiconductor manufacturers, electronic product manufacturers and electronic
manufacture service (EMS) providers—would appear to account for most if not all of the
sources of emissions from using liquid PFCs.

Figure 2 schematically depicts the relationship among these three segments. The figure
is intended to illustrate emission source boundaries; circle sizes qualitatively illustrate
relative sizes  of emissions from the sectors.

                                  Electronic product
       Vertically Integrating manufacturer
Figure 2. Relationship of liquid RFC users in
electronics product manufacture, showing
vertically integrated manufacture, semiconductor
The outer circle in Figure 2
denotes the entire
electronics  sector, which
would include military
contractors and system
integrators; the next largest
circle (within the outer
circle) denotes fully
vertically integrated
companies, i.e.,
companies, like IBM,
Motorola, Sony, Sharp,
Hitachi and Toshiba, for
example, who manufacture
ICs—and in doing so use
liquid RFC coolants—as
well as test and assemble
those and merchant ICs into commercial products, which may also use liquid PFCs
depending  on the testing method and surface mounting technology. Within that circle are
the semiconductor manufacturers (including foundries) which manufacture ICs (and also
use liquid RFC coolants) but which may either perform all testing and assembly or
subcontract portions to EMS providers, depicted by the dashed circles.

Figure 2 shows EMS providers as performing test or assembly services for vertically
integrated manufacturers and  semiconductor manufacturers or as providing both test
and assembly services to these manufacturers, which is denoted by T&A where the
dashed circles overlap. Finally, also illustrated in Figure 2 by the overlapping cooling and
testing segments, is the use of liquid RFC coolants in test equipment. Not shown in
Figure 2 are fabless companies; either 1C foundries or  EMS providers perform testing
and assembly services for fabless companies.

There are many users of liquid PFCs. As previously described there are many
technologies used in testing and assembly that do not use liquid PFCs, and no

information was obtained during this initial research that offered insights of their
historical, current or projected use compared to those that use liquid PFCs. In seeking
this information, some searches were made at websites of three trade associations that
represent semiconductor manufacturing industry: Semiconductor Equipment Materials
International (SEMI,, MicroElectronic Packaging and Test Engineering
Council (MEPTEC,, and Surface Mount Technology Association
(SMTA. SEMI, with its 2200 members represents the entire global
semiconductor industry. MEPTEC and SMTA are smaller entities that aim exclusively at
test and assembly. MEPTEC, a 25-year old association with 350 members (which
includes many major semiconductor manufacturers), aims exclusively at advancing 1C
assembly and testing technologies. SMTA, a 20-year old association with 535 members
(which includes a few relatively small semiconductor manufacturers) aims exclusively at
surface mount technologies.

This fragmented picture of electronics sector is also dynamic. While industry reports
often describe the accelerating trend toward more foundries and fabless companies,
there are fewer reports about the increasing practice of outsourcing both testing and
assembly to EMS providers. Three companies appear to have approximately 40 percent
of this global market, ASE Group, Amkor Technologies, Inc. and STATS ChipPac (a
recent merger between STATS, of Singapore, and ChipPac of Chandler, AZ) (Khadpe,
2004). Seven  of the 10 largest companies have headquarters in Taiwan, Singapore and
Malaysia (Khadpe, 2004). In 2003, for example, the share of the outsourced
semiconductor assembly and test (OSAT) market associated with IDMs accounted for
revenues of $6.5 billion (or approximately 25 percent of I DM assembly and test
expenses). By 2007, those revenues are projected to double to $13 billion, which would
then account for 30 percent of IDM assembly and test  expenses and which represents
an annual  compound average growth rate of 20 percent,  almost 3 times the
corresponding growth rate of the semiconductor market.  During this same period, the
total OSAT market is projected to grow from $26 billion in 2003 to $43 billion in 2007

Despite this fragmented and changing usage of liquid PFCs, it appears that, for the next
few years, the semiconductor industry will remain a major source of emissions. The
remainder of this
section addresses the
global distribution of
semiconductor fabs
that likely use  liquid
PFCs as coolants in
                            D US • Japan n Europe n Taiwan • South Korea n China • Others
                           Figure 3. Global distribution of 300 fabs in 2004 that likely use
                           gaseous and liquid PFCs during 1C manufacture.
TCUs and in 1C testing
and assembly.

According to the
January 2004 Edition of
SEMI's World Fab Watch, approximately 380 of 410 production fabs worldwide use
gaseous PFCs.23 Of the 380, approximately 300 fabs in 2004 are likely to use process
technologies that employ TCUs during 1C manufacture.24'25 Figure 3 shows the global
distribution of those
300 fabs. The
capacities of this
population of 300
production fabs span
             300 —            Source: January'CM Edition, World Fab
                           Figure 4. Distribution of 2004 fab capacities (wspm) that
                           likely use gaseous and liquid PFCs.
70,000 wafer (200 mm
equivalent) starts per
month. Figure 4 shows
the capacity distribution
of these 300 fabs. Note
that this distribution of capacities shows that 40 percent of the 300 production fabs are
large. The distribution shown in Figure 4 together with the information in Table 4 can be
used to estimate, roughly, global PFC emissions for 2004 — an estimate that includes
  The 410 fabs include production fabs in operation in 2004 that use silicon wafers and ICs
fabricated with feature sizes <1 urn.
24 The 300 fabs are those fabricating ICs with feature sizes <0.5 urn, the technology node when
TCUs began using liquid PFC coolants.
25 In this analysis fabs that do not use gaseous PFCs are excluded because it is reasonably
assumed that such fabs would use water/glycol TCUs rather than TCUs with liquid PFC HTFs.

gaseous and liquid RFC use during 1C manufacture. This rough global estimate starts
with 0.05 MMTCE from Table 4 (rounded to one significant figure to reflect the
roughness of these estimates), the total RFC emissions during manufacture from a large
fab. Assuming 0.05 MMTCE is the average total emissions for the 120 large fabs in
Figure 4 gives 6 MMTCE (=0.05 x 120) for the global emissions from this group of high-
volume fabs. The average capacity of the medium size fab is approximately 15,000
wspm (200 mm) or, if emissions are assumed to be proportional to fab size, 0.025
MMTCE per fab. This group of 95 fabs, therefore, is estimated to contribute 2.4 MMTCE
to the global total. The remaining group of small fabs contributes 0.6 MMTCE to the
world total (=0.007 MMTCE/fab x 85 fabs, where 0.007 MMTCE is emissions  of the
average size fab [4,000 wafer starts per month, wspm] in this group of small fabs). Thus,
in 2004, the total global (gaseous plus liquid) RFC emissions from semiconductor
manufacturing are estimated roughly at 9 MMTCE.26

The next section examines alternatives methods for estimating emissions from using
liquid PFCs.

The preliminary research identified three alternative methods for estimating annual
emissions from liquid RFC use. The first method relies on the share of total emissions—
17 percent on an MMTCE basis—first presented by T&T (2001) and apparently similar to
the corresponding share estimated from  reported emissions by Seiko Epson and
Mitsubishi. The second method uses T&T's estimate  of 500 gallons of liquid RFC lost for
their typical high-volume production fab, which could  be scaled to account for varying fab
sizes. The third method relies on manufacturers to  report estimates of evaporative
losses of liquid PFCs. The virtues of each method are presented in the remainder of this

7.1 Method 1:17 Percent of all PFC Emissions Originate from Liquid PFC Losses
26 An independently developed and arguably more precise estimate of global emissions is
provided by PEVM for 2004: 8.4 MMTCE, which only includes uncontrolled gaseous PFC
emissions. Thus, it appears that 0.6 MMTCE (or 7 percent = 0.6/9 of the total) might be attributed
to using liquid PFCs.

As discussed in Section 5 and shown in Table 3, emissions due to uses of liquid PFCs
contribute approximately 17 percent to total RFC emissions (from both liquid and
gaseous RFC usage). Using the 17 percent is simple, but unreliable. For the method to
be reliable, this (or any) fraction must remain constant over time and constant with
capacity. As shown by T&T (2001) the share not only cannot be assumed constant but is
also increasing as semiconductor manufacturers implement measures to reduce
emissions from using gaseous PFCs. The changing fraction is also evident in  reports of
Seiko Epson and Mitsubishi (cf. Table 3). In  addition, as shown at the end of the
previous section, evidence indicates the share decreases as  fab capacity decreases
(see footnote 27). Finally, even if these reasons could be overlooked, this method could
not account for using alternative gaseous or liquid PFCs, which would likely have GWPs
that differ with those reflected in the estimated 17 percent.

7.2 Method 2—500 Gallons Is Annual Loss per Fab. Across All Fab Sizes
The second method—assuming 500 gallons/fab for the loss across all fabs—while also
simple and capable of accounting for changing values for GWPs, has serious  limitations.
While this 500 gal/year estimate for a 1999-2000 vintage fab  is comparable to the
published figures for Japanese semiconductor manufacturers, it is unreasonable to
assume the losses for medium and small fabs would be the same.27  To overcome this
limitation, some method of scaling the 500 gallons figure is required. Second,  even if
some acceptable method of scaling were developed, the figure would likely not remain
constant over time. As T&T suggest, the 500 gallons amount was probably considerably
larger in the late 1990s, when semiconductor manufacturers failed to initially appreciate
the propensity of liquid PFCs to leak (because of their very low surface  tension) when
used simply as drop-in substitutes for water/glycol coolants.

7.3 Method 3—Estimating Fab Specific Annual Losses Of Liquid PFCs
The third method is methodologically simple and reasonably relies on manufacturers to
report annual losses. Companies representing approximately 70 percent of global
semiconductor manufacturing capacity already estimate and  report annual emissions of
22 Taking, for example, the estimate 4.8 x 10 Mtons for emissions from liquid PFCs for all of
Seiko Epson for 1998 equates to 1,000 gallons lost (using the properties of C6F14, GWP of 7400
[SAR] and density of 1.7 kg/I). Because Seiko Epson operated two large fabs during this period,
T&T's estimate of 500 gals/yr for one fab of comparable size and vintage is remarkable.

gaseous PFCs during 1C manufacture. In this method, annual losses are calculated
using the formula, for each liquid RFC used,

Annual Loss (gallons) = Annual emissions
                  = "Top-off' of TCUs, testing and assembly equipment + losses from retired
                     equipment - on site disposal from retired equipment
Mathematically, this can be written as mass of emissions as a function of RFC
       PFCi (kg) = Pi {1^(1) + Pit(l) + Cit (I) - Nit(l) - lit(l) - Dit(l)}
       Pi    = density of liquid PFCi. in kg/I, which is available from suppliers
       ljt-i(l)  = the inventory of liquid PFCi at the end of the previous period in
       Pit(l)   = purchases of liquid  PFCi during the period,
       Cit(l)   = nameplate capacity of retired equipment
       Nit(l)   = quantity of liquid PFCi for filling new equipment,
       lit(l)   = inventory of liquid PFCi at the end of the period in containers, and
       Dit(l)   = amount of PFCi disposed from retired equipment during the period.

This formula is methodologically rigorous under the assumption that it accurately reflects
the industry's practice for managing  its stores and uses of liquid PFC HTFs. It should be
noted that 3M has a policy of cost-free pickup and return (to 3M) of used high-GWP
HTFs in the U. S. provided quantities for return equal at least 30 gallons (see Thus, in practice disposal is likely unimportant. In addition, to the
extent material from decommissioned equipment is suitable for  new equipment,
recycling may be important. Tracking emissions in this manner is unlikely burdensome
because manufacturers likely carefully track liquid PFC purchases and inventory as well
as minimize extraneous losses due to the high cost of these liquids and semiconductor
manufacturers' practice of rigorous cost control.29

An uncertainty with any method is the choice for GWP. This initial research indicates
only 3M (which  used accepted IPCC methodologies for estimating emissions) has
29 This initial research did not reveal any pick-up and return policy offered by Solvay Solexis in
Europe, the headquarters' location or in the United States.

estimated GWPs for some of its Fluorinert and HFE (hydrofluoroether) fluids. However,
this initial research did not identify the procedure 3M follows to estimate GWPs. In
several instances (see T&T, for example), 3M suggests that users follow U. S.  EPA's
"recommendation" of using the GWP for C6F14 when no other value is available.
However, the reference 3M cites for the source of this "recommendation" contains no
explicit recommendation. In a footnote to Table 3-33 of the EPA, Inventory of
Greenhouse Gas Emissions and Sinks: 1990-2000 is a note that indicates "for
estimating purposes, the GWP value used for [a diverse collection of PFC/PFPEs for
solvent applications] was based on  C6F14" (EPA, 2002). It should also be mentioned that
the GWP for C6F14 from the Second Assessment Report is 7400 compared to the higher
value, 9000, provided in the Third Assessment Report.

This section proposes and summarizes additional work aimed at clarifying the
•   Distribution of users of liquid PFC HTFs in the electronic sector
•   Proportion of technologies that use liquid PFCs (and trends in those proportions) and
•   Modes and amounts of emissions from liquid PFC uses.
This additional work should provide robust knowledge on which to base improved
numerical estimates (with less uncertainty) of the contribution of liquid PFC emissions
from the electronic sector.

The lack of readily available past and current estimates on annual, global quantities
associated with liquid  PFC HTF use in the electronic sector frustrates efforts to compare
that use to similar uses in other sectors or different uses altogether. For example,
chillers are  prevalent in the manufacture of Pharmaceuticals but this initial  research
could not address the extent to which liquid PFC HTFs might be used. In addition, there
are direct medical applications of liquid PFCs in which all of the PFCs are lost to the
atmosphere. Knowledge of the share of liquid PFCs use in the electronic sector would
help resolve these uncertainties.

The lack of information about global use of liquid PFCs suggests that the uses  may be
too small— especially compared to  quantities of other specialty chemicals used in the

electronics sector—to attract the interest of industry and market analysts.30 However, the
high cost of liquid PFCs indicates these are high-value chemicals. Their unique physical
and chemical properties, together with 3M and Solvay Solexis' recent efforts to find and
market low-GWP substitutes, suggests the uses and sales of liquid PFCs are sufficiently
important to the future business of each enterprise to attract investment.

Much of the analysis in this report about the magnitude of emissions from liquid PFC use
rests on T&T's estimate that the annual loss from a typical large high-production fab is
500 gallons of liquid PFC, which  has the properties (density and GWP) of C6F14. The
only corroboration of 3M's 500 gallon estimate comes from comparing that quantity,
expressed as a fraction  of total emissions from a 1999-2000 vintage  typical high-volume
production fab, with three other estimates: one from AMD global operations and two
from Japanese electronic manufacturers.  In the case of the estimates published by the
Japanese manufacturers, the comparison of absolute quantities is encouraging though
still speculative since they reflect enterprise totals rather than fab-specific quantities.
Hence, the veracity and robustness of T&T's estimate warrants further investigation.

Further work also seems warranted regarding the share of liquid PFCs (and low-GWP
substitutes) used for cooling, TST, and reflow soldering. As indicated previously, claims
over the benefits of alternative methods are conflicting.  This confusion is expected given
that equipment manufacturers and  material suppliers made the claims. While it seems
intuitive that, in light of the high unit cost of liquid PFC, users might seek alternatives to
using liquid PFCs, it has been and is the practice of the electronic sector to maximize
throughput provided key technical requirements are achieved even if the cost of a
specific item is relatively high.31 However, during periods when demand is low,
30 Consider for example the likely largest current buyer of liquid PFCs, 1C equipment
manufacturers. Use T&T's estimate of 500 gallons/yrfor replacement. Then assume 10 new
fab/year being built (which seems generous at $2-3 billion each and global capital expenditures
for all new and existing fabs at $40 - 50 billion/year), with each having 290 tools that use liquid
PFCs (Figure 3, this report) at 25 gal/tool, on average. Add replacement and new sales. The
result is an annual global volume of roughly 2.5  million pounds of liquid PFCs. These 2.5 million
pounds can be compared to annual global NF3 supply of approximately 14 million pounds of NF3
in 2004, which is one of 7 gases used  in 1C and  LCD manufacture. Of course, this estimate of 2.5
million pounds liquid PFC consumption neglects other uses, which, even if they were comparable
to or a small multiple of the use by 1C manufacturers, would result in a higher estimate for the
liquid PFC volumes but still small compared to annual volumes of other specialty chemicals.
31 The use of NF3 in chamber cleaning is a noteworthy example of industry thinking. Initially, the
cost of NF3 exceeded the cost of the dominant alternative (C2F6) by roughly 20-fold. However, the


throughput matters less than costs, so electronic manufacturers employ a mix of
alternatives to hedge economic and technological risks.

Finally, additional information about the annual losses of liquid PFCs during 1ST and
VPR soldering is warranted. This review revealed not only wide variability—a factor of 10
in both 1ST and VPR soldering uses—but included claims that losses were being
reduced through equipment redesigns as well as, in the case of VPR soldering, through
the use of a secondary Fluorinert fluid.
benefits of NF3 more than compensated for its much higher cost. Today, production volumes of
NF3 are approaching those of C2F6.


3M, 1990; available

3M, 1991; U. S. Patent 49997032

3M, 2004; Product Literature, see

ACS, 2003. Product literature available at shock liquid.htm

Altera, 2002; "Reflow Soldering Guidelines for Surf ace-Mount Devices", Application Note
81, June, 2002 veer. 4.

Apell and Howell, 2003; "The Evolution of Mass Soldering, available at

Bar-Cohen, Avram, 1999; "Thermal Packaging for the 21st Century: Challenges and
Options, 5th Termini-International Workshop on Thermal Investigations of ICs and
Systems, Rome, Italy, October 3-6, 1999

Beu and Brown, "Analysis of IPCC Emissions Estimating Methodology", presented at
Partnership for RFC Emissions Reduction, Silicon Southwest, Austin, Texas, and 16
October, 2000

BOC Edwards, 2004. available at control/Chillers.cfm~content

Cincinnati Sub-Zero, 2003. Liquid-liquid thermal shock specifications available at ; for additional
product specifications see

Concoat Limited, 2004; "Delta-5 high-productivity in-line and batch vapor-phase
systems" available at

CTS, 2004; "Technology Advancement: Spray Cooling Opens the Door to Next
Generation  Heat Transfer, available at

Deutsche Bank, 2002. Analysts confidential report entitled "Semiconductor Capital
Equipment", March 2002.

DuPont, 2004; see

ESPEC, 2004 "New Chambers Make Liquid-to-Liquid Thermal Shock Affordable for
Semiconductor Testing" available at html?prid=27671&az=72744360

Gregory, M. and Leicht, G., 2004; "Vapor Phase Returns for Lead-Free Soldering"
presented at NEPCON Brighton Process Technology Seminar, 27 May 2004,
Brighton, England.

Harris and Hildebrandt, 2003; "Spray Cooling and Electronic Equipment", COTS
Journal, November, 2003, p. 51

Hewlett Packard, 2002; HP Labs, News, Archives, Using Injet Technology to Cool Chips,

Incopera, F. and DeWitt, D. 1996, Introduction to Heat Transfer, 3rd Ed. John Willey &
Sons, NY., 1996.

Intel, 1998, "Surface Mount Technology", Chapter?, Packaging Databook, 1998

Intel, 2004; "Assembly with Intel Components" available at, 2004;

Intersil, 2004; "Guidelines for Soldering Surface Mount Components to PC Boards",, 2004

IRTS, 2003. Integrated Reliability Test Systems, available at coe unit.asp?service=coe  rtcc&unit=9

Khadpe, S, 2004, "The Global Packaging Foundry Business 2003: Suppliers Are
Revving Up for a Boom Year", ChipScale Review,  May-June  2004 available at

MEFAS, 2004.  Reliability Qualifications and Testing Services,  available at detail.html

MICRON, 2004; "Recommended Solder Parameters", TN-00-15, 2004

Mitsubishi, 2002. "Environmental Sustainability Report: 2002,  Mitsubishi Electric
available at Report  2002.pdf

Mukherjee, S. and Mudawar, S., 2002, "Pumpless Loop for  Narrow Channel and Micro-
Channel Boiling, J. of Electronic Packaging, 125 (3), 431-441

Neves, 2003. "Highly Accelerated Thermal Shock (HATS) Testing for PCD Hole
Reliability, December 2003. available at
ipc  1103.pdf

Robinson, 2004; Robinson, K. "Impact of Lead-free on the SMT Soldering Equipment
Market", Surface Mount Technology (SMT), September, 2004

SDA, 2000; "Spray Cooling Technologies Market Investigation" Systems Design and
Analysis, 29 September 2000)

Seif, D. and Hermanns, S., 2002, "Progress and Challenges in PFC Reductions at
AMD", presented at Partnership for PFC Emissions Reduction,  Semicon  Southwest
2002, 14 October 2002, Austin,  Texas

Seiko Epson, 2002. "Environmental Report: April 2001 - March 2002" available at e.pdf

Sheppard, M. 1999. "Temperature Control for Semiconductor Cluster Tools"
Semiconductor International, 1 December  1999 available at
http://www. reed-

Simons, R. E. 1996; "Direct liquid immersion  cooling for high power density
microelectronics", Electronics Cooling, Vol 2,  May 1996 at

Smith Barney, 2004. Analyst confidential report entitled "Semiconductor Equipment:
Supply/Demand Model", June 14, 2004.

Solid State Systems Cooling, 2003. available at

Solvay Solexis, 2004. see

Tuma, 2003. Tuma, P. E.,"Fluorinated Heat Transfer Fluids for Spray Cooling: History
and Resources", presentation at Advanced Liquid Cooling Electronics Workshop,
Scottsdale, Arizona, 13 May 2003

Tuma, P.E. and Knoll, S., 2003; "A comparison of fluorinated and Dl/glycol heat transfer
fluids", Solid State Technology, 9 May 2003. available at

Tuma, P. E. and Tousignant, 2001; "Reducing Emissions of PFC Heat Transfer Fluids"
presented at Semicon West, Symposium on ESH Challenges and Analytical
Methodologies, San Francisco, CA 16 July 2001.
available at

Werner, 2005; K. Werner, e-mail of May 4, 2005 to C. S. Burton.

Wlson, 2003; J. Wlson, "Challenges in thermal Control of Military Electronic Systems",
Electronics Cooling, Vol.  9, no. 2, 2003, available at