Note: Descriptions are shown in the official language in which they were submitted.
~7~5~9
This invention relates to end point control in plasma
etching and particularly to end point control by the detection of a
substantial varlation in the intensity of an optical emission line
from the plasma. The invention herein described was made in the course
of or under a contract or subcontract thereunder with the Department
of the Army.
In plasma etching, for various reasons, such as
variations in starting temperature, plasma conditions and etching area,
there is normally considerable variation in etching from run to run.
It ts therefore dif~icult to ensure that etching is terminated at the
correct time (i.e. end point) for optimal etching, i.e., when the layer
to be etched ts removed but patterns in the layer are nolt overetched
and the underlytng layer is not significantly etched into.
The present invention provides for accurate end point
control by detection of a change in an emission line when the u~derlying
layer has been reached, with a change in material at that level.
The invent~on will be readily understood by the
following description of typical apparatus and some examples of
materials and em;ssion lines, by way of example, in conjunction with
the accompanytng drawings, in which:-
Figure 1 is a diagrammatic illustration of a plasmareaction chamber with associated apparatus;
Figure 2 is a diagrammatic illustration of reaction
chamber and emission detector;
Figure 3 is a typical emission signal for one etchant
gas for a particular type of wafer;
Ftgure 4 illustrates the use of dummy wafers for end
point detection of particular materials;
Figure 5 illustrates emission signals for photoresist
stripping.
A typical plasma etching apparatus is illustratPd
~L07157~39
in Figure 1, where the reaction chamber is seen in cross-section
normal to its axis. A suitable etchgas, typically CF~-~5%02, or
just 2 for plasma stripping of photoresist is introduced into
the cylindrical quartz reaction chamber 10 through a needle valve
12 and maintained at pressure P by the vacuum pump 13 which
evacuates the reaction chamber 10 through the throttle valve 14.
The etchgas flowrate is set by adjustment of the needle valve 12
and monitored accurately with the mass flow meter 15; the
pressure is set independently with the throttle valve 14 and
monitored with a precision pressure gauge 16. Flowrates of 5 to
500 cc/min. and pressures of 0.25 to 1.5 Torr are typical. A
plasma is generated in the reactor by applying power from the RF
power supply 17 to the capacitor plates 18 located round the
reaction chamber 10, power levels of 10 to 500 watts being
typical. An impedance matching network in the RF power supply
17 is used to optimize power coupling into the plasma. Etchrates
are made more uniform and controlled by use of a perforated metal
RF shield 19 which serves to confine the plasma to the annular
region outside it thereby reducing the temperature and temperature
nonuniformity at the workpieces 20 which are generally loaded in
quartz holders.
In a typical etching process, the workpieces are
loaded in the reaction chamber, the chamber is evacuated, the
etchgas is introduced, and etching begins when the RF power is
switched on. The reactor temperature rises during etching and,
because the etchrate is very temperature dependent, the thickness
etched in a specified time depends sensitively on the thermal
characteristics of the reactor and workpieces (i.e., starting
temperature~ thermal mass, and thermal time constant). For
material like Si, poly Si, and Si3N4 and Ti the etchrates are
high and depend also on the area of the workpieces exposed to the
~7~57~
etchgas, the separation between adjacent workpieces and on
workpieces position and number. The latter dependence occurs
because there is competition amongst the workpieces for
available etchant species; and accordingly the etchrate decreases
with increasing number of workpieces. Also workpieces in the
centre of a batch etch more slowly than the end wafers due to
the depletion effects.
As a result of the dependences on temperature,
plasma conditions, and work load, the times required for the
same etching application generally vary considerably from run
to run thus making it necessary to monitor the progress of etching
so that the etchcycle can be properly terminated. Irreparable
damage is done to the workpieces if they are overetched, so that
end point monitoring or control methods are important and must
be precise. For example, in the fabrication of semiconductor
devices, plasma etching is generally applied to the photoresist
masked etching of patterns in thin layers on slowly etching
substrates. The substrate then serves as a partial etchstop
and the effect of overetching is to cause undesirable etching
under the mask edges (undercutting) and excessive etching into
the substrate layer. In some cases the substrate layer is a thin
layer over a fast etching lower layer and penetration of the
etching into this lower layer can cause irreparable damage.
While it is possible to monitor etching by visual
observation, this is often difficult. Reliance is placed on
there being some visual change - such as a change in colour of
the surface being etched. This is not always available and
depends to a large extent on the perceptual accuity of the
operator. Also, constant observation by the operator is
necessary.
The present invention monitors the optical
_ ~7~ii7S~
emission from the plasma. It is possible to select emission lines
that are sensitive to changes in the material being etched,
thereby indicating the onset oF etching into an underlying layer
at the endpoint. One form of apparatus for monitoring the
plasma emission is illustrated in Figure 2.
The optical emission from the plasma at 21 (in
the reaction chamber 10) is viewed by the photodetector 22
through the filter 23, the passband of which is chosen to select
an emission line the intensity of which is particularly sensitive
to changes in the material being etched. The photodetector
signal is continuously monitored on a strip chart recorder 24,
and can also be input to a comparator circuit 25 which triggers
shut down of the etchcycle via an output through connection 26
to the RF power supply 17. A photomultiplier tube with an
electrometer output amplifier can be used as ~he detector but
solid state detectors are preferable as they are more compact
and less expensive.
Experiments using a monochromator showed that
a line at about 0.70 ~m wavelength changed significantly when the
wafer was changed from SiO2 to Si for etching in CF4+02. The
signal amplitude for this emission line was unchanged when the
SiO2 wafer was removed or when many SiO2 or Si3N4 wafers were
inserted. Addition of Si wafers caused large decreases in the
intensity of this emission line, the fractional decreases being
greatest for large wafer loads, high RF power, and low CF4+5%02
etchgas pressure. The Si/SiO2 signal ratio decreased slightly
with increasing temperature, and all signals decreased sharply
with addition of N2 to the etchgas. The decrease in the 0.70 ~m
emission line intensity with increased etchrate or waferload
suggests that the intensity of this line may depend on the number
of available Si etching radicals in the plasma which is depleted
-- ~C3 7~57~3?
as the etchrate or load is increased. Alternatively, this
emission line may be poisoned by the Si etching reaction products.
An example of the endpoint monitor signal
achieved in a typical polysilicon etching run is shown in
Figure 3. The original trace had better than 30:1 signal to
noise ratio, and the signal change of 40% was achieved for the
etching of a single photoresist patterned wafer of Poly-Si over
SiO2 at 50 watts in 0.25 Torr CF~5%02 etchgas at 50C. The
signal level at 30 is due to background illumination. The signal
increases at 31, when the plasma is ignited and reaches a level
at 33 characteristic of the Si etching. The signal then increases
again, a-t 34, as parts of the SiO2 substrate are exposed, and
finally the endpoint of the etching is marked by abrupt levelling
off of the time dependence of the signal at 35. Similar curves
for successive runs are found to be very reproducible in form and
amplitude. Also similar curves are observed for larger wafer
batches; the relative change in signal at the endpoint being
much greater, and the abrupt change in slope at the endpoint
being smeared out slightly due to nonuniformity in poly thickness
and etchrate across the batch. The etchrate decreases with
increasing wafer load and this can be compensated for by an
increase in power with a corresponding increase in the signal
amplitudes from the endpoint monitor. The peak 32 at the onset
of silicon etching generally occurs only when etching is preceded
by a short 2 plasma treatment used to improve etching uniformity
by removal of photoresist scum, the peak being attributed to the
etching of the thin SiO2 layer formed during the plasma oxidation
pre-clean.
Points 35 and 36 in Figure 3 correspond to the
termination points for two patterned wafers etched using the
method of endpoint control of the present invention. The wafer
1~'7~579
on which etching was terminated at 35 was clear of Poly-Si in all
exposed areas with negligible undercut. The wafer terminated
at 36 showed sufficiently little undercut to allow the etching
of good 2 ~m featuresO This shows that the endpoint is easily
recognized with a window, during which etching should be
terminated, that is wide enough to accommodate typical variations
in poly-silicon thickness or etchrate nonuniformity within a
batch load.
The 0.70 emission line may also be used for
monitoring of the plasma patterning of other structures, for
example Si3N4 on SiO2 on Si, typically used in silicon integrated
circuit manufacture. This is accomplished by using a dummy
wafer, as illustrated in Figures 4(a) and 4(b), Figure 4(a)
relating to Si3N4 on SiO2 and Figure 4(b) relating to SiO2 on Si.
In Figure 4(a), a wafer 40 comprises a silicon
substrate 41 on which is a layer of SiO2 42 on one surface and
a back protection layer 43 on the other surface. A layer of
Si3N4 44 is formed on the SiO2 layer 42 and a masking layer 45
is formed on the layer 44, defining an area 46 which is to be
etched. ProYided with wafer 40 is a dummy wafer 47. Wafer 47
has the silicon substrate 41, and back protection layer 43 but
the SiO2 layer 42 is omitted, the Si3N4 layer 44 being directly
on the substrate. The masking layer 45 is formed on layer 44,
with the area 46 defined. The Si3N4 layers 44 on each wafer is
the same thickness. One or more wafers 40 are loaded into the
reaction chamber with a dummy wafer 47. Etching is terminated
just after the optical emission intensity decrease associated
with penetration of etching through the Si3N4 layer 44 into the
silicon substrate 41 of the dummy wafer 40 occurs.
A similar approach is used for SiO2 over Si. As
illustrated in Figure 4(b) a wafer 50 has a silicon substrate
~L~7~579
51 having a SiO2 layer 52 on one surfac~ and a back protection
layer 53 on the other surface. A masking layer 54 is formed on
the SiO2 layer 52, defining an area 55 to be etched. A dummy
wafer 56 is also provided having the same formation as wafer 50
that is silicon substrate 51, SiO2 layer 52, back protection layer
53 and masking layer 54 defining area 55. The layers 52 are
of the same thickness but in this example the SiO2 layer 52 of
the dummy wafer 56 is given a slight pre etch - for example ~ 500
- 1000 ~ - prior to use as indicated at 57. The wafer, or wafers,
50 are loaded into the reaction chamber with a dummy wafer 56.
Etching is terminated when the Si etching emission signal level,
associated with the dummy wafer, is detected. Then a thin layer
of SiO2 remains over the substrate 51 of the wafer, or wafers, 50
and the etching to the Si/SiO2 interface can be completed with a
chemical dip etch that does not degrade pattern resolution, if
required. Silicon plasma etches so much more quickly than SiO2
that the plasma etching cannot be terminated at the Si/SiO2
interface with CF4+5%02 etchgas without unacceptable damage to
areas of the underlying silicon due to nonuniformity of etching
and of layer thickness.
To provide for nonuniformities of etch rate and
of layer thicknesses when using dummy wafers, 5i3N4 etching
should continue slightly past the endpoint signal, and the SiO2
etchback of SiO2 over Si dummy wafer must be larger than the
nonuniformities. For both examples, the dummy wafers must have
the same layer and masking characteristics as the batch wafers
and the wafers should be equally spaced in the reactor. The
back protection layers 43 and 53 are provided to avoid spurious
Si signals from the backs of the wafers.
Some major advantages of the new endpoint monitor
for marking the onset (or completion) of silicon layer etching,
1~17~57~
include:
1) The endpoint indicator is clear and precise thereby
requiring a minimum of operator judgement or skill. Moveover
the signal is reproducible and changes rapidly near the endpoint
thereby being suitable for use with a comparator circuit set at
say 80% of the S102 signal to trigger a time delayed termination
of the etchcycle. The setpoint can easily be calibrated with a
dummy load.
2) The method automatically compensates for changes in wafer
number, spacing, area exposed and layer thickness from batch to
batch.
3) The method also automatically compensates for batch to batch
variations in starting temperature, and plasma conditions. Thus,
it allows a relaxation in the required accuracy of setting of all
the plasma parameters.
~) A consequence of (3) is that the need for temperature
control and preheating is eliminated, and short, low temperature,
high RF power etchcycles with faster throughPut become feasible,
without compromising endpoint control accuracy. By allowing
fast etching at low temperature and high RF power, the invention
permits maintenance of good etchrate uniformity with minimal
photoresist erosion. This enables finer line patterning over a
more highly stepped layer than is normally possible.
Optical emission endpoint control also has
application in photoresist stripping, as illustrated in Figure 5.
In Figure 5, curve 60 corresponds to one 3" diameter wafer while
curve 61 corresponds to four 3" diameter wafers. The coordinates
of the curves are intensity of the emission line and stripping
time respectively. The stripping time is also related to plasma
temperature, as shown. The particular curves 60 and 61 are for
an emission line at a wavelength of about 0.30 ~m for stripping of
~7~57~
photoresist in an 0.5 Torr 2 plasma at an RF power of 300 watts
inside an etch tunnel. After an initial rise at 62 the signal
peaks slightly at 63, decreases, and then increases with increasing
temperature and stripping time to a maximum at 64 when parts of
the photoresist area are stripped away. The signal then
decreases until all the photoresist is stripped away at 65, and
then levels off. The endpoint is clearly visible even for partial
loads. The signal amplitude increases with increasing photoresist
area, reactor temperature and RF power, but decreases with
increasing 2 pressure. The emission line increases with
increasing stripping, suggesting that it is associated with a
reaction product.
The emission line at about 0.30 ~m is particularly
sensitive to photoresist stripping. ~lowever there are groups
of lines between about 0.250 and about 0.340 ~m and between
about 0.450 and about 0.485 ~m, and a single line at about 0.650
~m that have similar sensitivity.
The emission occurs from a species present
in the plasma, which species is associated with a change in
material being etched. Thus a signal produced by a change in
intensity of the emission can be used to indicate either the end
point of etching of a particular material, or of the end point
of etching some other material over a particular material - that
is indicate the beginning of etching of that particular material.
While the emission line at about 0.70 ~m
wavelength has been found to be particularly effective in the
plasma etching of Si on SiO2, or vice versa, it is appreciated
that with other materials other lines can prove effective.
Similarly, while CF4~5%02 has been described
as the etchgas in etching Si, the oxygen content can be varied
from zero as is well known, and CF4+5%02 has been chosen as a
g
~7~57~
convenient etchgas.
In Figures 1 and 2, what is known as a volume
loading apparatus is described. The invention can also be
used with a surface loading form of apparatus.
The process is suitable for the photoresists
commonly used in integrated circuit fabrication. For example,
typical photoresists are the Waycoat HR series supplied by
Philip A. Hunt Co.(Canada) Ltd., and the Shipley 1350 series
supplied by the Shipley Company Incorporated.
A key feature is the selection of the most useful
emission lines for greatest sensitivity. There can be various
emission lines showing useful characteristics and the method of
selection and detection can be varied.
-- 1 0 --
