|
|
| THE
EFFECTS OF CORROSIVE GASES ON METAL SURFACES |
|
Patricia
M. Clarke, Lab Manager, AIRCO ELECTRONIC GASES, Richard A. Hogle,
Regional Technical Manager, AIRCO ELECTRONIC GASES, Steve M. Lord,
SML ASSOCIATES
INTRODUCTION
Halogenated
gases like Tungsten Hexafluoride and Hydrogen Bromide are important
in Integrated Circuit Manufacturing. Due to the corrosive nature
of halogenated gases, special attention should be paid to their
handling. High purity WF6 and HBr are supplied typically in stainless
steel, pure nickel or nickel lined cylinders. Passive films formed
on the two surfaces are different. During use, these gases also
pass through plumbing constructed of nickel, Hastelloy, or stainless
steel, with a variety of surface finishes. It is important to
understand the differences between these materials and the chemical
interactions that occur on these surfaces, when exposed to halogenated
gases. A Fourier Transform Infrared (FTIR) Spectrometer was used
to evaluate chemical interactions with 316L electro-polished stainless
steel, mill finished nickel-200, chemically cleaned nickel-200,
and mill finished Hastelloy C-22 tubing surfaces. Purge and bake-out
performances of the tubing, and the resulting impurities generated
when the tubing was in contact with WF6 and HBr, were evaluated.
Specific impurities extracted from the tubing were identified.
This information will help in the design of purification systems
and process reactors.
Passivation:
The
basic chemistry of fluoride passivation is the fluorination of
a metal oxide surface with the rapid formation of an impervious
fluoride layer, reported by Cannon, et.al.4, to be 10-30A thick
at room temperature. Closer examination of this layer reveals
the presence of metal oxide under the fluoride layer4,5,6 and
the consequent need for very low oxygen content in the fluorinating
agent.1,5,6 Ohmi et.al.2 conducted direct fluorination of the
base metal by first removing the oxide with a 0.5% HF solution
and fluorinating at temperatures >20° C. The passive film,
thus produced, was thicker (>1OOO Angstroms); but, this technique
is impractical to use in fully-assembled reactor systems.
The
typical reaction using WF6 as the passivating gas is given:
MxOy
+ WF6 ® MFz + WOF4 + (WxOyFz + WxOy)
The
presence of volatile WOF4 is likely after passivation. Heated
evacuation will be needed to remove the compound from the system
once it has been formed.
HBr
has similar reaction characteristics with metal oxide surfaces.
Bromine and oxygen will transfer to the metal surface but to a
lesser degree due to the lower reactivity of the bromine. The
following reaction may occur:
MxOy
+ HBr® MxBry + H2O + H2
From
this, it can be seen that the HBr reaction can produce moisture
as a by-product when contacted with metal oxide surfaces. This
moisture in HBr then facilitates the corrosion of the surface,
creating further reaction products and accelerating the corrosion
of the tubing. Many by-products of these reactions may be volatile
and may enter the gas phase, therefore, entering the semiconductor
process tool.
Surface
Species:
Nickel-200,
316L stainless steel, and Hastelloy C-22 tubing have the following
compositions:
Table
1.
These
metals will form a wide variety of fluorides, bromides, hydrates,
hydroxides, etc. In this process, several compounds will be produced
that will end up in the gas phase.
Volatile
Species:
Trace
components of these metals can also form volatile fluorides and
oxyfluorides: silicon forms SiF4 (BP= -94.8° C); molybdenum
forms MOF5, (BP= 213.6° C), MoOF4 (BP= 186° C), and MOF6
(BP= 34° C); chromium forms a volatile oxyfluoride, CrO2F2
(BP= 34° C).
In
the case of HBr, several bromide compounds can be produced, similar
to the fluorides, but with much lower volatilities, due to the
larger bromine atom. Carbon can produce volatile impurities, such
as the CxHyBrz compounds (BP= 3.5° C to 190° C). Several
metal bromide hydrates are also of relatively high volatility
(for ex., FeBr3· H2O, Vapor Pressure @ 40° C is 11.9
Torr).
Hydrate
Formation:
Since
water is extremely important in the chemistry of halogenated compounds,
the moisture level is extremely critical to the chemistry. Free
water is adsorbed on the surface of most metal and metal oxides.
Much of this can be removed by baking of the surface in a dry
gas. Chemically bound water such as hydrates and hydroxides, however,
will not be removed by simple baking. For inert gases, these hydrates
which are chemically bound to the metal surface will not play
a role; but, in the case of halogenated compounds, these hydrates
will react with the gases, thus affecting the purity of the gas
entering the reactor.
Water
is held by metal surfaces in several forms. Adsorbed water is
held to the surface by loose bonds that can be broken by thorough
purging, evacuation, and mild heat. Metal oxides and fluorides
form hydrates and hydroxides which have thermodynamically strong
bonds. Table 2 shows the potential hydrates in these systems with
their hydrate heat of formation. The potential for hydrate formation
is the highest with the most negative Hf(hydration) - Hf(water)
value in the last column. AIF3 and Cr2O3 have high potential for
hydrates (-24.5 and -20 kcal/mole respectively; whereas, nickel
oxide and fluoride have a very low potential (-0.95 and -1.87
respectively).
The
relatively high Hf of the Cr2O3 hydrates indicates that chrome
oxides have a strong potential for attracting and holding water.
In contact with inert gases, the strongly bound water will not
enter the gas phase unless the surface is heated substantially
and the bonds are broken. This property helps the chromium protect
the more fragile Fe from being attacked by moisture to form rust.
In
the presence of a halogenated gas such as WF6 the hydrate is no
longer inert. The WF6 reaction with the hydrate to form HF and
tungsten oxides is thermodynamically favorable. Therefore, free
or adsorbed water can be reduced to very low levels; but, surface
reactions will still occur, producing HF and potentially harmful
WOF4.
Further
evidence of the reactions, with hydrates is found in some work
by Sematech in improving the speed of the clean cycle in tungsten
tools.9 After a long bake out cycle, an NF3 plasma was used to
rapidly remove water. The plasma resulted in an increased leak-back
rate that was greater than expected from the water measured by
RGA. A fluorine plasma will also react with the metal hydrates
(AIF3 has a particularly high potential for hydrate formation)
generating HF, NOx, H20 and other light gases. These species were
found by the RGA after the plasma cycle.9
It
can be noted, that chromium oxides, fluorides, bromides, and nickel
bromides have a relatively high potential for retention of water
as hydrates, whereas nickel oxides and fluorides have a low potential
for retention of water. This means that water is attracted by
the chromium surface in both systems and by the nickel surface
in the bromide system, thus holding water firmly. The nickel oxide
surface, in the fluoride system, has less potential for forming
hydrates, thus releases water more easily.
EXPERIMENTAL
To
examine the chemistry of the WF6 and HBr interaction with tubing
walls, a Fourier Transform Infrared (FTIR) cell was devised to
analyze the composition of the gas phase impurities (during pre
and post purge cycles, and during contact with the halogenated
gases). The cell consists of four polished nickel mirrors to reflect
the IR-Beam into the tubing, silver chloride windows to contain
the gas and 2 lengths of standard 3/8" tubing. The total
path length was 2 meters. This cell was placed in a BioRad FTS-40
FTIR with a water cooled, high energy source, and a high sensitivity
MCT detector.
The
tubing materials, examined in this series of tests, were 31 6L
stainless steel tubing that was electro-polished to a nominal
10 microinch Ra finish, mill finished nick eI-200 tubing, chemically
cleaned nickel-200 tubing which was the best available and had
an approximate surface of 25 microinch Ra, and mill finished Hastelloy
C-22 tubing. The tubing was used as received and no special cleaning
techniques were used for this series. This series of experiments
focused on the study of the gross characteristics of these tubes
which would be considered industry standard tubing.
Table
2.
The
prepurge cycle consisted of successive heating through several
stages up to 159° C. The cell was evacuated to <1 Torr,
isolated, and monitored for generation of gas phase species from
the tubing wall, using FTIR. This cell was then pulse purged with
30 psig helium (pulses ranged from 30 psig helium to 1 Torr vacuum,
10 times). This bake-out and pulse-purging step was repeated three
times.
After
the heated pulse purge and after the cell was cooled down to a
controlled 30° C, in each case, WF6 (or HBr) was introduced.
The cell was filled with approximately 50 Torr of WF6 (or HBr),
an FTIR spectrum was taken immediately upon filling, which was
used as the baseline (background). This baseline (background)
was then subtracted from subsequent spectra, taken during a four
hour contact period with the halogenated gas. The system was then
evacuated and pulse-purged 10 times to remove the WF6 (or HBr).
Lastly, the cell was isolated at <1 Torr vacuum and successively
heated to 159° C. FTIR spectra were taken at several temperatures
up to 159° C, revealing any gas phase impurities released
from the exposed tubing walls.
RESULTS
Figure
1 shows the moisture generated in the final heated pulse purge
cycle, before introduction of a halogenated gas. Note that 316L
electro-polished stainless steel has a higher capacity for holding
water; therefore, a longer bake-out time is necessary to remove
this moisture. The chemically cleaned nickel showed a higher retention
of moisture. This may be due to the chemical modifications made
during the chemical cleaning process. Chemical cleaning involves
aqueous solutions which may further hydrate the surface during
the chemical cleaning process. Mill finished Hastelloy C-22 and
nickel-200 gave off the least amount of water during the bake-out
cycle. The mill finished materials would be expected to dispel
water at a slower rate, due to the much higher surface area.
Figures
2-6 show the gas phase impurities generated by chemical reactions
between WF, and the metal surfaces, producing SiF4, CF4, C02,
SF6 and PF5. The carbon compounds are present due to surface contamination
from organic and carboxyl groups on the surface of the tubing
wall. 316L electro-polished stainless steel and chemically cleaned
nickel-200 showed the highest levels of these various impurities
generated from the surface of the tubing wall.
Figure
7 shows a comparison of FTIR spectra for PF5 in WF8 versus all
four materials. Nickel-200 and Hastelloy C-22 show very little
PF5 present; however, sizable peaks were seen in the chemically
cleaned nickel and the 316L stainless steel.
Figure
8 shows a comparison of FTIR spectra for WOF4 (after the tubing
was exposed to WF6 pulse purged, then baked to 159° C at <
1 Torr). A significant WOF4 peak was noted in the chemically cleaned
nickel. This again appears to be due to oxygen containing residues
on the surface of this tubing, due to the cleaning process. The
other materials did not show this characteristic. WOF4 has been
implicated in affecting chemical reactions in the tungsten CVD
process.10
When
the 316L electro-polished stainless steel and nickel-200 tubing
were exposed to HBr, several different reactions occurred. Figure
9 represents the volatile species generated (CxHyBrz). These compounds
appear to be reactions with organic contaminants on the surface
of the tubing. Maybe further prep and removal of organic materials
on the surface of the tubing may help alleviate this concern;
however, it is noted that these compounds were generated and will
end up in the gas phase of the reactor, potentially affecting
the etching process and affecting the surface chemistry of the
plasma etch.
The
interaction of HBr with moisture is shown in Figure 10. In the
period of the four hour contact of HBr with 316L electro-polished
stainless steel and nickel-200 tubing, the hydrogen bromide peak
decreased (shown as negative peaks in the plot) and the moisture
increased. This may be due to the chemical reaction of HBr with
native metal oxide on the surface.
HBr
+ MOx = MBr2 + H2O
Hydrates
may also be participating in this reaction where the hydrate bond
is being fractured by the HBr. As this continues, a corrosive
mixture can be produced. Moisture also has a strong effect on
the selective etch process. One can therefore infer that thorough
passivation of the metal surfaces would be required before operation.
DISCUSSION
There
are a variety of chemical reactions that are occurring on the
metal surfaces, creating gas phase impurities which can be carried
into the reactor system. Trace components in metal alloys, such
as silicon, carbon, sulfur, and phosphorus are extracted in the
fluoride system, and carbon in the bromide system. These materials
can affect the operation of the system, particularly when new
tubing and/or new gas distribution systems are installed. Over
time, it is probable that all of the extractable material will
be removed from the metal and the gas phase impurities will diminish;
however, the effect on the metallic microstructure, during subsequent
operations, is still unknown.
The
316L electro-polished stainless steel material has the largest
number of trace impurities, therefore exhibited the largest number
of gas phase impurities generated in the fluoride system. There
is also a large amount of water that can be expelled by the electro-polished
surface. This may be due to the high chromium oxide content, which
has a high potential for hydrate formation. Surface chemistry
studies such as ESCA will be necessary to determine this exactly.
The mill finished nickel-200 system exhibited the least amount
of gas phase impurities because there are less trace elements
to extract. Hastelloy C-22 also showed very few gas phase impurities,
although, it does release more water vapor during bake-out cycle.
Chemically cleaned nickel showed a variety of gas phase impurities.
Since the chemical composition of this tubing is the same as the
mill finished nickel tubing, it is likely that the chemical cleaning
process adds the impurities which are eventually turned into gas
phase contaminants. Acids such as phosphoric, sulfuric, acetic,
and nitric are used in the "chemical cleaning" process
and may leave residues that will eventually be converted into
SF6, PF5, and 0F4 in a fluoride system (CxHyBrz in a bromide system).
Chemical polishing of tubing surfaces has improved; however, it
apparently leaves behind chemical residues that can interact with
fluoride and bromide gases.
In
this study, the bromine would dynamically interact with many of
the same elements as fluorine; but, the by-products would not
be volatile under these conditions. The CxHyBrz compounds are
the only impurities apparent in the spectra. Thorough purging
with reactive gas, while by-passing the wafer chamber, would be
advised before starting the etch process.
CONCLUSIONS
Based
on these initial studies, there are several concerns with the
materials of construction of CVD reactor and gas handling equipment.
·
Stainless steel with its many components and trace impurities
creates more gas phase impurities. SiF4, CF4,PF5., SF6 and other
compounds that have not yet been identified, were seen in the
FTIR spectra upon introduction of WF6 (and CxHyBrz in HBr). Care
must be taken to fully purge these impurities at elevated temperatures
before they are introduced into the CVD system. Chemical interactions
between HBr and stainless steel need further investigation, due
to the lower volatile impurities which may not be revealed by
FTIR.
·
Since nickel-200 and Hastelloy C-22 are simpler chemical systems
and appear to have few chemical interactions, efforts should be
made to develop high integrity tubing or electro-polished tubing
in these materials, for use in corrosive applications. The higher
cost may be worth the benefits.
·
Chemical interactions between HBr and nickel (and HBr and Hastelloy)
are still being investigated, again, due to the lower volatile
impurities which may enter the gas phase at higher temperatures.
Figure
1.
Figure
2.
Figure
3.
BIBLOGRAPHY
Elliott,
Stephen, BOC internal memo 2Oth Mar 1991 re Australian HF Explosion
incident.
Ohmi,T et al, Fluorine Passivation of Stainless Steel, Corrosion
Science, Vol.31, pp 69-74,1990.
Wright, Lloyd, Personal communication, Lam Research.
Cannon, W.A., Passivation Reactions of Nickel and Copper Alloys
with Fluorine, Transactions of the Metallurgical Society of AIME,Vol
242, Aug 1968, pp 1835-1643.
Gillardeau, J. Some Aspects of the Fluoridation of Copper and
lron. Oxidation of Metals, Vol2, NO.311970, pp 319-330.
Gillardeau, J. & Macheteau Y. The Fluoridation Kinetics of
Iron. Oxidation of Metals, Vol 4, No.3, 1972 pp 141-149.
Mellor, W.; A Comprehensive Treatise on Inorganic and Theoretical
Chemistry' Vol l-XVI, Wiley and Sons N.Y 1931.
Gmelin, Leopold, Gmelin Handbook of Inorganic Chemistry, Springen
Vevlag.
Silvestri, T., et. al., "In-situ Purity Characterization
of a Single Wafer CVD Tungsten System", Advanced Metallization
for ULSI Applications 1991, Material Research Society, 1991.
Hogle, R., et. al., "Tungsten CVD Equipment Surface Chemistry",
Advanced Metallization for ULSI Applications 1992, Materials Research
Society, 1992.
|
