 |
Description  |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a remote plasma enhanced chemical vapor
deposition (RPECVD) apparatus and method for growing an epitaxial
semiconductor layer.
2. Discussion of Background
Plasma enhanced processes have figured prominently in research efforts to
lower process temperatures. In conventional plasma enhanced chemical vapor
deposition (PECVD), the parent gas molecules are dissociated into
precursor atoms and radicals which can deposit on substrates at lower
temperatures than in thermal chemical vapor deposition. The deposition
occurs at lower temperatures than purely pyrolytic processes because the
plasma supplied energy to break chemical bonds in the parent molecules
that would only be broken by thermal decomposition if the plasma were not
present. Parent molecule dissociation is accomplished in the plasma
through various processes involving collisions with electrons, ions,
photons, and excited neutral species. Unfortunately, the precursor species
are also subject to the same active environment which dissociated the
parent molecules. This can lead to further dissociation or reaction of gas
phase species to form more complicated radicals before the radicals can
condense on the substrate In a low pressure, low power silane (SiH.sub.4)
immersion plasma, Matsuda et al. Thin Solid Films 92,171 (1982), have
shown using mass spectroscopy that there are a host of gas phase species.
These species include H, H.sub.2, Si, SiH, SiH.sub.2, SiH.sub.3,
SiH.sub.4, Si.sub.2, Si.sub.2 H, Si.sub.2 H.sub.2, Si.sub.2 H.sub.3,
Si.sub.2 H.sub.4, and Si.sub.2 H.sub.5. The most dominant line in the mass
spectroscopy is the SiH.sub.2 line, even though it is ony 12% taller than
the SiH.sub.3 line and 125% taller than the Si.sub.2 H.sub.5 line. There
is a wide spectrum of precursor species incident on the growing film. A
further complication is that in conventional PECVD the substrate is
immersed in the plasma region. This results in a large flux of charged
species incident on the substrate during film deposition. The incident
energies of these ions may be as high as 160 eV in some immersion systems
(See Chapman, Glow Discharge Processes, John Willey & Sons, N.Y. 1980,
Chap. 4). This can lead to ion implantation, energetic neutral embedment,
sputtering, and associated damage This residual damage must be annealed
out during growth if high quality epitaxial layers are to be produced.
Thus, this damage imposes a minimum growth temperature, based on annealing
conditions below which high quality material cannot be obtained. Thus,
there are two major problems associated with conventional PECVD: adequate
control over incident gas phase species, and ion damage as a result of the
substrate being immersed in the plasma region.
RPECVD deposition of silicon nitride Si.sub.3 N.sub.4 and silicon SiO.sub.2
for gate insulators in (In, Ga) As FET devices has recently been disclosed
by Richard et al. J. Vac. Sci. Technol. A3(3), May/June 1985 (pages
867-872). According to this reference, to deposit SiO.sub.2, for example,
one reactant, O.sub.2, is excited in the plasma tube remote from the
semiconductor substrate. The other reactant, SiH.sub.4, enters the reactor
separately, near the substrate and is not excited to a plasma state. An
important point is that one of the reactants, O.sub.2, bearing one of the
component atoms of the SiO.sub.2, is introduced through the plasma tube.
The process is thought to follow the following reaction model. Monosilane
(SiH.sub.4) molecules interact with the metastable oxygen O.sub.x *(.sup.3
P.sub.j) flux resulting from the remote plasma. The lifetime of the
metastable oxygen is quite long, allowing pathlengths of 1-2 meters in the
RPECVD reactor using the SiO.sub.2 deposition parameters. (In contrast,
the pathlength of a typical metastable excited noble gas specie, e.g. He*,
used in the RPECVD epitaxial growth of semiconductor layers, according to
the present invention, is 5-30 cm.) This interaction leads to disiloxane,
(SiH.sub.3).sub.2 O, formation in the gas phase. On the heated substrate,
disiloxane is further oxidized by excess metastable oxygen, O*. This
oxidation removes H from the silyl groups, SiH.sub.3. Dehydrogenation is
accompanied by oxygen bridging of silicon atoms originally bound in
adjacent disiloxane molecules on the heated surface. An excess of the
plasma excited species is used to drive the dehydrogenation of the silyl
groups to completion, minimizing Si--H bonding. Silicon-poor films do not
form; thus the process is stable. For this case, CVD can be thought of as
a polymerization of disiloxane brought about by oxidation of the SiH bonds
of the silyl groups.
Important features of the SiO.sub.2 process described by the above-noted
Richard et al article are:
1) In the SiO.sub.2 process, O is activated by the plasma in the plasma
generation region and becomes incorporated in the deposited layers.
2) The interaction between the reactive species existing the plasma
generation region and the injected reactant results in the formation of
the chemical groups.
3) The lifetimes and therefore the pathlengths of the reactive species
exiting the plasma formation is quite long: for metastable oxygen the
pathlength is 1-2 meters.
4) The dielectric material formed, SiO.sub.2, is an amorphous material and
therefore has no long-range or crystalline order. For SiO.sub.2
deposition, metastable O* promotes the further oxidation of disiloxane
adsorbed on the substrate surface, which reduces the surface of adatoms
and enhances the formation of amorphous material.
Another prior art reference of interest is an article by Toyoshima et al,
Appl. Phys. Lett. 46(6), 15 March 1985, pp 584-586, which describes a
PECVD process to deposit hydrogenated amorphous silicon. However, the
deposited a-Si:H films retain from 5-30 atomic percent hydrogen in the
deposited layers, which is critical to the performance of a-Si:H, but
disastrous if one is trying to grow epitaxial Si layers. No process used
to deposit high quality a-Si:H films has proven successful in depositing
epitaxial Si layers.
SUMMARY OF THE INVENTION
Accordingly, one object of this invention is to provide a new and improved
apparatus and method for growing epitaxial semiconductor layers on a
substrate, which overcomes the problems in the prior art PECVD techniques
above-noted, including inadequacy of the control over incident gas phase
species, ion damage to the substrate, and the lack of excited metastable
gas species at the substrate to enhance surface mobility of the adatoms
and formation of the epitaxial layer.
Another object of this invention is to provide a novel apparatus and method
employing an improved RPECVD approach, by which epitaxial semiconductor
layers can be deposited on a substrate maintained at a relatively low
temperature.
Still a further object of this invention is to provide a novel RPECVD
apparatus and method for growing epitaxial diamond layers on a substrate.
These and other objects are achieved according to the invention by
providing a new and improved RPECVD apparatus and method for growing
semiconductor layers on a substrate wherein an intermediate feed gas,
which does not itself contain constituent elements to be deposited, is
first activated in an activation region to produce plural reactive species
of the feed gas. These reactive species are then spatially filtered to
remove selected of the reactive species, leaving only other, typically
metastable, reactive species which are then mixed with a carrier gas
including constituent elements to be deposited on the substrate. During
this mixing, the selected spatially filtered reactive species of the feed
gas chemically interacts, i.e. partially dissociates and activates, in the
gas phase, the carrier gas, with the process variables being selected so
that there is no back-diffusion of gases or reactive species into the feed
gas activation region. The dissociated and activated carrier gas along
with the surviving species of the feed gas then flows to the substrate.
The surviving reactive species of the carrier gas completely react and the
surviving metastable specie of the feed gas completely order the activated
carrier species on the substrate whereby the desired epitaxial
semiconductor layer is grown on the substrate.
BRIEF DESCRIPTION OF THE DRAWINGS
A more complete appreciation of the invention and many of the attendant
advantages thereof will be readily obtained as the same becomes better
understood by reference to the following detailed description when
considered in connection with the accompanying drawings, wherein:
FIG. 1 is a schematic cross-sectional side view of a flow tube
schematically illustrating key features of the present invention; and
FIG. 2 is a schematic side view illustrating in more detail a RPECVD
reaction chamber used according to the present invention.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring now to the drawings, wherein like reference numerals designate
identical or corresponding parts throughout the several views, it is first
noted that the reactor design and operating criteria discussed hereinafter
are based on the principle of a remote region of activation of a gas or
mixture of gases. The activated gas (gases) then plays several roles
leading to the deposition of the semiconducting film. Because of the
central importance of "remote region of activation" to the present
invention, this terminology is first defined referring to FIG. 1.
FIG. 1 shows schematically a section of a flow tube 12. A feed gas stream
(single gas, vapor, or mixture) enters at input inlet 10. In the region of
activation 14, the feed gas has its chemical reactivity increased.
Chemical reactivity of the feed gas can be increased in many ways. For
example, one or more components of the feed gas may be ionized; one or
more components of the feed gas may be dissociated into more reactive
species, such as converting water vapor into hydrogen and oxygen; or the
internal energy of the feed gas may be increased without ionization. This
can be accomplished by many methods. Some of these methods can be internal
to the flow tube. A sample of these internal methods might include
heaters, or catalytic surfaces, and electron or ion bombardment sources.
Some methods could be external to the flow tube. A sample of these
external methods might include a broad range optical sources (with an
appropriately transparent tube), microwave or radio frequency power
sources, or simple heaters Whatever the feed gas(es), the combined means
for activation, or the reactive species formed, in the activation region
14, energy is coupled into one or more gases, and that energy can
contribute to subsequent chemical reactions.
In experimental studies performed to date, an external radio frequency coil
14.sub.1, shown in FIG. 2, concentric with the flow tube was used to
activate the gas stream.
Referring to FIG. 1, the concept of a "remote" region of activation in the
present RPECVD technique will be described. By remote region of activation
is meant two things: (1) the substrate is not located in a remote region
of activation; (2) in any remote region of activation, only gas(es) from
the inlet of that region of activation is(are) present, other gas(es) that
may be present in other regions of the apparatus can not reach a remote
region of activation by diffusion or other processes that would allow such
gas(es) to enter through the exit of a region of activation. To ensure
this requires both a suitable reactor design and a proper selection of
operating parameters. In the flow system of the present invention, shown
in FIGS. 1 and 2, the design of the physical separation of the various
regions of the reactor, coupled with the flow velocity of the gas stream
(which of course depends on the selection of process parameters) in those
regions, determines whether back-diffusion of gases into a region of
activation can occur.
As shown in FIG. 1, the present invention includes a feed gas inlet 10
through which a feed gas is entered into a plasma tube 12. In an
activation region 14, the chemical reactivity of the feed gas is increased
to produce reactive species of the feed gas which pass downstream of the
exit plane 14.sub.2 in the downward direction shown in FIG. 1. Between the
exit plane 14.sub.2 and the carrier gas inlet 18, the feed gas reactive
species are filtered such that only the desired specie reaches the gas
including a constituent element to be deposited inlet 18 where it mixes
inlet 18 in a mixing and interaction region 20.
In a working embodiment of the invention used to date, a radiofrequency
coil 14.sub.1 concentric with the flow tube 12 has been used to create a
"plasma" (glow discharge) of the feed gas in the activation region 14.
Working examples have used either a pure noble gas plasma feed, as
discussed hereinafter, or noble gas mixtures with hydrogen. The plasma
environment in the activation region 14 contains many species, even with a
simple feed gas like helium In fact, the feed gas reactive species
produced in the activation region 14 include ions, electrons, and a host
of excited species all with different composite lifetimes which are
influenced by various factors The flow through the activation region 14
carries the species downstream towards the carrier gas inlet 18 and a
substrate 22 mounted in a deposition region 24 downstream of the inlet 18.
The distance that the various species can travel before they are
annihilated will depend on their composite lifetimes and the flow
velocity. According to the invention, the flow velocity of the feed and
carrier gases are controlled so as to control the relative abundance of
selected of the reacted species at a given distance downstream of the
region of activation, such as at the mixing and interaction region 20 at
the carrier gas inlet 18. Thus, by controlling the gas flow rates, and by
requiring the reactive species of the feed gas to pass from the exit plane
to the mixing and interaction region 20, a spatial filtering region 26 is
provided downstream of the exit plane 16, in which undesired reactor
species are annihilated and only selected of the reactive species are
passed downstream towards the mixing and interaction region 20.
Spatial filtering as above described involves two aspects. First, some
physical distance between the activation region which excites the feed gas
and the region where the carrier gas is introduced must exist. And second,
the lifetimes of the desired reactive species must be substantially longer
than the lifetimes of those species not desired Once these criteria are
established, spatial filtering occurs because the pumping velocity of the
reactive species determines how far downstream from the activation region
various species will travel before they decay or be annihilated. For
example, in a He discharge, electron impact excites He atoms into a host
of excited electronic states. These states include 2.sup.3 P, 2.sup.1 P,
3.sup.3 S, 3.sup.1 S, 3.sup.1 P, 3.sup.3 P, 3.sup.1 D, and 3.sup.3 D. All
these states all have energies greater than the metastable 2.sup.3 S
state. However, these states quickly decay to the ground state or one of
the lower metastable states, 2.sup.1 S or 2.sup.3 S exponentially with a
characteristic decay time. This decay time is less than 10.sup.-7 s. As
the species are pumped from the discharge regions, the metastables and
ground state He atoms are dominant. Of the two metastable states, the
2.sup.1 S state has the shorter decay time or effective lifetime
2.times.10.sup.-8 sec vs 6.times.10.sup.-3 sec for the 2.sup.3 S state.
Thus, the host of highly excited He states in the plasma region have been
spatially filtered to produce a desired flux of metastable 2.sup.3 S He
atoms at the entrance to the gas mixing region. The unwanted excited
species are completely attenuated exponentially along the length of the
spatial filter compared to a factor of 3-150 attenuation (for plug
velocities of 10-50 m/sec and spatial filter length of 0.3 m) for the
desirable metastable specie. For this particular spatial filter design and
system operating parameters, all activated gas feed species having
effective lifetimes less than 4.times.10.sup.-3 sec will be completely
annihilated in the spatial filter.
The spatial filtering region 16 also acts as a backstreaming isolation
region which in conjunction with the selected gas flow rates prevents
injected carrier gas from the inlet 18 from back diffusing to the exit
plane of 14.sub.2 of the activation region 14.
The flux of activated noble gas species (and by activated it is
specifically meant in the sense of internal energy and not kinetic energy)
partially dissociates and activates (in the gas phase) the carrier gas.
The flux of the activated noble gas species completely reacts and orders
the activated carrier species onto the substrate 22 and results in the
growth of an epitaxial semiconductor on the substrate. The flux of
activated spatially filtered noble gas species enhances surface reactivity
and reactant surface mobility in the growth of a single crystal epitaxial
layer. The technique of the invention as applied to surface effects can be
used in a low pressure process where the mean free path between the exit
plane 14.sub.2 of the activation region 14 and the substrate 22 is such
that no gas phase collisions occur.
Three examples of specific semiconductive materials grown using the RPECVD
technique according to the invention are next discussed. In these
examples, there is no attempt to limit the invention to these specific
features of remote region excitation technique, remote region feed gas,
reactant feed gas, or specific reactor system design.
In all three examples, reference is made to a schematic of a remote plasma
enhanced chemical vapor deposition reactor, shown in FIG. 2. This
representation of a RPECVD reactor primarily consists of a plasma tube 12
in which is located the region of activation 14, and an activation source
such as an rf coil 14.sub.1. The plasma tube 12 feeds into a deposition
chamber 20.sub.1 in which is located a gas dispersal ring 18, and the
substrate susceptor 28. Additional components include an electron gun 30,
phosphorous screen 32, and a manipulator arm 34 used to perform Reflection
High Energy Electron Diffraction (RHEED) characterizations of the
substrate 22 and the epitaxial semiconductor film deposited thereon The
plasma tube 12 used consists of a 7.6 cm inside diameter pyrex tube. The
plasma is driven by a 13.56-MHz rf generator with matching network The
substrates 22 are clamped to a graphite susceptor 28 heated internally by
a tungsten halogen lamp (not shown). Substrate temperatures are calibrated
using thermocouples (not shown) attached to the surface of a silicon
substrate. Gasses are introduced through two separate gas feeds, the
plasma feed gas inlet 12 and the carrier gas feed 18.sub.1 to the gas
dispersal ring 18, which serves as the carrier gas feed inlet. The plug
velocity of He or other noble gas through the 7.6 cm plasma tube 12 is
high enough to prevent back-diffusion of GeH.sub.4, SiH.sub.4, or
CH.sub.4. The plug velocities used are 3, 5, and 100 m/s for germanium,
silicon, and diamond depositions, respectively. Also shown is an outlet 36
for high vacuum pumping via a turbomolecular pump (not shown), an outlet
38 for pumping the process gasses using a roots blower (not shown)
together with a direct drive mechanical pump (not shown). Typical
pressures are less than 5.times.10.sup.-10 Torr minimum base pressure when
the process gasses are not flowing and 1-300 mTorr during epitaxial growth
of a semiconductor layer. The vacuum intake to the roots blower is
ballasted with a constant gas load to prevent antibackstreaming of oil
vapors. Examples illustrating use of this process to epitaxially grow
silicon, germanium, and diamond semiconductor layers are described below.
Epitaxial growth of germanium is accomplished by flowing 200 sccm of He
through the plasma tube and 20 sccm of 0.1% GeH.sub.4 in He through the
gas dispersal ring 18. The pressure is controlled at 200 mTorr. To
initiate deposition, 100 W is applied to the rf coil creating a He
discharge plasma in the activation region 14. The substrate temperature is
typically maintained between 225.degree.-450.degree. C., preferably at
300.degree. C., during growth.
Epitaxial growth is thought to occur through the following processes. The
rf energy coupled to the plasma tube establishes a He plasma in the
activation region 14. Through a variety of reactions many different
species of excited He atoms and ions are created in the plasma, each
having its own lifetime. These various species are caused to flow down
from the plasma tube toward the gas dispersal ring 18 and the substrate.
Each specie can be annihilated through a variety of mechanisms, and
therefore, each specie has an average time that it can survive, or
effective lifetime, until it is annihilated. This lifetime can be
translated into an average distance it will travel below the plasma tube
exit plane 14.sub.2 before it is destroyed. This distance is called the
pathlength. The pathlength of a specie is determined by the effective
lifetime of the specie and the plug velocity of the He gas flow.
Consequently, the system and the growth parameters can be designed and
chosen to cause undesired species to be spatially filtered in the spatial
filtering region 16 and the desired specie to interact with the reactant
molecules and to arrive at the substrate surface. In the present example
the specie desired to interact with the reactant GeH.sub.4 is the
metastable He(2 .sup.3 S). These metastables play three important roles in
the overall growth process:
1. They dissociate the germane molecules through inelastic gas phase
collisions;
2. They have inelastic collisions on the growth surface of the film which
enhances the surface mobility of the impinging species leading to epitaxy
at low surface temperatures; and
3. They can also play a role in dehydrogenation of surface reactants.
These three functions for the metastables will be discussed in further
detail below.
The metastable He(2 .sup.3 S) interacts with the GeH.sub.4 through
inelastic gas phase collisions, and creates several reaction products.
These products may include ionized and neutral GeH.sub.x species, where 0
<.times..ltoreq.4. The most probable products are GeH.sub.4 +, GeH.sub.3 +
and GeH.sub.3 ; and the desirable product is GeH.sub.3. As the radicals
condense on the substrate they must cross-link to form a germanium network
If this process is to form epitaxial layers of germanium, excess hydrogen
carried by the free radicals must be liberated and the reactant species
must have sufficient surface mobility to form an ordered solid. In the
RPECVD process hydrogen removal occurs when He metastables collide with
the growth surface.
Epitaxial growth of silicon proceeds much in the same manner as growth of
germanium. Again, growth is accomplished by flowing 200 sccm of He through
the plasma tube 12 and by flowing 100 sccm He and 1 sccm SiH.sub.4 through
the gas dispersal ring 18. A rf discharge plasma of 30 W is sustained
during deposition. The deposition process occurs at a total pressure in a
range of 50-300 mTorr, with 200 mTorr being preferred. The deposition rate
is approximately 0.01 nm/s on a Si(100) 1.times.1 surface at 520.degree.
C. Epitaxial growth has been achieved at temperatures as low as
200.degree. C. with best results occurring at about 400.degree.C. The role
of the metastable He in the epitaxial growth of silicon is thought to be
much the same as described above for the epitaxial growth of germanium.
Epitaxial growth of diamond may be accomplished by flowing a Noble gas (He,
Ar, or Xe) through the plasma tube and methane, CH.sub.4, through the gas
dispersal ring. One important factor that distinguishes growth of diamond
from growth of either germanium or silicon is the poly-phasic nature of
the deposited material. Depending upon the growth conditions, the
deposited layers may be diamond, graphite, amorphous or glassy carbon, or
mixtures of these materials. When a hydrocarbon such as methane is excited
in a plasma, radicals of the form CH.sub.x are formed. As in the silane
example, these radicals interact in the gas phase to form carbon-carbon
bonds. The added complication in the carbon case results from the ability
of carbon to form not one, but three hybridizations. Thus we get
carbon-carbon bonding of the ethane form (sp.sup.3 hybridization), of the
ethylene form (sp.sup.2 hybridization), and of the acetylene form (sp
hybridization). The parallel between these gas phase precursors and their
solid phase analogues is striking. Diamond (sp.sup.3 hybridization) has
ethane type bonding, graphite (sp.sup.2 hybridization) has ethylene type
bonding, and carbynes (sp hybridization) are chainlike compounds with
acetylene type bonding. To grow semiconducting diamond it is necessary to
preclude the incorporation of wrong bonds of graphite-like or carbyne-like
hybridization. The flux of gaseous precursors with incorrect hybridization
onto the film surface is inevitable if the undesirable methane radicals
(i.e., the ethylene and carbyne) are formed. Consequently, the design of
the growth reactor and the choice of the growth parameters must be
properly chosen to form precursors which upon condensation on the
substrate promote sp.sup.3 hybridization and diamond growth.
The growth of diamond proceeds technically in a similar way as does silicon
and germanium. Typically, 500 sccm of He flows through the plasma tube 12
with a 30 sccm dilute mixture of He, H.sub.2, and CH.sub.4 (4:10:1 by
volume) flowing from the gas dispersal ring 18. A rf discharge of 80 W is
sustained in the activation region 14 during deposition at a total
pressure range of 10-1000 mTorr, typically less than 100 mTorr. The
substrate temperatures is varied from 650.degree.-850.degree. C. The
quartz plasma tube size is 1.5 in. o.d. insuring a high plug velocity
necessary for transporting metastables and radicals to the substrate.
Using these growth parameters, diamond films have been grown at the rate
of approximately 2000 .ANG./hr.
While the process technically is very similar to the silicon and germanium
growth, the proper choice of noble gas and methane diluent is critical for
promoting diamond growth. Because the energy of the He metastable is so
high (.about.20 eV), the cross-section for collisional dissociation of the
CH.sub.4 molecule is low. Thus, the depositional precursor species created
by the He are CH.sub.4 +, CH.sub.3 +, or CH.sub.3, all of which are highly
saturated CH.sub.x radicals. Also the choice of methane diluent becomes
more pertinent. Unlike the growths of silicon and germanium where the
silane and germane were diluted in He, diamond growth is more facilitated
with hydrogen dilution. The hydrogen serves two roles. First as a source
of atomic hydrogen to the nucleating film, it more preferentially etches
the graphitic bonds than the diamond bonds. Second, it moderates the gas
phase chemistry promoting higher saturation of the CH.sub.x radicals.
In the growth of diamond, H.sub.2 is used in surface reactions such as the
etching of graphitic bonding units and in gas phase reactions to convert
sp and sp.sup.2 bonded hydrocarbon radials to sp.sup.3 bonding forms. For
this purpose the H.sub.2 source gs is activated by inversion to atomic
hydrogen, H(H.sub.2 +energy.fwdarw.2H). In the basic reactor design shown
in FIG. 2 this activation is carried out in one of two ways. By one
method, Case 1, the H.sub.2 gas enters with the carrier gas stream through
the gas disposal ring 18. The H.sub.2 is activated through interaction
with energetic species of the feed gas that have passed through the
spatial filter 14. Because this technique relies on energetic species of
the feed gas (ex. metastable He (2.sup.3 S)), to activate both the
hydrogen and the methane, dramatic reduction in deposition rate is
observed as H.sub.2 :CH.sub.4 flow ratios are increased beyond 20:1. This
limitation places a severe restriction on the range of gas mixtures for
which reasonably efficient diamond deposition rates can be achieved (rates
<1 .ANG./sec).
The second method (Case 2) of operation seeks to overcome this limitation.
The H.sub.2 is introduced with the noble gas feed through the plasma
region. This allows direct activation of the H.sub.2 (H.sub.2
+energy.fwdarw.2H) by the plasma. However if this scheme is used at powers
typical for pure noble gas plasmas (80 W), dramatic reduction in the
production of noble gas metastables by the plasma is observed. This is due
to the fact that H.sub.2 dissociation occurs at lower energies than He
metastable production. Thus the presence of H.sub.2 shifts the electron
energy distribution of the plasma region to lower energies where
metastable noble gas production is inefficient. If it is attempted to
overcome this limitation by brute force (i.e. just by increasing the
plasma power) a practical limitation is experienced in that high power
H.sub.2 noble gas plasmas etch the plasma tube releasing (for the usual
case of quartz or pyrex tubes) silicon, oxygen, and a variety of trace
contaminants that travel downstream with the gas flow and are incorporated
in the growing diamond film.
A comprehensive solution to the problem involves an important extension of
the concept of remote region of activation, that is, the use of multiple
remote regions of activation. In this specific case two separate plasma
tubes are operated. One with a H.sub.2 gas flow and plasma conditions
optimized for H atom production, and one with a noble gas flow (e.g., He)
with plasma conditions optimized for metastable production. This scheme
avoids the tube errosion and contamination problems noted in Case 2 above,
because mixed H.sub.2, noble gas plasmas are not formed and H.sub.2 plasma
power densities are significantly reduced (reduction factor >10). At the
same time the separate H.sub.2 plasma is more efficient (by orders of
magnitude) in the creation of atomic hydrogen as compared to the
metastable activated scheme (Case 1 above). In addition since the cross
section for energy exchange between excited metastable noble gas species
and atomic hydrogen is much less (order of magnitude) then the same cross
section for molecules H.sub.2, the multiple remote regions of activation
concept allows both higher deposition rates (rate .about.7.ANG./sec) and a
much broader range of accessible effective H:CH.sub.4 ratios than Case 1
above, thus allowing improvements both in film quality and deposition
rate.
It should be noted that this is one specific example of the broad concept
of multiple remote regions of activation, where such regions can differ in
many ways including source material (e.g., gas feed as in the example
above), means of activation (RF, thermal, etc.) spatial filter design, and
so on. Indeed, the multiple activation region concept permits optimization
of spatial filter designs which for mixed gas sources otherwise would of
necessity involve design compromises based on the differing criteria for
exclusion of the unwanted reactive species of two different parent gases.
Further commenting on the spatial filtering employed according to the
present invention, the design of the spatial filter in large part provides
the flexibility in the remote region of activation scheme. First, it is
noted that the region of spatial filtration works in both directions.
First it is optimized to transmit the desired excited species from the
region of activation (e.g., R.F. plasma) to the mixing region while
suppressing the transmission of other excited species created in the
region of activation. Second and equally important, it supports the remote
aspect of a region of activation by allowing the suppression of
back-diffusion of any species present in the deposition chamber. Thus
without a spatial filter the region of activation cannot be remote.
In order to realize the flexibility of the scheme, it is important to
realize that the spatial filter is not simply a time delay where merely
the intrinsic lifetime of the various reactive species is allowed to
change the relative distribution of reactive species at different planes
(times) downstream of the region of activation. If this was true, the
technique would be restricted to enhancing only the intrinsically longer
lived species and would be unable to separate species with nearly equal
intrinsic lifetimes.
Fortunately the intrinsic lifetime is only one component of the crucial
parameter, the effective lifetime. The effective lifetime can be
controlled in many cases. Consider the example presently disclosed herein
for epitaxial growth of semiconductors (Si, Ge, C), oxygen and oxygen
containing gases (e.g., H.sub.2 O) are common contaminants of feed gases
such as He. Oxygen is readily activated to a metastable state in an R.F.
plasma. In addition metastable oxygen species have lifetimes far longer
(factor of 100) than noble gas metastables. Thus a small amount of oxygen
or oxygen containing contaminants in the feed gas will have a
disproportionately large effect on the growing film. However by
incorporating a wall element 16.sub.1 providing an aluminum wall surface
in a portion of the spatial filter we effectively quench the metastable
oxygen while having no appreciable effect on the metastable noble gas
flux. Wall interactions are extremely important (dominate at pressure <10
Torr) in determining effective lifetimes, thus flow dimensions and
materials of construction of the spatial filter can be used to engineer
the relative effective lifetimes and the resultant transmission
characteristics of the spatial filter. Note that in this case (oxygen in
noble gas) the spatial filter is design | | |
