Classifications

• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L21/00—Processes or apparatus adapted for the manufacture or treatment of semiconductor or solid state devices or of parts thereof
  • H01L21/02—Manufacture or treatment of semiconductor devices or of parts thereof
  • H01L21/04—Manufacture or treatment of semiconductor devices or of parts thereof the devices having potential barriers, e.g. a PN junction, depletion layer or carrier concentration layer
  • H01L21/18—Manufacture or treatment of semiconductor devices or of parts thereof the devices having potential barriers, e.g. a PN junction, depletion layer or carrier concentration layer the devices having semiconductor bodies comprising elements of Group IV of the Periodic Table or AIIIBV compounds with or without impurities, e.g. doping materials
  • H01L21/185—Joining of semiconductor bodies for junction formation
  • H01L21/187—Joining of semiconductor bodies for junction formation by direct bonding
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L29/00—Semiconductor devices specially adapted for rectifying, amplifying, oscillating or switching and having potential barriers; Capacitors or resistors having potential barriers, e.g. a PN-junction depletion layer or carrier concentration layer; Details of semiconductor bodies or of electrodes thereof ; Multistep manufacturing processes therefor
  • H01L29/66—Types of semiconductor device ; Multistep manufacturing processes therefor
  • H01L29/66007—Multistep manufacturing processes
  • H01L29/66053—Multistep manufacturing processes of devices having a semiconductor body comprising crystalline silicon carbide
  • H01L29/66068—Multistep manufacturing processes of devices having a semiconductor body comprising crystalline silicon carbide the devices being controllable only by the electric current supplied or the electric potential applied, to an electrode which does not carry the current to be rectified, amplified or switched, e.g. three-terminal devices
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L29/00—Semiconductor devices specially adapted for rectifying, amplifying, oscillating or switching and having potential barriers; Capacitors or resistors having potential barriers, e.g. a PN-junction depletion layer or carrier concentration layer; Details of semiconductor bodies or of electrodes thereof ; Multistep manufacturing processes therefor
  • H01L29/66—Types of semiconductor device ; Multistep manufacturing processes therefor
  • H01L29/66007—Multistep manufacturing processes
  • H01L29/66075—Multistep manufacturing processes of devices having semiconductor bodies comprising group 14 or group 13/15 materials
  • H01L29/66227—Multistep manufacturing processes of devices having semiconductor bodies comprising group 14 or group 13/15 materials the devices being controllable only by the electric current supplied or the electric potential applied, to an electrode which does not carry the current to be rectified, amplified or switched, e.g. three-terminal devices
  • H01L29/66234—Bipolar junction transistors [BJT]
  • H01L29/66325—Bipolar junction transistors [BJT] controlled by field-effect, e.g. insulated gate bipolar transistors [IGBT]
  • H01L29/66333—Vertical insulated gate bipolar transistors
  • H01L29/66348—Vertical insulated gate bipolar transistors with a recessed gate
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L29/00—Semiconductor devices specially adapted for rectifying, amplifying, oscillating or switching and having potential barriers; Capacitors or resistors having potential barriers, e.g. a PN-junction depletion layer or carrier concentration layer; Details of semiconductor bodies or of electrodes thereof ; Multistep manufacturing processes therefor
  • H01L29/66—Types of semiconductor device ; Multistep manufacturing processes therefor
  • H01L29/66007—Multistep manufacturing processes
  • H01L29/66075—Multistep manufacturing processes of devices having semiconductor bodies comprising group 14 or group 13/15 materials
  • H01L29/66227—Multistep manufacturing processes of devices having semiconductor bodies comprising group 14 or group 13/15 materials the devices being controllable only by the electric current supplied or the electric potential applied, to an electrode which does not carry the current to be rectified, amplified or switched, e.g. three-terminal devices
  • H01L29/66363—Thyristors
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L29/00—Semiconductor devices specially adapted for rectifying, amplifying, oscillating or switching and having potential barriers; Capacitors or resistors having potential barriers, e.g. a PN-junction depletion layer or carrier concentration layer; Details of semiconductor bodies or of electrodes thereof ; Multistep manufacturing processes therefor
  • H01L29/66—Types of semiconductor device ; Multistep manufacturing processes therefor
  • H01L29/66007—Multistep manufacturing processes
  • H01L29/66075—Multistep manufacturing processes of devices having semiconductor bodies comprising group 14 or group 13/15 materials
  • H01L29/66227—Multistep manufacturing processes of devices having semiconductor bodies comprising group 14 or group 13/15 materials the devices being controllable only by the electric current supplied or the electric potential applied, to an electrode which does not carry the current to be rectified, amplified or switched, e.g. three-terminal devices
  • H01L29/66409—Unipolar field-effect transistors
  • H01L29/66416—Static induction transistors [SIT]
  • H01L29/66424—Permeable base transistors [PBT]
• • H—ELECTRICITY
  • H01—ELECTRIC ELEMENTS
  • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
  • H01L29/00—Semiconductor devices specially adapted for rectifying, amplifying, oscillating or switching and having potential barriers; Capacitors or resistors having potential barriers, e.g. a PN-junction depletion layer or carrier concentration layer; Details of semiconductor bodies or of electrodes thereof ; Multistep manufacturing processes therefor
  • H01L29/02—Semiconductor bodies ; Multistep manufacturing processes therefor
  • H01L29/12—Semiconductor bodies ; Multistep manufacturing processes therefor characterised by the materials of which they are formed
  • H01L29/16—Semiconductor bodies ; Multistep manufacturing processes therefor characterised by the materials of which they are formed including, apart from doping materials or other impurities, only elements of Group IV of the Periodic Table
  • H01L29/1608—Silicon carbide

Definitions

• the present invention relates to the field of semiconductors, and, more particularly, to a method of fabricating power semiconductor devices, and the devices produced by the method.
• a silicon controlled rectifier (SCR) or thyristor is a bistable semiconductor switching device formed from four layers of silicon.
• SCR silicon controlled rectifier
• thyristor bistable semiconductor switching device formed from four layers of silicon.
• MOS controlled thyristor is especially suited for resonant (zero voltage or zero current switching applications.
• the MCT has a forward voltage drop much like the SCR, and therefore enjoys greatly reduced conduction power loss.
• the MCT allows the control of high power circuits with very small amounts of input energy — a feature common to SCRs as well.
• turn-off is accomplished by turning on a highly interdigitated off-FET to short out one or both of the emitter-base junctions of a thyristor.
• IGBT insulated gate bipolar transistor
• MCT insulated gate bipolar transistor
• Available MCTs and IGBTs are useful at high switching frequency than is generally practice with power Darlington transistors, for example.
• both may be operated with junction temperatures of 150 °C and above, and operate in switching circuits having 600 volts or higher switch ratings.
• the wafer bonding has been for the purpose of replacing a thick, e.g. 100 m epitaxial layer growth.
• high temperature bonding anneals at temperatures of greater than about 1100° C are typically used to remove microvoids and bubbles. Both hydrophobic and hydrophilic bonding has been used.
• SUBSTITUTE SHEET (RUL£ 26) wherein two wafers are bonded together, and annealed at a temperature in a range of 800 to 1100°.
• An N-type wafer is doped N+ at a surface thereof and is bonded to a P+ wafer to define an N+ buffer region for the IGBT.
• a gate is formed on the upper surface and various diffusions are also made adjacent the gate to define an emitter/ collector encircling the gate.
• An emitter contact is formed on the diffusions and a collector contact is deposited on the lower surface of the wafer using conventional techniques.
• the relatively high temperature annealing and subsequent device processing steps may adversely affect the doping profile of the buffer layer. Accordingly, the turnoff speed may be reduced.
• the double-sided processing after annealing requires a relatively large number of process steps, and the substrates are subject to mechanical damage which may reduce yields.
• a first embodiment of a semiconductor device comprising a laterally extending semiconductor base, a buffer adjacent the base and having a first conductivity type dopant, and a laterally extending emitter adjacent the buffer and opposite the base and having a second conductivity type dopant.
• the buffer is relatively thin and has a first conductivity type dopant concentration greater than a second conductivity type dopant concentration in adjacent emitter portions to provide a negative temperature coefficient for current gain and a positive temperature coefficient for forward voltage for the device.
• the negative temperature coefficient for current gain provides reduces thermal run away and provides better short circuit protection for the device.
• the base may have a first conductivity type dopant in a concentration less than the concentration of first conductivity type dopant in the buffer.
• the buffer may have a thickness less than about 10 microns, and more preferably in a range of about 200 to 500 nanometers.
• the dopant concentration of the buffer is preferably greater than about 3xl0 18 cm "3 for one embodiment, and greater than about lxl0 17 cm "3 for another.
• At least one of the base and the emitter may comprise silicon, and the buffer may also comprise silicon in one embodiment. In another embodiment the buffer may comprise germanium.
• the semiconductor device may be formed according to low temperature bonding as described in detail below. Accordingly, in one embodiment the device further includes a bonded interface between the emitter and the buffer. The bonded interface may also be between the buffer and the base. The bonded interface is preferably substantially devoid of oxide.
• the emitter comprises an epitaxial portion adjacent the buffer and a second portion opposite the epitaxial portion.
• the semiconductor device may include a MOSFET current control device, or other current control device, formed in at least one of the first and second portions.
• Yet another device in accordance with the present invention includes a laterally extending localized lifetime killing portion between oppositely doped first and second laterally extending portions.
• the localized lifetime killing portion may comprise a plurality of laterally conf ned and laterally spaced apart lifetime killing regions.
• a bonded interface may be between the localized lifetime killing portion and either the first or second portions.
• the interface may be substantially devoid of oxide.
• the lifetime killing regions are preferably vertically spaced from the bonded interface by a predetermined distance, such as about 10 microns.
• Each of the lifetime killing regions may comprise at least one of defects and implanted impurities.
• the regions may in the form of circles of about 2-20 m in diameter and spaced about 5-20 m apart.
• each of the lifetime killing regions may comprise a strip region having a width of about 2 to 20 microns. The adjacent strip regions may be spaced about
• the semiconductor device may comprise a first laterally extending portion having a first conductivity type dopant; a second laterally extending portion on the first portion, the second portion also having the first conductivity type dopant; and at least one doped region of second conductivity type formed in the first portion adjacent an interface between the first and second portions and defining at least one PN junction.
• a conductive layer may be positioned between the at least one doped region and the second portion to lower a resistance of the PN junction.
• the conductive layer may be a metal or suicide, for example.
• One implementation of the PN junction may be to provide junctions spaced apart so as to define a vertical junction field-effect transistor.
• the conductive layer may also be arranged in a grid so that the device is a permeable base transistor.
• At least one of the first and second portions may be silicon.
• a bonded interface may be provided between the first and second portions. And the bonded interface is preferably substantially devoid of oxide.
• the invention is also directed to a semiconductor device comprising: a first laterally extending portion having a first conductivity type dopant; a second laterally extending portion on the first portion, the second portion also having the first conductivity type dopant; and a third laterally extending portion on the second portion and having a second conductivity type dopant.
• One of the first and second portions preferably has a dopant concentration greater than a dopant concentration of the third layer.
• the device preferably includes a first active control device on an outer surface of the first portion and a second active control device on an outer surface of the third portion.
• FIG. 1 is a flow chart illustrating a method of semiconductor device fabrication in accordance with the present invention.
• FIGS. 2-5 are cross-sectional views of one substrate being processed in accordance with the present invention.
• FIG. 6 is a cross-sectional view of an IGBT in accordance with the present invention and produced using the steps illustrated in FIGS. 2-5.
• FIG. 7 is a graph of resistance characteristics for an N-N hydrophobically bonded wafers as a function of anneal temperature.
• FIG. 8 is a graph of resistance versus inverse die area for N-N hydrophobically bonded wafers annealed at 400 C, and wherein the solid line represents expected resistance.
• FIG. 9 is a graph of resistance versus inverse die area of P-P hydrophobically bonded wafers annealed at 400 C, and wherein the solid line represents expected resistance.
• FIG.10 is a graph of forward and reverse current-voltage characteristics for twenty diodes fabricated from hydrophobically bonding P-type and N-type silicon wafers.
• FIG. 11 is a graph of diode ideality characteristics versus forward bias as a function of diode area for hydrophobically bonded P-N junctions.
• FIG. 12 is a graph of bond strength as a function of anneal time for hydrophobically bonded wafers annealed at 400 C, and wherein the dashed line indicates 800 ergs/ cm 2 as is needed for sawing and processing, and wherein the solid line is a least-squares fit to A+Blog(x).
• FIG. 13 is a cross-sectional view of a bonding P-N junction area between two substrates in accordance with the present invention.
• FIG.14 is a cross-sectional view of a pair of P-N junctions at the direct bond interface that may be used to form a vertical JFET that can switch the conduction of current across the bond interface.
• FIG. 15 is a cross-sectional view of a direct bonded IGBT including a thin N+ SiGe layer in accordance with the present invention.
• FIG. 16 is a cross-sectional view of a direct bonded IGBT including an ultra-thin ion implanted or epitaxially grown N+ buffer layer in accordance with the present invention.
• FIG. 17 is a graph of doping concentrations versus distance near the N+ buffer layer and P-type emitter anode of an IGBT or MCT in accordance with the present invention.
• FIG. 18 is a graph of doping concentrations versus distance near the N+ buffer layer and P-type emitter anode that further includes a P-type epitaxial layer grown on the P substrate in accordance with the present invention.
• FIG. 19 is a schematic cross-sectional view of the bond interface area and further illustrating localized recombination areas in accordance with the present invention.
• FIG. 20 is a cross-sectional view of an anode side substrate including an N+ buffer epitaxial layer in accordance with the present invention.
• FIG. 21 is a cross-sectional view of the anode side substrate as shown in FIG. 20 after further processing and being joined to a cathode side substrate.
• FIG. 22 is a cross-sectional view of an anode side substrate illustrating high energy implantation to form the N+ buffer layer in accordance with the present invention.
• FIG. 23 is a cross-sectional view of the -mode side substrate as shown in FIG. 22 after further processing and being joined to a cathode side substrate.
• FIG. 24 is a cross-sectional view of an anode side substrate including a N+ buffer layer near the P body diffusions in accordance with the present invention.
• FIG. 25 is a cross-sectional view of the anode side substrate as shown in FIG. 24 after further processing being joined to a cathode side substrate.
• FIG. 26 is a cross-sectional view of an anode side substrate being joined to an SOI substrate in accordance with the present invention.
• FIG. 27 is a cross-sectional view of the anode side substrate as shown in FIG. 26 being joined to a cathode side substrate.
• FIG. 28 is a cross-sectional view of an anode side substrate including an N+ buffer layer and base float zone mounted to an SOI substrate in accordance with the present invention.
• FIG. 29 is a cross-sectional view of the anode side substrate as shown in FIG. 28 after further processing and being joined to a cathode side substrate.
• first and second wafers are processed, with the processing blocks for the first wafer being identified in FIG. 1 with the suffix "a", and those to the second wafer designated with "b". Accordingly, only those steps for the first wafer will be described in detail, since those of skill in the art will readily appreciate similar steps may optionally also be performed in this embodiment on the second wafer also.
• a getting layer is formed on the second side, side B, of the first wafer at Block 52a.
• the gettering layer will be effective in removing contaminants, such as boron, for example, as will be readily appreciated by those skilled in the art.
• the gettering region will diffuse lifetime killing transition metals from the bulk of the wafer to the gettering site prior to thinning and direct bonding.
• typical gettering techniques include phosphorous diffusion, ion implantations or argon or carbon, and/ or polysilicon deposition on side B of the wafer.
• one or more active devices may be formed in a first side, i.e. side A, of the first wafer.
• the active devices may include one or more doped regions as may be formed by conventional techniques as will be readily appreciated by those skilled in the art.
• metal interconnects may also be formed, as described in greater detail below.
• the typical processing steps may include one or more of ion implantations, diffusions, metal depositions, polysilicon deposition, suicide formation, oxide growth, etc.
• the same or different types of MOSFET current controlled devices may be fabricated on the first and second substrates.
• the devices formed in accordance with the invention are directed to having current conduction in a vertical direction, that is, perpendicular to the interface formed between the bonded together surfaces as will be readily appreciated by those skilled in the art.
• a vertical direction that is, perpendicular to the interface formed between the bonded together surfaces
• the device has an N- base, then it would be desirable to inject a high density of holes (minority carriers) into the N- base to cause conductivity modulation.
• the condition for conductivity modulation is that the electron and hole densities in the base are equal.
• Conductivity modulation causes the electron density to increase much beyond its equilibrium value and thereby lowers the resistance of the N- base significantly as will be readily appreciated by those skilled in the art.
• the first wafer may be secured to a handling wafer or support film (Block 56a) and at Block 58a, the first wafer is thinned on side B, thereby removing the gettering layer and the contaminants contained therein.
• the wafer may be thinned by grinding, for example, to reduce the thickness to less than about 200 m, although even thinner wafers may be preferred for some applications.
• the handling wafer or support film may be removed after thinning.
• Side B may be polished and cleaned at Block 60a taking care to minimize hydrocarbon voids and to reduce oxygen at the ultimate bonding interface. If metal is exposed on the surface, such as metal bonding pads, it may be advantageous to protect the metal from chemicals used in the cleaning of the wafers.
• One possible technique is to deposit a protective insulator layer that is resistant to the chemicals. The insulator layer could be removed after the wafers are bonded.
• the polishing such as using chemical mechanical polishing (CMP), may be used so that the side B surface has a root-mean-square (RMS) surface roughness less than about 1 run. A surface roughness of less than about 10 nm is desirable for direct bonding the two substrates together.
• CMP chemical mechanical polishing
• the bonding could also be carried out at elevated temperatures of 200 to 400 C, for example, as will be readily appreciated by those skilled in the art.
• the cleaning is to remove hydrocarbons, organics, and metallic impurities from the surface.
• the cleaning process generally may use chemicals such as those used in RCA clean and piranha cleans as would be readily appreciated by those skilled in the art. Plasmas, UV, ozone, and laser irradiations may also be used to clean the surface prior to bonding.
• etching may be performed using a dilute hydrofluoric (HF) acid. It may also be desirable to minimize the native oxide regrowth prior to bonding.
• HF dilute hydrofluoric
• silicon one approach is to hydrogen terminate the silicon surface using a dilute HF etch followed by either no water rinse or minimal water rinse.
• Another more complicated approach to minimize native oxide is to direct bond the two wafers in either a vacuum, or in an ambient such as purified nitrogen, argon, or hydrogen that minimal concentrations of oxygen.
• Possible bond anneal ambients include nitrogen, oxygen, argon, hydrogen.
• One possible mechanism by which the bond strength increases with anneal time is that the hydrogen diffuses laterally along the bond interface and out of the wafer. The ambient may affect how readily the hydrogen diffuses laterally.
• the native oxide can impede the current flow across the interface.
• the hydrophobic approach may be susceptible to hydrocarbon absorption whereas a hydrophilic bonding approach in which there is a thin oxide at the interface may not be susceptible to hydrocarbon absorption.
• the oxide layer may be desirably less than about 1 run for satisfactory operation.
• a side B of the first wafer may then optionally be implanted, such as for lifetime killing and/ or with dopants for layers in the power device as will also be described in greater detail below.
• implanted such as for lifetime killing and/ or with dopants for layers in the power device as will also be described in greater detail below.
• ion implantation of protons, helium, carbon, argon, oxygen, etc. may be used.
• Lifetime killing metals such as platinum or gold may also be implanted or diffused into the surface.
• the temperature required to diffuse the metals may be generally greater than about 450 C, therefore, it may not be possible to have aluminum on the back surf ace during the anneal.
• the forward voltage versus turn-off time it may be desirable in power devices to have localized regions of lifetime killing rather than uniform lifetime killing.
• a photolithography step, or metal mesh may be used to confine higher energy protons laterally within certain regions.
• a photolithographic step may be used to define the location of the ion implanted dopant.
• the wafer typically includes a plurality of individual die thereon, these die can be tested (Block 64a) and the results later used to correlate with the second substrate to thereby increase the overall process yield.
• Yet another aspect of the invention relates to cutting the wafer along the outer streets (Block 66a). This will allow the precise alignment of the first and second wafers at Block 68.
• the wafers once properly aligned, may be bonded by bringing same together at a center point and allowing atomic bonding to tend to bring the wafers together extending outwardly from the center. In some embodiments, a relatively high or ultrahigh vacuum may be desirable for the bonding process.
• the two wafers may also be aligned based upon the crystal orientations of the two wafers as will also be appreciated by those skilled in the art.
• a low temperature anneal may be performed.
• the temperature may be less than about 800 C if aluminum is to be added later, but may be 450 C or less if the aluminum metal interconnects are already provided. If a barrier metal layer is provided between the aluminum and the silicon substrate a higher temperature anneal, e.g., about 450-550 C may be permitted. Perhaps the best overall gains are achieved if the two wafers are completely processed prior to bonding.
• a significant requirement of the bond strength is that it be strong enough for a 400 C anneal to allow sawing or dicing. Accordingly, a bond interface energy of 800 ergs/ cm 2 is required based on experience. A 400 C anneal is potentially low enough to slow the formation of a Si-Al eutectic as will be appreciated by those skilled in the art. In other embodiments, laser cutting may be used so that a lower bond strength may be tolerated as will be appreciated by those skilled in the art.
• the individual device dies/ circuits may be diced from the bonded together wafers using conventional techniques as will be readily appreciated by those skilled in the art.
• the power switching devices produced in accordance with the present invention have current transport across the bonding interface, that is, perpendicular to the bonding interface.
• the method of the present invention allows the fabrication of double-sided MOSFET controlled power switching devices with a significant reduction in the number of sequential process steps by about a factor of two compared to conventional techniques.
• the direct bonding approaches allows current production recipes for fabricating single-side power devices to be used, and thus a separate process sequence does not need to be developed.
• the present invention avoids the critical control of thermal budgets as in conventional processing, since an anneal is optimized for one dopant on the top side of the substrate but will not likely be optimized for another on the back side.
• the conventional approach may also lead to yield losses from scratches, etc.
• the present invention overcomes these drawbacks and also allows for a gettering operation to remove metal impurities prior to bonding.
• the yield can also be optimized by mapping the working die in the two wafers and aligning the wafers for best yield.
• the direct bonding after processing of the invention can be used to implement high performance IGBTs, MOSFETs, and and MCTs, for example.
• the directed bonded devices may also have an ultra-thin N+ buffer layer which will provide significant improvements in the turn-off time compared to alternate approaches as described in greater detail below.
• the direct bonded IGBTs and MCTs will have a novel feature of positive temperature coefficient for forward voltage which arises from a negative temperature coefficient for current gain.
• a silicon MOSFET current control power device in a first substrate to a second substrate that includes SiC material.
• Other candidates for the material of the second substrate may include GaN, InP, and Ga As.
• Wide bandgap materials, such as SiC generally have a high critical field of electrical breakdown, and also have high saturated drift velocities. Thus, wide bandgap materials are often desirable to be used to support most of the high voltage drop across depletion layers in power devices.
• Another reason for selecting a material for the second substrate to be different than silicon is to provide a high thermal conductivity- Thus, SiC, which has a three times higher thermal conductivity than silicon may be used for the second substrate.
• two or more non-silicon substrates may be processed and bonded in accordance with the present invention.
• one aspect of the present invention is directed to an approach to achieve a double-sided MOSFET controlled power switching device using low temperature direct semiconductor-semiconductor wafer bonding by fabricating two MOSFET current control devices on two separate wafers, thinning the wafers from the backside to approximately 200 m, and then performing aligned bonding of the wagers.
• the greatest advantage of this approach occurs when the two wafers are almost completely processed prior to bonding.
• aluminum interconnects are present on the surface and the maximum allowed bonding anneal temperature is about 450 C. If a barrier metal is used between the aluminum and silicon junction, bond anneal temperature of about 450-550 C may be used.
• FIG. 2 Initial processing of a first substrate 80 is shown in FIG. 2.
• the first substrate 80 will be direct bonded to a second substrate 95 to produce a double-sided MCT 110 (FIG. 6) as will be appreciated by those skilled in the art.
• a gettering implant 91 is provided as described in detail above.
• various dopant regions are formed in the upper surface of the substrate 80, along with the illustrated second gate region 81.
• the illustrated processed portion further includes an N-type base 82, an N-type buffer layer 83 on the N-type base, and a P+ emitter
• the substrate 80 also includes an anode layer 86 and an N+ region 87.
• the first substrate 80 is joined to the handling substrate 90 or wafer, and the gettering layer 91 is removed by thinning to produce the intermediate structure shown in FIG.4.
• a lifetime killing implant 92 is schematically illustrated being formed in the first substrate 80 in FIG. 5.
• the thus processed first substrate 80 is joined to a second substrate 95 after cleaning, direct bonding, and the low temperature anneal.
• the second substrate 95 illustratively includes an N- type base 96, a P-type base 97 on the N-type base, an N+ emitter 98 on the N-type base, a first gate 99, a cathode layer 100, and the illustrated P+ region 101.
• the second substrate 95 also illustratively includes the lifetime killing implants 102.
• An interface 103 is schematically illustrated between the first and second substrates 80, 95.
• the first requirement is the need for near ideal current conduction across the bonding interface. This requires that there be minimal native oxide at the bonding interface. Prior researchers have demonstrated that by using hydrophobic bonding in which the silicon surfaces are hydrogen terminated, bonding interfaces with minimal native oxide can be obtained. Also special considerations should be given to reducing boron and heavy metal contamination during the surface cleaning operations. Another requirement is that bubbles and microvoids must be minimized at the bonding interface.
• Low temperature hydrophobically bonded wafers seem to be particularly susceptible to hydrocarbon generated voids, and thus special attention should also be paid to cleaning procedures that remove hydrocarbons. Yet another requirement is low minority carrier recombination at the bonding interface.
• FIGS. 7 through 12 results from bonding experiments show that N-type to N-type and P-type to P-type silicon ⁇ 100> wafers can be hydrophobically bonded using low temperature anneals without producing potential barriers in either the conduction or valence bands.
• the cleaning procedure that was used to achieve a hydrogen terminated surface included combination of O 2 plasma and piranha cleans followed by a 10:1 HF dip, without a water rinse following the HF dip. Electrical data for the N-type to N-type wafers for various anneal temperatures is shown in FIG. 7.
• FIG. 8 shows the resistance versus inverse area and a scatter plot of resistance values for N-type to N-type wafers annealed at 400 C for 9 hours.
• FIG. 9 shows a similar result for a P-type to P-type bond annealed at 400 C for 24 hours.
• FIG. 10 shows forward and reverse current-voltage characteristics for twenty diodes fabricated from hydrophobically bonded P-type to N-type wafers for a 600 C bond anneal.
• the leakage current density is approximately 40 nA/cm 2 for hydrophobically bonded wafers.
• FIG. 11 shows the calculated ideality factor for small forward biases for PN diodes with varying areas. The measurements show a strong dependence on area, with the smallest area diodes having the largest ideality factor values. Increases in ideality factor values above 1.0 are typically due to minority carrier recombination, either at the metallurgical junction or the perimeter of the device.
• the measured dependence on area shows that the high ideality factor is due to recombination that is occurring at the edges of the unpassivated, sawed mesas.
• the idealitv factor approaches a value of 1.0 as the area of the diode increases.
• These are the best ideality factors for low temperature bonded devices.
• a significant requirement, in addition to the electrical characteristics, is that the bond strength be strong enough for a 400 C anneal to allow sawing or dicing of the switching power devices. From experience, a bond interface energy of greater than 800 ergs/ cm 2 is required to provide sufficient bond strength.
• FIG. 12 shows that the bond strength increases logarithmically with anneal time, indicating a first order reaction for the bond kinetics at 400 C. A 400 C anneal was chose because it is potentially low enough to slow the formation of Si-Al eutectic.
• a bond interface energy of 1000 ergs/ cm 2 was obtained for a 9 hour 400 C anneal which is sufficient to allow sawing of wafers.
• a possible process to implement a low resistance metal or silicide strapped PN junction is to use a photolithography step to define an implant of P-type dopant 121 into the N-type substrate 122 of the intermediate structure 120 shown in FIG. 13.
• a second photostep is used to define the location of the metal or silicide strip 123 within the P-type dopant region 121.
• the resist mask can be used to etch about 100 run of silicon.
• tungsten silicide 123 About 30 nm of tungsten can be evaporated. The excess tungsten that is on the surface of the resist can be removed by lift-off, followed by an anneal to form tungsten silicide 123. As will be readily appreciated by those skilled in the art, an alternate approach may be to use a poUshing technique to polish silicide that is formed above the silicon surface back planar with the adjacent silicon surface. The silicide can also be used to lower the resistance of a blanket doping layer, such as lowering the P-type base resistance of a Gate Turn-off Thyristor.
• a grid of low resistance PN junctions can be used as gates of a vertical JFET 130.
• the pair of illustrated junctions 131, 132 may be used to modulate a current flow perpendicular to the junctions, that is, across the interface 134.
• a plurality of such junctions could also be formed.
• the depletion regions 135 are formed around the P doped regions 123 as will be appreciated by those skilled in the art.
• MOS gates could be formed on the sides of trenches and operate in the depletion mode in which the current is being conducted between the channels at zero source-to-gate bias, and the gate is biased to increase the depletion layer to turn off the device.
• a grid of silicide lines at the interface between the two substrates 125, 122 can be used to form a permeable base transistor in which reverse biased Schottky diodes are used to modulate the current flow perpendicular to the grid of silicide lines 123 as will be readily appreciated by those skilled in the art.
• Vias may be chemically or plasma etched from the top surface of either substrate to the silicide or metal layer using the silicide or metal layer as an etch stop layer.
• technique may be to laser drill a via through the upper substrate 125 (FIG. 13), and stop at the metal or silicide layer.
• Yet another aspect of the invention is the ability to epitaxially grow semiconductor layers on either or both of the substrates prior to bonding. If aluminum interconnects are on the substrates, the epitaxial growth should be at a temperature of less than 450 C, and less than 450 to 550 C if a barrier metal layer is used as described above.
• the completed IGBT 140 also includes an anode layer 142, a P+ substrate emitter layer 143 adjacent the anode, and the SiGe buffer layer 141 adjacent the interface 144.
• the upper substrate 150 includes an external emitter layer 151, a gate layer 152 and its underlying insulator layer 153.
• the upper substrate 150 also includes the N-type base 155 including the lifetime killing implant 156 as discussed above.
• the other doped regions of the upper substrate 150 will be readily understood by those skilled in the art without further description.
• the SiGe base-emitter heterojunction that is properly configured may also have negative temperature coefficient for current gain, and, thus, positive temperature coefficient for forward voltage. This characteristic will provide short circuit protection and help prevent thermal run away as will be readily appreciated by those skilled in the art.
• the N+ buffer layer may be fabricated by implanting a thin, about 200 nm thick, layer with a concentration of about 1X10 19 cm '3 of N- type dopant, such as arsenic, antimony or phosphorous into the surface of the P+ substrate.
• the P+ substrate may have a concentration of about 3X10 18 to 1X10 19 cm '3 of P-type dopant.
• the substrate can then be annealed at a temperature of from about 900 to 1000 C to anneal the defects created during ion implantation as would be readily appreciated by those skilled in the art.
• a techniques to prevent the injection is to reduce the injection efficiency of holes in this region.
• a photolithography step can be performed to define a thick N+ ion implant into the regions outside of the active area to thereby reduce the injection efficiency.
• Other techniques could also be used, such as to have a defined oxide barrier at the bonding interface 144.
• the thin epitaxial layers of SiGe or high N-type concentration dopant can provide key advantages for the high performance IGBTs or MCTs.
• thin, high concentration dopant layers can be used for the N+ buffer of an IGBT to achieve short turn-off time and negative temperature coefficient for current gain.
• the substrate was used as an N+ emitter, then the device would be processed to have a P+ ultra-thin buffer layer as will be readily appreciated by those skilled in the art.
• Electron Injection Efficiency J e /J h / arid
• the device physics for a P+ emitter with an N+ base buffer layer that is higher in concentration than the P+ emitter results in a negative temperature coefficient for current gain.
• the profiles are illustrated in FIG. 17 for the various portions of the device near the interface. The interface could be on either side of the N+ buffer as will be appreciated by those skilled in the art.
• the negative temperature coefficient for current gain implies that the current in an IGBT or MCT decreases as the temperature increases. The current decreasing at higher temperatures means that the forward voltage will increase. Thus, the IGBT and MCT will have a positive temperature coefficient for forward voltage.
• a positive temperature coefficient for forward voltage for both IGBTs and MCTs is important to prevent thermal runaway and providing short circuit protection.
• the N+ buffer layer it is important to have an N+ buffer concentration that is higher than the P+ emitter; however, the N+ buffer must also be thin enough to provide sufficient current gain for the IGBT and MCT P+ backside emitters.
• MOSFET current control devices has a particular advantage for making an IGBT or MCT with the N+ buffer concentration higher than the P+ substrate, and being relatively thin to produce acceptable current gain for the backside P+ emitter.
• a common approach used to presently make IGBTs or MCTs is to grow the N+ buffer using high temperature epitaxial growth. The high temperature epitaxial growth will diffuse the N+ dopant to make a thick buffer layer (10 to 20 m thick). Because there is a maximum allowed N+ buffer integrated doping concentration, it is generally necessary that the N+ concentration be lower than the P+ substrate concentration to obtain backside P+ emitter current gain. Also, high temperature (1100 to 1200 C) anneals are typically used to diffuse P-type dopant to produce deep P-type junctions for field termination.
• this anneal will also diffuse the N-type dopant increasing the N+ buffer width. Since the integrated N+ buffer dopant (concentration integrated over thickness) must be low enough to provide sufficient gain for proper IGBT and MCT operation, a thicker N+ buffer must have a lower concentration. The high temperature field termination anneal thus makes it difficult to have an N+ buffer with a concentration greater than the P+ emitter concentration.
• a preferred approach to achieve an N+ buffer with a dopant concentration higher than the dopant concentration in the P+ emitter is to ion implant N-type ions (arsenic, phosphorous, antimony) into a P+ substrate that has a concentration of approximately 3xl0 18 cm ⁇ 3 . Because the ion implanted N-type dopant concentration is greater than the P+ doping concentration, the N- type dopant will over compensate the P+ doping concentration, and a thin, N+ layer can be formed on the pre-bonded surface of the substrate.
• the N+ buffer implant can also be made into the prebonded surface of the other substrate. Since one of the substrates will have a thickness between 100 ⁇ m and 200 ⁇ m and may have metal interconnects on the surface thereof, it may generally be more difficult to perform the ion implant anneal.
• An alternate approach to make a thin N+ buffer with a concentration higher than the P+ emitter is to epitaxially grow the N+ buffer on the prebonded surface of either substrate before bonding.
• the P+ substrate concentration is too high (such that a P+ to N+ junction would have too low of a breakdown voltage, have too much leakage current, or be too high concentration so that it is difficult to have an N+ buffer concentration greater than the P+ concentration) then an alternate approach is to first grow a lower concentration P-type epitaxial layer on the P+ substrate as understood with reference to FIG.18. It is likely necessary to optimize the thickness and concentration of the P-type epitaxial layer. If the P-type epitaxial layer is thick enough (must be thicker than the diffusion length of electrons into the P+ emitter), then the doping concentration of the P-type epitaxial layer will determine the effective emitter concentration that determines the injection efficiency.
• the P-type epitaxial layer may be tens of microns thick and have a doping concentration of approximately lxl0 17 cm "3 .
• An approach for a thinner P-type epitaxial layer is to perform process steps to reduce the minority carrier lifetime in the P-type epitaxial layer.
• the diffusion length for electrons into the P-type emitter is determined by the recombination time for electrons in the P-type emitter.
• relatively low N+ buffer concentrations are required to meet the criterion that the N+ buffer concentration be greater than the P+ emitter concentration.
• An N+ buff er ion implant can now be made into the epitaxially grown P-type layer, or an N+ epitaxial layer grown on the P-type epitaxial layer growth. Direct bonding of the two substrates can now be performed.
• an alternate way to achieve negative temperature coefficient for current gain is to epitaxially grow a SiGe strained N+ buffer layer on the prebond surface of either substrate. Thin epitaxial layers of SiGe or high N-type concentration dopant can provide
• thin, high concentration dopant layers can be used for the N+ buffer of an IGBT to achieve short turn-of time and negative temperature coefficient for current gain.
• thin layers of SiGe can act as the N+ buffer layer for an IGBT or MCT. Since the N+ buffer layer will be thin, the IGBT will have fast turn-off time.
• a SiGe base-emitter heterojunction that is properly designed also has negative temperature coefficient for current gain and thus positive temperature coefficient for forward voltage. This characteristic will provide short circuit protection and help prevent thermal run away. If a first substrate is used as an N+ emitter, then the device would be processed to have a P+ buffer made into the second substrate.
• a first approach is an ultia-thinning approach, wherein the key requirement to achieve an N+ buffer concentration higher than P+ emitter concentration is to minimize temperature steps after the N+ buffer formation.
• a preferred fabrication approach is to:
• a technique to achieve almost complete activation of the implanted boron and phosphorous ions even with metal interconnects present on the front surface is to use repetitive short pulses of excimer laser illumination.
• An alternate technique to implement an N+ buffer with a higher concentration than the P+ emitter on the back surface of the thinned substrate is to epitaxially grow the N+ buffer and P+ emitter on the back surface at approximately 500 C. If there are barrier metals under the aluminum interconnect, then growth temperatures of 500
• MBE Molecular Beam Epitaxially growth techniques include Metal Organic Chemical Vapor Deposition (MOCVD) and Ultra-high Vacuum Chemical Vapor Deposition (UHVCVD) . It is also possible to deposit N+ and P+ amorphous silicon layers and then regrow the layer's single crystal layer at approximately 500 C using solid phase epitaxial regrowth.
• An alternate technique to form a P+ emitter on the back surface is to use a P+ polysilicon layer. This type of emitter sometimes has a thin native oxide between the polysilicon layer and the single crystal layer that can provide an increase in current gain and in some cases, less change in current gain with temperature.
• High anneal temperatures can be used to activate the ion implanted dopants on the back surface if metal interconnects are not present on the front surface in step 1 (in this case, the substrate is processed to just before metal deposition).
• the substrate is thinned, boron and phosphorous are implanted into the back surface, and an anneal at 800 to 900 C is used to activate the implanted dopants. It will be necessary to perform the process steps and photolithography steps on the front surface to complete the process to make an IGBT. The difficult aspects of this process are that the wafer is approximately 100 ⁇ m thick at this point.
• the thinned IGBT or MCT substrate could be permanently bonded or soldered to a metal substrate, and the remaining processing steps completed.
• a technique for making a relatively narrow N+ buffer layer is to implant an N+ buffer into the prebond surface of a low N-type dopant concentration substrate, anneal the implant dopant, and then direct bond the substrate to a P+ emitter substrate. It is also possible to implant the N+ buffer dopant into the P+ substrate and over compensate the P+ dopant concentration.
• the N-type substrate is then thinned to approximately 100 ⁇ m and the diffusion and process steps to make an IGBT or MCT are next performed.
• the highest temperature step is a 1100 to 1200 C anneal to diffuse boron to make a deep junction for field termination. This high temperature step will cause the N+ buffer to diffuse and thus increase the thickness of the N+ buffer. With proper design and anneal temperature steps, it may be possible to produce an IGBT or MOSFET with the N+ buffer concentration higher than the P+ emitter concentration.
• One approach to achieve an N+ buffer concentration higher than the P+ emitter concentration is to grow a relatively thick (10 ⁇ m) P-type epitaxial layer with a dopant concentration of approximately lxl0 17 cm 3 on the P+ substrate as illustrated in FIG.18.
• the low concentration N-type substrate with the N+ buffer implanted into the prebond surface is then
• the effective dopant concentration for the P-type emitter injection efficiency will be the P-type epitaxial layer dopant concentration rather than the P+ substrate dopant concentration.
• Yet another approach to achieve an N+ buffer concentration higher than the P+ emitter concentration is to grow a relatively thick (10 ⁇ m - 20 ⁇ m) P-type epitaxial layer with a dopant concentration of approximately lxl0 1 cm 3 on the P+ substrate as shown in FIG. 18. This epitaxial growth is followed by the epitaxial growth of an N+ buffer, and finally the epitaxial growth of the N-base layer. Since the epitaxial growth is a very high temperature process, it is difficult to achieve a thin N+ buffer and thus, it is difficult to achieve the condition that the N+ buffer concentration is greater than the P+ emitter concentration. The effective dopant concentration for the P-type emitter injection efficiency will be the P-type epitaxial layer dopant concentration rather than the P+ substrate dopant concentration.
• Another aspect of the invention relates to lateral localized lifetime killing near the direct bonded interface of IGBTs and MCTs.
• the device 170 also is formed from an upper substrate 172 bonded to the lower substrate 171 at the schematically illustrated interface 173.
• the lower substrate 171 illustratively includes an N+ doped portion 176 and an N doped portion 177.
• a photolithography step or metal mesh would be used to define high energy proton (or other lifetime killing implants, defect generation techniques, or transition metal diffusions) so that they are laterally confined to certain regions.
• the region of the power device that has lifetime killing in it typically has higher forward voltage since many of the injected carriers recombine in the lifetime killing region rather than transit from anode to cathode.
• By laterally confining the ion implanted lifetime killing there will be regions of the device 170 that have no recombination of carriers as they transit from the anode to the cathode, and thus this portion of the device will have low (or ideal) forward voltage.
• the minority carriers that are in the base layer typically are removed from the base layer either by diffusing to the emitter-base junction or by diffusing to recombination centers.
• the lifetime killing implants 175 are defined laterally, then the minority carriers in the base will diffuse laterally to the short lifetime recombination region. Since it is possible to photodefine resist that is approximately 10 ⁇ m thick to approximately 3 ⁇ m feature sizes, the lifetime killing region can consist of a grid of 2-3 ⁇ m diameter circular lifetime killing regions 175 that are buried approximately 10 ⁇ m from the prebond interface and are separated approximately every 10 ⁇ m laterally (parallel lines 2-3 ⁇ m wide separated every 10 ⁇ m is also
• the effective distance the minority carriers then have to travel laterally to recombine is approximately 5 ⁇ m. Because of this short distance, the recombination time will be short.
• PN junction area will not have lifetime killing, and an almost ideal turn-off time can be obtained while still achieving a fast turn-off time by having the minority carriers recombine laterally.
• the above described technique of laterally confining the lifetime killing also has similar advantages for non-bonded devices such as PN diodes. It is desirable for a diode to have a low forward voltage so that it is desirable to have much of the diode area not have any lifetime killing. It is also desirable to have a fast turn-off time. Since much of the stored charge in a diode is near the PN junction, laterally confined lifetime killing regions (potentially implemented by high energy helium implant) that are approximately 4 ⁇ m to 8 ⁇ m into the N- type base from the PN junction are desirable.
• the lifetime killing regions that are spaced approximately 10 ⁇ m apart laterally will provide a high percentage of the area that has no lifetime killing, but will yet provide fast turn-off time by allowing the carriers to diffuse laterally approximately 5 ⁇ m to recombine laterally. Lateral confined lifetime killing also has similar advantages for a thinned IGBT in which the P+ emitter anode is formed on the backside of an IGBT or MCT device structure on the cathode side.
• a common approach for fabricating a punchthrough IGBT is by epitaxially growing the N+ buffer and N- base layer on a P+ substrate. The processing steps for the diffusion and MOSFET control devices of the IGBT or MCT near the cathode are now performed. Because of the high temperature of the epitaxial layer growth (typically > 1000 C), lifetime killing, such as proton or HE implants or transition metal diffusion are typically performed after the epitaxial growth. There are several ion implantation lifetime killing techniques, however, than can remain as minority carrier recombination centers after the high temperature epitaxial growth.
• a key requirement for these lifetime killing techniques is that the ion implantation to cause defects at the surface on which epitaxial layers will be grown so that good quality epitaxial layers can be grown.
• One technique is to implant He ions with sufficient energy so that they are buried beneath the surface about 0.5 m and at sufficient doses (lxlO 16 cm “3 ) so that when the substrate is heated, the He gas expands and creates bubbles beneath the silicon surface. These bubbles will survive the epitaxial layer growth. The silicon side walls of the bubbles can then
• the approaches to localize the lifetime killing in the lateral direction is to use a photolightographically defined resist masking layer so that the He implants can define a grind of 2-3 m diameter circular lifetime killing regions that are buried about 0.5 m beneath the surface on which the epitaxial layer will be grown and are separated about every 10 to 20 m laterally. Parallel lines 2-3 m wide may be separated every 10 -20 m as an option.
• the process for forming and growing the epitaxial layer and lateral lifetime killing is thus:
• ion implantation lifetime killing species that can be used in a similar manner on the He implant described above are: a.) oxygen implant to create oxygen precipitates that are buried beneath the silicon surface that will act as recombination centers.
• the anneal to create oxygen precipitates typically involves a long time anneal at 650 C to nucleate the precipitates, an anneal at 950 C to grow the precipitates, and then possibly an anneal at 1100 C to grow stacking faults.
• the ion implantation dose typically required to create a high density of oxygen is typically less than about lxl 0 15 cm '3 and will not roughen the surface as much as is the case for the He implant that causes the bubbles. b).
• Carbon implants to create carbon precipitates beneath the surface c) .
• the following description relates to alternate approaches to implement: 1) an N+ buffer near the P-type body on the anode side of the device, 2) a positive temperature coefficient for forward voltage for a double-side power device, 3) the use of a silicon-on-insulator (SOI)
• a number of power switching applications only require forward blocking operation and do not require reverse blocking operation.
• active device structure and field termination are required on the cathode side of the device to achieve high breakdown voltage, however, only a low breakdown voltage device is required on the anode side of the device.
• the principle methods for implementing an N+ buffer in a double-sided device include those described herein for direct bonding after the substrates have been processed. There are several methods which are also appropriate to implementing a power switching device which is fabricated using conventional double-sided semiconductor processing.
• Epitaxial growth may be used on the anode side substrate prior to fabrication of the anode side active devices.
• an epitaxial layer including an N-base layer 182 and an N+ buffer layer 181 and finally a lower doped N layer is grown on a float zone N-type substrate 183 as shown in FIG. 20.
• a two doping concentration buffer may be desired for robustness.
• the two doping concentration buffer would include the epitaxial growth of a wide, lower doping concentration N-type buffer and then a thin N+ buffer.
• the N+ buffer layer will generally be located about 2 m to 20 m from the top surface.
• the MOSFET current control devices located on the anode side of the device can then be fabricated within the P-type body 185 as illustrated.
• the P body will also form the emitter of a PNP bipolar transistor consisting of the P-type body emitter, N+ buffer/ N-type base, and P-type collector on the cathode side of the device.
• the direct bond approach used to fabricate a double-sided power device including the above described N+ buffer is to fabricate the anode side substrate as described above, thin to about 10 m to 200 m, polish and clean the surface, hydrogen terminate the surface, and then direct bond to a thinned and polished cathode side substrate 190 as shown in FIG. 21.
• the direct bond approach can readily form the two-step N buffer as described above by having the N-type substrate concentration of the anode side substrate 180 be the desired concentration of the lower concentration N-type buffer of the two step N-type buffer.
• a double-sided power device with an N+ buffer near the anode side current control device can be fabricated by
• a high energy implant of phosphorous may be used to form the N+ buffer region 181' of the substrate 180' as shown in FIG. 22.
• the other portions of the substrate 180' are the same as those described above for FIGS. 20 and 21, and need no further discussion herein.
• FIG. 23 further illustrates the bonding of the two substrates 190, 180' and needs no further description.
• a positive temperature coefficient for forward voltage for a double-sided power device can be obtained by having an N+ buffer concentration greater than the P-type body emitter concentration as described above. In this case it is generally desirable to have the N+ buffer layer 210 adjacent or near the P-type body 202 as shown in the anode side substrate 200 in FIG. 24.
• the N+ buffer layer is formed on the N-base float zone substrate portion 203.
• FIG. 25 illustrates the direct bonding of the thus formed anode side substrate 200 to the illustrated cathode side substrate 210.
• An alternate approach to implement an N+ buffer is to use silicon-on-insulator (SOI) technology.
• SOI silicon-on-insulator
• an N+ ion implant to form the N+ buffer 221 is made into one surface of a the anode side substrate 220 as shown in the top portion of FIG. 26.
• This substrate 220 will later be bonded to the silicon substrate 227 with oxide 226 on the surface thereof to form the SOI substrate 225 as shown in the lower portion of FIG. 26.
• the active device portion of the anode side substrate 220 is next fabricated as seen in the upper portion of FIG.26.
• the silicon substrate and SOI layer Prior to direct bonding to form the double-sided power device, the silicon substrate and SOI layer are removed by protecting the front surface of the wafer, grinding to within 50 m of the oxide layer, chemically etching the silicon and stopping the etch at the oxide layer, and finally chemically etching the oxide layer.
• An advantage of the SOI substrate is that the surface roughness should be sufficiently small so that a polishing operation is not required.
• the previously fabricated anode side substrate 220 can be direct bonded to the previously fabricated cathode side substrate 230 as shown in FIG. 27.
• An ion implanted N+ buffer may be formed into the prebond surface of an ultra-thin previously fabricated anode side substrate.
• an N+ ion implant is made into one of the prebond surface of either the anode side or cathode side substrate. It is generally desirable that the anode side substrate be about 3 m to 20 m thick to form the N+ buffer near the P-body.
• the ultra-thin anode side substrate can be implemented by the SOI approach described immediately above, grinding, and polishing, hydrogen ion implant layer splitting, and electrochemical etchstop plus polishing.
• a thin anode side substrate can be formed by epitaxially growing an N-type base layer, N+ buffer, N-type base layer on a P-type substrate and forming an anode side active device as described above.
• the electrochemical etch stop approach typically the P-type substrate is etched with the etch stopping within the PN junction depletion layer.
• thin active side substrates can be formed. It is generally necessary that the surface be polished to obtain a small enough surface roughness to direct bond to the cathode side substrate.
• the electrochemical etch stop technique requires a method to make electrical contact to the front side of the device while at the same time protecting the front side of the wafer.
• a potential approach is to perform both functions by using a conductive polymer.
• Another SOI approach to fabricate thin substrates for direct bonding to form a double- sided power device is to fabricate one or both sides of a double-sided power device in the top silicon layer of an SOI substrate, remove the substrate and oxide, and direct bond two previously fabricated substrates to form the device.
• the primary advantage of this approach is that it is not necessary to polish the prebond surface prior to direct bonding.
• the SOI approach to direct bonded double-sided power devices is useful whether or not an N+ buffer is included and is even useful if only forming a one sided IGBT or MCT device.
• the typical process is to direct bond an oxidized surface of small surface roughness ( ⁇ 1 nm), and prime the surface finish of the silicon wafer to a silicon handle substrate.
• the silicon surface that is adjacent the buried oxide layer has a small surface roughness.
• the approach to use an SOI substrate to form a double-sided power device is then to polish the top silicon layer to the desired thickness generally in the range of about 3 to 100 m thickness, fabricate the power switching device in the top silicon layer, remove the silicon handle substrate, remove the oxide layer, potentially ion implant into the prebond surface, and then direct bond two previously fabricated substrates to form a double-sided power switching device.
• This process may be better understood with reference to FIGS. 28 and 29.
• the anode side substrate 230 is bonded to the SOI substrate 240, and in FIG. 29, afer the SOI substrate is removed, the anode side substrate 230 is joined to the cathode side substrate 250.
• some high voltage power devices may require a silicon substrate thickness of 2 mm. This is a relatively thick substiate that can be formed by bonding together four 0.5 mm thick substrates, and with the top and bottom substrates having the processing already performed prior to bonding in accordance with the invention.
• the buffer is relatively thin and has a first conductivity type dopant concentration greater than a second conductivity type dopant concentration in adjacent emitter portions to provide a negative temperature coefficient for current gain and a positive temperature coefficient for forward voltage for the device.
• the buffer may be silicon or germanium.
• a low temperature bonded interface may be between the emitter and the buffer or the buffer and the base.
• Another embodiment of a device may include a laterally extending localized lifetime killing portion between oppositely doped first and second laterally extending portions.
• the localized lifetime killing portion may comprise a plurality of laterally confined and laterally spaced apart lifetime killing regions.
• Another device may include one or more PN junctions.

Landscapes

• Engineering & Computer Science (AREA)
• Microelectronics & Electronic Packaging (AREA)
• Power Engineering (AREA)
• Computer Hardware Design (AREA)
• Manufacturing & Machinery (AREA)
• Physics & Mathematics (AREA)
• Condensed Matter Physics & Semiconductors (AREA)
• General Physics & Mathematics (AREA)
• Ceramic Engineering (AREA)
• Chemical & Material Sciences (AREA)
• Crystallography & Structural Chemistry (AREA)
• Electrodes Of Semiconductors (AREA)
• Insulated Gate Type Field-Effect Transistor (AREA)
• Thyristors (AREA)
• Junction Field-Effect Transistors (AREA)

Applications Claiming Priority (7)

Publications (1)

Family

ID=27365101

Family Applications (1)

Country Status (4)

Families Citing this family (25)

Family Cites Families (6)

• 1999
  • 1999-03-09 JP JP2000536101A patent/JP2002507058A/ja not_active Withdrawn
  • 1999-03-09 EP EP99912328A patent/EP1062692A1/en not_active Withdrawn
  • 1999-03-09 KR KR1020007010106A patent/KR20010041822A/ko not_active Application Discontinuation
  • 1999-03-09 WO PCT/US1999/005066 patent/WO1999046809A1/en not_active Application Discontinuation

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events