Semiconductor device and method of producing the same

Abstract

A semiconductor device is provided in which a semiconductor substrate can be prevented from being broken while elements can be prevented from being destroyed by a snap-back phenomenon. After an MOS gate structure is formed in a front surface of an FZ wafer, a rear surface of the FZ wafer is ground. Then, the ground surface is irradiated with protons and irradiated with two kinds of laser beams different in wavelength simultaneously to thereby form an N.sup.+ first buffer layer and an N second buffer layer. Then, a P+ collector layer and a collector electrode are formed on the proton-irradiated surface. The distance from a position where the net doping concentration of the N.sup.+ first buffer layer is locally maximized to the interface between the P+ collector layer and the N second buffer layer is set to be in a range of 5 m to 30 m, both inclusively.

Claims

1. A method of producing a semiconductor device, said method comprising the steps of: grinding a first principal surface of a semiconductor substrate of a first conductivity type as a first semiconductor layer; injecting light ions into a surface of the first semiconductor layer exposed by the grinding; injecting p-type conductivity ions into the surface of the first semiconductor layer exposed by the grinding; irradiating the surface injected with the light ions and p-type conductivity ions with two types of laser beams different in wavelength to thereby electrically activate the injected light ions and form an n-type first buffer layer and an n-type second buffer layer; forming a p-type collector layer from the injected p-type conductivity ions; and forming an electrode by laminating a metal film on the surface irradiated with the laser beams, wherein the injecting of the light ions into the surface of the first semiconductor layer exposed by the grinding is to an injection depth, the injection depth being, at a peak position of injection, at least 5 m to 30 m, both inclusively, plus a depth of the p-type collector layer, wherein a penetration length of a short-wavelength laser beam as one of the two types of laser beams into the semiconductor substrate is in a range of 0.3 m to 5 m, both inclusively, wherein a penetration length of a long-wavelength laser beam as the other of the two types of laser beams into the semiconductor substrate is in a range of 5 m to 30 m, both inclusively, wherein the long-wavelength laser beam is emitted from either an Al.sub.XGa.sub.1-XAs laser or an In.sub.XGa.sub.1-XAs laser in which X expresses a normalized composition ratio, wherein a doping concentration of the n-type second buffer layer is lower than a doping concentration of the n-type first buffer layer and a doping concentration of the p-type collector layer, and the doping concentration of the n-type second buffer layer is higher than a doping concentration of the first semiconductor layer, wherein the p-type collector layer is separated from the n-type first buffer layer by the n-type second buffer layer, wherein the irradiating with two types of laser beams electrically activates the light ions and provides the light ions as donors by changing a state of silicon to a melted state in a region to a melted-state depth of 30 m from the surface of the first semiconductor layer exposed by the grinding, and wherein a full width at half maximum of a peak doping concentration of the n-type first buffer layer is not smaller than 2.5 m.

2. A method of producing a semiconductor device according to claim 1, wherein the irradiating step performs simultaneous irradiation with the two types of laser beams different in wavelength.

3. A method of producing a semiconductor device according to claim 1, wherein the short-wavelength laser beam is emitted from an all-solid-state laser, a gallium nitride-including semiconductor laser, or a gas laser, the all-solid-state laser selected from the group consisting of a YAG 2 laser, a YVO4 2 laser, and a YLF 2 laser, and the gas laser selected from the group consisting of an excimer laser and a helium-neon laser.

4. A method of producing a semiconductor device according to claim 1, wherein the long-wavelength laser beam is emitted from a semiconductor laser containing gallium as part of its structure, a ruby laser, or a normal-frequency all-solid-state laser selected from the group consisting of a YAG laser, a YVO.sub.4 laser, and a YLF laser.

5. A method of producing a semiconductor device according to claim 1, wherein the light ions are selected from the group consisting of protons, helium ions, lithium ions, and oxygen ions.

6. A method of producing a semiconductor device according to claim 1, wherein the light ions are protons.

7. A method of producing a semiconductor device according to claim 5, wherein the injecting light ions step injects the light ions at an acceleration voltage in a range of 200 keV to 30 MeV, both inclusively.

8. A method of producing a semiconductor device according to claim 6, wherein the injecting light ions step injects protons at an acceleration voltage in a range of 200 keV to 2 MeV, both inclusively.

9. A method of producing a semiconductor device according to claim 1, wherein the p-type collector layer is formed by electrically activating the p-type conductivity ions.

10. A method of producing a semiconductor device, said method comprising the steps of: grinding a first principal surface of a semiconductor substrate of a first conductivity type as a first semiconductor layer; injecting light ions into a surface of the first semiconductor layer exposed by the grinding; injecting p-type conductivity ions into the surface of the first semiconductor layer exposed by the grinding; irradiating the surface injected with the light ions and p-type conductivity ions with two types of laser beams different in wavelength to thereby electrically activate the injected light ions and form an n-type first buffer layer and an n-type second buffer layer; forming a p-type collector layer from the injected p-type conductivity ions; forming an electrode by laminating a metal film on the surface irradiated with the laser beams; and forming a surface structure on a second surface of the semiconductor substrate, wherein a penetration length of a short-wavelength laser beam as one of the two types of laser beams into the semiconductor substrate is in a range of 0.3 m to 5 m, both inclusively, wherein a penetration length of a long-wavelength laser beam as the other of the two types of laser beams into the semiconductor substrate is in a range of 5 m to 30 m, both inclusively, wherein the injecting light ions step injects the light ions into the surface of the first semiconductor layer exposed by the grinding to an injection depth, the injection depth being, at a peak position of injection, at least 5 m to 30 m, both inclusively, plus a depth of the p-type collector layer, wherein the injecting light ions step introduces damage within the injection depth, the damage reducing carrier lifetime and carrier mobility, wherein the irradiating step, involving the irradiating with the two types of laser beams different in wavelength, provides protons as donors and reduces the damage within the injection depth at least by extinguishing deep levels within the injection depth without extinguishing shallow levels within the injection depth, the deep levels being deeper from the surface injected with the light ions than the shallow levels, wherein the light ions are protons, wherein a doping concentration of the n-type second buffer layer is lower than a doping concentration of the n-type first buffer layer and a doping concentration of the p-type collector layer, and the doping concentration of the n-type second buffer layer is higher than a doping concentration of the first semiconductor layer, wherein the p-type collector layer is separated from the n-type first buffer layer by the n-type second buffer layer, wherein the irradiating with two types of laser beams electrically activates the light ions and provides the light ions as donors by changing a state of silicon to a melted state in a region to a melted-state depth of 30 m from the surface of the first semiconductor layer exposed by the grinding, and wherein a full width at half maximum of a peak doping concentration of the n-type first buffer layer is not smaller than 2.5 m.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The invention will be described with reference to certain preferred embodiments and accompanying drawings, wherein:

(2) FIG. 1 is a view showing the configuration of a semiconductor device according to Embodiment 1 and the net doping concentration thereof;

(3) FIG. 2-1 is a view showing a process of producing the semiconductor device according to Embodiment 1;

(4) FIG. 2-2 is a view showing a process of producing the semiconductor device according to Embodiment 1;

(5) FIG. 2-3 is a view showing a process of producing the semiconductor device according to Embodiment 1;

(6) FIG. 2-4 is a view showing a process of producing the semiconductor device according to Embodiment 1;

(7) FIG. 2-5 is a view showing a process of producing the semiconductor device according to Embodiment 1 and a profile of net doping concentration;

(8) FIG. 3 is a view showing the size and net doping concentration of each portion from the center of an N.sup. drift layer to a collector electrode in the semiconductor device according to Embodiment 1;

(9) FIG. 4 is a characteristic graph showing the relationship between the collector-emitter voltage and the collector current amplification rate;

(10) FIG. 5 is a graph showing a result of conversion of the spreading resistance of the semiconductor device according to Embodiment 1 into a carrier concentration;

(11) FIG. 6 is a graph showing a result of measurement of the projected range of protons and the percentage of leakage current non-defective products, versus acceleration voltage when protons were injected;

(12) FIG. 7-1 is a characteristic graph showing temperature at a depth from the wafer surface, versus elapsed time from a point of time when laser irradiation was terminated;

(13) FIG. 7-2 is a characteristic graph showing temperature at a depth from the wafer surface, versus elapsed time from a point of time when laser irradiation was terminated;

(14) FIG. 7-3 is a characteristic graph showing temperature at a depth from the wafer surface, versus elapsed time from a point of time when laser irradiation was terminated;

(15) FIG. 8 is a graph showing a result of measurement of leakage current versus annealing temperature of an electric furnace in a semiconductor device according to the related art;

(16) FIG. 9 is a graph showing results of conversion of the spreading resistance values of the semiconductor device according to Embodiment 1 and the semiconductor device according to the related art into carrier concentration values;

(17) FIG. 10 is a graph showing the percentage of leakage current non-defective products versus a value obtained by multiplying a peak value of the net doping concentration and a full width at half maximum thereof;

(18) FIG. 11 is an explanatory graph showing a result of measurement of the projected range of helium ions and the percentage of leakage current non-defective products, versus acceleration voltage when helium ions were injected;

(19) FIG. 12 is an explanatory graph showing a result of measurement of the projected range of lithium ions and the percentage of leakage current non-defective products, versus acceleration voltage when lithium ions were injected;

(20) FIG. 13 is an explanatory graph showing a result of measurement of the projected range of oxygen ions and the percentage of leakage current non-defective products, versus acceleration voltage when oxygen ions were injected;

(21) FIG. 14 is a view showing the configuration of a semiconductor device according to Embodiment 3 and the net doping concentration thereof;

(22) FIG. 15 is a graph showing turn-off waveforms of the semiconductor device according to the invention and the semiconductor device according to the related art;

(23) FIG. 16 is a diagram showing a snubberless circuit;

(24) FIG. 17 is a view showing the configuration of a semiconductor device according to Embodiment 4 and the net doping concentration thereof;

(25) FIG. 18-1 is a diagram showing an example of use of IGBTs and diodes according to the invention;

(26) FIG. 18-2 is a diagram showing an example of use of IGBTs and diodes according to the invention;

(27) FIG. 19 is a view showing the configuration of a semiconductor device having a field stop layer according to the related art and the net doping concentration thereof; and

(28) FIG. 20 is a graph showing results of SRIM2006's calculation of the projected range Rp of protons in silicon irradiated with protons and the FWHM Rp thereof in a hydrogen distribution after the irradiation.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

(29) Preferred embodiments of the invention as to a semiconductor device and a producing method thereof will be described below in detail with reference to the accompanying drawings. Incidentally, in the following description of embodiments and all the accompanying drawings, like numerals refer to like constituent parts for the sake of omission of duplicate description.

(30) FIG. 1 is a view showing the configuration of a semiconductor device according to Embodiment 1 and the net doping concentration thereof. As shown in a section 100 of the semiconductor device in FIG. 1, for example, an N.sup.+ first buffer layer (N.sup.+ field stop layer) 2 as a second semiconductor layer, an N second buffer layer 12 as a third semiconductor layer and a P.sup.+ collector layer 3 as a fourth semiconductor layer are formed in this order on one principal surface (first principal surface) of an N.sup. drift layer 1 as a first semiconductor layer. A P base layer 4 as a fifth semiconductor layer is formed on the other principal surface (second principal surface) of the N.sup. drift layer 1. An N source layer 5 as a sixth semiconductor layer is formed in part of a surface layer of the P base layer 4 so as to be far from the N.sup. drift layer 1. A gate electrode 7 is formed through a gate insulating film 6 in a trench which passes through the N source layer 5 and the P base layer 4 and reaches the N.sup. drift layer 1. An emitter electrode 8 as a first electrode is formed on a surface of the P base layer 4 and the N source layer 5. A collector electrode 9 as a second electrode is formed on a surface of the P.sup.+ collector layer 3.

(31) As shown in a characteristic graph 110 of distance from the emitter electrode versus net doping concentration (log) in FIG. 1, the net doping concentration of the N.sup.+ first buffer layer 2 has a peak near the midpoint of the N.sup.+ first buffer layer 2 and higher than the net doping concentration of the N.sup. drift layer 1. The net doping concentration of the N second buffer layer 12 is higher than that of the N.sup. drift layer 1 but lower than the peak value of the net doping concentration of the N.sup.+ first buffer layer 2. The net doping concentration of the N second buffer layer 12 decreases with a gradient in the direction from the interface between the N second buffer layer 12 and the N.sup.+ first buffer layer 2 to the interface between the N second buffer layer 12 and the P.sup.+ collector layer 3. Both the net doping concentrations of the P.sup.+ collector layer 3 and the P base layer 4 are higher than the net doping concentrations of the N.sup. drift layer 1, the N.sup.+ first buffer 2 and the N second buffer layer 12.

(32) The net doping concentration and size when the semiconductor device according to Embodiment 1 was produced with a chip size of 8 mm8 mm, a 1200 V withstand voltage class and a rating current of 75 A are shown as an example. The size is based on the interface between the P base layer 4 and the emitter electrode 8 and expressed in distance from this interface except as otherwise noted.

(33) The distance to the interface between the P base layer 4 and the N.sup. drift layer 1 is 3 m. The distance to the interface between the P.sup.+ collector layer 3 and the collector electrode 9 is 140 m. The distance from the interface between the N second buffer layer 12 and the P.sup.+ collector layer 3 to the interface between the P.sup.+ collector layer 3 and the collector electrode 9, that is, the thickness of the P.sup.+ collector layer 3 is 0.5 m. The distance from the peak position of the net doping concentration of the N.sup.+ first buffer layer 2 to the interface between the P.sup.+ collector layer 3 and the collector electrode 9, that is, the projected range Rp of protons is 16 m.

(34) The net doping concentration of the P base layer 4 takes 510.sup.16 atoms/cc at the interface between the P base layer 4 and the emitter electrode 8, decreases in the direction of the N.sup. drift layer 1 and takes a value lower than 510.sup.13 atoms/cc at the interface between the P base layer 4 and the N.sup. drift layer 1. The net doping concentration of the P.sup.+ collector layer 3 takes 110.sup.18 atoms/cc at the interface between the P.sup.+ collector layer 3 and the collector electrode 9, decreases in the direction of the N second buffer layer 12 and takes a value lower than 510.sup.13 atoms/cc at the interface between the P.sup.+ collector layer 3 and the N second buffer layer 12. The net doping concentration of the N.sup. drift layer 1 is 510.sup.13 atoms/cc. The maximum value of the net doping concentration of the N.sup.+ first buffer layer 2 is 110.sup.15 atoms/cc. Although either atoms/cc or atoms/cm.sup.3 will be used as a unit of concentration in the following description, cc and cm.sup.3 are equivalent to each other.

(35) A process of producing the semiconductor device according to Embodiment 1 will be described below. Production of a semiconductor device (withstand voltage: 1200 V class, rating current: 75 A) having the size and net doping concentration shown in FIG. 1 will be described as an example. FIGS. 2-1 to 2-5 are views showing the producing process. First, as shown in FIG. 2-1, an FZ wafer 10 having a specific resistance of 40 to 80 cm (e.g. 55 cm) and a diameter of 6 inches is prepared as a semiconductor substrate. Then, a P base layer 4, an N source layer 5, a gate insulating film 6, a gate electrode 7, an insulating film 11 and an emitter electrode 8 are formed by a standard trench gate type MOS device forming process. For example, the material of the emitter electrode 8 is aluminum (Al).

(36) Then, as shown in FIG. 2-2, the rear surface of the FZ wafer 10 is ground or wet-etched to set the thickness of the FZ wafer 10 at a predetermined value. In the case of 1200 V class, the thickness of the FZ wafer 10 is typically 100 to 160 m at this point of time. For example, in Embodiment 1, the thickness of the FZ wafer 10 is 140 m at this point of time.

(37) Then, as shown in FIG. 2-3, the ground or wet-etched surface is irradiated with protons (H.sup.+). On this occasion, the acceleration voltage is, for example, 1 MeV, and the dose quantity of protons is, for example, 110.sup.14 atoms/cm.sup.2. On this occasion, the projected range Rp of protons is 16 m from the ion-irradiated surface. Then, boron ions (B.sup.+ or BF.sub.2) are injected into the rear surface of the FZ wafer 10. On this occasion, the acceleration voltage is, for example, 50 keV, and the dose quantity of boron ions is, for example, 110.sup.13 atoms/cm.sup.2.

(38) Then, as shown in FIG. 2-4, the ion-injected surface is irradiated with a YAG 2 laser (wavelength: 500 nm) and a GaAs semiconductor laser (wavelength: 808 nm) simultaneously. The energy for laser beam irradiation is, for example, 200 mJ/cm.sup.2 for the YAG 2 laser and 4000 mJ/cm.sup.2 for the semiconductor laser.

(39) Then, as shown in FIG. 2-5, crystal defects caused by proton irradiation are recovered so that a high-concentration region 13 is formed in a region of proton projected range RpFWHM Rp from the ion-injected surface. In the high-concentration region 13, a region of from RpRp to Rp+Rp serves as an N.sup.+ first buffer layer. The net doping concentration of a proton pass region (from the ion-injected surface to the position of RpRp) becomes higher than the bulk concentration. The proton pass region serves as an N second buffer layer. The bulk concentration is the net doping concentration of the FZ wafer 10.

(40) Then, a P-type impurity such as boron is electrically activated to form a P.sup.+ collector layer 3. Then, a polyimide film, for example, 5 m thick is applied on a surface of the P.sup.+ collector layer 3 and patterned to form a passivation film in an edge region not shown. Then, the surface of the P.sup.+ collector layer 3 is coated with aluminum, titanium (Ti), nickel (Ni) and gold (Au) successively to form a collector electrode 9 in ohmic contact with the P.sup.+ collector layer 3. Thus, a semiconductor device is completed. A portion of the FZ wafer between the P base layer 4 and the high-concentration region 13 serves as an N.sup. drift layer. The characteristic graph of FIG. 2-5 is a profile of the net doping concentration corresponding to the semiconductor device shown in FIG. 2-5.

(41) FIG. 3 is a graph showing the size and net doping concentration of each portion from the center of the N.sup. drift layer to the collector electrode in the semiconductor device according to Embodiment 1. As shown in FIG. 3, the distance to the peak position of the net doping concentration of the N.sup.+ first buffer layer is set as Xp with the PN junction between the N second buffer layer and the P+ collector layer as a base of measurement of the distance. In the semiconductor device according to Embodiment 1, Xp is in a range of 5 m to 30 m, both inclusively. This reason will be described later.

(42) FIG. 4 is a characteristic graph showing the relationship between collector-emitter voltage and collector current amplification rate. The collector current amplification rate in the condition that an avalanche current begins to flow because a voltage BV very near the withstand voltage of the device is applied between the collector and the emitter in an off state of the device is shown in FIG. 4.

(43) The collector current amplification rate I.sub.C/I.sub.L is given by the following expression (1).
I.sub.C/I.sub.L=1/(1.sub.PNP.sup.eff)(1)
in which I.sub.C is a collector current when a voltage BV is applied between the collector and the emitter, I.sub.L is a leakage current in the condition that the P.sup.+ collector layer has not been formed in the rear surface of the FZ wafer, and .sub.PNP.sup.eff is an effective amplification rate when the voltage BV is applied. When the voltage BV is applied between the collector and the emitter, a depletion layer reaches a range of from the N.sup. drift layer to the N.sup.+ first buffer layer so that a neutral region of zero space charge is sufficiently shortened in the N.sup.+ first buffer layer. The collector current amplification rate in this condition is .sub.PNP.sup.eff.

(44) The effective amplification rate .sub.PNP.sup.eff is given by the following expression (2) and has a value smaller than 1.0.
.sub.PNP.sup.eff=.sub.ETM(2)
in which .sub.E is a hole injection efficiency from the emitter side, T is a transport efficiency in the drift layer, and M is an amplification coefficient due to impact ionization.

(45) When a positive bias is applied to the collector in an off state of the IGBT, the depletion layer is spread from the emitter side into the drift layer so that the neutral region width in the drift layer is reduced as the applied voltage is increased. Accordingly, the effective base width of the PNP transistor is reduced and .sub.E increases, so that .sub.PNP increases. Such .sub.PNP changing according to the applied voltage is called effective amplification rate in this specification and expressed as .sub.PNP.sup.eff in distinction from static .sub.PNP.

(46) The transport efficiency T is given by the following expression (3).
T=1/(cos h(L.sub.CNZ/L.sub.P))(3)
in which L.sub.CNZ is the residual width of the neutral region, and L.sub.P is the extended length of holes. The neutral region residual width L.sub.CNZ is the width of the neutral region which has been not yet extinguished from the drift layer (the N.sup. drift layer, the N.sup.+ first drift layer and the N second buffer layer) by the depletion layer spread from the emitter side into the drift layer.

(47) The amplification coefficient M due to impact ionization is given by the following expression (4).
M=1/(1(V/V.sub.PP).sup.6)(4)
in which V is the applied voltage, and V.sub.PP is an ideal withstand voltage value of a diode formed between one stepped front surface and the other, PN junction surface parallel to the substrate surface.

(48) In FIG. 4, L.sub.CNZ is changed from 0.1 m to 30 m. In FIG. 4, V.sub.PP is 1450 V and .sub.E is 0.3. As shown in FIG. 4, the collector current amplification rate increases rapidly at a voltage value lower than V.sub.PP. This reason is commonly known and description thereof is omitted but the rapid increase of the collector current amplification rate is caused by an operation of opening the base of the PNP transistor.

(49) The withstand voltage of the PNP transistor is a voltage when the collector current amplification rate increases rapidly. Accordingly, the withstand voltage of the PNP transistor is approximately 1334.4 V when L.sub.CNZ is in a range of 0.1 m to 3 m but the withstand voltage of the PNP transistor increases when L.sub.CNZ is not smaller than 5 m. When L.sub.CNZ is 30 m, the withstand voltage of the PNP transistor is 1335.3 V. In comparison between the L.sub.CNZ range of 0.3 m to 3 m and the region of L.sub.CNZ of 30 m, the withstand voltage of the PNP transistor increases at a small rate which is about 0.1% of the applied voltage.

(50) A trench gate structure device according to the related art will be described here. In the case where phosphorus is used for forming a field stop layer having a thickness of 1 m or less in a trench gate structure device according to the related art, the distance from an end of a depletion layer to a P.sup.+ collector layer is about 0.3 m when the depletion layer reaches the field stop layer. For this reason, the effective injection efficiency is near 1. On this occasion, holes injected from the rear surface of the wafer pass through the depletion layer and reaches the front surface side of the wafer. On the front surface side of the wafer, a P base layer is in contact with an emitter electrode. Holes pass through the neighborhood of a gate electrode embedded in the trench in the middle of movement toward the interface between the P base layer and the emitter electrode.

(51) On this occasion, the trench embedded in the gate electrode reaches an N.sup. drift layer. When a high voltage is applied in an off state, electric potential is curved in accordance with the shape of the bottom of the trench so that electric field intensity increases in the bottom of the trench. The reason why electric field intensity increases is that the curvature of the bottom of the trench is higher than that of the P well layer of a planer gate type device. Accordingly, holes are temporarily concentrated into the bottom of the trench. Because further increase of electric field intensity in the bottom of the trench is caused by the concentration of holes as described above, an avalanche breakdown occurs so that electrons are generated. The electrons pass through the depletion layer toward the P.sup.+ collector layer side and reach the P.sup.+ collector layer.

(52) The avalanche breakdown causes flowing of a current even when increase of the voltage is a very small value. As a result, injection of holes from the P.sup.+ collector layer is accelerated to establish a positive feedback state. As a result, the collector current increases. At the same time, carriers in the depletion layer increase so that the electric field intensity distribution changes to make it impossible to keep the high voltage in the N.sup. drift layer.

(53) As a result, the collector-emitter voltage decreases in addition of increase of the current, so that a snap-back phenomenon occurs. The snap-back phenomenon is a phenomenon that the device deteriorates because of concentration of the current into one place of the largest positive feedback.

(54) The trench gate type structure is effective for obtaining a low-loss device. On the other hand, in order to suppress the snap-back phenomenon in the trench gate type device, it is necessary to increase the distance from the end portion of the depletion layer to the P.sup.+ collector layer in the condition that the depletion layer has reached the field stop layer and has been not spread any more, that is, the distance of the neutral region as much as possible.

(55) It is commonly known that the snap-back phenomenon does not occur in the related-art field stop layer formed by use of Se or S because the distance of the neutral region is about 5 m or more. This is because the positive feedback is suppressed sufficiently. It is however known that deterioration of the withstand voltage at flowing of a static avalanche current in the field stop layer formed by irradiation with light ions such as protons is caused by the snap-back phenomenon.

(56) Therefore, when L.sub.CNZ is set to be not smaller than 5 m and, preferably, not smaller than 10 m, the snap-back phenomenon can be avoided even in the trench gate structure device having the field stop layer formed by irradiation with protons. Accordingly, it is preferable that Xp is not smaller than 5 m because Xp must be equal to or larger than L.sub.CNZ.

(57) In the method of producing the semiconductor device according to Embodiment 1, a proton projected range Rp, for example, of 16 m can be obtained when proton irradiation is performed at an acceleration voltage, for example, of 1 MeV. Accordingly, a neutral region can be formed in a position 5 m or more far from the PN junction. For this reason, the snap-back phenomenon does not occur even when an avalanche current flows at the time of measuring the withstand voltage, so that the withstand voltage of the device does not deteriorate. The actual withstand voltage varies according to the junction terminal structure such as a known guard ring structure, an RESURF (Reduced Surface Field) structure, a combination structure of a guard ring structure and a field plate structure, etc. For example, assume that the junction terminal structure in the semiconductor device according to Embodiment 1 is formed from a combination structure of a guard ring structure and a field plate structure. Accordingly, there is exhibited a value almost equal to the withstand voltage value approximate to a planer junction including a PNP transistor.

(58) A sample produced as described above was attached to a mount at an angle of 252 and ground so that a section of a wafer is exposed. SSM2000 made by Solid State Measurement, Inc. was used for measuring the spreading resistance of the sample. FIG. 5 shows a result of conversion of the resistance value obtained by the measurement into a carrier concentration. In FIG. 5, the vertical axis shows the carrier concentration, and the horizontal axis shows the distance from the wafer surface (proton-irradiated surface), that is, the depth from the wafer surface. In FIG. 5, the circular mark (.circle-solid.) indicates the carrier concentration at the depth when a YAG 2 laser and a semiconductor laser were applied simultaneously, and the rhombic mark (.diamond-solid.) indicates the carrier concentration at the depth when only the semiconductor laser was applied.

(59) As shown in FIG. 5, it is found that the carrier concentration at the time of application of only the semiconductor laser is lower than at the time of simultaneous application of the YAG 2 laser and the semiconductor laser. This reason is that damage due to proton irradiation is not recovered and the carrier mobility is reduced so that protons are not activated at the time of application of only the semiconductor laser.

(60) FIG. 6 shows a result of measurement of the projected range of protons and the percentage of leakage current non-defective products, versus acceleration voltage when protons were injected. In FIG. 6, the left side of the vertical axis shows the projected range of protons, the right side of the vertical axis shows the percentage of leakage current non-defective products, and the horizontal axis shows the acceleration voltage. The dose quantity of protons (H.sup.+) is 110.sup.14 atoms/cm.sup.2. In FIG. 6, the thick solid line indicates the percentage of leakage current non-defective products when a YAG 2 laser (wavelength: 532 nm, penetration length p1: 1.5 m) and an AlGaAs laser (wavelength: 808 nm, penetration length p2: 10 m) as a GaAs semiconductor laser were applied simultaneously. The thin solid line indicates the percentage of leakage current non-defective products when only the YAG 2 laser was applied. The thick broken line indicates the percentage of leakage current non-defective products when only the AlGaAs laser was applied. The thin one-dot chain line indicates the projected range of protons.

(61) In this measurement, a voltage of 1200 V is applied between the gate electrode and the emitter electrode so that the resulting product is judged to be non-defective when the current density of the leakage current is lower than 1 A/cm.sup.2 but the resulting product is judged to be defective when the current density of the leakage current is not lower than 1 A/cm.sup.2. The percentage of leakage current non-defective products is the percentage of non-defective chips to the total number of chips in one wafer before the wafer is cut into individual chips by dicing after completion of the wafer producing process.

(62) As shown in FIG. 6, it is found that the percentage of leakage current non-defective products exceeds 95% at an acceleration voltage of 300 keV to 1.3 MeV when the YAG 2 laser and the AlGaAs laser are applied simultaneously. On the other hand, when only the YAG 2 laser is applied, the percentage of leakage current non-defective products is about 90% at an acceleration voltage of not higher than 100 keV but the percentage of leakage current non-defective products is about 50% at an acceleration voltage of not lower than 200 keV. When only the AlGaAs laser is applied, the percentage of leakage current non-defective products is about 50% at maximum.

(63) The projected range of protons at the acceleration voltage exhibiting a high percentage of leakage current non-defective products is in a range of about 3 m to 30 m. On the other hand, the penetration length p1 of the YAG 2 laser into silicon is about 1.5 m, and the penetration length p2 of the AlGaAs laser into silicon is about 10 m.

(64) Accordingly, when only the YAG 2 laser is applied, protons are not provided as donors in a region deeper than 1.5 m because the penetration length p1 of about 1.5 m is so small that protons cannot be activated in the region deeper than 1.5 m. In addition, a lot of damages (defects) due to ion injection remain. Because the defects are dispersed in the wafer, the depletion layer reaches the P.sup.+ collector layer to cause a punch-through phenomenon. As a result, chips large in leakage current are produced, so that the percentage of leakage current non-defective products is reduced.

(65) When only the AlGaAs laser is applied, the temperature of the front surface (opposite to the laser-irradiated surface) of the wafer increases to about 600 C. because the penetration length p2 takes a large value of about 10 m. For this reason, when, for example, an aluminum electrode is formed on a surface opposite to the laser-irradiated surface, the aluminum electrode is melted if the laser output is unchanged as it is. It is therefore necessary to limit the laser output so that the aluminum electrode is not melted. Accordingly, the penetration length is shortened so that provision of protons as donors and recovery of defects cannot be achieved in a region deeper than the penetration length. As a result, a punch-through phenomenon occurs, so that chips large in leakage current are produced to reduce the percentage of leakage current non-defective products.

(66) The reason why two kinds of lasers different in penetration length and wavelength are applied will be described. FIGS. 7-1 to 7-3 are characteristic graphs showing temperature at a distance from the wafer surface (laser-irradiated surface), that is, at a depth, versus elapsed time from a point of time when laser irradiation was terminated. FIG. 7-1 shows the case where the YAG 2 laser and the AlGaAs laser are applied simultaneously. FIG. 7-2 shows the case where only the YAG 2 laser is applied. FIG. 7-3 shows the case where only the AlGaAs laser is applied.

(67) The temperature distribution was obtained by calculation of difference so that laser beam absorption, heat conduction in solid, radiation from surface and natural convection heat radiation were contained.

(68) Laser beam absorption I.sub.OUT was calculated with laser transmittance characteristic as I(x)=10EXP(ux), linear absorption coefficient as u=11789.73 cm.sup.1, and reflectance of silicon to the YAG 2 laser as 38%. Accordingly, laser beam absorption I.sub.OUT satisfies I.sub.OUT(10.38)=10 for the laser output.

(69) Heat conduction dQ/dt in solid is given by the expression (5).
dQ/dt=S.Math..Math.T/x(5)
in which Q is a heat quantity, S is a sectional area, and is a heat conductivity.

(70) Radiation E from surface is given by the expression (6).
E=.Math..Math.T4(6)
in which is a Boltzmann constant, and is a radiativity. The natural convection heat radiation was calculated with a natural convection heat radiation coefficient as 10 W/m.sup.2K (assumed value).

(71) The region where the effect of laser annealing appears in a short time is a region having a higher temperature than the melting point of the substrate. For example, when the substrate is a silicon substrate, this is a region having a higher temperature than 1414 C. which is the melting point of silicon. When the YAG 2 laser and the AlGaAs laser are applied simultaneously, the temperature at a depth of 30 m with the elapsed time up to 300 ns exceeds the melting point (1414 C.) of silicon as shown in FIG. 7-1. The temperature at a depth of 50 m is 600 C. which is lower than the melting point (660 C.) of aluminum, so that aluminum is not melted in a range having a depth of 50 m or more.

(72) Though not shown, the temperature at a depth of 50 m with the elapsed time of 100 s or more is 400 C. or lower. Accordingly, the aluminum electrode formed on the front surface (opposite to the laser-irradiated surface) of the wafer is not melted. Generally, when annealing is performed at a temperature of 300-400 C. for several hours, protons can be provided as donors. In the case of laser annealing, it is however necessary to change the state of silicon to a melted state or a state close to the melted state because the time required for reducing the temperature after temperature rise is a very short time of 1 s. According to the invention, because the temperature at a depth of 30 m or less exceeds the melting point of silicon, protons can be provided as donors by laser annealing. Accordingly, it is understood that Xp is preferably not larger than 30 m.

(73) On the other hand, when only the YAG 2 laser is applied, the temperature at a depth of 5 m or less exceeds the melting point of silicon but becomes lower than the melting point of silicon with the elapsed time of 200 ns or more as shown in FIG. 7-2. Because protons can be provided as donors only at a depth of 5 m or less, protons cannot be provided as donors sufficiently when the projected range is deeper than 5 m. At an acceleration voltage of 100 keV or lower to set the projected range of protons to be 1 m or less, the first buffer layer serves as a field stop layer so that the percentage of leakage current non-defective products is not lower than 90% as shown in FIG. 6. However, when the acceleration voltage becomes higher than 100 keV, the donor integrated concentration of the N.sup.+ first buffer layer cannot reach a donor integrated concentration (1.210.sup.12 atoms/cm.sup.2 or higher) necessary for performing the function of the field stop layer and defects are dispersed in the wafer so that the depletion layer reaches the P.sup.+ collector layer in an off state to cause increase of the leakage current. For this reason, the percentage of leakage current non-defective products is reduced as shown in FIG. 6.

(74) When only the AlGaAs laser is applied, the temperature at a depth of 5 m or less exceeds the melting point of silicon as shown in FIG. 7-3. The temperature even at a depth of 1 m becomes lower than the melting point of silicon when the elapsed time is 200 ns or more. Accordingly, the depth where protons can be provided as donors is only 1 m or less. In this case, protons cannot be provided as donors even at an acceleration voltage of 100 keV, so that a lot of chips having the N.sup.+ first buffer layer incapable of performing the function of the field stop layer are formed. For this reason, the percentage of leakage current non-defective products is reduced as shown in FIG. 6.

(75) In this manner, a depth which is obtained by simultaneous irradiation with the YAG 2 laser and the AlGaAs laser so that the temperature at the depth exceeds the melting point of silicon becomes larger than the sum of depths which are obtained by single irradiation with the YAG 2 laser and the AlGaAs laser respectively so that the temperature at each of the depths exceeds the melting point of silicon. It is conceived that this is caused by occurrence of such a nonlinear synergistic effect that the melting state and temperature distribution generated by irradiation with a short-wavelength laser such as a YAG 2 laser are knocked on to a deeper region by irradiation with a long-wavelength laser such as an AlGaAs laser (hereinafter referred to as knock-on effect). The penetration length of the short-wavelength laser is about 1 m and, preferably, 0.3-5 m. Specifically, a solid-state pulse (all-solid-state) laser with a wavelength of about 500 nm such as YAG 2, YVO.sub.4 2 or YLF 2 is preferred, or a gas laser such as an excimer laser or a helium-neon (HeNe) laser or a gallium nitride (GaN) semiconductor laser may be used instead. Actually, the solid-state pulse laser is preferred to the gas laser because high output power is required.

(76) The penetration length of the long-wavelength laser is about 10 m and, preferably, 5-30 m, both inclusively. Specifically, a semiconductor laser with a wavelength of 700-3500 nm such as Al.sub.XGa.sub.1-XAs or In.sub.XGa.sub.1-XAs is preferred. The subscript X given to an element of the semiconductor laser indicates a stoichiometric composition ratio (normalized composition ratio). The wavelength can be controlled in the aforementioned range when the value of X is adjusted. As the long-wavelength laser, a ruby laser may be used instead. These lasers can perform continuous oscillation. Alternatively, a solid-state pulse laser of a normal frequency such as YAG, YVO.sub.4 or YLF may be used instead.

(77) Incidentally, another laser than the aforementioned lasers may be used as the short-wavelength laser or the long-wavelength laser if the same penetration length into silicon can be obtained.

(78) FIG. 8 shows a result of measurement of leakage current versus annealing temperature in a 1200 V class semiconductor device annealed in a related-art type electric furnace as a comparative example. In FIG. 8, the vertical axis shows leakage current, and the horizontal axis shows annealing temperature. Annealing in the electric furnace is different from laser annealing in that the whole of the wafer is heated to almost the same temperature by annealing in the electric furnace.

(79) In the related-art semiconductor device shown in FIG. 8, a field stop layer formed from phosphorus is provided to prevent the leakage current from changing in accordance with change in the concentration of proton donors in accordance with the annealing temperature. The time required for annealing in the electric furnace is, for example, 1 hour.

(80) As shown in FIG. 8, the leakage current of the device only irradiated with protons without heat treatment is about 20 A (as plotted at a left end of FIG. 8). The reason why the leakage current is large as described above is that the depletion layer reaches a proton-projected range region to thereby increase the level density of deep levels (about 0.2 eV from the mid-cap) due to a large deal of defects in the projected range region and reduce the lifetime of the projected range region (about 10 ns).

(81) On the other hand, when the temperature for heat treatment becomes higher than 300 C., the leakage current is reduced rapidly. At a temperature of 400 C. or higher, the leakage current converges to the order of 10 nA which is almost equal to the value of the leakage current without proton irradiation. This shows that at a temperature of 400 C. or higher, defects induced by proton irradiation are almost recovered to extinguish the deep levels to thereby raise the lifetime to a value equal to the bulk lifetime (50 s or longer). That is, the lifetime becomes sufficient long when the temperature of the projected range region reaches about 400 C.

(82) On the contrary, in the semiconductor device according to Embodiment 1, as shown in FIG. 7-1, a region deep by a smaller value than 50 m from the irradiated surface is heated to a temperature of 600 C. or higher by laser irradiation, and silicon is melted in a region deep by a smaller than 30 m from the irradiated surface. Accordingly, at least in the proton-projected range region, deep levels are extinguished and the lifetime is recovered to the value of the bulk lifetime. Moreover, shallow levels indicating donors (shallow donors) are not extinguished. This reason is that the melting temperature holding time is very short (shorter than 10 s) differently from annealing in the electric furnace. It is obvious from the above description that increase of conduction loss caused by reduction of the lifetime can be suppressed in the semiconductor device according to Embodiment 1 because shallow donors can be formed without reduction of the lifetime differently from the related-art type semiconductor device.

(83) Next, values of spreading resistance were measured in the semiconductor device according to Embodiment 1 and the semiconductor device annealed in the electric furnace according to the related art. In the semiconductor device according to the related art, annealing was performed in the electric furnace at 350 C. for 1 hour after the processes of FIGS. 2-1 to 2-3 were performed, that is, after proton irradiation. FIG. 9 shows results of conversion of the measured resistance values into carrier concentrations. In FIG. 9, the vertical axis shows a net doping concentration, and the horizontal axis shows a distance from the wafer surface (proton-irradiated surface), that is, a depth from the wafer surface. As shown in FIG. 9, the peak value of the carrier concentration in the invention is 4.0510.sup.15 atoms/cm.sup.3 which is equal to the peak value of the net doping concentration in the related art. On the other hand, the full width at half maximum of the net doping concentration in the invention is 3.8 m which is about 2.1 times as wide as the full width (1.8 m) at half maximum of the net doping concentration in the related art. This is caused by the aforementioned knock-on effect.

(84) Next, the percentage of leakage current non-defective products versus a value obtained by multiplying the peak value of the net doping concentration of the first buffer layer and the full width at half maximum thereof will be described. FIG. 10 is a graph showing the percentage of leakage current non-defective products versus a value obtained by multiplying the peak value of the net doping concentration and the full width at half maximum. The peak value of the net doping concentration is the peak value shown in FIG. 9. The decreasing rate of electric field intensity in the field stop layer is decided by the integrated concentration of the field stop layer. The concentration of the field stop layer formed by ion injection conforms almost to the Gaussian distribution. A method in which a value obtained by multiplying the peak concentration and the full width at half maximum and dividing the resulting product by 2 is approximated as the integrated concentration of the field stop layer on the assumption that the field stop layer is regarded as a triangle and the full width at half maximum and peak concentration of the field stop layer are regarded as the bottom side and height of the triangle, respectively, is used here as a method of calculating the integrated concentration of the field stop layer more easily and accurately. This value is hereinafter referred to as F value. FIG. 10 shows the F value calculated based on numerical values shown in FIG. 9. As shown in FIG. 10, when the F value is not smaller than 5.010.sup.11 atoms/cm.sup.2, the percentage of leakage current non-defective products becomes 90% or higher. Accordingly, the F value of not smaller than 510.sup.11 atoms/cm.sup.2 is preferred and the F value of not smaller than 610.sup.11 atoms/cm.sup.2 is further preferred. This reason is that the field stop layer stops the depletion layer when the F value is not smaller than 510.sup.11 atoms/cm.sup.2. In addition, when the peak value of the carrier concentration is not smaller than 110.sup.15 atoms/cm.sup.3, the full width at half maximum of not smaller than 2.5 m is preferred and the full width at half maximum of not smaller than 3 m is further preferred. This reason is that the field stop layer stops the depletion layer when the peak value of the carrier concentration is in these ranges.

(85) According to Embodiment 1, after a surface structure is formed in a front surface of a wafer and a rear surface of the wafer is ground, the rear surface of the wafer is irradiated with protons and then simultaneously irradiated with two kinds of lasers different in wavelength, so that the field stop layer can be formed in a position deep from the rear surface of the wafer without any influence on the surface structure formed in the front surface of the wafer. Accordingly, even when an avalanche current begins to flow based on application of a high voltage between the collector and the emitter in an off state of the device, the distance from an end portion of the depletion layer to the P.sup.+ collector layer becomes long because the distance from a position where the net doping concentration of the N.sup.+ field stop layer is locally maximized to the interface between the second N buffer layer and the P.sup.+ collector layer is long. For this reason, the current is not concentrated into one place of the device, so that the snap-back phenomenon can be prevented. Hence, the device can be prevented from being destroyed by the snap-back phenomenon.

(86) A semiconductor device according to Embodiment 2 will be described below. In the semiconductor device according to Embodiment 2, an N.sup.+ first buffer layer is formed by injection of other ions than protons, differently from Embodiment 1. FIGS. 11 to 13 are explanatory graphs showing results of measurement of the projected range of ions and the percentage of leakage current non-defective products versus the acceleration voltage at the time of injection of each of helium ions (He.sup.2+), lithium ions (Li.sup.+) and oxygen ions (O.sup.+). In FIGS. 11 to 13, the left side of the vertical axis shows the projected range of ions, the right side of the vertical axis shows the percentage of leakage current non-defective products, and the horizontal axis shows the acceleration voltage. The dose quantity of ions is 110.sup.14 atoms/cm.sup.2.

(87) In FIGS. 11 to 13, the thick solid line expresses the percentage of leakage current non-defective products when a YAG 2 laser (wavelength: 532 nm, penetration length p1: 1.5 m) and an AlGaAs laser (wavelength: 808 nm, penetration length p2: 10 m) as a GaAs semiconductor laser are applied simultaneously. The thin one-dot chain line expresses the projected range of ions.

(88) As shown in FIG. 11, when helium ions are injected, the range of the acceleration voltage exhibiting 90% or higher as the percentage of leakage current non-defective products is narrow (about 900 keV to 3 MeV) compared with the case of injection of protons shown in FIG. 6. This reason is that the knock-on effect hardly occurs because the projected range of helium ions as expressed by the thin one-dot chain line in FIG. 11 is shallower than the projected range of protons as expressed by the thin one-dot chain line in FIG. 6.

(89) As shown in FIG. 12, when lithium ions are injected, the range of the acceleration voltage exhibiting 90% or higher as the percentage of leakage current non-defective products is narrow (about 1.2 MeV to 3 MeV) compared with the case of injection of protons shown in FIG. 6.

(90) As shown in FIG. 13, when oxygen ions are injected, the range of the acceleration voltage exhibiting 90% or higher as the percentage of leakage current non-defective products is about 7 MeV to 20 MeV. To increase the acceleration voltage to 1 MeV or higher, for example, a tandem electrostatic accelerator is used. To increase the acceleration voltage to 10 MeV or higher, for example, a circular accelerator (cyclotron) is used.

(91) Incidentally, there is no existing practical equipment capable of injecting an element heavier than oxygen, such as phosphorus (P) or selenium (Se), into a required depth as the projected range because the projected range is not longer than 1 m even when injection is performed at an acceleration voltage of MeV class.

(92) It is hence to be understood that helium ions, lithium ions or oxygen ions are preferred as other elements to be injected than protons in accordance with Embodiment 2. As shown in FIG. 6 and FIGS. 11 to 13, it is found that the preferred acceleration voltage is 200 keV at minimum and 30 MeV at maximum and the preferred acceleration voltage for injection of protons is 2 MeV at maximum.

(93) FIG. 14 is a view showing the configuration and net doping concentration of a semiconductor device according to Embodiment 3. As shown in a section 200 of the semiconductor device in FIG. 14, the semiconductor device according to Embodiment 3 includes, in addition to the semiconductor device of Embodiment 1 shown in FIG. 1, an N.sup.+ cathode buffer layer 22 (seventh semiconductor layer) provided between the N second buffer layer 12 and the P.sup.+ collector layer 3 and being higher in concentration than the N second buffer layer 12. The other configuration is the same as in Embodiment 1 and like numerals refer to like parts for the sake of omission of duplicate description.

(94) The net doping concentration and size of each portion of the semiconductor device according to Embodiment 3 are exemplified as an example. Incidentally, only values different from the values exemplified in Embodiment 1 will be described. As shown in a characteristic graph 210 of distance from the emitter electrode versus net doping concentration (log) in FIG. 14, the net doping concentration of the N.sup.+ cathode buffer layer 22 is higher than the net doping concentration of the N.sup.+ first buffer layer 2.

(95) The distance from a position where the net doping concentration of the N.sup.+ first buffer layer 2 is locally maximized to the interface between the P.sup.+ collector layer 3 and the collector electrode 9, that is, the projected range Rp of protons is 18 m. The distance from the interface between the N second buffer layer 12 and the N.sup.+ cathode buffer layer 22 to the interface between the P.sup.+ collector layer 3 and the collector electrode 9, that is, the thickness of the N.sup.+ cathode buffer layer 22 is 1 m. The distance from the interface between the N.sup.+ cathode buffer layer 22 and the P.sup.+ collector layer 3 to the interface between the P.sup.+ collector layer 3 and the collector electrode 9, that is, the thickness of the P.sup.+ collector layer 3 is 0.5 m.

(96) According to Embodiment 3, the depletion layer spread from the emitter side into the drift layer can be stopped more surely by the N.sup.+ cathode buffer layer 22 before the depletion layer reaches the P.sup.+ collector layer 3. Accordingly, the incomplete formation of the buffer layer can be prevented, so that the percentage of leakage current non-defective products can be improved more greatly.

(97) FIG. 15 is a graph showing turn-off waveforms of the semiconductor device according to Embodiments 1 to 3 and the semiconductor device including the field stop layer according to the related art. In FIG. 15, the thick solid line expresses the collector current I.sub.C and collector-emitter voltage V.sub.C of the semiconductor device according to Embodiments 1 to 3. The thin broken line expresses the collector current I.sub.C and collector-emitter voltage V.sub.C of the semiconductor device including the field stop layer according to the related art (see FIG. 19).

(98) All the turn-off waveforms shown in FIG. 15 are waveforms in a snubberless circuit 220 shown in FIG. 16. The snubberless circuit 220 is a single-phase chopper circuit in which the IGBT according to the invention is disposed in a D.U.T. (device under test). The load inductance Lm of a main circuit of the snubberless circuit 220 is 1 mH, and the circuit stray inductance Ls of the main circuit is 200 nH. The gate resistance of the snubberless circuit 220 is 10, and the gate drive voltage is 15 V.

(99) As shown in FIG. 15, the semiconductor device according to Embodiments 1 to 3 exhibits smooth switching waveforms without oscillation. The reason why oscillation does not occur is that stocked carriers are prevented from being exhausted at turn-off time. On the contrary, the semiconductor device according to the related art exhibits waveforms oscillated with 50 V increase in surge voltage because carriers are swept by spreading of the depletion layer so that stocked carriers are exhausted. This is a phenomenon the same as a so-called pinning effect of a space charge region as disclosed in JP-A-2003-318412.

(100) FIG. 17 is a view showing the configuration and net doping concentration of a semiconductor device according to Embodiment 4. As shown in a section 300 of the semiconductor device in FIG. 17, an N.sup.+ cathode buffer layer 32 as a second semiconductor layer and an N.sup.++ cathode layer 33 as a third semiconductor layer are formed in this order, for example, on one principal surface side (first principal surface side) of an N.sup. drift layer 31 as a first semiconductor layer. A P anode layer 34 as a fourth semiconductor layer is formed on the other principal surface side (second principal surface side) of the N.sup. drift layer 31. An anode electrode 38 as a first electrode is formed on a surface of the P anode layer 34. A cathode electrode 39 as a second electrode is formed on a surface of the N.sup.++ cathode layer 33. According to this structure, the semiconductor device according to Embodiment 4 performs the function of a diode.

(101) As shown in a characteristic graph 310 of distance from the anode electrode versus net doping concentration (log) in FIG. 17, the net doping concentration of the N+ cathode buffer layer 32 has a peak in a position near the interface between the N.sup.+ cathode buffer layer 32 and the N.sup. drift layer 31 and is higher than the net doping concentration of the N.sup. drift layer 31. The net doping concentration of the N+ cathode buffer layer 32 decreases with a gradient toward the interface between the N.sup.+ cathode buffer layer 32 and the N.sup.++ cathode layer 33. The net doping concentrations of the N.sup.++ cathode layer 33 and the P anode layer 34 are higher than the net doping concentrations of the N.sup. drift layer 31 and the N.sup.+ cathode buffer layer 32.

(102) In this manner, application of the invention to a diode permits achievement of an inverse recovery operation with suppression of oscillation as well as low loss. At the time of inverse recovery, the depletion layer is spread from the front surface side in the same manner as at the turn-off time, so that carriers are extinguished. In the semiconductor device according to Embodiment 4, sudden extinction of carriers can be suppressed to thereby achieve smooth inverse recovery without oscillation.

(103) On the other hand, in the IGBT or the like using the invention, turn-off with suppression of oscillation and low loss can be achieved as described in Embodiments 1 to 3. At the turn-off time, the depletion layer is spread from the front surface side of the wafer, so that carriers are extinguished. However, smooth turn-off without oscillation can be achieved because sudden extinction of carriers can be suppressed. Accordingly, a low-loss and soft-recovery diode and an IGBT capable of performing smooth turn-off without oscillation can be produced. In addition, in a power conversion apparatus such as a PWM inverter using an IGBT module having such characteristics, overvoltage destruction and occurrence of EMI noise can be suppressed.

(104) FIGS. 18-1 and 18-2 are diagrams showing applied examples of IGBTs and diodes according to Embodiments 1 to 4. Converter-inverter circuits shown in FIGS. 18-1 and 18-2 can control induction motors, servomotors, etc. efficiently and are widely used in industries, electric railroads, etc. FIG. 18-1 shows an example of application of diodes to a converter portion. FIG. 18-2 shows an example of application of IGBTs to a converter portion. In each of the examples, IGBTs are applied to an inverter portion. A free wheeling diode (FWD) is connected in parallel with each IGBT.

(105) In the above, the invention is not limited to the aforementioned embodiments and may be changed variously. For example, various values such as sizes, concentrations, voltage values, current values, treating conditions such as temperature and time, etc. described in the embodiments are only exemplary, so that the invention is not limited to those values. Although the respective embodiments have been described on the case where the first conductivity type is an N type and the second conductivity type is a P type, the invention can be effected also in the case where the first conductivity type is a P type and the second conductivity type is an N type.

(106) Moreover, the invention can be applied not only to a 1200 V class but also to a 600 V class, a 1700 V class or a withstand voltage class higher than the 1700 V class. For example, in the case of a 1700 V class, the specific resistance of the wafer is 80-200 cm and the final thickness of the wafer is 120-200 m. In the case of a 3300 V class, the specific resistance of the wafer is 200-500 cm and the final thickness of the wafer is 250-400 m.

(107) As described above, the semiconductor device and the method of producing the semiconductor device according to the invention are useful for a power semiconductor device and particularly adapted to a diode or an IGBT having soft recovery characteristics as well as high-speed and low-loss characteristics and having environmental friendliness.

(108) The invention has been described with reference to certain preferred embodiments thereof. It will be understood, however, that modifications and variations are possible within the scope of the appended claims.

(109) This application is based on, and claims priority to, Japanese Patent Application No: 2008-013018, filed on Jan. 23, 2008. The disclosure of the priority application, in its entirety, including the drawings, claims, and the specification thereof, is incorporated herein by reference.