US20030116083A1 - Enhanced n-type silicon material for epitaxial wafer substrate and method of making same - Google Patents

Abstract

An enhanced n⁺ silicon material for epitaxial substrates and a method for producing it are described. The enhanced material leads to improved gettering characteristics of n/n⁺ epitaxial wafers based on these substrates. The method for preparing such n⁺ silicon material includes applying a co-doping of carbon to the usual n dopant in the silicon melt, before growing respective CZ crystals. This improves yield of enhanced n⁺ silicon material in crystal growing and ultimately leads to device yield stabilization or improvement when such n/n⁺ epitaxial wafers are applied in device manufacturing.

Description

RELATED APPLICATION DATA
• This application is a division of copending U.S. patent application Ser. No. 09/354,994, filed Jul. 16, 1999, now U.S. Pat. No. ______, to issue Dec. 10, 2009.[0001]
FIELD OFF THE INVENTION
• This invention relates generally to the field of preparing silicon substrate wafers for use in the formation of semiconductor devices such as power discrete or power integrated circuits. [0002]
BACKGROUND OF THE INVENTION
• Semiconductor devices are built either into polished or epitaxial silicon wafers. The latter consists of an epitaxial (epi) layer on top of a polished wafer substrate. Epi layers typically contain low concentrations of electrically active dopants, usually phosphorous (n-type conductivity), or boron (p-type conductivity), typically close to 10[0003] ¹⁵ atoms cm⁻³. Substrates in many cases contain high concentrations of dopant atoms, which may be phosphorous, antimony, or arsenic (n-type) or boron (p-type), typically in the range 10¹⁸-10¹⁹ atoms cm⁻³. Dopant levels close to the solubility limit for respective dopant species are needed to lower the resistivity of epi wafer substrates, an important requirement for state-of-the-art power device applications. Silicon material containing such high levels of n-dopant is generally called n⁺ material. Such material, cut in slices from respective n⁺ crystals, is used for preparing n⁺ substrates for ultimate n/n⁺ epi wafers.
• Oxygen is incorporated into crowing crystals applying the Czochralski (CZ) technique through dissolution of the fused silica or quartz (SiO[0004] ₂) crucibles used for holding the silicon melt. The molten silicon reacts with the SiO₂ crucible wall to form SiO. Some of the SiO evaporates from the melt at the temperature and pressure commonly used for silicon crystal growth. However, some remains in the melt and may be incorporated into the growing crystal. As the melt is solidified, the contact area between the melt and the crucible wall decreases while the area of melt surface available for evaporation of SiO remains substantially constant until near the end of the crystal growth. Consequently, the concentration of oxygen in the melt and therefore the concentration incorporated into the crystal tends to decrease with increasing distance from the seed end of the crystal. Without any countermeasures, this leads to an axial oxygen profile which typically displays decreasing oxygen concentration toward the tail-end of the crystal. The presence of high concentrations of n-type dopants in the silicon melt enhances evaporation of SiO during crystal growing and thereby further reduces the amount of oxygen incorporated into a growing n⁺ crystal, leading to an axial oxygen profile decreasing heavily toward the tail-end of such a crystal. Without any state-of-the-art countermeasures, after reaching a certain percentage of the total length of such a CZ n⁺ crystal, the oxygen concentration typically drops below the level required to generate adequate oxygen precipitation when such material is later processed in thermal device manufacturing steps. The length of the crystal at which the oxygen level drops below that required for adequate oxygen precipitation is called the critical crystal length abbreviated L_c.
• Oxygen precipitation in epi wafer substrates is the prerequisite for internal gettering (IG) typically applied for controlling the degradation of device yield by way of heavy metal contamination during the thermal device manufacturing steps. Such degradation is described in an article by A. Borghesi, B. Pivac, A. Sassella and A. Stella entitled Oxygen Precipitation in Silicon, published in the Journal of Applied Physics, Vol. 77, No. 9, May 1, 1995, pp.4169-4244, at 4206-07. Effective IG has been observed at oxygen precipitation related bulk defect densities in the order of 10[0005] ⁹ atoms cm⁻³. This bulk defect density is critical for effective IG and is referred to hereinafter as the critical defect number N_c. Epitaxial n/n⁺ wafers based on such high defect density n⁺ substrates exhibit superior IG related leakage resistance and thereby potentially improved device yield. Thermally induced oxygen precipitation during device processing is suppressed in the case of n-type dopant atoms in epi wafer substrates which creates the necessity to introduce large quantities of oxygen into a crystal. It has been determined experimentally by the inventors hereof that CZ crystals with arsenic concentrations in the order to 10¹⁹ atoms cm⁻³ need approximately 8×10¹⁷ atoms cm⁻³ oxygen (ASTM 121-83 calibration) in order to reach the N_c necessary for effective IG. Without any state of the art countermeasures, L_c is less than 10% of the total crystal length in this case. In order to essentially increase L_c, effort heretofore has been generally directed at reducing the axial variation of oxygen incorporation. Currently used techniques aiming at axially homogenizing the oxygen level include adjusting crystal pull speed and utilizing crystal and crucible rotation, all in conjunction with controlling gas flow and pressure in the puller chamber. Another technique is the application of defined magnetic fields during crystal growth. These countermeasures are technically sophisticated and/or associated with high cost.
• The presence of carbon in silicon wafers has long been known to enhance the precipitation of oxygen. For example, Ahlgren et al. European Application No. 84109528.4 at page 7, lines 26 to 33 teaches that silicon with carbon concentration in excess of 4 ppma (2×10[0006] ¹⁷ atoms cm⁻³) (ASTM 123-76 calibration) can induce substantial oxygen precipitation in silicon containing less than 28 ppma(1.4×10¹⁸ atoms cm⁻³) oxygen (ASTM 121-79 calibration) after a thermal treatment that would induce negligible oxygen precipitation at lower concentrations of carbon. It appears that that work refers to the addition of carbon by the usual means as set forth above. Thus, the work accepts the carbon which is introduced as a necessary “evil” in consequence of the available equipment used in 1984 and sampling the carbon content along the length of the grown crystal to determine what portion can be advantageously used. Such carbon introduction is uncontrolled and mainly due to the graphite parts used in the puller construction. In current state of the art crystal pullers it is possible to maintain carbon at levels below 5×10¹⁵ atoms cm⁻³ in spite of the use of graphite heaters and insulation. Moreover, the European application makes no mention of the presence of n-type or p-type doping materials and it is directed to lightly doped silicon crystals for substrates.
• Developments aimed at reducing carbon contamination in crystal growth were originally driven by experimental evidence of detrimental device impact of carbon if present in certain concentration levels within critical device regions of wafers. In the case of epi wafer substrates it is highly unlikely that carbon would enter a critical device regions (typically located in epi layers deposited on top of a substrate) because carbon is a slow diffuser in silicon. Even so, current epi wafer specifications typically still call for carbon concentrations below 10[0007] ¹⁶ atoms cm⁻³.
SUMMARY OF THE INVENTION
• The present invention is directed to a process for growing silicon crystals wherein predetermined amounts of carbon are added in a controlled fashion to produce the level of oxygen precipitation desired. This process can be effective in n[0008] ⁺ doped silicon epi substrates at carbon levels significantly lower than 2×10¹⁷ atoms cm⁻³. Rapidly increasing carbon concentration is observed only toward the tail-end of carbon co-doped crystals because its incorporation into the crystal is controlled by the segregation behavior.
• Such carbon doping of CZ silicon at a very low concentration can strongly increase the oxygen precipitation in heavily n-doped materials. Moreover, there is a relationship between co-doped carbon, oxygen concentration and bulk defect density after annealing, enabling predetermination of the amount of carbon to be added to achieve the bulk defect level necessary for effective internal gettering. The established methodology allows development of simple and low-cost crystal growing processes leading to enhanced n-type silicon material for epitaxial wafer substrates.[0009]
BRIEF DESCRIPTION OF THE DRAWINGS
• Additional objects and features of the invention will be more readily apparent from the following detailed description and appended claims when taken in conjunction with the drawing, in which: [0010]
• FIG. 1 is a lateral sectional view of a pulling apparatus using the Czolchralski technique suitable for use with the present invention; [0011]
• FIG. 2 is a graph showing the axial distribution of oxygen concentration along the length of a heavily arsenic doped crystal co-doped with carbon in accordance with the invention; [0012]
• FIG. 3 is a graph similar to FIG. 2 but showing the axial distribution of carbon concentration along the length of a heavily arsenic doped crystal co-doped with carbon in accordance with the invention; [0013]
• FIG. 4 is a graph showing percentage of the grown crystal suitable for internal gettering as a function of the amount of carbon added to the melt; and [0014]
• FIG. 5 is a graph showing the required oxygen concentration for internal gettering as a function the amount of carbon added to the melt.[0015]
DESCRIPTION OF THE PREFERRED EMBODIMENT
• FIG. 1 shows one of several forms of a pulling apparatus using the Czolchralski technique suitable for use with the invention. A [0016] quartz crucible 11 is positioned inside a pulling chamber 13. The quartz crucible 11 is attached to a rotatable bottom shaft 15. A heater 17 is provided around the quartz crucible 11 for heating and controlling the temperature of a melt M in the quartz crucible 11. The melt is primarily silicon but includes dopants. A heat retaining tube 19 is provided between the heater 17 and the pulling chamber 13. An annular supporting member 21 is attached at the top surface of the heat retaining tube 19.
• To produce a silicon monocrystal by the CZ technique, polycrystalline silicon and the desired dopant(s), for example, P, B, Sb or As, are placed in the [0017] quartz crucible 11. A seed crystal is attached to and supported by a bracket 29 on a pulling shaft. The pulling chamber 13 is evacuated to a vacuum, and the heater 17 melts the polycrystalline silicon and the dopant(s). An inert carrier gas, such as argon, is passed through an inlet 31 into the pulling chamber 13 around the quartz crucible 11 and out the discharge 33. At the same time, the seed crystal is immersed in the melt in the quartz crucible 11. The pulling shaft then withdraws the seed crystal at a predetermined speed while rotating relative to the quartz crucible 11.
• FIGS. 2 and 3 show axial distributions of oxygen and carbon concentration in a 100 mm diameter silicon crystal doped with arsenic in the order of 10[0018] ¹⁹ atoms cm⁻³. Before growing this crystal, 150 mg carbon were initially added to the molten silicon charge of 30 kg. The inverse concentration characteristics of oxygen and carbon are evident from the two graphs. Carbon co-doped n⁺ crystals, even without employing means to maintain high oxygen incorporation throughout the crystal length, yield significantly higher in terms of potentially high-defect-density material needed for manufacturing n⁺ substrates for ultimately producing leakage-resistant n/n⁺ epitaxial wafers.
• FIG. 4 shows the crystal yield increase as a function of carbon added to the silicon melt. From FIG. 4 it is obvious that there is established a simple method for determining the amount of carbon which must be added to the initial molten silicon charge to achieve a desired minimum level of bulk defect density over a defined portion or the entire length of the crystal grown from said charge employing standard growing technique and by applying a defined wafer annealing procedure. As an example, in laboratory tests, it was found that substrate material with arsenic concentrations in the order of 10[0019] ¹⁹ atoms cm⁻³ reaches the N_c limit (for effective internal gettering) at an oxygen concentration of approximately 7.5×10¹⁷ atoms cm⁻³ if only light carbon doping of 1.9×10¹⁶ atoms cm⁻³ is applied. This is a substantial improvement over the 8.0×10¹⁷ atoms cm⁻³ oxygen needed with the typical state of the art carbon concentration level <10¹⁶ atoms cm⁻³. Oxygen can be further reduced to 6.25×10¹⁷ atoms cm⁻³ if the carbon concentration at the seed end of the crystal is increased to 4.3×10¹⁶ atoms cm⁻³. Consequently, in carbon-doped crystals there is no, or an extremely reduced, need to increase L_c by reducing the axial oxygen variation (increasing the oxygen concentration toward the crystal tail).
• Upon review of the graph shown in FIG. 5 it is seen that there is established a simple method to estimate the oxygen concentration needed over a range of carbon co-doping levels. The data points “a”, “b” and “c” in FIG. 4 and corresponding data points “d”, “e” and “f” in FIG. 5 were derived from tests comparing three heavily arsenic-doped crystals of 100 mm diameter. The arsenic concentration in these crystals was targeted to increase from 1.8×10[0020] ¹⁹ atoms cm⁻³ (crystal seed) to 3.8×10¹⁹ atoms cm⁻³ (crystal tail). The corresponding amount of arsenic dopant was added to 30 kg charges of poly-silicon after melting the silicon charge. The first crystal was grown without intentionally adding carbon (data points “a” and “d” respectively in FIGS. 4 and 5). The second crystal was grown after adding only 50 mg high-purity carbon to the melt (data points “b” and “e” respectively in FIGS. 4 and 5), and for the third crystal 150 mg high-purity carbon was added (data points “c” and “f” respectively in FIGS. 4 and 5). No additional countermeasures for homogenizing the axial oxygen profile were applied. As a result, the three crystals with varying carbon levels have a comparable axial oxygen profile: the oxygen concentration falls from 8.3×10¹⁷ atoms cm⁻³ at the crystal seed to 4.0×10¹⁷ atoms cm⁻³ at the crystal tail.
• Summarizing, critical bulk defect density levels, needed for effective internal gettering in substrates for epitaxial wafers, can be reached at significantly lower oxygen levels in respective crystal material, as compared to material without carbon doping. Applying precipitation testing on wafers from these crystals (evaluation of N[0021] _c as a function of crystal location), it was found that carbon co-doping clearly increases the critical crystal length with oxygen precipitation characteristics needed for effective internal gettering (N_c>10⁹ atoms cm⁻³). There is a nearly linear increase of the high-precipitation portion of these crystals with carbon co-doping (FIG. 4). For example, more than 50% of the total length of a crystal exceeds L_c when 150 mg of carbon is added to the initial 30 kg charge of silicon. The oxygen concentration necessary to generate effective internal gettering is coupled with the added carbon in a well-defined manner (FIG. 5). This means carbon co-doping can be applied for oxygen precipitation control in n⁺ material used for epi wafer substrates, instead of sophisticated and/or expensive measures to increase and axially homogenize the oxygen concentration in such crystals.

Claims (13)

1. An enhanced n⁺ silicon crystal material for epitaxial substances having:

a dopant of n-type material at a concentration of at least 10¹⁸ atoms cm⁻³; and

a co-dopant of carbon at a concentration of at least 10¹⁶ atoms cm⁻³.

2. An enhanced n⁺ silicon crystal material for epitaxial substances as defined in claim 1 wherein said dopant of n-type material is arsenic at a concentration of at least 10¹⁹ atoms cm⁻³.

3. An enhanced n⁺ silicon crystal material for epitaxial substances as defined in claim 1 wherein said dopant of n-type material is antimony.

4. An enhanced n⁺ silicon crystal material for epitaxial substances as defined in claim 1 wherein said dopant on n-type material is phosphorus at a concentration of at least 10¹⁹ atoms cm⁻³.

5. An enhanced n⁺ silicon crystal material for epitaxial substances as defined in claim 1 in which the co-dopant of carbon is present in a concentration of at least 1.9×10¹⁶ atoms cm⁻³.

6. An enhanced n⁺ silicon crystal material for epitaxial substances as defined in claim 1 in which the co-dopant of carbon is present in a concentration sufficient in the presence of said n⁺ doping concentration to produce oxygen-precipitation-related bulk defects in the material of at least 10⁹ defects cm⁻³.

7. A wafer of an enhanced n⁺ silicon crystal material according to claim 1, including:

an epitaxial layer on a major surface of the substrate wafer, thereby providing an active device layer for active device layer formation decoupled from defects in the substrate wafer,

the epitaxial layer having an n-type dopant concentration at least 3 orders of magnitude less than the n⁺ doping concentration.

8. A method for preparing n⁺ silicon material comprising the steps of:

applying a dopant of phosphorus to silicon melt in an amount sufficient to produce an n⁺ doping concentration of at least 10¹⁸ atoms cm⁻³;

applying a co-dopant of carbon to the silicon melt in an amount effective to promote oxygen precipitation in the silicon material in the presence of said n⁺ doping concentration; and

applying a seed crystal to said melt and growing a crystal therefrom by withdrawing the seed in the Czochralski technique, wherein said co-dopant of carbon is at a concentration of at least 10¹⁶ atoms cm⁻³.

9. A method for preparing n⁺ silicon material as defined in claim 1 wherein said dopant of phosphorus produces an n-type doping concentration of at least 10¹⁹ atoms cm⁻³.

10. A method for preparing n⁺ silicon material as defined in claim 1 wherein said co-dopant of carbon is at a concentration of at least 10¹⁶ atoms cm⁻³.

11. A method for preparing n⁺ silicon material comprising the steps of:

applying a dopant of phosphorus to silicon melt in an amount sufficient to produce an n⁺ doping concentration of at least 10¹⁹ atoms cm⁻³;

applying a co-dopant of carbon to the silicon melt; and

applying a seed crystal to said melt and growing a crystal therefrom by withdrawing the seed in the Czochralski technique;

the co-dopant of carbon being applied in an amount sufficient to produce a carbon concentration of at least 1.9×10¹⁶ atoms cm⁻³.

12. A method for preparing n⁺ silicon material comprising the steps of:

applying a dopant of phosphorus to silicon melt in an amount sufficient to produce an n⁺ doping concentration of at least 10¹⁹ atoms cm⁻³;

applying a co-dopant of carbon to the silicon melt; and

applying a seed crystal to said melt and growing a crystal therefrom by withdrawing the seed in the Czochralski technique;

the co-dopant of carbon being applied in a concentration sufficient in the presence of said n⁺ doping concentration to produce oxygen-precipitation-related bulk defects in the material of at least 10⁹ defects cm⁻³.

13. A method for preparing n⁺ silicon material comprising the steps of:

applying a dopant of phosphorus to silicon melt in an amount sufficient to produce an n⁺ doping concentration of at least 10¹⁹ atoms cm⁻³;

applying a co-dopant of carbon to the silicon melt in an amount sufficient to produce a carbon concentration in the n⁺ silicon material of at least 1.9×10⁶ atoms cm⁻³; and

applying a seed crystal to said melt and growing a crystal therefrom by withdrawing the seed in the Czochralski technique;

slicing the crystal into wafers and manufacturing substrate wafers therefrom; and

forming an epitaxial layer on a major surface of the substrate wafers, thereby providing an active device layer for active device layer formation decoupled from defects in the substrate wafer,

the epitaxial layer having an n-type dopant concentration at least 3 orders of magnitude less than the n⁺ doping concentration.

Images