Controlling Oxygen Precipitation In III Wafers IV Wafers

These
include such wafer pre-treatments as the socalled “Hi-Low-High”
treatments in which oxygen is first out-diffused at
high temperatures (to create a low oxygen content surface
layer) followed by a generally long, low temperature
treatment to (re-) nucleate oxygen clusters. This is
followed by another high temperature treatment to grow
them into precipitates. Other approaches have included
attempts to very narrowly specify oxygen concentration
and crystal growth process or even the segments of the
crystal from which wafers should be taken for specific
application. All of this tailoring of silicon material
to specific applications has lead to a vast amount of
complication and rigidity in the silicon industry and
quite a lot of confusion over whether silicon is an
engineered material or a commodity.

The
MDZ wafer is a serious departure from previous attempts
at solving the problem of oxygen precipitation control.
The RTP family of processes that produce MDZ wafers
result in a radical change of the wafer’s material
properties. At its core is the enormous effect that
vacancies have on the control of the nucleation processes
of oxygen in silicon. The process which produces the
MDZ wafer installs a useful vacancy concentration profile
(or template) into a silicon wafer which subsequently
takes over the control of the wafer’s oxygen precipitation
behavior from all of the difficult factors important
in conventional silicon: crystal growth, IC application,
and even oxygen concentration itself. An MDZ wafer is
a wafer which is programmed through the RTP treatment
to behave in a well defined, ideal manner in any application,
sweeping aside an entire raft of technological difficulties.

An
illustration of the huge effect that vacancies have
on oxygen precipitation behavior
is shown in Fig. 1 in which the dependence of the resulting
oxygen precipitate density on vacancy concentration
is shown.

Fig.
2 A schematic illustration of the difference between
conventional methods of installing denuded zones (DZ)
in silicon wafers via oxygen out-diffusion and renucleation
and a new method based on the installation of tailored
vacancy concentration profiles

The
core concept of the MDZ wafer is to utilise this very
strong dependence and engineer a profile of vacancies
into a silicon wafer. The very steep, switch-like dependence
of precipitate density on vacancy concentration means
that a profile of vacancy concentration rising from
the surface and going through the threshold value produces
a rather sharply layered structure with a highly precipitating
bulk underneath a non-precipitating surface layer. The
threshold for this layered design lies at a vacancy
concentration of about 1012 cm-3. Figure 2 schematically
illustrates the design of such a wafer and compares
it to a conventional oxygen out-diffusion approach to
the problem of forming a denuded zone.

Vacancy
concentration profiles

Vacancies
may be introduced into silicon wafers at high temperatures
by a number of different mechanisms. Two examples are
nitridation or through simple high temperature Frenkel
pair generation. Generating large concentrations of
vacancies in wafers at high temperatures is not a difficult
task at all. The problem lies in fighting the tendency
of the wafer to return to equilibrium during the cooling
of the wafer and keeping them in the wafer in sufficiently
large concentrations to be useful. This is where RTP
comes in.

The
simplest procedure for installing a useful profile of
vacancies in a silicon wafer relies solely on Frenkel
pair generation and the close proximity (relative to
vacancy diffusion lengths) of the two wafer surfaces.
Heating a thin wafer to a high temperature T results
in the rapid equilibration of the vacancy-interstitial
system.

First,
Frenkel pairs – vacancies and self-interstitials
in equal amounts - are produced. This - very fast –
reaction leads to a recombination generation equilibrium.
The product CiCv of the two concentrations acquires
the equilibrium product value Ci*Cv*, with Ci = Cv =
(Ci*Cv*)1/2 at this stage. Were the sample to be cooled
at this point under the condition of equal concentration,
the vacancies and interstitials would merely mutually
annihilate each other completely in the reverse process
of their generation resulting in no vacancy concentration
enhancement by the time the samples reach room temperature.

This
is averted by the next stage of the process in which
the two point defect concentrations separately equilibrate
to their respective equilibrium values via a process
which takes advantage of the thinness of silicon wafers.
Useful vacancy incorporation is made possible by the
fact that both Ci and Cv approach their (different)
equilibrium values, Ci*and Cv*, due to an exchange with
the wafer surfaces (considered as ideal sink/source
of point defects). This coupled process is controlled
mainly by diffusion of self-interstitials which are
the faster diffusers – the two concentrations being
comparable. The total time to achieve this complete
equilibrium in a standard wafer (ca. 700 m thick) was
found to be extremely short, less than several seconds
at 1250oC. In other words, wafers are thin with respect
to interstitial diffusivity on an RTP time scale. The
speed of the equilibration is, in fact, a measure of
interstitial diffusivity. This also implies that the
interstitial diffusivity is high, on the order of 2.5x10-4cm2/s.

After
this equilibration, the vacancies become the dominant
species since Cv*>Ci*. On subsequent cooling the
point defects quickly recombine, only now, some of the
vacancies survive. During the cooling of the equilibrated
vacancy and interstitial solutions, the majority vacancies
consume the minority interstitials until there are effectively
no more to left consume leaving behind an excess concentration
of surviving vacancies equal to the initial concentration
difference at the anneal temperature, .C= Cv*-Ci*.

The
proximity of the wafer surfaces during cooling adds
the final component to the process. During cooling,
in addition to recombination, vacancies will out diffuse
to the wafer surfaces where the local (equilibrium)
vacancy concentration is rapidly decreasing. However,
if the cooling rate is fast enough – in the range
of about 40 to 100K/s (readily accessible through using
RTP) - the middle of the wafer will be not affected
by vacancy out-diffusion, and the vacancy species will
be present there in the concentration .C.

The
regions near the surfaces will be strongly affected.
At lower temperatures the profile is effectively frozen-in
by the binding of vacancies to oxygen which becomes
complete by about 900oC [17]. At this point the vacancies
convert from their relatively mobile state (free mono-vacancies)
to their relatively immobile state, O2V. The profiling
of the vacancy concentration through out-diffusion achieved
in a given RTP heat treatment is largely the result
of the cooling conditions of the wafer above this temperature.

In
the present discussion, the most important thing is
that the near-surface regions of a quenched wafer are
depleted of vacancies, by vacancy out-diffusion during
the cooling stage of RTP. In the near-surface zones
the vacancy concentration is below Cv*, and oxygen precipitation
is suppressed in a practical sense as a result of prohibitively
long incubation times (the tabula rasa effect). The
rapid increase in temperature during the ramp-up to
the process temperature serves to dissolve all pre-existing
oxygen clusters. In the middle of a wafer the precipitation
is strong, due to the presence of vacancies in concentration
over Cv*. Such a precipitation profile is precisely
what is required for ideal IG. The width of DZ (precipitation-denuded
zone) is easily controlled by the cooling rate.

Faster
cooling rates mean a shorter vacancy diffusion length,
and thus a narrower DZ. If the cooling rate is too slow,
such as occurs in conventional furnace annealing, the
high temperature vacancy concentration profile will
be allowed to fully relax throughout the sample thickness
to its equilibrium value near the binding temperature,
which is well below the threshold for precipitation
enhancement. At this point the entire wafer thickness
becomes of the tabula rasa type.

Precipitation control

Installing
a vacancy concentration profile which rises from the
wafer surface into the bulk of the wafer crossing the
critical concentration Cv* at some desired depth is
the core of the concept behind the MDZ wafer. The installed
vacancies have full control of the oxygen precipitation
behavior of the wafer. An example of a depth distribution
of oxygen precipitates produced by vacancy concentration
control is shown in Fig 3.

Fig.
3. (a) Depth profiles of platinum diffusion profiles
(~ vacancy concentration) measured at 730 and 800°C,
calculated vacancy concentrations, and measured oxygen
precipitate densities in an RTP treated sample processed
at 1250°C. The oxygen precipitate density axis is
scaled to correspond to the vacancy-precipitate density
calibration and (b) an etched cross section of a silicon
wafer containing such a profile following a precipitation
heat treatment (800°C 4 hours + 1000°C 16 hours)
showing the depth profile of oxygen precipitates resulting
from an RTP-installed vacancy concentration profile.
The bulk density of precipitates in this example is
1 x 1010 cm-3

The
second is the cooling rate controlling the depth of
the DZ and, at slower rates, eventually the vacancy
concentration (and with it, resulting precipitate density)
at the center of the wafer. Fig 4 illustrates the cooling
rate effect. The minimum Tp for an effective straightforward
RTP treatment corresponds to Cv*(Tp) – Ci*(Tp)
/1012 cm-3, the threshold value. This happens around
1150°C. The quenched-in vacancy concentration (and
with it bulk precipitate density) rises with increasing
Tp. The bulk precipitate density reaches about 1010
cm-3 by about 1250°C. In RTP treatments, the concentrations
of intrinsic point defects in silicon wafers can be
dynamically and very rapidly altered means other than
temperature and cooling rate. Ambient is also very important
in these technologies. Examples include interstitial
injection from the wafer surfaces during oxidation,
and vacancy enhancements due to nitridation effects.
One example is the complete elimination of the effect
when oxygen ambient concentrations exceed about 2000
ppma. Another example is the creation of an inverse
vacancy profile in pure nitrogen or ammonia ambients.
An example of this is shown in Fig 5. There are many
advantages to using a RTP-vacancy based system to control
oxygen precipitation behavior in silicon wafers: simplicity,
ability to copy-exact, ease-ofuse and reliability of
results. Figure 6 shows just

Figure
5. An example of an ambient effect on quenched in vacancy
concentration and subsequent precipitation behavior.
The circles in a) are a vacancy concentration profile
in an argon ambient. b) is an etched cross section of
the resulting precipitate profile. The squares in a)
are the result of the same treatment performed in nitrogen.
c) illustrates the resulting precipitate depth profile