 |
Description  |
|
|
TECHNICAL FIELD
The present invention relates to a magnetron sputtering cathode having
improved transfer efficiency and target utilization which is suitable for
the coating of flat disk-shaped substrates such as silicon wafers or rigid
magnetic disks.
BACKGROUND
Sputter coating is a process where atoms of a solid target are ejected by
the bombardment of energetic ions onto the target. The collection of these
sputtered atoms on a nearby object, called a substrate, coats the
substrate with the target material. The source of the bombarding ions is
commonly a gas discharge, where collisions between electrons and neutral
gas atoms results in the generation of electron and gas ion pairs, the
ions having a positive charge. A negatively charged electrode (cathode)
placed in the gas discharge attracts the positive ions causing the ion
bombardment responsible for sputtering.
The target is consumed by the sputtering process and requires periodic
replacement. A cathode assembly supports the target, provides water
cooling, sets up a magnetic field in the region of the gas discharge and
shields non-target portions of the cathode from unwanted ion bombardment.
Most sputtering systems operate with the target at a negative potential,
with a grounded metal chamber acting as an anode. The gas discharge is
usually made from argon gas at pressures in the range of 1 to 20
millitorr. (Atmospheric pressure is 760 Torr).
Argon is the gas of choice because of its chemical inertness, relatively
large atomic mass, and relatively low cost. Electrical gas discharges can
be achieved with any gas, but if a chemically reactive gas is chosen, it
will react with atoms sputtered from the target to yield a coating which
is the reaction product of the two constituents. When this is
intentionally done, the process is termed reactive sputtering. An example
of reactive sputtering is the sputtering of a titanium target in a
nitrogen-argon gas mixture to yield a coating of titanium nitride.
Residual atmospheric gas contaminants present in the gas discharge will
also react with the coating material resulting in its contamination. Since
this is to be avoided, many sputtering systems are evacuated in the region
of the discharge to pressure levels of 1.times.10.sup.-7 Torr or less
prior to introduction of the ion providing gas.
Another method of minimizing coating contamination is achieved by
increasing the coating rate. This method is effective because at a given
constant residual gas pressure, the degree of coating purity is directly
proportional to coating rate. Thus, a doubling of the coating rate has the
same effect as halving the residual gas pressure.
Many sputter coating applications require that the temperature of the
substrate be regulated to achieve optimum coating quality. The substrate
support is accordingly provided with a heating and/or cooling means.
Similarly, coating quality can also be improved if the substrate is
subjected to a moderate negative bias with regard to the gas discharge.
This method, termed bias sputtering, causes positive argon ion bombardment
of the coating during its growth, which bombardment can have beneficial
effects on coating porosity, stress and conformality.
The sputtering process may be used as a material removal or surface
cleaning method. For this use the discharge chamber is not equipped with a
sputtering cathode. Instead, the object to be cleaned or etched becomes
part of the primary system cathode. The ensuing ion bombardment of the
substrate removes surface contaminants and may also remove some of the
bulk atoms of the object. This process is termed sputter-etching and is
frequently used in a sputtering system as a preparatory step prior to the
deposition of target material onto a substrate.
A measure of transport efficiency for a sputter coating target is the
fraction of the material sputtered off the target which reaches the
substrate that is being coated. Thus, 50% efficiency represents the
condition where half the material which sputters off the target due to ion
bombardment coats the substrate. The remaining sputtered material usually
coats other parts of the sputtering process chamber. There are many
reasons why it is very desirable to increase transport efficiency.
Firstly, many sputter coating materials are costly and since the unused
sputtered material cannot effectively be recovered, this represents a
significant additional coating cost.
Secondly, the unused sputtered material usually coats other sputter chamber
components where, after a time, it causes problems such as particulate
generation or electrical shorting of insulators. Consequently, the coating
process must be interrupted for purposes of removing this extraneous
material. Such maintenance typically requires that many hours elapse
before coating can resume and therefore also represents a significant
operating cost.
Thirdly, typical target volumes, being limited by cathode design
constraints, require that the sputter coating process be interrupted for
purposes of spent target replacement. Here again, the coating process is
disrupted.
Fourthly, the deposition rates achievable with sputtering processes are
inherently low, and are frequently limited by the size and cost of
available power supplies, or limited by heat dissipation factors inherent
in cathode designs. These factors limit the rate at which material can be
sputtered off the target source. Thus, an improvement in transport
efficiency allows for an improvement in coating rate, within the above
limitations and consequently results in coating process productivity
improvements.
Fifthly, sputter coating quality is frequently impaired by the
incorporation of atmospheric gases such as oxygen, nitrogen and water
vapor. This gas incorporation rate is dependent upon the relative arrival
rates of gas atoms versus the sputtered atoms at the substrate. A high
sputtered atom arrival rate therefore favors increased coating purity.
The low rate of arrival of sputtered atoms requires that very stringent
measures be taken in the design, construction, and vacuum pumping of the
sputtering apparatus to assure lower atmospheric gas impurity arrival
rates. These measures lead to equipment cost increases, and usually also
imply additional maintenance and operating costs. In many instances
sputter coating quality is limited by the available state-of-the-art
vacuum technology and practice.
U.S. Pat. Nos. 4,428,816 to Class et al and 4,472,259 to Class et al
disclose magnetron sputtering cathodes having improved transfer
efficiency. Such cathodes are suitable for the coating of substrates only
when the substrate is caused to travel past the cathode in a linear
fashion. Such cathodes are suitable therefore only in a sputtering
apparatus which includes a scanning mechanism. The absence of such
substrate motion, results in unacceptable thickness non-uniformity of the
substrate coating.
There are reasons why the requirement for substrate translational motion is
a disadvantage. First among these is the fact that many coating processes,
including "planarized" aluminum coatings require close control of
substrate temperature, and in addition require the application of radio
frequency (RF) substrate bias power during the coating process. The
application and control of substrate heating combined with the
requirements of Rf power plus translation pose serious design and handling
problems which increase equipment cost and degraded performance
reliability.
The achievement of coating thickness uniformity combined with high transfer
efficiency is difficult. This is because magnetron discharges are
inherently non-uniform in their degree of ionization (plasma density).
Consequently, the rate at which material is emitted from the surface of
the sputtering target is also non-uniform, tending to be localized in
linearly continuous regions of the target known as "racetrack" regions. As
the substrate is brought closer to these racetrack regions, transfer
efficiency (the fraction of the target that reaches the substrate) is
improved but coating thickness uniformity is impaired.
The deposition profile at a substrate that is being coated by material
sputtered from a racetrack pattern of a planar target can be modeled using
the Knudsen cosine law. The cosine law states that the maximum emission of
sputtered material from an emitting surface occurs along the direction
which is perpendicular to the surface. The emission in any direction which
is inclined at an angle, .theta., with regard to the surface
perpendicular, is less than this maximum emission by the cosine of the
angle .theta.. For flat, disk-shaped substrates, the obvious choice for a
coating source is a cathode which yields a ring shaped racetrack which is
the emitting source of the coating material. Thus, the sputtering target
can be a planar disk, with a circular racetrack on it, and the substrate
placed nearby, with its surface parallel to the target plane, and its
center coaxial with the centerline of the racetrack. The deposition
properties of such ring emitting sources have been documented in the
literature. See, for example, L. Holland, Vacuum Deposition of Thin Films,
Chapman & Hall Ltd., London, 1963, pages 152-156 and L. Massel and R.
Glang, Handbook of Thin Film Technology, McGraw-Hill, New York, 1970,
pages 1-56 to 1-59.
These studies show that there is an optimum spacing between source and
substrate which yields the greatest deposition rate uniformity. For
separations greater than this optimum, the deposition rate is greatest at
the point where the axial centerline of the target ring intercepts the
substrate plane. From this maximum, the rate decreases monotonically with
radial distance from the central point, yielding a dome-shaped deposition
pattern. For source to substrate separation less than the optimum the
deposition pattern assumes a saddle-like shape which shows a maximum at
the substrate radial position equal to the ring-source radius. This
maximum becomes more pronounced as the source to substrate distance is
further decreased.
The practical application of these principles to a ring shaped magnetron
source requires that the zone of uniform deposition be approximately 10%
greater than the substrate diameter. For these reasons, the achievement of
uniform deposition on a substrate having a radius R.sub.s requires that
the racetrack radius be approximately 1.1 R.sub.s and the source to
substrate distance be equal to approximately 1.3 R.sub.s. Such sources
typically have a transfer efficiency of 15 to 18%. Such sources are
therefore limited in transfer efficiency because any attempt at decreasing
the source to substrate distance, results in a non-uniform, saddle-shaped
deposition profile.
Another limitation associated with this magnetron sputtering configuration
is the manner by which progressive erosion of the target influences both
the coating thickness profile as well as the utilization efficiency of the
target. It is known that the localized ion bombardment associated with the
magnetron geometry causes a target surface to be locally sputtered away.
Thus, the target surface develops a localized groove in the racetrack
region. As sputtering progresses, this groove becomes progressively
narrower and deeper. As a consequence, the angular range over which
sputtered atoms leave the target surface is narrowed i.e., becomes "over
cosine". Thus, the coating rate at the substrate progressively increases
for those portions of the substrate immediately adjacent to the racetrack
region of the target and progressively decreases elsewhere. The resulting
progressive coating thickness non-uniformity limits the useful coating
life of the target, with the attendant costs associated with target
replacement.
This problem is accentuated by a reduction in target to substrate distance.
This arises because, at the greater substrate distance, collisions between
neutral gas atoms and the sputtered atoms have the effect of broadening
the angular range over which the sputtered atoms travel after leaving the
target surface. This gas scattering has the effect of compensating for the
deposition profile narrowing associated with a deepening target racetrack
groove. As the target to substrate distance decreases, the opportunity for
gas also decreases, thereby accentuating the aforementioned problems of
film thickness non-uniformity and target utilization.
Another undesirable aspect of the development of a deep and narrow
racetrack groove is that this causes the target to become eroded through
its useful thickness before an optimum volume of the target material is
used to achieve substrate coating. Consequently, more frequent target
replacement with its attendant cost is mandated. For these reasons, a
means of achieving a broad target erosion pattern is desirable.
Another factor which enters in the practical application of ring shaped
magnetron sputtering sources is related to the conformality of the
coating. Semiconductor wafer substrates are not flat when viewed on a
microscopic scale. Instead, they have micrometer sized features such as
steps, and square sided holes which have a depth equal to the lateral
dimensions. One example would be a square hole measuring one micrometer in
length and width and one micrometer in depth. Another example would be a
groove measuring one micrometer in width and depth, with a length measured
in tens or hundreds of micrometers. There are many reasons why it is
desirable that the coating should conformally cover these features, i.e.
that the coating thickness at the bottom or side-wall of a step or hole be
equal to that at the top. This is a problem with sputter deposition
because the majority of sputtered atoms which leave the target travel to
the substrate without gas scattering and in a straight line path.
Furthermore, most sputtered atoms move only short distances on the
substrate surface after arrival. As a consequence, if a given substrate
region does not lie on a direct line of sight path to an emitting region
of the target, the sputtered atoms from that region of target are shadowed
and do not contribute to the build-up of a coating in the given substrate
region. Alternatively, some substrate regions may have a broad line of
sight view of the target and therefore achieve coating thicknesses which
are much greater than the average. As a consequence, the sputter coating
of a vertical step can result in a layer which is thick and overhangs the
top of the step leaving only a thin coating on the sidewall and bottom of
a step. Also, for substrates coated using the ring configuration, it is
not unusual to find that a substrate step which faces radially outward
experiences a different coating coverage from one which faces radially
inward. From this perspective it is desirable that the target emitting
region be very broad, thereby affording the greatest possible viewing
angle to all substrate sites.
The aforementioned problems have been partially addressed by a variety of
methods disclosed in prior issued patents. It has been known for some time
that the shape of the magnetic field needed to establish the target
racetrack has a direct influence on the width of the racetrack groove.
This magnetic field is produced by opposite magnetic poles situated behind
or adjacent the target. These cause arch shaped magnetic flux "lines" to
emerge from the target in the vicinity of the north magnetic pole, arch
over and then reenter the target in the vicinity of the south magnetic
pole. Discharge electrons are trapped by the combination of magnetic field
and the strongly negative potential of the target surface. This
combination of electric and magnetic field also induces a sideways
magnetron drift to the electrons. To prevent the loss of these electrons,
the arch shaped magnetic field is made to close on itself, thereby forming
an endless "tunnel" adjacent to the target face. The discharge is confined
in this region and tends to be most highly ionized in those places where
the arching magnetic field is substantially parallel to the target face.
Accordingly, a highly arched field produces a narrow racetrack, and a
gently curved arch produces a desirable broad racetrack. U.S. Pat. No.
4,162,954 to Morrison discusses these target racetrack characteristics.
Similarly, U.S. Pat. No. 4,457,825 to Lamont, Jr., discloses this art as
applied to a ring source. The '825 patent also discloses the use of a ring
shaped target with inward sloping faces like the frustum of a cone.
Recalling the cosine law, one might expect such a target to emit more
sputtered material radially inward. Accordingly, with such a target one
can achieve a closer target-substrate spacing without the uniformity loss
associated with the saddle-shaped deposition profile.
The characteristics of such a source are described in an article by J. C.
Helmer in the Journal of Vacuum Science and Technology, Second Series,
Volume 4, Number 3, Part 1, May/June 1986, pages 408-412. At an operating
pressure of 6mTorr argon gas, this cathode exhibits a 21.9% transfer
efficiency, and yields an aluminum sputter deposition rate of
approximately 1800 angstroms per minute at an applied D.C. power of 1563
watts. From the Helmer reference it may be inferred that the useful
deposition zone radius, R.sub.s is yielded by a racetrack radius of equal
dimension, i.e., R.sub.s =RR, and the substrate spacing is approximately,
875 R.sub.R. A closer spacing than this causes the familiar saddle shaped
deposition profile and deposition uniformity loss.
Another method of achieving a broad erosion pattern on a planar target is
disclosed in U.S. Pat. No. 4,444,643 to C. B. Garrett. Here the entire
field forming magnet structure is mechanically moved to cause the
associated racetrack discharge to continually traverse the target face.
The approach has the disadvantage of requiring mechanical motion with the
inherent issue of reliability. No substrate spacing is referenced in the
patent. Planar magnetron devices of this class are known to require a
minimum substrate spacing of 5 cm to 6 cm to minimize electron bombardment
of the substrate. (See for example, the article by W. Class and R.
Hieronymi, Solid State Technology, December, 1982, pages 55 to 61.) At
this spacing a substantial amount of sputtered material is lost from an
annular region within 6cm from the target periphery.
Yet another method for achieving a broad erosion pattern is described in
U.S. Patent No. 4,401,539 to Abe et al where an auxiliary magnetic field
is used to displace the position of a racetrack produced by a primary,
field-producing electromagnet. The auxiliary magnetic field is achieved by
the use of an electromagnet coil which shares one of the two cylindrical
pole pieces of the primary, field-producing electromagnet. The primary,
field-producing electromagnet, having cylindrical symmetry, causes a
circular racetrack to be generated on the face of a disk-shaped planar
target. The activation of a current in the auxiliary magnet field coil,
having a polarity opposite to the current in the primary coil causes a
reduction in the diameter of the racetrack. This diameter reduction is
proportional to the magnitude of the auxiliary coil current. A programmed
coil current modulation applied to this auxiliary electromagnet coil
permits the achievement of a racetrack having a diameter which
synchronously varies with auxiliary coil current. Accordingly, the
auxiliary coil current is used to cause a broad erosion pattern on the
target as well as to compensate for the coating thickness non-uniformities
which develop as target erosion progresses.
The characteristics of this source are:
R.sub.R =0.85 R.sub.s to 1.0 R.sub.s
and target-to-substrate spacing is 1.3 R.sub.s. No transfer efficiency
characteristics are available for this source, but the geometry predicts
that it is between 15 and 20%. The primary benefits of this approach
reside in improved target utilization and film thickness uniformity.
The aforementioned sources all have disadvantages in film step coverage
because of the limited angular range with which sputtered atoms arrive at
the substrate. This range is limited by the size of the target erosion
zone as well as the separation between substrate and target since, for a
given erosion zone width, the angular arrival range varies inversely with
the spacing.
U.S. Pat. No. 4,604,180 issued to Y. Hirikawa addresses this by combining
two separate concentric ring sources. The inner ring source is achieved on
a disk shaped flat target, and the outer ring source is achieved on a
target with an inwardly facing surface shaped as the frustum of a cone.
The relative sputter emission rate of each ring source is controlled by
separately applying D.C. power to each ring source. Alternately, the rings
are maintained at a common potential, and the power splitting between ring
sources controlled by varying the coil current in the electromagnet.
U.S. Pat. Nos. 4,606,806 to J. Helmer, 4,595,482 and 4,627,904 to D. Mintz
and No. 4,569,746 to M. Hutchinson similarly disclose separately powered
concentric ring sources, one planar and the other the frustum of a cone,
where the magnetic field producing means are two electromagnet coils which
couple to a ferromagnetic pole-piece structure. The coils share one pole
piece and the coil currents are such that the magnetic flux induced by the
coils is always additive in the shared pole piece. As a consequence, two
separate magnetron discharges are formed; one adjacent to the planar ring
shaped target, and the other adjacent to the conical ring shaped target.
Deposition uniformity is achieved by separately adjusting the electrical
power to the separate discharges such that at a given substrate spacing,
the deposition contributions add to give the desired uniformity.
As revealed in U.S. Pat. No. 4,627,904 to D. Mintz, the power to the outer
ring target is five to twelve times greater than that applied to the inner
target. As will be shown in the description of the inventive cathode
disclosed here, such a power splitting will only give a uniform deposition
profile for a substrate radius R.sub.s when the outer ring racetrack
radius R.sub.R is of approximately equal radius, and the substrate spacing
is approximately equal to 0.875 R.sub.R. This is very similar to the
configuration described in U.S. Pat. No. 4,457,825 to Lamont Jr. Similar
transport efficiencies are therefore predicted. This configuration will
not permit extended operation at closer substrate spacing because only
limited means is provided for minimization of the associated pair of
racetrack grooves, which at a closer spacing would result in a deposition
pattern having dual peaks with an intermediate valley associated with the
annular non-sputtering region between the inner and outer targets.
Another means of overcoming the shortcomings of a single ring emitting
source is disclosed in U.S. Pat. No. 4,622,121 to U. Wegman, et al. Here,
again, two separate annular sputtering zones are established | | |
