Implantation through a screen oxide is known to knock oxygen atoms into
the silicon crystal. These oxygen atoms (recoils) build a layer in the
crystalline silicon which is rich in oxygen within a few lattice distances.
We use this knock-in effect to produce ultra shallow profiles in the
sub 
Å range [Wim91a], [Orl91b], [Sub91].
Figure 2.5-4 shows the basic scheme for the knock-in
implantation. A layer of a material (target 
) which is rich in some
dopant species 
, e.g. borosilicate glass (BSG) with 6% Boron, is
deposited on the substrate, e.g. silicon. The ion bombardment 
with a
projected range preferably within the deposited layer kicks out some of
the dopants (recoils). Such recoils can be knocked into the substrate
directly (Figure 2.5-4 left) or can kick out other dopants
(Figure 2.5-4 right). The knocked-in species remain within a
thin layer (
Å) at the surface of the substrate. The
concentrations achieved can be a few 
of dopant for
implantation dose of bombarding species
(Figure 2.5-5).
The simulations were performed with the Monte Carlo module of
PROMIS. The Monte Carlo ion implantation module is able to
calculate implantations of any bombarding species into any target
which is assumed to be amorphous and may contain species 
, 
, ...
The trajectories of any recoil can be followed. One bombarding ion is likely
to produce several dozen recoils.
We have simulated implantations through glass targets containing a certain
percentage of boron, phosphorus and germanium with silicon and germanium as
bombarding species. For the practical application we need to know how deep
the knocked-in profiles for different recoil species and different
bombarding species are, and how we can control the peak concentration and
the depth of the knocked-in profiles.
The knocked-in amount of dopant increases linearly with the percentage
of in the layer and also linearly with the dose of the bombarding
species . We have implanted different doses of silicon into a layer of
BSG/PSG containing some percentage of boron/phosphorus into a silicon
substrate. The dopant concentration at the glass/
interface in
Figure 2.5-6 is divided by the implanted silicon dose.
Concentrations as high as several 
can be gained with
reasonable silicon doses of .
Figure 2.5-7 shows the dependence of the interface
concentration on the implantation energy using as bombarding species
for knocked-in boron, phosphorus and germanium profiles. The concentrations
are normalized by the implanted silicon dose (
.
The knocked-in profiles are very shallow in the silicon substrate. The
profiles decrease by two orders of magnitude within a few nanometer from the
glass/ interface (depth = 
), and decrease more moderate
deeper into the substrate. To explain this behavior we would have to analyze
energy and angle distribution of the knocked-in ions at the interface.
A typical knocked-in boron profile is a few nanometers shallower than a knocked-in phosphorus profile. Using a target containing boron and phosphorus, e.g., BPSG, in a practicable ratio differential knock-in technique can be applied to produce ultra shallow n-p-n junctions (Figure 2.5-8 left) by just one implantation step.
Another application consists of combining the differential knock-in technique
with band-gap engineering [Sub91]. While forming the deposited target
we might also add germanium together with other dopants. During the
implantation process germanium will be knocked-in along with the dopants
into silicon. During the subsequent anneal (RTA) the germanium is
incorporated into the silicon lattice forming a 
lattice
affecting the transistor performance by the altered band-gap structure.
The knock-in technique might be an alternative to the expensive molecular
beam epitaxy (MBE) technique.
For effectively influencing the band gap structure we require germanium
concentrations of the same order of magnitude as the silicon atomic density
. Typical knocked-in germanium concentrations are below
, therefore, the knock-in method for band gap engineering
is limited to non-glass target layers.
However, the application of glasses as target materials is questionable,
since a huge amount of oxygen is knocked into the crystalline silicon
substrate (see Figure 2.5-9). For a silicon dose of
the simulation shows that the oxygen concentration in the
silicon target is as high as 
within the top few
nanometers and for several 
.
Knocked-in profiles are very shallow and can be controlled in the
concentration. Surface concentrations of several can be
obtained with reasonable implantation doses. The depth of the profiles is
typically in the Å range and cannot be influenced easily.
Silicate glasses are not applicable as target material, because of the
inacceptable high oxygen concentration.
For the application as knock-in source for ultra shallow profiles other target materials have to be found, which are easily removable from the crystalline silicon. Applications of knock-in techniques where the target material need not be removed for further processing are possible. For instance to improve the adhesive strength and the electrical properties of contacts.
