Patent Application
20090004458
The present invention generally relates to epitaxial semiconductor structures, especially epitaxial silicon wafers used in the manufacture of electronic components, and to methods for their preparation. More specifically, the epitaxial structures comprise a single crystal silicon substrate that is heavily doped with an N-type dopant (N+) or a P-type dopant (P+) and an epitaxial layer which is lightly doped with an N-type or P-type dopant, wherein the heavily doped substrate comprises a region near the lightly doped epitaxial layer having a high concentration of structures capable of suppressing or preventing silicon interstitial diffusion toward the epitaxial layer, thereby reducing silicon interstitial diffusion into said epitaxial layer.
Single crystal silicon, the starting material for most processes for the fabrication of semiconductor electronic components, is commonly prepared by the Czochralski process, wherein a single seed crystal is immersed into molten silicon and then grown by extraction. As molten silicon is contained in a quartz crucible, it is contaminated with various impurities, among which is mainly oxygen. As such, oxygen is present in supersaturated concentrations in the wafers sliced from single crystal silicon grown by this method.
During the thermal treatment cycles typically employed in the fabrication of electronic devices, oxygen precipitate nucleation centers may form and ultimately grow into large oxygen clusters or precipitates. Depending upon their location, such precipitates can be beneficial or detrimental. When present in active device regions of the wafer, they can impair the operation of the device. When present outside these regions, oxygen precipitates may serve as a gettering site for metals.
Various approaches have been used to manage oxygen precipitation behavior in wafers. For example, in U.S. Pat. No. 5,994,761, Falster et al. disclose a process for installing a non-uniform concentration of vacancies in a wafer in a rapid thermal annealer whereby in a subsequent oxygen precipitation heat-treatment, oxygen precipitates form in the vacancy-rich regions but not in the vacancy-lean regions. In U.S. Pat. No. 6,336,968, Falster discloses a process in which non-oxygen precipitating wafers are prepared by rapid thermally annealing the wafers in an oxygen-containing atmosphere or by slow-cooling the wafers through the temperature range at which vacancies are relatively mobile.
While these techniques have proven useful, to-date, for typical silicon wafers, epitaxial wafer structures comprising highly doped substrates present somewhat different challenges. For example, uncontrolled oxygen precipitation in heavily doped substrates can lead to the generation of relatively large concentrations of silicon self-interstitials at high temperatures because of their emission during oxygen precipitate growth. Relatively large concentrations of silicon self-interstitials, in turn, tend to promote diffusion of dopant (or other impurities) from the highly doped substrate into the more lightly doped device layer, thereby potentially altering critical characteristics, such as avalanche breakdown voltage, in some power devices. As such, if the number of silicon interstitials near the epitaxial layer were reduced, diffusion of dopant into the device layer would be retarded.
Among the various aspects of the present invention is a process for suppressing silicon self-interstitial diffusion from a heavily doped substrate into a lightly doped epitaxial layer in epitaxial silicon structures and the resulting structures, per se.
Briefly, therefore, one aspect of the present invention is an epitaxial silicon wafer comprising a single crystal silicon substrate that is heavily doped with an N-type dopant (N+) or a P-type dopant (P+) (i.e., having a resistivity of less than 5 mQ*cm) and an epitaxial layer which is lightly doped with an N-type or P-type dopant (i.e., having a resistivity of greater than about 10 mQ*cm). The heavily doped substrate comprises a silicon self-interstitial sink layer having a population of dislocation loops capable of suppressing diffusion of silicon self-interstitials into the epitaxial layer, wherein the layer of dislocation loops is at a depth of at least about 250 Å from the interface between the substrate and the epitaxial layer.
The present invention is further directed to a process for preparing such a wafer.
Other objects and features of this invention will be in part apparent and in part pointed out hereinafter.
With respect to the Figures, corresponding reference characters indicate corresponding parts throughout the several views of the drawings.
In accordance with one aspect of the present invention, N/N+, P/P+, N/P+, or P/N+ epitaxial silicon wafers may be prepared with a region comprising dislocation loops capable of serving as a sink for silicon self-interstitials near the epitaxial layer, thereby suppressing diffusion into the epitaxial layer. This region is referred to herein as a “silicon self-interstitial sink layer,” the “interstitial sink layer” or more simply, the “sink layer.”
Referring to
In one embodiment, silicon substrate 10 has an essential absence of oxygen precipitate nuclei to further reduce the potential for diffusion of dopants and other impurities into epitaxial layer 11. That is, oxygen precipitates will not form in this embodiment during an oxygen precipitation heat treatment (e.g., annealing the wafer at a temperature of 800° C. for four hours and then at a temperature of 1000° C. for sixteen hours).
Referring again to
Czochralski-grown silicon typically has an oxygen concentration within the range of about 5×1017 to about 9×1017 atoms/cm3 (ASTM standard F-121-83). In general, the single crystal silicon wafer may have an oxygen concentration falling anywhere within or even outside the range typically attainable by the Czochralski process.
In one embodiment, the single crystal silicon wafer comprises an axially symmetric region which has radial width of at least about 30% the length of the radius of the wafer and has no detectable agglomerated intrinsic point defects at a detection limit of 3×103 defects/cm3. The axially symmetric region may contain vacancies or silicon self-interstitials as the predominant intrinsic point defect.
Depending upon the cooling rate of the single crystal silicon ingot from the melting point of silicon (about 1410° C.) through the range of about 750° C. to about 350° C., oxygen precipitate nucleation centers may form in the single crystal silicon ingot from which the silicon wafer is sliced. In one embodiment, the silicon wafer possesses such nucleation centers. In another embodiment, the silicon wafer does not.
The single crystal silicon wafer is heavily doped with one or more N-type or P-type dopants. Typical N-type dopants include phosphorous, arsenic, and antimony. In one embodiment, the dopant is phosphorous. In another embodiment, the dopant is arsenic, while in yet another, the dopant is antimony. In a further embodiment, two or more of phosphorous, arsenic, and antimony are used as dopants. Typical P-type dopants include boron, aluminum, and gallium. In one embodiment, the dopant is boron. In another, the dopant is aluminum, while in yet another, the dopant is gallium. In a further embodiment, two or more of boron, aluminum, and gallium are used as dopants. Regardless of the dopant(s), the total concentration of dopant(s) is such that the wafer has a resistivity of less than about 5 mQ·cm, such material typically being referred to as N+ or P+ silicon. In one embodiment, the dopant concentration is sufficient to provide the wafer with a resistivity of less than about 3 mQ·cm. In certain embodiments, resistivities of less than about 2 mΩ·cm will be preferred. In yet other embodiments, the dopant concentration is sufficient to provide the substrate with a resistivity of less than about 1 mQ·cm.
The resistivity values noted above correspond to an N-type dopant concentration that is generally greater than about 1.24×1019 at/cm3. For example, the heavily doped wafer may have N-type dopant(s) present in a concentration greater than about 2.25×1019 at/cm3, such as greater than about 3.43×1019 at/cm3. In one preferred embodiment, the heavily doped wafer has N-type dopant(s) present in a concentration greater than about 7.36×1019 at/cm3. Similarly, the resistivity values noted above correspond to a P-type dopant concentration that is generally greater than about 2.1×1019 at/cm3. For example, the heavily doped wafer may have P-type dopant(s) present in a concentration greater than about 3.7×1019 at/cm3, such as greater than about 5.7×1019 at/cm3. In one preferred embodiment, the heavily doped wafer has P-type dopant(s) present in a concentration greater than about 1.2×1020 at/cm3.
Referring again to
In general, dislocation loops may be formed in a series of steps comprising inducing damage to the crystal lattice structure of a substrate to create a layer of amorphous material and annealing the damaged structure. This processing forms the dislocation loops of the interstitial sink layer, typically adjacent to the layer of amorphous material. The dislocation loops developed in this manner may be, for example, perfect dislocation loops, Frank dislocation loops, or a combination thereof. Referring again to
Regardless of the means by which they are formed, the concentration of dislocation loops in the silicon self-interstitial sink layer is at least about 1×108 loops/cm2. Preferably, the concentration of dislocation loops in the silicon self-interstitial sink layer is at least about 1×109 loops/cm2, such as at least about 5×109 loops/cm2, or even at least about 1×1010 loops/cm2.
Referring now to
The ions, I+ in
In general, the dislocation loops will form at the end of range of the implanted ions, with the end of range generally being deeper into the substrate than the projected range. That is, dislocation loops form at the edge of the layer of amorphous material farthest from face 13. As such,
Accordingly, target depth, D, is such that the subsequently formed dislocation loops do not extend to the surface upon which the epitaxial layer is formed or to the corresponding substrate/epitaxial layer interface. This target depth may be very close to or just below face 13, but D is more typically at least about 250 Å. In one embodiment, D is at least about 500 Å. In another embodiment, D is at least about 750 Å. In a further embodiment, D is about 1000 Å. In yet a further embodiment, D is greater than about 1000 Å.
Implantation depth may be affected, at least in part, by the ionic species implanted, since lighter ions tend to penetrate further into substrate 10 for a given implantation energy. Thus, for example, at an implant energy of 50 keV, silicon ions will have an average implant depth of about 750 Å, whereas germanium ions will have an average implant depth of 400 Å. In general, ions are preferably implanted using at least about 30 keV, such as at least about 40 keV, or even at least about 50 keV. In one application, ions are implanted using at least about 45 keV and less than about 55 keV.
Generally, dislocation loops form at the end of range of the implanted ions upon subsequent anneal if sufficient energy is used to implant a sufficient concentration of ions to form an amorphous layer of silicon. Typically, the dislocation loops may form at a depth of about 100 Å to 300 Å below the implanted ions, although the exact depth may be more or less. In general, it is more difficult to form amorphous material using lower mass elements. Accordingly, a much higher concentration of low mass elements must be used to induce sufficient damage, whereas lower concentrations of high mass elements are sufficient to form amorphous silicon. For example, when the implanted ions are silicon ions, the implanted dose is preferably at least about 2×1014 atoms/cm2, such as at least about 5×1014 atoms/cm2, or even at least about 1×1015 atoms/cm2. In one preferred embodiment, the implanted ion dose is at least about 2×1015 atoms/cm2. By comparison, when the implanted ions are the higher mass germanium ions, the implanted dose is preferably at least about 6×1013 atoms/cm2, such as at least about 1×1014 atoms/cm2, or even at least about 5×1014 atoms/cm2. In one preferred embodiment, the implanted ion dose is at least about 1×1015 atoms/cm2.
The implanted substrate is then annealed for a time and at a temperature to convert the crystallographic damage caused during ion implantation into dislocation loops 15, as shown in
In one embodiment, the ion implantation and anneal process is followed by one or more successive ion implantation and anneal processes, typically with successively lower implantation energy levels for the same ion, or with different ion(s). In this embodiment, the silicon self-interstitial sink layer comprises multiple strata of dislocation loops, which may increase the capacity for silicon self-interstitial consumption.
Referring again to
Regardless of when it is formed, epitaxial layer 11 is preferably grown by chemical vapor deposition; such processes are described, for example, in U.S. Pat. No. 5,789,309. Doping of the epitaxial layer may take place after or during the epitaxial layer growth process. Regardless of the doping method, the resulting epitaxial layer has a dopant concentration to provide the epitaxial layer with a resistivity of at least about 10 mΩ·cm, such as at least about 100 mΩ·cm. For example, the epitaxial layer will typically have a resistivity of between about 100 mΩ·cm and about 100 Ω·cm. In one application, the epitaxial layer will have a resistivity of between about 300 mΩ·cm and about 10 Ω·cm.
As an alternative means of characterizing the epitaxial layer, the epitaxial layer will typically have a dopant concentration of less than about 4.8×1018 at/cm3, such as between about 4.3×1013 at/cm3 and about 7.8×1016 at/cm3. In one application, the epitaxial layer has a dopant concentration between about 4.4×1014 at/cm3 and about 1.9×1016 at/cm3.
The epitaxial layer is doped, as described, with one or more of either N-type or P-type dopants. In an embodiment wherein N-type doping is preferred, the N-type dopants are selected, for example, from the group consisting of phosphorous, arsenic, and antimony. Typically, the N-type dopant will be phosphorous, arsenic, or both phosphorous and arsenic. In one embodiment, the dopant is phosphorous. In another, the dopant is arsenic. In yet another embodiment, both phosphorous and arsenic are used as dopants. In an embodiment wherein P-type doping is preferred, the P-type dopants are selected, for example, from the group consisting of boron, aluminum, and gallium. In one embodiment, the P-type dopant is boron.
One advantage of epitaxial deposition is that existing epitaxial growth reactors can be used in conjunction with a direct dopant feed during epitaxial growth. That is, the dopant can be mixed with the carrier gas to dope the deposited epitaxial layer.
In one embodiment, the wafer is also subjected to an oxygen precipitate nuclei dissolution step to improve oxygen precipitation behavior in the substrate. Stated differently, restricting or preventing oxygen precipitation in the substrate may yield an even greater degree of control over diffusion of dopant (and other impurities) into the more lightly doped epitaxial layer. The dissolution step may be performed before, after, or as part of the formation of the silicon self-interstitial sink layer, so long as the process parameters do not compromise the integrity of the dislocation loops. Preferably, this process is performed before the silicon self-interstitial sink layer is formed.
The highly doped wafer is subjected to a heat treatment step to cause dissolution of any pre-existing oxygen clusters and any pre-existing oxidation-induced stacking faults (OISF) nuclei in the substrate. Preferably, this heat treatment step is carried out in a rapid thermal annealer (RTA) in which the wafer is rapidly heated to a target temperature, then annealed at that temperature for a relatively short period of time. In general, the wafer is rapidly heated to a temperature in excess of 1150° C., preferably at least 1175° C., typically at least about 1200° C., and, in some embodiments, to a temperature of about 1200° C. to 1275° C. The wafer will generally be maintained at this temperature for at least one second, typically for at least several seconds (e.g., at least 3), and potentially for several tens of seconds (such as between about 10 and about 60 seconds, e.g., 20, 30, 40, or 50 seconds) depending upon the concentration, type, and size of any pre-existing defects.
The rapid thermal anneal may be carried out in any of a number of commercially available RTA furnaces in which wafers are individually heated by banks of high power lamps. Rapid thermal annealer furnaces are capable of rapidly heating a silicon wafer, e.g., they are capable of heating a wafer from room temperature to 1200° C. in a few seconds. One such commercially available RTA furnace is the 3000 RTP available from Mattson Technology (Freemont, Calif.).
In addition to dissolving a variety of pre-existing oxygen clusters and OISF nuclei, the annealing step will increase the number density of crystal lattice vacancies in the wafer. Information obtained to date suggests that certain oxygen-related defects, such as ring OISF, are high temperature nucleated oxygen agglomerates catalyzed by the presence of a high concentration of vacancies. Furthermore, in high vacancy regions, oxygen clustering is believed to occur rapidly at elevated temperatures, as opposed to regions of low vacancy concentration where behavior is more similar to regions in which oxygen precipitate nucleation centers are lacking. Because oxygen precipitation behavior is influenced by vacancy concentration, therefore, controlling the density of vacancies in the heat-treated wafer limits or even avoids oxygen precipitation in a subsequent oxygen precipitation heat treatment. Advantageously, the (number) density of vacancies in the annealed wafer can be controlled by limiting the cooling rate from the annealing temperature, by including a sufficient partial pressure of oxygen in the annealing atmosphere, or by doing both.
The vacancy concentration in the annealed wafer may be controlled, at least in part, by controlling the atmosphere in which the heat-treatment is carried out. Experimental evidence obtained to date suggests that the presence of a significant amount of oxygen suppresses the vacancy concentration in the annealed wafer. Without being held to any particular theory, it is believed that the rapid thermal annealing treatment in the presence of oxygen results in the oxidation of the silicon surface, creating an inward flux of silicon self-interstitials. This inward flux of self-interstitials has the effect of gradually altering the vacancy concentration profile by causing Frankel pair recombinations to occur, beginning at the surface and then moving inward.
Regardless of the mechanism, the annealing step is carried out in the presence of an oxygen-containing atmosphere in one embodiment. That is, the anneal is carried out in an atmosphere containing oxygen gas (O2), water vapor, or an oxygen-containing compound gas which is capable of oxidizing an exposed silicon surface. The atmosphere may thus consist entirely of oxygen or oxygen compound gas, or it may additionally comprise a non-oxidizing gas, such as argon. However, when the atmosphere is not entirely oxygen, the atmosphere will preferably contain a partial pressure of oxygen of at least about 0.001 atmospheres (atm.), or 1,000 parts per million atomic (ppma). More preferably, the partial pressure of oxygen in the atmosphere will be at least about 0.002 atm. (2,000 ppma), still more preferably 0.005 atm. (5,000 ppma), and still more preferably 0.01 atm. (10,000 ppma).
Intrinsic point defects (vacancies and silicon self-interstitials) are capable of diffusing through single crystal silicon, with the rate of diffusion being temperature dependent. The concentration profile of intrinsic point defects, therefore, is a function of the diffusivity of the intrinsic point defects and the recombination rate as a function of temperature. For example, the intrinsic point defects are relatively mobile at temperatures in the vicinity of the temperature at which the wafer is annealed in the rapid thermal annealing step, whereas they are essentially immobile for any commercially practical time period below or at temperatures of as much as 700° C. Experimental evidence obtained to-date suggests that the effective diffusion rate of vacancies slows considerably, such that vacancies can be considered to be immobile for any commercially practical time period, at temperatures less than about 700° C. and perhaps less than about 800° C., 900° C., or even 1,000° C.
Accordingly, in one embodiment the concentration of vacancies in the annealed wafer is controlled, at least in part, by controlling the cooling rate of the wafer through the temperature range in which vacancies are relatively mobile. Such control is exercised for a time period sufficient to reduce the number density of crystal lattice vacancies in the cooled wafer prior to cooling the wafer below the temperature range in which vacancies are relatively mobile. As the temperature of the annealed wafer is decreased through this range, the vacancies diffuse to the wafer surface and become annihilated, leading to a change in the vacancy concentration profile. The extent of such change depends on the length of time the annealed wafer is maintained at a temperature within this range and the magnitude of the temperature, with greater temperatures and longer diffusion times generally leading to increased diffusion. In general, the average cooling rate from the annealing temperature to the temperature at which vacancies are practically immobile (e.g., about 950° C.) is preferably no more than 20° C. per second, more preferably no more than about 10° C. per second, and still more preferably no more than about 5° C. per second.
Alternatively, the temperature of the annealed wafer following the high temperature anneal may be reduced quickly (e.g., at a rate greater than about 20° C./second) to a temperature of less than about 1150° C. but greater than about 950° C., and then held for a time period that is dependent upon the holding temperature. For example, several seconds (e.g., at least about 2, 3, 4, 6 or more) may be sufficient for temperatures near 1150° C., whereas several minutes (e.g., at least about 2, 3, 4, 6 or more) may be required for temperatures near 950° C. to sufficiently reduce the vacancy concentration.
Once the annealed wafer is cooled to a temperature outside the range of temperatures at which crystal lattice vacancies are relatively mobile, the cooling rate does not appear to significantly influence the precipitating characteristics of the wafer and, as such, does not appear to be narrowly critical.
Conveniently, the cooling step may be carried out in the same atmosphere in which the heating step is carried out. Suitable atmospheres include, e.g., nitriding atmospheres (i.e., atmospheres containing nitrogen gas (N2) or a nitrogen-containing compound gas that is capable of nitriding an exposed silicon surface, such as ammonia); oxidizing (oxygen-containing) atmospheres; non-oxidizing, non-nitriding atmospheres (such as argon, helium, neon, carbon dioxide); and combinations thereof
While the rapid thermal treatments employed herein may result in the out-diffusion of a small amount of oxygen from the surface of the front and back surfaces of the wafer, the resulting annealed wafer has a substantially uniform interstitial oxygen concentration as a function of distance from the silicon surface. For example, the annealed wafer will have a substantially uniform concentration of interstitial oxygen from the center of the wafer to regions of the wafer that are within about 15 microns of the silicon surface, more preferably from the center of the silicon to regions of the wafer that are within about 10 microns of the silicon surface, even more preferably from the center of the silicon to regions of the wafer that are within about 5 microns of the silicon surface, and most preferably from the center of the silicon to regions of the wafer that are within about 3 microns of the silicon surface. In this context, a substantially uniform oxygen concentration shall mean a variance in the oxygen concentration of no more than about 50%, preferably no more than about 20%, and most preferably no more than about 10%.
In one embodiment, the epitaxial layer is formed in conjunction with the annealing step detailed above. In this embodiment, the annealing step is carried out in the epitaxial reactor. Upon completing the annealing step and epitaxial layer formation, the cooling atmosphere, cooling rate, or both the cooling atmosphere and rate are controlled as detailed above. That is, in one variation of this embodiment, the atmosphere after the anneal and epitaxial layer formation is an oxygen-containing atmosphere that is capable of oxidizing an exposed silicon surface. Specifically, the atmosphere will preferably contain a partial pressure of oxygen of at least about 0.001 atmospheres (atm), or 1,000 parts per million atomic (ppma). More preferably, the partial pressure of oxygen in the atmosphere will be at least about 0.002 atm (2,000 ppma), still more preferably 0.005 atm (5,000 ppma), and still more preferably 0.01 atm (10,000 ppma).
In other variations of this embodiment, the cooling rate of the wafer is controlled with or without controlling the cooling atmosphere. Specifically, the cooling rate is controlled such that the average cooling rate from the annealing temperature to the temperature at which vacancies are practically immobile (e.g., about 950° C.) is preferably no more than 20° C. per second, more preferably no more than about 10° C. per second, and still more preferably no more than about 5° C. per second. Alternatively, the temperature may be reduced quickly (e.g., at a rate greater than about 20° C./second) to a temperature of less than about 1150° C. but greater than about 950° C., and then held for a time period between several seconds to several minutes, depending upon the holding temperature. For example, at least about 2, 3, 4, 6 seconds or more may be sufficient for temperatures near 1150° C., whereas at least about 2, 3, 4, 6 minutes or more may be required for temperatures near 950° C.
In one embodiment, a polysilicon layer is deposited on the backside of the highly doped substrate before the annealing step described above. The grain boundaries of the polysilicon layer serve as a gettering site for dopant. In general, the polysilicon layer may be deposited by any means conventionally known in the art. For example, the polysilicon layer may be deposited by chemical vapor deposition using silane (SiH4) gas and arsenic doping, as more fully described in U.S. Pat. No. 5,792,700 or 5,310,698.
Silicon structures manufactured according to this invention may be used in various technologies. For example, the silicon structure of this invention is suitable for use in the manufacture of power devices, such as power diodes, thyristors, and, in particular, power MOSFETs and JFETs. This list is in no way intended to be restrictive or comprehensive.
The following non-limiting example is provided to further illustrate and explain the present invention. The invention should not be limited to any of the details provided herein.
A single crystal silicon wafer doped with about 7.86×1019 phosphorus atoms/cm3 is exposed to an ion implantation process wherein silicon ions are implanted into the front surface of the wafer. The silicon ions are implanted with an energy level of 50 keV such that the substrate has a concentration of about 2×1015 atoms/cm2 at an average distance of about 1000 Å from the front surface. The ion implanted, highly doped substrate is then annealed at about 900° C. for about 120 minutes to form about 1.3×1010 dislocation loops/cm2 at about 1000 Å from the front surface. An epitaxial layer is then formed on the front surface of the substrate, the epitaxial layer being doped with less than about 4.8×101 phosphorus atoms/cm3. Transmission electron microscopic analysis at 75,000 times magnification reveals the presence of the dislocation loops, as seen in
When introducing elements of the present invention or the preferred embodiments(s) thereof, the articles “a”, “an”, “the”, and “said” are intended to mean that there are one or more of the elements. The terms “comprising”, “including”, and “having” are intended to be inclusive and mean that there may be additional elements other than the listed elements. Moreover, unless explicitly noted otherwise, reference to the heavily doped substrate as “N+” or “P+” should be understood to also refer to substrates having doping levels conventionally referred to as N++ and N+++ or P++ and P+++, respectively. Also, unless explicitly noted otherwise, reference to the lightly doped epitaxial layer as “N” or “P” should be understood to also refer to substrates having doping levels conventionally referred to as N− or P−, respectively.
In view of the above, it will be seen that the several objects of the invention are achieved and other advantageous results attained.
As various changes could be made in the above products and methods without departing from the scope of the invention, it is intended that all matter contained in the above description and shown in the accompanying drawings shall be interpreted as illustrative and not in a limiting sense. The above description of the various embodiments is intended only to acquaint others skilled in the art with the invention, its principles, and its practical application so that others skilled in the art may adapt and apply the invention in its numerous forms, as may be best suited to the requirements of a particular use.
