U.S. patent application number 10/809070 was filed with the patent office on 2004-09-30 for method and device for the production of a silicon single crystal, silicon single crystal, and silicon semiconductor wafers with determined defect distributions.
This patent application is currently assigned to Siltronic AG. Invention is credited to Ammon, Wilfried Von, Schmidt, Herbert, Virbulis, Janis, Weber, Martin, Wetzel, Thomas.
Application Number | 20040192015 10/809070 |
Document ID | / |
Family ID | 32991939 |
Filed Date | 2004-09-30 |
United States Patent
Application |
20040192015 |
Kind Code |
A1 |
Ammon, Wilfried Von ; et
al. |
September 30, 2004 |
Method and device for the production of a silicon single crystal,
silicon single crystal, and silicon semiconductor wafers with
determined defect distributions
Abstract
A method for the production of a silicon single crystal by
pulling the single crystal, according to the Czochralski method,
from a melt which is held in a rotating crucible, the single
crystal growing at a growth front, heat being deliberately supplied
to the center of the growth front by a heat flux directed at the
growth front. The method produces a silicon single crystal with an
oxygen content of from 4*10.sup.17 cm.sup.-3 to 7.2*10.sup.17
cm.sup.-3 and a radial concentration change for boron or phosphorus
of less than 5%, which has no agglomerated self-point defects.
Semiconductor wafers are separated from the single crystal. These
semiconductor wafers have may have agglomerated vacancy defects
(COPs) as the only self-point defect type or may have certain other
defect distributions.
Inventors: |
Ammon, Wilfried Von;
(Hochburg/Ach, AT) ; Virbulis, Janis; (Babites
pag., LV) ; Weber, Martin; (Kastl, DE) ;
Wetzel, Thomas; (Haiming, DE) ; Schmidt, Herbert;
(Halsbach, DE) |
Correspondence
Address: |
WILLIAM COLLARD
COLLARD & ROE, P.C.
1077 NORTHERN BOULEVARD
ROSLYN
NY
11576
US
|
Assignee: |
Siltronic AG
|
Family ID: |
32991939 |
Appl. No.: |
10/809070 |
Filed: |
March 25, 2004 |
Current U.S.
Class: |
438/502 ;
438/478 |
Current CPC
Class: |
C30B 29/06 20130101;
C30B 15/14 20130101; C30B 15/203 20130101; C30B 15/206 20130101;
Y10T 117/1068 20150115 |
Class at
Publication: |
438/502 ;
438/478 |
International
Class: |
H01L 021/30; H01L
021/46; C30B 001/00; H01L 021/20; H01L 021/36 |
Foreign Application Data
Date |
Code |
Application Number |
Mar 27, 2003 |
DE |
103 13 940.0 |
Aug 28, 2003 |
DE |
103 39 792.2 |
Claims
What is claimed is:
1. A method for the production of a silicon single crystal
comprising the steps of: pulling the single crystal, according to
the Czochralski method, from a melt which is held in a rotating
crucible, the single crystal growing at a growth front; and
supplying heat deliberately to the center of the growth front by a
heat flux directed at the growth front.
2. The method as claimed in claim 1, wherein a curvature of the
growth front is reduced or increased.
3. The method as claimed in claim 1, wherein an axial temperature
gradient G.RTM.) at the growth front is regulated, r extending from
0 as far as a radius of the growing single crystal.
4. The method as claimed in claim 1, wherein a temperature
distribution, in which a radial variation of a temperature gradient
Gs.RTM.) in the melt is less than 15%, is produced in a region with
an extent of up to 5 cm below the growth front and at least 90% of
a diameter of the single crystal.
5. The method as claimed in claim 1, wherein the heat flux is
produced by a heat source, which deliberately increases the
temperature at a center of a bottom of the crucible compared with
the temperature at an edge of the bottom of the crucible.
6. The method as claimed in claim 5, wherein a bottom heater is
arranged below the crucible, and thermal insulation is used to
ensure that the bottom heater heats the center of the bottom of the
crucible more strongly than the edge of the bottom of the
crucible.
7. The method as claimed in claim 5, wherein the heat source is
arranged at the center of the bottom of the crucible.
8. The method as claimed in claim 5, wherein the temperature of the
crucible at the center of the bottom of the crucible is increased
by at least 2 K relative to the temperature at the edge of the
bottom of the crucible.
9. The method as claimed in claim 1, wherein a heat source is
arranged below the growth front in the melt.
10. The method as claimed in claim 1, wherein the heat flux is
produced by iso-rotation of the single crystal and the crucible,
the crucible being rotated with at least 10% of a rotational speed
of the single crystal.
11. The method as claimed in claim 10, wherein the melt is exposed
to a CUSP magnetic field.
12. The method as claimed in claim 10, wherein the melt is exposed
to a traveling magnetic field.
13. The method as claimed in claim 1, wherein the heat flux is
produced by an electromagnetic field to which the melt is exposed,
at least 10% of a wall area of the crucible being shielded against
an effect of the electromagnetic field on the melt.
14. The method as claimed in claim 13, wherein the heat flux is
produced by a traveling magnetic field.
15. The method as claimed in claim 14, wherein a rotational
symmetry of the electromagnetic field is broken by a partial
shielding of the field.
16. The method as claimed in claim 1, wherein the heat flux is
produced by applying a positive electrical voltage of more than 100
volts to the crucible.
17. The method as claimed in claim 1, wherein additional heat is
supplied to a phase boundary of the single crystal, to the
atmosphere surrounding the phase boundary and to the melt.
18. The method as claimed in claim 1, wherein the growing single
crystal is cooled by a cooling device.
19. The method as claimed in claim 1, wherein a fluctuation of a
pull rate when pulling a silicon single crystal with a diameter of
at least 200 mm, with a pull rate at which neither defects due to
agglomerated vacancies nor defects due to agglomerated interstitial
atoms are created, is at least .+-.0.02 mm/min while the single
crystal is being pulled over a length of at least 30 mm.
20. A silicon single crystal with an oxygen content of from
4*10.sup.17 cm.sup.-3 to 7.2*10.sup.17 cm.sup.-3 and a radial
concentration change for boron or phosphorus of less than 5%, which
has no agglomerated self-point defects.
21. The single crystal as claimed in claim 20, which is doped with
nitrogen and/or carbon.
22. The single crystal as claimed in claim 20, with a radial oxygen
concentration variation (ROV) of at most 5%.
23. Semiconductor wafers separated from a single crystal as claimed
in claim 20.
24. Silicon semiconductor wafers with agglomerated vacancy defects
(COPs) as an only self-point defect type, said defects having a
variation in their average diameter of less than 10% and being
present on a circular surface of the semiconductor wafers, the
diameter of the circular surface being at least 90% of the diameter
of the semiconductor wafers.
25. Silicon semiconductor wafers with agglomerated vacancy defects
(COPs) as a defect type, these defects being covered with an oxide
layer whose thickness is less than 1 nm.
26. The semiconductor wafers as claimed in claim 25, wherein said
defects have an average diameter of less than 50 nm.
27. Silicon semiconductor wafers which are free of agglomerated
self-point defects and have two or more mutually separated axially
symmetric regions, in which unagglomerated vacancies dominate as
the defect type.
28. Silicon semiconductor wafers which are free of agglomerated
self-point defects and have two or more mutually separated axially
symmetric regions, in which unagglomerated interstitial silicon
atoms dominate as the defect type.
29. Silicon semiconductor wafers with agglomerated interstitial
atoms (LPITs) as a defect type, wherein said agglomerated
interstitial atoms are so small that no secondary dislocations are
also present.
30. Silicon semiconductor wafers having at least one region with
agglomerated vacancy defects (COPs) as a defect type, said defects
being covered with an oxide layer whose thickness is less than 1
nm, and at least one region with agglomerated interstitial atoms
(LPITs) as the defect type, wherein said agglomerated interstitial
atoms are so small that no secondary dislocations are also
present.
31. Semiconductor wafers as claimed in claim 30, the agglomerated
vacancy defects having an average diameter of less than 50 nm.
32. A device for the production of a single crystal according to
the Czochralski method, comprising: a crucible which contains a
melt; a heating device surrounding the crucible; a magnetic
instrument surrounding the crucible and producing a static or
dynamic magnetic field; a heat source arranged above the melt and
supplying heat to the phase boundary of the single crystal, to the
gas phase and to the melt; a cooling device surrounding the single
crystal; a heat shield surrounding the single crystal; and a
control unit which causes iso-rotation of the single crystal and
the crucible.
Description
BACKGROUND OF THE INVENTION 1. Field of the Invention
[0001] The invention relates to a method for the production of a
silicon single crystal by pulling the single crystal, according to
the Czochralski method, from a melt which is held in a rotating
crucible, with the single crystal growing at a growth front. The
invention also relates to a silicon single crystal and to
semiconductor wafers which are separated therefrom. 2. The Prior
Art
[0002] The production of single crystals which have a diameter of
200 mm or more represents a significant challenge, particularly
since it is very difficult to deliberately adjust the radial
crystal properties within a very narrow tolerance range. This
applies to the concentration of impurities or dopants, and
especially to the crystal defects and self-point defects, and
agglomerates thereof. Self-point defects include interstitial
silicon atoms (silicon self-interstitials) and vacancies, which are
formed at the growth front of the single crystal. They very
substantially determine the radial and axial defect distributions
occurring later in the single crystal, and they also affect the
impurity distributions which occur. For example, vacancies
contribute to the precipitation of oxygen. When they exceed a size
of about 70 nm, oxygen precipitates form oxygen-induced stacking
faults (OSFs). The vacancies themselves can combine into
agglomerates and form so-called COPs (crystal originated
particles). Agglomerates of interstitial atoms form local crystal
dislocations, which are also referred to as LPITs (large etch pits)
because of the detection method which is used. The oxygen
concentrations and the thermal conditions at the growth front and
in the solidifying single crystal determine the nature and
distribution of the crystal defects and impurities.
[0003] The thermal conditions when pulling the single crystal. are
dictated by the heat sources, i.e. the heating elements which are
used, and the heat of crystallization released during
solidification. The heat energy is transferred to the single
crystal by radiation, heat conduction and heat transport, for
example via the flows in the melt. The removal of heat in the
vicinity of the growth front is determined to a large extent by the
heat radiated from the edge of the single crystal and by the
thermal dissipation in the single crystal. Overall, the thermal
budget can be affected by the design of the pulling system, i.e.
via the geometrical arrangement of the thermally conductive parts,
the heat shields and by additional heat sources. The process
conditions, for example growth rate, pressure, quantity, type and
flow of inert gases through the pulling system furthermore
contribute substantially to the thermal balance. Increasing the
pressure or the quantity of inert gases, for example, will reduce
the temperature. Faster pull rates produce more heat of
crystallization.
[0004] The flows which transport heat in the melt are extremely
difficult to predict. The heating elements, generally arranged in a
ring around the crucible, produce a convective flow in the melt.
Together with the counter-rotation conventionally used for the
single crystal and crucible, this leads to pattern of movement in
the melt which is distinguished by an upward melt flow at the edge
of the crucible and a downward melt flow below the growing single
crystal.
[0005] As experiments have shown, the movement of the melt also
depends on the degree and direction of the rotation of the crucible
and the single crystal. For example, iso-rotation and
counter-rotation produce very different convection patterns.
Crystal pulling with iso-rotation has already been studied
(Zulehner/Huber in Crystals 8, Springer Verlag, Berlin Heidelberg
1982, pp 44-46). Counter-rotation is generally preferred because,
compared to iso-rotation, it leads to a less oxygen-rich material
and significantly more stable conditions for the crystal growth.
The iso-rotation variant is not generally used on an industrial
scale.
[0006] The flows which transport heat and oxygen in the melt can
also be affected by the forces due to applied electromagnetic
fields. Static or dynamic fields make it possible to alter the
degree and direction of the flows in the melt, so that different
oxygen contents can be obtained. They are primarily used for oxygen
control. Magnetic fields are used in a number of variants, for
example in the form of static magnetic fields (horizontal, vertical
and CUSP magnetic fields), single-phase or polyphase alternating
fields, rotating magnetic fields and traveling magnetic fields.
According to U.S. patent application Ser. No. 2002/0092461 A1, for
example, a traveling magnetic field is used in order to control the
incorporation of oxygen into the single crystal. Recent numerical
simulations for the effect of magnetic fields on the movement of
the melt are presented, for example, in "Numerical investigation of
silicon melt flow in large diameter CZ-crystal growth under the
influence of steady and dynamic magnetic fields", Journal of
Crystal Growth, 230 (2001) 92-99.
[0007] The radial temperature distribution at the growth front of
the crystal is extremely important for the crystal properties. It
is determined essentially by the heat radiated from the edge of the
single crystal. As a rule, a much more pronounced temperature drop
is therefore observed at the edge of the single crystal than at its
center. The axial temperature drop is usually denoted by G (axial
temperature gradient). Its radial variation G.RTM.) very
substantially determines the self-point defect distribution, and
therefore the other crystal properties as well. The radial change
of the temperature gradient G due to the thermal budget is
generally determined by numerical simulation calculations. The
radial variation of the temperature gradient can be experimentally
determined from the behavior of the radial crystal defect
distribution for different growth rates.
[0008] The ratio V/G.RTM.) is of crucial importance in terms of the
creation of crystal defects, G.RTM.) being the axial temperature
gradient at the growth front of the single crystal and depending on
the radial position (the radius r) in the single crystal, and V
being the rate at which the single crystal is pulled from the melt.
If the ratio V/G is more than a critical value k1, then vacancy
defects (vacancies) predominantly occur; these can agglomerate and
then be identified, for example, as COPs (crystal originated
particles). Depending on the detection method, they are sometimes
referred to as LPDs (light point defects) or LLSs (localized light
scatterers). Because of the usually decreasing radial profile of
V/G, the largest COPs most commonly occur at the center of the
crystal. They generally have a diameter of about 100 nm, and
therefore cause problems for component fabrication. The size and
number of the COPs is determined by the initial concentrations of
the vacancies, the cooling rate and the presence of impurities
during agglomeration. For example, the presence of nitrogen leads
to a shift of the size distribution toward smaller COPs with a
larger defect density.
[0009] If the ratio V/G is lower than a critical value k2, which is
less than k1, then self-point defects are predominantly found in
the form of interstitial atoms (silicon self-interstitials), which
can also produce agglomerates and are microscopically seen as
dislocation loops. These are often referred to as A swirls, and the
smaller form as B swirls, or as LPIT defects (large etch pits) for
short because of their appearance. The size of LPITs lies in the
range up to 10 .mu.m. As a rule, not even epitaxial layers can
cover up these defects perfectly. These defects as well can also
impair the functionality of the electronic components fabricated on
silicon wafers.
[0010] In the broadest sense, the region in which neither
agglomeration of vacancies nor agglomeration of interstitial atoms
takes place, i.e. in which V/G lies between k1 and k2, is referred
to as a neutral zone or perfect region. The value of V/G at which
the crystal changes from excess vacancies to excess interstitials
naturally lies between k1 and k2, and is given in the literature as
the critical limit C.sub.crit=1.3*10.sup.-3 cm.sup.2 min.sup.31 1
K.sup.31 1 (Ammon, Journal of Crystal Growth, 151, 1995, 273-277).
In a more specific sense, however, distinction is also made between
a region in which there are still free unagglomerated vacancies and
a particular region of free interstitial atoms. The vacancy region,
also referred to as the v region (vacancies), is distinguished in
that if the oxygen content of the single crystal is high enough,
oxidation-induced stacking faults are created there, while the I
region (interstitials) remains fully fault-free. In this more
specific sense, therefore, only the I region is actually a perfect
crystal region.
[0011] Large ingrown oxygen precipitates with a diameter of more
than about 70 nm can be revealed as oxygen-induced stacking faults
(OSFs). To this end, the semiconductor wafers cut from the single
crystal are subjected to a special heat treatment, which is
referred to as wet oxidation. The growth rate of the oxygen
precipitates created during the crystal. pulling, which are
sometimes also referred to as grown BMDs (bulk micro-defects), is
promoted by vacancies in the silicon lattice. OSFs are therefore
encountered primarily in the v region.
[0012] The single crystal would be virtually defect-free if the
pulling conditions can be adjusted so that the radial profile of
the defect function V/G.RTM.) lies within the critical limits for
COP or LPIT formation. This is not easy to achieve, however,
especially when single crystals with a comparatively large diameter
are being pulled, because the value of G then depends significantly
on the radial position r. In general, owing to the radiative heat
losses, the temperature gradient G is very much greater at the edge
of the single crystal than at the center.
[0013] The radial profile of the defect function V/G.RTM.), or of
the temperature gradient G.RTM.), can lead to there being several
defect regions on a semiconductor wafer cut from a single crystal.
COPS preferentially occur at the center. The size distribution of
the agglomerated vacancies is dictated by the cooling rate of the
single crystal in the vicinity of the growth front. The size
distribution of the COPs can be altered from a few large COPs to
many small, less perturbing COPs by a high cooling rate (more than
2 K/min), or short dwell times in the temperature range from the
melting point to about 1100.degree. C., or by doping the melt with
nitrogen. Furthermore, a radial size distribution such that smaller
defects are formed with increasing radius is found in the COP
region. The COP region is followed by an oxygen-induced stacking
fault ring (OSF), due to the interaction of vacancies and oxygen
precipitates. Outside this is a fully defect-free region, which is
in turn bounded by a region with crystal defects consisting of
interstitial agglomerates (LPITs). At the edge of the single
crystal, the interstitial atoms diffuse as a function of the
thermal conditions, so that a centimeter-wide defect-free ring may
also be created there.
[0014] The crystal defect regions that occur have already been
discussed at length, in relation to the radial V/G profile, by
Eidenzon/Puzanov in Inorganic Materials, vol. 33, No. 3, 1997, pp
219-255. This article has already indicated possible ways of
producing defect-free material. Both the cooling rate in
temperature range during agglomeration, the effect of nitrogen
doping and methods such as oscillating growth rate are referred to
in this context.
[0015] To a certain extent, radial homogenization of V/G(r) can be
achieved by using passive or active heat shields in the vicinity of
the solidification front, as proposed for example in U.S. Pat. No.
6,153,008. Most publications relate to an effect on the cooling
behavior due to modified heat shields. With the known prior art,
however, sufficient radial V/G homogenization for the production of
perfect silicon, especially with large crystal diameters, cannot be
achieved in this way. By means of impurities, for example nitrogen
or carbon, but also oxygen, the size and positioning of the defect
distribution can be influenced so that the precipitation of
impurities such as oxygen, can also be influenced. It is therefore
of great importance to be able to deliberately produce and control
both axial and radial impurity profiles.
SUMMARY OF THE INVENTION
[0016] It is an object of the present invention to provide a method
which makes it possible to deliberately set up the defect
distributions required by the customer in a single crystal, even
with large crystal diameters, so that as many semiconductor wafers
as possible can be obtained with the specified properties of the
single crystal. Semiconductor wafers which have only COPs,
especially those with a predetermined size and density
distribution, and semiconductor wafers which have no agglomerates
of self-point defects (perfect silicon), are of particular interest
in this context. Nevertheless, semiconductor wafers with a stacking
fault ring (ring wafers), with both self-point defect types or with
only one self-point defect type, together with a predetermined
oxygen concentration or particular oxygen precipitation, may be
specified by the customer.
[0017] The invention relates to a method for the production of a
silicon single crystal by pulling the single crystal, according to
the Czochralski method from a melt which is held in a rotating
crucible, the single crystal growing at a growth front, wherein
heat is deliberately supplied to the center of the growth front by
a heat flux directed at the growth front.
[0018] The invention also relates to a silicon single crystal. with
an oxygen content of from 4*10.sup.17 cm.sup.-3 to 7.2*10.sup.17
cm.sup.-3 and a radial concentration change for boron or phosphorus
of less than 5%, which has no agglomerated self-point defects, and
which is optionally doped with nitrogen and/or carbon. The radial
variation of the oxygen concentration (ROV) is preferably at most
5%, particularly preferably at most 2%.
[0019] The invention also relates to silicon semiconductor wafers
with agglomerated vacancy defects (COPs) as the only self-point
defect type, these defects having a variation in their average
diameter of less than 10% and being present on a circular surface
of the semiconductor wafers, the diameter of the circular surface
being at least 90% of the diameter of the semiconductor wafers.
[0020] Lastly, the invention also relates to semiconductor wafers
with certain other defect distributions. When analyzing the pulling
tests which were carried out, it was found that insufficient radial
homogenization of the ratio V/G.RTM.) is correlated with an
inadequate heat supply from the melt to the center of the growth
front. In the past, the importance of the heat supply from the melt
for the production of perfect silicon was not understood. According
to the present invention, it is recommended that heat be
deliberately supplied to the center of the growth front so that,
per unit time, more heat reaches the center of the growth front
than the edge region of the growth front surrounding the center.
This can be achieved by a heat source acting on the center of the
growth front and/or by an upward melt flow at the center of the
melt. Besides the importance of an axial heat flux directed at the
growth front, it was also found that an isothermal temperature
distribution in the melt, parallel to the growth front, in a region
of up to 5 cm below the growing single crystal, is particularly
advantageous for radial homogenization. Expressed in terms of an
axial temperature gradient Gs.RTM.) in the melt, a temperature
distribution in which a radial variation of the temperature
gradient in the melt is no more than 15% should be produced in a
region with an extent of up to 5 cm below the growth front and at
least 90% of the diameter of the single crystal. The radial
variation of Gs.RTM.) is preferably less than 10%, and particularly
preferably less than 3%. The present invention therefore provides
conditions suitable for deliberate defect control or for the
production of perfect silicon.
[0021] Especially with a view to the production of perfect silicon,
tests have shown that the method according to the invention is
particularly tolerant with respect to fluctuations of the pull
rate. For instance, it is possible to pull. silicon single crystals
with a diameter of at least 200 mm, which have no agglomerated
point defects, even if the pull rate fluctuates by .+-.0.02 mm/min,
particularly preferably .+-.0.025 mm/min or more, the fluctuation
range referring to a single crystal of at least 30 mm. This fact
increases the yield significantly, without the need to provide
additional error-prone regulatory means to control the pull
rate.
[0022] According to one embodiment of the invention, a heat flux
directed at the center of the melt is produced in the form of an
upward melt flow by iso-rotation of the crucible and the growing
single crystal, the crucible being rotated with at least 10% of the
rotational speed of the single crystal. Since the oxygen content of
the single crystal. is increased to technically inexpedient
concentrations by this, however, it is preferable to counteract the
incorporation of oxygen into the crystal lattice by applying a
magnetic field. For example, traveling magnetic fields (TMFs) which
produce an upward or downward flow parallel to the crucible wall,
or static CUSP fields which cause a reduction of the melt movement
in the vicinity of the crucible edge, are suitable for this. With
these magnetic fields, the oxygen content can be reduced to less
than 6.0*10.sup.17 cm.sup.-3 and the growth conditions can
simultaneously be stabilized. Currents of up to 3000 A with up to
50 coil turns are preferably used to generate the required magnetic
fields.
[0023] According to another embodiment of the invention, a heat
flux directed at the center of the growth front may also be
produced by a heat source which deliberately increases the
temperature at the center of the bottom of the crucible compared
with a temperature at the edge of the bottom. The temperature of
the crucible is higher at the center of the crucible bottom, i.e.
in the region above which the center of the growth front of the
single crystal lies, by at least 2 K, preferably at least 5 K and
particularly preferably at least 10 K, than the temperature at the
edge of the crucible bottom. One embodiment of the invention
therefore provides for the use of a heating resistor which is
fitted at the center of the crucible bottom, or on the crucible
shaft under the center of the crucible bottom. Instead of a heating
resistor, it is also possible to use an induction coil which is
operated at medium to high frequency (50 Hz to 500 Hz). The
electromagnetic forces due to the coil drive an upward flow
directed at the center of the growth front. The melt is also heated
from the center of the crucible bottom. Depending on the
geometrical arrangement, heating powers in the range of from 1 kW
to 60 kW, will be required.
[0024] According to another embodiment of the invention, a bottom
heater, which is conventionally present in pulling systems for the
production of single crystals with diameters of at least 200 mm, is
used for deliberately heating the melt from the center of the
crucible bottom, thermal insulation being used to ensure that the
bottom heater heats the center of the crucible bottom more strongly
than the edge of the crucible bottom. To this end, a concentric gap
filled with thermally insulating material is provided in an outer
region of the baseplate and/or the outer crucible, so that the
quartz crucible is thermally insulated more strongly in the outer
region. The baseplate carries the crucible and a graphite outer
crucible surrounding the latter. When heating is carried out with
the bottom heater, therefore, heat is supplied to the melt
essentially only at the center of the quartz-crucible bottom
because of the annular thermal insulation in the baseplate or the
outer crucible. For example, graphite sheets or graphite felts are
suitable as an insulator material for filling the gap in the
baseplate and/or in the outer crucible. The necessary bottom heater
power is preferably in the range of from 20 kW to 80 kW, which is
higher than the conventional powers. Thermal insulation may also be
integrated into the crucible shaft, so as to minimize the downward
thermal dissipation via the crucible shaft.
[0025] Another embodiment according to the invention for
deliberately supplying heat to the center of the growth front
consists in fitting a heat source below the center of the crystal
growing in the melt. This may be done using an electrically
operated graphite heating element embedded in quartz, or by means
of a heating element which is protected from the melt by using
other process-compatible materials.
[0026] According to another embodiment of the invention, a heat
flux directed at the center of the growth front is produced by an
electromagnetic field, to which the melt is exposed and which is
partially shielded by shielding at least 10% of the area of a wall
of the crucible against an effect of the electromagnetic field on
the melt. A particularly preferred way of producing such a heat
flux consists in using a traveling magnetic field. The forces due
to the field depend on the material of the shielding and on the
amplitude and frequency of the electric current which flows through
the coils producing the magnetic field. Metallic materials may be
used as magnetic shielding, for example copper plates with a
thickness in the centimeter range, which are arranged between the
magnetic coils and the crucible, and which hence remove some of the
area of the crucible wall and the melt lying behind it from the
effect of the magnetic field. Shielding which consists of two
mutually opposed plates, each with a vertex angle of 90.degree.,
has proved particularly effective. Frequencies of from 10 Hz to
about 1000 Hz are preferably used. A frequency range of from 30 Hz
to 100 Hz is particularly suitable when using a traveling magnetic
field with partial shielding in the form of rectangular copper
plates. Currents of up to 500 A with up to 50 coil turns are
preferably used to produce such a traveling field. Fast crucible
rotations of at least 3 rpm reduce the effect of the magnetic
field, so that the intended supply of additional heat to the growth
front can be influenced by the speed of the crucible rotation. The
amount of melt respectively present in the crucible should
furthermore be taken into account, since different melt flow
patterns may be formed as a function of this. The necessary
conditions, i.e. the ratio of magnetic field, shielding and pulling
process parameters, for example the crucible rotation, will each be
determined roughly by experiments and approximate simulation
calculations, as a function of the amount of melt present in each
case.
[0027] The aforementioned embodiments of the invention may be
combined with measures which are already known and which are
suitable for homogenizing the axial temperature gradient G(r).
Preferred combinations are ones in which heat is additionally
supplied to the phase boundary, which is formed by the growing
single crystal, the atmosphere surrounding it, and the melt. This
may, for example, be done by using a heat shield described in U.S.
Pat. No. 6,153,008. It is particularly preferable to use a heating
element on the lower edge of the heat shield, which is described in
that patent application. A cooler acting on the single crystal may
furthermore be fitted over the heating element, as described for
example in U.S. Pat. No. 5,567,399. This makes it possible to
increase the pull rate and to further adjust the radial
homogenization of G.RTM.). The accelerated cooling associated with
this furthermore makes the remaining COPs significantly smaller.
The size of these COPs can thereby be brought below a critical
value, below which these defects no longer have any effect on the
component function.
BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Other objects and features of the present invention will
become apparent from the following detailed description considered
in connection with the accompanying drawings. It is to be
understood, however, that the drawings are designed as an
illustration only and not as a definition of the limits of the
invention.
[0029] In the drawings, wherein similar reference characters denote
similar elements throughout the several views:
[0030] FIG. 1 schematically represents the principle of the method
according to the invention;
[0031] FIG. 2 shows profiles of the ratio V/G.RTM.) as a function
of the radius of the single crystal;
[0032] FIG. 3 shows the typical melt flows occurring in the
conventional Czochralski method (with counter-rotation of the
single crystal and the crucible);
[0033] FIG. 4 shows the profile typically resulting therefrom for
the axial temperature gradient Gs.RTM.) in the melt;
[0034] FIGS. 5 and 6 respectively show melt flow patterns and the
profile of the axial temperature gradient Gs.RTM.) as are
encountered when carrying out the method according to the
invention;
[0035] FIGS. 7 to 13 show various arrangements of preferred
embodiments of the invention;
[0036] FIG. 14 shows an arrangement according to FIG. 11, in which
a heating element and a cooling element are also provided; and
[0037] FIGS. 15 to 17 relate to examples according to the
invention, and show the distribution of defect types on various
crystal regions.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
[0038] Referring now in detail to the drawings, FIG. 1
schematically represents the principle of the method according to
the invention. The single crystal 1 is growing at a growth front 2,
to the center of which a heat flux 3 is deliberately supplied
through the melt. With the aid of the additional axial heat supply
indicated, it is possible to carry out homogenization of the radial
temperature gradient G.RTM.) at the growth front which is
sufficient for the production of perfect silicon, or to set up a
temperature gradient G.RTM.) required for deliberate defect
control, for single crystals with large diameters of at least 200
mm. The quality of the homogenization of G.RTM.) is dictated by the
temperature distribution in the melt. It is particularly preferable
for the axial temperature gradient Gs.RTM.) set up in the melt to
have the smallest possible radial variation in the melt, so as to
obtain the indicated isothermal temperature distribution 7 parallel
to the growth front.
[0039] The effectiveness of the method according to the invention
is demonstrated by the profiles represented in FIG. 2 for the ratio
V/G.RTM.) as a function of the radius of the single crystal, for
single crystals with a diameter of 300 mm. The thermal flow in the
melt toward the center of the growth front, which is obtained with
iso-rotation of the single crystal and the crucible according to
the invention, leads to a very significant radial homogenization of
V/G.RTM.), shown as curve (c), while attempted homogenization by
means of heat protection shields according to different designs (a)
and (b), which are not in accordance with the invention, are not
sufficient for the production of perfect silicon. The subsequent
figures contrast the effect of the central heat flux according to
the invention (FIG. 5 and FIG. 6) compared with conventional melt
convection (FIG. 3 and FIG. 4) in the form of results from model
calculations.
[0040] FIG. 3 shows the typical melt flows occurring in the
conventional Czochralski method (with counter-rotation of the
single crystal and the crucible), which are distinguished by an
axial flow directed downward at the crucible bottom. In this case,
the temperature conditions represented in FIG. 4 are obtained at a
few centimeters below the growth front. Gs.RTM.) exhibits a
pronounced change as a function of the radius. The radial change of
Gs.RTM.) within the crystal diameter is about 17%.
[0041] The conditions are significantly different when carrying out
the method according to the invention, for example according to the
embodiment in which the melt is exposed to an asymmetric traveling
field generated by means of two shields, which shield at least 10%
of the wall area of the crucible. The melt flow patterns
represented in FIG. 5 show an axial melt flow directed at the
growth front. The heat transport due to the melt flow-leads to a
significantly different temperature distribution in the melt below
the growing single crystal (FIG. 6) compared with FIG. 4. A
significantly more homogeneous temperature gradient Gs.RTM.) is
found in the melt, which provides the desired axial homogenization
of self-point defects and impurities and dopants in the single
crystal. The radial variation of Gs.RTM.) is less than 15% in a
silicon melt. For the conditions on which FIG. 6 is based, 7% was
determined on average.
[0042] The subsequent figures, FIG. 7 to FIG. 13, represent various
arrangements of preferred embodiments of the invention. Heating
elements play a central role in FIG. 7 to FIG. 10; these may be
designed as electrical heating resistors, induction heaters or
possibly radiation heaters, and are arranged at respectively
different positions below the growing single crystal. Each heating
element functions as a heat source, which produces a heat flux
directed at the center of the growth front of the single crystal.
In order to reinforce the effect of the heating elements, thermally
insulating elements 6, which may be graphite sheets or graphite
felts, may be fitted in a ring below the quartz crucible, although
not under the center of the crucible bottom. They impede off-axial
supply of heat to the melt. In order to focus the heating effect in
the melt flow directed at the center of the growth front, elements
with high or extremely high thermal conductivity, for example made
of graphite or other process-compatible materials, may be
incorporated at the center of the crucible bottom. The energy
supplied by means of the heating elements is in each case adapted
to the geometrical and process-related situation, and must for
example be readjusted according to the residual amount of melt in
the crucible, which decreases as the crystal grows.
[0043] FIG. 7 schematically shows the arrangement which, in
addition to a conventional main heater 4, has an additional heating
element 8 which is arranged as a crucible bottom heater below the
graphite outer crucible 5 and produces a heat flux 3 directed
upward at the center of the growth front 2 of the single crystal 1
by means of the thermal insulation 6. The thermal insulation 6 may
be integrated in the outer crucible and/or the baseplate, which
carries the outer crucible. The heating power of the additional
crucible bottom heater 8 should preferably be more than 2% of the
heating power of the main heater, in order to produce an effective
heat flux. The crucible bottom heater may, for example, be designed
as an electrical heating resistor made of graphite, and may
optionally be configured so that it can be translated. The
necessary heating power must be adapted to the respective amount of
melt (depending on the length of crystal that has already
solidified). It is in the range of more than 5 kW.
[0044] FIG. 8 represents other design features which lead to
improved heat transfer at the crucible center. For instance, the
central heat flux may be enhanced by means of an increased material
base at the quartz-crucible center, for example by a central
thickening 12 of the outer crucible. An insulating element 16 may
be inserted in order to prevent thermal dissipation via the
crucible shaft.
[0045] In the arrangement according to FIG. 9, an additional
heating element 9 producing a heat flux is integrated at the bottom
of the outer crucible 5. In this embodiment, it is possible to use
either an inductively operated heating element or a resistive
heating element, or a combination of the two.
[0046] In the arrangement according to FIG. 10, the heat flux
required according to the invention at the center of the growth
front is produced by a heating element 10 arranged in the melt,
below the growth front of the growing single crystal. To this end,
for example, it is possible to use a quartz-clad graphite heater,
for example a heater with the meandering structure of heating zones
which is represented on an enlarged scale.
[0047] With an arrangement according to FIG. 11, an intended heat
flux 3 directed at the center of the growth front is produced by
iso-rotation of the single crystal and the crucible. To this end,
the speed of the crucible rotation must be set to a value of at
least 10% of the speed of the crystal rotation. A preferred flow
pattern 11 is set up in the melt. During the pulling process,
additional variations of the crucible or crystal rotation may be
necessary in order to compensate for the varying thermal budget.
The generally high oxygen contents in the melt due to the
iso-rotation of the crucible and the single crystal can be reduced
by magnetic fields acting on the melt primarily in the edge region
of the crucible. Static, magnetic and CUSP fields are particularly
expedient, and facilitate oxygen contents lower than 6.0*10.sup.17
cm.sup.-3 in the melt without compromising the process conditions
according to the method.
[0048] With an arrangement according to FIG. 12, an intended heat
flux 3 directed at the center of the growth front is produced by a
static electric field between the crucible and the single crystal.
To this end, a positive voltage of more than 100 volts must be
applied to the crucible. A preferred flow pattern 11 is set up in
the melt.
[0049] Other preferred embodiments according to the invention
relate to the use of electromagnetic fields which produce a heat
flux directed perpendicularly to the growth front due to the forces
which they exert on the melt, the forces on the melt being limited
by shielding at least 10% of the wall area of the crucible. The
coils producing the magnetic field may be arranged outside or
inside the crystal pulling system. A preferred embodiment of this
type comprises a partially screened traveling magnetic field. FIG.
13 represents a suitable arrangement with a single crystal 1
growing at a growth front 2, a heat flux 3 produced by the effect
of the traveling field and directed at the center of the growth
front, and an annular heating element 4 arranged around the
crucible. A preferred flow pattern 11 is set up in the melt. The
traveling field is produced by a magnet 13, which is also arranged
in a ring around the heating. element 4. With a coil of up to 50
turns and a coil diameter of more than 500 mm, it has been found
that electrical currents of from more than 100 A to 500 A are
particularly suitable for producing the magnetic field. For partial
shielding of the traveling magnetic field, there are two mutually
opposed shields 14 fitted radially inside the magnetic coil, by
which the rotational symmetry of the field is broken so that
somewhat different conditions are formed in the direction of the
shields than perpendicularly thereto. The shields preferably
consist of copper and each have a vertex angle of 90.degree.. They
shield at least 10% of the wall area of the crucible.
[0050] As a particularly preferred embodiment of the invention,
FIG. 14 shows the combination of the embodiment according to FIG.
11 with an additional heat source 19, with the aid of which
additional heat is supplied to the phase boundary of the single
crystal, to the atmosphere surrounding this and to the melt. The
heat source 19 preferably comprises a heating resistor designed as
a ring, which surrounds the single crystal 1 in the vicinity of the
phase boundary. Powers of more than 5 kW are preferably delivered
to the heat source 19, so that the temperature gradient G.RTM.) at
the phase boundary of the single crystal is homogenized. The heat
source is connected via insulation to a conventional heat shield
18, which ensures sufficient shielding of the single crystal from
the heat radiation of the melt, and thereby also influences the
temperature distribution in the single crystal. To this end, heat
shields geometrically shaped according to the requirements are
used, which may consist of a plurality of layers of graphite,
graphite felt, molybdenum or combinations thereof. An additional
cooling device 17 is arranged above the heat source 19. The cooling
device 17 provides a further way of adjusting the necessary
temperature distribution. Furthermore, the cooling device increases
the gradient G overall, which makes it possible to use a faster
pull rate, for example more than 0.36 mm/min for perfect 300 mm
crystals. Static or dynamic magnetic fields are produced in the
melt by means of the magnetic coils 13 arranged around the
crucible, so that the necessary melt flows transporting heat and
oxygen can be set up accurately.
[0051] Of course, the present invention also covers other
combinations of the described embodiments and features, even if
such combinations have not been explicitly mentioned.
[0052] For instance, another preferred embodiment is based on the
one represented in FIG. 14, but instead of the annular heater 19,
it is equipped with features such as the partial thermal insulation
6 shown in FIG. 8 or the heating element 9 in the vicinity of the
crucible bottom as disclosed in FIG. 9. This embodiment makes it
possible to pull single crystals with a diameter of 300 mm or more
at a comparatively fast pull rate of at least 0.6 mm/min, with the
radial temperature gradient deviating by no more than 10% from the
critical value C.sub.crit. It is hence possible to produce single
crystals with an increased output, the agglomerated self-point
defects of which, due to their small size and composition, lead to
no productivity losses in the fabrication process of the electronic
components, or to significantly reduced productivity losses.
[0053] The particular uses of the invention will be presented below
with reference to three examples, which relate to the production of
silicon semiconductor wafers using a device according to FIG.
14.
[0054] The unagglomerated self-point defect regions were determined
with the aid of a charge-carrier lifetime measurement (.mu.PCD). To
this end, for example, axial sections in the single crystal are
smoothly etched, cleaned and heat-treated for 4 hours at
800.degree. C. and 16 hours at 1000.degree. C., and a lifetime
measurement is carried out followed by image processing. The
vacancy regions are thereby detectable, since there is a modified
lifetime due to the oxygen precipitates which are formed.
[0055] FIG. 15 illustrates the distribution determined with the aid
of .mu.PCD measurements for an axial crystal section. The single
crystal was pulled with an increasing pull rate. The radial region
which appears to be structureless due to the reduced oxygen
precipitation characterizes the region where interstitial silicon
atoms dominate, while vacancy defects are predominant in the other
regions. As the pull rate increases, transition is observed from
agglomerated interstitial atoms, the LPITs 30, through
unagglomerated interstitial atoms 31 to the unagglomerated
vacancies 32. A silicon wafer taken from the single crystal at the
section position A can therefore have radial regions of vacancies
32 as represented in FIG. 15, even at the wafer edge. For oxygen
contents of more than about 5*10.sup.17 1/cm.sup.3, the
resulting,sequence of regions can likewise be determined accurately
with the aid of the oxygen-induced stacking faults. With the method
according to the invention, the thermal conditions can be set up so
that any desirable predetermined radial defect distribution is
possible.
[0056] The distribution of the agglomerated self-point defects
(COPs or LPITs) was analyzed by the conventional method by means of
Secco etching (21 minutes at 30.degree. C.) on ground, smoothly
etched test wafers using a light microscope. The conventional
light-scattering methods were also employed on polished wafers
(SP1), so that larger COPs with a diameter of more than 90 nm could
be detected. In order to detect smaller COPs, these were enlarged
to the measurability range before the light scattering measurement
by means of a standard etching method, an SC1 treatment (3 hours in
hydrogen peroxide and aqueous ammonia), or a gate oxide integrity
test (GOI test) was carried out. Detailed size analyses of the
agglomerated self-point defects were carried out by means of
transmission electron microscopy studies (TEM). The measurement
methods have already been described many times and at length in the
literature, for example in H. Bender, J. Vanhellemont, R. Schmolke,
"High Resolution Structure Imaging of Octahedral Void Defects in
As-grown Czochralski Silicon" Japan J. Appl. Phys. 36 (1997), L
1217-L 1220, Part 2, No. 9A/B. The examples presented below relate
to primarily (100)-oriented single crystals with a diameter of 300
mm. The results are, however, readily applicable to other
orientations, for example (110) or (113), and larger diameters.
EXAMPLE 1
[0057] A silicon single crystal was produced, from which
semiconductor wafers with the following properties could be
separated:
[0058] With respect to agglomerated self-point defects, the single
crystals with a diameter of 300 mm had only agglomerated vacancy
defects (COPs), these defects having an average diameter of less
than 50 nm and being covered with an oxide layer, the thickness of
which was less than 1 nm. The thickness of the oxide layer was
usually more than 2 nm.
[0059] FIG. 16 shows such a polished and SC1-treated 300 mm silicon
wafer, which was examined for small vacancy agglomerates with a
diameter of less than 50 nm (small COPs) by means of laser
scattering methods. The defect distribution was also confirmed by
measurements on axial crystal sections and by GOI studies.
[0060] The particular advantage of such semiconductor wafers is
that the defects (small COPs) do not cause problems in the
fabrication of electronic components because, due to the small size
of the agglomerates and the small thickness of the oxide layer,
they can be erased by a heat treatment, at least in the regions
where the components are integrated. The heat treatment need not
necessarily be carried out separately, since the semiconductor
wafers are in any case exposed to the requisite temperatures of at
least 1000.degree. C. at the start of component fabrication.
[0061] The process parameters employed are generally derived from
the following formula, which gives the size distribution of the
COPs. 1 V COPs .infin. [ 1 q ( 1 - C crit ( V / G ) ) ] 3 2
[0062] The size distribution of the COPs is given by the volume
V.sub.COPs.
[0063] q is the cooling rate of the crystal at the solidification
front in the temperature range of from 1100.degree. C. to about
1000.degree. C. As described above, the defect function V/G.RTM.)
characterizes the crystallization process in respect of the defects
which occur, the critical limit being C.sub.crit=1.3*10.sup.-3
cm.sup.2 min.sup.-1 K.sup.-1.
[0064] After the typical proportionality factor has been determined
for a crystallization process, the size distribution of the COPs
can therefore be adjusted by means of V/G and the cooling rate.
[0065] For the example in question, a cooling rate in the
temperature range of from 1100.degree. C. to 950.degree. C. was
determined at about 0.8.degree. C./min from associated model
simulation calculations. To this end, with a cylindrical crystal
position of 50 cm, a ratio of the crucible rotation to the crystal
rotation of 1:2 was used, together with a heat supply from the main
heater 4, the bottom heater 8 and the ring heater 19 in the ratio
of about 1:0.3:0.2. The ratio V/G was up to 10% more than
C.sub.crit.
EXAMPLE 2
[0066] A silicon single crystal was produced, from which
semiconductor wafers with the following properties could be
separated:
[0067] The semiconductor wafers were free of agglomerated
self-point defects and two or more mutually separated axially
symmetric regions, in which unagglomerated vacancies dominate as
the defect type. The semiconductor wafers therefore have the
properties of a silicon wafer corresponding to the section A in
FIG. 15. The particular advantage of producing such semiconductor
wafers is that the process management during the production of the
single crystal is simplified, because less outlay is required on
control technology. This is because there is a particularly wide
process window in respect of the allowed variation of V/G. In the
case of such semiconductor wafers, the oxygen precipitation
occurring in the vacancy region can furthermore be adjusted
accurately to the requirements of the component fabrication.
[0068] In this example, at the section position which was about 30
cm along the cylindrical crystal length, a ratio of the crucible
rotation to the crystal rotation of 1:2.4 was used, together with a
heat supply from the main heater 4, the bottom heater 8 and the
ring heater 19 in the ratio 1:0.3:0.24. The ratio V/G was close to
C.sub.crit.
[0069] EXAMPLE 3
[0070] This example relates to semiconductor wafers with a defect
distribution similar to that of the semiconductor wafers in Example
2, with the difference that unagglomerated interstitial silicon
atoms dominate as the defect type in the two or more mutually
separated axially symmetric regions. The process management during
the production of the single crystal is simplified in the case of
such semiconductor wafers as well, for the reasons mentioned
above.
[0071] This distribution is illustrated in FIG. 17. The deliberate
control of the heat distribution at the solidification front even
makes it possible for a region 31 dominated by interstitial atoms
to be produced at the center, or to be alternate with a
vacancy-rich annular region 32 in a radial sequence.
[0072] The described distribution was achieved by a stronger heat
supply at the center of the solidification front. To this end, the
required heat flux was produced for each crystal position by means
of the heating powers, the crystal displacement and the crucible
and crystal rotations, together with the process pressure and the
argon flow.
[0073] In this example, at the section position which was about 45
cm along the cylindrical crystal length, a ratio of the crucible
rotation to the crystal rotation of 1:2.1 was used, together with a
heat supply from the main heater 4, the bottom heater 8 and the
ring heater 19 in the ratio 1:0.4:0.24. The ratio V/G was close to
C.sub.crit in this case as well.
[0074] With the aid of the proposed invention, it is moreover
possible to obtain semiconductor wafers in which agglomerates of
interstitial atoms (LPITs) occur as the defect type in regions with
dominance of the interstitial atoms, although their size is so
small that no secondary dislocations leading to A swirls are
formed.
[0075] An example of this involves silicon semiconductor wafers
with agglomerated interstitial atoms (LPITs) as the defect type,
although their size is so small that there are still no secondary
dislocations.
[0076] Another example involves silicon semiconductor wafers having
at least one region with agglomerated vacancy defects (COPs) as the
defect type, these defects having an average diameter of less than
50 nm and being covered with an oxide layer whose thickness is less
than 1 nm, and at least one region with agglomerated interstitial
atoms (LPITs) as the defect type, although their size is so small
that there are still no secondary dislocations.
[0077] Accordingly, while only a few embodiments of the present
invention have been shown and described, it is obvious that many
changes and modifications may be made thereunto without departing
from the spirit and scope of the invention.
* * * * *