Stacked structure and method for manufacturing a nitride semiconductor substrate

Abstract

This nitride semiconductor substrate has a diameter of 2 inches or more and has a principal plane in which a crystal plane having the nearest low index is a (0001) plane, wherein the ratio of FWHM2{10-12} to FWHM1{10-12} is 80% or more.

Description

Technical Field

[0001] The present invention relates to a nitride semiconductor substrate, a stacked structure, and a method for manufacturing a nitride semiconductor substrate.

Background Art

[0002] A method is known in which a substrate made of a single crystal of a group III nitride semiconductor is used as a base substrate (seed substrate), and a crystal layer made of a single crystal of a group III nitride semiconductor is further grown on a main surface of the base substrate where the closest low-index crystal plane is the (0001) plane. According to this method, at least one nitride semiconductor substrate can be obtained by slicing a crystal layer grown to a predetermined thickness (for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] An object of the present invention is to improve the crystal quality of a nitride semiconductor substrate.

Means for Solving the Problems

[0005] According to one aspect of the present invention, a nitride semiconductor substrate having a diameter of 2 inches or more and a main surface where the closest low-index crystal plane is the (0001) plane, FWHM1 {10-12} the ratio of FWHM2 {10-12} is 80% or more is provided. However, FWHM1 {10-12} and FWHM2 {10-12}are the full width at half maximum of the diffraction of the {10-12} plane measured by X-ray rocking curve measurement, respectively, FWHM1 {10-12} Under the incident conditions for measuring {10-12} , from the X-ray source of Cu, through an X-ray mirror that makes the X-rays parallel light, a monochromator with two reflections of Ge(220), and an incident-side aperture with a width of 1.4 mm in the ω direction as the rotation angle direction around the rotation axis of the goniometer and a length of 12 mm in the direction parallel to the rotation axis in this order, the center of the main plane is irradiated with X-rays of Cu Kα1, FWHM1 {10-12} Under the light-receiving conditions for measuring {10-12} , with the light-receiving side slit open and without passing through the analyzer crystal, X-rays are received by a detector having an aperture with a width of 14.025 mm in the ω direction, FWHM2 {10-12} The incident conditions for measuring {10-12} are the same as the incident conditions for measuring FWHM1 {10-12} and FWHM2 {10-12} Under the light-receiving conditions for measuring {10-12} , X-rays are received by the detector through an analyzer crystal with three reflections of Ge(220) having an entrance aperture with a width of 6.54 mm in the ω direction.

[0006] According to another aspect of the present invention, a nitride semiconductor substrate having a main plane with a diameter of 2 inches or more and the closest low-index crystal plane being the (0001) plane, the maximum-minimum difference of FWHM1 measured by diffracting equivalent crystal planes represented by the {10-12} plane from three directions rotated by 60° each in the circumferential direction around the normal line of the center of the main plane {10-12} is 9 arcsec or less a nitride semiconductor substrate is provided.

[0007] According to still another aspect of the present invention, a nitride semiconductor substrate having a main plane with a diameter of 2 inches or more and the closest low-index crystal plane being the (0001) plane, FWHM1 {10-12} is 50 arcsec or less A nitride semiconductor substrate is provided.

[0008] According to still another aspect of the present invention, a base substrate made of a single crystal of a group III nitride semiconductor, having a polished main surface, and the low-index crystal plane closest to the main surface being the (0001) plane; a stacked unit provided above the main surface of the base substrate, including a low oxygen concentration region made of a single crystal of a group III nitride semiconductor and a high oxygen concentration region made of a single crystal of a group III nitride semiconductor provided on the low oxygen concentration region; a top low oxygen concentration region made of a single crystal of a group III nitride semiconductor provided above the stacked unit; and the oxygen concentration in the high oxygen concentration region is higher than the oxygen concentrations in the low oxygen concentration region and the top low oxygen concentration region, respectively; the stacked unit is provided in a plurality of repetitions in the thickness direction between the base substrate and the top low oxygen concentration region. A stacked structure is provided.

[0009] According to still another aspect of the present invention, a method for manufacturing a nitride semiconductor substrate using a vapor phase growth method, comprising: (a) preparing a base substrate made of a single crystal of a group III nitride semiconductor, having a polished main surface, and the low-index crystal plane closest to the main surface being the (0001) plane; (b) homoepitaxially growing a single crystal of a group III nitride semiconductor flat above the main surface; (c) forming a plurality of recesses composed of inclined interfaces other than the (0001) plane on the surface of the flat homoepitaxial growth crystal, gradually expanding the inclined interfaces as crystal growth progresses, and growing a three-dimensional growth layer characterized by disappearing the (0001) plane from the crystal growth interface at least once; (d) further epitaxially growing a single crystal of a group III nitride semiconductor on the three-dimensional growth layer to disappear the inclined interfaces and grow a planarization layer having a polished surface composed of the (0001) plane. having, performing a plurality of cycles including said (c) and said (d), A method for manufacturing a nitride semiconductor substrate is provided.

Advantages of the Invention

[0010] According to the present invention, the crystal quality of the nitride semiconductor substrate can be improved.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Figure 23

Figure 24

Embodiments for Carrying Out the Invention

[0012] <One Embodiment of the Present Invention> Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

[0013] (1) Method for Manufacturing Nitride Semiconductor Substrate Using FIGS. 1 to 9, the method for manufacturing a nitride semiconductor substrate according to this embodiment will be described. FIG. 1 is a flowchart showing the method for manufacturing a nitride semiconductor substrate according to this embodiment. FIGS. 2(a) to (g), FIGS. 3(a) to (c), FIGS. 5(a) to 6(b), and FIGS. 7 to 9 are schematic cross-sectional views showing a part of the method for manufacturing a nitride semiconductor substrate according to this embodiment. FIG. 4 is a schematic perspective view showing a part of the method for manufacturing a nitride semiconductor substrate according to this embodiment. Note that FIG. 4 corresponds to the perspective view at the time of FIG. 3(b) and shows a part of the three-dimensional growth layer 30 growing on the base substrate 10. Also, in FIG. 5(b), the thin solid line indicates the crystal plane during growth, and in FIGS. 3(c) to 6(b) and FIGS. 7 to 9, the dotted line indicates the dislocation.

[0014] The method for manufacturing a nitride semiconductor substrate according to this embodiment is a method for manufacturing a nitride semiconductor substrate using a vapor phase growth method, (a) preparing a base substrate made of a single crystal of a group III nitride semiconductor, having a polished main surface, and the low-index crystal plane closest to the main surface being the (0001) plane; (b) epitaxially growing a single crystal of a group III nitride semiconductor having a top surface with the (0001) plane exposed above the main surface of the base substrate, causing a plurality of recesses formed by inclined interfaces other than the (0001) plane to occur on the top surface, gradually expanding the inclined interfaces as going above the main surface of the base substrate, disappearing the (0001) plane from the top surface at least once, and growing a three-dimensional growth layer; (c) epitaxially growing a single crystal of a group III nitride semiconductor on the three-dimensional growth layer, disappearing the inclined interfaces, and growing a planarization layer having a polished surface; and performing a plurality of cycles including (b) and (c).

[0015] Specifically, as shown in FIG. 1, the method for manufacturing a nitride semiconductor substrate according to this embodiment has, for example, a base substrate preparation step S100, a three-dimensional growth step S200, a planarization step S300, an execution number determination step S400, a main growth step S500, a slicing step S600, and a polishing step S700.

[0016] (S100: Base Substrate Preparation Step) First, in the base substrate preparation step S100, a base substrate 10 made of a single crystal of a group III nitride semiconductor is prepared. In this embodiment, as the base substrate 10, for example, a gallium nitride (GaN) free-standing substrate is prepared.

[0017] In the following, in a crystal of a group III nitride semiconductor having a wurtzite structure, the <0001> axis (for example, the

[0001] axis) is referred to as the "c-axis", and the (0001) plane is referred to as the "c-plane". Note that the (0001) plane may be referred to as the "+c plane (group III element polar plane)", and the (000-1) plane may be referred to as the "-c plane (nitrogen (N) polar plane)". Further, the <1-100> axis (for example, the [1-100] axis) is referred to as the "m-axis", and the {1-100} plane is referred to as the "m-plane". Note that the m-axis may be denoted as the <10-10> axis. Further, the <11-20> axis (for example, the [11-20] axis) is referred to as the "a-axis", and the {11-20} plane is referred to as the "a-plane".

[0018] In the lower base plate preparation step S100 of the present embodiment, for example, the lower base plate 10 is manufactured by the VAS (Void-Assisted Separation) method.

[0019] Specifically, the lower base plate preparation step S100 includes, for example, a substrate preparation step S110 for crystal growth, a first crystal layer formation step S120, a metal layer formation step S130, a void formation step S140, a second crystal layer formation step S150, a peeling step S160, a slicing step S170, and a polishing step S180.

[0020] (S110: Substrate Preparation Step for Crystal Growth) First, as shown in FIG. 2(a), a substrate 1 for crystal growth (hereinafter, may be abbreviated as "substrate 1") is prepared. The substrate 1 is, for example, a sapphire substrate. Note that the substrate 1 may be, for example, a Si substrate or a gallium arsenide (GaAs) substrate. The substrate 1 has, for example, a main surface 1s that becomes a growth surface. The low-index crystal plane closest to the main surface 1s is, for example, the c-plane 1c.

[0021] In the present embodiment, the c-plane 1c of the substrate 1 is inclined with respect to the main surface 1s. The c-axis 1ca of the substrate 1 is inclined at a predetermined off-angle θ0 with respect to the normal of the main surface 1s. The off-angle θ0 within the main surface 1s of the substrate 1 is uniform over the entire main surface 1s. The off-angle θ0 within the main surface 1s of the substrate 1 affects the off-angle θ3 at the center of the main surface 10s of the lower base plate 10 described later.

[0022] (S120: First Crystalline Layer Formation Step) Next, as shown in Fig. 2(b), for example, by using the metalorganic vapor phase epitaxy (MOVPE) method, trimethylgallium (TMG) gas as a Group III source gas, ammonia gas (NH3) as a nitrogen source gas, and monosilane (SiH4) gas as an n-type dopant gas are supplied to a substrate 1 heated to a predetermined growth temperature, so that a first crystalline layer (underlying growth layer) 2, which is a low-temperature growth GaN buffer layer and a Si-doped GaN layer, is grown on the main surface 1s of the substrate 1 in this order. At this time, the thickness of the low-temperature growth GaN buffer layer and the thickness of the Si-doped GaN layer are, for example, 20 nm and 0.5 μm, respectively.

[0023] (S130: Metal Layer Formation Step) Next, as shown in Fig. 2(c), a metal layer 3 is deposited on the first crystalline layer 2. As the metal layer 3, for example, a titanium (Ti) layer is used. Also, the thickness of the metal layer 3 is, for example, 20 nm.

[0024] (S140: Void Formation Step) Next, the above-described substrate 1 is put into an electric furnace and placed on a susceptor having a predetermined heater. After the substrate 1 is placed on the susceptor, the substrate 1 is heated by the heater and heat treatment is performed in an atmosphere containing hydrogen gas or a hydride gas. Specifically, for example, heat treatment is performed for 20 minutes at a predetermined temperature in a hydrogen (H2) gas stream containing 20% NH3 gas. Note that the heat treatment temperature is, for example, 850°C or higher and 1,100°C or lower. By performing such heat treatment, the metal layer 3 is nitrided to form a metal nitride layer 5 having a high density of fine holes on the surface. Also, by performing the above-described heat treatment, a part of the first crystalline layer 2 is etched through the holes of the metal nitride layer 5 to form a high density of voids in the first crystalline layer 2.

[0025] Thereby, as shown in Fig. 2(d), a void-containing first crystalline layer 4 is formed.

[0026] (S150: Second Crystalline Layer Formation Step) Next, for example, by the hydride vapor phase epitaxy (HVPE) method, a gallium chloride (GaCl) gas, an NH3 gas, and a dichlorosilane (SiH2Cl2) gas as an n-type dopant gas are supplied to a substrate 1 heated to a predetermined growth temperature, so that an Si-doped GaN layer is epitaxially grown as a second crystal layer (full-growth layer) 6 on the void-containing first crystal layer 4 and the metal nitride layer 5. Note that, as the n-type dopant gas, instead of the SiH2Cl2 gas, a germanium tetrachloride (GeCl4) gas or the like may be supplied to epitaxially grow a Ge-doped GaN layer as the second crystal layer 6.

[0027] At this time, the second crystal layer 6 grows on the void-containing first crystal layer 4 and the metal nitride layer 5 through the holes of the metal nitride layer 5 from the void-containing first crystal layer 4. Some of the voids in the void-containing first crystal layer 4 are filled by the second crystal layer 6, but the other part of the voids in the void-containing first crystal layer 4 remains. A flat void is formed between the second crystal layer 6 and the metal nitride layer 5 due to the voids remaining in the void-containing first crystal layer 4. This void will cause the peeling of the second crystal layer 6 in the peeling step S160 described later.

[0028] Also, at this time, the second crystal layer 6 is grown while inheriting the orientation of the substrate 1. That is, the off-angle θ1 in the main plane of the second crystal layer 6 becomes uniform over the entire main plane, similar to the off-angle θ0 in the main plane 1s of the substrate 1.

[0029] Also, at this time, the thickness of the second crystal layer 6 is, for example, 600 μm or more, preferably 1 mm or more. Note that the upper limit value of the thickness of the second crystal layer is not particularly limited, but from the viewpoint of improving productivity, it is preferable that the thickness of the second crystal layer 6 is 50 mm or less.

[0030] (S160: Peeling step) After the growth of the second crystal layer 6 is completed, in the process of cooling the HVPE apparatus used to grow the second crystal layer 6, the second crystal layer 6 naturally peels off from the substrate 1 with the void-containing first crystal layer 4 and the metal nitride layer 5 as the boundary.

[0031] At this time, in the second crystal layer 6, tensile stress is introduced due to the attraction between the initial nuclei generated during its growth process. Therefore, due to the tensile stress generated in the second crystal layer 6, internal stress acts on the second crystal layer 6 such that its surface side is concave. Also, the dislocation density on the main surface (surface) side of the second crystal layer 6 is low, while the dislocation density on the back surface side of the second crystal layer 6 is high. Therefore, due to the difference in dislocation density in the thickness direction of the second crystal layer 6 as well, internal stress acts on the second crystal layer 6 such that its surface side is concave.

[0032] As a result, as shown in Fig. 2(f), after the second crystal layer 6 is peeled off from the substrate 1, it warps such that its surface side becomes concave. Therefore, the c-plane 6c of the second crystal layer 6 is curved in a concave spherical shape with respect to the plane perpendicular to the normal direction at the center of the main surface 6s of the second crystal layer 6. The off-angle θ2 formed by the c-axis 6ca with respect to the normal to the center of the main surface 6s of the second crystal layer 6 has a predetermined distribution.

[0033] (S170: Slicing process) Next, as shown in Fig. 2(f), for example, the second crystal layer 6 is sliced by a wire saw along a cutting plane SS that is substantially perpendicular to the normal direction at the center of the main surface 6s of the second crystal layer 6.

[0034] Thereby, as shown in Fig. 2(g), a base substrate 10 as an as-sliced substrate is formed. At this time, the thickness of the base substrate 10 is set to, for example, 450 μm. Note that the off-angle θ3 of the base substrate 10 may change from the off-angle θ2 of the second crystal layer 6 due to the slicing direction dependency.

[0035] (S180: Polishing process) Next, both surfaces of the base substrate 10 are polished by a polishing apparatus. Thereby, the main surface 10s of the base substrate 10 is mirror-finished.

[0036] Through the above base substrate preparation process S100, a base substrate 10 made of a single crystal of GaN is obtained.

[0037] The diameter of the lower base substrate 10 is, for example, 2 inches or more. Also, the thickness of the lower base substrate 10 is, for example, 300 μm or more and 1 mm or less.

[0038] The main surface 10s of the lower base substrate 10 has, for example, a main surface (substrate surface) 10s that becomes an epitaxial growth surface. In the present embodiment, the low-index crystal plane closest to the main surface 10s is, for example, the c-plane (+c-plane) 10c.

[0039] The c-plane 10c in the lower base substrate 10 is curved in a concave spherical shape with respect to the main surface 10s. Here, the "spherical shape" means a curved surface approximated by a sphere. Also, the "spherical approximation" here means being approximated within a predetermined error range with respect to a perfect spherical surface or an ellipsoidal spherical surface.

[0040] In the present embodiment, the c-plane 10f of the lower base substrate 10 is in a curved surface shape approximated by a sphere in each of the cross section along the m-axis and the cross section along the a-axis. The radius of curvature of the c-plane 10c in the lower base substrate 10 is, for example, 1 m or more and less than 10 m.

[0041] The off-angle θ3 formed by the c-axis 10ca with respect to the normal line of the center of the main surface 10s of the lower base substrate 10 has a predetermined distribution.

[0042] In this embodiment, the magnitude of the off-angle θ3 at the center of the main surface 10s of the underlying substrate 10 is, for example, 1° or less, preferably 0.4° or less. If the magnitude of the off-angle θ3 at the center of the main surface 10s exceeds 1°, three-dimensional growth of the three-dimensional growth layer 30 may be difficult to occur depending on the first growth conditions in the three-dimensional growth process S200 described later. For this reason, it becomes difficult to disappear the c-plane 30c. On the other hand, in this embodiment, by setting the magnitude of the off-angle θ3 at the center of the main surface 10s to 1° or less, the three-dimensional growth layer 30 can be easily three-dimensionally grown in the three-dimensional growth process S200 described later. Thereby, the c-plane 30c can be easily disappeared. Further, by setting the magnitude of the off-angle θ3 at the center of the main surface 10s to 0.4° or less, the three-dimensional growth layer 30 can be three-dimensionally grown under relatively wide growth conditions, and the c-plane 30c can be stably disappeared.

[0043] Note that, from the viewpoint of the three-dimensional growth of the three-dimensional growth layer 30, the smaller the magnitude of the off-angle θ3 at the center of the main surface 10s, the better. However, if the magnitude of the off-angle θ3 at the center of the main surface 10s is too close to 0°, the surface of the three-dimensional growth layer 30 may be excessively rough. For this reason, the magnitude of the off-angle θ3 at the center of the main surface 10s is preferably, for example, 0.1° or more.

[0044] Note that the magnitude and direction of the off-angle θ3 at the center of the main surface 10s of the underlying substrate 10 can be adjusted, for example, by the magnitude and direction of the off-angle θ0 of the crystal growth substrate 1 used in the above-described VAS method and the slice angle and slice direction in the slice process S170.

[0045] Also, in this embodiment, the root mean square roughness RMS of the main surface 10s of the underlying substrate 10 is, for example, less than 1 nm.

[0046] Also, in this embodiment, since the underlying substrate 10 is manufactured by the above-described VAS method, the dislocation density on the main surface 10s of the underlying substrate 10 is low. Specifically, the dislocation density on the main surface 10s of the underlying substrate 10 is, for example, 3×106 cm -2 1×10 or more 7 cm -2 is less than.

[0047] (S200: Three - dimensional growth process) When the lower base plate 10 is prepared as shown in Fig. 3(a), as shown in Fig. 3(b), Fig. 3(c), and Fig. 4, a single crystal of a group - III nitride semiconductor having a top surface 30u with the c - plane 30c exposed is epitaxially grown above the main surface 10s of the lower base plate 10. Thereby, the three - dimensional growth layer 30 is grown.

[0048] At this time, a plurality of recesses 30p surrounded by inclined interfaces 30i other than the c - plane are formed on the top surface 30u of the single crystal. As going above the main surface 10s of the lower base plate 10, the inclined interface 30i is gradually expanded, and the c - plane 30c is gradually reduced. Thereby, the c - plane 30c is made to disappear from the top surface 30u at least once. As a result, a three - dimensional growth layer 30 in which the inclined interface 30i widely exists on the surface is formed.

[0049] Also, at this time, in the three - dimensional growth process S200 of the first cycle, as shown in Fig. 3(a), the lower base plate 10 in a state where neither the formation of the mask layer on the main surface 10s nor the formation of the uneven pattern on the main surface 10s is subjected to any pattern processing is used. Here, the "mask layer" means, for example, a mask layer having a predetermined opening used in the so - called ELO (Epitaxial Lateral Overgrowth) method. Also, the "uneven pattern" here means, for example, at least one of trenches and ridges obtained by directly patterning the main surface of the lower base plate used in the so - called pendulum epitaxy method. The height difference of the uneven pattern here is, for example, 100 nm or more. In the three - dimensional growth process S200 of the first cycle, a single crystal of a group - III nitride semiconductor is directly epitaxially grown on the main surface 10s of the lower base plate 10 having the above - mentioned structure.

[0050] Also, at this time, in the three-dimensional growth step S200 of the first cycle, the three-dimensional growth layer 30 is three-dimensionally grown so as to deliberately roughen the main surface 10s of the underlying substrate 10. Note that, even if the three-dimensional growth layer 30 forms such a growth pattern, it is grown as a single crystal as described above. In this regard, the three-dimensional growth layer 30 is different from a so-called low-temperature growth buffer layer that is formed as amorphous or polycrystalline on a hetero-substrate such as sapphire before epitaxially growing a group III nitride semiconductor on the hetero-substrate.

[0051] Also, at this time, in the three-dimensional growth layer 30, an inclined interface growth region 70 (gray portion in the figure) is formed using an inclined interface 30i other than the c-plane as a growth surface. Further, as will be described later, the area occupied by the inclined interface growth region 70 in the cross-sectional plane along the main surface 10s of the underlying substrate 10 in the three-dimensional growth layer 30 is, for example, 80% or more.

[0052] In the present embodiment, as the three-dimensional growth layer 30, for example, a layer made of the same group III nitride semiconductor as the group III nitride semiconductor constituting the underlying substrate 10 is epitaxially grown. Specifically, for example, by the HVPE method, the underlying substrate 10 is heated, and a GaN layer is epitaxially grown as the three-dimensional growth layer 30 by supplying GaCl gas and NH3 gas to the heated underlying substrate 10.

[0053] Here, in the three-dimensional growth step S200, in order to cause the above-described growth process to occur, for example, the three-dimensional growth layer 30 is grown under a predetermined first growth condition.

[0054] First, with reference to FIG. 10(a), the reference growth conditions under which neither the inclined interface 30i nor the c-plane 30c expands or contracts will be described. FIG. 10(a) is a schematic cross-sectional view showing the growth process under the reference growth conditions under which neither the inclined interface nor the c-plane expands or contracts.

[0055] In Fig. 10(a), the thick solid line indicates the surface of the three-dimensional growth layer 30 per unit time. The inclined interface 30i shown in Fig. 10(a) is the most inclined interface with respect to the c-plane 30c. Also, in Fig. 10(a), let the growth rate of the c-plane 30c in the three-dimensional growth layer 30 be G c0 and let the growth rate of the inclined interface 30i in the three-dimensional growth layer 30 be G i and let the angle formed between the c-plane 30c and the inclined interface 30i in the three-dimensional growth layer 30 be α. Also, in Fig. 10(a), assume that the three-dimensional growth layer 30 grows while maintaining the angle α formed between the c-plane 30c and the inclined interface 30i. Note that the off-angle of the c-plane 30c of the three-dimensional growth layer 30 is negligible compared to the angle α formed between the c-plane 30c and the inclined interface 30i.

[0056] As shown in Fig. 10(a), when neither the inclined interface 30i nor the c-plane 30c expands or contracts, the locus of the intersection point between the inclined interface 30i and the c-plane 30c is perpendicular to the c-plane 30c. From this, the reference growth condition where neither the inclined interface 30i nor the c-plane 30c expands or contracts satisfies the following equation (a). G c0 = G i / cosα ···(a)

[0057] Next, with reference to Fig. 10(b), the first growth condition where the inclined interface 30i expands and the c-plane 30c contracts will be described. Fig. 10(b) is a schematic cross-sectional view showing the growth process under the first growth condition where the inclined interface expands and the c-plane contracts.

[0058] Also in Fig. 10(b), similar to Fig. 10(a), the thick solid line indicates the surface of the three-dimensional growth layer 30 per unit time. Also, the inclined interface 30i shown in Fig. 10(b) is also the most inclined interface with respect to the c-plane 30c. Also, in Fig. 10(b), let the growth rate of the c-plane 30c in the three-dimensional growth layer 30 be G c1 and let the progress rate of the locus of the intersection point between the inclined interface 30i and the c-plane 30c in the three-dimensional growth layer 30 be R1. Also, of the angles formed between the locus of the intersection point between the inclined interface 30i and the c-plane 30c and the c-plane 30c, let the smaller angle be αR1 Let it be so. When the angle formed between the R1 direction and the G i direction is α’, then α’ = α + 90 - α R1 holds. Note that the off-angle of the c-plane 30c of the three-dimensional growth layer 30 is assumed to be negligible compared to the angle α formed between the c-plane 30c and the inclined interface 30i.

[0059] As shown in Fig. 10(b), the progress rate R1 of the locus of the intersection point between the inclined interface 30i and the c-plane 30c is expressed by the following formula (b). R1 = G i / cosα’ ···(b)

[0060] Also, the growth rate G of the c-plane 30c in the three-dimensional growth layer 30 c1 is expressed by the following formula (c). G c1 = R1sinα R1 ···(c)

[0061] By substituting formula (b) into formula (c), G c1 is expressed by the following formula (d) using G i . G c1 = G i sinα R1 / cos(α + 90 - α R1 ) ···(d)

[0062] For the inclined interface 30i to expand and the c-plane 30c to shrink, it is preferable that α R1 < 90°. Therefore, the first growth condition for the inclined interface 30i to expand and the c-plane 30c to shrink preferably satisfies the following formula (1) according to formula (d) and α R1 < 90°. G c1 > G i / cosα ···(1) However, as described above, G i is the growth rate of the inclined interface 30i that is most inclined with respect to the c-plane 30c, and α is the angle formed between the inclined interface 30i that is most inclined with respect to the c-plane 30c and the c-plane 30c.

[0063] Or, G under the first growth condition c1 is preferably considered to be greater than G under the reference growth condition c0 From this, it can also be considered that by substituting the formula (a) into G c1 >G c0 the formula (1) can be derived.

[0064] Note that since the growth condition for expanding the inclined interface 30i most inclined with respect to the c-plane 30c is the most severe condition, if the first growth condition satisfies the formula (1), other inclined interfaces 30i can also be expanded.

[0065] Specifically, for example, when the inclined interface 30i most inclined with respect to the c-plane 30c is the {10-11} plane, α = 61.95°. Therefore, the first growth condition preferably satisfies, for example, the following formula (1'). G c1 >2.13G i ···(1’)

[0066] Or, as described later, for example, when the inclined interface 30i is the {11-2m} plane with m≧3, since the inclined interface 30i most inclined with respect to the c-plane 30c is the {11-23} plane, α = 47.3°. Therefore, the first growth condition preferably satisfies, for example, the following formula (1”). G c1 >1.47G i ···(1”)

[0067] As the first growth condition of the present embodiment, for example, the growth temperature in the three-dimensional growth step S200 is made lower than the growth temperature in the flattening step S300 described later. Specifically, the growth temperature in the three-dimensional growth step S200 is, for example, 980°C or higher and 1,020°C or lower, preferably 1,000°C or higher and 1,020°C or lower.

[0068] Further, as the first growth condition of the present embodiment, for example, the ratio of the partial pressure of NH3 gas as the nitrogen source gas to the partial pressure of GaCl gas as the group III source gas in the three-dimensional growth step S200 (hereinafter, also referred to as "V / III ratio") may be made larger than the V / III ratio in the planarization step S300 described later. Specifically, the V / III ratio in the three-dimensional growth step S200 is, for example, 2 or more and 20 or less, preferably 2 or more and 15 or less.

[0069] Actually, as the first growth condition, at least one of the growth temperature and the V / III ratio is adjusted within the above range so as to satisfy formula (1).

[0070] Note that other conditions among the first growth conditions of the present embodiment are as follows, for example. Growth pressure: 90 to 105 kPa, preferably 90 to 95 kPa Partial pressure of GaCl gas: 1.5 to 15 kPa Flow rate of N2 gas / Flow rate of H2 gas: 0 to 1

[0071] Here, the three-dimensional growth step S200 of the present embodiment is classified into two steps, for example, based on the shape during the growth of the three-dimensional growth layer 30. Specifically, the three-dimensional growth step S200 of the present embodiment includes, for example, an inclined interface expansion step S220 and an inclined interface maintenance step S240. Through these steps, the three-dimensional growth layer 30 has, for example, an inclined interface expansion layer 32 and an inclined interface maintenance layer 34.

[0072] (S220: Inclined Interface Expansion Step) First, as shown in FIGS. 3(b) and 4, the inclined interface expansion layer 32 of the three-dimensional growth layer 30 made of a single crystal of a group III nitride semiconductor is epitaxially grown directly on the main surface 10s of the lower substrate 10 under the above-described first growth conditions.

[0073] In the initial stage of growth of the inclined interface growth layer 32, the inclined interface growth layer 32 is step-flow grown (two-dimensional growth) with a predetermined thickness using the c-plane 30c as the growth surface in the normal direction of the main surface 10s of the underlying substrate 10 (the direction along the c-axis). That is, the single-crystalline inclined interface growth layer 32 is flatly homoepitaxially grown. Here, a part of the inclined interface growth layer 32 grown with the c-plane 30c as the growth surface is also referred to as the "initial layer". By this growth, an initial layer having a mirror-finished surface is formed with a predetermined thickness. At this time, the initial layer is grown continuously, for example, in the direction along the main surface 10s of the underlying substrate 10, that is, over the entire main surface 10s of the underlying substrate 10. Also, at this time, the thickness of the initial layer is, for example, 1 μm or more and 100 μm or less, preferably 1 μm or more and 20 μm or less.

[0074] Thereafter, by gradually growing the inclined interface growth layer 32 under the first growth conditions, as shown in FIGS. 3(b) and 4, a plurality of concave portions 30p composed of inclined interfaces 30i other than the c-plane are formed on the top surface 30u of the inclined interface growth layer 32 where the c-plane 30c is exposed. The plurality of concave portions 30p composed of inclined interfaces 30i other than the c-plane are randomly formed on the top surface 30u. Thereby, an inclined interface growth layer 32 in which the c-plane 30c and the inclined interfaces 30i other than the c-plane are mixed on the surface is formed.

[0075] Here, the "inclined interface 30i" means a growth interface inclined with respect to the c-plane 30c, and includes low-index facets other than the c-plane, high-index facets other than the c-plane, or inclined surfaces that cannot be represented by plane indices. The facets other than the c-plane are, for example, {11-2m}, {1-10n}, etc. However, m and n are non-zero integers.

[0076] In this embodiment, by using the above-described underlying substrate 10 and adjusting the first growth conditions so as to satisfy the formula (1), as the inclined interface 30i, for example, a {11-2m} plane with m≧3 can be generated. Thereby, the inclination angle of the {11-2m} plane with respect to the c-plane 30c can be made gentle. Specifically, the inclination angle can be made 47.3° or less.

[0077] By further growing the inclined interface expansion layer 32 under the first growth condition, as shown in FIGS. 3(b) and 3(c), in the inclined interface expansion layer 32, the inclined interfaces 30i other than the c-plane are gradually expanded and the c-plane 30c is gradually reduced as going upward above the main surface 10s of the lower base substrate 10. At this time, as going upward above the lower base substrate 10, the inclination angle formed by the inclined interface 30i with respect to the main surface 10s of the lower base substrate 10 is gradually reduced. As a result, finally, most of the inclined interfaces 30i become the {11-2m} planes where m≥3 as described above.

[0078] When the inclined interface expansion layer 32 is further grown, the c-plane 30c of the inclined interface expansion layer 32 disappears from the top surface 30u, and the outermost surface (uppermost surface) of the inclined interface expansion layer 32 is composed only of the inclined interfaces 30i.

[0079] In this way, by generating a plurality of recesses 30p composed of the inclined interfaces 30i other than the c-plane on the top surface 30u of the inclined interface expansion layer 32 and disappearing the c-plane 30c, as shown in FIG. 3(c), a plurality of valley portions 30v and a plurality of top portions 30t are formed on the surface of the inclined interface expansion layer 32. Each of the plurality of valley portions 30v is an inflection point convex downward on the surface of the inclined interface expansion layer 320 and is formed above the position where each of the inclined interfaces 30i other than the c-plane occurs. On the other hand, each of the plurality of top portions 30t is an inflection point convex upward on the surface of the inclined interface expansion layer 320 and is formed at or above the position where the c-plane 30c (finally) disappears across a pair of inclined interfaces 30i that expand in opposite directions. The valley portions 30v and the top portions 30t are alternately formed in the direction along the main surface 10s of the lower base substrate 10.

[0080] In this embodiment, at the initial stage of growth of the inclined interface expansion layer 32, on the main surface 10s of the lower base substrate 10, the inclined interface expansion layer 32 is grown to a predetermined thickness with the c-plane 30c as the growth surface without generating the inclined interface 30i, and then, on the surface of the inclined interface expansion layer 32, the inclined interface 30i other than the c-plane is generated. As a result, the plurality of valley portions 30v are formed at positions away from the main surface 10s of the lower base substrate 10 upward.

[0081] Due to the growth process of the inclined interface expansion layer 32 as described above, the dislocations bend and propagate as follows. Specifically, as shown in Fig. 3(c), a plurality of dislocations extending in the direction along the c-axis in the lower base substrate 10 propagate from the lower base substrate 10 in the direction along the c-axis of the inclined interface expansion layer 32. In the region of the inclined interface expansion layer 32 where the c-plane 30c is the growth surface, the dislocations propagate from the lower base substrate 10 in the direction along the c-axis of the inclined interface expansion layer 32. However, in the inclined interface expansion layer 32, when the growth interface where the dislocations are exposed changes from the c-plane 30c to the inclined interface 30i, the dislocations bend and propagate in a direction substantially perpendicular to the inclined interface 30i. That is, the dislocations bend and propagate in a direction inclined with respect to the c-axis. As a result, in the processes after the inclined interface expansion step S220, dislocations are locally collected above the approximate center between the pair of top portions 30t. As a result, the dislocation density on the surface of the planarization layer 40 described later can be reduced.

[0082] At this time, in this embodiment, when any cross-section perpendicular to the main surface 10s of the lower base substrate 10 is viewed, the average distance (also referred to as the "average distance between the closest top portions") L between the pair of top portions 30t that are the closest among the plurality of top portions 30t with one of the plurality of valley portions 30v in between is, for example, more than 100 μm along the direction of the main surface 10s of the lower base substrate 10. The average distance L between the closest top portions is the distance when the cross-section where the c-plane 30c disappears from the crystal growth interface is viewed.

[0083] When fine hexagonal pyramid-shaped crystal nuclei are generated on the main surface 10s of the lower base plate 10 from the initial stage of the inclined interface expansion process S220, etc., if the average distance L between the closest tops is 100 μm or less, in the processes after the inclined interface expansion process S220, the distance at which dislocations bend and propagate becomes short. For this reason, dislocations cannot be sufficiently collected above the approximate center between a pair of tops 30t in the inclined interface expansion layer 32. As a result, there is a possibility that the dislocation density on the surface of the flattening layer 40 described later is not sufficiently reduced. On the other hand, in the present embodiment, by making the average distance L between the closest tops exceed 100 μm, in the processes after the inclined interface expansion process S220, the distance at which dislocations bend and propagate can be ensured to be at least more than 50 μm. Thereby, dislocations can be sufficiently collected above the approximate center between a pair of tops 30t in the inclined interface expansion layer 32. As a result, the dislocation density on the surface of the flattening layer 40 described later can be sufficiently reduced.

[0084] On the other hand, in the present embodiment, the average distance L between the closest tops is made less than 800 μm. When the average distance L between the closest tops is 800 μm or more, the height from the main surface 10s of the lower base plate 10 to the valley portion 30v and the top portion 30t of the inclined interface expansion layer 32 becomes excessively high. For this reason, in the flattening process S300 described later, the thickness until the flattening layer 40 becomes mirror-finished becomes thick. On the other hand, in the present embodiment, by making the average distance L between the closest tops less than 800 μm, the height from the main surface 10s of the lower base plate 10 to the valley portion 30v and the top portion 30t of the inclined interface expansion layer 32 can be lowered. Thereby, the flattening layer 40 can be mirror-finished quickly.

[0085] Also, at this time, in the inclined interface expansion layer 320, based on the difference in the growth surface during the growth process, a c-plane growth region 60 that has grown with the c-plane 30c as the growth surface and an inclined interface growth region 70 (gray portion in the figure) that has grown with an inclined interface 30i other than the c-plane as the growth surface are formed.

[0086] Also, at this time, in the c-plane growth region 60, a valley portion 60a is formed at the position where the inclined interface 30i has occurred, and a peak portion 60b is formed at the position where the c-plane 30c has disappeared and terminated. Also, in the c-plane growth region 60, on both sides sandwiching the peak portion 60b, a pair of inclined portions 60i are formed as the locus of the intersection points of the c-plane 30c and the inclined interface 30i.

[0087] Also, at this time, by satisfying the first growth condition with the formula (1), the angle formed by the pair of inclined portions 60i when looking at a cross-section passing through the centers of two adjacent valley portions 60a is, for example, 70° or less.

[0088] Details of these regions will be described later.

[0089] (S240: Inclined Interface Maintenance Step) After disappearing the c-plane 30c from the surface of the inclined interface expansion layer 32, the growth conditions in the inclined interface maintenance step S240 are maintained with the above-described first growth condition in the same manner as in the inclined interface expansion step S220.

[0090] As a result, as shown in Fig. 5(a), while maintaining the state where the inclined interface 30i occupies more on the surface than the c-plane 30c, the growth of the three-dimensional growth layer 30 is continued over a predetermined thickness. As a result, an inclined interface maintenance layer 34 is formed on the inclined interface expansion layer 32.

[0091] Here, in the three-dimensional growth step S200, in order to surely bend the propagation direction of the dislocation and reduce the dislocation density as described above, it is important that the c-plane 30c has disappeared at least once when looking at the history of the growth interface at an arbitrary position of the three-dimensional growth layer 30. For this reason, it is desirable that the c-plane 30c disappears at least once at an early stage of the three-dimensional growth step S200 (for example, the above-described inclined interface expansion step S220).

[0092] However, in the inclined interface maintaining step S240, after the c-plane 30c has disappeared at least once, the c-plane 30c may reappear on a part of the surface of the inclined interface maintaining layer 34. However, it is preferable to mainly expose the inclined interface 30i on the surface of the inclined interface maintaining layer 34 so that the area ratio occupied by the inclined interface growth region 70 in the surface cross-section along the main surface 10s of the lower base substrate 10 is 80% or more. Note that the higher the area ratio occupied by the inclined interface growth region 70 in the surface cross-section, the better, and it is preferably 100%.

[0093] Also, at this time, under the first growth conditions, by growing the inclined interface maintaining layer 34 with the inclined interface 30i as the growth surface, as described above, at the position where the inclined interface 30i is exposed in the inclined interface expanding layer 32, the dislocations that bent and propagated in the direction inclined with respect to the c-axis continue to propagate in the same direction in the inclined interface maintaining layer 34.

[0094] Also, at this time, the inclined interface maintaining layer 34 grows with the inclined interface 30i as the growth surface, so that substantially the entire inclined interface maintaining layer 34 becomes a part of the inclined interface growth region 70.

[0095] Through the three-dimensional growth step S200 described above, a three-dimensional growth layer 30 having an inclined interface expanding layer 32 and an inclined interface maintaining layer 34 is formed.

[0096] In one cycle of the three-dimensional growth step S200, the height from the main surface 10s of the lower base substrate 10 to the top 30t of the three-dimensional growth layer 30 (the maximum height in the thickness direction of the three-dimensional growth layer 30) is, for example, more than 100 μm and less than 1.5 mm.

[0097] (S300: Planarization step) After growing the three-dimensional growth layer 30 in which the c-plane 30c has disappeared, as shown in FIG. 5(b), a single crystal of a group III nitride semiconductor is further epitaxially grown on the three-dimensional growth layer 30.

[0098] At this time, as going upward above the main surface 10s of the lower base plate 10, the inclined interface 40i is gradually reduced, and the c-plane 40c is gradually enlarged. Thereby, the inclined interface 30i formed on the surface of the three-dimensional growth layer 30 disappears. As a result, a planarization layer 40 having a mirror-finished surface is grown. Here, the "mirror" referred to herein means a surface where the maximum value of the height difference between adjacent unevenness on the surface is equal to or less than the wavelength of visible light.

[0099] In the present embodiment, as the planarization layer 40, for example, a layer mainly composed of the same group-III nitride semiconductor as the group-III nitride semiconductor constituting the three-dimensional growth layer 30 is epitaxially grown. In the planarization step S300, a silicon (Si)-doped GaN layer is epitaxially grown as the planarization layer 40 by supplying GaCl gas, NH3 gas, and dichlorosilane (SiH2Cl2) gas as an n-type dopant gas to the lower base plate 10 heated to a predetermined growth temperature. Note that, as the n-type dopant gas, GeCl4 gas or the like may be supplied instead of SiH2Cl2 gas.

[0100] Here, in the planarization step S300, in order to cause the above-described growth process to occur, for example, the planarization layer 40 is grown under a predetermined second growth condition.

[0101] With reference to FIG. 11, the second growth condition in which the inclined interface 40i is reduced and the c-plane 40c is enlarged will be described. FIG. 11 is a schematic cross-sectional view showing a growth process under the second growth condition in which the inclined interface is reduced and the c-plane is enlarged. FIG. 11 shows a process in which the planarization layer 40 grows on the three-dimensional growth layer 30 where the inclined interface 30i most inclined with respect to the c-plane 30c is exposed.

[0102] Also in FIG. 11, similar to FIG. 10(a), the thick solid line indicates the surface of the planarization layer 40 per unit time. Further, in FIG. 11, the growth rate of the c-plane 40c in the planarization layer 40 is G c2 and the growth rate of the inclined interface 40i in the planarization layer 40 is G iLet the progress rate of the locus of the intersection point of the inclined interface 40i and the c-plane 40c in the planarization layer 40 be R2. Also, among the angles formed by the locus of the intersection point of the inclined interface 40i and the c-plane 40c and the c-plane 30c, let the smaller angle be α R2 Let it be. When the angle formed by the R2 direction and the G i direction is α”, then α” = α - (90 - α R2 ). Also, in FIG. 11, assume that the planarization layer 40 grows while maintaining the angle α formed by the c-plane 30c and the inclined interface 30i in the three-dimensional growth layer 30. Note that the off-angle of the c-plane 40c of the planarization layer 40 is assumed to be negligible compared to the angle α formed by the c-plane 30c and the inclined interface 30i.

[0103] As shown in FIG. 11, the progress rate R2 of the locus of the intersection point of the inclined interface 40i and the c-plane 40c is represented by the following equation (e). R2 = G i / cosα” ···(e)

[0104] Also, the growth rate G c2 of the c-plane 40c in the planarization layer 40 is represented by the following equation (f). G c2 = R2sinα R2 ···(f)

[0105] By substituting equation (e) into equation (f), G c2 is represented by the following equation (g) using G i . G c2 = G i sinα R2 / cos(α + α R2 - 90) ···(g)

[0106] For the inclined interface 40i to shrink and the c-plane 40c to expand, it is preferable that α R2 < 90°. Therefore, the second growth condition for the inclined interface 40i to shrink and the c-plane 40c to expand preferably satisfies the following equation (2) according to equation (g) and α R2 < 90°. G c2 < G i / cosα ···(2) However, as described above, G i is the growth rate of the inclined interface 40i that is most inclined with respect to the c-plane 40c, and α is the angle formed between the inclined interface 40i that is most inclined with respect to the c-plane 40c and the c-plane 40c.

[0107] Alternatively, when the growth rate of the c-plane 30c in the flattening layer 40 under the reference growth conditions is G c0 , G c2 under the second growth conditions is preferably considered to be smaller than G c0 under the reference growth conditions. From this also, by substituting the formula (a) into G c2 <G c0 , the formula (2) can be derived.

[0108] Note that since the growth conditions for reducing the inclined interface 40i that is most inclined with respect to the c-plane 40c are the most severe conditions, if the second growth conditions satisfy the formula (2), other inclined interfaces 40i can also be reduced.

[0109] Specifically, when the inclined interface 40i that is most inclined with respect to the c-plane 40c is the {10-11} plane, the second growth conditions preferably satisfy the following formula (2'). G c2 <2.13G i ···(2’)

[0110] Alternatively, for example, when the inclined interface 30i is the {11-2m} plane with m≧3, since the inclined interface 30i that is most inclined with respect to the c-plane 30c is the {11-23} plane, the second growth conditions preferably satisfy, for example, the following formula (2”). G c2 <1.47G i ···(2”)

[0111] As the second growth condition of the present embodiment, the growth temperature in the planarization step S300 is set higher than the growth temperature in the three-dimensional growth step S200, for example. Specifically, the growth temperature in the planarization step S300 is, for example, 990°C or higher and 1,120°C or lower, preferably 1,020°C or higher and 1,100°C or lower.

[0112] Also, as the second growth condition of the present embodiment, the V / III ratio in the planarization step S300 may be adjusted. For example, the V / III ratio in the planarization step S300 may be made smaller than the V / III ratio in the three-dimensional growth step S200. Specifically, the V / III ratio in the planarization step S300 is, for example, 1 or more and 10 or less, preferably 1 or more and 5 or less.

[0113] Actually, as the second growth condition, at least one of the growth temperature and the V / III ratio is adjusted within the above ranges so as to satisfy formula (2).

[0114] Note that other conditions among the second growth conditions of the present embodiment are as follows, for example. Growth pressure: 90 to 105 kPa, preferably 90 to 95 kPa Partial pressure of GaCl gas: 1.5 to 15 kPa Flow rate of N2 gas / Flow rate of H2 gas: 1 to 20

[0115] By epitaxially growing a planarization layer 40 made of a single crystal of a group III nitride semiconductor on the three-dimensional growth layer 30 under the above-described second growth conditions, as shown in FIG. 5(b), as going upward from the three-dimensional growth layer 30, the c-plane 40c can be expanded while the inclined interfaces 40i other than the c-plane are reduced.

[0116] Specifically, due to growth under the second growth conditions, the planarization layer 40 grows from the inclined interface 30i of the inclined interface maintaining layer 34 in a direction along the direction perpendicular to the c-axis with the inclined interface 40i as the growth surface (i.e., the in-plane direction or the lateral direction). As the planarization layer 40 grows laterally, above the top 30t of the inclined interface maintaining layer 34, the c-plane 40c of the planarization layer 40 begins to be exposed again. Thereby, a planarization layer 40 is formed in which the c-plane 40c and the inclined interfaces 40i other than the c-plane are mixed on the surface.

[0117] When the planarization layer 40 is further grown laterally, the c-plane 40c gradually expands, and the inclined interfaces 40i of the planarization layer 40 gradually shrink. Thereby, the recesses 30p formed by a plurality of inclined interfaces 30i on the surface of the three-dimensional growth layer 30 are gradually filled.

[0118] After that, when the planarization layer 40 is further grown, the inclined interfaces 40i of the planarization layer 40 completely disappear, and the recesses 30p formed by a plurality of inclined interfaces 30i on the surface of the three-dimensional growth layer 30 are completely filled. Thereby, the surface of the planarization layer 40 becomes a mirror surface (flat surface) composed only of the c-plane 40c.

[0119] At this time, during the growth process of the three-dimensional growth layer 30 and the planarization layer 40, the dislocation density can be reduced by locally collecting dislocations. Specifically, the dislocations that bend and propagate in a direction inclined with respect to the c-axis in the three-dimensional growth layer 30 continue to propagate in the same direction in the planarization layer 40. As a result, in the planarization layer 40, dislocations are locally collected at the junction of adjacent inclined interfaces 40i above the approximate center between a pair of tops 30t. Among the plurality of dislocations collected at the junction of adjacent inclined interfaces 40i in the planarization layer 40, the dislocations having opposite Burgers vectors disappear at the time of meeting. In addition, a part of the plurality of dislocations collected at the junction of adjacent inclined interfaces 40i forms a loop, and the propagation in the direction along the c-axis (that is, the surface side of the planarization layer 40) is suppressed. Among the other parts of the plurality of dislocations collected at the junction of adjacent inclined interfaces 40i in the planarization layer 40, the propagation direction is changed again from the direction inclined with respect to the c-axis to the direction along the c-axis, and propagates to the surface side of the planarization layer 40. By thus disappearing a part of the plurality of dislocations or suppressing a part of the plurality of dislocations from propagating to the surface side of the planarization layer 40, the dislocation density on the surface of the planarization layer 40 can be reduced. Further, by locally collecting dislocations, a low dislocation density region can be formed above the portion of the planarization layer 40 where the dislocations propagate in a direction inclined with respect to the c-axis.

[0120] Also, at this time, in the planarization layer 40, as the c-plane 40c gradually expands, the c-plane growth region 60 grown with the c-plane 40c as the growth plane is formed while gradually expanding as it goes upward in the thickness direction again.

[0121] On the other hand, in the planarization layer 40, as the inclined interface 40i gradually shrinks, the inclined interface growth region 70 gradually shrinks as it goes upward in the thickness direction and terminates at a predetermined position in the thickness direction. Due to such a growth process of the planarization layer 40, in a cross-sectional view, a valley portion 70a of the inclined interface growth region 70 is formed at the position where the c-plane 40c reappears. Also, in the process of gradually filling the recess formed by the inclined interface 40i, in a cross-sectional view, a peak portion 70b of the inclined interface growth region 70 is formed at the position where the inclined interface 40i disappears.

[0122] In the planarization step S300, since the surface of the planarization layer 40 becomes a mirror surface composed only of the c-plane 40c, the height in the thickness direction (the maximum height in the thickness direction) of the planarization layer 40 is, for example, equal to or greater than the height from the valley portion 30v to the top portion 30t of the inclined interface maintaining layer 34.

[0123] By one cycle including the above three-dimensional growth step S200 and the planarization step S300, a stacked unit 50 having a three-dimensional growth layer 30 and a planarization layer 40 is formed. Hereinafter, a portion having a c-plane growth region 60 below the inclined interface growth region 70 and the inclined interface growth region 70, excluding the c-plane growth region 60 above the inclined interface growth region 70, may also be referred to as the stacked unit 50. Also, the stacked unit 50 formed by the first cycle may be referred to as the "first stacked unit 51".

[0124] (S400: Number of implementation determination step) After the first cycle is completed, the number of times of implementing the cycle including the three-dimensional growth step S200 and the planarization step S300 is determined.

[0125] When the number of implementations is less than a predetermined N times (No in S400), the cycle including the three-dimensional growth step S200 and the planarization step S300 is repeated as follows.

[0126] (S200: Three-dimensional growth step) In the cycles after the second cycle, as shown in Fig. 6(a), the three-dimensional growth layer 30 is grown again on the planarization layer 40 of the first stacked unit 51. At this time, the three-dimensional growth layer 30 is grown under the same first growth conditions as in the three-dimensional growth step S200 of the first cycle.

[0127] Also in the three-dimensional growth step S200 of the cycles after the second cycle, similar to the three-dimensional growth step S200 of the first cycle, for example, an inclined interface expansion step S220 and an inclined interface maintenance step S240 are performed.

[0128] (S220: Inclined Interface Expansion Step) As shown in Fig. 6(a), the inclined interface expansion layer 32 of the three-dimensional growth layer 30 made of a single crystal of a group III nitride semiconductor is epitaxially grown on the planarization layer 40 of the first stacked unit 51.

[0129] Also in the cycles after the second cycle, at the initial stage when the inclined interface expansion layer 32 grows, with the normal direction of the surface of the planarization layer 40 of the first stacked unit 51 (the direction along the c-axis), the inclined interface expansion layer 32 is grown with the c-plane 30c as the growth plane. That is, also at the initial stage of the cycle, with the c-plane 30c as the growth plane, the above-mentioned initial layer is grown over the entire planarization layer 40 of the first stacked unit 51.

[0130] Thereafter, by gradually growing the inclined interface expansion layer 32 under the first growth conditions, a plurality of concave portions 30p composed of inclined interfaces 30i other than the c-plane are generated on the top surface 30u where the c-plane 30c of the inclined interface expansion layer 32 is exposed, and the c-plane 30c is made to disappear at least once. Thereby, a plurality of valley portions 30v and a plurality of top portions 30t are formed on the surface of the inclined interface expansion layer 32.

[0131] In addition, also in the cycles after the second cycle, the plurality of concave portions 30p are randomly formed on the top surface 30u of the inclined interface expansion layer 32. Therefore, in a plan view, the positions of the valley portions 30v and the top portions 30t on the surface of the inclined interface expansion layer 32 in the cycles after the second cycle do not necessarily coincide with the positions of the valley portions 30v and the top portions 30t on the surface of the inclined interface expansion layer 32 in the first cycle.

[0132] Due to the growth process of the inclined interface expansion layer 32 as described above, the dislocations are bent and propagated again as follows. Specifically, as shown in Fig. 6(a), a plurality of dislocations remaining on the surface of the planarization layer 40 of the first stacked unit 51 propagate from the planarization layer 40 of the first stacked unit 51 in the direction along the c-axis of the inclined interface expansion layer 32. In the region of the inclined interface expansion layer 32 that grows with the c-plane 30c as the growth plane, the dislocations propagate from the planarization layer 40 of the first stacked unit 51 in the direction along the c-axis of the inclined interface expansion layer 32. However, in the inclined interface expansion layer 32, when the growth interface where the dislocations are exposed changes from the c-plane 30c to the inclined interface 30i, the dislocations bend and propagate in a direction substantially perpendicular to the inclined interface 30i. That is, the dislocations are bent and propagated again in a direction inclined with respect to the c-axis. As a result, also in the cycles after the second cycle, dislocations are locally collected above the approximate center between the pair of top portions 30t.

[0133] At this time, in the inclined interface expansion step S220 of the cycles after the second cycle, when any cross section perpendicular to the main surface 10s of the lower base plate 10 is viewed, the average distance L between the closest tops is made longer, for example, than that in the inclined interface expansion step S220 of the first cycle. Thereby, in the cycles after the second cycle, the distance by which the dislocations bend and propagate can be made longer than that in the first cycle. As a result, on the surface of the planarization layer 40 in the cycles after the second cycle, a dislocation-free density region with a non-overlapping 50-μm square can be easily formed, and the density of the dislocation-free region can be increased.

[0134] Also, at this time, in the inclined interface expansion step S220 in each of the plurality of cycles, the average distance L between the closest tops is gradually increased, for example, as the plurality of cycles are repeated. As a result, as the plurality of cycles are repeated, the distance over which dislocations bend and propagate can be gradually increased. Consequently, as the plurality of cycles are repeated, the dislocation-free region on the surface of the planarization layer 40 can be gradually widened, or the density of the dislocation-free region on the surface of the planarization layer 40 can be gradually increased.

[0135] Also, at this time, in the inclined interface expansion step S220 in each of the cycles after the first cycle as well, when looking at an arbitrary cross section perpendicular to the main surface 10s of the underlying substrate 10, the average distance L between the closest tops is set to be, for example, at least more than 100 μm, similar to the first cycle. As a result, similar to the first cycle, dislocations can be sufficiently collected above the approximate center between a pair of tops 30t in the three-dimensional growth layer 30.

[0136] Also, at this time, in the inclined interface expansion step S220 of at least the last cycle among the plurality of cycles, when looking at an arbitrary cross section perpendicular to the main surface 10s of the underlying substrate 10, it is preferable to set the average distance L between the closest tops to be, for example, more than 200 μm. As a result, in the steps after the inclined interface expansion step S220 of the last cycle, the distance over which dislocations bend and propagate can be ensured to be at least more than 100 μm. Consequently, a dislocation-free region with at least a 100-μm side can be formed on at least a part of the surface of the planarization layer 40 of the last cycle.

[0137] Note that at this time, in the inclined interface expansion step S220 of the cycles after the first cycle as well, it is preferable to set the average distance L between the closest tops to be, for example, less than 800 μm. As a result, in the planarization step S300 of the cycles after the first cycle as well, the planarization layer 40 can be mirror-finished quickly.

[0138] (S240: Inclined Interface Maintenance Step) Even in the cycles after the second cycle, after disappearing the c-plane 30c from the surface of the inclined interface expansion layer 32, the growth conditions in the inclined interface maintenance step S240 are maintained under the above-described first growth conditions, similar to the inclined interface expansion step S220.

[0139] As a result, as shown in Fig. 6(b), while maintaining the state where the inclined interface growth region 70 occupies an area of 80% or more of the cross-sectional area along the surface, the growth of the three-dimensional growth layer 30 is continued over a predetermined thickness. As a result, an inclined interface maintenance layer 34 is formed on the inclined interface expansion layer 32.

[0140] Through the three-dimensional growth step S200 of the cycles after the second cycle as described above, a three-dimensional growth layer 30 having an inclined interface expansion layer 32 and an inclined interface maintenance layer 34 is formed again.

[0141] (S300: Planarization step) Even in the cycles after the second cycle, when growing the three-dimensional growth layer 30 in which the c-plane 30c has disappeared at least once, as shown in Fig. 7, an epitaxial growth of a planarization layer 40 made of a single crystal of a group III nitride semiconductor is performed on the three-dimensional growth layer 30. At this time, the planarization layer 40 is grown under the second growth conditions similar to the planarization step S300 of the first cycle. As a result, as going upward from the three-dimensional growth layer 30, the c-plane 40c is expanded while the inclined interface 40i other than the c-plane is reduced.

[0142] After that, when further growing the planarization layer 40, the inclined interface 40i of the planarization layer 40 completely disappears, and the recess 30p formed by a plurality of inclined interfaces 30i on the surface of the three-dimensional growth layer 30 is completely filled. As a result, the surface of the planarization layer 40 becomes a mirror surface (flat surface) composed only of the c-plane 40c.

[0143] At this time, even in the cycles after the second cycle, by locally collecting dislocations in the growth processes of the three-dimensional growth layer 30 and the planarization layer 40, the dislocation density can be further reduced. That is, the dislocation density on the surface of the planarization layer 40 in the cycles after the second cycle can be reduced compared to the dislocation density on the surface of the planarization layer 40 in the first cycle.

[0144] In this way, by repeatedly performing a plurality of cycles including the three-dimensional growth step S200 and the planarization step S300, the dislocation density can be gradually reduced according to the number of times the cycle is repeated.

[0145] By the cycles after the second cycle including the three-dimensional growth step S200 and the planarization step S300 as described above, the laminated unit 50 having the three-dimensional growth layer 30 and the planarization layer 40 is formed again. Hereinafter, the laminated unit 50 formed by the second cycle may be referred to as the "second laminated unit 52".

[0146] (S400: Implementation times determination step) When the second cycle is completed, the number of times of performing the cycle including the three-dimensional growth step S200 and the planarization step S300 is determined again.

[0147] When the number of implementation times reaches a predetermined N times (Yes in S400), the cycle including the three-dimensional growth step S200 and the planarization step S300 is terminated.

[0148] The number of times of performing the cycle is, for example, 2 or more and 5 or less. By setting the number of times of performing the cycle to 2 or more, the effect of reducing dislocations by repeating the cycle can be sufficiently obtained. On the other hand, the effect of reducing dislocations appears dramatically up to 3 times, and after 3 times, it becomes difficult to obtain the effect of reducing dislocations as the number of times increases. For this reason, it is preferable that the number of times of performing the cycle is 5 or less.

[0149] In this embodiment, for example, N = 2 and the cycle is terminated.

[0150] Note that the c-plane growth region in which the c-plane 40c of the planarization layer 40 gradually expands and grows in the last cycle becomes a part of the uppermost c-plane growth region 80 described later.

[0151] (S500: Main growth step (c-plane growth step)) After the flattening process S300 of the last cycle, as shown in FIG. 8, on the flattened layer 40 with a mirror-finished surface, the main growth layer 44 is formed over a predetermined thickness with the c-plane as the growth plane.

[0152] At this time, the growth conditions in the main growth process S500 are maintained under the above-described second growth conditions in the same manner as in the flattening process S300 of each of the above cycles. Thereby, the main growth layer 44 can be grown in a step-flow manner with the c-plane as the growth plane.

[0153] Also, at this time, the radius of curvature of the c-plane of the main growth layer 44 can be made larger than the radius of curvature of the c-plane 10c of the underlying substrate 10. Thereby, the variation in the off-angle of the c-axis with respect to the normal of the surface in the main growth layer 44 can be made smaller than the variation in the off-angle of the c-axis 10ca with respect to the normal of the main surface 10s in the underlying substrate 10.

[0154] Also, at this time, by growing the main growth layer 44 with only the c-plane as the growth plane without exposing the inclined interface 40i, the entire main growth layer 44 becomes the uppermost c-plane growth region 80 described later.

[0155] In the main growth process S500, the thickness of the main growth layer 44 is, for example, 300 μm or more and 10 mm or less. By setting the thickness of the main growth layer 44 to 300 μm or more, at least one or more substrates 100 can be sliced from the main growth layer 44 in the slicing process S600 described later. On the other hand, by setting the thickness of the main growth layer 44 to 10 mm, when the final thickness is 650 μm and slicing a 700-μm-thick substrate 100 from the main growth layer 44, at least 10 substrates 100 can be obtained even considering a kerf loss of about 200 μm.

[0156] Through the processes from the three-dimensional growth process S200 to the main growth process S500 of the first cycle described above, the laminated structure 90 of the present embodiment is formed.

[0157] Note that the processes from the three-dimensional growth process S200 in the first cycle to the main growth process S500 are continuously performed in the same chamber without exposing the underlying substrate 10 to the atmosphere. Thereby, it is possible to suppress the formation of an unintended high oxygen concentration region (a region having an oxygen concentration higher than that of the inclined interface growth region 70) at the interface between the three-dimensional growth layer 30 and the planarization layer 40, the interface between the respective stacked units 50, the interface between the planarization layer 40 and the main growth layer 44, and the like.

[0158] (S600: Slicing process) Next, as shown in FIG. 9, for example, the main growth layer 44 is sliced by a wire saw along a cut surface substantially parallel to the surface of the main growth layer 44. Thereby, at least one nitride semiconductor substrate 100 (also referred to as substrate 100) as a sliced substrate is formed. At this time, the thickness of the substrate 100 is, for example, 300 μm or more and 700 μm or less.

[0159] At this time, the radius of curvature of the c-plane 100c of the substrate 100 can be made larger than the radius of curvature of the c-plane 10c of the underlying substrate 10. Note that at this time, the radius of curvature of the c-plane 100c of the substrate 100 can be made larger than the radius of curvature of the c-plane 40c of the main growth layer 44 before slicing. Thereby, the variation in the off-angle θ of the c-axis 100ca with respect to the normal of the main surface 100s of the substrate 100 can be made smaller than the variation in the off-angle of the c-axis 10ca of the underlying substrate 10.

[0160] (S700: Polishing process) Next, both surfaces of the substrate 100 are polished by a polishing apparatus. Note that at this time, the final thickness of the substrate 100 is, for example, 250 μm or more and 650 μm or less.

[0161] Through the above processes S100 to S700, the substrate 100 according to the present embodiment is manufactured.

[0162] (Process for manufacturing a semiconductor laminate and process for manufacturing a semiconductor device) Once the substrate 100 is manufactured, for example, a semiconductor functional layer made of a group III nitride semiconductor is epitaxially grown on the substrate 100 to fabricate a semiconductor laminate. After fabricating the semiconductor laminate, electrodes and the like are formed using the semiconductor laminate, and the semiconductor laminate is diced to cut out chips of a predetermined size. Thereby, a semiconductor device is fabricated.

[0163] (2) Laminated structure Next, with reference to FIG. 8, the laminated structure 90 according to the present embodiment will be described.

[0164] The laminated structure 90 of the present embodiment has, for example, a base substrate 10, a plurality of laminated units 50, and a top c-plane growth region (top low oxygen concentration region) 80.

[0165] The laminated units 50 are provided, for example, in a plurality and repeatedly in the thickness direction between the base substrate 10 and the top c-plane growth region 80. Each of the plurality of laminated units 50 has, for example, a c-plane growth region (low oxygen concentration region 60) and an inclined interface growth region (high oxygen concentration region) 70.

[0166] The c-plane growth region 60 is a region grown with the c-plane 30c as a growth surface. In the c-plane growth region 60, the incorporation of oxygen is suppressed as compared with the inclined interface growth region 70. Therefore, the oxygen concentration in the c-plane growth region 60 becomes lower than the oxygen concentration in the inclined interface growth region 70. Specifically, the oxygen concentration in the c-plane growth region 60 is, for example, 5×10 16 cm -3 Hereinafter, preferably 3×10 16 cm -3 or less.

[0167] The c-plane growth region 60 is provided above the main surface 10s of the base substrate 10. Since the c-plane 30c has disappeared at least once during the growth process of the three-dimensional growth layer 30, the c-plane growth region 60 is not continuous from the base substrate 10 side to the top c-plane growth region 80.

[0168] The c-plane growth region 60 has, for example, in a cross-sectional view, a plurality of valley portions 60a and a plurality of ridge portions 60b. Here, each of the valley portion 60a and the ridge portion 60b means a part of the shape observed based on the difference in emission intensity when the cross-section of the laminated structure 90 is observed with a fluorescence microscope or the like, and does not mean a part of the outermost surface shape generated during the growth of the three-dimensional growth layer 30. Each of the plurality of valley portions 60a is formed at a downwardly convex inflection point in the c-plane growth region 60 in a cross-sectional view, at the position where the inclined interface 30i is generated. At least one of the plurality of valley portions 60a is provided at a position away from the main surface 10s of the lower base substrate 10 upward. On the other hand, each of the plurality of ridge portions 60b is formed at an upwardly convex inflection point in the c-plane growth region 60 in a cross-sectional view, at the position where the c-plane 30c (finally) disappears and terminates with a pair of inclined interfaces 30i that expand in opposite directions to each other. The valley portions 60a and the ridge portions 60b are alternately formed in the direction along the main surface 10s of the lower base substrate 10.

[0169] When looking at an arbitrary cross-section perpendicular to the main surface 10s of the lower base substrate 10, the average distance between the pair of ridge portions 60b that are closest to each other among the plurality of ridge portions 60b with one of the plurality of valley portions 60a in between is separated in the direction along the main surface 10s of the lower base substrate 10, which corresponds to the average distance L between the closest tops of the above-described three-dimensional growth layer 30, and is, for example, more than 100 μm. Hereinafter, this distance is also referred to as the "average distance between the closest ridge portions".

[0170] The c-plane growth region 60 has a pair of inclined portions 60i provided as a locus of the intersection of the c-plane 30c and the inclined interface 30i on both sides sandwiching one of the plurality of ridge portions 60b. Here, the inclined portion 60i means a part of the shape observed based on the difference in emission intensity when the cross-section of the laminated structure 90 is observed with a fluorescence microscope or the like, and does not mean the outermost inclined interface 30i generated during the growth of the three-dimensional growth layer 30.

[0171] When looking at a cross-section passing through the centers of two adjacent valley portions 60a, the angle formed by a pair of inclined portions 60i is, for example, 70° or less, preferably 20° or more and 65° or less. The fact that the angle formed by the pair of inclined portions 60i is 70° or less means that, under the first growth condition, the growth rate G i of the inclined interface 30i that is most inclined with respect to the c-plane 30c among the three-dimensional growth layers 30, relative to the growth rate G c1 of the c-plane 30c among the three-dimensional growth layers 30. The ratio G c1 / G i was high. As a result, it is possible to easily generate inclined interfaces 30i other than the c-plane. Consequently, at the position where the inclined interface 30i is exposed, it becomes possible to easily bend dislocations. Also, by setting the angle formed by the pair of inclined portions 60i to 70° or less, it is possible to easily generate a plurality of valley portions 30v and a plurality of top portions 30t above the main surface 10s of the underlying substrate 10. Furthermore, by setting the angle formed by the pair of inclined portions 60i to 65° or less, it is possible to more easily generate inclined interfaces 30i other than the c-plane, and it is possible to more easily generate a plurality of valley portions 30v and a plurality of top portions 30t above the main surface 10s of the underlying substrate 10. Note that by setting the angle formed by the pair of inclined portions 60i to 20° or more, it is possible to suppress an increase in the height from the valley portion 30v to the top portion 30t of the three-dimensional growth layer 30, and it is possible to suppress an increase in the thickness until the planarization layer 40 becomes mirror-like.

[0172] On the other hand, the inclined interface growth region 70 is a region that has grown with the inclined interface 30i or the inclined interface 40i as the growth surface. In the inclined interface growth region 70, oxygen is more easily incorporated compared to the c-plane growth region 60 and the uppermost c-plane growth region 80 described later. For this reason, the oxygen concentration in the inclined interface growth region 70 becomes higher than the respective oxygen concentrations of the c-plane growth region 60 and the uppermost c-plane growth region 80. Note that the oxygen incorporated into the inclined interface growth region 70 is, for example, oxygen that has unintentionally entered the vapor phase growth apparatus, or oxygen released from members (such as quartz members) constituting the vapor phase growth apparatus. Specifically, the oxygen concentration in the inclined interface growth region 70 is, for example, 9×10 17 cm -3 or more and 5×10 19cm -3 is as follows.

[0173] The inclined interface growth region 70 is provided, for example, on the c-plane growth region 60 in the same-layer stacking unit 50. The inclined interface growth region 70 is continuously provided along the main surface 10s of the underlying substrate 10. That is, when a plurality of cross-sectional views are taken by cutting the three-dimensional growth layer 30 along the main surface 10s of the underlying substrate 10, it is desirable that a cross-section that does not include the c-plane growth region grown with the c-plane 30c as the growth surface exists in at least a part of the thickness direction of the three-dimensional growth layer 30.

[0174] The lower surface of the inclined interface growth region 70 is formed, for example, following the shape of the c-plane growth region 60 in the same-layer stacking unit 50.

[0175] The upper surface of the inclined interface growth region 70 has, for example, a plurality of valleys 70a and a plurality of ridges 70b in a cross-sectional view. Here, each of the valleys 70a and the ridges 70b means a part of the shape observed based on the difference in emission intensity when the cross-section of the laminated structure 90 is observed with a fluorescence microscope or the like, and does not mean a part of the outermost surface shape generated during the growth of the planarization layer 40. As described above, the plurality of valleys 70a of the inclined interface growth region 70 are formed at positions where the c-plane 40c is regenerated in a cross-sectional view. Also, the plurality of valleys 70a of the inclined interface growth region 70 are each formed above the plurality of ridges 60b of the c-plane growth region 60 in a cross-sectional view. On the other hand, as described above, the plurality of ridges 70b of the inclined interface growth region 70 are formed at positions where the inclined interface 40i disappears and terminates in a cross-sectional view. Also, the plurality of ridges 70b of the inclined interface growth region 70 are each formed above the plurality of valleys 60a of the c-plane growth region 60 in a cross-sectional view.

[0176] Also, the boundary surface along the main surface 10s of the underlying substrate 10 at the upper end of the inclined interface growth region 70 is the boundary surface between the stacking units 50, and is the boundary surface at the position where the inclined interface 40i disappears and terminates. Hereinafter, the boundary surface along the main surface 10s of the underlying substrate 10 at the upper end of the inclined interface growth region 70 of a predetermined stacking unit 50 is also simply referred to as the "boundary surface of the stacking unit 50".

[0177] In the c-plane growth region 60 in the second and higher stacked units 50, for example, it is provided above the inclined interface growth region 70 in the lower stacked unit 50.

[0178] The uppermost c-plane growth region 80 is provided, for example, above the inclined interface growth region 70 in the uppermost stacked unit 50.

[0179] The uppermost c-plane growth region 80 is a region grown with the c-plane as the growth surface. In the uppermost c-plane growth region 80, oxygen uptake is suppressed as compared with the inclined interface growth region 70. For this reason, the oxygen concentration in the uppermost c-plane growth region 80 becomes lower than the oxygen concentration in the inclined interface growth region 70. The oxygen concentration in the uppermost c-plane growth region 80 is, for example, 5×10 16 cm -3 or less, preferably 3×10 16 cm -3 or less as in the c-plane growth region 60.

[0180] Here, in the present embodiment, when looking at an arbitrary cross section perpendicular to the main surface 10s of the lower base plate 10, the average distance between the closest ridges in the second and higher stacked units 50 among the plurality of stacked units 50 is, for example, longer than that in the first stacked unit 50.

[0181] Further, in the present embodiment, when looking at an arbitrary cross section perpendicular to the main surface 10s of the lower base plate 10, the average distance between the closest ridges in each of the plurality of stacked units 50 gradually becomes longer as going to the upper layer of the plurality of stacked units 50, for example.

[0182] In addition, in the present embodiment, during the growth process of the three-dimensional growth layer 30 in each cycle, at the position where the inclined interface 30i other than the c-plane is exposed, dislocations bend and propagate in a direction substantially perpendicular to the inclined interface 30i. As a result, in the planarization layer 40, part of the plurality of dislocations disappears, or part of the plurality of dislocations is suppressed from propagating to the surface side of the c-plane expansion layer 42. Thereby, the dislocation density at the interface of each stacked unit 50 is reduced compared to the dislocation density on the main surface 10s of the underlying substrate 10.

[0183] In addition, in the present embodiment, also in the cycles after the second cycle, by locally collecting dislocations during the growth processes of the three-dimensional growth layer 30 and the planarization layer 40, the dislocation density at the interface of the stacked units 50 of the second layer and above is lower than the dislocation density at the interface of the stacked unit 50 of the first layer.

[0184] In addition, in the present embodiment, by repeatedly performing a plurality of cycles including the three-dimensional growth step S200 and the planarization step S300, the dislocation density is gradually reduced according to the number of cycles repeated. Specifically, the dislocation density at each interface of the plurality of stacked units 50 is gradually reduced as going to the upper layer of the plurality of stacked units 50.

[0185] As a result of these, in the present embodiment, the dislocation density is rapidly reduced in the thickness direction.

[0186] Here, let the dislocation density on the main surface 10s of the underlying substrate 10 be N0, and the dislocation density (average dislocation density) at the interface of the uppermost stacked unit 50 be N. On the other hand, when a crystal layer of a group III nitride semiconductor is epitaxially grown with a thickness equal to the thickness from the main surface 10s of the underlying substrate 10 of the present embodiment to the interface of the uppermost stacked unit 50 with only the c-plane as the growth surface on the main surface 10s of the underlying substrate 10 (hereinafter, also referred to as "the case of c-plane thick film growth (c-plane limited growth)"), let the dislocation density on the surface of the crystal layer be N'.

[0187] In the case of c-plane thick film growth, the dislocation density on the surface of the crystal layer tended to be inversely proportional to the thickness of the crystal layer. Specifically, in the case of c-plane thick film growth, when the thickness of the crystal layer was 2 mm, the reduction rate of the dislocation density determined by N’ / N0 was approximately 0.5.

[0188] In contrast, in the present embodiment, the reduction rate of the dislocation density determined by N / N0 is smaller than, for example, the reduction rate of the dislocation density determined by N’ / N0 in the case of c-plane thick film growth.

[0189] Specifically, in the present embodiment, the thickness per layer of the stacking unit 50 is, for example, more than 200 μm and 1.5 mm or less, preferably 1.2 mm or less. When the number of stacked layers of the stacking unit 50 is two, the thickness from the main surface 10s of the lower base substrate 10 to the interface of the second stacking unit 52 is, for example, 3 mm or less, preferably 2.4 mm or less. In this case, the reduction rate of the dislocation density determined by the above-mentioned N / N0 is, for example, 0.15 or less, preferably 0.10 or less, more preferably 0.08 or less.

[0190] In the present embodiment, the lower limit value of the reduction rate of the dislocation density is not limited because the smaller it is, the better. However, when the lower base substrate 10 having an average dislocation density of 3×10 6 cm -2 is used, it is difficult to make the average dislocation density at the interface of the second stacking unit 52 10 3 cm -2 or less. Therefore, the reduction rate of the dislocation density is, for example, 3×10 -4 or more.

[0191] In addition, in the present embodiment, the entire surface of the stacked structure 90 is configured to be aligned with the +c plane, and each stacked unit 50 and the uppermost c-plane growth region 80 do not include an inversion domain (inversion domain), respectively. In this regard, the stacked structure 90 of the present embodiment is different from a stacked structure formed by the so-called DEEP (Dislocation Elimination by the Epitaxial-growth with inverse-pyramidal Pits) method, that is, it is different from a stacked structure in which an inversion domain is included in a core located at the center of a pit.

[0192] (3) Nitride semiconductor substrate (nitride semiconductor self-supporting substrate, nitride crystal substrate) Next, with reference to FIG. 12, the nitride semiconductor substrate 100 according to the present embodiment will be described. FIG. 12(a) is a schematic top view showing the nitride semiconductor substrate according to the present embodiment, (b) is a schematic cross-sectional view along the m-axis of the nitride semiconductor substrate according to the present embodiment, and (c) is a schematic cross-sectional view along the a-axis of the nitride semiconductor substrate according to the present embodiment. Here, the direction along the m-axis is defined as the x-direction, and the direction along the a-axis is defined as the y-direction.

[0193] In the present embodiment, the substrate 100 obtained by slicing the present growth layer 44 by the above-described manufacturing method is, for example, a self-supporting substrate made of a single crystal of a group III nitride semiconductor. In the present embodiment, the substrate 100 is, for example, a GaN self-supporting substrate.

[0194] The diameter of the substrate 100 is, for example, 2 inches or more. Further, the thickness of the substrate 100 is, for example, 300 μm or more and 1 mm or less.

[0195] The conductivity of the substrate 100 is not particularly limited. However, when manufacturing a semiconductor device as a vertical Schottky barrier diode (SBD) using the substrate 100, the substrate 100 is, for example, n-type, the n-type impurity in the substrate 100 is, for example, Si or germanium (Ge), and the n-type impurity concentration in the substrate 100 is, for example, 1.0×10 18 cm -3 or more and 1.0×1020 cm -3 is as follows.

[0196] The substrate 100 has a main surface 100s that serves as an epitaxial growth surface, for example. In the present embodiment, the low-index crystal surface closest to the main surface 100s is, for example, the c-plane 100c.

[0197] Note that the main surface 100s of the substrate 100 is, for example, mirror-polished, and the root mean square (RMS) roughness of the main surface 100s of the substrate 100 is, for example, less than 1 nm.

[0198] Also, in the present embodiment, the impurity concentration in the substrate 100 obtained by the above-described manufacturing method is lower than that of a substrate obtained by a method such as the flux method or the ammonothermal method.

[0199] Specifically, the hydrogen concentration in the substrate 100 is, for example, 1×10 17 cm -3 less than, preferably 5×10 16 cm -3 or less.

[0200] Also, in the present embodiment, since the substrate 100 is formed by slicing the main growth layer 44 grown with the c-plane 40c as the growth surface, it does not include the inclined interface growth region 70 grown with the inclined interface 30i or the inclined interface 40i as the growth surface. That is, the entire substrate 100 is composed of a low oxygen concentration region.

[0201] Specifically, the oxygen concentration in the substrate 100 is, for example, 5×10 16 cm -3 or less, preferably 3×10 16 cm -3 or less.

[0202] Also, in the present embodiment, the substrate 100 does not include a polarity inversion region (inversion domain), for example, as described above.

[0203] (Curvature of the c-plane and variation in off-angle) As shown in FIGS. 12(b) and (c), in the present embodiment, the c-plane 100c, which is the low-index crystal plane closest to the main surface 100s of the substrate 100, is curved in a concave spherical shape with respect to the main surface 100s, for example, due to the manufacturing method of the substrate 100 described above.

[0204] In the present embodiment, the c-plane 100c of the substrate 100 has a curved surface shape approximated by a spherical surface in each of the cross section along the m-axis and the cross section along the a-axis, for example.

[0205] In the present embodiment, since the c-plane 50f of the substrate 100 is curved in a concave spherical shape as described above, at least a part of the c-axis 100ca is inclined with respect to the normal of the main surface 100s. The off-angle θ, which is the angle formed by the c-axis 100ca with respect to the normal of the main surface 100s, has a predetermined distribution within the main surface 100s.

[0206] Among the off-angles θ of the c-axis 100ca with respect to the normal of the main surface 100s, the direction component along the m-axis is denoted as "θ m ", and the direction component along the a-axis is denoted as "θ a ". Note that θ 2 =θ m 2 +θ a 2 holds.

[0207] In the present embodiment, since the c-plane 100c of the substrate 100 is curved in a concave spherical shape as described above, the off-angle m-axis component θ m and the off-angle a-axis component θ a can be approximately expressed by a linear function of x and a linear function of y, respectively.

[0208] Specifically, for example, X-ray rocking curve measurements of the (0002) plane are performed at each position on a straight line passing through the center within the main plane 100s. When the peak value of the angle ω formed between the X-ray incident on the main plane 100s and the main plane 100s (hereinafter also referred to as the peak angle ω) is plotted against the position on the straight line, the peak angle ω can be approximated by a linear function of the position. Here, the "peak angle ω" refers to the angle formed between the X-ray incident on the main plane 100s and the main plane 100s, and is the angle at which the diffraction intensity is maximum. The radius of curvature of the c-plane 100c can be obtained by the reciprocal of the slope of the linear function approximated as described above.

[0209] In this embodiment, the radius of curvature of the c-plane 100c of the substrate 100 is, for example, larger than the radius of curvature of the c-plane 10c of the base substrate 10 used in the above-described manufacturing method of the substrate 100.

[0210] Specifically, the radius of curvature of the c-plane 100c of the substrate 100 is, for example, 23 m or more, preferably 30 m or more, and more preferably 40 m or more.

[0211] For reference, even in the case of c-plane thick film growth, the radius of curvature of the c-plane in a substrate sliced from a crystal layer having the same thickness as the thickness from the main plane 10s of the base substrate 10 to the interface of the uppermost stacked unit 50 in this embodiment may be larger than the radius of curvature of the c-plane 10c of the base substrate 10. However, in the case of c-plane thick film growth, when the thickness of the crystal layer is 2 mm, the radius of curvature of the c-plane in the substrate sliced from the crystal layer is about 11 m, which is about 1.4 times the radius of curvature of the c-plane 10c of the base substrate 10.

[0212] In this embodiment, the upper limit value of the radius of curvature of the c-plane 100c of the substrate 100 is not particularly limited because the larger it is, the better. When the c-plane 100c of the substrate 100 is substantially flat, it may be considered that the radius of curvature of the c-plane 100c is infinite.

[0213] In addition, in the present embodiment, since the radius of curvature of the c-plane 100c of the substrate 100 is large, the variation in the off-angle θ of the c-axis 100ca with respect to the normal of the main surface 100s of the substrate 100 can be made smaller than the variation in the c-axis 10ca off-angle of the underlying substrate 10.

[0214] Specifically, when performing X-ray rocking curve measurement on the (0002) plane of the substrate 100 and measuring the off-angle θ of the c-axis 100ca with respect to the normal of the main surface 100s based on the diffraction peak angle of the (0002) plane, the variation determined by the maximum-minimum difference in the magnitude of the off-angle θ within a diameter of 2 inches (50.8 mm) from the center of the main surface 100s is, for example, 0.127° or less, preferably 0.097° or less, and more preferably 0.073° or less.

[0215] For reference, in the underlying substrate 10 fabricated by the above-described VAS method, the variation in the off-angle of the c-axis 10ca determined by the above-described measurement method is approximately 0.38°. Also, in the case of c-plane thick film growth, when the thickness of the crystal layer is the same as the thickness from the main surface 10s of the underlying substrate 10 of the present embodiment to the interface of the uppermost stacked unit 50 (for example, 2 mm), in the nitride semiconductor substrate obtained from the crystal layer, the variation in the off-angle of the c-axis determined by the above-described measurement method is approximately 0.26°.

[0216] In the present embodiment, since the lower limit value of the variation in the off-angle θ of the c-axis 100ca of the substrate 100 is preferably as small as possible, it is not particularly limited. When the c-plane 100c of the substrate 100 is substantially flat, it may be considered that the variation in the off-angle θ of the c-axis 100ca of the substrate 100 is 0°.

[0217] In addition, in the present embodiment, since the curvature of the c-plane 100c becomes isotropically smaller with respect to the main surface 100s of the substrate 100, the direction dependence of the radius of curvature of the c-plane 100c is small.

[0218] Specifically, the difference between the absolute value of the radius of curvature of the c-plane 100c in the direction along the a-axis and the absolute value of the radius of curvature of the c-plane 100c in the direction along the m-axis obtained by the above-described measurement method is, for example, 90% or less, preferably 50% or less, more preferably 20% or less of the larger of these values.

[0219] Also, in the present embodiment, when the peak angle ω is approximated by a linear function of the position within the main plane 100s in the X-ray rocking curve measurement of the c-plane 100c, the error of ω with respect to the linear function of the position is small. The error of ω in the present embodiment can be made smaller than, for example, a substrate obtained from a crystal layer grown on a base substrate patterned by an ELO method using a mask layer or the like, or a substrate obtained from a planarization layer when the c-plane does not disappear in the three-dimensional growth process with one cycle.

[0220] Specifically, the error of the measured peak angle ω with respect to the linear function approximated as described above is, for example, 0.05° or less, preferably 0.02° or less, more preferably 0.01° or less. Note that since at least some of the peak angles ω may coincide with the linear function, the minimum value of the error is 0°.

[0221] (Dark spot) Next, the dark spots on the main plane 100s of the substrate 100 of the present embodiment will be described. Here, the "dark spot" means a point with low emission intensity observed in the observation image of the main plane 100s in a multiphoton excitation microscope or the cathodoluminescence image of the main plane 100s, and includes not only dislocations but also non-emission centers caused by foreign substances or point defects. Note that the "multiphoton excitation microscope" is sometimes called a two-photon excitation fluorescence microscope.

[0222] In this embodiment, since the substrate 100 is manufactured using the bottom substrate 10 made of high-purity GaN single crystal produced by the VAS method, there are few non-luminescent centers caused by foreign matters or point defects in the substrate 100. Therefore, when the main surface of the substrate 100 is observed by a multiphoton excitation microscope or the like, 95% or more, preferably 99% or more, more preferably 100% of the dark spots are not non-luminescent centers caused by foreign matters or point defects, but dislocations.

[0223] Further, in this embodiment, by the above-described manufacturing method, the dislocation density on the surface of the main growth layer 44 is reduced compared to the dislocation density on the main surface 10s of the bottom substrate 10. Thereby, the dislocations are also reduced on the main surface 100s of the substrate 100 formed by slicing the main growth layer 44.

[0224] Moreover, in this embodiment, by performing the three-dimensional growth process S200 or the like using the bottom substrate 10 in a state where no pattern processing is performed by the above-described manufacturing method, on the main surface 100s of the substrate 100 formed by slicing the main growth layer 44, no region with an excessively high dislocation density due to the pattern processing of the bottom substrate 10 is formed, and regions with a low dislocation density are uniformly formed.

[0225] Specifically, in this embodiment, when the dislocation density is obtained from the dark spot density by observing the main surface 100s of the substrate 100 at a field of view of 250 μm square with a multiphoton excitation microscope, there is no region where the dislocation density exceeds 3×10 6 cm -2 , and regions where the dislocation density is less than 1×10 6 cm -2 exist in 80% or more, preferably 90% or more, more preferably 95% or more of the main surface 100s.

[0226] Note that when the manufacturing method of this embodiment is used, the upper limit value of the ratio of the region where the dislocation density is less than 1×10 6 cm -2 is 100%.

[0227] Furthermore, in the present embodiment, by repeatedly performing a plurality of cycles including the three-dimensional growth step S200 and the planarization step S300, the dislocation density on the main surface 100s of the substrate 100 is reduced according to the number of cycles repeated. That is, the dislocation density on the main surface 100s of the substrate 100 in the present embodiment is lower than the dislocation density on the main surface of the substrate formed by slicing the present growth layer when only one cycle is performed.

[0228] Specifically, in the present embodiment, when observing the main surface 100s of the substrate 100 at a field of view of 250 μm square by a multiphoton excitation microscope and obtaining the dislocation density from the dark spot density, there is no region where the dislocation density is 1×10 6 cm -2 or more, and regions where the dislocation density is less than 7×10 5 cm -2 exist in 80% or more, preferably 90% or more, more preferably 95% or more of the main surface 100s.

[0229] Note that when the manufacturing method of the present embodiment is used, the upper limit value of the ratio of the region where the dislocation density is less than 7×10 5 cm -2 is 100%.

[0230] Also, in the present embodiment, the average dislocation density of the entire main surface 100s of the substrate 100 is, for example, less than 7×10 5 cm -2 , preferably less than 5.5×10 5 cm -2 , and more preferably less than 3×10 5 cm -2 or less.

[0231] Also, the main surface 100s of the substrate 100 in the present embodiment includes, for example, a dislocation-free region of at least 50 μm square based on the average distance L between the nearest tops in the three-dimensional growth step S200 described above.

[0232] In addition, in the present embodiment, since the dislocation density is reduced over the entire main surface 100s of the substrate 100, the dislocation-free regions with a size of 50 μm square are scattered, for example, over the entire main surface 100s of the substrate 100.

[0233] In addition, in the present embodiment, by performing a plurality of cycles and making the average distance L between the closest tops in each cycle longer than that in the first cycle, the density of the dislocation-free regions is increased. Specifically, the main surface 100s of the substrate 100 has, for example, 100 dislocation-free regions with a size of 50 μm square per cm 2 or more, preferably 1000 per cm 2 or more, more preferably 1600 per cm 2 or more, still more preferably more than 4800 per cm 2 exceeding, most preferably more than 10000 per cm 2 at the above density.

[0234] Note that the upper limit value of the density of the non-overlapping dislocation-free regions with a size of 50 μm square is 40000 per cm 2 based on the measurement method.

[0235] In addition, in the present embodiment, in the inclined interface expansion step S220 of at least the last cycle among the plurality of cycles, by making the average distance L between the closest tops exceed 200 μm, at least a part of the main surface 100s of the substrate 100 has, for example, a dislocation-free region with a size of at least 100 μm square. The main surface 100s of the substrate 100 has, for example, 100 non-overlapping dislocation-free regions with a size of 100 μm square per cm 2 or more, preferably 250 per cm 2 at the above density.

[0236] Note that the upper limit value of the density of the non-overlapping dislocation-free regions with a size of 100 μm square is 10000 per cm 2 based on the measurement method.

[0237] In addition, in the present embodiment, even in a region other than the dislocation-free region (for example, between a pair of adjacent dislocation-free regions), dislocations do not excessively concentrate, and the dislocation density is low. Specifically, the number of dislocations present in the quadrangular region sandwiched between the opposing sides of the closest pair of 50-μm square dislocation-free regions is, for example, 330 or less, preferably 90 or less, more preferably 70 or less, still more preferably 33 or less, and most preferably 18 or less.

[0238] For reference, in a substrate obtained by a conventional manufacturing method that does not perform a special process for collecting dislocations, the size of the dislocation-free region is smaller than 50 μm square, or the density of 50-μm square dislocation-free regions is lower than 100 pieces / cm 2 As the density of 50-μm square dislocation-free regions is low, the number of dislocations present between a pair of dislocation-free regions also increases. Furthermore, there is a high possibility that 100-μm square dislocation-free regions are not formed.

[0239] Next, the Burgers vector of dislocations in the substrate 100 of the present embodiment will be described.

[0240] In the present embodiment, since the dislocation density on the main surface 10s of the base substrate 10 used in the above-described manufacturing method is low, when the three-dimensional growth layer 30 and the planarization layer 40 are grown on the base substrate 10, it is less likely that a plurality of dislocations combine (mix). As a result, in the substrate 100 obtained from the planarization layer 40, generation of dislocations having a large Burgers vector can be suppressed.

[0241] Specifically, in the substrate 100 of the present embodiment, for example, there are many dislocations with a Burgers vector of any one of <11-20> / 3, <0001>, or <11-23> / 3. Here, the "Burgers vector" can be measured, for example, by the large-angle convergent electron diffraction method (LACBED method) using a transmission electron microscope (TEM). A dislocation with a Burgers vector of <11-20> / 3 is an edge dislocation, a dislocation with a Burgers vector of <0001> is a screw dislocation, and a dislocation with a Burgers vector of <11-23> / 3 is a mixed dislocation in which an edge dislocation and a screw dislocation are mixed.

[0242] In the present embodiment, when 100 dislocations are randomly extracted from the main surface 100s of the substrate 100, the ratio of the number of dislocations with a Burgers vector of any one of <11-20> / 3, <0001>, or <11-23> / 3 is, for example, 50% or more, preferably 70% or more, more preferably 90% or more. Note that, in at least a part of the main surface 100s of the substrate 100, there may be dislocations with a Burgers vector of 2<11-20> / 3 or <11-20>, etc.

[0243] (Regarding the X-ray rocking curve measurement of c-plane diffraction with different widths of the incident-side slit in the ω direction) Here, the inventor has found that by performing the X-ray rocking curve measurement of c-plane diffraction while varying the width of the incident-side slit in the ω direction, it is possible to simultaneously evaluate both the crystal quality factor constituting the substrate 100 of the present embodiment and the curvature (warpage) of the above-described c-plane 100c.

[0244] First, the influence of the crystal quality factor in the X-ray rocking curve measurement will be described.

[0245] The full width at half maximum of the diffraction pattern in X-ray rocking curve measurement is greatly affected by crystal quality factors such as the level of dislocation density, the level of mosaicity, the magnitude of stacking fault density, the magnitude of basal plane dislocation density, the magnitude of point defect (such as vacancy) density, the magnitude of the in-plane fluctuation amount of lattice constant, and the distribution of impurity concentration. When these crystal quality factors are not good, the fluctuation of the diffraction angle in X-ray rocking curve measurement becomes large, and the full width at half maximum of the diffraction pattern becomes large.

[0246] Next, with reference to Fig. 13(a), the influence of the curvature of the c-plane 100c in X-ray rocking curve measurement will be described. Fig. 13(a) is a schematic cross-sectional view showing the diffraction of X-rays with respect to the curved c-plane.

[0247] Let the width of the incident-side slit in the ω direction be dI, the irradiation width (footprint width) of the X-rays irradiated on the main surface of the substrate be b, and the Bragg angle of the crystal be θ B When, the irradiation width b of the X-rays on the main surface of the substrate is obtained by the following formula (h). b = dI / sinθ B ···(h) Note that "ω" and "ω direction" respectively refer to the "rotation angle" and "rotation angle direction" around the rotation axis of the goniometer in X-ray rocking curve measurement. In the measurement of the c-plane diffraction, "ω" and "ω direction" can also be considered as the substrate rotation angle and the substrate rotation direction when the substrate 100 placed on the substrate stage is rotated with the axis passing through the center of the substrate stage and parallel to the mounting surface of the substrate stage as the central axis.

[0248] As shown in Fig. 13(a), when the c-plane of the substrate is curved, when the radius of curvature of the c-plane is R and half of the central angle formed by the curved c-plane in the range of the irradiation width b of the X-rays is γ, the radius of curvature R of the c-plane is very large with respect to the irradiation width b of the X-rays. From this, the angle γ is obtained by the following formula (i). γ = sin -1 (b / 2R) ≒ b / 2R ···(i)

[0249] At this time, at the incident-side end (the right-side end in the figure) of the region on the c-plane of the substrate irradiated with X-rays, the diffraction angle with respect to the main surface of the substrate is θ B +γ = θ B +b / 2R.

[0250] On the other hand, at the light-receiving-side end (the left-side end in the figure) of the region on the c-plane of the substrate irradiated with X-rays, the diffraction angle with respect to the main surface of the substrate is θ B -γ = θ B -b / 2R.

[0251] Therefore, due to the difference between the diffraction angle with respect to the main surface of the substrate at the incident-side end of the c-plane of the substrate and the diffraction angle with respect to the main surface of the substrate at the light-receiving-side end of the c-plane of the substrate, the fluctuation of the diffraction angle of the X-rays with respect to the curved c-plane is b / R.

[0252] Figures 13(b) and (c) are diagrams showing the fluctuation of the diffraction angle of the (0002) plane with respect to the radius of curvature of the c-plane. Note that the vertical axis of Figure 13(b) is on a logarithmic scale, and the vertical axis of Figure 13(c) is on a linear scale.

[0253] As shown in Figures 13(b) and (c), when the width dI in the ω direction of the incident-side slit of the X-rays is increased, that is, when the irradiation width b of the X-rays is increased, the fluctuation of the diffraction angle of the (0002) plane increases according to the irradiation width b of the X-rays. Also, as the radius of curvature R of the c-plane becomes smaller, the fluctuation of the diffraction angle of the (0002) plane gradually increases. Further, the difference in the fluctuation of the diffraction angle of the (0002) plane when the irradiation width b of the X-rays is varied becomes larger as the radius of curvature R of the c-plane becomes smaller.

[0254] For example, when the width dI in the ω direction of the incident-side slit is narrow, in the fluctuation of the diffraction angle of the (0002) plane, the influence due to the curvature of the c-plane is small, and the influence due to the above-described crystal quality factors is dominant. On the other hand, for example, when the width dI in the ω direction of the incident-side slit is wide, in the fluctuation of the diffraction angle of the (0002) plane, both the influence due to the crystal quality factors and the influence due to the curvature of the c-plane are superimposed. Therefore, if the X-ray rocking curve measurement is performed while varying the width dI in the ω direction of the incident-side slit, it becomes possible to simultaneously evaluate both the above-described crystal quality factors and the curvature (warpage) of the c-plane over the region irradiated with X-rays.

[0255] Here, the characteristics when performing the X-ray rocking curve measurement of the c-plane diffraction in the substrate 100 of the present embodiment will be described.

[0256] Hereinafter, when irradiating the main surface 100s of the substrate 100 with X-rays of Cu Kα1 through a two-crystal monochromator of the Ge(220) plane and an incident-side slit and performing the X-ray rocking curve measurement of the (0002) plane diffraction, the full width at half maximum of the (0002) plane diffraction when the width in the ω direction of the incident-side slit is 1 mm is defined as "FWHMa", and the full width at half maximum of the (0002) plane diffraction when the width in the ω direction of the incident-side slit is 0.1 mm is defined as "FWHMb".

[0257] In the substrate 100 of the present embodiment, all of the above-described crystal quality factors such as the high or low dislocation density, the high or low mosaicity, the large or small stacking fault density, the large or small basal plane dislocation density, the large or small point defect (such as a vacancy) density, the large or small in-plane fluctuation amount of the lattice constant, and the distribution of the impurity concentration are good.

[0258] As a result, in the substrate 100 of the present embodiment, when performing the X-ray rocking curve measurement of the (0002) plane diffraction with the width in the ω direction of the incident-side slit being 0.1 mm, the full width at half maximum FWHMb of the (0002) plane diffraction is, for example, 80 arcsec or less, preferably 50 arcsec or less, more preferably 32 arcsec or less.

[0259] Further, in the substrate 100 of the present embodiment, as described above, all of the above-described crystal quality elements are good over a wide range of the main surface 100s.

[0260] As a result, when X-ray rocking curve measurement of the (0002) plane diffraction is performed at a plurality of measurement points set at 5 mm intervals (between the center and the outer edge) within the main surface 100s of the substrate 100 of the present embodiment, with the width in the ω direction of the incident-side slit being 0.1 mm, for example, in 90% or more, preferably 95% or more, more preferably 100% of all the measurement points, the full width at half maximum FWHMb of the (0002) plane diffraction is 80 arcsec or less, preferably 50 arcsec, more preferably 32 arcsec or less.

[0261] Further, in the substrate 100 of the present embodiment, the in-plane variation of the above-described crystal quality elements is small. For this reason, when X-ray rocking curve measurement is performed with the width in the ω direction of the incident-side slit widened, the diffraction pattern of the (0002) plane tends to be less likely to become narrower than the diffraction pattern of the (0002) plane when X-ray rocking curve measurement is performed with the width in the ω direction of the incident-side slit narrowed.

[0262] As a result, in the substrate 100 of the present embodiment, the full width at half maximum FWHMa of the (0002) plane diffraction when the width in the ω direction of the incident-side slit is 1 mm can be, for example, equal to or greater than the full width at half maximum FWHMb of the (0002) plane diffraction when the width in the ω direction of the incident-side slit is 0.1 mm.

[0263] Note that even when the crystal quality elements of the substrate 100 are good, there may be a case where FWHMa < FWHMb in a state where FWHMb is very small.

[0264] In addition, in the substrate 100 of the present embodiment, as described above, not only are there few dislocations over a wide range of the main surface 100s, but all of the above-described crystal quality elements are well-balanced and good. Furthermore, the curvature of the c-plane 100c of the substrate 100 is small, and the radius of curvature of the c-plane 100c is large. Due to these factors, in the substrate 100 of the present embodiment, even if the width in the ω direction of the incident-side slit is widened and the X-ray rocking curve measurement of (0002) plane diffraction is performed, over the region irradiated with X-rays, the above-described crystal quality elements are well-balanced and good, and because the radius of curvature of the c-plane is large, the fluctuation of the diffraction angle of the (0002) plane does not increase so much. Therefore, even if the X-ray rocking curve measurement is performed with different widths in the ω direction of the incident-side slit, the difference in the fluctuation of the diffraction angle of the (0002) plane becomes small.

[0265] As a result, at a predetermined measurement point (for example, the center of the main surface) of the substrate 100 of the present embodiment, the difference FWHMa - FWHMb obtained by subtracting the full width at half maximum FWHMb of (0002) plane diffraction when the width in the ω direction of the incident-side slit is 0.1 mm from the full width at half maximum FWHMa of (0002) plane diffraction when the width in the ω direction of the incident-side slit is 1 mm is, for example, 30% or less (0% or more) of FWHMa, preferably 22% or less.

[0266] Also, when the X-ray rocking curve measurement of (0002) plane diffraction is performed with different widths in the ω direction of the incident-side slit at a plurality of measurement points set at 5 mm intervals within the main surface 100s of the substrate 100 of the present embodiment (between the center and the outer edge), for example, at 95% or more, preferably 100% of all the measurement points, FWHMa - FWHMb is, for example, 30% or less of FWHMa, preferably 22% or less.

[0267] Note that in the substrate 100 of the present embodiment, even if FWHMa < FWHMb, |FWHMa - FWHMb| / FWHMa is 30% or less. Also, in the substrate 100 of the present embodiment, FWHMa may be almost equal to FWHMb and |FWHMa - FWHMb| / FWHMa may be 0%.

[0268] Further, in the substrate 100 of the present embodiment, even if the width of the incident-side slit in the ω direction is widened and the X-ray rocking curve measurement is performed, the variation in the above-described crystal quality factors is small over the region irradiated with X-rays, so that the diffraction pattern has a single peak.

[0269] For reference, a substrate manufactured by a conventional manufacturing method (hereinafter also referred to as a conventional substrate) will be described. The conventional manufacturing method referred to here is, for example, a conventional VAS method, a method of growing a thick film with a c-plane as a growth surface, the above-described DEEP method, a THVPE (Tri-halide vapor phase epitaxy) method, an ammonothermal method, a flux method, and the like.

[0270] In the conventional substrate, at least any one of the above-described crystal quality factors is not better than that of the substrate 100 of the present embodiment. For this reason, FWHMb in the conventional substrate is larger than that of the substrate 100 of the present embodiment.

[0271] In the conventional substrate, in-plane variation of at least any one of the above-described crystal quality factors may occur. For this reason, the diffraction pattern of the (0002) plane when the width of the incident-side slit in the ω direction is widened and the X-ray rocking curve measurement is performed may be wider than the diffraction pattern of the (0002) plane when the width of the incident-side slit in the ω direction is narrowed and the X-ray rocking curve measurement is performed. As a result, in the conventional substrate, FWHMa < FWHMb may occur.

[0272] In the conventional substrate, the radius of curvature of the c-plane is smaller than that of the substrate 100 of the present embodiment. When the width of the incident-side slit in the ω direction is widened, at least a part of the region irradiated with X-rays necessarily includes a portion where at least any one of the crystal quality factors is not better than that of the substrate 100 of the present embodiment. For this reason, the difference FWHMa - FWHMb in the conventional substrate is larger than that of the substrate 100 of the present embodiment.

[0273] In a conventional substrate, in-plane variations of at least any of the above-described crystal quality factors may occur. When the width of the incident-side slit in the ω direction is increased, in at least a part of the region irradiated with X-rays, there may be portions where the fluctuations in the diffraction angle of the (0002) plane are different. For this reason, the diffraction pattern of the (0002) plane when the width of the incident-side slit in the ω direction is increased may have a plurality of peaks.

[0274] As described above, the conventional substrate may not satisfy the above-described requirements for (0002) plane diffraction defined for the substrate 100 of the present embodiment.

[0275] (Regarding X-ray rocking curve measurement of an asymmetric reflection surface with different measurement conditions) Furthermore, the inventor has found that by performing X-ray rocking curve measurement of an asymmetric reflection surface with different measurement conditions, the crystal quality of the substrate 100 of the present embodiment can be appropriately evaluated.

[0276] First, with reference to FIGS. 14 and 15, the measurement region MA when performing X-ray rocking curve measurement of an asymmetric reflection surface will be described. FIGS. 14 and 15 are a schematic perspective view and a schematic plan view showing X-ray rocking curve measurement of an asymmetric reflection surface in the present embodiment, respectively. Note that a part of the main surface 100s of the substrate 100 is temporarily shown as a quadrangle.

[0277] In each figure and the following description, two directions orthogonal to each other along the main surface 100s of the substrate 100 are defined as the "X direction" and the "Y direction", respectively, and the normal direction of the main surface 100s of the substrate 100 is defined as the "Z direction". Also, the direction of the rotation axis of the goniometer is defined as the "y direction", the direction orthogonal to the rotation axis of the goniometer and coinciding with the X direction of the substrate 100 is defined as the "x direction", and the direction orthogonal to the x direction and the y direction is defined as the "z direction".

[0278] Also, as described above, let the rotation angle and the rotation angle direction around the rotation axis of the goniometer be “ω” and “ω direction”, respectively. Also, let the angle at which the rotation axis of the goniometer is inclined with respect to the main surface of the substrate 100 (that is, the inclination angle of the y direction with respect to the Y direction) be “χ”.

[0279] Also, let the width of the incident-side aperture in the ω direction be “d I ”, and let the length of the incident-side aperture in the direction parallel to the rotation axis of the goniometer be “L I ”. On the other hand, let the width of the light-receiving side aperture in the ω direction be “d D ”, and let the length of the light-receiving side aperture in the direction parallel to the rotation axis of the goniometer be “L D ”. In the measurement described later, the light-receiving side aperture corresponds to the aperture of the detector or the entrance aperture of the analyzer crystal.

[0280] Also, let the region where X-rays are irradiated from the X-ray source to the substrate 100 be “X-ray irradiation region IA”, the region where the detector can receive X-rays from the substrate 100 be “light-receivable region DA”, and the region where information regarding the crystal quality of the substrate 100 can actually be obtained be “measurement region MA”. In each figure, the trajectory of the X-rays from the incident side to the main surface 100s and the virtual trajectory of the X-rays from the detector side to the main surface 100s are shown in light gray, the X-ray irradiation region IA and the light-receivable region DA are shown in gray, and the measurement region MA is shown in dark gray.

[0281] As shown in FIGS. 14 and 15, in the X-ray rocking curve measurement of the asymmetric reflection surface, the rotation axis of the goniometer is inclined at an angle χ with respect to the main surface 100s of the substrate 100. For this reason, the X-ray irradiation region IA on the main surface 100s of the substrate 100 in this measurement extends along a direction inclined with respect to the Y axis of the main surface 100s. As a result, the irradiation region IA in this measurement becomes wider than the measurement region along the Y axis in the X-ray rocking curve measurement of the c-plane diffraction described above.

[0282] On the one hand, as shown in FIGS. 14 and 15, a light-receiving region DA as a region where the detector can receive X-rays from the substrate 100 can be obtained, for example, as a virtual X-ray irradiation region when it is assumed that X-rays are irradiated from the detector side toward the substrate 100. The light-receiving region DA on the main surface 100s of the substrate 100 in this measurement extends in a direction symmetric to the X-ray irradiation region IA with respect to the Y-axis of the main surface 100s.

[0283] As a result, as shown in FIGS. 14 and 15, a measurement region MA where information regarding the crystal quality of the substrate 100 can actually be obtained is a region where the X-ray irradiation region IA based on the incident conditions and the light-receiving region DA based on the light-receiving conditions overlap.

[0284] Here, the influence due to the imperfection of the crystal constituting the substrate 100 (i.e., the fluctuation of the diffraction angle) is the width of the component corresponding to the ω direction on the main surface 100s of the substrate 100, that is, the width W in the X direction of the measurement region MA MA depends on.

[0285] As shown in FIG. 15, the width W in the X direction of the measurement region MA MA is obtained by the following formula (j).

[0286]

Equation

[0287] When the width d in the ω direction of the incident-side aperture I is fixed, the width W in the X direction of the measurement region MA MA is compared. For example, when the width d in the ω direction of the light-receiving side aperture D is narrow, the width W in the X direction of the measurement region MA MA becomes narrow. For this reason, in the fluctuation of the diffraction angle of the asymmetric reflecting surface, the influence due to the curvature of the crystal plane is small, and the influence due to the above-described crystal quality factors becomes dominant. On the contrary, for example, when the width d in the ω direction of the light-receiving side aperture D is wide, the width W in the X direction of the measurement region MA MAbecomes wider. Therefore, in the fluctuation of the diffraction angle of the asymmetric reflection surface, both the influence of the crystal quality factor and the influence of the curvature of the crystal plane are superimposed. Accordingly, if the X-ray rocking curve measurement is performed while varying the width d D in the ω direction of the light receiving side aperture, it becomes possible to simultaneously evaluate both the above-described crystal quality factor and the curvature of the crystal plane over the entire measurement region MA.

[0288] Next, in the substrate 100 of the present embodiment, the characteristics when the X-ray rocking curve measurement of the {10-12} plane diffraction is performed as the asymmetric reflection surface will be described.

[0289] Hereinafter, FWHM1 {10-12} and FWHM2 {10-12} are the half-value widths of the {10-12} plane diffraction measured by the X-ray rocking curve measurement, respectively. FWHM1 {10-12} Under the incident conditions for measuring, from a Cu X-ray source, an X-ray mirror that makes the X-rays parallel light, a monochromator with double reflection of Ge(220), and the width d I in the ω direction as the rotation angle direction around the rotation axis of the goniometer is 1.4 mm and the length L I in the direction parallel to the rotation axis is 12 mm, the X-rays of Cu Kα1 are irradiated to the center of the main surface 100s of the substrate 100 through the incident side aperture in this order. FWHM1 {10-12} Under the light receiving conditions for measuring, with the light receiving side slit open and without passing through the analyzer crystal, the X-rays are received by a detector having an aperture (light receiving side aperture) with a width d D in the ω direction of 14.025 mm. FWHM2 {10-12} The incident conditions for measuring are the same as the incident conditions for measuring FWHM1 {10-12} . FWHM2 {10-12} Under the light receiving conditions for measuring, the X-rays are received by a detector through a Ge(220) triple reflection analyzer crystal having an entrance aperture (i.e., light receiving side aperture) with a width d D in the ω direction of 6.54 mm.

[0290] In the substrate 100 of the present embodiment, when performing X-ray rocking curve measurement of {10-12} plane diffraction, the theoretical diffraction angles are χ = 43.19° and ω = 24.05°, respectively.

[0291] Under the above-described conditions, the width W in the X direction of the measurement region MA MA becomes the following value. FWHM1 {10-12} When measuring W MA : approximately 18.92 mm FWHM2 {10-12} When measuring W MA : approximately 9.74 mm

[0292] Incidentally, hereinafter, like when measuring FWHM1 {10-12} the condition of making the width W in the X direction of the measurement region MA relatively wide may be abbreviated as the "wide measurement region condition", and like when measuring FWHM2 MA the condition of making the width W in the X direction of the measurement region MA relatively narrow may be abbreviated as the "narrow measurement region condition". {10-12} When measuring W MA

[0293] In the substrate 100 of the present embodiment, as described above, over a wide range of the main surface 100s, all of the above-described crystal quality elements are well-balanced and good, and the curvature of the crystal plane of the substrate 100 is small. In the substrate 100 of the present embodiment, even when performing X-ray rocking curve measurement of {10-12} plane diffraction under the wide measurement region condition, over the entire measurement region MA, the above-described crystal quality elements are well-balanced and good, and the curvature of the crystal plane is small, so the fluctuation of the diffraction angle of the {10-12} plane does not become so large.

[0294] As a result, in the substrate 100 of the present embodiment, FWHM1 under the wide measurement region condition {10-12} is, for example, 50 arcsec or less, preferably 40 arcsec or less.

[0295] In addition, in the substrate 100 of the present embodiment, regardless of the crystal orientation, all of the above-described crystal quality elements are well-balanced and good, and the curvature of the crystal plane of the substrate 100 is small. In the substrate 100 of the present embodiment, even if X-ray rocking curve measurements of diffraction of equivalent crystal planes are performed from different directions under wide measurement region conditions, the variation in the diffraction angle fluctuation of the {10-12} plane does not become very large.

[0296] As a result, in the substrate 100 of the present embodiment, FWHM1 that measures the diffraction of the equivalent crystal plane represented by the {10-12} plane from three directions rotated by 60° each in the circumferential direction about the normal line at the center of the main surface 100s {10-12} The maximum-minimum difference is, for example, 9 arcsec or less.

[0297] Furthermore, in the substrate 100 of the present embodiment, as described above, over a wide range of the main surface 100s, all of the above-described crystal quality elements are well-balanced and good, and the curvature of the crystal plane of the substrate 100 is small. Therefore, even if X-ray rocking curve measurements of {10-12} plane diffraction are performed with different measurement conditions, the variation in the diffraction angle fluctuation of the {10-12} plane becomes small.

[0298] As a result, at the center of the main surface 100s of the substrate 100 of the present embodiment, the ratio of FWHM2 {10-12} to FWHM1 {10-12} is, for example, 80% or more.

[0299] Note that the "ratio of FWHM2 {10-12} to FWHM1 {10-12} " here refers to the value obtained by (FWHM2 {10-12} / FWHM1 {10-12} ) × 100.

[0300] On the other hand, even if the above-described crystal quality elements and the curvature of the crystal plane satisfy the above requirements, if there are defects in the surface state such as pits in a part of the main surface 100s of the substrate 100, FWHM1 under wide measurement region conditions {10-12} is larger than FWHM2 {10-12}may become smaller. That is, FWHM1 {10-12} with respect to FWHM2 {10-12} may exceed 100%.

[0301] On the other hand, in the present embodiment, it is preferable that there are no defects in the surface state on the main surface 100s of the substrate 100. That is, the ratio of FWHM2 {10-12} to FWHM1 {10-12} is preferably, for example, 100% or less.

[0302] For reference, in a conventional substrate, at least one of the above-described crystal quality elements is not good. Therefore, when measurement is performed under wide measurement region conditions, at least a part of the measurement region MA necessarily includes a portion where at least one of the crystal quality elements is not good. Further, in the conventional substrate, the curvature of the crystal plane is larger than that of the substrate 100 of the present embodiment. As a result, FWHM1 {10-12} in the conventional substrate is larger than that of the substrate 100 of the present embodiment.

[0303] In addition, in the conventional substrate, in-plane variation of at least one of the above-described crystal quality elements may occur. Therefore, the maximum-minimum difference of FWHM1 {10-12} of the equivalent crystal plane diffraction measured from three directions in the conventional substrate is larger than that of the substrate 100 of the present embodiment.

[0304] In addition, in the conventional substrate, as described above, at least one of the crystal quality elements is not good and the curvature of the crystal plane is larger than that of the substrate 100 of the present embodiment. Therefore, the ratio of FWHM2 {10-12} to FWHM1 {10-12} in at least one direction of measurement of the conventional substrate is smaller than that of the substrate 100 of the present embodiment.

[0305] As described above, the conventional substrate may not satisfy the above-described requirements for {10-12} plane diffraction defined for the substrate 100 of the present embodiment.

[0306] (4) Effects obtained by this embodiment According to this embodiment, one or more of the following effects can be obtained.

[0307] (a) In the three-dimensional growth step S200, by generating an inclined interface 30i other than the c-plane on the surface of the single crystal constituting the three-dimensional growth layer 30, at the position where the inclined interface 30i is exposed, the dislocation can be bent and propagated in a direction substantially perpendicular to the inclined interface 30i. Thereby, the dislocations can be locally collected. By locally collecting the dislocations, the dislocations having mutually opposite Burgers vectors can be made to disappear. Alternatively, by the locally collected dislocations forming loops, the propagation of the dislocations to the surface side of the flattening layer 40 can be suppressed. In this way, the dislocation density on the surface of the flattening layer 40 can be reduced. As a result, a substrate 100 with a reduced dislocation density compared to the lower substrate 10 can be obtained.

[0308] (b) In the three-dimensional growth step S200, the c-plane 30c is made to disappear from the top surface 30u of the three-dimensional growth layer 30 at least once. Thereby, a plurality of valley portions 30v and a plurality of top portions 30t can be formed on the surface of the three-dimensional growth layer 30. As a result, the dislocations propagating from the lower substrate 10 can be surely bent at the position where the inclined interface 30i in the three-dimensional growth layer 30 is exposed.

[0309] Here, consider the case where the c-plane remains in the three-dimensional growth step. In this case, in the portion where the c-plane remains, the dislocations propagating from the lower substrate propagate in a substantially vertically upward direction without being bent and reach the surface of the second layer. Therefore, above the portion where the c-plane remains, the dislocations are not reduced, and a high dislocation density region is formed.

[0310] On the other hand, according to the present embodiment, in the three-dimensional growth step S200, by disappearing the c-plane 30c from the top surface 30u of the three-dimensional growth layer 30 at least once, the surface of the three-dimensional growth layer 30 can be composed only of inclined interfaces 30i other than the c-plane, and a plurality of valleys 30v and a plurality of peaks 30t can be formed on the surface of the three-dimensional growth layer 30. Thereby, the dislocations propagating from the lower base substrate 10 side can be surely bent over the entire surface of the three-dimensional growth layer 30. By surely bending the dislocations, it becomes easier to disappear a part of the plurality of dislocations, or it becomes difficult for a part of the plurality of dislocations to propagate to the surface side of the planarization layer 40. As a result, it becomes possible to reduce the dislocation density over the entire main surface 1s of the substrate 100 obtained from the present growth layer 44.

[0311] (c) In the present embodiment, a cycle including the three-dimensional growth step S200 and the planarization step S300 is repeated a plurality of times. By one cycle including the three-dimensional growth step S200 and the planarization step S300, as described above, the dislocations propagating from the lower base substrate 10 side can be bent and the dislocations can be locally collected. By further performing the next cycle, the dislocations remaining in the planarization layer 40 in the previous cycle can be further bent and the dislocations can be locally collected. Thereby, the dislocation density on the surface of the planarization layer 40 in the next cycle can be reduced compared to the dislocation density on the surface of the planarization layer 40 in the previous cycle. By repeating the cycle in this way, the dislocation density can be gradually reduced according to the number of times the cycle is repeated.

[0312] (d) Even if the c-plane 30c remains in the three-dimensional growth step S200 of any one of the plurality of cycles, in the cycles after the next time, the c-plane 30c can be disappeared at least once above the portion where the c-plane 30c remains. Thereby, above the portion where the c-plane 30c remains, the dislocations can be surely bent. As a result, it becomes possible to stably reduce the dislocation density.

[0313] (e) In the inclined interface expansion step S220 of the cycles after the second cycle, when looking at an arbitrary cross-section perpendicular to the main surface 10s of the underlying substrate 10, the average distance L between the closest tops is made longer than that in the inclined interface expansion step S220 of the first cycle. Thereby, in the cycles after the second cycle, the distance over which dislocations bend and propagate can be made longer than that in the first cycle. As a result, on the surface of the planarization layer 40 in the cycles after the second cycle, a dislocation-free density region with a 50-μm side that does not overlap can be easily formed, and the density of the dislocation-free region can be increased.

[0314] (f) In the inclined interface expansion step S220 in each of the plurality of cycles, the average distance L between the closest tops is gradually increased as the plurality of cycles are repeated, for example. Thereby, as the plurality of cycles are repeated, the distance over which dislocations bend and propagate can be gradually increased. As a result, as the plurality of cycles are repeated, the dislocation-free region on the surface of the planarization layer 40 can be gradually widened, or the density of the dislocation-free region on the surface of the planarization layer 40 can be gradually increased.

[0315] (g) In the inclined interface expansion step S220 in each of the plurality of cycles, when looking at an arbitrary cross-section perpendicular to the main surface 10s of the underlying substrate 10, by setting the average distance L between the closest tops to be more than 100 μm, the distance over which dislocations bend and propagate can be ensured to be at least more than 50 μm. Thereby, dislocations can be sufficiently concentrated above the approximate center between a pair of tops 30t in the three-dimensional growth layer 30. As a result, the dislocation density on the surface of the planarization layer 40 can be sufficiently reduced.

[0316] (h) Further, in the inclined interface expansion step S220 of at least the last cycle among the plurality of cycles, when looking at an arbitrary cross-section perpendicular to the main surface 10s of the lower base plate 10, it is preferable that the average distance L between the closest tops is, for example, more than 200 μm. Thereby, in the steps after the inclined interface expansion step S220 of the last cycle, it is possible to ensure that the distance for dislocations to bend and propagate is at least more than 100 μm. As a result, it is possible to form a dislocation-free region with at least a 100-μm side length on at least a part of the surface of the planarization layer 40 of the last cycle.

[0317] (i) In the three-dimensional growth step S200 in each of the plurality of cycles, by adjusting the first growth conditions so as to satisfy formula (1), it is possible to generate a {11 - 2m} plane with m ≥ 3 as the inclined interface 30i. Thereby, the inclination angle of the {11 - 2m} plane with respect to the c-plane 30c can be made gentle. Specifically, the inclination angle can be made 47.3° or less. By making the inclination angle of the {11 - 2m} plane with respect to the c-plane 30c gentle, the period of the plurality of tops 30t can be lengthened. Specifically, when looking at an arbitrary cross-section perpendicular to the main surface 10s of the lower base plate 10, the average distance L between the closest tops can be made more than 100 μm.

[0318] Incidentally, for reference, usually, when an etch pit is generated in a nitride semiconductor substrate using a predetermined etchant, an etch pit composed of a {1 - 10n} plane is formed on the surface of the substrate. On the other hand, on the surface of the three-dimensional growth layer 30 grown under predetermined conditions in the present embodiment, it is possible to generate a {11 - 2m} plane with m ≥ 3. Therefore, it is considered that in the present embodiment, an inclined interface 30i peculiar to the manufacturing method is formed as compared with a normal etch pit.

[0319] (j) In the three-dimensional growth process S200 in each of a plurality of cycles, after disappearing the c-plane 30c from the surface of the three-dimensional growth layer 30, while maintaining a state where the inclined interface 30i occupies a large part of the surface, the growth of the three-dimensional growth layer 30 is continued over a predetermined thickness. Thereby, the c-plane 30c can be surely disappeared over the entire surface of the three-dimensional growth layer 30. For example, even if the timing at which the c-plane 30c disappears on the surface of the three-dimensional growth layer 30 in the inclined interface expansion process S220 is shifted and the c-plane 30c remains in a part of the inclined interface expansion layer 32, the c-plane 30c can be surely disappeared.

[0320] Also, after the c-plane 30c disappears, by continuing the growth by the inclined interface 30i of the three-dimensional growth layer 30, it is possible to sufficiently secure the time for bending the dislocation at the position where the inclined interface 30i is exposed. Here, if the c-plane growth is performed immediately after the c-plane disappears, there is a possibility that the dislocation is not sufficiently bent and propagates in a substantially vertical direction toward the surface of the second layer. On the other hand, in the present embodiment, by sufficiently securing the time for bending the dislocation at the position where the inclined interface 30i other than the c-plane is exposed, the dislocation in the vicinity of the top 30t of the three-dimensional growth layer 30 can be surely bent, and the propagation of the dislocation in a substantially vertical direction from the underlying substrate 10 toward the surface of the planarization layer 40 can be suppressed. Thereby, the concentration of the dislocations above the top 30t of the three-dimensional growth layer 30 can be suppressed.

[0321] (k) By the manufacturing method of the present embodiment, the radius of curvature of the c-plane 100c of the substrate 100 can be made larger than the radius of curvature of the c-plane 10c of the underlying substrate 10. Thereby, the variation in the off-angle θ of the c-axis 100ca with respect to the normal of the main surface 100s of the substrate 100 can be made smaller than the variation in the c-axis 10ca off-angle of the underlying substrate 10.

[0322] As one of the reasons why the radius of curvature of the c-plane 100c of the substrate 100 can be increased, for example, the following reasons can be considered.

[0323] As described above, in the three-dimensional growth step S200, the three-dimensional growth layer 30 is three-dimensionally grown with the inclined interface 30i other than the c-plane as the growth surface, thereby forming the inclined interface growth region 70. In the inclined interface growth region 70, oxygen is more easily incorporated compared to the c-plane growth region 60. For this reason, the oxygen concentration in the inclined interface growth region 70 becomes higher than the oxygen concentration in the c-plane growth region 60. That is, the inclined interface growth region 70 can be considered as a high oxygen concentration region.

[0324] In this way, by incorporating oxygen into the high oxygen concentration region, the lattice constant of the high oxygen concentration region can be made larger than the lattice constants of other regions outside the high oxygen concentration region (Reference: Chris G. Van de Walle, Physical Review B vol.68, 165209 (2003)). In the underlying substrate 10 or the c-plane growth region 60 grown with the c-plane 30c of the three-dimensional growth layer 30 as the growth surface, due to the curvature of the c-plane 10c of the underlying substrate 10, stress concentrated toward the center of curvature of the c-plane is applied. On the other hand, by relatively increasing the lattice constant of the high oxygen concentration region, a stress that expands the c-plane 30c in the lateral direction can be generated in the high oxygen concentration region. Thereby, the stress concentrated toward the center of curvature of the c-plane 30c below the high oxygen concentration region and the stress that expands the c-plane 30c of the high oxygen concentration region in the lateral direction can be offset.

[0325] By obtaining such a stress offset effect by the three-dimensional growth layer 30 in this way, the radius of curvature of the c-plane 100c of the substrate 100 obtained from the planarization layer 40 can be made larger than the radius of curvature of the c-plane 10c of the underlying substrate 10 obtained by the conventional VAS method.

[0326] (l) In the substrate 100 obtained by the manufacturing method of the present embodiment, the dislocation density can be lowered, the off-angle variation can be made as small as possible, and all of the above crystal quality elements that determine the full width at half maximum of the X-ray rocking curve measurement can be made well-balanced and good.

[0327] As a result, regarding the X-ray rocking curve measurement of (0002) plane diffraction, in the substrate 100 of the present embodiment, FWHMb can be set to 32 arcsec or less. Further, in the substrate 100 of the present embodiment, even when the width of the incident-side slit in the ω direction is 1 mm, the radius of curvature of the c plane is large over the entire region irradiated with X-rays, and due to the above-described crystal quality factors being well balanced, (FWHMa - FWHMb) / FWHMa can be set to 30% or less.

[0328] Furthermore, regarding the X-ray rocking curve measurement of {10-12} plane diffraction, in the substrate of the present embodiment, FWHM1 {10-12} can be set to 50 arcsec or less. Also, in the substrate of the present embodiment, the maximum-minimum difference of FWHM1 {10-12} measured by diffracting equivalent crystal planes represented by the {10-12} plane from three directions can be set to 9 arcsec or less. Also, in the substrate 10 of the present embodiment, the ratio of FWHM2 {10-12} to FWHM1 {10-12} can be set to 80% or more.

[0329] <Other Embodiments> The embodiments of the present invention have been specifically described above. However, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the gist thereof.

[0330] In the above-described embodiment, the case where the base substrate 10 is a GaN free-standing substrate has been described. However, the base substrate 10 is not limited to a GaN free-standing substrate, and for example, it may be a free-standing substrate made of a group III nitride semiconductor such as aluminum nitride (AlN), aluminum gallium nitride (AlGaN), indium nitride (InN), indium gallium nitride (InGaN), aluminum indium gallium nitride (AlInGaN), that is, Al x In y Ga 1-x-y N (0 ≦ x ≦ 1, 0 ≦ y ≦ 1, 0 ≦ x + y ≦ 1).

[0331] In the above-described embodiment, the case where the substrate 100 is a GaN free-standing substrate has been described. However, the substrate 100 is not limited to the GaN free-standing substrate. For example, it may be a free-standing substrate made of a group-III nitride semiconductor such as AlN, AlGaN, InN, InGaN, AlInGaN, etc., that is, Al x In y Ga 1-x-y It may be a free-standing substrate made of a group-III nitride semiconductor represented by the composition formula of N (0 ≦ x ≦ 1, 0 ≦ y ≦ 1, 0 ≦ x + y ≦ 1).

[0332] In the above-described embodiment, the case where the substrate 100 is n-type has been described. However, the substrate 100 may be p-type or semi-insulating. For example, when manufacturing a semiconductor device as a high electron mobility transistor (HEMT) using the substrate 100, it is preferable that the substrate 100 has semi-insulating properties.

[0333] In the above-described embodiment, in the polishing step S180 of the lower substrate preparation step S100, the case where the main surface 10s of the lower substrate 10 is mirror-finished has been described. However, in the polishing step S180, the main surface 10s of the lower substrate 10 may be coarsely polished while maintaining the so-called epi-ready state in which a single crystal of a group-III nitride semiconductor can grow epitaxially. Specifically, the root mean square roughness RMS of the main surface 10s of the lower substrate 10 is, for example, set to be 1 nm or more and 10 nm or less. By setting the RMS of the main surface 10s of the lower substrate 10 within the above range, when growing the three-dimensional growth layer 30 on the lower substrate 10 in the first cycle of the three-dimensional growth step S200, the generation of the inclined interface 30i other than the c-plane on the surface of the three-dimensional growth layer 30 can be promoted. Also, by setting the RMS of the main surface 10s of the lower substrate 10 within the above range, in the first cycle of the three-dimensional growth step S200, it is possible to suppress the surface of the three-dimensional growth layer 30 from becoming excessively rough and suppress the reduction in the average distance L between the nearest tops in the three-dimensional growth layer 30.

[0334] Alternatively, for example, while maintaining good crystal quality of the bulk portion in the lower base substrate 10, crystal strain introduced by processing such as the slicing step S170 and the polishing step S180 of the lower base substrate 10 may be left to remain on the main surface 10s side of the lower base substrate 10. Specifically, when X-ray rocking curve measurement is performed with the incident angle with respect to the main surface 10s of the processed lower base substrate 10 being 2°, the full width at half maximum (FWHM) of the (10-10) plane diffraction is, for example, made larger than the half width of the lower base substrate 10 before processing and set to be 60 arcsec or more and 200 arcsec or less. By setting the FWHM of the (10-10) plane diffraction within the above range, in the three-dimensional growth step S200 of the first cycle, due to the crystal strain on the main surface 10s side of the lower base substrate 10, the stable crystal plane appearing on the surface of the three-dimensional growth layer 30 can be changed. As a result, an inclined interface 30i other than the c-plane can be generated on the surface of the three-dimensional growth layer 30. Further, by setting the FWHM of the (10-10) plane diffraction within the above range, in the three-dimensional growth step S200 of the first cycle, it is possible to suppress an excessive amount of dislocations from occurring in the three-dimensional growth layer 30 due to the crystal strain on the main surface 10s side of the lower base substrate 10.

[0335] In the above-described embodiment, the case where the three-dimensional growth process S200 of the first cycle is performed immediately after the lower base plate preparation process S100 has been described. However, the present invention is not limited to this case. In the initial process S190, after the lower base plate preparation process S100 and before the three-dimensional growth process S200 of the first cycle, an initial layer may be grown directly on the main surface 10s of the lower base plate 10 with the c-plane as the growth surface. At this time, the growth rate of the initial layer in the initial process S190 is made lower than the growth rate of the three-dimensional growth layer 30 in the three-dimensional growth process S200. Thereby, the initial layer can be grown by step-flow on the main surface 10s of the lower base plate 10. By growing the initial layer by step-flow, even if there are abnormal portions such as rough portions or portions with different crystallinity in a part of the main surface 10s of the lower base plate 10, regardless of the state of the main surface 10s of the lower base plate 10, the surface morphology and crystallinity of the initial layer can be made substantially uniform over the entire surface. By making the surface morphology and crystallinity of the initial layer substantially uniform, in the three-dimensional growth process S200 of the first cycle, the occurrence state of the inclined interface 30i such as the inclination angle of the inclined interface 30i with respect to the c-plane 30c can be made substantially uniform over the entire surface of the three-dimensional growth layer 30. Specifically, the formation of a region with a short nearest-neighbor top distance in a part of the surface of the three-dimensional growth layer 30 can be suppressed, and the nearest-neighbor top distance can be made substantially uniform over the entire surface of the three-dimensional growth layer 30. As a result, the formation of a region with a high dislocation density in a part of the surface of the planarization layer 40 can be suppressed, and the dislocation density can be made low over the entire surface of the planarization layer 40.

[0336] In the above-described embodiment, the case where the growth temperature is mainly adjusted as the first growth condition in the three-dimensional growth process S200 of each cycle has been described. However, if the first growth condition satisfies the formula (1), as the first growth condition, growth conditions other than the growth temperature may be adjusted, or the growth temperature and growth conditions other than the growth temperature may be adjusted in combination.

[0337] In the above-described embodiments, in the planarization step S300 of each cycle, the case where the growth temperature is mainly adjusted as the second growth condition has been described. However, if the second growth condition satisfies formula (2), as the second growth condition, growth conditions other than the growth temperature may be adjusted, or a combination of the growth temperature and growth conditions other than the growth temperature may be adjusted.

[0338] In the above-described embodiments, the case where the growth conditions in the inclined interface maintenance step S240 of each cycle are maintained under the above-described first growth conditions in the same manner as in the inclined interface expansion step S220 has been described. However, if the growth conditions in the inclined interface maintenance step S240 satisfy the first growth condition, the growth conditions in the inclined interface maintenance step S240 may be made different from the growth conditions in the inclined interface expansion step S220.

[0339] In the above-described embodiments, in the implementation number determination step S400, the case where the cycle is terminated when the number of implementations reaches 2 has been described. However, the number of implementations may be more than 2.

[0340] In the above-described embodiments, in the three-dimensional growth step S200 of the cycles after the first cycle, the case where the three-dimensional growth layer 30 is grown under the same first growth conditions as in the three-dimensional growth step S200 of the first cycle has been described. However, if the growth conditions in the three-dimensional growth step S200 of the cycles after the first cycle satisfy the first growth condition, the growth conditions of this step may be made different from the growth conditions in the three-dimensional growth step S200 of the first cycle.

[0341] In the above-described embodiments, in the planarization step S300 of the cycles after the first cycle, the case where the planarization layer 40 is grown under the same second growth conditions as in the planarization step S300 of the first cycle has been described. However, if the growth conditions in the planarization step S300 of the cycles after the first cycle satisfy the second growth condition, the growth conditions of this step may be made different from the growth conditions in the planarization step S300 of the first cycle.

[0342] In the above-described embodiment, the case where the growth conditions in the main growth step S500 are maintained under the second growth conditions, similar to the planarization step S300 of each cycle, has been described. However, if the growth conditions in the main growth step S500 satisfy the second growth conditions, the growth conditions in this step may be made different from the growth conditions in the planarization step S300 of each cycle.

[0343] In the above-described embodiment, the case where the second crystal layer 6 or the main growth layer 44 is sliced using a wire saw in the slicing step S170 and the slicing step S600 has been described. However, for example, an outer peripheral blade slicer, an inner peripheral blade slicer, an electric discharge machine, or the like may be used.

[0344] In the above-described embodiment, the case where the substrate 100 is obtained by slicing the main growth layer 44 in the laminated structure 90 has been described. However, the present invention is not limited to this case. For example, the laminated structure 90 may be used as it is to manufacture a semiconductor laminate for manufacturing a semiconductor device. Specifically, after manufacturing the laminated structure 90, in the semiconductor laminate manufacturing step, a semiconductor functional layer is epitaxially grown on the laminated structure 90 to manufacture a semiconductor laminate. After manufacturing the semiconductor laminate, the back surface side of the laminated structure 90 is polished, and the underlying substrate 10 and the plurality of laminated units 50 in the laminated structure 90 are removed. Thereby, a semiconductor laminate having the main growth layer 44 and the semiconductor functional layer, similar to the above-described embodiment, can be obtained. According to this case, the slicing step S600 and the polishing step S700 for obtaining the substrate 100 can be omitted.

[0345] In the above-described embodiment, the case where the manufacturing process is terminated after manufacturing the substrate 100 has been described. However, the substrate 100 may be used as the base substrate 10, and the steps S200 to S700 may be performed again. Thereby, a substrate 100 with a further reduced dislocation density can be obtained. Also, a substrate 100 with a further reduced variation in the off-angle θ of the c-axis 100ca can be obtained. Further, the steps S200 to S700 using the substrate 100 as the base substrate 10 may be defined as one cycle, and the cycle may be repeated a plurality of times. Thereby, the dislocation density of the substrate 100 can be gradually reduced according to the number of times the cycle is repeated. Also, the variation in the off-angle θ of the c-axis 100ca in the substrate 100 can be gradually reduced according to the number of times the cycle is repeated.

Example

[0346] Hereinafter, various experimental results that support the effects of the present invention will be described. In the following, the "nitride semiconductor substrate" may be simply abbreviated as "substrate".

[0347] (1) Experiment 1 (1-1) Fabrication of nitride semiconductor substrate The substrates of Samples 1 to 4 were fabricated as follows. Note that a stacked structure was also fabricated for Sample 1.

[0348] <Fabrication conditions of the nitride semiconductor substrate of Sample 1> In Sample 1, cycles including a three-dimensional growth process and a planarization process were performed twice. (Base substrate) Material: GaN Fabrication method: VAS method Diameter: 2 inches Thickness: 400 μm Low-index crystal plane closest to the main surface: c-plane No pattern processing such as a mask layer on the main surface. Off-angle at the center of the main surface: 0.4° in the m direction

[0349] [First cycle] (Three-dimensional growth layer) Material: GaN Growth method: HVPE method First growth condition: The growth temperature was set to be 980°C or higher and 1,020°C or lower, and the V / III ratio was set to be 2 or higher and 20 or lower. At this time, at least one of the growth temperature and the V / III ratio was adjusted within the above ranges so that the first growth condition satisfied formula (1). (Planarization layer) Material: GaN Growth method: HVPE method Growth temperature: 1,050°C V / III ratio: 2 Note that the above second growth condition satisfies formula (2). Thickness from the main surface of the underlying substrate to the interface of the first stacked unit (thickness of the first stacked unit): approximately 800 μm

[0350] [Second cycle] (Three-dimensional growth layer) Growth conditions similar to those of the three-dimensional growth layer in the first cycle (Planarization layer) Growth conditions similar to those of the planarization layer in the first cycle Thickness from the interface of the first stacked unit to the interface of the second stacked unit (thickness of the second stacked unit): approximately 800 μm (Main growth layer) Growth conditions similar to those of the planarization layer in each cycle Thickness: approximately 2 mm Note that in the laminated structure for observation, the thickness of the main growth layer was set to be approximately 200 μm. (Slicing conditions) Thickness of the substrate: 400 μm Kerf loss: 200 μm

[0351] (Fabrication conditions of the nitride semiconductor substrate of Sample 2) In Sample 2, one cycle including the three-dimensional growth process and the planarization process was used. That is, the cycle was not repeated. (Underlying substrate) Equivalent to the underlying substrate used in Sample 1 (Three-dimensional growth layer) Growth conditions similar to those of the three-dimensional growth layer in the first cycle of Sample 1 (Planarization layer) Growth conditions similar to those of the planarization layer in the first cycle of Sample 1 Thickness from the main surface of the underlying substrate to the interface on the planarization layer: approximately 800 μm (This growth layer) Growth conditions similar to those of the above-mentioned planarization layer Thickness: approximately 2 mm (Slicing conditions) Same as Sample 1.

[0352] (Fabrication conditions of the nitride semiconductor substrate of Sample 3) The method by the above-mentioned "c-plane thick film growth" was used. (Underlying substrate) Material: GaN Fabrication method: VAS method Diameter: 62 mm Thickness: 400 μm The low-index crystal plane closest to the main surface: c-plane Off-angle at the center of the main surface: 0.5° in the m-axis direction No pattern processing such as a mask layer on the main surface. (Crystal layer) Material: GaN Growth method: HVPE method Growth temperature: 1050 °C V / III ratio: 2.8 Growth time: 15 hours (Processing) Grinding: The cylindrical region was removed to obtain a cylindrical region with a diameter of 56 mm. Slicing: 5 pieces with a thickness of 630 μm Beveling: The diameter was set to 50.8 mm. Polishing: The thickness was set to 400 - 450 μm.

[0353] (Fabrication conditions of the nitride semiconductor substrate of Sample 4) The substrate of Sample 4 was fabricated by the same conventional VAS method as the underlying substrate. For Sample 4, except that the absolute value of the off-angle and the off-direction are different from those of the underlying substrate, the curvature radius of the c-plane, the dislocation density, etc. are equivalent to those of the underlying substrate.

[0354] (1-2) Evaluation (Observation by fluorescence microscope) Using a fluorescence microscope, the cross-section of the laminated structure before slicing the substrate of Sample 1 was observed.

[0355] (Observation by multiphoton excitation microscope) Using a multiphoton excitation microscope, the main surfaces of the substrates of Samples 1 to 3 and the substrate of Sample 4 corresponding to the underlying substrate were observed. At this time, the dislocation density was measured by measuring the dark spot density over the entire main surface every 250 μm in the field of view. It should be noted that it has been confirmed by measuring while shifting the focus in the thickness direction that all of the dark spots in these substrates are dislocations. Also, at this time, the ratio of the number of regions where the dislocation density is less than 1×10 6 cm -2 or less or less than 7×10 5 cm -2 was determined with respect to the total number of measurement regions at a field of view of 250 μm square.

[0356] (X-ray rocking curve measurement of (0002) plane diffraction) For each of the substrates of Samples 1 to 3 and the substrate of Sample 4 corresponding to the underlying substrate, the following two types of X-ray rocking curve measurements of (0002) plane diffraction were performed.

[0357] For the X-ray rocking curve measurement, "X'Pert-PRO MRD" manufactured by Spectris was used, and as the monochromator on the incident side, the company's "Hybrid Monochromator" was used. The hybrid monochromator has, in order from the X-ray source side, an X-ray mirror and two crystals of the Ge(220) plane. In this measurement, first, the X-rays emitted from the X-ray source are made into parallel light by the X-ray mirror. Thereby, the number of photons of the X-rays used (i.e., the X-ray intensity) can be increased. Next, the parallel light from the X-ray mirror is made into monochromatic light of Cu Kα1 by two crystals of the Ge(220) plane. Next, the monochromatic light from the two crystals of the Ge(220) plane is narrowed to a predetermined width through the incident slit and incident on the substrate. When the full width at half maximum of the rocking curve of the (0002) plane of a perfect crystal GaN is measured using the hybrid monochromator, it is 25.7 arcsec as obtained by simulation. That is, this full width at half maximum is the theoretical measurement limit when measured with the above optical system.

[0358] In addition, the X-rays incident on the substrate in this measurement are made into parallel light directed toward the substrate side in a cross-section perpendicular to the rotation axis of the goniometer (i.e., a cross-section along the ω direction). However, the incident X-rays are not parallel light in a cross-section along the rotation axis of the goniometer. For this reason, between the time when the X-rays reach the substrate from the slit, the width of the X-rays in the ω direction is almost constant, but the width in the direction perpendicular to the ω direction of the X-rays spreads. Therefore, in the X-ray rocking curve measurement, the full width at half maximum of the X-rays diffracted by a predetermined crystal plane depends on the width in the ω direction where the X-rays become parallel light among the incident slits.

[0359] On the other hand, the receiving side was set to open. The window width in the ω direction of the detector on the receiving side was set to 14.025 mm. In the above optical system, since the goniometer radius is 420 mm, it is possible to measure the variation of the Bragg angle of ±0.95°.

[0360] (X-ray Rocking Curve Measurement 1) The width of the incident-side slit in the ω direction was set to 0.1 mm, and X-ray rocking curve measurements of the (0002) plane were performed for each of the substrates of Samples 1 to 3 and the substrate of Sample 4 corresponding to the lower base substrate. At this time, among the main planes of each substrate, at a plurality of measurement points set at 5 mm intervals on a straight line passing through the center and along the m-axis direction, and on a straight line passing through the center and along the a-axis direction orthogonal to the m-axis, the measurement was performed. At this time, X-rays were incident from the side defined as positive as the position in the main plane of the substrate. As a result of the measurement, the peak angle ω formed by the X-rays incident on the main plane and the main plane was plotted against the position on the straight line, and the peak angle ω was approximated by a linear function of the position. The radius of curvature of the c-plane was obtained from the reciprocal of the slope of the linear function.

[0361] Also, at each measurement point, the full width at half maximum FWHMb of the (0002) plane diffraction was obtained when the width of the incident-side slit in the ω direction was 0.1 mm.

[0362] (X-ray Rocking Curve Measurement 2) The width of the incident-side slit in the ω direction was set to 1 mm, and X-ray rocking curve measurements were performed for each of the substrates of Samples 1 to 3 and the substrate of Sample 4 corresponding to the lower base substrate. At this time, among the main planes of each substrate, at a plurality of measurement points set at 5 mm intervals on a straight line passing through the center and along the m-axis direction, and on a straight line passing through the center and along the a-axis direction orthogonal to the m-axis, the measurement was performed. As a result of the measurement, at each measurement point, the full width at half maximum FWHMa of the (0002) plane diffraction was obtained when the width of the incident-side slit in the ω direction was 1 mm. Furthermore, at each measurement point, the ratio of FWHMa - FWHMb to FWHMa was obtained.

[0363] In X-ray rocking curve measurements 1 and 2, when X-rays were incident at a Bragg angle of 17.28° of the (0002) plane with respect to the main plane of each substrate, when the width of the slit in the ω direction was 0.1 mm, the X-ray footprint was approximately 0.337 mm, and when the width of the slit in the ω direction was 1 mm, the X-ray footprint was approximately 3.37 mm.

[0364] (1-3) Results The results are shown in Tables 1 to 5.

[0365]

Table 1

[0366]

Table 2

[0367]

Table 3

[0368]

Table 4

[0369]

Table 5

[0370] <Sample 4: corresponding to the lower base substrate> Referring to Table 1, Table 5, and FIG. 21(b), the results of the substrate of Sample 4 corresponding to the lower base substrate fabricated by the conventional VAS method will be described. FIG. 21(b) is a diagram showing the normalized X-ray diffraction pattern when the X-ray rocking curve measurement of (0002) plane diffraction is performed with different incident-side slits for the nitride semiconductor substrate of Sample 4. Note that FIG. 21(b) shows the measurement results in the direction along the a-axis. Also, in the figure, "Line width" means the above-mentioned X-ray footprint.

[0371] (Dislocation) As shown in Table 1, in the substrate of Sample 4 corresponding to the lower base substrate, the average dislocation density was 1×10 6 cm -2 or more. In the main plane of the substrate of Sample 4, there was no region where the dislocation density was less than 1×10 6 cm -2 .

[0372] Also, although not shown in the figures, in the substrate of Sample 4, dislocations were uniformly dispersed in the plane. Therefore, over the entire substrate of Sample 4, the size of the dislocation-free region was smaller than 50 μm square, and a 50 μm square dislocation-free region was not formed.

[0373] (X-ray rocking curve measurement results) As shown in Table 1, the radius of curvature of the c-plane in the substrate of Sample 4 was less than 10 m.

[0374] As shown in Table 5, in the substrate of Sample 4, the variation in the off-angle of the c-axis within a diameter of 40 mm was about ±0.24°.

[0375] Also, as shown in Table 5, in the substrate of Sample 4, at all measurement points, when the width of the incident-side slit in the ω direction was 0.1 mm, FWHMb was more than 32 arcsec.

[0376] As shown in Fig. 21(b), in the substrate of Sample 4, the X-ray diffraction pattern when the width of the incident-side slit in the ω direction was 1 mm was broader than the X-ray diffraction pattern when the width of the incident-side slit in the ω direction was 0.1 mm.

[0377] Therefore, as shown in Table 5, in the substrate of Sample 4, at all measurement points, (FWHMa - FWHMb) / FWHMa was more than 30%.

[0378] <Sample 3: c-plane thick film growth> Referring to Table 1, Table 4, and Fig. 20, the results of the substrate of Sample 3 obtained by c-plane thick film growth will be described. Fig. 20 is a view of the main surface of the nitride semiconductor substrate of Sample 3 observed using a multiphoton excitation microscope.

[0379] (Dislocations) As shown in Table 1, in the substrate of Sample 3 obtained by c-plane thick film growth, since the dislocation density decreased in inverse proportion to the thickness of the crystal layer, the average dislocation density was reduced compared to that of Sample 4 equivalent to the base substrate.

[0380] However, in the substrate of Sample 3, the proportion of the region where the dislocation density was less than 1×10 6 cm -2 was less than 90%, and the proportion of the region where the dislocation density was less than 7×10 cm -2 was less than 80%.

[0381] Also, as shown in Fig. 20, in the substrate of Sample 3, the dislocations were uniformly dispersed in the plane. In addition, even in the regions not shown in the figure, the dislocation distribution was the same as that in Fig. 20. For this reason, over the entire substrate of Sample 3, the size of the dislocation-free region was smaller than 50 μm square, and a 50 μm square dislocation-free region was not formed.

[0382] Thus, even when using the method of Sample 3, by which a high-quality substrate can be obtained as a conventional method, in the obtained substrate, the above-mentioned low dislocation density region was not formed, and a 50 μm square dislocation-free region was not formed. For this reason, it is considered that a 50 μm square dislocation-free region is not formed even in a substrate produced by other conventional manufacturing methods that do not perform a special process for collecting dislocations.

[0383] (X-ray rocking curve measurement results) As shown in Table 1, in the substrate of Sample 3, due to the effect of thickening the crystal layer, the radius of curvature of the c-plane was slightly larger than that of the base substrate.

[0384] Also, as shown in Table 4, in the substrate of Sample 3 obtained by c-plane thick film growth, the variation in the off-angle of the c-axis within a diameter of 40 mm was improved compared to the substrate of Sample 4 and was about ±0.074°. Also, the FWHMb of the substrate of Sample 3 was improved compared to the FWHMb of the substrate of Sample 4.

[0385] However, on the substrate of Sample 3, a plurality of portions where FWHMb exceeded 32 arcsec were found. Also, on the substrate of Sample 3, (FWHMa - FWHMb) / FWHMa greatly exceeded 30% at all measurement points.

[0386] Thus, the substrate of Sample 3, which is a relatively high-quality substrate as a conventional substrate, had improved dislocation density and off-angle variation compared to the substrate of Sample 4, which is a comparable substrate. However, the substrate of Sample 3 did not have a single point that satisfied the condition of the full width at half maximum, i.e., FWHMb ≤ 32 arcsec and (FWHMa - FWHMb) / FWHMa ≤ 30%. It is considered that at least one of the above-mentioned crystal quality factors was not as good as those of the substrates of Samples 1 and 2 in the substrate of Sample 3.

[0387] Therefore, since even the substrate of Sample 3, which is a relatively high-quality substrate as a conventional substrate, does not satisfy the above-mentioned condition of the full width at half maximum, it is considered that the substrates fabricated by other conventional manufacturing methods also do not satisfy the above-mentioned condition of the full width at half maximum.

[0388] <Sample 2: 1 cycle> Referring to Table 1, Table 3, and FIG. 19, the results of the substrate of Sample 2 with one cycle will be described. FIG. 19 is a view of the main surface of the nitride semiconductor substrate of Sample 2 observed using a multiphoton excitation microscope.

[0389] (Dislocation) As shown in Table 1, in the substrate of Sample 2 with one cycle, the average dislocation density was reduced compared to those of the substrates of Samples 3 and 4. In the substrate of Sample 2, the ratio of the region where the dislocation density was less than 1×10 6 cm -2 was more than 90%.

[0390] However, in the substrate of Sample 2, a region where the dislocation density was 1×10 6 cm -2 or more existed about 5% of the main surface. Also, in the substrate of Sample 2, the dislocation density was 7×10 5 cm-2 The ratio of the area less than [a certain value] was less than 80%.

[0391] As shown in FIG. 19, the main surface of the substrate of Sample 2 included at least a dislocation-free region with a size of 50 μm square. Also, in the substrate of Sample 2, dislocation-free regions with a size of 50 μm square were scattered over the entire main surface.

[0392] Also, in the substrate of Sample 2, at least one dislocation-free region with a size of 50 μm square existed within all fields of view with a size of 250 μm square. The main surface of the substrate of Sample 2 had non-overlapping dislocation-free regions with a size of 50 μm square at a density of 1000 pieces / cm 2 or more.

[0393] However, there was no dislocation-free region with a size of 100 μm square within the main surface of the substrate of Sample 2.

[0394] (X-ray rocking curve measurement results) As shown in Table 1, for the substrate of Sample 2, the radius of curvature of the c-plane was larger than those of the substrates of Samples 3 and 4, and was 23 m or more.

[0395] Also, as shown in Table 3, for the substrate of Sample 2, the variation in the off-angle of the c-axis within a diameter of 40 mm was smaller than those of the substrates of Samples 3 and 4, and was about ±0.03°.

[0396] Also, for the substrate of Sample 2, at all measurement points, FWHMb was 32 arcsec or less. Also, for the substrate of Sample 2, at all measurement points, (FWHMa - FWHMb) / FWHMa was 30% or less.

[0397] <Sample 1: 2 cycles> Referring to Table 1, Table 2, FIG. 16, FIG. 17, FIG. 18, and FIG. 21(a), the results of the substrate of Sample 1 with two cycles will be described. FIG. 16 is a diagram showing an observation image obtained by observing a cross-section of the laminated structure of Sample 1 with a fluorescence microscope. FIGS. 17 and 18 are diagrams showing the main surface of the nitride semiconductor substrate of Sample 1 observed using a multi-photon excitation microscope. In FIGS. 17 and 18, the white solid-line square portions indicate dislocation-free regions with a side length of 50 μm, and the white dotted-line square portions indicate dislocation-free regions with a side length of 100 μm. FIG. 21(a) is a diagram showing a normalized X-ray diffraction pattern when X-ray rocking curve measurements of (0002) plane diffraction were performed on the nitride semiconductor substrate of Sample 1 with different incident-side slits. The measurement conditions for FIG. 21(a) are the same as those for FIG. 21(b) described above.

[0398] (Laminated structure) As shown in FIG. 16, the laminated structure of Sample 1 had two laminated units and a top low-oxygen concentration region based on the difference in oxygen concentration due to the difference in growth surfaces. Each of the two laminated units had a low-oxygen concentration region and a high-oxygen concentration region.

[0399] The low-oxygen concentration region was not continuous from the bottom substrate side to the top low-oxygen concentration region in the thickness direction.

[0400] The low-oxygen concentration region of the first laminated unit had a portion corresponding to an initial layer grown on the bottom substrate with the c-plane as the growth surface in the initial stage of growth. The initial layer was formed with a predetermined thickness over the entire main surface of the bottom substrate.

[0401] The low-oxygen concentration region of the first laminated unit had a plurality of valleys and a plurality of peaks in cross-sectional view. The average distance between the closest peaks in the first laminated unit was approximately 135 μm.

[0402] The high oxygen concentration region of the first stacked unit was provided above the low oxygen concentration region of the first stacked unit. The high oxygen concentration region of the first stacked unit was provided continuously along the main surface of the underlying substrate. That is, it was confirmed that the c-plane disappeared at least once during the growth process of the first stacked unit.

[0403] The low oxygen concentration region of the second stacked unit was provided above the high oxygen concentration region of the first stacked unit.

[0404] The low oxygen concentration region of the second stacked unit also had a plurality of valleys and a plurality of peaks in a cross-sectional view. The average distance between the closest peaks in the second stacked unit was longer than the average distance between the closest peaks in the first stacked unit and was approximately 209 μm.

[0405] The high oxygen concentration region of the second stacked unit was provided above the low oxygen concentration region of the second stacked unit. The high oxygen concentration region of the second stacked unit was provided continuously along the main surface of the underlying substrate. That is, it was also confirmed that the c-plane disappeared at least once during the growth process of the second stacked unit.

[0406] In the stacked structure of Sample 1, the thickness from the main surface of the underlying substrate to the interface of the second stacked unit was 1690 μm. From the results in Table 1, the reduction rate of the dislocation density obtained by the above N / N0 was 0.07.

[0407] (Dislocation) As shown in Table 1, in the substrate of Sample 1 with two cycles, the average dislocation density was reduced compared to those of the substrates of Samples 2 to 4.

[0408] In the substrate of Sample 1, the ratio of the region where the dislocation density was less than 1×10 6 cm -2 was 100%. That is, no region where the dislocation density was 1×10 6 cm -2 or more existed on the main surface.

[0409] Also, in the substrate of Sample 1, the ratio of the area where the dislocation density was less than 7×10 5 cm -2 was 80% or more.

[0410] As shown in FIGS. 17 and 18, the main surface of the substrate of Sample 1 included a dislocation-free region with a size of at least 50 μm square. Also, in the substrate of Sample 1, a plurality of 50-μm square dislocation-free regions were scattered over the entire main surface.

[0411] Also, in the substrate of Sample 1, a plurality of 50-μm square dislocation-free regions were present within all fields of view of 250 μm square. That is, the density of non-overlapping 50-μm square dislocation-free regions on the main surface of the substrate of Sample 1 was higher than that of Sample 2. Specifically, the main surface of the substrate of Sample 1 had a density of non-overlapping 50-μm square dislocation-free regions of 10,000 or more per cm 2 .

[0412] Also, in the substrate of Sample 1, the number of dislocations present within the quadrilateral region sandwiched between the opposing sides of the closest pair of dislocation-free regions was 90 or less.

[0413] Furthermore, as shown by the white dotted square portions in FIGS. 17 and 18, the main surface of the substrate of Sample 1 included a 100-μm square dislocation-free region. The main surface of the substrate of Sample 1 had a density of non-overlapping 100-μm square dislocation-free regions of 100 or more per cm 2 .

[0414] (X-ray rocking curve measurement results) As shown in Table 1, in the substrate of Sample 1, the radius of curvature of the c-plane was larger than those of the substrates of Samples 3 and 4, and was 23 m or more.

[0415] Also, as shown in Table 2, in the substrate of Sample 1, the variation in the off-angle of the c-axis within a diameter of 40 mm was smaller than those of the substrates of Samples 3 and 4, and was about ±0.006°.

[0416] Also, in the substrate of Sample 1, when the width of the incident-side slit in the ω direction was 0.1 mm, the FWHMb at all the measurement points was 32 arcsec or less.

[0417] As shown in Fig. 21(a), even when the width of the incident-side slit in the ω direction was increased from 0.1 mm to 1 mm in the substrate of Sample 1, the spread of the X-ray diffraction pattern was small.

[0418] As a result, as shown in Table 2, in the substrate of Sample 1, (FWHMa - FWHMb) / FWHMa was 30% or less at all the measurement points.

[0419] That is, in the X-ray rocking curve measurement of the (0002) plane diffraction in the substrate of Sample 1, results as good as or better than those of the substrate of Sample 2 were obtained.

[0420] (2) Experiment 2 (2-1) Preparation of nitride semiconductor substrate Samples 1, 3, and 4 in the above-mentioned Experiment 1 were prepared and the following evaluations were performed.

[0421] (2-2) Evaluation (X-ray rocking curve measurement of {10-12} plane diffraction) For each of the substrate of Sample 1, the substrate of Sample 3, and the substrate of Sample 4 corresponding to the base substrate, the following two types of X-ray rocking curve measurements of {10-12} plane diffraction were performed.

[0422] The same apparatus as in Experiment 1 was used for the X-ray rocking curve measurement. As described in the above embodiment, FWHM1 {10-12} was measured under the wide measurement region condition, and FWHM2 {10-12} was measured under the narrow measurement region condition.

[0423] FWHM1 {10-12}Under the incident conditions when measuring, from a Cu X-ray source, an X-ray mirror that makes the X-rays parallel light, a monochromator with two reflections of Ge(220), and a width d in the ω direction as the rotation angle direction around the rotation axis of the goniometer I is 1.4 mm and a length L in the direction parallel to the rotation axis I are 12 mm, the incident side aperture was irradiated with Cu Kα1 X-rays to the center of the main surface 100s of the substrate 100 through these in this order. That is, the measurement in Experiment 2 was performed with the incident side slit removed in Experiment 1. FWHM1 {10-12} Under the light receiving conditions when measuring, the light receiving side slit was opened and the X-rays were received without passing through the analyzer crystal. In the detector that receives the X-rays, the width d in the ω direction of the aperture (the width d in the ω direction of the light receiving side aperture) D , and the length of the aperture in the direction parallel to the rotation axis (the length of the light receiving side aperture) L D were both set to 14.025 mm. FWHM2 {10-12} The incident conditions when measuring FWHM2 {10-12} were the same as the incident conditions when measuring FWHM1. FWHM2 {10-12} Under the light receiving conditions when measuring, the X-rays were received through an analyzer crystal with three reflections of Ge(220). In the analyzer crystal, the width d in the ω direction of the entrance aperture (the width d in the ω direction of the light receiving side aperture) D , and the length of the entrance aperture in the direction parallel to the rotation axis (the length of the light receiving side aperture) L D were set to 6.54 mm and 14.025 mm, respectively.

[0424] (2-3) Results The results are shown in Table 6.

[0425]

Table 6

[0426] <Sample 4: Equivalent to the lower base plate> Referring to Table 6 and FIG. 24, the results of the substrate of Sample 4 corresponding to the underlying substrate will be described. FIG. 24 is a diagram showing the normalized X-ray diffraction pattern when X-ray rocking curve measurements of {10-12} plane diffraction are performed on the nitride semiconductor substrate of Sample 4 under different measurement conditions.

[0427] As shown in Table 6 and FIG. 24, for the substrate of Sample 4, FWHM1 {10-12} was large under the wide measurement region condition, exceeding 700 arcsec. Also, for the substrate of Sample 4, the variation in FWHM1 {10-12} measured from the diffraction of equivalent crystal planes in three directions was large, and the maximum-minimum difference exceeded 150 arcsec.

[0428] As shown in FIG. 24, for the substrate of Sample 4, the pattern of {10-12} diffraction under the wide measurement region condition was broader than the pattern of {10-12} diffraction under the narrow measurement region condition.

[0429] Therefore, as shown in Table 6, for the substrate of Sample 4, in each of the three directions, the ratio of FWHM2 {10-12} to FWHM1 {10-12} was less than 40%.

[0430] <Sample 3: c-plane thick film growth> Referring to Table 6 and FIG. 23, the results of the substrate of Sample 3 will be described. FIG. 23 is a diagram showing the normalized X-ray diffraction pattern when X-ray rocking curve measurements of {10-12} plane diffraction are performed on the nitride semiconductor substrate of Sample 3 under different measurement conditions.

[0431] As shown in Table 6 and FIG. 23, for the substrate of Sample 3, when measured from the direction 150° with respect to the a-axis, FWHM1 {10-12} was large under the wide measurement region condition, exceeding 70 arcsec. Also, for the substrate of Sample 3, there was a variation in FWHM1 {10-12} measured from the diffraction of equivalent crystal planes in three directions, and the maximum-minimum difference exceeded 20 arcsec.

[0432] As shown in FIG. 23, in the substrate of Sample 3, when measured from a direction of 150° with respect to the a-axis, the pattern of {10-12} diffraction under the wide measurement region condition was broader than the pattern of {10-12} diffraction under the narrow measurement region condition.

[0433] Therefore, as shown in Table 6, in the substrate of Sample 3, when measured from a direction of 150° with respect to the a-axis, FWHM1 {10-12} with respect to FWHM2 {10-12} was less than 80%.

[0434] <Sample 1: 2 cycles> Referring to Table 6 and FIG. 22, the results of the substrate of Sample 1 will be described. FIG. 22 is a diagram showing the normalized X-ray diffraction pattern when X-ray rocking curve measurements of {10-12} plane diffraction were performed on the nitride semiconductor substrate of Sample 1 under different measurement conditions.

[0435] As shown in Table 6 and FIG. 22, in the substrate of Sample 3, FWHM1 {10-12} under the wide measurement region condition was 40 arcsec or less. Also, in the substrate of Sample 3, the variation in FWHM1 {10-12} measured for diffraction of equivalent crystal planes from three directions was small, and the maximum-minimum difference was 9 arcsec or less.

[0436] As shown in FIG. 22, in the substrate of Sample 1, even under the wide measurement region condition, the broadening of the {10-12} diffraction pattern was small.

[0437] Therefore, as shown in Table 6, in the substrate of Sample 1, the ratio of FWHM1 {10-12} to FWHM2 {10-12} was 80% or more.

[0438] Also, in the substrate of Sample 1, since there were no defects on the main surface, the ratio of FWHM1 {10-12} to FWHM2 {10-12} was 100% or less.

[0439] (3) Summary of the results of Samples 1 and 2 According to Samples 1 and 2 above, in the three-dimensional growth process, the first growth conditions were adjusted to satisfy Equation (1). As a result, during the growth process of the three-dimensional growth layer, the c-plane could be surely eliminated. By surely eliminating the c-plane, at the position where the inclined interface in the three-dimensional growth layer was exposed, the dislocations could be surely bent. As a result, it was confirmed that the dislocation density on the main surface of the substrate could be efficiently reduced.

[0440] Furthermore, according to Sample 1, a cycle including a three-dimensional growth process and a planarization process was repeated a plurality of times. As a result, in the second cycle, the dislocations remaining in the planarization layer in the first cycle were further bent, and the dislocations could be locally collected. As a result, it was confirmed that the dislocation density of Sample 1 with two cycles could be reduced compared to the dislocation density of Sample 2 with one cycle.

[0441] Also, according to Sample 1, the average distance between the nearest ridges in the second stacking unit was longer than the average distance between the nearest ridges in the first stacking unit. As a result, in the second cycle, the distance by which the dislocations bent and propagated could be made longer than in the first cycle. As a result, it was confirmed that the density of the dislocation-free regions with a size of 50 μm square could be increased.

[0442] Also, according to Sample 1, the average distance between the nearest ridges in the second stacking unit was approximately 209 μm. As a result, in the second cycle, the distance by which the dislocations bent and propagated could be ensured to be at least more than 100 μm. As a result, it was confirmed that at least a part of the surface of the planarization layer in the second cycle could form a dislocation-free region with a size of at least 100 μm square. Also, the density of the non-overlapping dislocation-free regions with a size of 100 μm square on the main surface was 100 pieces / cm 2 It was confirmed that the above could be achieved.

[0443] According to Samples 1 and 2, due to the stress cancellation effect in the high oxygen concentration region, the radius of curvature of the c-plane of the substrate can be made larger than the radius of curvature of the c-plane of the underlying substrate, and it was confirmed that the variation in the off-angle of the c-axis in the substrate could be made smaller than the variation in the off-angle of the c-axis in the underlying substrate.

[0444] Also, according to Samples 1 and 2, not only could the dislocation density be reduced and the off-angle variation be made smaller, but all of the above-described crystal quality factors that determine the full width at half maximum could be made well-balanced and good.

[0445] As a result, regarding the X-ray rocking curve measurement of (0002) plane diffraction, it was confirmed that for the substrates of Samples 1 and 2, FWHMb could be made 32 arcsec or less. Furthermore, in Samples 1 and 2, even when the width of the incident-side slit in the ω direction was 1 mm, over the entire region irradiated with X-rays, due to the large radius of curvature of the c-plane and the well-balanced and good above-described crystal quality factors, it was confirmed that (FWHMa - FWHMb) / FWHMa could be made 30% or less.

[0446] Furthermore, regarding the X-ray rocking curve measurement of {10-12} plane diffraction, for the substrate of Sample 1, it was confirmed that FWHM1 {10-12} could be made 50 arcsec or less. Also, for the substrate of Sample 1, it was confirmed that the maximum-minimum difference in FWHM1 {10-12} measured by diffracting equivalent crystal planes represented by the {10-12} plane from three directions could be made 9 arcsec or less. Also, for the substrate 10 of Sample 1, it was confirmed that the ratio of FWHM2 {10-12} to FWHM1 {10-12} could be made 80% or more.

[0447] <Preferred Embodiment of the Present Invention> Hereinafter, the preferred embodiments of the present invention will be appended.

[0448] (Appended Note 1) A nitride semiconductor substrate having a main surface with a diameter of 2 inches or more and the closest low-index crystal plane being the (0001) plane, FWHM1 {10-12} The ratio to FWHM2 {10-12} is 80% or more Nitride semiconductor substrate. However, FWHM1 {10-12} and FWHM2 {10-12} are each the full width at half maximum of {10-12} plane diffraction measured by X-ray rocking curve measurement, FWHM1 {10-12} Under the incident conditions for measuring, from a Cu X-ray source, through an X-ray mirror that makes the X-rays parallel light, a Ge(220) double reflection monochromator, and an incident-side aperture with a width of 1.4 mm in the ω direction as the rotation angle direction around the rotation axis of the goniometer and a length of 12 mm in the direction parallel to the rotation axis in this order, irradiate the center of the main surface with Cu Kα1 X-rays, FWHM1 {10-12} Under the light-receiving conditions for measuring, with the light-receiving side slit open and without passing through an analyzer crystal, receive the X-rays with a detector having an aperture with a width of 14.025 mm in the ω direction, FWHM2 {10-12} The incident conditions for measuring are the same as the incident conditions for measuring FWHM1 {10-12} , FWHM2 {10-12} Under the light-receiving conditions for measuring, receive the X-rays with the detector through a Ge(220) triple reflection analyzer crystal having an entrance aperture with a width of 6.54 mm in the ω direction.

[0449] (Appendix 2) FWHM1 {10-12} The ratio to FWHM2 {10-12} is 100% or less The nitride semiconductor substrate according to Appendix 1.

[0450] (Appendix 3) The FWHM1 obtained by measuring the diffraction of the equivalent crystal plane represented by the {10-12} plane from three directions rotated by 60° each in the circumferential direction about the normal line of the center of the main surface {10-12} has a maximum-minimum difference of 9 arcsec or less The nitride semiconductor substrate according to Appendix 1 or 2

[0451] (Appendix 4) A nitride semiconductor substrate having a main surface with a diameter of 2 inches or more and the closest low-index crystal plane being the (0001) plane, The FWHM1 obtained by measuring the diffraction of the equivalent crystal plane represented by the {10-12} plane from three directions rotated by 60° each in the circumferential direction about the normal line of the center of the main surface {10-12} has a maximum-minimum difference of 9 arcsec or less Nitride semiconductor substrate However, FWHM1 {10-12} is the half-value width of the {10-12} plane diffraction measured by X-ray rocking curve measurement, FWHM1 {10-12} Under the incident conditions for measuring FWHM1, X-rays from a Cu X-ray source are irradiated onto the center of the main surface through an X-ray mirror that makes the X-rays parallel light, a Ge(220) double-reflection monochromator, and an incident-side aperture with a width of 1.4 mm in the ω direction as the rotation angle direction around the rotation axis of the goniometer and a length of 12 mm in the direction parallel to the rotation axis in this order, FWHM1 {10-12} Under the light-receiving conditions for measuring FWHM1, the X-rays are received by a detector having an aperture with a width of 14.025 mm in the ω direction with the light-receiving side slit open and without passing through an analyzer crystal

[0452] (Appendix 5) FWHM1 {10-12} is 50 arcsec or less The nitride semiconductor substrate according to any one of Appendices 1 to 4

[0453] (Appendix 6) A nitride semiconductor substrate having a main surface with a diameter of 2 inches or more and the nearest low-index crystal plane being the (0001) plane, FWHM1 {10-12} is 50 arcsec or less nitride semiconductor substrate. However, FWHM1 {10-12} is the full width at half maximum of the {10-12} plane diffraction measured by X-ray rocking curve measurement, FWHM1 {10-12} Under the incident conditions for measuring FWHM1, X-rays are irradiated onto the center of the main surface through, in this order, an X-ray mirror that makes the X-rays parallel light from a Cu X-ray source, a Ge(220) double-reflection monochromator, and an incident-side aperture having a width of 1.4 mm in the ω direction as the rotation angle direction around the rotation axis of the goniometer and a length of 12 mm in the direction parallel to the rotation axis, with Cu Kα1 X-rays. FWHM1 {10-12} Under the light-receiving conditions for measuring FWHM1, the X-rays are received by a detector having an aperture with a width of 14.025 mm in the ω direction with the light-receiving side slit open and without passing through an analyzer crystal.

[0454] (Appendix 7) When the main surface of the nitride semiconductor substrate is observed with a multi-photon excitation microscope at a field of view of 250 μm angle and the dislocation density is determined from the dark spot density, there is no region on the main surface where the dislocation density is 1×10 6 cm -2 or more, and a region where the dislocation density is less than 7×10 5 cm -2 exists in 80% or more of the main surface, The main surface has non-overlapping dislocation-free regions of 50 μm angle at a density of 1000 pieces / cm 2 or more The nitride semiconductor substrate according to any one of Appendices 1 to 6.

[0455] (Appendix 8) A nitride semiconductor substrate having a main surface with a diameter of 2 inches or more and the nearest low-index crystal plane being the (0001) plane, When observing the main surface of the nitride semiconductor substrate at a field of view of 250 μm square by a multiphoton excitation microscope and obtaining the dislocation density from the dark spot density, there is no region on the main surface where the dislocation density is 1×10 6 cm -2 or more, and a region where the dislocation density is less than 7×10 5 cm -2 exists in 80% or more of the main surface. The main surface has non-overlapping dislocation-free regions of 50 μm square at a density of 1000 / cm 2 or more. Nitride semiconductor substrate.

[0456] (Appendix 9) The number of dislocations existing within the rectangular region sandwiched between the opposing sides of the closest pair of the dislocation-free regions is 90 or less. The nitride semiconductor substrate according to Appendix 7 or 8.

[0457] (Appendix 10) The main surface has non-overlapping dislocation-free regions of 100 μm square at a density of 100 / cm 2 or more. The nitride semiconductor substrate according to any one of Appendices 7 to 9.

[0458] (Appendix 11) When irradiating the main surface with Cu Kα1 X-rays through a two-crystal monochromator of the Ge(220) plane and an incident slit, and performing X-ray rocking curve measurement of (0002) plane diffraction, when the width in the ω direction of the incident slit is 0.1 mm, the full width at half maximum FWHMb of the (0002) plane diffraction is 32 arcsec or less, and the difference FWHMa - FWHMb obtained by subtracting FWHMb from the full width at half maximum FWHMa of the (0002) plane diffraction when the width in the ω direction of the incident slit is 1 mm is 30% or less of FWHMa. The nitride semiconductor substrate according to any one of Appendices 1 to 10.

[0459] (Appendix 12) A nitride semiconductor substrate having a main surface with a diameter of 2 inches or more and the nearest low-index crystal plane being the (0001) plane, When irradiating the main surface with X-rays of Cu Kα1 through a two-crystal monochromator of the Ge(220) plane and an incident-side slit and performing X-ray rocking curve measurement of (0002) plane diffraction, When the width of the incident-side slit in the ω direction is 0.1 mm, the full width at half maximum FWHMb of the (0002) plane diffraction is 32 arcsec or less, The difference FWHMa - FWHMb obtained by subtracting FWHMb from the full width at half maximum FWHMa of the (0002) plane diffraction when the width of the incident-side slit in the ω direction is 1 mm is 30% or less of FWHMa A nitride semiconductor substrate.

[0460] (Appendix 13) FWHMa - FWHMb is 30% or less of FWHMa at a plurality of measurement points set at predetermined intervals in the <1-100> axis direction and the <11-20> axis direction within the main surface. The nitride semiconductor substrate according to Appendix 11 or 12.

[0461] (Appendix 14) When performing X-ray rocking curve measurement of the (0002) plane diffraction with the width of the incident-side slit in the ω direction being 0.1 mm at a plurality of measurement points set at 5 mm intervals within the main surface, in 90% or more of all the measurement points, the full width at half maximum FWHMb of the (0002) plane diffraction is 32 arcsec or less. The nitride semiconductor substrate according to any one of Appendices 11 to 13.

[0462] (Appendix 15) A base substrate made of a single crystal of a group III nitride semiconductor, having a polished main surface, and the nearest low-index crystal plane to the main surface being the (0001) plane, A stacked unit provided above the main surface of the base substrate, having a low oxygen concentration region made of a single crystal of a group III nitride semiconductor and a high oxygen concentration region made of a single crystal of a group III nitride semiconductor provided on the low oxygen concentration region. Provided above the lamination unit, there is a top low-oxygen concentration region made of a single crystal of a group-III nitride semiconductor, and it includes the oxygen concentration in the high-oxygen concentration region is higher than the oxygen concentration in each of the low-oxygen concentration region and the top low-oxygen concentration region, a plurality of the lamination units are repeatedly provided in the thickness direction between the lower base substrate and the top low-oxygen concentration region a laminated structure.

[0463] (Supplementary Note 16) The upper surface of the low-oxygen concentration region in each of the plurality of lamination units has a plurality of valleys and a plurality of peaks, in the second and higher lamination units among the plurality of lamination units, when looking at an arbitrary cross section perpendicular to the main surface, the average distance by which a pair of the closest peaks among the plurality of peaks are separated in the direction along the main surface across one of the plurality of valleys is longer than that in the first lamination unit among the plurality of lamination units The laminated structure according to Supplementary Note 15.

[0464] (Supplementary Note 17) In each of the plurality of lamination units, the average distance between the pair of the closest peaks gradually becomes longer as going to the upper layer of the plurality of lamination units The laminated structure according to Supplementary Note 15 or 16.

[0465] As an aspect of the manufacturing method of the present invention, for example, the following supplementary notes are cited.

[0466] (Supplementary Note 18) A method for manufacturing a nitride semiconductor substrate using a vapor phase growth method, comprising: (a) preparing a lower base substrate made of a single crystal of a group-III nitride semiconductor, having a polished main surface, and the low-index crystal plane closest to the main surface being the (0001) plane; (b) Growing a single crystal of a group III nitride semiconductor having a top surface with the (0001) plane exposed epitaxially above the main surface of the lower base substrate, forming a plurality of recesses composed of inclined interfaces other than the (0001) plane on the top surface, gradually expanding the inclined interface as going above the main surface of the lower base substrate, disappearing the (0001) plane from the top surface at least once, and growing a three-dimensional growth layer; (c) Growing a single crystal of a group III nitride semiconductor epitaxially on the three-dimensional growth layer, disappearing the inclined interface, and growing a planarization layer having a mirror-finished surface; having performing the cycle including the above (b) and (c) a plurality of times A method for manufacturing a nitride semiconductor substrate.

[0467] (Appendix 19) In the above (b) in each of the plurality of cycles, forming the plurality of recesses on the top surface of the single crystal and disappearing the (0001) plane, thereby forming a plurality of valleys and a plurality of tops on the surface of the three-dimensional growth layer; In the above (b) in the cycles after the second cycle among the plurality of cycles, when looking at an arbitrary cross section perpendicular to the main surface, making the average distance between the pair of closest tops among the plurality of tops across one of the plurality of valleys be longer in the direction along the main surface than that in the above (b) of the first cycle among the plurality of cycles The method for manufacturing a nitride semiconductor substrate according to Appendix 18.

[0468] (Appendix 20) In the above (b) in each of the plurality of cycles, gradually increasing the average distance between the pair of closest tops as repeating the plurality of cycles The method for manufacturing a nitride semiconductor substrate according to Appendix 19.

[0469] (Appendix 21) In each of the plurality of cycles, in the above (b), the average distance between the pair of closest tops is set to be more than 100 μm The method for manufacturing a nitride semiconductor substrate according to Appendix 19 or 20.

[0470] (Appendix 22) In the above (b) of at least the last cycle among the plurality of cycles, the average distance between the pair of closest tops is set to be more than 200 μm The method for manufacturing a nitride semiconductor substrate according to any one of Appendices 19 to 21.

[0471] (Appendix 23) In each of the plurality of cycles, in the above (b), the average distance between the pair of closest tops is set to be less than 800 μm The method for manufacturing a nitride semiconductor substrate according to claim 21 or 22.

[0472] (Appendix 24) In the above (a), the root mean square roughness of the main surface of the underlying substrate is set to be 1 nm or more The method for manufacturing a nitride semiconductor substrate according to any one of Appendices 18 to 23.

[0473] (Appendix 25) In the above (a), the crystal strain introduced by the processing of the underlying substrate is left on the main surface side of the underlying substrate, when X-ray rocking curve measurement is performed with the incident angle with respect to the main surface of the processed underlying substrate being 2°, the half-value width of the (10 - 10) plane diffraction is made larger than the half-value width of the underlying substrate before processing and is set to be 60 arcsec or more and 200 arcsec or less The method for manufacturing a nitride semiconductor substrate according to any one of Appendices 18 to 24.

[0474] (Appendix 26) In each of the plurality of cycles, in the above (b), After disappearing the (0001) plane from the surface, while maintaining the state where the inclined interface occupies more of the surface than the (0001) plane, the growth of the three-dimensional growth layer is continued over a predetermined thickness. The method for manufacturing a nitride semiconductor substrate according to any one of Appendices 18 to 25.

[0475] (Appendix 27) (d) After the (c) of the last cycle among the plurality of cycles, a step of forming a main growth layer over a predetermined thickness on the planarized layer whose surface has been mirror-finished, with the (0001) plane as the growth surface; (e) A step of slicing at least one nitride semiconductor substrate from the main growth layer; further comprising The method for manufacturing a nitride semiconductor substrate according to any one of Appendices 18 to 26.

[0476] (Appendix 28) In the (a), preparing the underlying substrate in which the (0001) plane is curved in a concave spherical shape with respect to the main surface; In the (e), making the radius of curvature of the (0001) plane of the nitride semiconductor substrate larger than the radius of curvature of the (0001) plane of the underlying substrate. The method for manufacturing a nitride semiconductor substrate according to Appendix 27.

[0477] (Appendix 29) In the (b) in each of the plurality of cycles, as the inclined interface, a {11 - 2m} plane where m ≥ 3 is generated. The method for manufacturing a nitride semiconductor substrate according to any one of Appendices 18 to 28.

[0478] Also, as another aspect of the manufacturing method of the present invention, for example, the following appendices may be applicable.

[0479] (Appendix 30) A method for manufacturing a nitride semiconductor substrate using a vapor phase growth method, (a) Prepare a substrate made of a single crystal of a group III nitride semiconductor, having a polished main surface, and the low-index crystal plane closest to the main surface is the (0001) plane. (b) Flatly homoepitaxially grow a single crystal of a group III nitride semiconductor above the main surface. (c) Generate a plurality of recesses formed by inclined interfaces other than the (0001) plane on the surface of the flat homoepitaxial growth crystal, and gradually expand the inclined interfaces as crystal growth progresses, so that the (0001) plane disappears from the crystal growth interface at least once, and grow a three-dimensional growth layer. (d) Further epitaxially grow a single crystal of a group III nitride semiconductor on the three-dimensional growth layer to eliminate the inclined interface and grow a planarization layer having a polished surface composed of the (0001) plane. Having Perform the cycle including the above (c) and (d) multiple times. A method for manufacturing a nitride semiconductor substrate.

[0480] (Appendix 31) In the above (c) of the second and subsequent cycles among the plurality of cycles, When looking at an arbitrary cross section perpendicular to the main surface when the (0001) plane disappears from the crystal growth interface, among the plurality of valleys and the plurality of peaks existing on the surface of the three-dimensional growth layer, the average distance between the closest pair of peaks sandwiching one valley is longer than the average distance of the above (c) of the previous cycle in the direction along the main surface. The method for manufacturing a nitride semiconductor substrate according to Appendix 30.

[0481] Note that the above Appendices 20 to 29 can be made subordinate to the above Appendix 30 or 31 by replacing (b) with (c), (c) with (d), (d) with (e), and (e) with (f).

Explanation of symbols

[0482] 10 Substrate 30 First layer 40 Second layer 50 Nitride semiconductor substrate (substrate)

Claims

1. A bottom substrate made of a single crystal of a group III nitride semiconductor, having a polished main surface, wherein the low-index crystal plane closest to the main surface is the (0001) plane; A stacked unit provided above the main surface of the bottom substrate, comprising a low-oxygen concentration region made of a single crystal of a group III nitride semiconductor, and a high-oxygen concentration region made of a single crystal of a group III nitride semiconductor provided on the low-oxygen concentration region; An uppermost low-oxygen concentration region made of a single crystal of a group III nitride semiconductor, provided above the stacked unit; Comprising: The oxygen concentration in the high-oxygen concentration region is higher than the oxygen concentrations in the low-oxygen concentration region and the uppermost low-oxygen concentration region, respectively; A plurality of the stacked units are repeatedly provided in the thickness direction between the bottom substrate and the uppermost low-oxygen concentration region; When any cross section perpendicular to the main surface is viewed, the upper surface of the low-oxygen concentration region in each of the stacked units has a plurality of valleys and a plurality of peaks; In the second and higher stacked units of the stacked units, When any cross section perpendicular to the main surface is viewed, the average distance by which a pair of the closest peaks among the plurality of peaks, sandwiching one of the plurality of valleys, are separated in the direction along the main surface is longer than that in the first stacked unit of the stacked units; Stacked structure.

2. A method for manufacturing a nitride semiconductor substrate using a vapor phase growth method, comprising: (a) preparing a bottom substrate made of a single crystal of a group III nitride semiconductor, having a polished main surface, wherein the low-index crystal plane closest to the main surface is the (0001) plane; (b) epitaxially growing a single crystal of a group III nitride semiconductor flat above the main surface; (c) forming a plurality of recesses composed of inclined interfaces other than the (0001) plane on the surface of the flat homoepitaxial growth crystal, and gradually expanding the inclined interfaces as crystal growth progresses to cause the (0001) plane to disappear from the crystal growth interface at least once, thereby growing a three-dimensional growth layer; (d) further epitaxially growing a single crystal of a group III nitride semiconductor on the three-dimensional growth layer to eliminate the inclined interfaces and grow a planarization layer having a polished surface composed of the (0001) plane; Having: Performing a plurality of cycles including the steps (c) and (d); In the step (c) in each of the cycles, By forming the plurality of concave portions and eliminating the (0001) plane, when any cross-section perpendicular to the main surface is viewed, a plurality of valleys and a plurality of peaks are formed on the surface of the three-dimensional growth layer. In the (c) of the cycles after the second cycle among the cycles, When any cross-section perpendicular to the main surface is viewed, the average distance by which the pair of closest peaks among the plurality of peaks are separated in the direction along the main surface across one of the plurality of valleys is made longer than that in the (c) of the first cycle among the cycles. A method for manufacturing a nitride semiconductor substrate.

3. In the (b), Above the main surface, the single crystal is homoepitaxially grown with the (0001) plane as the growth surface over the entire main surface. The method for manufacturing a nitride semiconductor substrate according to claim 2.

Images