Patent Application
20060005761
The present invention generally relates to producing semiconductor grade single crystal silicon that can be used in the manufacture of electronic components and the like. More particularly, the present invention relates to processes for producing silicon ingots by controlling the shape of the melt-solid interface in a selected thermal environment.
Single crystal silicon, which is the starting material for most processes for the fabrication of semiconductor electronic components, is commonly prepared by the so-called Czochralski (“Cz”) method. In this method, polycrystalline silicon (“polysilicon”) is charged to a crucible and melted, a seed crystal is brought into contact with the molten silicon and a single crystal is grown by slow extraction. Silicon crystals grown from a melt may be grown with an excess of one or the other type of intrinsic point defect, either crystal lattice vacancies (“V”) or silicon self-interstitials (“I”) and may have regions of both or of neither. It has been suggested that the dominant point defect type is determined near solidification and, if the dominant-point defect concentrations reach a level of critical supersaturation in the system and the mobility of the point defects is sufficiently high, a reaction, or an agglomeration event, will likely occur. Agglomerated intrinsic point defects in silicon can severely impact the yield potential of the material in the production of complex and highly integrated circuits.
Those skilled in the art are familiar with the ratio of the pull rate v of the crystal to an axial thermal gradient G being indicative of the type of intrinsic point defect likely to occur in the growing crystal. For example, when the pull rate is high lattice vacancies are the dominant point defects. Alternatively, when pull rate is low silicon self-interstitials are the dominant point defects. Thus, during a dynamic growth process (i.e., where v/G may vary as a function of the radius and/or axial length of the crystal) point defects within the growing silicon crystal can change from being vacancy dominated to interstitial dominated, or vice versa. Moreover, there is an identifiable vacancies/self-interstitials (V/I) transition associated with such a change.
In accordance with the present invention, it has been discovered that substantially defect-free single silicon crystals can be produced by closely controlling the shape of the molten silicon/silicon crystal (melt/crystal or melt-solid) interface. In particular, by controlling the melt-solid interface shape according to a target melt-solid shape profile as a function of axial length, a region substantially free of agglomerated defects may be produced. Further, by selecting a smooth seed lift profile that is determined using the V/I transition pull rate values, perfect silicon material may be produced over substantially all of the crystal body length. The target interface shape is unique to the crystal hotzone design and position along the axial length of the ingot.
In accordance with one aspect of the invention, a method for use in combination with a crystal growing apparatus for growing a monocrystalline ingot according to the Czochralski process is provided. The crystal growing apparatus includes a heated crucible containing a semiconductor melt from which the ingot is pulled. The ingot is grown on a seed crystal pulled from the melt. The method includes determining a set point for an operating parameter of the crystal growing apparatus as a function of a length of the ingot during pulling. The set point is specified by a pre-defined melt-solid interface shape profile that represents a desired shape of a melt-solid interface between the melt and the ingot during pulling as a function of the length of the ingot and an operating condition affecting the melt. The method also includes adjusting the operating condition of the crystal growing apparatus according to the determined operating parameter set point to control the shape of the melt-solid interface while the ingot is being pulled from the melt.
In accordance with another aspect of the invention, a method for defining a melt-solid interface shape profile is provided. The profile is used in combination with a crystal growing apparatus for growing a monocrystalline ingot according to the Czochralski process. The crystal growing apparatus includes a heated crucible containing a semiconductor melt from which the ingot is pulled. The ingot is grown on a seed crystal pulled from the melt. The melt-solid interface shape profile represents a desired shape of a melt-solid interface between the melt and the ingot during pulling as a function of the length of the ingot. The method includes selecting a plurality of axial positions along the length of a model ingot and defining a plurality of melt-solid interface shapes for each of the identified axial positions. The method also includes determining a thermal model of a hotzone of the crystal growing apparatus for each of the axial positions and each of the melt-solid interface shapes. The method also includes defining a velocity profile representative of a ramped pull rate. The method also includes determining point defects concentration fields in all regions of interest of the model ingot using a point defect simulator. The point defect model is responsive to the velocity profile and the thermal model for identifying a V/I transition for each of the plurality of defined melt solid interface shapes for each of the plurality of identified axial positions. The method further includes identifying a target melt-solid interface shape corresponding to a substantially flat V/I transition for each of the plurality of identified axial positions.
In accordance with yet another aspect of the invention, a system for use in combination with a crystal growing apparatus for growing a monocrystalline ingot according to the Czochralski process is provided. The crystal growing apparatus includes a heated crucible containing a semiconductor melt from which the ingot is pulled. The ingot is grown on a seed crystal pulled from the melt. The apparatus includes a memory storing a pre-defined melt-solid interface profile. The melt-solid interface profile represents a desired shape of a melt solid interface between the melt and the ingot during pulling as a function of a length of the ingot and an operating condition affecting the melt. A processor is responsive to the pre-defined melt-solid interface profile for determining a set point for an operating parameter of the crystal growing apparatus as a function of the length of the ingot during pulling. A controller is responsive to the determined operating parameter set point to adjust the operation condition of the crystal growing apparatus according to the determined operating parameter set point to control the shape of the melt-solid interface while the ingot is being pulled from the melt.
Alternatively, the invention may comprise various other methods and apparatuses.
Other features will be in part apparent and in part pointed out hereinafter.
Corresponding reference characters indicate corresponding parts throughout the several views of the drawings.
Single crystal silicon, which is the starting material for most processes for the fabrication of semiconductor electronic components, is commonly prepared by the so-called Czochralski (“Cz”) method. In this method, polycrystalline silicon (“polysilicon”) is charged to a crucible and melted, a seed crystal is brought into contact with the molten silicon and a single crystal is grown by slow extraction. The present invention proposes a method and system for producing silicon crystal ingots suitable for use in device manufacturing. In some embodiments, the method and system of the present invention may be used to produce a silicon crystal ingot, a substantial portion or all of which is substantially free of agglomerated intrinsic point defects. That is, silicon ingots, a substantial portion or all of which is has a density of detectable defects of less than about 1×104 defects/cm3, less than about 5×103 defects/cm3, less than about 1×103 defects/cm3 or even no detectable agglomerated intrinsic point defects). In other embodiments, the present invention may produce silicon ingots having substantially no agglomerated defects that are larger than about 60 nm in diameter.) More specifically, the present invention can control the shape of the molten silicon and crystal (melt-solid) interface during crystal growth to limit and/or suppress the formation of agglomerated intrinsic point defects. The melt-solid interface shape may be concave, convex in shape relative to the crystal, and may even be both concave and convex in shape relative to the crystal (e.g., a “sea gull wing” shape). It has been discovered that the melt-solid interface shape is an important parameter for controlling defects during crystal growth.
The shape of melt-solid interface can be affected significantly by melt convection. Convection refers to the process of heat transfer in a liquid by the movement of the liquid itself. In general, there are two types of convection: natural convention and forced convection. Natural convection is when the movement of the melt is due, for example, solely to the presence of heaters giving rise to density gradients. Forced convection is when the movement of the melt is due to an external agent such as magnetic fields in the crucible, or rotational speed and or direction of the ingot and/or crucible. Accordingly, the shape of the melt-solid interface may be controlled by controlling one or more of at least four (4) operational parameters, which can be used alone, or in combination, to achieve the desired shape of melt-solid interface shape. These parameters include (1) heat flux, which determines the melt boundary temperature field, generated by either a change in lower heater power or by a change in lower insulation level or efficiency (e.g., adjust bottom heater output); (2) magnetic field strength; (3) rotation of the crucible; and (4) rotation of the silicon crystal.
For instance, because magnetic fields can affect the flow pattern in electrically conducting fluids such as silicon melt, a magnet (e.g., Cusp type, Vertical type and Horizontal type) can be used to affect melt convection, and, thus change the temperature distribution in the melt, which, in turn, can affect on the melt-solid interface shape. Similarly, crucible and crystal rotation influences the flow pattern in of the, and, thus, affect the temperature distribution in the melt, which, again, affects the melt-solid interface shape. Conventionally, the crucible and crystal are rotated in opposite directions when growing a silicon single crystal. This process is referred to herein as counter-rotation. Although controlling the difference between the rotational rate of the crucible and the rotational rate crystal in a counter rotation process may be used to affect on the shape the melt-solid interface, it has been discovered by the inventors that a so-called iso-rotation process provides for greater control of the melt-solid interface shape. In an iso-rotation process, both the crucible and crystal (i.e., seed) are rotated in the same direction and the difference between the rotational rate of the crucible and the rotational rate crystal may be used to affect on the shape the melt-solid interface.
Silicon crystals grown from a melt may be grown with an excess of one or the other type of intrinsic point defect, either crystal lattice vacancies (“V”) or silicon self-interstitials (“I”) and may have regions of both or of neither. According to one or more embodiments of the present invention, the crystal/melt interface shape may be controlled during the growth of the crystal to control the initial distribution of point defects at solidification front at several degrees or tens of degrees (around 100 degrees) from the melt-solid interface. It has been suggested that the dominant point defect type is determined near solidification and, if the dominant-point defect concentrations reach a level of critical supersaturation in the system and the mobility of the point defects is sufficiently high, a reaction, or an agglomeration event, will likely occur. Agglomerated intrinsic point defects in silicon can severely impact the yield potential of the material in the production of complex and highly integrated circuits. By controlling the shape of the melt-solid interface, the agglomeration reaction may be greatly reduced and even avoided entirely, producing silicon substantially free of agglomerated intrinsic point defects.
Those skilled in the art are familiar with the ratio of the pull rate v of the crystal to an axial thermal gradient G being indicative of the type of intrinsic point defect likely to occur in the growing crystal. For example, when the pull rate is high lattice vacancies are the dominant point defects. Alternatively, when pull rate is low silicon self-interstitials are the dominant point defects. Thus, during a dynamic growth process (i.e., where v/G may vary as a function of the radius and/or axial length of the crystal) point defects within the growing silicon crystal can change from being vacancy dominated to interstitial dominated, or vice versa. Moreover, there is an identifiable vacancies/self-interstitials (V/I) transition associated with such a change. It has been discovered that a large-defect free region exists between agglomerated vacancy defects and agglomerated interstitial type defects. Significantly, the V/I transition occurs within this defect-free region. That is, this defect free region corresponds to the transition region from an excess vacancy dominant region to an excess interstitial dominant region. The defect-free region may be vacancy dominated and/or interstitial dominated material. The defect-free region does not contain critical excess point defects to form any defects and generally includes the V/I transition.
Moreover, it has been discovered that when the identified V/I transition has a preferred shape, or profile, the crystal is substantially free of agglomerated defects at this transition. For instance, a substantially flat V/I transition perpendicular to the pull axis under dynamic growth simulations corresponds to a portion of the crystal substantially free of agglomerated defects.
According to the present invention, a target interface shape is determined for a particular hotzone at various positions along the length of the crystal to determine the target interface shape profile for the given hotzone. By suppressing or otherwise controlling the agglomeration reactions that produce agglomerated defects, rather than simply limiting the rate at which such defects form, or attempting to annihilate some of the defects after they have formed, a method acting to suppress or control agglomeration reactions yields a silicon substrate that is substantially free of undesirable amounts or sizes of agglomerated intrinsic point defects. Such a method also affords single crystal silicon wafers having epi-like yield potential, in terms of the number of integrated circuits obtained per wafer, without having the high costs associated with an epitaxial process.
According to one embodiment of the invention, the melt-solid interface shape is controlled within a certain range or percentage of a height deviation ratio (HDR). For example, the molten silicon is controlled for a 200 mm crystal such that the height deviation ratio (HDR) ratio between crystal center and edge is about plus or minus about 11%, preferably about plus or minus about 9%, more preferably plus or minus about 7%, and most preferably plus or minus about 5%. Where the height deviation ratio is determined from the following equation:
HDR=[Hc−He]/Radius×100, (1);
where Hc is height of crystal center from melt surface and He is height of crystal edge from melt surface. For crystals having a diameter other than 200 mm, the maximum height deviation ratio is gradually decreased by a slope of 0.06 by the crystal radius. In another embodiment, an acceptable range or operating window is determined as described in APPENDIX A.
Referring now to
During both heating and crystal pulling, a crucible drive unit (i.e., motor) 45 rotates crucible 19 (e.g., in the clockwise direction). The crucible drive unit 45 may also raise and/or lower the crucible 19 as desired during the growth process. For example, crucible drive unit 45 raises crucible 19 as the melt 29 is depleted to maintain its level, indicated by reference character 47, at a desired height. A crystal drive unit 49 similarly rotates the cable 37 in a direction opposite the direction in which crucible drive unit 45 rotates crucible 19 (e.g., in the counter-clockwise direction) or in the same direction as the crucible drive (e.g., iso-rotation). In embodiments using iso-rotation, the crystal drive unit 49 may rotate the cable 37 in the same direction in which crucible drive unit 45 rotates crucible 19 (e.g., in the clockwise direction) In addition, the crystal drive unit 49 raises and lowers crystal 31 relative to the melt level 47 as desired during the growth process.
In one embodiment, crystal growth apparatus 13 preheats the seed crystal 35 by lowering it nearly into contact with the molten silicon of melt 29 contained by crucible 19. After preheating, crystal drive unit 49 continues to lower seed crystal 35 via cable 37 into contact with melt 29 at its melt level 47. As seed crystal 35 melts, crystal drive unit 49 slowly withdraws, or pulls, it from the melt 29. Seed crystal 35 draws silicon from melt 29 to produce a growth of silicon single crystal 31 as it is withdrawn. Crystal drive unit 49 rotates crystal 31 at a reference rate as it pulls crystal 31 from melt 29. Crucible drive unit 45 similarly rotates crucible 19 at another reference rate in either the opposite direction (counter-rotation) or in the same direction (i.e., iso-rotation) relative to crystal 31. A control unit 51 of
Referring now to
Referring now to
In one embodiment, the defect model is a dynamic defect modeler 310 linked to the thermal model 308, and generates a point defect profile for an ingot being pulled from silicon 31 melt according to a predetermined pull rate profile 320. The defect model 310 may be either a point defect modeler or an agglomerated point defect modeler. One commercially available steady state point defect model interfaces with the software package CrysVUn available from the Fraunhofer Institute for Integrated Circuits located in Erlangen, Germany. An unsteady state point defect simulator has been published by Brown et al. (Journal of crystal Growth, 2001).In this case, the predetermined pull rate is representative of a ramped pull rate that forces the ingot to transition from vacancy dominated to interstitial dominated, and hence generate a V/I transition profile (i.e., point defect profile).
Referring briefly to
In another embodiment, the defect model is a static agglomerated defect modeler such as developed at the Massachusetts Institute of Technology Cambridge, Mass. and described in “Modeling the Linkages between Heat Transfer and Microdefect Formation in Crystal Growth: Examples of Czochralski Growth of Silicon and Vertical Bridgrnan Growth of Bismuth Germanate,” T. Mori, Ph. D. Thesis, Massachusetts Institute of Technology, 2000.
In the case of a static agglomerated defect model, both pull rate and the melt-solid interface shape may be analyzed by the V/I simulator to identify target melt-solid interface shapes and target pull rates at each of the various axial positions.
Moreover, the V/I simulator 300 can be fine tuned by growing a crystal 31 according to the defined profiles and comparing simulated V/I transitions to actual V/I transitions in a grown crystal 31 grown in an actual growth process 328 to determine a tuning factor 330. For example, by logging actual operating parameters during the actual growth process, and evaluating actual V/I transitions in the grown crystal the model can be fine tuned to improve the accuracy of predicted V/I transitions. In one embodiment, copper decoration and secco etching processes are used to reveal the actual V/I transitions crystal. Furthermore, actual interface shape measurements may be taken from test crystals by axial cutting of full-width thin samples, fully precipitating the oxygen in these samples, etching to remove the denuded zone and to clean the surface, and decorative etching to show the precipitation variation at the as grown solid-liquid interface. Based on actual transition seed lift values identified in grown crystals using metallic thermal precipitation of microdefects and subsequent decorative etching methods, a smooth seed lift profile may then be defined to achieve perfect silicon production over a majority of the crystal body length.
Referring now to
Referring next to
In one embodiment, the CPU 71 is responsive to the desired melt-solid interface shape profile 302 and the determined length by the storage of a desired heater power set point or profile. The PLC 69 is responsive to the heater power set point to generate a heater control signal 90 to supply to the heater power supply 82. The heater power supply 82 is responsive to the heater control signal 90 to control the current being supplied to the bottom heater 56 (e.g., resistance heater 21) surrounding the crucible 19 to control the temperature of the melt. More specifically, the heater control signal 90 controls the power being supplied to bottom heater 56 to control the temperature of the melt and a temperature profile along the crucible wall, and, thus, control the melt-solid interface shape. It has been discovered that the height of the interface increases with the heat flux from the bottom heater 56. As a result, a desired change in center height, as indicated by 502, of the interface height can be achieved by increasing power supplied to the bottom heater 56 by an amount that yields the desired change in height. For example, increasing the power to the bottom heater by 10 kilowatts, increases the height of the interface shape approximately 6 mm in a 28″ crucible, in a chosen hotzone configuration. (See
In another embodiment, the CPU 71 is responsive to the desired melt-solid interface shape profile 302 and the determined length by the storage of a determined magnet power set point. The PLC 69 is responsive to the magnet power set point to generate a magnet control signal 92 to supply to the magnet power supply 85. The magnet power supply 85 is responsive to the magnet control signal 92 to control the current being supplied to coils of a magnet 57 surrounding the crucible 19 to control the magnet field being applied to the melt 29. In particular, the application of a magnetic field provides a means of controlling the melt-solid interface shape along with the oxygen concentration. It has been discovered that by, decreasing the magnetic field strength, the melt-solid interface height increases. As a result, increasing, or decreasing, the power supplied to the magnet by an amount that yields the desired change in the magnetic field can achieve a desired change in center height 502 of the interface. For example, decreasing the magnetic field fifteen percent (15%) by decreasing the current, and thus power, to the magnet, increases the center height 502 of the interface shape approximately 2.5 mm. (See
In one embodiment, the CPU is responsive to the desired melt-solid interface shape profile 302 and the determined length by the storage of a determined crucible rotational speed set point and crystal rotational speed set point. The PLC 69 is responsive to linearly compute intermediate crucible rotational speed set point and the crystal rotational speed set points, based on the target values at selected crystal lengths to generate a crucible rotation control signal 94 and crystal rotation control signal 94 to supply to the crucible drive unit 45 and crystal drive unit 49, respectively. As a result, the relative rotational rate, or delta rotation, between the crucible 19 and crystal 31 is controlled to control the interface shape. Where the relative rotation rate refers to the absolute difference between the absolute crystal rotation rate and absolute crucible rotation rate, (i.e., ∥seed rotation−|crucible rotations. It has been discovered that, an increase in delta rotation increases the interface height; while a decrease in delta rotation decreases the interface height 502. As a result, a desired change in the center height of the interface height can be achieved by controlling the rotational speed differential between the crystal and the crucible. For example, increasing the rotational speed differential by two (2) revolutions per minute, increases the center height 502 of the interface shape approximately 4.5 mm. (See
In another embodiment, the crucible 19 is rotated in one direction and the crystal 31 is rotated in the same direction to improve the range over which the melt-solid interface shape can be controlled.
Although the invention is described above in reference to controlling various parameters individually, the present invention includes controlling two or more of the parameters to achieve the desired melt-solid interface shape.
As described above in reference to
Referring now to
In operation, a system for use in combination with a crystal growing apparatus for growing a monocrystalline ingot according to the Czochralski process has a memory storing a pre-computed set of process parameter setpoint values at various crystal lengths, that achieve or nearly achieve a desired melt-solid interface profile. The melt-solid interface profile represents a desired shape of a melt solid interface between the melt and the ingot during pulling as a function of a length of the ingot. A set of process gains allow estimation of operating parameters of the crystal growing apparatus as a function of the length of the ingot during pulling. Also, a controller is responsive to the determined operating parameter set point to adjust the operation condition of the crystal growing apparatus according to the determined operating parameter set point to control the shape of the melt-solid interface while the ingot is being pulled from the melt.
In defining the melt-solid interface shape profile, the process parameters may be selected based on actual oscillatory seed lift used to grow a crystal. In this instance, the actual interface shape may be confirmed, by a method of precipitation, bulk etching, and decorative etching or lifetime measurement. This embodiment of the present invention permits adjustment of the interface shape on subsequent crystals to closely achieve the desired interface shape at various crystal lengths. The seed lift profile may be changed from an oscillatory profile to a smooth profile based on the seed lift values experimentally determined where low or zero large-microdefect density is present (i.e., at the V/I transition boundary). The production of substantially radially-perfect or radially low defect silicon may be achieved, where small seed lift bias or profile adjustments are used to maintain low or zero large-microdefect levels, particularly where no interstitial loops or large vacancy clusters (as identified by D-defect or other measurements) occur.
The cooling rates through the temperature between 1150° C. and 800° C. determine the maximum possible acceptable operating window for the process, for a given V/I transition. The operating window is defined in terms of pull-rates. Suppose the mean pull-rate of a crystal at a location is x mm/min, a maximum operating window of y mm/min means that substantially large-microdefect free crystal at the chosen location can be produced within the pull-rate range between x+y/2 mm/min and x−y/2 mm/min. This is determined as follows.
A crystal is grown with a varying pull-rate profile 700, as shown in
Any circular disc cut from the shown cylindrical ingot at any axial location is still not completely substantially microdefect free. Since wafers produced from the ingot are circular discs, real operating window is defined by the width of the ingot that is substantially microdefect free everywhere along the radial location, i.e., from its center to the edge. This is called the real operating window. An acceptable interface shape for a given cooling rate can vary this real operating window. An acceptable interface shape giving an acceptable operating window of at least 0.005 mm/min, preferably 0.01 mm/min, and most preferably 0.02 mm/min or higher is determined first by the point defect simulator, and then by the actual crystal growth. Any interface allowing the production of substantially large-microdefect free circular discs is defined as acceptable.
This application in claims priority from U.S. Provisional Application No. 60/577,722, filed on Jun. 7, 2004.
| Number | Date | Country | |
|---|---|---|---|
| 60577722 | Jun 2004 | US |
