Fault inspection method, fault inspection device, and method for producing silicon carbide chip

Abstract

A fault inspection method is provided by which a basal plane dislocation (BPD) converted into a threading edge dislocation (TED) at the interface between a substrate layer and a buffer layer can be detected in a short time by a PL inspection.　The fault inspection method comprises: an irradiation step in which the whole of a silicon carbide substrate 100 is irradiated with first ultraviolet light L1, which expands a Shockley stacking fault (SSF) from a conversion point at which a BPD has been converted to a TED, at a wavelength and a UV irradiance which make the first ultraviolet light L1 reach at least the interface I between a substrate layer 101 and a buffer layer 102; a visualization step in which the whole of the silicon carbide substrate 100 is irradiated with second ultraviolet light L2, which has a lower irradiance than the first ultraviolet light L1, to cause the silicon carbide substrate 100 to emit PL light and thereby perform visualization; an imaging step in which an image of the silicon carbide substrate 100 that is emitting the PL light is taken; and a determination step in which it is deemed that in the PL light image taken in the imaging step, a BPD converted into a TED is present in a region having an image of an SSF.

Description

Defect inspection method, defect inspection device, and silicon carbide chip manufacturing method

[0001] The present invention relates to a defect inspection method and defect inspection device for detecting basal plane dislocations (BPDs) that have converted into threading edge dislocations (TEDs) at the interface between a substrate layer and a buffer layer in a silicon carbide substrate having a substrate layer, a buffer layer, and a drift layer, and a method for manufacturing a silicon carbide chip.

[0002] Silicon carbide is a wide-gap semiconductor and has recently attracted attention in the field of power devices. Silicon carbide substrates used in power devices are typically manufactured by epitaxial growth, in which a drift layer is grown on a substrate. However, due to the large difference in atomic radii between carbon and silicon atoms, silicon carbide is more susceptible to crystal defects in the drift layer than gallium nitride, another wide-gap semiconductor. Crystal defects are a cause of forward current degradation, which affects device performance, and therefore pose a problem in the manufacture of power devices.

[0003] Forward current degradation in power devices is thought to be caused by carrier capture at basal plane dislocations (hereinafter referred to as "BPDs") in the drift layer, and the recombination energy of the captured carriers causes Shockley stacking faults (hereinafter referred to as "SSFs"). Therefore, a technology has been developed to suppress the formation of BPDs in the drift layer during the epitaxial growth process. For example, by setting the off-angle to 4°, more than 95% of the crystal defects inherited from BPDs in the substrate layer to the drift layer formed by epitaxial growth can be structurally converted into threading edge dislocations (hereinafter referred to as "TEDs").

[0004] In silicon carbide substrates used in power devices, in order to suppress the expansion of SSF in the drift layer, a buffer layer with a short carrier lifetime is inserted at the interface between the substrate layer and the drift layer, thereby reducing the carrier density injected into the drift layer when current is applied.

[0005] However, in power devices for automotive applications, a large current exceeding the rated current may flow due to special circumstances, such as when the device derails. In such cases, SSFs may expand from BPDs converted to TEDs at the interface between the buffer layer and the drift layer, and from BPDs converted to TEDs at the interface between the substrate layer and the buffer layer. Furthermore, even if the current density is small, SSFs may expand due to long-term use. Therefore, in order to further improve the stability of power devices for automotive applications, it is desirable to detect, in the inspection of silicon carbide substrates, not only BPDs inherited in the drift layer, but also BPDs converted to TEDs at the interface between the buffer layer and the drift layer, and BPDs converted to TEDs at the interface between the substrate layer and the buffer layer.

[0006] A defect inspection method using photoluminescence (hereinafter referred to as "PL") has been proposed as a technique for detecting crystal defects in a silicon carbide substrate (see, for example, Patent Document 1). The defect inspection method described in Patent Document 1 irradiates the entire silicon carbide substrate with ultraviolet light, selects candidate regions where BPDs exist, and irradiates the candidate regions with high-intensity ultraviolet light to expand SSFs from BPDs in the drift layer and BPDs in the buffer layer that have been converted to TEDs. The candidate regions where SSF defect images occur are detected as regions where BPDs exist. In the defect inspection method of Patent Document 1, carriers generated by irradiation with high-intensity ultraviolet light are captured by BPDs in the drift layer and BPDs converted to TEDs at the interface between the buffer layer and the drift layer, expanding the SSFs. As a result, not only BPDs in the drift layer but also BPDs in the buffer layer that have been converted to TEDs can be detected by PL inspection.

[0007] International Publication No. 2021 / 177127

[0008] However, in the epitaxial growth method, as described above, the conversion position from BPD to TED may be at the interface between the buffer layer and the drift layer, or at the interface between the substrate layer and the buffer layer. Here, the shapes of the BPD in the drift layer and the SSFs expanded from the BPD converted to TED at the interface between the buffer layer and the drift layer are triangular, whereas the shape of the SSFs expanded from the BPD converted to TED at the interface between the substrate layer and the buffer layer is elongated strip-like. Therefore, the defect area of ​​the SSFs expanded from the BPD converted to TED at the interface between the substrate layer and the buffer layer is larger than that of the SSFs expanded from the BPD converted to TED at the interface between the buffer layer and the drift layer, and this is more likely to cause product defects.

[0009] In this regard, the defect inspection method of Patent Document 1 inspects BPDs present in the drift layer and BPDs converted to TEDs at the interface between the buffer layer and the drift layer, but does not anticipate the inspection of BPDs converted to TEDs at the interface between the substrate layer and the buffer layer. Furthermore, the defect inspection method of Patent Document 1 selects candidate regions and irradiates only the candidate regions with high-intensity ultraviolet light to expand SSFs from BPDs, which results in a large number of steps and a long time required for defect inspection. Silicon carbide substrates have become larger in diameter thanks to recent improvements in manufacturing technology, and the inspection time problem becomes even more pronounced when the defect inspection method of Patent Document 1 is applied to large-diameter silicon carbide substrates.

[0010] The present invention has been made in consideration of the above problems, and aims to provide a defect inspection method and defect inspection device that can detect BPDs converted to TEDs at the interface between a substrate layer and a buffer layer in a short time by PL inspection. It is also an object of the present invention to provide a silicon carbide chip manufacturing method that manufactures silicon carbide chips from a silicon carbide substrate inspected by the defect inspection method of the present invention.

[0011] A characteristic configuration of the defect inspection method according to the present invention for solving the above problems is as follows: A defect inspection method for detecting basal plane dislocations (BPDs) that have converted into threading edge dislocations (TEDs) at an interface between a substrate layer and a buffer layer in a silicon carbide substrate having a substrate layer, a buffer layer, and a drift layer, comprising: an irradiation step of irradiating the entire silicon carbide substrate with first ultraviolet light that expands Shockley stacking faults (SSFs) from a conversion point at which the basal plane dislocations (BPDs) have converted into threading edge dislocations (TEDs), at a wavelength and UV irradiance that reaches at least the interface between the substrate layer and the buffer layer; a visualization step of irradiating the entire silicon carbide substrate with second ultraviolet light having an irradiance lower than that of the first ultraviolet light, thereby visualizing the silicon carbide substrate by causing it to emit photoluminescence; and an imaging step of imaging the silicon carbide substrate emitting photoluminescence. and a determination step of determining that a basal plane dislocation (BPD) that has converted into a threading edge dislocation (TED) is present in an area where a defect image of a Shockley stacking fault (SSF) occurs in the PL light emission image captured in the photographing step.

[0012] According to the defect inspection method of this configuration, the irradiation step irradiates the entire silicon carbide substrate with first ultraviolet light at a wavelength and UV irradiance that reaches at least the interface between the substrate layer and the buffer layer. Carriers generated by the irradiation of the first ultraviolet light are captured by BPDs converted to TEDs at the interface between the substrate layer and the buffer layer, allowing SSFs to expand from the conversion point of the BPDs to TEDs. Next, the visualization step irradiates the entire silicon carbide substrate with second ultraviolet light, allowing the BPDs converted to TEDs at the interface between the substrate layer and the buffer layer to be detected by PL inspection. Furthermore, since the irradiation step irradiates the entire silicon carbide substrate with the first ultraviolet light to expand the SSFs, the number of steps is reduced, enabling defects to be detected in a short time.

[0013] In the defect inspection method according to the present invention, it is preferable that in the irradiating step, the silicon carbide substrate is heated to 100 to 200° C. and then irradiated with the first ultraviolet light.

[0014] According to the defect inspection method of this configuration, by irradiating the silicon carbide substrate with the first ultraviolet light while it is heated to 100 to 200° C., the carrier generation efficiency is improved, SSFs can be expanded in a short time, and the inspection time can be shortened. Furthermore, by setting the temperature during irradiation with the first ultraviolet light to a temperature expected to be the operating temperature (100 to 200° C.) in an automobile engine compartment or the like, it is possible to appropriately detect BPDs that are converted to TEDs at the interface between the substrate layer and the buffer layer in an automotive power device and that may cause SSF expansion.

[0015] In the defect inspection method according to the present invention, the first ultraviolet light has a wavelength of 325 to 365 nm and a UV illuminance of 10 W / cm ２ It is preferable that this is equal to or greater than this.

[0016] According to the defect inspection method of this configuration, the first ultraviolet light has a wavelength of 325 to 365 nm and a UV illuminance of 10 W / cm ２ For these reasons, sufficient carriers can be generated in the substrate layer, and the SSF can be appropriately expanded from the BPDs converted to TEDs at the interface between the substrate layer and the buffer layer. Furthermore, by setting the wavelength and UV irradiance within appropriate ranges, BPDs that are likely to have SSFs expanded among the BPDs converted to TEDs at the interface between the substrate layer and the buffer layer can be appropriately detected.

[0017] In the defect inspection method according to the present invention, it is preferable that the first ultraviolet light is a light beam having a concentric illuminance distribution in which the illuminance is relatively high in the center and relatively low in the peripheral area, and that the first ultraviolet light is irradiated in pulses while being shifted at regular intervals so that there are continuous circular areas in which the illuminance of the light beam is equal to or higher than a predetermined value.

[0018] According to the defect inspection method of this configuration, when first ultraviolet light, which has a higher illuminance toward the center of the light beam, is irradiated onto the silicon carbide substrate, the light is irradiated in pulses while being shifted at regular intervals so that circular areas where the illuminance is above a predetermined level are continuous.Therefore, the first ultraviolet light with the illuminance required to expand the SSF can be irradiated onto the entire silicon carbide substrate while averaging the illuminance distribution, and the SSF can be uniformly expanded from the BPD converted to TED at the interface between the substrate layer and the buffer layer.

[0019] In the defect inspection method according to the present invention, it is preferable that in the irradiating step, the first ultraviolet light is irradiated using a galvano scanner.

[0020] According to the defect inspection method having this configuration, the first ultraviolet light is irradiated using a galvano scanner, so that the silicon carbide substrate can be irradiated with the first ultraviolet light with high precision and uniformity.

[0021] Preferably, the defect inspection method according to the present invention further comprises a shrinking step of shrinking Shockley stacking faults (SSFs) by heating the silicon carbide substrate in which Shockley stacking faults (SSFs) have expanded from a conversion point at which basal plane dislocations (BPDs) present at the interface between the substrate layer and the buffer layer are converted to threading edge dislocations (TEDs).

[0022] According to the defect inspection method of this configuration, the silicon carbide substrate after inspection can be directly commercialized by heating and contracting the SSFs expanded from the conversion point where BPDs present at the interface between the substrate layer and the buffer layer are converted to TEDs. Therefore, the defect inspection method of this configuration enables 100% inspection in the manufacture of power devices.

[0023] The defect inspection device according to the present invention for solving the above problems has a characteristic configuration, and is a defect inspection device for detecting basal plane dislocations (BPDs) that have converted into threading edge dislocations (TEDs) at an interface between a substrate layer, a buffer layer, and a drift layer in a silicon carbide substrate, the defect inspection device comprising: a defect expansion irradiation means that irradiates the entire silicon carbide substrate with first ultraviolet light, which expands Shockley stacking faults (SSFs) from the conversion point at which the basal plane dislocations (BPDs) have converted into threading edge dislocations (TEDs), at a wavelength and UV irradiance that reaches at least the interface between the substrate layer and the buffer layer; a visualization irradiation means that irradiates the entire silicon carbide substrate with second ultraviolet light, which has an irradiance lower than that of the first ultraviolet light, to cause the silicon carbide substrate to emit photoluminescence; and an imaging means that images the silicon carbide substrate emitting photoluminescence. and a determination means for determining that a basal plane dislocation (BPD) that has converted into a threading edge dislocation (TED) is present in an area where a defect image of a Shockley stacking fault (SSF) occurs in the PL light emission image captured by the imaging means.

[0024] According to the defect inspection device of this configuration, the defect expansion irradiation means irradiates the entire silicon carbide substrate with first ultraviolet light at a wavelength and UV irradiance that reaches at least the interface between the substrate layer and the buffer layer. Carriers generated by the irradiation of the first ultraviolet light are captured by BPDs converted to TEDs at the interface between the substrate layer and the buffer layer, and SSFs can be expanded from the conversion point of the BPDs to TEDs. Next, the visualization irradiation means irradiates the entire silicon carbide substrate with second ultraviolet light, allowing the BPDs converted to TEDs at the interface between the substrate layer and the buffer layer to be detected by PL inspection. Furthermore, because the defect expansion irradiation means irradiates the entire silicon carbide substrate with the first ultraviolet light to expand the SSFs, the number of processes is reduced, enabling defects to be detected in a short time.

[0025] A characteristic configuration of the method for manufacturing a silicon carbide chip according to the present invention for solving the above-mentioned problems is a method for manufacturing a silicon carbide chip from a silicon carbide substrate in which the Shockley stacking faults (SSFs) have been shrunk by the defect inspection method described above, comprising: a mapping step of creating mapping data in which regions determined in the determination step to contain basal plane dislocations (BPDs) that have converted into threading edge dislocations (TEDs) are mapped onto the entire silicon carbide substrate; a circuit formation step of forming a circuit on the silicon carbide substrate; a cutting step of cutting the silicon carbide substrate into a plurality of chips; and a excluding step of excluding defective chips that include regions determined to contain basal plane dislocations (BPDs) that have converted into threading edge dislocations (TEDs) based on the mapping data.

[0026] According to the method for manufacturing a silicon carbide chip of this configuration, the area determined to contain BPD converted to TED at the interface between the substrate layer and the buffer layer is mapped across the entire silicon carbide substrate, and after the circuit formation process and cutting process, defective chips containing the area are excluded based on the mapping data.This makes it possible to manufacture silicon carbide chips that are free of crystal defects that cause forward current degradation and are highly stable.

[0027] FIG. 1 is a schematic diagram of a silicon carbide substrate, where (a) is a cross-sectional view of the silicon carbide substrate, and (b) is a perspective view showing the expansion of SSFs from BPDs converted to TEDs at the interface between the substrate layer and the buffer layer. FIG. 2 is a micrograph of a silicon carbide substrate irradiated with ultraviolet light at a wavelength and UV irradiance sufficient to reach the interface between the substrate layer and the buffer layer, after KOH etching. (a) is a photograph of the drift layer surface, and (b) is a photograph of the substrate layer surface from which the buffer layer and drift layer have been removed by polishing. FIG. 3 is a schematic diagram of a defect inspection device of the present invention. FIG. 4 is an explanatory diagram of the first ultraviolet light of the present invention, where (a) is an explanatory diagram of the irradiance distribution, and (b) is an explanatory diagram of pulse irradiation. FIG. 5 is a flowchart showing the steps of the defect inspection method and the method for manufacturing a silicon carbide chip of the present invention. FIG. 6 is a micrograph (PL emission image) of a silicon carbide substrate irradiated with first ultraviolet light under different temperature conditions and then subjected to PL emission.

[0028] Hereinafter, embodiments of the defect inspection method, defect inspection device, and silicon carbide chip manufacturing method of the present invention will be described in detail with reference to the drawings. However, the present invention is not intended to be limited to the configurations described below. Note that the thickness relationships of the layer structure of the silicon carbide substrate shown in Figure 1 have been appropriately exaggerated or simplified for ease of explanation, and do not strictly reflect the magnitude relationships of the actual thicknesses of the individual layers.

[0029] [Silicon Carbide Substrate] Before describing the defect inspection method, defect inspection device, and silicon carbide chip manufacturing method of the present invention, a silicon carbide substrate used in the manufacture of power devices will be described. FIG. 1 is an explanatory diagram schematically showing a silicon carbide substrate 100, where (a) is a cross-sectional view of the silicon carbide substrate 100 and (b) is a perspective view showing the extension of SSFs from BPDs converted to TEDs at the interface I between a substrate layer 101 and a buffer layer 102. In FIG. 1(b), a portion of the silicon carbide substrate 100 is cut out to show a plane parallel to the basal plane. The basal plane is a plane perpendicular to the c-axis used to identify the crystal structure. In FIG. 1(b), the TEDs present in the cut-out region are shown by dashed lines.

[0030] The silicon carbide substrate 100 has a structure in which a substrate layer 101, a buffer layer 102, and a drift layer 103 are stacked. The substrate layer 101 is a silicon carbide single crystal (4H-SiC) substrate that serves as a base for thin film manufacturing by epitaxial growth. The buffer layer 102 and the drift layer 103 are thin films made of silicon carbide single crystal (4H-SiC) deposited on the substrate layer 101 by epitaxial growth. The buffer layer 102 is doped with impurities such as nitrogen and vanadium at a high concentration, and functions to suppress minority carriers from reaching the drift layer 103 in a power device. The drift layer 103 is doped with impurities at a lower concentration than the buffer layer 102, and functions as the drift layer 103 of a SiC-MOSFET.

[0031] In the silicon carbide substrate 100, the BPDs converted to TEDs at the interface I between the substrate layer 101 and the buffer layer 102 (hereinafter referred to as "latent BPDs") have a short carrier lifetime in the substrate layer 101, and therefore do not emit PL light even when the silicon carbide substrate 100 is irradiated with excitation light.

[0032] The mechanism by which forward current degradation occurs in power devices manufactured using silicon carbide substrate 100 is that carriers generated by current flow are captured by BPDs and recombine, causing SSFs to expand from the BPDs due to recombination energy, resulting in changes in device characteristics. At this time, the BPDs in drift layer 103, the BPDs converted to TEDs at the interface between buffer layer 102 and drift layer 103, and the SSFs expanding from latent BPDs expand from BPDs formed parallel to the basal plane along planes parallel to the basal plane, becoming triangular or band-like. The SSFs expanding from latent BPDs in substrate layer 101 become elongated band-like as shown by the arrows in FIG. 1( b ), and have a larger area than triangular SSFs, thereby significantly affecting the power device.

[0033] Here, the latent BPDs themselves in the substrate layer 101 do not emit PL light even when the silicon carbide substrate 100 is irradiated with excitation light, but the SSFs extending from the latent BPDs emit PL light when irradiated with excitation light. Therefore, by extending the SSFs from the latent BPDs in the substrate layer 101 before capturing the PL light and extracting a defect image of the band-shaped SSFs in the captured image of the PL light (hereinafter referred to as a "PL light image"), it is believed possible to detect an area containing latent BPDs that may cause forward current degradation when a large current is passed through the power device. Therefore, in order to perform PL inspection of the silicon carbide substrate 100 at the wafer stage before manufacturing the power device, chemical etching was performed to confirm whether it was possible to extend the SSFs from the latent BPDs by irradiating the entire silicon carbide substrate 100 with ultraviolet light at a wavelength and UV irradiance that reached at least the interface I between the substrate layer 101 and the buffer layer 102.

[0034] 2 shows micrographs of KOH etching of a silicon carbide substrate 100 irradiated with ultraviolet light at a wavelength and UV irradiance sufficient to reach the interface I between the substrate layer 101 and the buffer layer 102, where (a) is a photograph of the drift layer surface and (b) is a photograph of the substrate layer surface from which the buffer layer and drift layer have been removed by polishing. When the silicon carbide substrate 100 is immersed in molten potassium hydroxide (KOH), the surface of the silicon carbide substrate 100 is eroded by the alkali. However, crystal defects such as dislocations are regions where the atomic arrangement is disordered, and therefore are chemically more active than the surrounding area. Therefore, they are eroded more quickly than normal crystal portions, and are selectively etched from the dislocation cores, forming etched depressions called etch pits. The type of crystal defect can be detected based on the shape and size of the etch pits. As shown in FIG. 2( a), KOH etching was performed on silicon carbide substrate 100 irradiated with ultraviolet light at a wavelength and UV irradiance that reached interface I between substrate layer 101 and buffer layer 102, and TED and extended SSFs were observed on the surface of drift layer 103. Here, when buffer layer 102 and drift layer 103 were removed by polishing, BPDs that were conversion points to TED were observed on the surface of substrate layer 101, as shown in FIG. 2( b). From this, it is considered that SSFs were extended from BPDs (i.e., latent BPDs) that were converted to TED at interface I between substrate layer 101 and buffer layer 102 by irradiating silicon carbide substrate 100 with ultraviolet light at a wavelength and UV irradiance that reached interface I between substrate layer 101 and buffer layer 102.

[0035] Based on this knowledge, the present inventors have developed a defect inspection method and defect inspection device that can detect potential BPDs by irradiating the entire silicon carbide substrate 100 with ultraviolet light at a wavelength and UV irradiance that reaches at least the interface I between the substrate layer 101 and the buffer layer 102 at the wafer stage before manufacturing a power device.

[0036] [Defect Inspection Apparatus] Figure 3 is a schematic diagram of a defect inspection apparatus 1 of the present invention. The defect inspection apparatus 1 is an apparatus that inspects a silicon carbide substrate 100 placed on a mounting table 6 by utilizing PL light generated when a semiconductor is irradiated with excitation light. As shown in Figure 3, the defect inspection apparatus 1 includes a defect expansion irradiation means 2, a visualization irradiation means 3, an imaging means 4, and a determination means 5. The operations of the defect expansion irradiation means 2, the visualization irradiation means 3, the imaging means 4, and the determination means 5 are controlled by a control means 7. Each component of the defect inspection apparatus 1 will be described in detail below.

[0037] <Irradiation means for expanding defects> The irradiation means for expanding defects 2 is a means for irradiating the entire silicon carbide substrate 100 with first ultraviolet light L1, which expands SSFs from the conversion point where BPDs are converted to TEDs, at a wavelength and UV irradiance that reaches at least the interface I between the substrate layer 101 and the buffer layer 102. The wavelength of the first ultraviolet light L1 is preferably 325 to 365 nm, more preferably 355 nm. The UV irradiance of the first ultraviolet light L1 is 10 W / cm ２ More than 100 W / cm is preferable. ２ The wavelength of the first ultraviolet light L1 is 325 to 365 nm, and the UV illuminance is 10 W / cm ２ As a result, the first ultraviolet light L1 reaches at least the interface I between the substrate layer 101 and the buffer layer 102, generating carriers in the substrate layer 101 and expanding the SSF from the latent BPD. ２ If the wavelength is smaller than this, the SSF expansion rate will be slow, which may hinder the rapid inspection desired in the power device manufacturing process. Thus, by setting the wavelength and UV irradiance within an appropriate range, it is possible to appropriately detect potential BPDs that may cause SSF expansion. The first ultraviolet light L1 may be of any type, as long as it has a wavelength and UV irradiance that can reach at least the interface I between the substrate layer 101 and the buffer layer 102. Examples of such a laser include a He—Cd laser, a YAG laser, and an InGaN laser. Among these, it is preferable to use a YAG laser.

[0038] 4A and 4B are explanatory diagrams of the first ultraviolet light L1 of the present invention, where (a) is an explanatory diagram of the illuminance distribution, and (b) is an explanatory diagram of pulse irradiation. In the first ultraviolet light L1, the outline of the light beam is shown by a broken line, and the outline of a portion of the light beam having an illuminance equal to or higher than a predetermined level is shown by a solid line. The portion having an illuminance equal to or higher than a predetermined level is defined as a portion where the UV illuminance of the first ultraviolet light L1 reaches at least the interface I between the substrate layer 101 and the buffer layer 102, and is 10 W / cm. ２ This is the region above. As shown in FIG. 4( a), the first ultraviolet light L1 is a light beam having a concentric illuminance distribution in which the illuminance at the center O is relatively high and the illuminance at the periphery is relatively low. Therefore, the peripheral region of the first ultraviolet light L1 (the region sandwiched between the solid and dashed lines) does not contribute to expanding the SSF from the latent BPD. Therefore, as shown in FIG. 4( b), the first ultraviolet light L1 is preferably pulse-irradiated while being shifted at regular intervals so that a circular region where the illuminance of the light beam is equal to or higher than a predetermined value is continuous. This allows the first ultraviolet light L1 to be irradiated to the entire silicon carbide substrate 100 with the illuminance required to expand the SSF while averaging the illuminance distribution, thereby enabling the SSF to be uniformly expanded from the latent BPD. The pulse irradiation of the first ultraviolet light L1 preferably has an irradiation time per pulse of approximately 10 to 15 nsec and a pulse irradiation interval of 10 msec or more. The moving speed of the first ultraviolet light L1 is preferably 20 to 50 cm / s.

[0039] The first ultraviolet light L1 is preferably irradiated using a galvanometer scanner. In this embodiment, as shown in FIG. 3 , the defect expansion irradiation means 2 includes a light source 21, a focusing unit 22, and a galvanometer scanner 23. The galvanometer scanner 23 has an X-axis galvanometer mirror 23x, a Y-axis galvanometer mirror 23y, and an actuator (not shown) that rotates the X-axis galvanometer mirror 23x and the Y-axis galvanometer mirror 23y. The light source 21 generates the first ultraviolet light L1. The focusing unit 22 focuses the first ultraviolet light L1 output from the light source 21 and adjusts it to a desired spot diameter. The galvanometer scanner 23 irradiates the first ultraviolet light L1 onto the silicon carbide substrate 100 while changing the irradiation direction of the first ultraviolet light L1 output from the focusing unit 22. In this way, the defect expansion irradiation means 2 irradiates the first ultraviolet light L1 using the galvano scanner 23, and therefore can highly accurately and uniformly irradiate the first ultraviolet light L1 onto the silicon carbide substrate 100. As a result, carriers generated in the silicon carbide substrate 100 by irradiation with the first ultraviolet light L1 are captured by latent BPDs, and expand the SSF from the conversion point where the BPDs are converted into TEDs.

[0040] <Visualization Irradiation Means> The visualization irradiation means 3 is a means for irradiating the entire silicon carbide substrate 100 with second ultraviolet light L2 having lower illuminance than the first ultraviolet light L1, thereby causing the silicon carbide substrate 100 to emit PL. The UV illuminance of the second ultraviolet light L2 is 1 W / cm ２ The following is preferable. The wavelength of the second ultraviolet light L2 is preferably 365 nm or less. The second ultraviolet light L2 excites the silicon carbide substrate 100 to emit PL light. By irradiating the entire silicon carbide substrate 100 with the second ultraviolet light L2 using the visualization irradiation means 3, PL light L3 is emitted from the expanded SSF, making it possible to detect latent BPDs. The second ultraviolet light L2 has a lower illuminance than the first ultraviolet light L1, and the light source can be, for example, an ultraviolet laser or a mercury-xenon lamp. Under the above irradiation conditions, the silicon carbide substrate 100 is excited to emit PL light L3 without further expanding the SSF from the conversion point where BPDs are converted to TEDs.

[0041] <Photographing Means> The photographing means 4 is a means for photographing the silicon carbide substrate 100 emitting PL light. The photographing means 4 is a digital camera having an image sensor in which solid-state imaging elements such as CCDs (Charged-Coupled Devices) and CMOSs ​​(Complementary Metal-Oxide-Semiconductors) are arranged in a two-dimensional array, and the image sensor detects light incident on the photographing means 4. The photographing means 4 is preferably configured to be movable so that any position on the silicon carbide substrate 100 can be photographed from the normal direction of the substrate surface. The photographing means 4 photographs a PL luminescence image including PL light L3 after irradiation with second ultraviolet light L2 from the visualization irradiation means 3. In the PL luminescence image, a bright linear defect image appears at the position where the SSF is present. The photographing means 4 records the photographed image in a storage device (not shown), such as a hard disk.

[0042] <Determination Means> The determination means 5 determines whether or not a potential BPD exists in an area where an SSF defect image occurs in the PL emission image captured by the imaging means 4. The determination means 5 is a computer having a CPU, memory, storage, etc., and its determination function is realized by the CPU reading and executing a program recorded in the memory. Specifically, the determination means 5 performs image analysis of the PL emission image to extract an SSF defect image, which is a bright linear defect image, and determines that a BPD exists in the area where this defect image occurs. The defect image extraction can be performed, for example, by edge detection processing, etc.

[0043] <Placement Table> The placement table 6 has a built-in heater and heats the silicon carbide substrate 100 placed on its upper surface. When the defect expansion irradiation means 2 irradiates the entire silicon carbide substrate 100 with the first ultraviolet light L1, the placement table 6 is preferably heated so that the temperature of the silicon carbide substrate 100 reaches 100 to 200°C, more preferably 100 to 150°C. Heating the silicon carbide substrate 100 to 100 to 200°C improves carrier generation efficiency, and irradiation with the first ultraviolet light L1 allows SSFs to be expanded in a short time, thereby shortening the inspection time. Furthermore, by setting the temperature during irradiation with the first ultraviolet light L1 to a temperature expected to be used in an automobile engine compartment, for example (100 to 200°C), potential BPDs that may cause forward current degradation in automotive power devices can be appropriately detected.

[0044] <Control means> The control means 7 is a computer having a CPU, memory, storage, etc., and the CPU reads and executes a program recorded in the memory, thereby realizing the function of controlling the operations of the defect expansion irradiation means 2, the visualization irradiation means 3, the imaging means 4, and the determination means 5.

[0045] [Defect Inspection Method] Figure 5 is a flowchart showing the steps of the defect inspection method and silicon carbide chip manufacturing method of the present invention. In the defect inspection method of the present invention, the irradiation step, visualization step, photography step, and judgment step are performed in order using the defect inspection device 1 of Figure 3, and a shrinkage step is also performed as an optional step. In addition, in the silicon carbide chip manufacturing method of the present invention, following each step of the defect inspection method, a mapping step, circuit formation step, cutting step, and exclusion step are performed in order. In the flowchart of Figure 5, each step of the defect inspection method and silicon carbide chip manufacturing method is indicated by the symbol "S".

[0046] <Irradiation Step: S1> In the irradiation step, the defect expansion irradiation means 2 irradiates the silicon carbide substrate 100 with first ultraviolet light L1 at a wavelength and UV irradiance that reaches at least the interface I between the substrate layer 101 and the buffer layer 102 (S1). In the irradiation step, the first ultraviolet light L1 is preferably irradiated while the silicon carbide substrate 100 is heated to 100 to 200° C., and more preferably while the silicon carbide substrate 100 is heated to 100 to 150° C. By irradiating the silicon carbide substrate 100 with the first ultraviolet light L1 while the silicon carbide substrate 100 is heated to 100 to 200° C., carrier generation efficiency is improved, SSFs can be expanded in a short time, and inspection time can be shortened. Furthermore, by setting the temperature during irradiation with the first ultraviolet light L1 to a temperature that is expected to be the operating temperature (100 to 200°C) in the engine compartment of an automobile, it is possible to appropriately detect potential BPDs in an automotive power device that may cause SSF expansion. The first ultraviolet light L1 in the irradiation step has a wavelength of 325 to 365 nm and a UV illuminance of 10 W / cm. ２ It is preferable that the wavelength is 355 nm and the UV irradiance is 100 W / cm or more. ２More preferably, the above conditions are satisfied. By irradiating the first ultraviolet light L1 under the above irradiation conditions, sufficient carriers can be generated in the substrate layer 101, allowing the SSF to be appropriately expanded from the latent BPDs. Furthermore, by setting the wavelength and UV irradiance within the above ranges, latent BPDs that may cause SSF expansion can be appropriately detected. Furthermore, in the irradiation step, the first ultraviolet light L1 is a light beam having a concentric irradiance distribution in which the irradiance is relatively high in the center O and relatively low in the peripheral portion, and it is preferable that pulse irradiation be performed while shifting the light beam at regular intervals so that circular regions where the irradiance of the light beam is equal to or higher than a predetermined value are continuous. This allows the entire silicon carbide substrate 100 to be irradiated with the first ultraviolet light L1 at an irradiance required for SSF expansion while averaging the irradiance distribution, thereby allowing the SSF to be uniformly expanded from the latent BPDs. Furthermore, in the irradiation step, the first ultraviolet light L1 is preferably irradiated using a galvanometer scanner 23. By using the galvano scanner 23 for irradiation, the first ultraviolet light L1 can be irradiated uniformly and with high precision over the entire silicon carbide substrate 100. In S2, the control means 7 determines whether there is an unirradiated region that has not been irradiated with the first ultraviolet light L1. If there is an unirradiated region (S2: Yes), the control means 7 controls the galvano scanner 23 and returns to S1 to repeatedly perform the irradiation step on the unirradiated region. If there is no unirradiated region (S2: No), the process proceeds to the visualization step.

[0047] <Visualization Step: S3> In the visualization step, the visualization irradiation means 3 irradiates the entire silicon carbide substrate 100 with second ultraviolet light L2, thereby causing the silicon carbide substrate 100 to emit PL (S3). The second ultraviolet light L2 in the visualization step has a wavelength of 365 nm or less and a UV illuminance of 1 W / cm ２ By irradiating the second ultraviolet light L2 under the above irradiation conditions, the silicon carbide substrate 100 is excited and can emit PL light L3 without further expanding the SSF from the conversion point where BPD is converted to TED.

[0048] <Photographing Step: S4> In the photographing step, the silicon carbide substrate 100 emitting PL light is photographed (S4). Generally, in a power device such as a SiC-MOSFET, the drift layer 103 is formed to a thickness of about 5 to 10 μm and the off angle is set to 4°. Therefore, the width of the band-shaped SSF in the drift layer 103 photographed from the normal direction of the substrate surface is about 70 to 140 μm. Therefore, the PL light emission image is photographed by photographing a band-shaped SSF having a width of about 70 to 140 μm at a size corresponding to a plurality of pixels. In the photographing step, the entire silicon carbide substrate 100 may be photographed at once, or the silicon carbide substrate 100 may be divided into a plurality of regions and photographed.

[0049] <Determination Step: S5> In the determination step, it is determined whether or not a potential BPD exists in an area where an SSF defect image occurs in the PL image (S5). The defect image can be extracted, for example, by edge detection processing. By analyzing the PL image, an SSF defect image, which is a bright linear defect image, is extracted, and it is determined that a BPD exists in the area where this defect image occurs.

[0050] <Shrinkage Step: S6> In the shrinkage step, the silicon carbide substrate 100 is heated to shrink the expanded SSFs in the silicon carbide substrate 100 (S6). The silicon carbide substrate 100 is heated by, for example, a heating furnace, separately from the heater built into the mounting table 6. The temperature of the silicon carbide substrate 100 in the shrinkage step is set to a temperature higher than the heating temperature of the silicon carbide substrate 100 during the irradiation step, for example, 300 to 1000°C. By heating at the above temperature, the expanded SSFs shrink and the defects are repaired, so that the silicon carbide substrate 100 after inspection can be directly commercialized. Therefore, the defect inspection method of the present invention enables 100% inspection in the manufacture of power devices.

[0051] [Method for Manufacturing Silicon Carbide Chip] In the method for manufacturing a silicon carbide chip, after the shrinking step in the defect inspection method is performed, the steps of mapping, circuit formation, cutting, and removal are performed in this order.

[0052] <Mapping Step: S7> In the mapping step, mapping data is created by mapping the regions determined to contain potential BPDs in the determination step onto the entire silicon carbide substrate 100 (S7). The mapping data is created, for example, by the control means 7 receiving position information of the potential BPDs from the determination means 5 and performing mapping. By creating the mapping data, the positions of the potential BPDs can be confirmed.

[0053] <Circuit Forming Step: S8> In the circuit forming step, a circuit is formed on silicon carbide substrate 100 (S8). In the circuit forming step, source, drain, and gate terminals are formed on silicon carbide substrate 100, for example.

[0054] <Cutting Step: S9> In the cutting step, silicon carbide substrate 100 is cut into a plurality of chips (S9). In the cutting step, silicon carbide substrate 100 can be cut using, for example, blade dicing using a metal blade, stealth dicing using a laser, or the like.

[0055] <Exclusion step: S10> In the exclusion step, defective chips including regions determined to contain potential BPDs are excluded based on the mapping data created in the mapping step (S10). By excluding defective chips, it is possible to manufacture highly stable silicon carbide chips that do not contain crystal defects that cause forward current degradation.

[0056] As described above, in the defect inspection method and defect inspection device of the present invention, the entire silicon carbide substrate 100 is irradiated with first ultraviolet light L1 at a wavelength and UV irradiance that reaches the interface I between the substrate layer 101 and the buffer layer 102. Therefore, SSF can be extended from the BPDs of the drift layer 103 and the BPDs of the buffer layer 102 that have been converted to TED, and it has become possible to detect latent BPDs that have been difficult to detect until now by PL inspection.

[0057] Next, an embodiment of the defect inspection apparatus of the present invention will be described below.

[0058] In the following tests, when ultraviolet light was irradiated onto the entire silicon carbide substrate by the defect expansion irradiation means, the silicon carbide substrate was heated to a temperature range of 100 to 200°C, simulating the temperature used in the engine compartment of an automobile, and the ease of BPD detection was compared (Examples 1 to 3). Similar tests were also conducted on a silicon carbide substrate that was not heated at 20°C (Comparative Example 1) and a silicon carbide substrate that was heated to a high temperature of 250°C (Comparative Example 2).

[0059] Example 1 A silicon carbide substrate was a thin film substrate (4-inch wafer) of silicon carbide single crystal manufactured by epitaxial growth. The silicon carbide substrate was heated to 100°C, and the entire silicon carbide substrate was irradiated with first ultraviolet light, and then irradiated with second ultraviolet light having a lower irradiance than the first ultraviolet light. The first ultraviolet light had a wavelength of 355 nm and a UV irradiance of 150 W / cm. ２ The second ultraviolet light had a wavelength of 313 nm and a UV irradiance of 0.3 W / cm. ２ An ultra-high pressure mercury lamp (1000 psi) was used as the light source. The PL-emitting silicon carbide substrate was photographed with a camera to obtain a PL image. The PL image was captured using a digital camera equipped with a cooled CCD image sensor with 1 million pixels, a 10x microscope objective lens, an exposure time of 5 seconds, and a 420 nm bandpass filter.

[0060] Examples 2 and 3, and Comparative Examples 1 and 2 A silicon carbide substrate similar to that of Example 1 was used, and PL images were obtained in the same manner as in Example 1, except that the temperature during irradiation with the first ultraviolet light was set as shown in FIG. 6 .

[0061] 6 shows micrographs (PL images) of silicon carbide substrates that were irradiated with first ultraviolet light under different temperature conditions and then allowed to emit PL light. As shown in FIG. 6, by irradiating the silicon carbide substrate with first ultraviolet light while it was heated to 100 to 200° C., it was possible to appropriately detect latent BPDs that may cause SSF expansion among the BPDs converted to TEDs at the interface between the substrate layer and the buffer layer (Examples 1 to 3). In contrast, in Comparative Example 1, in which the silicon carbide substrate was not heated, and Comparative Example 2, in which the silicon carbide substrate was heated to too high a temperature, SSF did not expand appropriately even when irradiated with the first ultraviolet light, and therefore latent BPDs could not be identified.

[0062] The defect inspection method, defect inspection device, and silicon carbide chip manufacturing method of the present invention can be used to inspect silicon carbide substrates in the manufacturing process of power device elements such as SiC-MOSFETs.

[0063] REFERENCE SIGNS LIST 1 Defect inspection device 2 Defect expansion irradiation means 3 Visualization irradiation means 4 Photography means 5 Determination means 100 Silicon carbide substrate 101 Substrate layer 102 Buffer layer 103 Drift layer L1 First ultraviolet light L2 Second ultraviolet light I Interface O Center

Claims

1. A defect inspection method for detecting basal plane dislocations (BPDs) that have been converted to threading edge dislocations (TEDs) at an interface between a substrate layer and a buffer layer in a silicon carbide substrate having a substrate layer, a buffer layer, and a drift layer, comprising: an irradiation step of irradiating the entire silicon carbide substrate with first ultraviolet light, which expands Shockley stacking faults (SSFs) from a conversion point at which the basal plane dislocations (BPDs) have been converted to threading edge dislocations (TEDs), at a wavelength and UV illuminance that reaches at least the interface between the substrate layer and the buffer layer; a visualization step of irradiating the entire silicon carbide substrate with second ultraviolet light having a lower illuminance than the first ultraviolet light, to visualize the silicon carbide substrate by causing it to emit photoluminescence; and an imaging step of imaging the silicon carbide substrate emitting photoluminescence. and a determination step of determining that a basal plane dislocation (BPD) that has been converted into a threading edge dislocation (TED) is present in an area in which a defect image of a Shockley stacking fault (SSF) occurs in the PL light emission image captured in the imaging step.

2. The defect inspection method according to claim 1, wherein in the irradiating step, the silicon carbide substrate is irradiated with the first ultraviolet light while being heated to 100 to 200°C.

3. The first ultraviolet light has a wavelength of 325 to 365 nm and a UV illuminance of 10 W / cm ２ The defect inspection method according to claim 1, 4. A defect inspection method as described in claim 1, wherein the first ultraviolet light is a light beam having a concentric illuminance distribution in which the illuminance is relatively high in the center and relatively low in the peripheral area, and the light beam is irradiated in pulses while being shifted at regular intervals so that there are continuous circular areas where the illuminance of the light beam is equal to or higher than a predetermined value.

5. A defect inspection method according to claim 4, wherein in the irradiation step, the first ultraviolet light is irradiated using a galvano scanner.

6. The defect inspection method according to any one of claims 1 to 5, further comprising a shrinkage step of shrinking Shockley stacking faults (SSFs) by heating a silicon carbide substrate in which Shockley stacking faults (SSFs) have expanded from a conversion point at which a basal plane dislocation (BPD) present at an interface between the substrate layer and the buffer layer is converted into a threading edge dislocation (TED).

7. A defect inspection device for detecting basal plane dislocations (BPDs) that have been converted into threading edge dislocations (TEDs) at an interface between a substrate layer and a buffer layer in a silicon carbide substrate having a substrate layer, a buffer layer, and a drift layer, comprising: a defect expansion irradiation means for irradiating the entire silicon carbide substrate with first ultraviolet light, which expands Shockley stacking faults (SSFs) from a conversion point at which the basal plane dislocations (BPDs) have been converted into threading edge dislocations (TEDs), at a wavelength and UV illuminance that reaches at least the interface between the substrate layer and the buffer layer; a visualization irradiation means for irradiating the entire silicon carbide substrate with second ultraviolet light having a lower illuminance than the first ultraviolet light, to cause the silicon carbide substrate to emit photoluminescence; and an imaging means for imaging the silicon carbide substrate emitting photoluminescence. and a determination means for determining that a basal plane dislocation (BPD) that has been converted into a threading edge dislocation (TED) is present in an area in which a defect image of a Shockley stacking fault (SSF) occurs in the PL light emission image captured by the imaging means.

8. A method for manufacturing silicon carbide chips, comprising the steps of: a mapping step of creating mapping data in which an area determined in the determination step to contain basal plane dislocations (BPDs) that have been converted into threading edge dislocations (TEDs) is mapped onto the entire silicon carbide substrate; a circuit formation step of forming a circuit on the silicon carbide substrate; a cutting step of cutting the silicon carbide substrate into a plurality of chips; and a removal step of removing defective chips including areas determined in the determination step to contain basal plane dislocations (BPDs) that have been converted into threading edge dislocations (TEDs) based on the mapping data.

Images