JP2009010120A - Silicon carbide bipolar semiconductor device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To suppress a stacking fault and an area magnification thereof which make a defective nucleus existing near a front surface of an element as a starting point, and thereby to suppress an increase in forward voltage, in a silicon carbide bipolar semiconductor device with a mesa structure. <P>SOLUTION: The silicon carbide bipolar semiconductor device is constituted by laminating the following in order: a first conductivity-type silicon carbide single crystal substrate; a first conductivity-type silicon carbide drift layer; a second conductivity-type silicon carbide charge injection layer; and a second conductivity-type high doping layer with a doping density higher than that of the silicon carbide charge injection layer. This device is also characterized in that there is provided the mesa structure from the silicon carbide drift layer to an element surface at the high doping layer side, and the thickness of the high doping layer is 15 nm to 1.5 μm, and the maximum doping density in the high doping layer is 1×10<SP>20</SP>cm<SP>-3</SP>to 2×10<SP>21</SP>cm<SP>-3</SP>. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a bipolar semiconductor device, and more particularly to a silicon carbide bipolar semiconductor.

  Wide-gap semiconductor materials such as silicon carbide (SiC) have various excellent characteristics such as about 10 times higher dielectric breakdown strength than silicon (Si), and high breakdown voltage that requires high withstand voltage characteristics. It attracts attention as a suitable material for power semiconductor devices.

  Bipolar semiconductor elements such as pn diodes, bipolar transistors, GTO, and GCT have higher built-in voltages than unipolar semiconductor elements such as Schottky diodes and MOSFETs, but their resistance is greatly increased by conductivity modulation of the drift layer by minority carrier injection. Becomes smaller. Therefore, bipolar semiconductor elements are used to reduce loss in high voltage and large current regions such as power applications.

  When these bipolar semiconductor elements are made of SiC, it is possible to achieve performance that is significantly superior to Si elements. For example, in the case of a pn diode composed of SiC, in the case of a high breakdown voltage element of 10 kV, the forward voltage is about 1/3 lower than that of a pn diode composed of Si, and the reverse recovery time corresponding to the off-time speed is about The speed is as high as 1/20 or less, and the power loss can be reduced to about 1/5 or less, which can greatly contribute to energy saving.

It has also been reported that power loss is similarly reduced for SiC bipolar elements other than SiC pn diodes, such as SiC npn transistors, SiC SIAFETs, and SiC SIJFETs (Non-Patent Document 1). In addition, SiCGTO using a p-type semiconductor layer having an opposite polarity as a drift layer has also been reported (Non-Patent Document 2).

As conventional pn diodes using SiC, there are a planar high-voltage diode (Non-patent Document 3), a mesa-structured diode (Non-patent Document 4), and the like. FIG. 4 is a cross-sectional view schematically showing a pn diode having a mesa structure. In this pn diode, a pn junction between the p ⁺ type charge injection layer 4 for injecting charges and the n ⁻ type drift layer 3 is formed by an epitaxial growth technique, and then mesa is formed by reactive ion etching or the like for element isolation. A structure 10 is formed. The mesa structure 10 can be formed, for example, by removing both end portions of the p ⁺ type charge injection layer 4 having a thickness of about 3 μm formed on the entire surface of the n ⁻ type drift layer 3 to a depth of 4 μm by mesa etching. it can. In FIG. 4, an oxide film 7 for protecting the device surface, an anode electrode 8 provided on the p ⁺ type charge injection layer 4, and a cathode electrode 9 provided on the back surface of the n ⁺ substrate are shown. Yes. After the mesa structure 10 is formed, a pn diode having a high breakdown voltage (for example, 8.3 kV) can be obtained by forming the electric field relaxation structure 6 for relaxing the electric field concentration.
International Publication No. 03/038876 Pamphlet Edited by Hiroyuki Matsunami “Semiconductor SiC Technology and Applications” Nikkan Kogyo Shimbun, March 31, 2003, pages 218-221 Materials Science Forum Volume 389-393 2000 1349-1352 Proceedings of International Conference on Silicon Carbide, III-Nitride and Related Materials 1997, 136-137 Materials Science Forum Volume 483-485 2005 969-972 Journal of Applied Physics Volume 95 No. 3 2004 pages 1485-1488 Journal of Applied Physics Volume 92 No. 8 2004 4699-4704 Journal of Crystal Growth Volume 262 2004 130-138 Journal of Applied Physics Volume 96 No. 9 2004 4916-4922

In the pn diode having the mesa structure as shown in FIG. 4, the p ⁺ type charge injection layer 4 as the second conductivity type charge injection layer is not formed by ion implantation unlike the planar structure pn diode, but the epitaxial growth method is used. Therefore, there are relatively few crystal defects formed in the p ⁺ type charge injection layer 4 and its periphery. Therefore, such a pn diode has a relatively high charge injection efficiency during forward bias and a relatively low on-voltage.

On the other hand, the element surface of the pn diode having the mesa structure as shown in FIG. 4 (for example, the surface of the p ⁺ type charge injection layer 4 in FIG. 4) is affected by process steps such as wafer handling, oxidation treatment, and etching treatment. The defect nucleus 23 as shown in FIG. 5 is likely to be locally generated near the element surface as compared with the inside of the bulk. These defect nuclei 23 are converted into stacking faults 24 by the recombination energy of electrons 21 and holes 22 generated during energization (Non-Patent Documents 5 to 7 above). The stacking fault 24 occurs as a planar defect having a shape such as a triangle, and its area increases with an increase in energization time.

Here, a specific description will be given with reference to FIG. 5. Generally, the electron-hole pair is composed of electrons 21 from the n ⁻ -type drift layer 3, which is the first conductivity type drift layer, and the second conductivity type charge injection layer. It is easy to form near the pn interface 11 where holes 22 from a certain p ⁺ type charge injection layer 4 meet. However, some of the electrons 21 reach the vicinity of the surface of the p ⁺ type charge injection layer 4 and may encounter holes 22 there. At this time, the energy released when the electron / hole pairs are generated is given to the defect nuclei 23 localized in the vicinity of the surface of the p ⁺ -type charge injection layer 4. Defects 24 are generated, and the area of stacking faults further increases.

  Since the region of the stacking fault 24 acts as a high resistance region when energized, the forward voltage of the bipolar element increases as the area of the stacking fault 24 increases. As a result, the power loss increases.

Also, in the pn diode having a conductivity type opposite to that of the pn diode shown in FIG. 4, that is, a pn diode in which the p-type is the first conductivity type and the n-type is the second conductivity type, some holes are n ⁺
In some cases, it reaches the vicinity of the surface of the type charge injection layer 4 and encounters electrons there. At this time, the energy released when the electron / hole pairs are generated is given to the defect nuclei localized in the vicinity of the surface of the n ⁺ -type charge injection layer 4, and the stacking faults start from the defect nuclei. And the area expansion of stacking faults proceeds. Therefore, in this case as well, the power loss increases as in the case of the pn diode shown in FIG.

  An object of the present invention is to suppress the occurrence of the above-described stacking fault and its area expansion in a silicon carbide bipolar semiconductor device having a mesa structure, thereby suppressing an increase in forward voltage.

  Therefore, as a result of intensive studies to solve the above-described problems, the present inventors formed a highly doped layer on the surface layer of the second conductivity type charge injection layer of the silicon carbide bipolar semiconductor device, and formed electrons from the drift layer. (Or holes) of electrons (or holes) that could not be trapped in the electric field injection layer are trapped in the highly doped layer, thereby suppressing generation of stacking faults and area expansion, thereby reducing power loss. A method for obtaining a silicon carbide bipolar semiconductor device has been found and the present invention has been completed.

The silicon carbide bipolar semiconductor device of the present invention is
Than the first conductivity type silicon carbide single crystal substrate, the first conductivity type silicon carbide drift layer formed by the epitaxial growth method, the second conductivity type silicon carbide charge injection layer formed by the epitaxial growth method, and the silicon carbide charge injection layer A highly doped layer of the second conductivity type having a high doping density is laminated in this order,
Having a mesa structure from the silicon carbide drift layer to the element surface on the highly doped layer side;
The thickness of the high doping layer is 15 nm to 1.5 μm, and the maximum doping density in the high doping layer is 1 × 10 ²⁰ cm ⁻³ to 2 × 10 ²¹ cm ⁻³ .

  The highly doped layer is preferably formed by ion implantation. Furthermore, it is preferable to use aluminum, nitrogen or phosphorus as the ion species in the ion implantation.

  According to the silicon carbide bipolar semiconductor device of the present invention, the occurrence of stacking faults and the expansion of the area thereof are suppressed, thereby suppressing an increase in forward voltage.

  Hereinafter, an example of a silicon carbide bipolar semiconductor device (pn diode) and a method for manufacturing the same according to the present invention will be described with reference to FIG. This figure is for explanation, and the actual dimensions and the like of the semiconductor device will be understood by those skilled in the art based on the description in this specification and the prior art.

Hereinafter, the case where the first conductivity type is n-type and the second conductivity type is p-type will be mainly described.
A silicon carbide bipolar semiconductor device of the present invention includes a first conductivity type silicon carbide single crystal substrate, a first conductivity type silicon carbide drift layer formed by an epitaxial growth method, and a second conductivity type silicon carbide charge injection formed by an epitaxial growth method. A layer, and a second conductivity type highly doped layer having a higher doping density than the silicon carbide charge injection layer, are laminated in this order,
Having a mesa structure from the silicon carbide drift layer to the element surface on the highly doped layer side;
The thickness of the high doping layer is 15 nm to 1.5 μm, and the maximum doping density in the high doping layer is 1 × 10 ²⁰ cm ⁻³ to 2 × 10 ²¹ cm ⁻³ .

<Each conductive layer>
First conductivity type silicon carbide substrate (n ⁺ substrate) 2 is an n-type SiC single crystal substrate obtained by slicing a SiC bulk crystal obtained by a sublimation method (modified Rayleigh method), a CVD method or the like.

The first conductive type drift layer (n ^- -type drift layer) 3, on the n ⁺ substrate 2, n is the same crystal type and the n ⁺ substrate 2 by an epitaxial growth method ^- growing type SiC single crystal film To form. The n ⁻ type drift layer 3 contains nitrogen or the like as a dopant. In order to obtain a high withstand voltage, it is necessary to reduce the doping density of the first conductivity type layer. Therefore, in the semiconductor device of the present invention, the donor density of the n ⁻ type drift layer 3 is preferably 1 ×.
10 ^{14 to} 6 × 10 ¹⁴ cm ⁻³ .

A second conductivity type charge injection layer (p ⁺ type charge injection layer) 4 is formed on the n − type drift layer 3. The p ⁺ type charge injection layer 4 contains boron, aluminum, or the like as a dopant. The p ⁺ type charge injection layer 4 is formed by growing a p ⁺ type SiC single crystal film by an epitaxial growth method. In the present invention, the acceptor density of the p ⁺ type charge injection layer 4 is preferably 1 × 10 ^{18 to} 5 × 10 ¹⁹ cm ⁻³ .

Next, a second conductivity type high doping layer (p ⁺⁺ type high doping layer) 5 having a high doping density is formed on the upper surface layer of the second conductivity type charge injection layer (p ⁺ type charge injection layer) 4. As shown in FIG. 2, most of the electrons 21 that have entered the p ⁺ -type charge injection layer 4 through the pn interface 11 from the n ⁻ -type drift layer 3 become holes 22 in the charge injection layer 4. Although trapped, a part of the electrons 21 are not trapped by the holes 22 in the charge injection layer 4 and further move toward the surface of the device. Here, by providing the p ⁺⁺ type highly doped layer 5, the electrons 21 not captured by the holes 22 in the charge injection layer 4 are allowed to reach the p ⁺⁺ type highly doped layer 5 before reaching the vicinity of the device surface. Can be more reliably captured by the holes 22 present in the. Therefore, the generation of electron / hole pairs in the vicinity of the defect nuclei 23 localized in the vicinity of the surface of the p ⁺ type charge injection layer 4 is suppressed, and the generation of the stacking fault 24 and the area expansion can be significantly suppressed. The increase of the forward voltage is suppressed, and the loss is reduced.

From this point of view, the doping density of the p ⁺⁺ type highly doped layer 5 needs to be sufficiently higher than that of the p ⁺ type charge injection layer 4, so the minimum value thereof may be 1 × 10 ²⁰ cm ^−3. preferable. On the other hand, since the solid solubility limit of SiC, which is one of the dopants of the p-type conductive layer, with respect to SiC is about 2 × 10 ²¹ cm ⁻³ , the maximum value is preferably 2 × 10 ²¹ cm ^−3. (Non-patent document 8).

In forming this p ⁺⁺ type highly doped layer 5, boron, aluminum or the like is introduced into the p ⁺ type charge injection layer 4 as a dopant, but aluminum is introduced because it shows a more stable doping density profile under high temperature heat treatment. It is particularly preferred.

Furthermore, provision in the present invention, the thickness of the p ⁺⁺ type highly doped layer 5, the relationship between the dopant total of the total number and the p ⁺ -type charge injection layer 4 of the dopant contained in the p ⁺⁺ type highly doped layer 5 can do. Here, in order to simplify the conditions for the definition, the total number of dopants contained in the p ⁺⁺ type highly doped layer 5 is set to be equal to or greater than the total number of dopants in the p ⁺ type charge injection layer 4 That is, the product (dose amount) of the doping density and thickness of the p ⁺⁺ type highly doped layer 5 is set to be larger than the product (dose amount) of the doping density and thickness of the p ⁺ type charge injection layer 4. For the case, the thickness of the p ⁺⁺ type highly doped layer 5 is considered. Under these conditions, the p ⁺⁺ type highly doped layer 5 has an ability equal to or higher than that of the p ⁺ type charge injection layer 4 due to the effect of trapping electrons.

For example, when the thickness of the p ⁺ type charge injection layer 4 is 3 μm, the dose of the p ⁺ type charge injection layer 4 is 3 × 10 ^{14 to} 1.5 × 10 ¹⁶ cm ⁻² . In contrast, if the doping density of the p ⁺⁺ type highly doped layer 5 is 1 × 10 ²⁰ cm ^-3, in order to obtain an equivalent dose of the p ⁺ -type charge injection layer 4, p ⁺⁺ type highly doped The thickness of the layer 5 needs to be 30 nm to 1.5 μm. On the other hand, when the doping density of the p ⁺⁺ type highly doped layer 5 and 2 × 10 ²¹ cm ^-3, in order to obtain a p ⁺ -type charge injection layer 4 and the same or higher dose, p ⁺⁺ type highly doped The thickness of the layer 5 may be 1.5 nm to 75 nm or more.

In the present invention, the p ⁺⁺ type highly doped layer 5 can be formed by ion implantation. In this case, the minimum energy that can be implanted is about 10 keV, so p ⁺⁺
It is difficult to obtain a p ⁺⁺ type highly doped layer 5 in which the type highly doped layer 5 has a thickness of less than 15 nm. In consideration of the thickness of SiC consumed in the alloying reaction with the electrode metal, it is desirable that the thickness of the p ⁺⁺ type highly doped layer 5 be 50 nm or more.

On the other hand, when the ion implantation layer having a high doping density is made thick, the dose is increased, which may cause defects. Therefore, it is desirable that the thickness of the p ⁺⁺ type highly doped layer 5 is 500 nm or less.

From the above examination, it is preferable that the maximum doping density of the high doping layer 5 is 1 × 10 ²⁰ to 2 × 10 ²¹ cm ⁻³ . The thickness is preferably 15 nm to 1.5 μm, and particularly preferably 50 nm to 500 nm.

In the present invention, the “thickness of the second conductivity type highly doped layer (p ⁺⁺ type highly doped layer) 5” means that the second conductivity type highly doped layer (p ⁺⁺ type highly doped layer) 5 is In the case where ions are implanted from the surface of the second conductivity type charge injection layer (p ⁺ type charge injection layer) 4, the doping density distribution in the depth direction (in the case of the multistage ion implantation (implantation) method) , The distance (depth) from the surface to the position giving the maximum doping density in the doping density distribution formed by the maximum ion implantation energy).

<Mesa structure>
The mesa structure of the silicon carbide bipolar semiconductor device according to the present invention can be formed, for example, by the following method.

A silicon oxide film having a thickness of about 10 μm is formed on the p ⁺⁺ type highly doped layer 5 by, for example, a CVD method (Chemical Vapor Deposition). A photoresist film corresponding to the mesa shape is formed on the silicon oxide film by photolithography. Subsequently, the portion without the photoresist film, that is, the exposed silicon oxide film is removed with hydrofluoric acid. By this operation, a silicon oxide film corresponding to the mesa shape is formed on the surface of the p ⁺⁺ type highly doped layer 5. Using this silicon oxide film as a mask, the exposed region of the p ⁺⁺ type highly doped layer 5 is, for example, formed by reactive ion etching (RIE) from the p ⁺⁺ type highly doped layer 5 to the p ⁺ type charge injection layer 4, The mesa structure 10 having a height width of 4 μm is formed by removing up to a part of the n ⁻ -type drift layer 3. The RIE mask is not limited to the silicon oxide film, and may be aluminum, nickel, or the like.

Next, it is desirable to form the electric field relaxation structure 6 in order to alleviate the electric field concentration around the mesa. The electric field relaxation structure 6 can be formed by the following procedure, for example.
A thickness of 2 is formed on the p ⁺⁺ type highly doped layer 5 and the p ⁺ type charge injection layer 4 by, eg, CVD.
A silicon oxide film of about μm is formed. A photoresist film corresponding to the shape of the electric field relaxation structure is formed on the silicon oxide film by photolithography. Subsequently, the portion of the silicon oxide film that does not have a photoresist film, that is, a portion that has an electric field relaxation structure is removed by hydrofluoric acid. By this operation, a silicon oxide film corresponding to the shape of the electric field relaxation structure is formed on the surfaces of the p ⁺⁺ type highly doped layer 5 and the p ⁺ type charge injection layer 4. Using this silicon oxide film as a mask, the electric field relaxation structure 6 is formed by ion implantation using boron or aluminum as a dopant.

The thickness of the electric field relaxation structure 6 is preferably 10 nm to 1 μm, more preferably 100 nm.
nm to 1 μm. The doping density of the electric field relaxation structure 6 is preferably 1 × 10 ¹⁶ to 1 × 10 ¹⁸ cm ⁻³ .

After ion implantation, it is preferable to activate the implanted ions by heat treatment at 1600 ° C. to 1800 ° C. in an atmosphere of Ar or the like.
<Electrode>
The anode electrode 8 and the cathode electrode 9 used in the silicon carbide bipolar semiconductor device according to the present invention can be formed, for example, by the following method.

First, an oxide film for protecting the element surface is formed. For example, an oxide film having a thickness of about 40 nm is formed by a thermal oxidation method.
Subsequently, an unnecessary silicon oxide film is removed by using a photolithography technique and hydrofluoric acid to expose SiC. Next, using a sputtering method or the like, a metal thin film of nickel (about 350 nm thick) is formed on the lower surface of the n ⁺ substrate 2 for the cathode electrode 9, and nickel is formed on the upper surface of the p ⁺⁺ type highly doped layer 5 for the anode electrode 8. Metal thin films having a thickness of about 50 nm are formed. Thereafter, unnecessary nickel is removed from the anode electrode 8 by using a photolithography technique and an etching solution. Furthermore, after forming a metal thin film, these electrodes are alloyed by heat treatment to form ohmic electrodes.

  Further, for the purpose of alleviating mechanical stress due to bonding and reducing the influence of spreading resistance of the anode electrode 8, titanium (thickness: about 50 nm) and aluminum (thickness: 1 μm) are formed on the surface of the anode electrode 8 using a sputtering method or the like. A metal thin film may be formed.

In the present invention, the crystal type and crystal plane of the SiC single crystal are not particularly limited, and the effects of the present invention can be obtained for various types.
As described above, the silicon carbide bipolar semiconductor device in which the n-type is the first conductivity type and the p-type is the second conductivity type has been described. However, the silicon carbide bipolar semiconductor device according to the present invention has the opposite conductivity type, A silicon carbide bipolar semiconductor device in which the p-type is the first conductivity type and the n-type is the second conductivity type may be used. In this case, in the silicon carbide bipolar semiconductor device according to the present invention, the p ⁺ type substrate 2, the p ⁻ type drift layer 3, the n ⁺ type charge injection layer 4, and the n ⁺⁺ type highly doped layer 5 They are stacked in order. In this case, the ions introduced into the n ⁺ type charge injection layer 4 when the n ⁺⁺ type highly doped layer 5 is formed are nitrogen and phosphorus, and the electric field relaxation structure 6 is formed. The same applies to ions introduced into the p ⁻ -type drift layer 3.

Hereinafter, the present invention will be described by way of examples. However, the present invention is not limited to these examples, and various modifications and changes can be made without departing from the scope of the present invention.
[Example]
EXAMPLES Hereinafter, although an Example demonstrates this invention, this invention is not limited to this Example.

[Example 1]
The pn diode shown in FIG. 1 was produced.
(1) Formation of Conductive Layer First, an n ⁺ substrate 2 (donor density 7 ×) which is an n-type 4H—SiC single crystal substrate obtained by slicing a SiC bulk crystal obtained by a sublimation method (improved Rayleigh method) 10 ¹⁸ cm ^-3, on the thickness of 400 [mu] m), the same crystal form and the n ⁺ substrate 2 by an epitaxial growth method by growing a n- type SiC single crystal film, having a thickness of 60 [mu] m n- type drift layer 3 was formed. This n − -type drift layer 3 contains nitrogen as a dopant, and the donor density of the n − -type drift layer 3 is 1 × 10 ^{14 to} 6 × 10 ¹⁴ cm ⁻³ .

Then, by growing a p ⁺ -type SiC single crystal layer by epitaxial growth method on the n- type drift layer 3 was formed p ⁺ type conductive layer 4 having a thickness of 3 [mu] m. The p ^{+ type} conductive layer 4 contains aluminum as a dopant, and the acceptor density of the p ^{+ type} conductive layer 4 is 1 × 10 ^{18 to} 5 ×.
It was in the range of 10 ¹⁹ cm ⁻³ .

After the p ^{+ type} conductive layer 4 was formed, the p ⁺⁺ type highly doped layer 5 was formed by ion implantation of aluminum. The ion implantation conditions are multistage implantation with a total dose of 2.3 × 10 ¹⁶ cm ⁻² at an implantation energy of 20 to 150 keV, a depth of 250 nm, and a doping density of 1 × 10 ²¹ cm ^−3. A box profile was formed.

(2) Formation of Mesa Structure A silicon oxide film having a thickness of about 10 μm was formed on the p ⁺⁺ type highly doped layer 5 by the CVD method. A photoresist film corresponding to the mesa shape was formed on the silicon oxide film by photolithography. Subsequently, the portion without the photoresist film, that is, the exposed silicon oxide film was removed with hydrofluoric acid. Using this silicon oxide film as a mask, the exposed region of the p ⁺⁺ type highly doped layer 5 is subjected to reactive ion etching (RIE) from the p ⁺⁺ type highly doped layer 5 to the p ⁺ type charge injection layer 4, and The mesa structure 10 having a height width of 4 μm was formed by removing a part of the n ⁻ -type drift layer 3.

(3) Formation of electric field relaxation structure In order to alleviate electric field concentration at the mesa peripheral part, ion implantation is performed using aluminum as a dopant in the n ^{− type} conductive layer 3 over a range from the pn junction interface to the outside of the mesa corner part of about 50 μm. Thus, the electric field relaxation structure 6 was formed. This ion implantation was performed under the condition that the dose amount was 3.2 × 10 ¹⁴ cm ⁻² at a multistage implantation energy of 15 to 350 keV. The profile in the depth direction with respect to the doping density of aluminum implanted under these conditions is a box profile that becomes 6 × 10 ¹⁸ cm ⁻³ from the vicinity of the implantation surface to a depth of 500 nm. After the ion implantation, the implanted ions were activated by heat treatment at 1600 ° C. to 1800 ° C. in an atmosphere such as Ar.

Next, in order to protect the element surface, an oxide film 7 having a thickness of about 40 nm was formed on the entire element surface on the side where the p ⁺⁺ type highly doped layer 5 exists.
Subsequently, as described above, an unnecessary silicon oxide film was removed by using a photolithography technique and hydrofluoric acid to expose the surface of the p ⁺⁺ type highly doped layer 5. Next, using a sputtering method or the like, a metal thin film of nickel (thickness 350 nm) is formed as the cathode electrode 9 on the lower surface of the n ⁺ substrate 2, and nickel (thickness) is formed as the anode electrode 8 on the upper surface of the p ⁺⁺ type highly doped layer 5. 50 nm) metal thin films were respectively formed. For the anode electrode 8, unnecessary nickel was removed by using a photolithography technique and an etching solution. Furthermore, after forming a metal thin film, these electrodes were alloyed by heat treatment to form ohmic electrodes.

Further, a metal thin film of titanium (thickness 50 nm) and aluminum (thickness 1 μm) was laminated on the surface of the anode electrode 8 in this order.
(4) Energization test of the obtained pn diode The following energization test was performed on the obtained pn diode. A cathode electrode of a pn diode was pasted on a copper plate using a high melting point solder, and an aluminum wire was bonded to the anode electrode using an ultrasonic bonding apparatus. After connecting a current source to the copper plate and the aluminum wire and deteriorating the direct current of 100 A / cm ² in the forward direction, an electroluminescence image (EL image) of this pn diode was obtained. By repeating the acquisition, the movement and shape change of the stacking fault 24 was observed, and the number of defect nuclei 23 serving as the starting point of the stacking fault 24 was evaluated. The evaluation results are shown in FIG.

[Comparative Example 1]
The ion implantation process for forming the p ⁺⁺ type highly doped layer 5 was not performed, and a process similar to the example was applied to the other processes, and a comparative pn diode was manufactured.

Thereafter, the same energization test as in Example 1 was performed on the comparative pn diode, and the evaluation results shown in FIG. 3 were obtained.
As shown in FIG. 3, in the pn diode (Comparative Example 1) in which the p ⁺⁺ type highly doped layer 5 is not introduced, the number of defect nuclei 23 found near the surface layer of the p ⁺ type charge injection layer 4 is 1 or less. The number of the defect nuclei in the pn diode (Example 1) in which the p ⁺⁺ type highly doped layer 5 was introduced was 1% or less, compared with 57% (13/23). It was 90% (18/20). As described above, it can be confirmed that the introduction of the p ⁺⁺ type highly doped layer 5 can suppress the expansion of stacking faults starting from the defect nucleus 23 in the vicinity of the surface layer of the p ⁺ type charge injection layer 4. It was.

In addition, Joseph. J. et al. SUMAKERIS et al, between p ⁺ -type charge injection layer and the n- type drift layer, a doping density of ^{1 × 10 17 ~1 × 10 19} cm -3, a thickness of about 1.5 [mu] m p ⁺
A type buffer layer is provided, and an invention for recombining electrons diffused from the n ⁻ drift layer with holes is performed (Patent Document 1). On the other hand, in the present invention, p ^{++ is formed on} the surface layer of the p ⁺ type charge injection layer.
By providing the type highly doped layer, the electrons which are minority carriers are recombined, and by increasing the doping density of the p ⁺⁺ type highly doped layer to 1 × 10 ²⁰ to 2 × 10 ²¹ cm ⁻³ , p The thickness of the ⁺⁺ type highly doped layer can be reduced to 1.5 μm or less, and an ohmic contact can be formed using the p ⁺⁺ type highly doped layer to obtain a low resistance anode electrode. I can expect. Moreover, in the said patent document 1, the suppression effect of the stacking fault from the surface layer vicinity is not evaluated quantitatively. In this respect, the present invention relates to Joseph. J. et al. It is clearly different from those of SUMAKERIS and others.

  In the present embodiment, a pn diode has been described as an example, but other silicon carbide bipolar semiconductor devices such as a thyristor, a gate turn-off thyristor (GTO), an insulated gate bipolar transistor (IGBT), a bipolar junction transistor (BJT), etc. The present invention can be applied.

It is a schematic cross section in one mode of a silicon carbide bipolar semiconductor device of the present invention. It is a figure for demonstrating the effect | action of the silicon carbide bipolar semiconductor device of this invention. It is a figure which shows the evaluation result of the pn diode produced by the Example and the comparative example. It is sectional drawing of the conventional pn diode which formed the mesa structure. It is a figure explaining the production | generation mechanism of the stacking fault in the conventional pn diode.

Explanation of symbols

1: pn diode 2: n ⁺ substrate 3: n ⁻ type drift layer 4: p ⁺ type charge injection layer 5: p ⁺⁺ type highly doped layer 6: electric field relaxation structure 7: oxide film 8: anode electrode 9: cathode electrode 10: mesa structure 11: pn junction interface 21: electron 22: hole 23: defect nucleus 24: stacking fault

Claims (3)

Than the first conductivity type silicon carbide single crystal substrate, the first conductivity type silicon carbide drift layer formed by the epitaxial growth method, the second conductivity type silicon carbide charge injection layer formed by the epitaxial growth method, and the silicon carbide charge injection layer A highly doped layer of the second conductivity type having a high doping density is laminated in this order,
Having a mesa structure from the silicon carbide drift layer to the element surface on the highly doped layer side;
Carbonization characterized in that the thickness of the highly doped layer is 15 nm to 1.5 μm and the maximum doping density in the highly doped layer is 1 × 10 ²⁰ cm ⁻³ to 2 × 10 ²¹ cm ^−3. Silicon bipolar semiconductor device.   2. The silicon carbide bipolar semiconductor device according to claim 1, wherein the highly doped layer is formed by ion implantation.   The silicon carbide bipolar semiconductor device according to claim 1, wherein the highly doped layer is formed by ion implantation of aluminum, nitrogen, or phosphorus.

Images