Stainless Steel for Semiconductor Applications

bentgalaxyΗμιαγωγοί

1 Νοε 2013 (πριν από 3 χρόνια και 5 μήνες)

168 εμφανίσεις

Stainless Steel for Semiconductor
Applications
Sunniva R. Collins
Research Metallurgist
Swagelok/Nupro Company
4800 E. 345th St.
Willoughby, OH 44094
440/951-7100
ABSTRACT
Figure 1 SEM image of a semiconductor chip at ap-
AISI 316L is the standard alloy used in several
proximately 30× × × ×
product forms for components in gas delivery systems
for semiconductor manufacturing. At present, customer
capital expenditure to construct a fab is usually recov-
and market requirements are driving processing from
ered within two to three years of starting operations.
AOD/VAR to VIM/VAR, and restricting chemistries to
a narrow range. This paper will discuss industry re-
Semiconductor chips are manufactured by a
quirements, and effects on product physical metallurgy
multi-step process involving chemical vapor deposition,
and manufacturability (machining, electropolishing, and
photolithography, chemical stripping, and other proc-
welding).
esses, to create an engineered structure on the surface of
a specially prepared silicon wafer. An example is shown
in Fig. 1. Many of these processes involve gases or va-
1. INTRODUCTION
por phases.
The gases used in the manufacture of semicon-
The semiconductor equipment industry is a
ductor chips include both bulk and specialty gases, as
global US$55 billion equipment and materials indus-
shown in Table 1. These gases can also be categorized
(1)
try. This segment of the semiconductor industry sup-
as inert, corrosive, or toxic/pyrophoric. For each of
plies gas delivery systems and manufacturing equipment
these categories of gases, seal integrity (health and
to fabrication sites, or fabs. This market is projected to
safety) and contamination are industry concerns. For
grow in the next several years, as nearly 400 fabs will
example, arsine and phosphine, widely used in ion im-
be equipped between now and 2001, and another 170
plantation, are highly toxic, with threshold levels in the
are under construction. Much of the growth will take
ppm and ppb ranges, respectively. Silane, used in sili-
place in the Asia-Pacific region, with facilities and joint
con epitaxy and some silica deposition processes, burns
ventures under development in Taiwan, Singapore, Ko-
spontaneously on exposure to air. The halogen gases,
rea and China. The current cost of constructing a fab is
such as hydrogen chloride or hydrogen bromide, will
between US$1 and US$2 billion. Approximately three-
corrode stainless steel when moisture levels exceed a
quarters of that cost is due to the chip manufacturing
few ppm. Corrosion introduces the additional conse-
equipment (tools) and tubing or components for gas
quence of potential particle generation, which in turn
delivery, and only about one-quarter is due to facilities
may be responsible for yield reduction.
cost. Within one fab, there are usually 350 to 600 proc-
ess tools, along with several miles of piping to support
It is commonly accepted by chip manufacturers
tooling. Depending on the profitability of the product
that “a particle one-tenth the size of the thinnest line can
being manufactured, the (2)
do damage to a chip and cause failure.” As line widths
are approaching 0.25 µ m, the size of the “killer defect”
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619Table 1. Gases Used in Semiconductor Manufacturing
Category Gases
Corrosives
• ammonia • hydrogen chloride • silicon tetrafluoride
• boron trichloride • hydrogen fluoride • trichlorosilane
• boron trifluoride • phosphorous pentachloride • tungsten hexafluoride
• chlorine • phosphorous oxychloride • hydrogen bromide
• dichlorosilane • silicon tetrachloride
Toxics • arsine • germane • stibine
• chlorine • phosphine • the corrosives
• diborane • silane
Pyrophorics
• diborane • phosphine • silane
• dichlorosilane
Flammables • arsine • germane • silane
• diborane • hydrogen • trichlorosilane
• dichlorosilane • phosphine • stibine
Bulk Gases • CDA (clean dry air) • hydrogen • nitrogen
• argon • helium • oxygen
becomes even smaller, about 0.025 µ m. (For compari- ardous gases contained. The assembled systems are leak
tested and dried down to very low moisture levels prior
son, the thickness of an average human hair is between
to service.
50 and 120 µ m.) These stringent manufacturing re-
quirements mean that all components in contact with the
This paper will discuss current industrial re-
process gases must be engineered to minimize or elimi-
quirements of AISI 316L for semiconductor applica-
nate potential sources of system contamination.
tions, as well as some of the emerging materials discus-
sions in the industry, including:
There has been an increased emphasis on the
cleanliness and composition of the materials of con-
• low manganese 316L for welded assemblies used in
struction for gas delivery systems. The semiconductor
corrosive gas delivery systems
industry has standardized on AISI 316L stainless steel
for the majority of gas delivery applications. In terms of • cleanliness of materials produced by different melt
performance, cost and availability, this alloy is the op- methods
timum choice. 316L is available in a wide variety of • chromium passivation for enhanced corrosion resis-
product forms, including bar stock (rounds, hexes, tance
squares, flats, and rectangles), pipe and tube stock
• use of other alloys, such as ferritic stainless steels.
(seamless and welded and drawn), and near-net shape
components (forgings and castings).
2. CHEMISTRY REQUIREMENTS AND
Components for use in semiconductor gas de-
RESTRICTIONS
livery systems and manufacturing equipment are typi-
cally machined, electropolished and welded. Semicon-
ductor-grade tubing is drawn in a variety of sizes, usu-
316L stainless steel is mostly iron, with sig-
ally from seamless tube stock. It is also electropolished.
nificant alloying additions of chromium, which gives
Prior to assembly into systems or equipment, compo-
the metal its “stainless” or corrosion-resistant character-
nents and tubing are specially cleaned to remove all
istics, and nickel, which stabilizes the austenite and
contaminants from the surfaces that will come into con-
makes the metal nonmagnetic and tough. Table 2
tact with the system fluid. The systems and equipment
shows the standard compositional ranges for AISI
are assembled with fittings with metal-to-metal seals,
316L, as well as the proposed SEMI compositions for
and by autogenous welding. Once assembled, the sys-
use in semiconductor applications. (SEMI is an interna-
tems and equipment must be leak-tight to keep out con-
tional trade organization, with headquarters in
tamination that may affect the process, and to keep haz-
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619Table 2. Standard Chemistries for 316L, and Proposed SEMI I and II Grades
CMn P S Si Cr Ni Mo N
AISI 316L 0.030 2.00 0.045 0.030 0.75 16.00- 10.00- 2.00- 0.10 max
max max max max max 18.00 14.00 3.00
SEMI I* 0.030 1.00 0.040 0.003 0.75 16.00- 12.00- 2.00- NS
max max max max max 18.00 15.00 3.00
SEMI II* 0.020 0.30 0.040 0.003 0.50 16.00- 12.00- 2.00- NS
max max max max max 18.00 15.00 3.00
Al ≤ 0.020%
Mountain View, CA, serving corporations in the semi-
conductor equipment and materials industry). The
nickel contents of 12 to 15% are given in order to ac- 2.2 Chemistry effects on electropolishing
comodate the Japanese equivalent for AISI 316L, JIS G
4303, which allows this range. Please note that these are
proposed chemistries, as the understanding of this in- Semiconductor gas delivery components and
dustry about the effects of these limits is still evolving. tubing are typically electropolished and passivated.
SEMI I grade is meant for use in bulk gas delivery sys- Electropolishing is performed to improve the appear-
tems, while SEMI II grade is intended for corrosive ance of the part and to smooth the wetted surfaces. Sur-
service. As Table 2 indicates, several elements are re-
face finishes of 10 µ in. Ra max. or better are required.
stricted for semiconductor applications. Elements not Electropolishing and passivation also improve the sur-
listed in Table 2 may also be restricted by end users for face chemistry of the part, enhancing the passive oxide
a variety of manufacturability concerns. These elements
film and removing any free iron from the surface. Elec-
have been found to have effects on machinability, elec-
tropolishing will usually attack any second phases pre-
tropolishing, and most importantly, weldability. sent in the alloy, such as nonmetallic inclusions or
sigma phase. Electropolished components and tubing
are inspected visually and at low magnifications (10×).
2.1 Chemistry effects on machinability
An excessive number of pits or a frosted appearance are
cause for rejection.
The surface finish requirement of the semicon-
Sulfide inclusions are preferential sites for
ductor industry calls for fine finish machining. Many
pitting during electropolishing and for corrosion in ser-
components must have a surface roughness measure-
vice. During electropolishing, sulfide inclusions act as
ment of better than 10 µ in. Ra prior to electropolishing.
anodes and are sacrificially removed by the process. A
low volume percent of sulfide inclusions is desirable.
Machinability is enhanced by sulfur. Sulfur is The limitations on sulfur to 0.003% and below are in-
present to some limited degree in every melt. The sulfur tended to address this issue.
content can also be directly correlated to the volume of
sulfide inclusions in the alloy. However, because of the
Molybdenum increases the ability of the alloy
effects of sulfide inclusions on electropolishing and to resist pitting, especially to chloride solutions. The
corrosion, it is usually limited to under 0.012%. The current range for molybdenum, within the 2 to 3%
trend for sulfur will continue to decrease, as shown by
specified by AISI, is around 2 to 2.5%. The trend ap-
the 0.003% max. level currently under discussion by
pears to be increasing toward the upper end of the
SEMI. range, from 2.5 to 3% molybdenum for some of the
newer formulations for this alloy.
Carbon has effects on hardenability, which
indirectly impacts machinability. Carbon hardens steel,
Residual or unspecified elements, such as alu-
but is kept to 0.03% max. in 316L in order to minimize
minum and titanium, are present in the melt as a result
carbide precipitation on the grain boundaries during of deoxidation practices, refractories, and the type of
welding. The SEMI standards are reducing carbon to a
scrap used. These elements have some effect on
maximum of 0.02%. The reasons for this are not clear,
but it may be an additional effort to avoid carbides.
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619Figure 2 Acceptable weld at approximately 10× × × × Figure 3 Rejectable weld at approximately 10× × × ×
the stability and hardness of the phases present, and can for welding. As discussed earlier, the sulfide inclusions
also be found in nonmetallic inclusions. Oxides, ni- are also preferential sites for corrosion and for pitting
trides, and carbides act as cathodes during electro- during electropolishing. Reduction of sulfur to levels
polishing, and will cause significant pitting. For this near 0.001% causes extreme changes in the weldability.
reason, the minimization of these types of inclusions The dynamics of the weld pool shift so that heat is
(aluminum oxides and titanium carbonitrides, for exam- transferred from the center of the weld outward rather
ple) are desirable. Limitations on aluminum and tita- than inward, so the weld bead becomes very wide and
nium, as well as on the interstitial elements, are an at- shallow.
tempt to control or eliminate the formation of oxides,
nitrides, and carbides. When welding heats with different sulfur con-
tents together, where the spread is more than 0.010%
(for example, 0.001% S to 0.012% sulfur), the weld arc
2.3 Chemistry effects on weldability will usually deflect toward the low sulfur heat. This arc
wandering can cause incomplete penetration of the
weld, and is also a cause for rejection. Usually an at-
The majority of the restrictions on chemistries tempt is made to match heats within a spread of
for 316L for semiconductor applications are due to ±0.005% sulfur.
weldability issues. Many of these restrictions have to do
with the slag that may form on welds, or the redeposited
metallic vapors that may potentially form a corrosion
2.3.2 Chromium equivalents and nickel equivalents
site on the internal surfaces. The redeposited weld va-
por plume is an issue to this industry, because it is not
possible to remove this layer once the system is assem-
Seemingly minor changes in the chemistries
bled. Welds are visually inspected, and excessive heat
can alter the way in which 316L solidifies during weld-
coloration or slag formation are causes for rejection.
ing. Possible solidification modes for 316L are illus-
(4)
Welds are expected to be smooth, straight, flat or
trated in Fig. 4. These modes include austenitic, aus-
slightly beaded, with an ID width approximately one to
tenitic-ferritic, ferritic-austenitic, and ferritic. The ma-
(3)
two times the thickness of the tubing. Figures 2 and 3
jority of 316L chemistries will provide welds that are
show examples of acceptable and rejectable welds.
austenitic, austenitic-ferritic, or ferritic-austenitic. The
austenitic weld will solidify completely to austenite and
no further high-temperature transformations occur. The
2.3.1 Sulfur effects on weldability
austenitic-ferritic weld solidifies as austenite and delta
ferrite is formed from the melt retained between the
austenite dendrites. In the ferritic-austenitic
Sulfur strongly affects weldability. At levels
approaching 0.012%, it reduces the heat input necessary
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619A AF FAFAF
Table 3. Maximum solubility of slag elements in
α/ α/ α/ α/δ δ δ δFe (bcc) and γ γ γ γFe (fcc)
L L L L L
Element Solubility in Solubility in
ferrite, wt. % austenite, wt. %
Ca 0.024 0.016
Si 10.9 1.9
Al 30 0.95
Ti 8.7 ~1
Zr 11.7 ~1
L = liquid = austenite = ferrite
Figures 5 and 6 show SEM and backscatter
electron images of weld slag on a rejectable weld.
Figure 4 Solidification Modes in AISI 316L
When the weld slag is analyzed using EDS, concentra-
A = austenitic; AF = austenitic-ferritic; FA = ferritic-
tions of calcium, aluminum, silicon, and other elements
austenitic; F = ferritic
are found. Table 3 lists the maximum solubility in both
ferrite (bcc) and austenite (fcc) of some elements com-
(8)
monly found in weld slag. In all cases, these elements
weld, ferrite solidifies first and austenite forms between
have a higher solubility in ferrite than in austenite.
the ferrite dendrites. The austenite subsequently grows
These data indicate that ferritic-austenitic solidification
into the ferrite, resulting in a significant decrease in the
is the preferred mode for the reduction or elimination of
volume fraction of ferrite. At room temperature, the
weld slag.
weld is substantially austenite, with a small volume of
retained ferrite.
Welding technique may have some effect on
the solidification mode, since it can affect the weld
The competition between ferrite-promoting
metal composition through dilution and nitrogen
elements and austenite-promoting elements can be de-
pickup. However, the overall effect of solidification
scribed by the chromium and nickel equivalents. There
conditions is of secondary importance, and solidifica-
are several commonly used chromium and nickel
(9)
tion mode is largely determined by chemistry. Under
(5)
equivalents, but the equations developed by Hammar
practical solidification conditions, the ferritic-austenitic
(6)
and Svensson show an excellent correlation between
solidification mode occurs when Cr eq/Ni eq = 1.5 ±
composition and solidification mode, especially for
0.03.
austenitic stainless steels. The Hammar and Svensson
equivalents are as follows:
A drawback to using these equations is that
several of the elements are not commonly reported on
the heat certifications, including niobium, titanium,
• Cr eq =Cr+1.37 Mo +1.5 Si+2Nb +3Ti
copper and nitrogen. Nitrogen in particular will have
• Ni eq = Ni + 0.31 Mn + 22 C + 14.2 N + Cu
strong effects on the ratio, as indicated by its 14.2
multiplicative factor. At very low levels, nitrogen con-
tent is determined by melt processing route. Typical
Using these equations, solidification mode can be pre-
levels for AOD/VAR material are around 0.04% nitro-
(7)
dicted by the ratio of Cr eq/Ni eq. ForCreq/Ni eq <
gen. Other elements used in the calculation of these
1.5, the solidification mode is austenitic or austenitic-
ratios, particularly silicon, manganese, and carbon, are
ferritic. From 1.5 ≤ Cr eq/Ni eq ≤ 2.0, solidification is
being restricted, as shown by the proposed SEMI chem-
ferritic-austenitic, and for values of Cr eq/Ni eq > 2.0,
istries listed in Table 2. Other factors will also affect
solidification is ferritic. Welds acceptable to the semi-
weldability, and must also be considered. These include
conductor industry appear to have a ferritic-austenitic
level of cold work, grain size, and sulfur levels as dis-
solidification mode. Acceptable welds have no slag, and
cussed in Section 2.3.1.
will be very slightly magnetic, indicating some retained
ferrite.
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619Figure 5 SEM image of weld slag at approximately
Figure 6 Backscatter electron image of weld slag
100× × × ×
showninFig.5
Slag elements are rejected by the soldification front
The contrast indicates that the slag phase and the
of the weld
weld metal have significantly different chemical
compositions
2.3.3 Effects of other elements on weldability
welding, tends to vaporize and redeposit downstream on
the surface. The location and thickness of the rede-
posited layer depends on the flow of purge gases. Also,
Some of the other elements that are restricted
the plume or deposit can be greatly minimized by the
for weldability issues include manganese, silicon, alu-
use of a high purity shield/purge gas. The redeposited
minum, and copper. Manganese stabilizes the austenite,
layer is extremely thin and nonadherent. Its color can
and has effects on hardenability. It is also present in the
range from a light straw to a deep blue; the color de-
steel to “trap” the sulfur in manganese sulfide (MnS)
pends on the thickness of the redeposited layer. It can
inclusions. Manganese has been identified as contribut-
be removed by a hot deionized water rinse.
ing to the surface discoloration that occurs during the
welding of fluid system components. Low manganese
The redeposited layer has been reported to be
316L (< 0.3% manganese max.) is discussed at length
the preferential site for corrosion in the presence of
in the next section. Restrictions on silicon and alumi-
corrosive media such as the halogen gases. The corro-
num appear to be due to their presence in weld slag.
sion cell mechanism in which the redeposited layer is a
Silicon increases oxidation resistance and contributes
necessary and sufficient condition for corrosion to oc-
slightly to the hardness. When the oxygen level is low,
cur has not been described, nor has this layer been iso-
it is dissolved in the alloy in solid solution. A judicious
lated as a cause from other metallurgical changes in the
balance of chromium and nickel equivalents should
vicinity of the weld that may also act as preferential
reduce slag issues. Copper also tends to redeposit
corrosion sites. For example, the microstructure of a
downstream of the weld, and may be restricted by some
weld is significantly different from the microstructure of
end users as a potential source of weld contamination,
tubing or bar stock. A heat-affected zone (HAZ) is
or a potential corrosion site.
found on either side of the weld. The microstructure and
properties of the HAZ are altered by the heat input dur-
ing welding, and will also have different corrosion
2.3.4 Low manganese 316L
characteristics than either the weld or the base mate-
(10, 11)
rial.
Manganese has a higher vapor pressure than
the other constituent metals in stainless steel, and during
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619Several articles have discussed the advantages known to have a detrimental effect on corrosion resis-
(12-
of lowering manganese content in the base material. tance.
14)
Initial work required that manganese levels be re-
duced to 0.05% manganese max., but this restriction has
recently been relaxed to 0.3% manganese, possibly in 3. MELTING/REFINING METHODS
response to the economics of low manganese 316L pro-
duction. Other researchers have performed corrosion
testing of welds in stainless steel tubing with varying The following paragraphs describe current
levels of manganese and changes in welding parame- commercial processes by which 316L stainless steel is
(15-17) (18, 19)
ters. It is not clear what portion of the improve- made. These methods include:
ments reported are due to reduced manganese, or to
improved welding techniques. No study to date in the • Argon oxygen decarburization (AOD): primary
open technical literature has shown a connection be-
melting and refining
tween low manganese 316L and improved product yield
• Vacuum oxygen decarburization (VOD): primary
in semiconductor manufacturing. Additionally, in the
melting and refining
unwelded condition, a clear improvement in corrosion
• Vacuum induction melting (VIM): primary melting
resistance for low manganese 316L over standard 316L
• Vacuum arc remelting (VAR): secondary melting
has not been shown.
• Electroslag remelting (ESR): secondary melting
• Electron beam melting (EBM): primary melting
Although reducing the manganese content in
316L at the melt stage is one potential solution to man-
AISI 316L for semiconductor applications is
ganese vaporization during welding, this approach ig-
produced by a primary melting method, or by a primary
nores the beneficial effects of manganese on stainless
melt method followed by a secondary melt method.
steel. Manganese contributes to the machinability and
Common processing paths are AOD, AOD/VAR,
hardenability of stainless steel. It is also an austenite
VOD/VAR or VIM/VAR. Material processed via AOD
stabilizer. Manganese improves resistance to solidifica-
or VOD is usually acceptable for bulk gas delivery,
tion cracking and fissuring in fully austenitic welds.
which is relatively non-corrosive service. System de-
signers will often specify a “double melt” or “double
Removing manganese from the melt also ig-
vacuum melt” for corrosive service applications. These
nores the economics of stainless steel manufacturing. At
represent from 5 to 20% of all applications, but are re-
levels of manganese above 0.3%, standard furnace
sponsible for the majority of research into materials of
charge materials can be used, so there should be no
construction for this industry.
additional cost. Once this level drops below 0.3%, how-
ever, more expensive furnace charge materials must be
used, such as electrolytic iron, which can significantly
3.1 Argon oxygen decarburization (AOD) and vacuum
increase the cost of the material.
oxygen decarburization (VOD)
Any changes in chemistry to remove manga-
nese from 316L must recognize that sulfur levels must
Argon oxygen decarburization (AOD) is a
also be kept very low. If manganese is not available to
process that was originally developed and commercial-
trap the sulfur in MnS inclusions, another type of sul-
ized in the late 1960’s to reduce material and operating
fide inclusionmay form, suchasironsulfidesorcom-
costs and to increase the productivity of stainless steel
plex chromium sulfides. The effects of these types of
production. The majority of stainless steel in the world
inclusions in corrosive ultra-high purity semiconductor
is made using this process. It is energy efficient and
applications have not been fully explored. Also, once
material efficient, and is part of the reason why a pound
sulfur levels get very low (below 0.005%, or 50 ppm),
of stainless steel costs less in constant dollars today
the heat input necessary to achieve a weld increases
than it did thirty years ago.
dramatically. The low manganese/low sulfur weld is
typically two to three times as wide as a conventional
A furnace charge melted in an electric arc fur-
autogenous weld. Because the heat input is so much
nace (EAF) is moved to an AOD vessel. A slag layer,
higher, the HAZ is also larger. Although reducing man-
consisting of molten flux and impurities from the melt,
ganese content may reduce the weld vapor redeposited
forms on top of the furnace charge. This slag layer pro-
near the weld, the tradeoff is a larger HAZ, which is
tects the molten metal from the air and removes impuri-
ties. Argon, oxygen, and nitrogen are injected into the
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619melt from the bottom at sonic velocities. The gas blow The VIM furnace charge is usually smaller
from the bottom allows the removal of carbon to low than an AOD charge. The VIM crucible has a powder
levels without excessive chromium oxidation. One of refractory lining that is rammed into place between the
the advantages of this procedure is that low levels of crucible wall and a steel lining. During the first heat, the
carbon can be obtained from high carbon furnace steel lining becomes part of the melt and the rammed
charges, with comparably little loss of chromium by refractory lining is sintered and densified by the heat of
oxidation. The submerged blowing also provides excel- the melt. The rammed crucible lining undergoes a
lent slag/metal and gas/metal contact, total utilization of higher degree of erosion than do brick-lined ladles.
the injected oxygen with the bath, and the removal of Therefore, this lining is replaced frequently in compari-
dissolved gases (hydrogen and nitrogen) and nonmetal- son to refractories in the EAF/AOD process.
lic inclusions. Metallurgical refining processes such as
decarburization, deoxidation via the gas phase, dephos-
phorization, and desulfurization can be carefully con- 3.3 Vacuum arc remelting (VAR)
trolled, and result in excellent control of the chemistry
of the melt. The electric arc furnace, the ladles, and the
AOD vessel are lined with refractory bricks. The fur- Vacuum arc remelting (VAR) is a secondary
naces, ladles and vessels are relined on a scheduled melt process in which a consumable electrode (an ingot
maintenance basis. cast from an AOD melt or a VIM melt) is melted con-
tinuously by means of a DC arc under a high vacuum of
-4 -2
Vacuum oxygen decarburization (VOD) is 10 to 10 mbar. The molten material solidifies in a
similar to argon oxygen decarburization, except that the water-cooled copper mold. Vacuum arc remelting pro-
process occurs under a rough vacuum of 10 to 100 duces a homogeneous, virtually segregation-free ingot.
mbar, and bottom blowing is restricted to small amounts The VAR process removes dissolved gases such as ni-
of inert gas (argon), mostly for stirring purposes. The trogen and hydrogen, reduces the amount of undesirable
resulting material is similar in cleanliness to AOD mate- tramp elements with high vapor pressures (arsenic, lead,
rial. tellurium, selenium, bismuth, silver, copper), and re-
duces oxygen/oxide levels. The process takes from 12
to 24 hours. During VAR, a portion of the ingot is so-
3.2 Vacuum induction melting (VIM) lidifying, a portion is molten, and a portion is getting
ready to melt. The molten pool is approximately as
deep as it is wide.
Vacuum induction melting (VIM) is a primary
melt process in which a furnace charge of clean scrap or Solidification occurs at a very consistent rate,
known charge material is melted under a fine vacuum of resulting in a controlled microstructure. Since the so-
-3
10 to 1 mbar in an inductively heated crucible. VIM lidification microstructure carries through to the final
came into common usage in the production of superal- bar product, this will yield bar stock that is homogene-
loys, which are highly reactive with oxygen when mol- ous with very little segregation of elements. The solidi-
ten. Under a vacuum, oxidation reactions and thus oxide fication front rejects the impurities and inclusions, so
inclusions are almost completely avoided. they tend to float out. Once solidification is complete,
the ingot is cropped to remove the region of last solidi-
Since VIM is strictly a melting process, and fication.
refining is limited, a much purer furnace charge must be
used. The metallurgical refining processes such as de-
carburization, deoxidation via the gas phase, dephos- 3.4 AOD/VAR vs. VIM/VAR
phorization, and desulfurization do not occur to the
same extent as in EAF/AOD. Metallurgical refining is
limited to the purely pressure-dependent reactions as in A general perception exists among members of
the case of carbon, oxygen, nitrogen, and hydrogen re- the semiconductor industry that VIM/VAR material is
moval, and the evaporation of elements with high vapor cleaner than AOD/VAR, and is therefore the next logi-
pressures such as arsenic, copper, lead, bismuth, tellu- cal step in the drive for ultra-pure materials. When the
rium, antimony, and tin. The furnace charge require- industry moved toward VAR materials in the early
ments and the maintenance of the vacuum make this a 1990’s, there was strong evidence that material cleanli-
more expensive process than EAF/AOD. ness was significantly improved by vacuum arc remelt-
ing, and that materials-related issues such as stringers
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619and pin-hole leaks could be minimized. The same type of the melt may take place. This melt method will re-
of evidence does not exist to show the cleanliness bene- duce sulfur and nonmetallic inclusions. The remaining
fit of VIM over AOD as a primary melt method; for inclusions are very small and are evenly distributed in
heats of identical chemistries, it has not been proven the remelted ingot. ESR is not commonly applied to
that VIM/VAR will provide a cleaner 316L than 316L materials used for semiconductor applications. It
AOD/VAR, especially at the low levels of nonmetallic will probably never be specified widely, as the percep-
inclusions currently seen. tion in the industry is that a vacuum melt or remelt is
necessary to ensure clean steel.
Using currently available analytical tech-
niques, it is not possible to determine absolutely, with- Electron beam melting (EBM) is a vacuum
out prior knowledge, the primary melt method of a melting process routinely used for the production of
VAR heat. The combination of low sulfur (0.005% and reactive and refractory metals, such as tungsten, molyb-
below) and low oxygen (0.010% and below) may indi- denum, tantalum, niobium, hafnium, vanadium, zirco-
cate a double vacuum melted material. nium, and titanium. In electron beam melting, the feed-
stock is melted by impinging high energy electrons. For
A VIM melt may see only one refractory lin- 316L, continuous flow melting has been applied. This is
ing, as opposed to as many as four for an AOD melt. a two-stage process in which the first step (material
The refractories have been identified as a potential feeding, melting and refining) occurs in a water-cooled
source of inclusions causing pin-hole leaks in some of copper trough, ladle or hearth. In the second step, the
the thin-walled products. The tradeoff is that refracto- melt soldifies in a water-cooled copper crucible. Impu-
ries and slag are important in metallurgical refining. rities are segregated by flotation or sedimentation. The
VIM is a more expensive process than AOD for the purity and properties of EBM steels are in some re-
following reasons: spects better than those of VAR and ESR steels, but the
processing costs are much higher, and the heat quanti-
ties are smaller. The essential advantage of the electron
• The heat size is usually smaller.
beam melting of steel is the drastic reduction of impuri-
• The cycle time for melting a heat is longer (6 hours
ties and interstitial elements.
for VIM versus 2 hours for AOD).
-3 0
• VIM requires pulling a fine vacuum of 10 to 10
A recent study examined the corrosion resis-
mbar. AOD occurs at ambient pressures by shield-
tance of welds in tubing manufactured by different melt
ing the melt with argon.
(17)
processing methods. Welds in tubing manufactured
• Furnace charge materials are usually more expen-
from EBM 316L showed fewer corrosion initiation sites
sive for VIM than for AOD, due to higher purity
than welds in tubing manufactured from 316L VOD,
requirements, since less refining occurs in the VIM
VOD/VAR, and VOD/VAR with low manganese. Stan-
process.
dard ASTM corrosion tests were used. All heats did
• VIM requires higher energy input than AOD.
show corrosion initiation, however, and the greatest
• VIM requires higher maintenance than AOD.
improvement in the reduction of corrosion pits was real-
ized by going from VOD to VOD/VAR. The low man-
ganese VOD/VAR was not significantly better than the
3.5 Other melt methods
VOD/VAR with 1.68% manganese. EBM 316L, due to
its small heat sizes and relatively high cost, will proba-
bly remain a specialty offering.
The following paragraphs describe electroslag
remelting and electron beam melting, other methods by
which 316L may be melted and refined.
Electroslag remelting (ESR) is a secondary
4. CORROSION IN SEMICONDUCTOR SYSTEMS
melting process in which the ingot is built up in a water
cooled mold by melting a consumable electrode im-
mersed in a superheated slag. During melting, the metal
In ultra-high-purity semiconductor gas delivery
is refined and cleaned of impurities because of the ex-
systems, corrosion depends on many factors, such as:
cellent contact of the molten metal and the slag. As with
vacuum arc remelting, the resulting ingot is fully dense
• gas concentration and purity
and homogeneous. The ESR process is usually carried
• moisture content
out under a normal air atmosphere, so some oxidation
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619passive film. Most frequently, the function of passiva-
• temperature
tion is to remove free iron, oxides, and other surface
• localized inhomogeneities in material
contamination. Once the surface is cleaned and the bulk
• system flow rates
composition of the stainless steel is exposed to air, the
• time of exposure
passive film forms immediately. In components and
• frequency of exposure
tubing for semiconductor service, passivation is used
after electropolishing to enrich the passive oxide film
Most of these factors are controlled by system operating
by removing free iron ions and iron oxide. (This passive
parameters and protocols. However, a great deal of em-
oxide film, by the way, is on the order of 20 to 100 nm
phasis has been placed on material cleanliness, surface
thick, or less than 1/1000th the thickness of a human
treatments, and new alloys, to minimize the potential
hair. The film can range from a couple of hundred to
for corrosion.
about one thousand atoms in thickness.)
The mechanism of corrosion, especially as
applied to corrosive ultra-high purity service, has been
(20-23)
4.2 Chromium oxide passivation technology
examined by several researchers. The level of
moisture at which corrosion starts has not been deter-
mined absolutely, but Wang et al. indicate that 1 ppm
Chromium oxide (Cr O ) passivation is a tech-
2 3
moisture in 100% HCl is a realistic moisture level in
(20)
nology intended for use in welded assemblies exposed
service. In a study by Smudde et al., HBr containing
to halogen gases. During the welding process, the elec-
0.5 ppm (500 ppb) water by volume does not attack
(21)
tropolished surface of the components is destroyed at
electropolished 316L tubing. For HBr, 100 ppb
(22)
the weld and disturbed in the heat affected zones. If a
moisture is considered a realistic service level.
complete chromium oxide passive layer can be formed
on the surface of the weld and adjoining regions, it is
A recent investigation evaluated the effects of
reasoned that the corrosion resistance of this area can
purity, materials and surface finish on gas system per-
be improved.
formance. The study used typical operating parameters
found in the most corrosive service in semiconductor
(23)
Chromium oxide passivation is based on the
grade gas delivery. Twenty-seven gas sticks were
constructed as models of typical gas distribution sys- steady-state diffusion of chromium from the bulk mate-
tems in modern semiconductor facilities and tools. Nine rial to the surface of the part to form a chromium oxide
layer. A dry passivation treatment using a gas mixture
sticks were manufactured from standard components
of 10% hydrogen, 1 ppm oxygen and the balance argon
(316L, 10 µ in. Ra max.), nine from ultrahigh purity
at 500°C for one hour is reported to result in the forma-
components (316L, 7 µ in. max.), and nine from a vari-
tion of a passive Cr O layer on an electrochemically
ety of advanced alloy components. The sticks were
2 3
(24)
buffed, electropolished 316L surface. This type of
tested at three sites in the United States in HBr service
treatment is an attempt to develop a method to passivate
with varying levels of moisture in the system. This ac-
the internal wetted surfaces of a gas delivery system
celerated corrosion life approach applies a dosage re-
after the system has been welded and leak tested. The
sponse model (dosage times time, or ppb H O⋅day), in
2
results of a five-day corrosion test of treated tubing ex-
which the equivalent life of a test can be accelerated by
posed to HCl gas (containing 1.4 ppm moisture) at 5
increasing the moisture content. Using this model, 10
2
kg/cm and 100°C showed no sign of corrosion on the
days exposure at 100 ppb moisture is the equivalent of
100 days exposure at 10 ppb moisture. The researchers’ passivated surface. However, at 200°C, the corrosion
proceeded so rapidly that the original surface was not
results show that standard 10 µ in. Ra max. gas compo-
recognizable. Although chromium oxide passivation has
nents performed well over the equivalent of 12 years of
been achieved under laboratory conditions, its stability
cycling, and six years of corrosion in HBr application at
as a corrosion-resistant surface treatment appears to be
100 ppb of moisture. In fact, they performed as well as
very temperature-dependent. Its functional range will
the ultrahigh purity and advanced alloys.
need to be determined before it can be applied to fin-
ished components and assemblies.
4.1 Passivation
4.3 Ferritic stainless steel
A passive oxide film forms spontaneously on
stainless steels in the presence of oxygen. It is not nec-
essary to chemically treat a stainless steel to obtain the
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619Ongoing research into optimizing chromium which 316L does not provide the desired performance.
oxide passivation technology suggests that ferritic Ferritic stainless steels have not been specified to date
rather than austenitic stainless steel should be used as in semiconductor systems. Issues associated with manu-
the base material in gas delivery systems. Researchers facturability or use by this industry have not been fully
claim to have achieved “a 100% Cr O passivated film explored.
2 3
(without Fe O ) on an electropolished surface of ferritic
2 3
stainless steel with a chromium concentration greater
(25)
than 26%.”
CONCLUSIONS
Ferritic stainless steels are commonly known
as the “straight-chromium” types, and are characterized
by being ferromagnetic. This family of alloys can offer Specifications for semiconductor gas delivery
moderate to outstanding corrosion resistance. The high- systems are based largely on engineering judgment.
chromium ferritic alloys are sensitive to notch brittle- Materials requirements are driven by manufacturability
ness on slow cooling to ordinary temperatures. Ferritic (machinability, electropolishing, and weldability) and
stainless steels can be severely embrittled by interstitial applications (weldability and corrosion) issues. Many
contamination during welding by nitrogen, carbon, hy- restrictions placed on AISI 316L used in semiconductor
drogen or oxygen. These elements have detrimental applications are attempts by end users to guarantee that
effects on ductility and toughness, and are present due the 316L they receive will function in their systems.
to improper or contaminated gas shielding. Inert back- Some of these restrictions, such as the limits on sulfur
ing gases and high-purity shielding gases are necessary or the specification of VAR as a secondary melting
to produce acceptable welds. method, are based on data or operating experience, and
have shown real improvements in reducing leaks and
A ferritic stainless steel with a chromium oxide extending system lives. However, other areas of re-
passivation coating for use in piping systems has been search, such as low manganese 316L, electron beam
(26-28)
patented. The closest ASTM equivalent ferritic melting, or chromium oxide passivation, have evolved
stainless steel is XM-27 (UNS No. S44627), also called into a quest for “ultra-pure, corrosion-free” 316L.
E-Brite or E-Brite 26-1. The nominal chemical compo-
sition for XM-27 is 26% chromium, 1% molybdenum, The expense, both up-front and hidden, of
0.015% nitrogen max., and 0.01% carbon max. This these “ultra-pure” materials should be evaluated care-
alloy is considered a third-generation super-ferritic, fully. These materials are usually only available from a
which exhibits excellent resistance to chloride induced limited number of sources. They should be specified
stress corrosion cracking. The major commercial appli- only if evidence verifies significant benefits in terms of
cation for this grade is in heat exchangers and piping increased system life (or improved product yield in
systems for chloride-bearing aqueous solutions and semiconductor manufacturing). Otherwise, materials
seawater. specifications that limit chemistries, melt methods, and
surface treatments only increase cost without adding
There are significant technological barriers to value.
the adoption of ferritic stainless steels for use in semi-
conductor gas delivery systems. First, dry chromium
passivation will need to be perfected, and will need to
occur at temperatures near ambient. Second, weldability
issues will need to be addressed. Although subassem-
blies are often welded under clean room conditions,
final setup involves field welding, and welding parame-
ters cannot be as closely controlled. Finally, manufac-
turability issues, such as machining and electropolish-
ing, will also need to be addressed. Ferritic stainless
steel may someday take its place alongside Inconel∗,
Monel∗, Hastelloy+, and nickel, alloys currently speci-
fied for the small number of corrosive applications in
∗ INCO Alloys International, Inc.
+ Haynes International, Inc.
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-61912. S. Miyoshi, Y. Shirai, T. Kojima, T. Ohmi,
REFERENCES “Ultraclean Tube Welding Technology,” Proc. 20th
Symposium on ULSI Ultra Clean Technology, 1994,
pp. 244-249.
1. Market information from: SEMI International,
805 East Middlefield Rd., Mountain View, CA, 94043. 13. H. Nakashima, M. Ohsawa, T. Ohtsuka, M.
Nakatani,Y.Baba, K. Sato,T.Ohmi, “Fume-Free
2. “Introduction to Semiconductor Fabrication,” Welding Technology for Manufacturing Equipment,”
Applied Materials, January 1995. Proc. 20th Symposium on ULSI Ultra Clean Technol-
ogy, 1994, pp. 250-261.
3. “Guideline for Welding Stainless Steel Tubing
for Semiconductor Applications,” SEMI Standard F3- 14. S. Miyoshi, T. Kojima,T.Suenaga,T.Ohmi,
90, SEMI International, Mountain View, CA, 1990, 5 “Metal Fume-Free Welding Technology,” Microcon-
pp. tamination Conference Proceedings, 1993, pp. 606-615.
4. N. Suutala, “Solidification of Austenitic 15. S. Krishnan, S. Grube, O. Laparra, A. Tud-
Stainless Steels,” Third Scandinavian Symposium in hope, “Site-Specific Corrosion in Gas Delivery Tubing
Materials Science, 20-21 June 1983, Oulu, Finland; Exposed to Semiconductor Grade HCl,” Cleanrooms
Acta Univ. Oulu, C 26,Metallurg.4, 1983, pp. 53-60. Supplement, October 1995, pp. S11-S15.
5. N. Suutala, “Effect of Manganese and Nitro- 16. S. Krishnan, “Effect of Design and Choice of
gen on the Solidification Mode in Austenitic Stainless Materials on the Integrity of a Gas Delivery System,”
Steel Welds,” Met. Trans. A, Vol. 13, Dec. 1982, pp. Workshop on Gas Distribution Systems,SEMICON
2121-2130. West 96, July 1996, pp. 81-92.
6. O. Hammar and U. Svennson, in Solidification 17. C. Burton, “Effect of Stainless Steel Melt
and Casting of Metals, The Metals Society, London, Methods on the Corrosion Resistance of Welds in UHP
1979, pp. 401-410. Gas Lines,” Workshop on Gas Distribution Systems,
SEMICON West 97, July 1997, pp. F-1-F-17.
7. N. Suutala and T. Moisio, “Use of Chromium
and Nickel Equivalents in Considering Solidification 18. ASM Metals Handbook, 9th Ed., Vol. 15:
Phenomena in Austenitic Stainless Steels,” in Solidifi- Casting, D. M. Stafanescu, Ed., ASM International,
cation Technology in the Foundry and Casthouse,The Materials Park, OH, 1988.
Metals Society, London, 1983, pp. 310-314.
19. Vacuum Metallurgy, A. Choudhury, ASM
8. Binary Alloy Phase Diagrams,T.B.Massal- International, Materials Park, OH, 1990.
ski, Ed., American Society for Metals, Metals Park,
OH, 1986. 20. H.-C. Wang, G. Doddi, S. Chesters, “Estimat-
ing the Lifetime of Electropolished Stainless Steel
9. N. Suutala, “Effect of Solidification Condi- (EPSS) Tubing in Corrosive Gas Service,” Journal of
tions on the Solidification Mode in Austenitic Stainless the IES, July/August 1994, pp. 28-31.
Steel Welds,” Met. Trans. A, Vol. 14, Feb. 1983, pp.
191-197. 21. G. H. Smudde,Jr.,W.I.Bailey, B. S. Felker,
M. A. George, J. G. Lanagan, “Materials Selection for
10. N. Suutala and M. Kurkela, “Localized Corro- HBr Service,” Corrosion Science, Vol. 37, No. 12,
sion Resistance of High Alloy Austenitic Stainless 1995, pp. 1931-1946.
Steels and Welds,” Proc. Conf. Stainless Steels ‘84,
Goteberg, 3-4 Sept. 1984; Institute of Metals, London, 22. S. M. Fine,J.T.McGuire,B.-S. Choi,T.Bzik,
1985, pp. 240-247. K. T. Crofton, A. Melnyk, M. Perez, D. P. Sheriff, “Op-
timizing the UHP Gas Distribution System for a Plasma
11. G. J. Davies and J. G. Garland, “Solidification Etch Tool,” Solid State Technology, March 1996, pp.
Structures and Properties of Fusion Welds,” Intl. Met. 71-81.
Rev., 1975, Vol. 20, pp. 83-106. 23. A. Glew,M.Crockett, A. Johnson,W.Dax, R.
Binder, J. Riddle, J. Ammenheuser, “Accelerated Life
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619Testing of Gas System Performance and Reliabil-
ity,”SEMATECH Technology Transfer 97043272-TR, 27. T. Ohmi, “ Method of Forming Oxide Passiva-
SEMATECH, Inc., May 31, 1997, 62 pp. (available full tion Film Having Chromium Oxide Layer on the Sur-
text on the Worldwide Web at www. se- face Thereof, and Stainless Steel Having Excellent Cor-
matech.org/public/docubase/abstract/3272atr.htm). rosion Resistance,” International Patent No. 94 08065,
April 14, 1994, 30 pp.
24. T. Ohmi,A.Ohki,M.Nakamura, K. Kawada,
T. Watanabe, Y. Nakagawa, S. Miyoshi, S. Takahashi, 28. T. Ohmi and S. Miyoshi, “Stainless Steel Having
M. S. K. Chen, “The Technology of Chromium Oxide Chromium Oxide Passivation Coating for Piping Sys-
Passivation on Stainless Steel Surface,” Journal of the tem,” International Patent No. 95 18246, July 6, 1995,
Electrochemical Society, Vol. 140, No. 6, June 1993, 17 pp.
pp. 1691-1699.
25. T. Ohmi, “Corrosion-Free Cr O Passivated
2 3
Gas Tubing System for Specialty Gases,” Cleanrooms ACKNOWLEDGMENTS
Supplement, October 1995, pp. S18-S22.
The author wishes to thank Swagelok/Nupro Company
26. T. Ohmi, “ Method of Forming Passive Oxide for permission to publish this paper. Helpful discus-
Film Based on Chromium Oxide on Stainless Steel,” sions with colleagues at Swagelok are also gratefully
International Patent No. 93 10274, May 27, 1993, 28 acknowledged.
pp.
Presented at the 39th Mechanical Working and Steel Processing Conference of the Iron and Steel Society, Indianapolis, IN, 19-22
October 1997; published in the Conference Proceedings 1998; pp. 607-619