Note: Descriptions are shown in the official language in which they were submitted.
1 1 6933~
This invention relates to the continuous growth of crystalline -
bodies from a melt, and more particularly to a system for automa~ic
- - string-stabilized ribbon growth.
Gne of the most important applications for crystal growing is
in the area of semiconductor substrate material. This substrate material
is available in ribbon form in which ribbons of large-grain polycrystalline
or single crystalline material are grown. The ribbons, usually of sili-
con, have wide application in the semiconductor industry and are especially
well adapted for use as solar cell substrates.
Several processes are described in the literature for the growth
of crystalline ribbon from melt. In one process capillary action is used
to feed molten material up through a die which is used to shape the grow-
ing ribbon. In order to exercise control over this process it is necessary
to control the heat removal from the ribbon surface, the pulling speed,
the average die temperature, and the precise t~emperature distribution
along the entire die, most speclflcally at the edges. This process
suffers principally from a lack of growth stability since the position
of the edges of the rlbbon are difficult to control and the ribbon oten
: "freezes" to the die when the die temperature or some other growth variable
momentarily fluctuates.
,
.
:;
..
~'.~i
~ -2- ; ~
b \ ~ 3 6 ~ 3 3 ~
An additional problem encountered with this technique is that all
the material flowing up through the die is incorporated into the solidified
ribbon. Since this material includes impurities, unacceptable impurity
levels can exist in the solidified ribbon. In contrast, when pulling a
ribbon directly from a melt impurities tend to become segregated at the
interface between the melt and the solidified ribbon, such that the impuri-
ties are rejected back into the melt and do not become incorporated into the
solidified ribbon. Since the use of a die partially prevents such segrega-
tion of impurities, pulling a ribbon directly from a melt is a preferred pro-
cess.
U.S. patent 3,129,061 describes another process ~or ribbon growth
which is referred to as the web dendritic growth process. In this process
dendrites, which are of the same material as the ribbon, are grown into a
melt supercooled at the ribbon edges in order to stabiliz0 the ribbon edge
position. The principal problem associated with this technique is the high
degree of temperature control needed to maintain dendritic growth at the
ribbon edges while maintaining conventional growth along the ribbon "web".
~See also "Thermal Analysis of Solidiication in Web-Dendritic Ribbon Growth",
by Harrill, Rhodes, Faust, and Hilborn, Journal of Crystal Grow-th, Vol. ~4;
pps. 34-44, 1978). Other web denclritic techniques are illustrated in U.S.
patents 3,298,795; 3,031,403; and 3,370,927. The last of these patents is
directed to the angular pulling of continuous dendritic crystals.
By way of further background, a technique for growing a matrix
structure of silicon is illustrated in an article by Theodore P. Ciszek and
Guenter H. Schwuttke, entitled "Inexpensive Silicon Sheets for Solar Cells",
NASA Tech Briefs, Winter 1977, pps. 432-433. In this technique a graphite
screen is dipped into a liquid silicon bath and is then pulled from the bath.
This produces a patterned sheet or film of liquid silicon which solidifies in
the screen to produce a textured semicrystalline composite. It will be
appreciated that the technique described by
~ 1~93~
1 Ciszek, et al is not a crystaI growing technique because silicon is first captured by
2 capiIlary action in the graphite screen and is then held until such time as it
3 solidifies This differs from the growth of ribbon from a melt in which
4 crystallization takes place at the face of the melt as the ribbon is withdrawn. In
ribbon growth, grain boundaries generally exist perpendicular to the plane of the
6 ribbon, while the Ciszek, et al technique generally produces randomly oriented
7 grain boundaries which act as random carrier traps for impurities, resulting in
8 devices which are less uniform in performance. The Ciszek et al technique also
9 results in undesirably sma11 grain sizes and in a non-uniform central web which is
encumbered by the grid structure. Because of the use of the grid, the product ofll Ciszek, et al is not a flat sheet, and if the grid embedded in the semiconductor
12 ribbon is not very precisely placed, shorting of p-n junctions created by diffusion
13 can occur.
14 With respect to prior art crystal growing furnaces, reference is made to
U. S. patents 3,639,718 and 3,865,554. Both of these patents describe furnaces
16 suitable for batch-type crystal growth using the Czochralski method. These
17 ~ furnaces are designed for processes requiring critical monitoring during crystal
18 ` pulling and are not designed or adapted for continuous ribbon growth. More
19 specifically, these furnaces are designed to carry out crystal growing with temper-
atures maintaii~ed to + 0.lC, with visual inspection and automatic control being a
21 prerequisite for obtaining uniform crystal properties.
SUMMARY OF THE INVENTION
22 In the present invention, a ribbon with an unobstructed central web is
23 grown direetly from ~the melt, with edge positions determined and stabilized by
24 wetted strings or strands moving continuously in parallel up throu~h the melt.
~5 During the ribbon drawing process, the strands are frozen into the growing ribbon.
lJ69336
These strands, unlike dendrites, are of a material different from the melt. For
2 ~ example in the growth of silicon ribbons, graphite, carbon, silicon carbide or quartz
3 strings are preferred, and are preferably wetted by the melt. It will be appreciated
4 that the above mentioned strings span a range of wettabilities such that the
amount of material attaching to the strings can vary while nonetheless resulting in
6 a string frozen into the ribbon.
7 In this string stabilized growth (SSG) process, the shape of a surface
8 tension controlled meniscus is defined on the bottom by the melt, on the top by the
9 interface with the growing crystal, and at the edges by the wetted strings. The
parameters that must be controlled in the prooess are pulling speed, rate of heat
11 removal from ribbon surfaees, and average melt temperature. Due to the capillary
12 ~ indueed edge definition, it is not necessary to maintain precise control of the
13 temperature of the melt at the ribbon edges vis-a-vis the ribbon center.
14 ~ In accordance with this mvention string stabilized ribbons may be grown
over a wide range of angles relative ~to the melt surface. Growth at a relatively
,
I6 low angle to the melt has the primary advantage of increasing the area of the
17 soIidification interface, thereby inereasing the maximum possible growth speed.
18 Continuous introduction of the string to the melt may be accomplished by a
19 number of techniques. In one embodiment, a crucible is provided having small
apertures in its bottom and through which the strings are introduced, with the
21 surface tension of the melt usually being sufficient to counteract the gravitational
22 head to retain the molten material in the crucible and prevent leakage through the
23 apertures. GravitQtionally produced leskage can alternatively be prevented by gas
24 pressure or by electromagnetic forces utilized to balance the gravitational head of
the liquid.~ ~
26 The transference of Impurities from the strings into the melt is controlled
27 either by minimizing the time during whlch the strings are immersed in the melt or
3 3 6
by freezing the melt around the strings as they enter the melt, thereby freezing in
2 impurities and confining them to the areas occupied by the strings. The freezing
3 technique can be accomplished by passage of an electrical current through the
4 strings, across the interface between the string and the melt, and through the melt
to form a Peltier or thermoelectric junction between the string and the melt to
6 achieve localized Peltier cooling.
7 Unlike ribbon forming systems which utilize a die, when pulling a ribbon
8 directly from a melt, a substantial portion of the impurities within the melt are
9 rejected back into the melt during the solidification process because they are
10 ~ frozen out of the ribbon as the ribbon solidifies at the melt-ribbon interface. In a
11 non-replenished system, should this occur for any length of time, the concentration
12 of impurities within the melt increases.
13 As part of the invention, the impurity concentration of the melt at the
14 crystal growing site is minimized by electrom~gnetic stirring and melt dumping to
15 ' reduce the impurity concentration in the ribbon. In one embodiment, minimizing
16 the amount of irnpurities introduced from the melt into the ribbon is achieved by
17 maintaining a constant flow of melt perpendicular to the plane of the growing
18 ribbon. This permits convection and diffusion to carry awQy the impurities that are
19 rejected at the growth interface. Additionally, because of the segregation of
impurities at the growth interface which concentrates the impurities in the melt RS
21 opposed to the ribbon, continuous melt flushing or~ dumping is effective to reduce
22 impurity concentration in the ribbon. Flushing off a portion of the melt reduces
23 the impurity concentration in the melt within the crucible and thus results in a
24 lower impurity concentration in the solidified ribbon.
An Improved furnace is provided for automatic continuous ribbon growth in
26 which close temperature control or visual inspection is not necessary, and in which
27 temperature variations of as much as 10C can be tolerated. The furnace Is
:
- 6 -
I ~ 6~3~
provided with a crucible supported by posts, selected opposing posts being provided
with pins adapted to fit into radially extending slots in the bottom of the crucible
3 for accommodating thermal expansion and contraction of the crucible while
4 maintaining angular position. A centrally located positioning tube is provided for
locating the center of the crucible within the furnace. In one embodiment, a
6 thermocouple is conveniently disposed in one end of the positioning tube for
7 crucible temperature monitoring.
8 Heater rods are mounted throu~h the base of the crucible for more
9 efficient and controllable energy transfer to the melt than is possible with
conventional heating elements which surround the crucible. In one embodiment,
11 the heater rod ends are in sliding fit engagement with electrical contact blocks
12 positioned on either side OI the crucible, with the sliding fit accommodating
13 thermal expansion and contraction of the rods. Moreover, non-uniformities in the
14 temperature profile of the melt can be corrected by changing the size and shape of
individual heater rods.
16 String introduction tubes are provided at the string introduction points in
17 the bottom of the crucible to prevent by capillary action, any melt leakage at
18 these points. In order to prevent freezing of the molten material within a tube, the
19 tube is recessed within the crucible such that the bottom portion of the crucible
surrounds the tube, with heating provided by radiative transfer from the sur-
21 rounding crucible walls. String positioning or guide tubes are disposed below the
22 crucible in alignment with the string introduction tubes to facilitate strhlg
23 introduction.;
24 A melt flushing tube is provided to permit controlled removal or flushing of
molten material as the melt is replenished. ~ Such removal is controlled in one
26 embodiment by selective freezmg and thawing of the melt in the tube; in another
27 embodiment by~cQpillsry action; and in 8 further embodlment by controlled suction.
3 ~
The flatness of the ribbon can be maintained for angled growth by a
2 straight-edged wetted stabilization member within the melt which acts to control
3 the meniscus shape to prevent curvature or bowing of the ribbon. In an inclined
4 surface embodiment in which strings are pulled vertically through an inclined surface
down which pumped melt is flowing, an abrupt change in melt flow direction is
6 ~ provided to prevent curvature or bowing of the ribbon. The melt is pumped to the
7 top of the inclined surface in one embodiment by electromagnetic pumping.
BRIEF DESCRIPTION OF THE DRAWINGS
8 The present invention is more fully set forth below in the following
g j detailed description and in the accompanying drawings, in which:
Fig. 1 is a diagrammatic illustration o- a prior art ribbon growth process in
11 ~ which a die is used to feed molten material up from a melt;
12 , Fig. 2 is a diagrammatic illustration of a prior ~rt we~dendritic growth
13 process;
14 ~ Fig. 3 is a diagrammatic illustration of one embodiment o~ the subject
invention,~ in which edge definition is a result of the use of wetted strings or
lS strands;
17 Fig. 4 is a diagrammatic illustration of the subject process illustrating the
18 use of a seed crystal for start up;
:
19 i Pigs. SA - 5D are diagrammatic illustrations of crystal growth at angles to
the vertical, for;string supported systems; ~ ~ ~
21 ~ Fig. ~6 ~ is ~a cross-sectional Illuetration showing one method of string
22 introduction in which a pulley o~ like direction changing device is located within
23 the melt;
24 Pig. 7 is a cross-sectional illustration showing another method of string
introduction in which the string is introduced into the melt through an aperture in
26 the bottom of the meIt crucible, with melt ~ontainment vill surface tension;
-- 8 --
-
1 ~ 69336
Fig. 8 is a cross-sectional illustration of string feed and melt containment
2 utilizing a pressurized chamber to counteract gravitational force;
3 Fig. 9 is a cross-sectional illustration of string feed and melt containment
4 ~ utilizing electromagnetic forces to counteract gravitationally induced flow;
Fig. 10 is a diagrammatic illustration of thermoelectric melt freezing
6 resulting in impurity trapping at the string;
7 Fig. 11 is a diagrammatic illustration of impurity control with electro-
8 magnetic stirring combined with melt flushing;
g Fig. 12 is an isometric view of one embodiment of an improved furnace for
use in the string stabllized growth of ribbon;
11 Figs. 13A -13F illustrate the crucible assembly for the furnace of Fig. 12,
12 ~ in exploded view, and as partially and fully assembled;
13 , Fig. 14 is a partial cross-sectional view illustrating string introduction and
14 ribbon pulling apparatus;
15 ' Fig. lS is a cross-sectional illustration of a string positioning tube for use
16 ' in the furnace of Fig. 12;
17 , Fig. 16 is a diagrammatic illustration of a heater rod and connection block
18 for use in the furnace of Fig. 12;
19 , Fig. 17 is R cross-sectional view of a string introduction tube surrolmded by
crucible walls to prevent the freezing of melt in the string introduction tube;
21 Figs. 18, 19~ and 20 illustrate meniscus control flanges for preve~lting
22 meniscus creep;
23 Fig; 21 is a cross-sectional diagram illustrating suction control for melt
24 dumping;
Fig. 22 is a diagrammatic illustration showing the use of a eapillary tube
26 for melt dumping control;
27 Fig. 23 is R diagrammatic illustratlon representing conditions for vertical
28 ribbon growth;
~'
_ 9
; : . . .
~^
1 ~ 69336
Figs. 24 and 25 are diagramrnatic i31ustrations representing conditions for
2 ~angled growth, illustrating measurement of meniscus height and curvature;
3 Fig. 26 is a diagrammatic illustration of the trough-shaped response of the
ribbon to angled growth;
Fig. 27 is a diagrammatic illustration showing the effect on ribbon {latness
6 due to changes in meniscus radii;
7 Figs. 28A and 28B are diagrammatic illustrations showing the effect of
8 interposing a stabilizing device in the melt to promote flat ribbon growth;
g , Fig. 29 is a diagrammatic illustration of a wetted stabilization member for
10 preventing ribbon curvature or bowing during angled growth;
11 Figs. 30A and 30B illustrate the attachment of the lower meniscus to the
12 wetted stabilization member;
13 Fig. 31 is a cross-sectional and diagrammatic illustration of an inclined
14 surface ribbon growth technique in which a separate inclined surface is used;
¦ Fig. 32 is a cross-sectional and diagrammatic i~lustration of an inclined
6 surface rlbbon growth technique in which the inclined surface is an integral part of
17 the cruciblei
18 Fig. 33 is a cross-sectional and diagrammatic illustration of the inclined
19 surface growth technique showing the use of a flow discontinuity to enable flat
ribbon growth;
21 ` Fig. 34 is a cross-sectional and diagrammatic illustration of a portion of
22 Pig. 33 showing the~region around the flow discontinuity;
23 Figs. 35 and 36 are cross-sectional and diagrammatic illustrations of two
24 alternative methods of forming flow discontinuities for the inclined surfaoe growth
technique of Figs. 31 and 32; and,
; 26 Fig. 37 is a diagrammatic i~lustration of an electromagnetic melt pumping
27 technique for use In the method of Pig. 31.
: . :
- 10 -
3 3 6
DE~AILED DESCRlPrION
Pig. 1 illustrates the prior art process for the growth of crystalline ribbon
2 utilizing a die 1û, formed of spaced members 11 defining between them a capillary
3 filled slot 12. Surface tension forces advance molten material up through slot 12
4 and a ribbon 14 is built up at a meniscus 16 at the outlet 18 of the die. In this
; 5 process of building up a ribbon, a number of parameters must be controlled so that
6 the ribbon is formed uniformly. Moreover, further drawbacks in the utilization of
7 a die include the difficulty in attaining growth stability due to edge position
8 dislocation and the inability to concentrate impurities in the melt as the ribbon is
9 formed.
Referring now to Fig. 2, edge stability has been promoted in the past
11 through the utilization of dendrites which are grown from a seed 20 which has a
12 laterally extending portion 22. Using suitable growing conditions, edge dendrites
13 24 are formed with a central web 26 ~extending therebetween. Latera] growth
14 ~ b~ins when the seed is dipped into a melt 28 which is supercooled at the positions
15 ~ wheie the dendrites are to be grown. In one embodiment the melt consists of liquid
I6 silicon. When pulling commences, the two coplanar dendrites grow downward into
17 the melt from the ends of the seed. The melt is pulled up between the dendrites by
18 surface tension to form the web. Solidifieation in the web occurs above the~melt
19 surface while the dendrltes grow below the ~surface,; with the growth of the
20~ dendrltes re`quiring the ~melt to~ be supercooled~at the ribbon edges. ~As mentioned
` 21 hereinbefore, the principal problem associated wlth this technique is the high
: : ~ :: : : :
22 degree of temperature control~`needed~to maintain dendritic growth ~t the~rlbbon~
23 edges while maintaining conventional growth ~along the ribbon web.
24 In contradistlncbon to ~the prlar art ~techn~ques of Figs. 1 and 2, the subject
invention proYides, as shown in Fig. 3, a thln, relatively wide crystalline body 30
.
26 grown directly from a melt 32 of the same material, with the ribbon edges defined
:: : :
:
: :
.
- 11 -
1 1 ~933B
by wetted strings 34 pulled up through the melt. As the ribbon is formed the
2 strings are frozen into the growing ribbon, with a thin film of melt 35 frozen about
3 strings 34. The central part of the ribbon is defined by the capillary shaping action
4 of the meniscus, here illustrated at 36. The growing ribbon or web 38 is contained
between the strings and solidfies at interface 40. It should be noted that while Fig.
6 3 illustrates a case where the diameter of the string is approximately equal to the
7 ribbon thickness, the same technigue may be used to grow ribbon the thickness of
8 ~ which is either greater or smaller than the diameter of the string.
9 Fig. 4 illustrates how the growth process is initiated. A seed 42 with
strings 34 attached is lowered into melt 32, contact is made, and a meniscus is
11 established. Growth is then initiated by pulling the strings and seed upward. Note,
12 the seed/string unit may include a piece of previously grown ribbon.
13 The growth of the crystalline ribbon itself will now be examined in detail.
14 The growth of a thin sheet directly from the melt is controlled by a combination of
geometrical and thermal constraints, as explained below. The geometrical
16 constraints are dictated by the meniscus geometry. The local radius of curvature
17 of the meniscus along the face of the ribbon is determined by a balance of the
18 gravitationally induced pressure head and surface tension induced pressure drop
19 across the liqwd-gas interface. Laplace's equation applied at all points of the
20 ~ ribbon except near the edges yields:
21 R= ~O ~ ~ (Eq.1)
22 where R is the locaI radius of curvature of the meniscus in a plane perpendicular to
23 the ribbon,~ is the surface tension of the liquid, ~ is the mass density of the
24 liquid, g is the local acceleration of gravity, and h is the elevation of a point above
the melt surface. The menlscus must join the melt surfaee continuously, that is, it
26 must have zero slope at the melt surface. In addition, the meniscus must meet the
27 growing crystal at some specific angle, this angle determined by the properties of
28 the material being grown. For silicon, this angle is approximately 11~.
-- 12 -
;
~ 3 fi933~
The Laplace equation and the two boundary conditions for the meniscus
2 may be used to find an expression for the meniscus height, i.e. elevation of the
3 ; interface above the melt, required for constant thickness growth. For the vertical
4 growth of silicon:
Meniscus Height = 1.09 ~ (Eq. 2)
6 A similar expression may be found for ribbon growth of any material at any angle
7 to the melt.
8 ; Thus, steady state, constant thickness growth requires that the menîscus
9 establish itself at a very specific level for~ each growth situation. A meniscus
which is higher than this value will lead to a ribbon of decreasing thickness, while a
11 meniscus lower than this value will lead to an increasing thickness.
12 ; Whereas the meniscus determines the rate of change of the ribbon
13 thickness, thermal considerations primarily determine the meniscus height itself
14 An instantaneously valid energy balance must be maintained at the intsrface
between heat conduction up the meniscus from the melt, heat of fusion generated
16 i by the solidifying crystal at the interface and heat conducted up and through the
17 ribbon from the interface. The heat conducted up the meniscus depends upon the
18 temperature of the melt and the meniscus height. The heat of fusion generated
1- 19 ' depends primarily on the ribbon thickness and the growth~ speed, while the heat
, conducted up and through the ribbon depends upon the ribbon thickness and the
21 , manner of the heat removal from the ribbon surface. Thus, the requirement of an
22 instantaneous energy balance at the interface defines a fairly complex relationship
.~
23 between all ;relevant growth variables. This relatlonship, in combination with the
24 geometrical considerations discussed earlier, dictates the physics of the growth of
25 ` ribbon from the melt.
26 The process may be shown to be stable for essentially all configurations. In
27 steady state growth, geometrical rejuirements such as angles o' meniscus attach-
- . :
:`
- 13 -
.
1 ~ 6933~
ment, curvature of meniscus surfaces, etc, fix the meniscus height and hence, the
2 heat flux up the meniscus for a given melt temperature. Thus, the steady state
3 ribbon thickness is principally fixed by an interaction of the heat of fusion
4 generated at the interface with the heat conducted up and through the ribbon, with
the system reactin~ in a stable manner to adjust the thickness of ~he ribbon being
6 grown until the proper heat balance is attained. This discussion applies to both
7 vertical growth of a ribbon, as well as growth at an angle to the melt.
8 The width of the ribbon is of course defined by the capillary attachment to
9 the edge stabilizing strings. This surface tension attachment provides for a wide
, latitude of allowed temperatures at the ribbon edges, as the edge position is no
11 longer determined primarily by thermal effects. The strings also contribute
12 greatly to the stability of the ribbon when produced at angles to the vertical.
13 I Figs. 5A through 5D show cross sectional views of ribbon growth at various
14 1 angles to the melt. These approximately scale drawings are for a material with an
~ angle, h of attachment of the meniscus to the solid of 30. Fig. 5A shows
16 ~ i vertical growth from the melt; Fig. 5B shows growth at 60 from the horizontal;
17 ~ Fig. 5C shows growth at 30 from the horizontal; and Fig. 5D shows growth at 159
18 from the horizontal.
19 1 These figures serve to illustrate how the boundary conditions of meniscus
~ attachment to the growing crystal may be met while growing at any angle to the
21 melt in steady state growth. It should be noted that while the interfaces 40 in
22 , Figs. 5A through 5~ are shawn as straight lines, they need not be straight. In fact,
23 they may assume any shape, concave up, concave down, etc., as dictated by the
24 thermal conditions. The end points of the interface are where the triple junctions
of solid, liguid and gas exist. Geometrically, the distance between these end
26 points, in combination with the angle of growth from the melt, determines the
27 steady state ribbon thickness being grown.
~"
I 1 69336
It will be appreciated that the subject string edge stabilization technique
2 greatly enhances the ability to grow ribbon at any angle to the melt. The main
3 advantage of pulling at an angle is increased growth speed due to increased
4 ` interface area.
'.
INTRODUCTION OF THE STRING TO THE MELT
Figs. 6 through 9 illustrate methods of introducing the edge stabilizing
6 string to the melt. In Fig. 6 string 34 is passed around a pulley 44 which is totally
7 submerged in meIt 32. The string enters the melt, passes around the pulley and
8 then emerges from the meit where it is incorporated into the growing ribbon. The
9 direction of movement of the string is illustrated by arrows 46.
In Fig. 7 a relatively small string aperture 50 in the bottom of a crucible 52
1l 1 accommodates the string. The gravitational~ pressure head of the melt is
12 I counteracted by surface tension of the melt attached to both the crucible and the
13 , string moving upwardly, as illustrated by arrow 54.
li ~ In another embodiment, as illustrated in Fig. 8, a small chamber 60 is
provided immediately below a string aperture 62 in a crucible 64. This chamber is
16 ~ pressurized with gas to counteract the gravitational head of melt 66 in the
17 ' crucible. Chamber 60 has a gas inlet channel 68 communicating with a suitable
18 pressurized~gas source (not~shown). ~ ~
-19 In a still iurther embodlment, shown in FIg. 9,;electromagnetic ~`orces are
induced in the molten material to counteract the gravitationally induced pressure.
21 These forces act in~the dlrection of arrows 70. In thls method, a magnetlc field,
~2 indicated by X's, is created perpendicular to a string guide tube 71 depending from
23 a crucible 72. A current generated by connecting a DC source 76 across strlng
24 guide tube 71 is then passed through the melt in this area in a direction
i.
perpendicular to both the axis of the guide tube and the applied magnetic field.
:
- 15 -
1 ~ 6933G
The result is an upwardly directed electromagnetic force induced directly in the2 melt. This ~echnique is used with materials that are electrical conductors when
3 molten and where the crucible offers some resistance so that current will flow
4 transverse to the tube axis. The magnetic field may be applied by current loops,
and the field may be introduced indirectly through the crucible.
.
REDUCTION OF STRING CARRIED IMPURITIES
.
6 It will be appreciated that any string, no matter how well purified, will
7 carry with it some impurities. As the string passes through the melt, these
8 ` impurities enter the melt and cause its contamination.
9 Two techniques may be used simultaneously to reduce the level of such
impurity introduction. First, the residence time of the string within the melt is
11 minimized by making the melt depth as small as possible. However, no matter how
12 short the residence time, some Impurities will enter the melt Irom the string. This
13 may be avoided by freezing a layer of melt to the string as it enters the melt.
14 ~In one embodiment, the string is cooled before it enters the melt, thus
freezing the melt to it on contact. In one method, the string is cooled by passing
16 ~ cooled gas over it as it enters the melt. However, after a small amount of time,
17 ~ the string will warm up to the melt temperature, and the newly frozen material
18 will melt, thereby requiring geometry in which the residence time of the string in
19 the melt is minimized as by reducing the melt depth.
In another embodiment, a thermoelectric effect between the string and the
21 melt is used to freeze the silicon around the string. In this technique, illustrated in
22 Fig. 10, a current is passed through the string 34 in the region below the melt. This
:
23 current is generated by a DC source 80 which is connected to crucible 82 via line
24 86 and to string 34 via line 88 and conductive rollers 84. In operation, current
passes along the string, through the junction between the string and the melt~
-- 16 --
1 ~ 6~336
through the melt and crucible, then back to the source through line 86. It will be
2 appreciated that any two dissimilar conductors will have between them a relative
3 Peltier coefficient, a measure of the thermoelectric potential. If a current is
4 i caused to flow across such a junction, heat is either evolved or withdrawn at the
j~mction, depending on the direction of current flow. In the case of silicon and6 graphite, graphite may be doped (e.g. with boron) so as to have a significant Peltier
7 coefficient ~with respect to silicon. Thus, the Peltier effect may be used to
8 withdraw a small amount of heat directly from the string melt interface.
9 Assuming sufficient current, the Peltier effect will cause a small amount of
- 10 material to quickly freeze around the string as it enters the melt, thus sealing in
j
11 the impurities.
12 , In practice, both techniques, cooling the string and Peltier cooling, may be
!
13 , used. However, it should be noted that the passing of current through the string
14 i will cause a certain amount of Joule heating in the portion of the string leading to
15 I the melt.
SEGREGATION OF IMPURITIES INTO THE MELT
16 i In Fig. 11, whenever solidification takes place at a well defined interface,
17 ~ a phenomenon known as Impurity segregation takes place. Because of considera-
18 tions of chemi~al equilibrium, only a certain fraction~of the impurities in the melt
19 may be incorporated into the~solida the rèmalnder ~belng rejected at the interface.
20 This segregation is used ~to advantage in almost all~ crystal growth processes in
. ~
21 order to purify the material~being grown.
22 In order to maximlz~e the beneficlal effects of impurity segregation, it is
23 necessary to maintam a flow of melt by the interface in order to ~arry away the
24 rejected impurlties. Thus, the present invention includes maintaining a flow melt
25 under the growing ribbon, substantially perpendicular to the plane of the ribbon.
- 17 -
11 1 6933~;
However, segregation of impurities into the melt is not sufficient to insure
2 significant purification of the material in a continuous growth process. This is
3 because, in a continuous growth process using replenishment of the melt, the
4 impurities rejected at the growth interface continually build up in the melt. As the
level of impurities in the melt builds up, the amount of impurities incorporated into
6 the ribbon also increases. In the present invention, this problem is solved by
; 7 maintainin~ a constant flush or dumping of the melt. The functioning of this
8 technique may be explained as follows:
g Considering an impurity for which the segregation coefficient is .001, one
part in a thousand of the impurities arriving at the interface is incorporated into
11 Ithe crystal. Assumlng that the raw material has a 10 parts per million (ppm)
12 ~impurity level, if the melt is composed only of such material, as will be the case
; 13 ,initially, one would expect to grow a crystal of .01 ppm impurity level due to the
`
14 , segregation coefficient of .001. However, with the passage of time, the impurities
15 ! accumulate in the melt and~eventually the crystal will have 10 ppm of impurity,
16 t i.e., the impurity level of the incoming material. At this point, since the crystal
7 thas an impurlty concentration of 10 ppm, the melt will actually have 10,000 ppm of
18 impurity. However, if 1/lOth of the melt is continually dumped in the steady state~
19 with replenishment of fresh stock with 10 ppm impurity concentration, it can be
20 ` shown that the melt will have approximately lOI~ ppm of impurity and the crystal
21 will have O.l~ppm of impurity. Thus, a high degree of purification is maintained
22 ~ even in the steady stQte. ;~ ~
23 F~g. 11 illustrates the manner of continual replenlshment, circulation and
24 flushing of the melt. Means for reducing melt depth to minimize string residence
time is also illustrated. The replenishment~system is indicated schematically by a
26 tube 90 through which small bits of raw material are dropped into the melt. The
27 circulation of the melt under the growth interface is accomplished by substantially
- 18 -
r -
1 ~ 89~3~
circular motion of the melt, with one circulation eddy on either side of the ribbon
midline as illustrated by dotted arrows 92. In the case of an electrically
3 conducting melt, this circulation can be accomplished by Induced electromagnetic
4 forces in the liquid melt. For example~ if the crucible is made of graphite and the
melt is silicon, leads g4 can be attached from a DC source 95 to the crucible
bottom near the ribbon edges adjacent string introduction apertures 114 at
7 contacts 96. The substantially higher electrical conductivity of the liquid silicon
8 with respect to the graphite causes the majority of the current to flow through the
g liquid silicon as illustrated by dotted arrow 98. A magnetic field is simultaneously
10 ~ applied perpendicular to melt surface 102, such that the induced forces in the melt
11 causes it to ciroulate in approximately the manner indicated. The direction of the
12 magnetic field is indicated by arrows B. This magnetic field may be applied by
13 external magnets or, more ccnveniently, by the field induced by the current passing
14 I through the main furnace heater if a heater coil is used.
A hollow barrier 110 is disposed in crucible 112 to reduce the impurity
16 level by reducing the melt depth, with the barrier having the aforementioned string
17 introduction apertures 114 therethrough.
18 Melt dumping may be accomplished through the use of a melt flushing tube
19 120 projecting downwardly from crucible bottom 122. Dumping may be controlled
2û through the seiective sealing available through induced electromagnetic forces
21 described previously. Alternatively, the tube may be sealed by maintaining the tip
22 at a temperature slightly below the melting temperature of the melt. A slight
23 application of heat to the tip by means of a small resistance heater or the like,
24 suffices to thaw the maleriel in the tip and allow it to ilow out of the crucible.
. .
.
:. .
- 19 - .
~ `~
. ` I 3 ~g33~
FURNAC~ CONSTRUCTION
Referring to Fig. 12, a furnace 130 is illustrated having a base 132 to which
2 is secured a main watèr-cooled housing 134 defining therein a vertically extending
3 chamber 136. The furnace is provided with an angularly offset viewing extension
4 138 having a sealed viewing port 140 of glass or other light transmissive material, a
loading port 142 through which a cruclble 144 and all furnace components may be
6 inserted, and a housing extension 150 projecting radially outward from main
- 7 ` housing 134. Within chamber 136 a heater contact block 200 is disposed to retain
8 ~ heater rods 198 which run through the base of the crucible to heat the melt carried
9 ` therein (See Fig. 13A). Water cooled electrical conductors 156 and 158 provide
` electrical power to heater contact blocks.
11 Housing extension 150 includes a vacuum pump extension 152, which may
12 ~ be used either for evacuating the housing or for introduaing an inert gas which
flows over the crucible and exits from the top of chamber 136 through the aperture
14 I provided for the ribbon's exit. Coils 154 are arranged to circulate a cooling liquid
I to dissipate heat from chamber 136. Ribbon pulling apparatus generally indicated
16 , at 160, located on top of main housing 134, includes rolls 162 which are powered by
17 1 an auxiliary motor 164, and which are adapted to ~ngage a ribbon 166 to pull it
18 1 upward as indicated by arrow 172. A fixture 174 is provided to hold a seed crystal
19 utilized to initiate ribbon growth and to which the ribbon is affixed.
Fig. 13A is an exploded view of the apparatus contained within the
21 chamber 136. Crucible 144 is mounted to a pedestal 170 which has a top surface
.
22 172 into whicù apertures 174 of various diameters are machined to accommodate
23 screw threaded, press fit or slide fit pieces A lower insulation pack 176 Is placed
24 on pedestal 170 and is maintained in position by locator posts 178 which extend
through apertures 179 in insulation pack 176. Crucible 144 is mounted over
26 insulation pack 176 and is positioned and supported by support posts 180 ~ 186
.
: .
- 20 -
1 3 B9336
.
which project through apertures 188 in the lower insulation pack. Lower insulation
2 pack 176 is also provided with apertures 190 through which string positioning tubes
... . . .
3 (Fig. 14) project, these tubes serving to guide the stabilizing strings into crucible
4 144 via apertures 192 in the crucible bottom. Additionally, an aperture 194 is
provided in the lower insulation pack to accomodate a central thermocouple
6 protection tube ~Fig. 14) housing a thermocouple. Pedestal 170 is provided with
7 corresponding apertures lgO~ and 194' to accommodate the string positioning tubes
8 and the thermocouple protection tube.
g With respect to the crucible, crucible 144 is provided with channels 196
which act as radiative cavities for heater rods 198 which pass therethrough. The
11 heater rods are supported by heater contact blocks 200 apertured to accommodate
12 ends 202 of the heater rods in Q sliding fit, as will be described. The contact blocks
13 are supported on pedestal 170 by support posts 204 which are electrically insulated
14 from pedestal 170 by sleeved bushings 206. Support posts 204 are made of
graphite, as are locator posts 178, with bushings 206 being made of borsn nitride.
16 ~ The crucible is made of graphite in one embodiment, it being understood that the
17 ; crucible can be made of a variety of materials ranging in wettability from that
18 associated with carbon to that associat~d with quartz Ol boron nitride. The
19 insulation packs in one embodiment are made of Fiberform which is a trademark of
2~ Fiber Materials, Inc., of Biddeford, Maine. Generally, Fiberform is formed from a
21 colloidal suspension of graphite.
22 An upper insulation pack 210 which has been suitably apertured so as to
23 receive locator posts 178 and heater rods 198 surrounds crucible 144. This
24 insulation pack is provided with a central cut out portion 212 adapted to surround
the outer walls of cruciùle 144. Graphite nuts 214 bolted to locator posts 178
26 clamp the entire structure together. Double graphite radiation shields 216 spaced
27 by graphite washers 218 are mainteined in horizontal spaced adjacency on surface
.
- 21 -
I 1 ~9336
220 at the top of the upper insulation pack by raised portions 222 integrally formed
2 ~ in the upper insulstion pack. Sufficient space is provided between the shields to
3 accommodate the emerging ribbon. The radiation shields are made of solid
4 graphite as opposed to~iberform so that there is no particulate matter to fall into
the melt.
6 The manner of assembly of the apparatus of Fig. 13A is depicted in Figs.
7 13B -13E. Referring to Fig. 13B, pedestal 170 is first provided with locator posts
8 178, support posts 180-186 (support post 180 not shown in this Figure); and
9 contact block support posts 204. Contact blocks 200 are shown mounted on their
respective support posts. In Fig. 13C, lower insulation block 176 is positioned over
11 posts 178 and crucible 144 is mounted thereabove between contact blocks 200. As
12 shown in Fig. 13D, upper insulation block 210 is then placed over lower insulation
13 block 176 thereby surrounding crucible 144 as illustrated. Referring to Fig. 13E,
14 heater rods 198 are passed through the apertures in eontact block 200, through
upper insulation block 210, through the crucible channels and through the apertures
16 in the opposing contact block. Shields 216 are then bolted to the top of upper
17 insulation block 210 in spaced adjacency as illustrated.
18 Referring now to Fig. 13F, crucible 144 has its underside provided with
19 aforementioned apertures 192 for the general purpose of string introduction. A
countersunk aperture 230 is adapted to accommodate a thermocouple protection
21 tube 248 and a thermocouple housed therewithin, thermocouple leads being
22 illustrated at 2ds9. The thermocouple protection tube serves three functions. It
23 assures X, Y positioning of the crucible relative to the aforementioned pedestal. It
24 is provided with a short length of screw thread which acts to hold the crucible
down against supports 180 - 186. Finally, it houses the main furnace control
26 thermocouple.
~^-r~ rk
- 22 --
r -- ~
1 1 6~336
The bottom surface of crucible 144 is provided with radially extending
2 channels 232 which are adapted to receive locator pins 234 projecting above the
3 top surfaces 236 of support posts 180 and 186. When crucible 144 is positioned such
4 that its bottom surface contacts the top surfaces 236 of support posts 182 and 184,
S pins 234 project into channels 232. The purpose of this crucible support structure
is to assure angular positioning of the crucible and to accommodate linear thermal
expansion and contraction sf the crucible. It will be appreciated that the crucible
8 may expand by as much as 1% from room temperature to operating temperature,
g ; which in the case of silicon is 1420 C.
Referring now to Fig. 14, the string feed system is illustrated in which
11 ~ strings 242 are initially wound on spools 240. The spools can be biased or driven by
12 ~ motors or springs which oppose the feed direction to provide the requisite tension.
13 The spools are suspended from supports 244 which depend from a plate 246
14 mounted to the bottom flange of the furnace~ PedastaI 170 is also mounted to
plate 246, with string positior~ing tubes 252, thermocouple protection tube 248
16 extending through plate 246 and then through mating apertures 190' and 194' in
17 pedestal 170. Also butted to plate 246 is a small gas cup 254. This cup is provided
18 with argon independently of the main furnace thereby establishing a small plenum
19 chamber around all apertures in the furnace bottom. This assures that any flow of
gas into the furnace through these apertures will be inert gas as opposed to
21 surrounding air, flow bemg induced by a "chimney" effect in which hot less dense
22 gas in the furnace causes an upward flow~ ~
23 Ribbon pu~llng is accomplished by a ribbon drive unit 260 mounted on top of
24 the furnace in which ribbon 262 is positioned between opposing rollers of drive unit
260. Driven roller 264 is paired with an idler roller (not shown) and idler roller 266
26 is paired with a driven roller (not shown). Driving roller 264 results in the rotation
27 of gear 268 which drives an offset gear 270 to power the driven roller of the lower
28 set.
-- 23 -
.
~ ~ 69336
1 The string positioning tubes 252 of l~ig. 14 may be given the configuration
2 illustrated in Fig. 15 where their relationship to a string introduction tube 271 of a
3 crucible 272 is shown. A cylindrical hollow rod 273 is provided with inserts 274.
4 ` The inserts have a hollow central channel 276 and countersunk ends 278. Moreover,
string introduction tube 271 may include a countersunk portion 279 to facilitate
6 string introduction.
7 ` Referring now to Fig. 16, a portion 2B0 of a heater rod is shown inserted
8 into contact block 200. The heater rods in one embodiment, are turned from round
g graphite stock to provide a central rod portion 282 of decreased diameter and an
enlarged shaft end 284 having a circumferential contact rib 286. The shaft end
11 with contact rib engages a channel 288 in block 200 in a sliding spring fit. End 284
12 flnd rib 286 are divided into four quadrants by slots 290 in the horizontal and
13 vertical planes. This acts to create four integral graphite springs and provide
14 ~flexibility for the aforementioned sliding spring fit. I
'The rod thus includes a type of sliding connector to accommodate thermal
16 expansion along the length of the rod due to heating. This expansion may be on the
17 order of 2% such that if the rods were clamped to ths block, intolerable bowing
18 would occur. A peg 292 is positioned at each end of each heater rod to retain the
19 ~ heater rod in proper position to prevent the heater eod from "working out" of the
connector due to small osclllations in length during thermal cycling.
FURNACE OPERATION
21 In operation, a voltage is impressed across the heater rods in parallel via
22 water cooled current lead~ 156 and 158. Thus, current is fed in through current
23 lead 156 through heater contact block 200, divides through the parallel heater rods,
24 is coDected by the socond heater contact block 200, end leeves by current leed ISB.
- 24 -
1 1 ~9336
The furnace- is constructed to minimize the power consumed during
2 operation. To this end heater rods 198 are disposed inside channels 196 within the
3 crucible itself, to assure that all the power dissipated by the heater rod is received
4 directly by the crucible, rather than a portion of it radiating directly to the
insulation pack as would be the case for external heater elements. The result is
6 that during operation with molten silicon at 1430C, the total furnace power
7 consumption is 3500 watts. The attainment of this low power consumption is
8 further explained by noting that in the traditional furnace, the crucible is
g completely surrounded by a large graphite resistance heater. This heater must run
10 at a tempe.ature higher than the crucible operating temperture in order to
11 communicate heat to it. In additlon, the traditional large graphite resistance
12 heater has a larger surface area than the crucible it surrounds. These two factors
13 combine to yield a furnace power dissipation far in excess of the minimum
14 ' requirement of keeping the cruclble hot. By utilizing heater rods enclosed by the
15 I crucible, an approach may be made to the theoretically minimum power usage
16 associated with a given amount of surrounding insulation.
17 Further, the use of heater rods facilitates reshaping of the temperture
18 profile across the crucible so as to attain a nearly uniform profile. Such
19 modifications may be made by changing the overall resistance of one or more
20 elements with respect to the others with which it is in parallel. A larger diameter
21 heater rod will have lower resistance and will therefore dissipate more heat than
22 the others, in a parallel arrangement. The individual heater~ rods may also be
23 shaped along their lengths, larger lower resistance portions dissipating less heat
24 than an adjacent narrower portion because of the series arrangement. ~inally, the
25 heater rods may be repositioned, thereby effecting substantial changes in profile.
26 Once a satisfactory profile is obtained, it is highly reproducible, requiring only the
27 substitution of a similarly machined heater element for a spent element.
- 25 -
~ ~`~
~ ~ 6~33~
During start-up, silicon or other working material is melted in crucible 144
2 by supplying power to heater rods 198. The strings are then introduced into the
3 bottom of string:positioning tubes 252 and are fed up through these tubes and
-4 through apertures 142 in the bottom of crucible 144. The proper positioning of
tubes 252 relative to crucible 144 is assured by the crucible locating means
6 comprised of thermocouple protection tube 248 and crucible support posts 180 -186
7 described previously. Note that the thermocouple protection tube determines the
8~ X- Y positioning of the crucible at a point directly between the two string
9 introduction apertures 192, thereby minimizing any location errors of apertures 192
with respect to tubes 252. Easy upward motion of the strings is further insured by
11 the use of countersinks as described with reference to Pig. 15.
12 Other convenlent aspects of furnace use are the ability to withdraw
13 positioning tubes 252 from the furnace during a run. This becomes necessary in the
'I '
14 l event of a string jamming or as droplets freeze within the tube. In such an event,
the tube may be simply withdrawn from the bottom of the furnaçe, cleaned and
16 replaced. Similarly, the thermocouple may be withdrawn from tube 248 and
17 ~ replaced during a run with minimum elapsed time.
CAPILLARY RETENTION OF MELT
.. . .
WITHIN APERTURES IN THE CRUCIBLE BOTTOM
18 What is now described is the capillary retention of melt at an orifice in the
19 crucible bottom. In one embodiment, capillary retention may be effected by using
a downwardly projecting cylindrical tube at the orifice, in which the tube is
21 fabricated of the wetted, or partially wetted crucible material. Such an arrange-
22 ment is shown in Eig. 17 and is described in connection with string introduction.
23 The present discusslon rssumes no string present.
.
.
- 26 -
-
1 1 6933G
In principle, this tube functions by balancing the gravitional head of the
2melt with capillary forces associated with the bead of melt that forms at the
3bottom of the tube. For any given configuration and liquid, there will exist a
4maximum head that may be supported in this manner. With H the distance from
5the free melt surface to the bottom of the tube and r the outside radius of the6cylindrical tube, the maximum head supportable may be found by assuming that the
7drop beneath the projection is hemispherical with radius r.
8It follows that:
9 rHmax ~0 9 ~Eq. 3)
,
or
11 ~ p ~ r ~Eq. 4)
12Thus, if the melt level is such that the distance from the free surface to
13the bottom of the tube is less than HmaX no leakage will occur. However, if this
14distance should increase beyond HmaX, leakage will occur until the melt level is
15lowered and the distance from melt surface to the bottom of the tube is HmaX~ At
16 ~ this point, the leakage will stop.
17For any given material r~Hmax is a constant determined by material
18parameters alone. Por silicon, r D HmaX = Z~ -0.587cm2. EIowever, the
19designer may exerclse control over HmaX by changing the radius, r9 of the tube.
20For example, decreasing this radius will increase the msxlmum head that is
` 21supported.
, ;: '
:
. '
--27 -
1 ~ ~9336
.
. . i
CAPILLARY RETENTION O~ MELT DURING STRING INTRODUCTION
The string introduction technique illustr~ted in Fig. 7 may represent an
2 unstable condition if using a wetted crucible since when the silicon wets the
3 bottom of the crucible, its wetting area increases and therefore the head that may
4 be supported at the introduction aperture will decrease. This will lead to the
S ; possibility of melt leakage. When a sma~l downward cylindrical tube is utili2ed as
6 outlined above, the tube defines the limit on the wetting area of the liquid.
7 For the case of string introduction, HmaX is made significantly greater
8 than the maximum expected distance ~ between the melt surface and the tube end.
9 Typically, for à system where H = 1.3cm, a string introduction tube is used where r
= 0.16 cm, resulting in a maximum stable head HmaX of 3.67 cm.
11 The provision of a string within the tube has little effect on the amount of
12 head that may be supported. This may be understood as follows: Referring to ~ig.
13 1 17, string 294 is traveling upward as indicated by arrow Z96. The type of wetting
14 ~ which will exist at the bottom of string introduction tube 271 is illustrated by melt
droplet 300. This type of wetting of the string will actually increase the maximum
16 head allowed as more surface is available for meniscus attachment. Even
17 downward motion of the string will have little effect on the supportable head, as
18 the small radius of curvature around a small string guarantees that significant
19 , heads may be supported regardless of the angle of attachment to the string. This
type string introduction tube may be used both for vertical and angled growth.
21 Depending on the precise nature of the lnteraction between melt and
22 crucible material, the droplet at the bottom of the string introduction tube may
23 creep up the sides of the tube, resulting in a lower maximum supportable head due
2~ to a larger wetted area~ This process may be kept from continuing by the
embodiments of either Fig. 18 or Fig. 19. In Fig. 18 a small annular indentation
26 302 serves to stop the advancement of the melt up the projection because of
- 28 -
':
1 ~ 6g33~
meniscus attachment properties which cause the meniscus to attach at the
.
2 ~ discontinuity formed at the junction of a vertically ruMing surface met by an
3 inwardly projectlng surface. In Pig. 19 an annular ridge or flange 304 serves the
4 same purpose.
Referring to Fig. 20j a string introduction tube 306 may be provided with a
6 ridge or flange 308 on its un~erportion so as to accommodate an angular
7 orientation of the string with respect to the melt surface and so as to provide
8 ~ increased leakage protection. Additionally, the configurations of Figs. 18 and 19
g may also be used for angled growth. Note that the improvements embodied in Fig.
18, 19 and 20 are not restricted to string introduction ports, but may also apply to
11 suction draining and capillary overflow melt dumping described hereinaftèr. I
1,
PREVENTION OF MELT FREEZING
12 ~ The string introduction tubes of Figs. 18, 19 and 20, if left exposed, can
13 result in the freezing of the melt within the tubes. In order to maintain the
14 temperatures of the tubes above the melt temperature, a portion 310 of the; crucible 272 may be extended to completely surround the projection tube thereby
16 ~ forming a radiative cavity around the tube as illustrated. In one emodiment the
17 bottom of the crucible is counter-bored at the string introduction aperture and a
18 smaller string introduction tube is mounted concentrically in the counter-bore so
19 that the surrounding crucible walls are in spaced adjacency with respect to the
string introduction tube.
SUCTION DRAINING OF MELT
21 With respect to melt dumping and more particularly with respect to
22 Fig. 21, the melt as described hereinabove, may be dumped by providing the23 crucible with a downwardly projecting tube and by controlling the outflow from the
- 29 -
- ``
1 :1 6933~; ~
.
tube by ireezing and thawing the material at the tip. This assumes the "spigot" to
2 be long enough, i.e., extend sufficiently below the melt so that capillarity cannot
3 prevent the liquid from leaving the tip when the tip is thawed out.
4 However, as illustrated in Fig. 21, an alternative is to keep the molten
silicon in by capillarity in which rH~2~ g, and apply suction at the tip when it is
6 desired to withdraw silicon. In order to accomplish this, an exit conduit 312
7 communicates between the crucible and a dump receptacle 314 which may be
8 evacuated through a conduit 316 by vacuum apparatus (not shown). Suction is
9 applied to the receptacle whenever it is desired to withdraw melt from the
crucible. This is accomplished in a semi-batch or nearly continuous mode as
11 desired. The result is that solidified rejected silicon is deposited as illustrated at
12 318, with a relatively long length of conduit 312 being provided. Recepticle 314
13 may be made of graphite and it may be severable as illustrated by the flange/bolt
14 attachment system diagrammatically illustrated at 320.
:1,
CAPI~LARY OVERFLOW MELT DUMPING
,
Referring now to Fig. 22, crucible lg4 is provided with a melt dumping
16 tube 311. Since string introduotion tubes 330 have a predetermined maximum head
17 that will be supported, any head above this maximum will cause a slow leakage
18 until a steady state level is attained. If tube 311 is either longer or has a larger
- 19 diameter as compared with tubes 330, it acts essentially in the rnanner of an
overflow valve.
21 The maxlmum head, HmaX~ supported by a tube is determined by its outside
22 diameter. Thus, two tubes of equal outside diameters support the same maximum
23 amount of head. By makmg tube 311 longer than string introductlon tubes 330?
24 when the melt level is increased, the melt level at which leakage occurs at tube
311 will be reached before the level at which melt would leak from tubes 330. Of
~`
- 30 -
. .
-
1 1 69336
course tubes 330 and 311 may be of different diameters and thus support different
2 maximum melt heads. Thus the relative diameters and lengths may be adjusted so
3 that leakage occurs first at tube 311. This leakage will continue until the melt
4 level is reduced sufficiently to allow the melt dumping tube 311 to seal by capillarity.
6 Capillary overflow melt dumping is capable of very precise melt level
7 control, since the melt rises past the point where H = HmaX leakage will begin and
8 continue until H = HmaX at which point It wlll stop. Thus, the melt level is
9 automatically controlled in a manner requiring no melt height measurement or
external feedback. Further, it should be noted that the melt will overflow in a well
11 controlled manner drop by drop. This may be understood by noting that for H12 slightly greater than HmaX, the capillary forces at the bottom of the melt dumping
13 tube are almost, but not quite, balancing the forces from the gravitational head.
14 Thus, the melt is being forced to flow out by only a small pressure differential and
15 ~ will therefore flow out in a well controlled manner. In one embodiment, the melt
16 replènishment rate is made to exceed the growth rate by, for instance, 20%. Thus,
17 20% of the melt will be continually dumped thereby leading to considerable melt
18 E)urification.
19 Note that any suitable ¢ombination of tube length and radius may be used.
In practlce, however, the radius oE the melt dumping tube will roughly determine21 the size of the effluent drops. Relatively f~ne drops are desired so that they do not
22 clog up area 336. This establishes an upper limit on tube radius. In one23 embodiment r = 0.2 cnn, rnd H = 2.98 cm.
`'; ' '
'
- 31 -
~ 1 ~9336
GROWTH AT AN ANGLE TO THE FREE MELT SURFACE
THEORETICAL CONSIDERATIONS
It can be demonstrated that when an edge stabilized ribbon is pulled from
2 the melt at an angle, the ribbon will assume a trough shape. It has been found that
"flat" ribbon can be grown by providing means which permit the angular orientation
of the meniscus surface at its lower attachment point to be something other than
its normal tangential angle, e.g., tangential to the melt surface. The ability to
6 grow flat ribbon may be explained by the effects of geometrical and thermal
7 constraints on the growth of a thin sheet or ribbon.
8 The geometrical constraints arise primarily from the action of surface
g ` tension on the meniscus and the allowable angles of attachment of the meniscus to
10 the melt ribbon, etc. The thermal constraints stem from material parameters such
11 as heat of fusion and thermal conductivity, as well as from the heat withdrawal
12 i mechanisms present, e.g., radiative or conductive. With respect to these con-
13 straints and their interactions, a more detailed discussion of the geometrical
14 constraints than is described in connection with Figs. 5A- SD is now described.
15 Although this discussion applies to growth at any angle to the melt, it will be
16 illustrated by reference to a cross-section through a ribbon grown vertically as
17 shown in Fig. 23. As noted previously, the meniscus must join the melt surface
18 smoothly as illustrated by reference character 400, so as to join the melt surface
19 tangentially; that i9, the meniscus must have zero slope at a free horizontal melt
20 surface. Thus, the angle of attachment of the meniscus to the mel~ is zero. This
21 may be understood physically by noting that the meniscus is in fact composed OI
22 nothing but melt. Therefore, there is no physical boundary that would proYide a
23 basis for discontinuity in slope of the liquid surface. However, whenever the
24 meniscus attaches to a solid body such as a solidified ribbon, a non-zero angle of
25 attachment will, in general, be formed. The meniscus must meet the growing
-32-
:` , .
I 1 6933~;
crystal at a speciflc angle ~" the angle determined by the properties of the
2 material being grown. For silicon this angle is approximately 11. The line of
3 ~' attachment of the meniscus to the growing crystal is referred to herein as the
4 interface edge depicted by reference character 402.
In computing the precise shape of the meniscus surface, these conditions of
6 meniscus slope constitute the boundary conditions. Between the ends, meniscus
7 shape is determined by the influence of capillary forces as embodied in Laplace's
8 equation:
bp~gh=~(R + R ) (Eq. S)
g where~p is the pressure drop being supported across the liquid gas interface by
10 I capillarity.~O, g, h, and ~ are defined hereinbefore, and Rl and R2 are the
11 principal radii of curvature of the meniscus surface. Whenever a "flat" ribbon is
12 ~ being grown, either vertically or at an angle, one of the principal radii, R2, is
13 1 infinite. In this case equation 5 reduces to the equation 1, where R1 = R. Since it
14 is generally desired to grow flat ribbon, the reduced form equation 1 is generally
15 applicable. However, in order to fully understand angled growth, use is made of
16 ' the general equation 5. As may be seen from both equations 1 and 5, the local radii
~7 of curvature of the meniscus surface are a function of the height h of the point in
18 question above the free melt surface. Thus, in simple vertical growth, the radius
19 of curvature of the surface of the meniscus is continually decreasing as one moves
2~ up the meniscus from the melt surface to the growth interface.
21 In order to clarify the following discussion several simplifications are made
22 with no loss of generality: The angle of attachment of the meniscus to the ribbon,
23 or interface attachment angle ~b~" is taken as 2ero degrees; and the meniscus24 surface is taken to have constant curvature, thus ignoring the dependence of
25 surface curvature on height h.
-3
~ ~ 693.~3
.
`: I
Referring to Fig. 24, with respect to growth at an angle,6~a, to the melt
2 ~ surface, this figure presents a cross-sectional view of the growth of flat ribbon
3 from the melt surface with the aforementioned assumptions of zero interface
4 attachment angle and a constant radius of curvature for the men}scus surfaces. As
: -
previously noted, interface edges 402 constitute the line of attachment of the
6 meniscus to the growing ribbon. As may be seen from ~ig. 24, a very large
7 distance L exists between tbese interface edges as measured along the ribbon
8 length Referring back to Pig. 23, symmetry dictates that this distance be zero for
9 the simple case of vertical growth of ribbon from the melt surface. However, in
the angled growth of Fig. 24 a separation of the two interface edges exists with a
11 resultant vast enlargement of the growth interface. It is this enlarged growth
1~ interface and its orlentation substantially parallel to the rihbon which brings about
13 the principal advantages of angled growth, to wit, higher speed due to a l~rger area
14 for heat removal, lower thermal stresses due to the extraction of heat perpen-
dieul~r to, rather than parallel to the pulling direction, and the greater tendency of
. , ,
16 any unwanted crystal structure such as grain boundaries, to grow only a short
17 distance due to propagation toward the ribbon surface as opposed to propagation
18 along the ribbon length.
19 As illustrated in Fig. 24, the meniscus heights on the two sides of a ribbon
grown al: an angle to the melt will be different. The term meniscus height is more
21 rigorously defined as the vertical distance between the free melt surface and the
22 interface edge on the corresponding side of the ribbon. It is with a qualitative;
23 understanding of the origin of this difference in meniscus heights that angled
24 growth can be understood. ~ ~
Fig. 25, serves to illustrate the critical parameters of the interface of Fig.
26 24. As CaD be seen from~ Fig. ~25, the right and left meniscus surfaces are segments
27 of a circular arc. If vertical ribbon were being grown, the rlbbon edges would be
.
:
- 34 -
~ 3 6933G
defined by the dotted lines 410, moved closer together. In the case of the right2 hand meniscus surface of Fig~ 24, the point of tangency to the ribbon surface is
3 moved further out on the arc, from point 412 for vertical growth to point 414 for
4 angled growth. In the case of the left hand meniscus surface, the point of tangency
S to the ribbon surface is closer to the melt. This is illustrated by the difference in
6 position between point 416 for vertical growth and point 418 for angled growth.
7 Thus, the right hand meniscus height, MR, is increased over the corresponding
height for vertical growth, while the left hand meniscus height, ML is decreased.
9 The result is an enlarged growth interface and a large distance between interface
edges as measured along the ribbon.
11 An important problem arises, however, with respect to this growth12 conIiguration. Up to this point, only the influence of geometrical constraints has
13 been considered. Heat transfer and thermal constraints must be considered as
14 well. Note, that the entire growth interface is by definition at the same
temperature, the melting point of the material in question. This large isothermal
16 surface imposes a severe restriction on the temperature of the surface of the
17 ribbon near the interface as it must be at or near the melting point of the material
18 ~ over a large area.
19 From Fig. 25 a specific distance between interface edges is defined for
each angle of growth to the melt surface. However, the above discussion makes it21 clear that the thermal constraints will not necessarib be satisfied by this growth
22 geometry. In general, the thermal constraints will favor a smaller distance
23 between interface edges than that dictated by geometric constraints alone.
24 However both geometric and thermal constraints must simultaneously be satisfied.
In fact the ribbon responds by growing in a trough shaped rnanner as illustrated in
26 Fig. 26.
.
33~
The simultaneous satisfaction of both geometric and thermal constraints by
2 trough shaped growth may be understood from the complete form of Laplace's
3 equation, Eq. 5. Referring again to Fig. 26, the trough shaped ribbon 420, and
4 consequently curved meniscus 422 gives rise to a finite curvature R2 of the ribbon.
This is in direct contrast to the growth of flat ribbon where R2 = infinity.
6 The radii of curvature R1R and R1L of the meniscus surfaces as measured
7 in vertical cross-section is positive on both sides of the ribbon. That is, the
8 meniscus surface is concave in a vertical section on both the left or upper, and
9 right or lower, ribbon surfaces. However, R2, the principal radius of curvature due
to the ribbon troughing, is of Opposiee sign on the two sides of the ribbon. The left
11 side is concave, having a positive curvature, while the right side is convex, having a
12 negative curvature.
13 ` From Eq. 5, (i/Rl + 1/R2) must remain unchanged for a given height h. On
14 the left side of the ribbon, a finite and positive R2 implies that 1/R1 must be
smaller if the sum (l/Rl ~ l/R2) is to remain constant. Thus, RlL must take on a16 larger positive value than it had~ when R2 was infinite in the flat ribbon case. On
. . .
17 the right side, a finite and negative R2 implies that 1/R1 must take on a larger
18 value for the sum ~I/RI ~ l/R2) to remain constant. Thus RlR must take on a
19 smaller positive value than it had when R2 was infinite.
ao In summary, in the transition from growth of flat ribbon to growth of
21 trough-shaped ribbon, the geometrical constraint of capillarity requires that on the
22 upper surface the radius of curvature R1 increase, while on the lower surface it
23 ~ must decrease.
24 The effect of this change in radii may be seen from ~ig. 27 which shows
vertical cross-sections of angled growth through the middle of a ribbon 430 which
26 is growing flat (dotted lines) and though the middle of R ribbon 432 growing in a
27 trough shape. As illustrated~the trough-shaped ribbon exhibits a larger upper radius
.
'
- 36 -
-
1 1 6~33~
of curvature 434 than the flat ribbon curvature 436, and a smaller lower radills of
2 curvature 438 than the llat ribbon curvature 440. The result is a decreased
3 distance between interface edges as illustrated by arrows 442 and 444. Thus the
4 growth of trough-shaped ribbon 432 is the natural response of the system in an
effort to satisfy both geometric and thermal constraints.
FLAT RIBBON GROWTH
6 It has been found that by permitting the angular orientation of the
7 meniscus surface at its lower attachment point to change from the normal
8 ~ tangential angle, e.g., tangential to the melt surface, the ribbon will grow flat.
9 Fig. 28A is a vertical cross-section through a ribbon grown at an angle to
the melt surface. In Fig. 28~, a device 450 has~been inserted which allows for awide range of meniscus attachment angles at lower attachment point 451 of
12 ' meniscus surface 452. The result of the steep angle of attachment of the right side
13 , meniscus is a substantial reduction in the distance L between interface edges. This
14 ~ is the desired condition for flat ribbon growth and is achieved by relaxation of the
;l ,
~~ constraint on the angle of lower meniscus attachment, which allows for flexibility
16 in meeting the thermal constraints while growing flat ribbon. While flat ribbon
17 ' growth is described for ribbons pulled at an angle, permitting variable attachment
18 angles is usefui to prevent bowing in the nèar vertical growth situation. The
19 - following sections:deal in whole or in part with practical embodiments of device
2û 450 of Fig. 28B.
WETTED STABILIZATION MEMBERS
21 As discussed previously, angled ~growth may ~e stabilized by providing
.
22 ~ means which allow for a wide range of meniscus attachment angles. One means of
23 ~ providing or such a range of menlscus attachment angles is to use a wetted
':
':
:
- 37 -
. . . , :
1 1 8933B
stabilization member or boat, illustrated at 330 in Fig. 29. Stabilization member
~ 330 is carried in crucible 144 at the top surface of the melt, here illustrated by
3 reference character 332. In the embodiment illustrated, boat 330 is U-shaped, with
4 a straight edge 331 running in a horizontal plane parallel to the plane of ribbon 166.
As can be seen, ribbon 166 is drawn at an angle to the vertical and a meniscus 333
6 is formed adjacent to edge 331. In Fig. 29, meniscus 333 attaches to edge 331 and
7 ~ to a line 334 on the ribbon at which the ribbon freezes.
8 It will be appreciated that in order for meniscus 333 to form with the
g appropriate curvature, there is finite spacing between the plane occupied by ribbon
166 and straight line 331. Boat 330 or similar structure is wetted by the melt and
11 one of its functions is to assure that there is a region opposite the lower meniscus
12 attachment point where there is no melt or liquid. In fact, the edge of the crucible
13 can serve as a line of attachment.
14 i Figures 30A and 30B illustrate the operation of stabilization member 330.
i
Fig. 30A is a vertical cross-sectlon of the growing ribbon of Fig. 29, and Fig. 30B is
16 an enlarged view of the indieated area of Fig. 30A.
17 ' As seen from Fig. 30A, at the point 338 of attachment of the lower side
18, meniscus surface 340 to the stabilization member, the angle of surface 340 is
19 1 approximate]y 60 with respect to the horizontal. This is in contrast to the case of
2~ i growth directly from the free melt surface wherein this angle would be zero at the
21 melt surface. ~ ~
22 Fig. 30B illustrates one means by which the angle of the meniscus surface
23 is allowed to vary with respect to the horizontal. As shown in this figure edge 331
24 of ooat 330 is inherently slightly rounded. Any pair of materials such as a silicon
25 melt and a graphite stabilization member determine an angle of attachment of the
26 melt to the stabllization member, typically 30 for the ~materials mentioned.
27 However the angle of the meniscus surface with respect to the horizontal can be
-38-
.
~ 1 fi~33~
made to change because the meniscus is free to attach to the corner of the
2 stabilization member anywhere along its curved edge. Thus the meniscus surface
3 may attach to edge 331 over a range of angles.
ANGLED GROWTH WITH A VERTICALLY PULLED RIBBON
.
4 Fig 31 presents an alternative embodiment for the growth of ribbon at an
angle to the melt surface. In thls embodiment a pump 400 raises molten material
6 from melt 402 in a crucible 404 to the top 406 of surface 408 of inclined ramp 409.
7 ~ The molten material at 406 travels down surface 408, some of it crystallizing into
8 a growing ribbon 410 and most of it returning to melt 402.
9 In this embodiment, each string 412 passes up through a corresponding
integral standpipe 414 and then passes through a capillary retention tube 416 at
11 aperture 418 in ramp 409, finally emerging through the flowing melt here
12 illustrated at 420. It will be appreciated that the flow down the plane is
13 ~ maintained at a rate significantly higher than that required by the growing cr~stal,
14 with the excess flow returning to the melt where it is recirculated through the
system.
16 Growth from an inclined surface has several advantages over angled growth
17 directly from the melt surface. First, excellent circulation of the melt exists
,
18 under the growth interface because of the flowing melt. This results in effective
` ~ 19 removal of impurities rejected at the growth interface. Secondly, all the
, ~ ~
advantages of angled growth are combined with the manufacturing and handling
.~
21 convenience of a vertically pulled ribbon.
22 An alternative embodiment is shown in Fig. 32, In which a ramp 421 is
23 formed integral to a crucible 422. In this configuration, string 423 enters through a
24 deep counterbore 424 and then passes through a capillary retention tube 426, the
interior of which communicates with the top surface 428 of ramp 421. The string
~ :
:~ :
--39 -
~ J ~933~
then passes through flowing melt 430, pumped at 432 from melt 434, with ribbon
2 436 formed in a vertical direction as illustrated. This configuration offers the
3 potential for enhanced mechanical stability and greater thermal uniformity as
4 compared with the configuration of Fig. 31.
s The flow down the inclined surface is primarily controlled by gravity and
6 fluid viscosity. It should be noted that the surface of the liquid flowing down ramp
7 421 is a free surface, and hence is at the ambient pressure. In fact, in
8 understanding the mechanics of growth of vertical ribbon from an inclined surface,
g the inclined surface technique may be regarded as a variation on growth of ribbon
at an angle to the melt. The situation is somewhat altered since the motion of the
11 fluid down the ramp may have an effect on the growth and since the free melt
12 ~ surface is no longer perpendicular to the gravity vector. However, the funda-
13 I mental qualitative arguments developed previously in reference to growth of ribbon
14 i at an angle to the melt surface apply to this growth configuration. Inspection of
Figs. 31 and 32 show that the ribbon Is growing with an elongated interface, much
~6 as in Fig. 24. In fact, growth from an inclined surface may be viewed as roughly
17 , parallel to angled growth of ribbon from a horizontal melt surface, all rotated
!
18 through an angle such that the ribbon is growing vertically, e.g., angle~i (Fig. 31) =
19 ~9 a (Fig- 24)-
' In the case of growth at an angle to the horizontal melt surface it was
21 noted that often the ribbon will naturally grow in a trough shape. The solution to
22 this problem illustrated schematically in Fig. 28B is to allow the angle of contact
23 of the meniscus to the melt surface to assume a value oth~r than tangential to the
24 melt surface. Vertical growth of ribbon from an inclined surface will engender the
same problem of ribbon troùghing. The solution to the problem in this case follows
26 theprincipleofFig. ~8R.
.
- 40 -
i_~
1 ~ 69336
Fig. 33 presents a cross-sectional view of an embodiment of growth at an
2 angle to the melt surface which allows for the growth of flat ribbon. In this
3 embodiment ramp 421 is configured with an abrupt discontinuity at 440 in top
4 surface 428 such that the inclination of the surface changes abruptly alon~ a
horizontal line parallel to the plane of ribbon 436. Over a large range of flow
6 parameters, flow 430 will conform very closely to this discontinuity resulting in a
7 sharp change of ~dlrection of the melt surface. As shown in detail in ~ig. 34, a
8 meniscus 442 may attach to melt surface 444 at any location around the corner
9 ; formed at 440 by the flow discontinuity. Thus, while the meniscus surface must
always meet the local melt surîace tangentially, as it changes position around the
11 corner formed by the flow discontinuity, the angle ~ of the meniscus surface at
12 its point of attachment will change as measured with respect to a reference line
13 such as horizontal 446. Thus, the flow discontinuity of Fig. 34 performs in much
14 ~' the same manner as the capillary stabillzation boat of Fig. 29.
1 It will be appreciated that other geometries are possible which utilize a
16 , flow discontinuity to accomplish edge stabilization. The essential requirement is
17 ', to create a free melt surface with an abrupt change of direction. Fig. 35
18 illustrates an embodiment where melt is pumped up a central flow channel 460 to
19 , the apex in a tent-shaped ramp 462 and then flows down the sloped sides of 464 and
466 of ramp 462. The abrupt change in flow direction in the region marked by
21 reference character 468 acts to establish a range of attachment angles and allows
22 for a larger change in the angle of the meniscus surface with respect to the
23 horizontal than does the configuration of Fig. 33. Finally, if only a srnall allowable
24 range of angular variation of the meniscus surface with respect ~to the horizontal is
found to be sufficient to stabilize the growth, it may be effected by the
26 configuration illustrated in Fig. 36. In this conflguration, flow 470 conforms to a
27 small hollow channel 472 machined into the top surface 474 of ramp 476~ thus
--41 --
I 3 6933~
providing for a small range of angle selection by the meniscus surface at its point
2 of attachment to the melt.
3 It will be appreciated that in certain flow configurations an upwardly
4 projecting ridge can replace channel 472 of the Fig. 36 embodiment so as to
provide a flow discontinuity at which a range of attachment an~les are possible.6 In summary, in the inclined surface embodiments a flow discontinuity is
7 provided for permitting a range of attachment angles so that flat ribbon may be
8 grown. The flow discontinuity is established in a horizontal direction in a plane
9 parallel to that of the growing ribbon and may be generated by various methods for
1~ producing corners, apexes, troughs and ridges.
11 All the conflguretions just described require a pump to raise the melt to
12 the top of the ramp. As illustrated in Fig. 37, electromagnetic pumping may be
13 utilized in which current is passed thiough a tube 500 in the direction shown by
14 arrow 502. Here the tube is immersed in ~melt 504. A crossed magnetic field as
' illustrated by X's 506 causes a pumping action to result in pump flow 508 in the
16 tube upon application of current.
17 Having above indicated a preferred embodiment of the present invention, it
18 , will occur to those skilled in the art that modifi~ations and alternatives can be
19 practiced within the spirit of the invention. It is accordingly intended to define the
scope of the invention only as indicated in the following claims.
`
. :
: `
,
:
- 42 -
.
