Preparation method of epitaxial structure, epitaxial structure and semiconductor device

Abstract

The invention provides a preparation method of an epitaxial structure, the epitaxial structure and a semiconductor device, and the preparation method comprises the steps: placing one surface of a substrate on a graphite disc, and sequentially preparing a nucleating layer, a buffer layer, a high-resistance layer, a channel layer, an insertion layer and a barrier layer on the other surface of the substrate; a convex surface and a concave surface are arranged on the surface of the graphite plate, the concave surface is arranged in the middle area of the graphite plate, and the convex surface is arranged on at least one side of the concave surface; the buffer layer comprises at least one of a multi-layer AlGaN structure and a multi-layer superlattice structure, the temperature distribution of the epitaxial structure is optimized, the thickness of the epitaxial structure is changed and the component distribution trend of the epitaxial structure is adjusted by changing the distance from the upper surface of the graphite disc to the lower surface of the epitaxial structure by using the graphite disc with a convex surface and a concave surface, and the performance of the epitaxial structure is improved. And through the arrangement of the buffer layer structure, the epitaxial structure with excellent thickness uniformity and component uniformity is prepared.

Description

Technical Field

[0001] This application relates to the field of semiconductor technology, specifically to a method for preparing an epitaxial structure, the epitaxial structure, and a semiconductor device. Background Technology

[0002] Gallium nitride (GaN) epitaxial wafers are the core material for GaN power devices, and their quality directly determines the device's performance, reliability, and cost. With the surge in demand for high-efficiency, low-power semiconductor devices in consumer electronics and new energy vehicles, the market for GaN power devices and chips is expanding rapidly, driving epitaxial wafer manufacturing technology towards high throughput and low cost. Against this backdrop, how to achieve high-quality, uniform fabrication of large-size epitaxial wafers through optimizing epitaxial growth processes and hardware design has become a key technical challenge that the industry urgently needs to solve.

[0003] Currently, the large-scale production of gallium nitride epitaxial wafers mainly relies on metal-organic chemical vapor deposition (MOCVD) technology, with the core objective of shortening the growth cycle and achieving larger sizes (such as upgrading from 6 inches to 8 inches). Larger sizes offer significant advantages in terms of capacity and cost: the theoretical device yield of 8-inch wafers is approximately 78% higher than that of 6-inch wafers, and the manufacturing cost per chip can be reduced by 20% to 30%. However, while existing horizontal planetary MOCVD equipment (such as the Aixtron G5 series) performs excellently in the fabrication of 6-inch epitaxial wafers, achieving high uniformity in thickness and composition, it faces significant challenges in the growth of 8-inch epitaxial wafers. For example, the edge warping of 8-inch wafers is exacerbated, leading to localized temperature imbalances, which in turn cause problems such as "M-shaped" thickness distribution and "W-shaped" composition distribution. Summary of the Invention

[0004] This application provides a method for preparing an epitaxial structure, an epitaxial structure, and a semiconductor device to solve the problems of current epitaxial wafers exhibiting an "M-shaped" thickness distribution and a "W-shaped" composition distribution.

[0005] In a first aspect, this application provides a method for preparing an epitaxial structure, comprising: placing one side of a substrate on a graphite disk, and sequentially preparing a nucleation layer, a buffer layer, a high-resistivity layer, a channel layer, an insertion layer and a barrier layer on the other side of the substrate; The graphite disk has a convex surface and a concave surface, the concave surface is located in the middle region of the graphite disk, and the convex surface is located on at least one side of the concave surface; The buffer layer includes at least one of a multilayer AlGaN structure and a multilayer superlattice structure.

[0006] Optionally, the convex surface satisfies at least one of the following conditions: (1) The convex surface is annularly raised, and the inner ring of the annularly raised surface is located on the outer periphery of the concave surface; (2) The protrusion of the convex surface is 30μm~100μm.

[0007] Optionally, the concave surface satisfies at least one of the following conditions: (1) The concave surface is spherical or ellipsoidal in shape; (2) The indentation of the concave surface is 15μm~60μm.

[0008] Optionally, the concave surface satisfies at least one of the following conditions: (1) The graphite disk is made of graphite; (2) The surface of the graphite disk is provided with a SiC coating, and the thickness of the SiC coating is 50nm~150nm.

[0009] Optionally, the buffer layer satisfies at least one of the following conditions: (1) The multilayer AlGaN structure is configured as at least two layers of Al stacked sequentially. X Ga 1-X N-structure, where 0.1 ≤ X ≤ 0.9; (2) Along the direction from the substrate to the high-resistivity layer, the aluminum content in the multilayer AlGaN structure decreases. (3) The multilayer superlattice structure is configured as a stacked y-layer superlattice structure, where y≥2; (4) The material of the superlattice structure is at least one of AlN / AlGaN, AlGaN / AlGaN, and AlGaN / GaN; (5) Along the direction from the substrate to the high-resistivity layer, the aluminum content of the multilayer superlattice structure decreases.

[0010] Optionally, the prepared epitaxial structure satisfies at least one of the following conditions: (1) The thickness of the substrate is 0.7 mm to 1.2 mm; (2) The nucleation layer material is AlN, and the thickness is 150nm~300nm; (3) The thickness of the buffer layer is 2000nm~3500nm; (4) The high-resistivity layer material is carbon-doped GaN with a thickness of 1000nm~1800nm; (5) The channel layer material is intrinsic GaN with a thickness of 150nm~500nm; (6) The insertion layer material is AlN, and the thickness is 0.5nm~1.5nm; (7) The barrier layer material is AlGaN with a thickness of 10nm~20nm.

[0011] Optionally, the preparation method satisfies at least one of the following conditions: (1) The conditions for forming the buffer layer are: The reaction temperature was 1000~1040℃, the reaction pressure was 50~60mbar, the hydrogen flow rate was 80~120slm, the ammonia flow rate was 3~10slm, the trimethylgallium flow rate was 30~300sccm, the trimethylaluminum flow rate was 300~800sccm, the ethylene flow rate was 0~400sccm, and the total growth time was 60~200min. (2) The conditions for forming the high-resistivity layer are: The reaction temperature was 980~1020℃, the reaction pressure was 60~200mbar, the hydrogen flow rate was 80~130slm, the ammonia flow rate was 10~30slm, the trimethylgallium flow rate was 250~400sccm, the ethylene flow rate was 150~400sccm, and the growth time was 20~50min. (3) The conditions for forming the channel layer are: The reaction temperature was 1030~1060℃, the reaction pressure was 200~400mbar, the hydrogen flow rate was 90~110slm, the ammonia flow rate was 30~45slm, the trimethylgallium flow rate was 100~300sccm, and the growth time was 15~40min. (4) The conditions for forming the insertion layer are: The reaction temperature was 1030~1060℃, the reaction pressure was 50~150mbar, the hydrogen flow rate was 80~110slm, the ammonia flow rate was 10~35slm, the trimethylaluminum flow rate was 40~120sccm, and the growth time was 0.5~2min. (5) The conditions for the formation of the barrier layer are: The reaction temperature was 1030~1060℃, the reaction pressure was 50~150mbar, the hydrogen flow rate was 80~110slm, the ammonia flow rate was 10~35slm, the trimethylgallium flow rate was 30~60sccm, the trimethylaluminum flow rate was 40~120sccm, and the growth time was 2~5min.

[0012] According to a second aspect of this application, an epitaxial structure is provided, comprising the epitaxial structure prepared by the preparation method described in the first aspect above.

[0013] Optionally, the epitaxial structure is 8 inches in size.

[0014] Based on the above technical solutions According to a third aspect of this application, a semiconductor device is provided, comprising an epitaxial structure prepared by the preparation method described in the first aspect, or an epitaxial structure described in the second aspect.

[0015] The beneficial effects of this application are as follows: Based on the above technical solution and the above preparation method, a graphite disk with convex and concave structures is used to optimize the temperature distribution of the epitaxial structure by changing the distance from the upper surface of the graphite disk to the lower surface of the epitaxial structure. This achieves temperature field supplementation for the nucleation layer, buffer layer, high-resistivity layer, channel layer, insertion layer and barrier layer during the growth process, thereby changing the thickness of the epitaxial structure and adjusting the component distribution trend of the epitaxial structure. Furthermore, by setting the buffer layer structure, an epitaxial structure with excellent thickness uniformity and component uniformity is prepared.

[0016] Other features and advantages of this application will be described in detail in the following detailed description section. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a process flow diagram of an epitaxial structure preparation method provided in a specific embodiment of this application; Figure 2 A schematic diagram of the extensional structure provided for a specific embodiment of this application; Figure 3 A schematic diagram of the graphite disk provided for a specific embodiment of this application; Figure 4 The total thickness distribution curves of the epitaxial structures prepared in Examples 1-1 and 2-1 provided for specific embodiments of this application; Figure 5 Distribution curves of Al mass content in the barrier layer of the epitaxial structures prepared in Examples 1-1 and Comparative Examples 2-1 for specific embodiments of this application; Wherein: 00, substrate; 10, nucleation layer; 20, buffer layer; 30, high-resistivity layer; 40, channel layer; 50, insertion layer; 60, barrier layer; 70. Graphite disk; 701. Concave surface; 702. Convex surface; 703. Outer wall. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0020] The inventors' current research has revealed significant challenges in the growth of 8-inch epitaxial wafers. For example, increased edge warping of 8-inch epitaxial wafers leads to imbalances in local temperature distribution, resulting in problems such as an "M-shaped" thickness distribution and a "W-shaped" composition distribution.

[0021] Therefore, in a first aspect, this application provides a method for preparing an epitaxial structure based on graphite disk growth, comprising: placing one side of a substrate on a graphite disk, and sequentially preparing a nucleation layer, a buffer layer, a high-resistivity layer, a channel layer, an insertion layer and a barrier layer on the other side of the substrate; The graphite disk has a convex surface and a concave surface, the concave surface is located in the middle region of the graphite disk, and the convex surface is located on at least one side of the concave surface; The buffer layer includes at least one of a multilayer AlGaN structure and a multilayer superlattice structure.

[0022] Through the above technical solution, the method for preparing epitaxial structures based on graphite disk growth utilizes graphite disks with convex and concave structures. By changing the distance from the upper surface of the graphite disk to the lower surface of the epitaxial structure, the temperature distribution of the epitaxial structure is optimized. This achieves temperature field supplementation for the nucleation layer, buffer layer, high-resistivity layer, channel layer, insertion layer, and barrier layer during the growth process, thereby changing the thickness of the epitaxial structure and adjusting the component distribution trend of the epitaxial structure. Furthermore, by setting the buffer layer structure, an epitaxial structure with excellent thickness uniformity and component uniformity is prepared.

[0023] The inventors speculate that the temperature control mechanism of the graphite disk is as follows: The concave surface in the middle region increases the distance between the graphite disk and the epitaxial structure, reducing heat transfer and preventing excessive temperature. Meanwhile, the convex surface on at least one side of the concave surface reduces the distance between the graphite disk and the epitaxial structure in the corresponding edge region, increasing heat transfer to compensate for the lower temperature caused by warping at the edge. Through this differentiated heat regulation, the temperature field during the epitaxial wafer growth process is balanced, thereby adjusting the growth rate and composition distribution trend, and solving the problem of thickness “M-shaped” and composition “W-shaped” distribution caused by temperature field imbalance.

[0024] Therefore, by using the above-mentioned temperature control mechanism, the temperature distribution of the epitaxial wafer is optimized by changing the distance from the upper surface of the graphite disk to the lower surface of the epitaxial wafer, thereby changing the growth rate distribution and composition distribution. The original "W-shaped" aluminum composition distribution (high at the edges, low and fluctuating in the center) caused by temperature field imbalance is adjusted to a "one-line" distribution; the original "M-shaped" thickness distribution (thick at the center and edges, thin in the middle annular region) caused by temperature field imbalance is transformed into a "one-line" distribution, thus solving the problem of temperature field imbalance caused by edge warping during the growth of 8-inch epitaxial wafers.

[0025] Meanwhile, the buffer layer structure (multilayer AlGaN and / or superlattice structure) further improves the crystallization quality and stress management capability of the material. Through the synergistic effect of the structural optimization of the graphite disk and the setting of the buffer layer, the thickness uniformity and composition uniformity of the entire epitaxial structure are significantly improved.

[0026] For example, such as Figure 1 As shown, a method for preparing an epitaxial structure based on graphite disk growth includes the following steps: Provides gallium nitride epitaxial graphite disks; An eight-inch single-crystal silicon substrate is provided on the graphite disk; A nucleation layer is formed on one side of an eight-inch single-crystal silicon substrate using in-situ MOCVD. A buffer layer is formed on the side of the nucleation layer away from the silicon substrate by in-situ MOCVD. A high-resistivity layer is formed on the side of the buffer layer away from the nucleation layer by in-situ MOCVD. On the side of the high-resistivity layer away from the buffer layer, a channel layer is formed by in-situ MOCVD. An insertion layer is formed on the side of the channel layer away from the high-resistivity layer by in-situ MOCVD. A barrier layer is formed on the side of the insertion layer away from the channel layer using in-situ MOCVD.

[0027] Understandably, in-situ MOCVD involves introducing organometallic compounds (as metal source precursors) and other reactive gases (such as group V hydrides in group III-V compounds) into a reaction chamber under a carrier gas (usually hydrogen, nitrogen, etc.). Under certain process conditions such as temperature and pressure, these gases undergo thermal decomposition or chemical reaction on the substrate surface. The resulting solid products are deposited on the substrate surface, gradually forming thin films or epitaxial layers with specific structures and properties, while the gaseous byproducts generated by the reaction are discharged from the reaction chamber.

[0028] Therefore, MOCVD, which utilizes gas-phase chemical reactions to achieve controllable growth of materials on the substrate surface, can precisely control the composition, thickness, and crystal quality of the deposited layer by adjusting parameters such as reaction temperature, pressure, gas flow rate, and precursor concentration. It is suitable for preparing various semiconductor thin film materials (such as gallium nitride, gallium arsenide, etc.) and related device structures.

[0029] In some embodiments, such as Figure 3 As shown, a graphite disk 70 is provided with an annular outer wall 703. A concave surface 701 and a convex surface 702 are provided on the graphite disk surface inside the outer wall 703. The concave surface 701 is located in the central region of the graphite disk, and the convex surface 702 is continuously and smoothly connected to the outer wall 703. The concave surface 701 is located between the convex surface 702 and the outer wall 703. When the graphite disk is used, the substrate is fixed on the graphite disk surface inside the outer wall 703.

[0030] In some examples, the convex surface is shaped like a ring-shaped ridge, with the inner ring of the ring-shaped ridge located on the outer periphery of the concave surface. This matches the circular structure of the 8-inch epitaxial wafer, enabling the distance between the edge region of the epitaxial wafer (corresponding to the convex surface position) and the graphite disk to be distributed in a ring-shaped symmetrical manner. This avoids the local distance differences that may be caused by non-ring-shaped structures, thereby making the heat transfer at the edge more uniform, enhancing the ring-shaped symmetry of the temperature field, compensating for the problem of lower temperature caused by edge warping, and enhancing the correction effect on the "M-shaped" thickness distribution and "W-shaped" composition distribution, making the thickness and composition closer to the "I-shaped". At the same time, it adapts to the growth area of ​​the circular epitaxial wafer, reduces local stress concentration, and works synergistically with the buffer layer to improve the crystal quality.

[0031] For example, the annular bulge can take the form of a continuous closed ring, a multi-segment arc with uniform spacing (such as 3-4 segments of equal arc), or a double-layered ring with nested inner and outer rings.

[0032] In some examples, the protrusion of the convex surface is 30μm to 100μm, and the chord length of the convex surface is 40mm to 90mm.

[0033] Understandably, the convexity of a convex surface refers to the vertical height difference between the apex of the annular bulge and the reference plane of the graphite disk, that is, the height by which the convex surface protrudes outward relative to the reference plane of the graphite disk; the chord length of a convex surface refers to the distance between the endpoints of the different sides of the annular bulge that begin to bulge, that is, the straight-line distance between the starting points of two opposite bulges on the edge of the convex surface, that is, the chord length of the convex surface.

[0034] Within the convexity range, by controlling the vertical height difference between the apex of the annular raised surface and the reference plane of the graphite disk, the distance between the graphite disk and the edge region of the epitaxial wafer can be kept within a reasonable range. This avoids insufficient heat transfer and inability to effectively compensate for the low temperature caused by edge warping when the convexity is less than 30 μm, and also prevents excessive local heat transfer and new temperature field imbalance when the convexity is greater than 100 μm. The resulting gradient heat transfer can synergize with the heat distribution in the central concave region, accurately balance the temperature field of epitaxial wafer growth, and effectively correct the thickness "M-type" and composition "W-type" distribution, improve the uniformity of the epitaxial structure, and adapt to the stress management requirements of the buffer layer, reducing defects caused by temperature unevenness.

[0035] Understandably, the chord length of the convex surface is 40mm~90mm, ensuring a precise match between the coverage area of ​​the convex surface and the area of ​​temperature imbalance caused by warping at the edge of the epitaxial wafer. Within this range, the chord length can cover the core area at the edge of the epitaxial wafer that requires temperature compensation, ensuring that the heat transferred by the annular convex surface can effectively offset the low temperature caused by edge warping, while not extending excessively into non-edge areas, thus avoiding interference with the temperature field regulation of the central concave surface. In this way, it works in conjunction with the convexity of the convex surface to construct a symmetrical and reasonably gradient temperature field, significantly correcting the "M-shaped" thickness and "W-shaped" composition distribution of the epitaxial structure, and improving overall uniformity. If the chord length is less than 40 mm, the edge area covered by the convex surface is too small to fully compensate for the temperature imbalance caused by edge warping, which will result in significant thickness and composition deviations in local areas. If the chord length is greater than 90 mm, the excessive extension of the convex surface will break the temperature field balance between the center and the edge, which may cause new distribution anomalies in the area near the center due to excessive heat. At the same time, it will also increase the mismatch with the stress management of the buffer layer, which will reduce the quality stability of the epitaxial structure.

[0036] In some examples, the specific μm value of the convexity protrusion can be 30, 40, 50, 60, 70, 80, 90, 100, 35, 45, 65, 85, 95, or a value within any two of these ranges.

[0037] In some examples, the specific mm value of the convex chord length can be 40, 50, 60, 70, 80, 90, or a value within a range of any two of these values.

[0038] In some cases, the concave surface is spherical or ellipsoidal. A spherical concave surface forms a symmetrical distance gradient with a fixed curvature, while an ellipsoidal concave surface can adapt to asymmetric temperature field requirements by adjusting the major and minor axis parameters. Both types of concave surfaces avoid the abrupt changes in local distance that might occur with planar or stepped surfaces, thus achieving a smooth transition in heat transfer within the central region and precisely controlling the central temperature field to balance the heat compensation of the convex edge surface. Therefore, a spherical or ellipsoidal concave surface can make the temperature gradient in the central region more uniform, reducing thickness and composition fluctuations caused by local overheating or undercooling. Combined with the annular convex edge surface, it forms a globally symmetrical temperature field control, further enhancing the correction effect on "M-shaped" thickness and "W-shaped" composition distribution. Simultaneously, the curved structure can reduce the contact stress between the epitaxial wafer and the graphite disk, synergizing with the stress management effect of the buffer layer to improve the material's crystallization quality.

[0039] In some examples, the concave indentation is 15μm to 60μm, and the chord length of the concave surface is 40mm to 90mm. The purpose of setting the concave indentation to 15μm to 60μm is to adjust the distance between the central region of the graphite disk and the lower surface of the epitaxial wafer, thereby precisely reducing the temperature of the corresponding region to compensate for the abnormal center temperature caused by warping during the growth of the 8-inch gallium nitride epitaxial wafer, and thus improve the thickness and composition uniformity of the epitaxial wafer.

[0040] Understandably, the indentation of a concave surface refers to the vertical height difference between the lowest point of the spherical or ellipsoidal surface and the reference plane of the graphite disk; that is, the depth to which the concave surface is recessed relative to the reference plane of the graphite disk. The chord length of a concave surface refers to the distance between the endpoints on different sides of the spherical or ellipsoidal surface where the concavity begins, which is the straight-line distance between two opposite starting points of the concavity on the edge of the concave surface.

[0041] Within the concave area, adjusting the distance between the graphite disk in the central region and the lower surface of the epitaxial wafer allows for a suitable gradient balance between the heat transfer intensity in the central region and the convex edge. This avoids both excessively high central temperatures due to insufficient distance (leading to excessive compositional shift) and excessively low central temperatures due to excessive distance (exacerbating the "M-shaped" thickness distribution). This synergistically corrects the overall temperature field, improving thickness and compositional uniformity. If the concave depth is less than 15 μm, insufficient concave depth will result in an excessively small distance between the central region and the epitaxial wafer, leading to heat accumulation and potentially causing compositional anomalies (such as high aluminum content). If the concave depth is greater than 60 μm, excessive spacing will result in an excessively low central temperature, exacerbating the "M-shaped" thickness distribution (thicker at the center and edges, thinner in the middle). It may also disrupt the temperature field synergy with the convex edge, weakening the overall control effect and ultimately affecting the uniformity and crystal quality of the epitaxial structure.

[0042] Within the chord length range, the concave surface effectively covers the core area where temperature anomalies are caused by warping, ensuring that the distance adjustment generated by the indentation effectively balances the central temperature field. However, it does not extend excessively into the edge area, avoiding overlap with the control range of the convex surface and preventing temperature field interference. This, in conjunction with the concave indentation and convex parameters, forms a globally coordinated temperature gradient, enhancing the improvement effect on thickness and composition uniformity. If the chord length is less than 40 mm, the central area covered by the concave surface is too small to fully offset the central temperature anomaly, resulting in localized thickness and composition deviations. If the chord length is greater than 90 mm, excessive extension of the concave surface will encroach on the control area of ​​the convex edge, disrupting the temperature field balance between the center and edge, potentially causing new uneven distribution. Furthermore, it may mismatch with the edge structure of the epitaxial wafer, weakening the overall control synergy and ultimately affecting the quality of the epitaxial structure.

[0043] In some examples, the specific μm value of the concave indentation can be 15, 18, 25, 30, 40, 50, 60, or a value within a range of any two of these values.

[0044] In some examples, the specific mm value of the convex chord length can be 40, 50, 60, 70, 80, 90, or a value within a range of any two of these values.

[0045] For example, taking the flat surface of the graphite disk before processing as the reference plane, the convex surface is a curved surface that rises upward relative to the reference plane (the amount of protrusion = the vertical height of the highest point), and the concave surface is a curved surface that is recessed downward (the amount of indentation = the vertical depth of the lowest point); the annular convex shape is a continuous or segmented annular convex structure symmetrically distributed around the center of the graphite disk.

[0046] In some embodiments, the graphite disk is made of graphite, specifically high-purity graphite (C content greater than 99%). Graphite has excellent thermal conductivity and high-temperature stability, and can maintain stable thermal conductivity efficiency in the high-temperature environment (e.g., 1000℃~1060℃) during 8-inch epitaxial wafer growth. This ensures uniform heat transfer between the graphite disk and the substrate, providing a stable thermal conductivity basis for temperature field control on both convex and concave surfaces, and avoiding abnormal local thermal resistance due to impurities.

[0047] In some embodiments, the surface of the graphite disk is provided with a SiC coating with a thickness of 50nm~150nm. Utilizing the characteristics of SiC such as high temperature resistance, strong chemical stability and suitable thermal conductivity, it can isolate graphite from direct contact with reactive gases during high-temperature epitaxial growth, avoiding contamination of the epitaxial layer by impurities generated by the graphite material due to high-temperature reaction. At the same time, this thickness range can ensure uniform coating coverage to form effective protection without significantly hindering the thermal conduction between the graphite disk and the substrate, ensuring the stable transmission of the temperature field regulation effect of the convex and concave surfaces.

[0048] In some examples, the specific nm value of the SiC coating thickness can be 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, or a value within any two of these ranges.

[0049] In some embodiments, the multilayer AlGaN structure is configured as at least two Al layers stacked sequentially. X Ga 1-X The N structure, where 0.1 ≤ X ≤ 0.9, can be configured with 4, 3, or 2 layers of Al. X Ga 1-X N-structure; utilizing the multilayer AlGaN structure, the difference in aluminum content between different layers forms a gradient stress buffering mechanism. The aluminum content is in the range of 0.1 to 0.9, which can not only ensure the stability of each layer material and the lattice matching with adjacent layers, but also gradually release the stress between the substrate and the epitaxial layer through multilayer stacking, while controlling key performance parameters such as carrier concentration.

[0050] In some embodiments, the aluminum content in the multilayer AlGaN structure decreases along the direction from the substrate to the high-resistivity layer. The gradual change in aluminum content forms a progressive lattice-matching transition. There is a lattice mismatch between the substrate and the epitaxial layer. AlGaN with higher aluminum content has better lattice compatibility with the substrate. As it approaches the high-resistivity layer, the decreasing aluminum content makes the lattice constant of each layer gradually approach the requirements of the high-resistivity layer and the layer above it. The stress caused by the lattice mismatch is gradually released through a step-like transition.

[0051] For example, the multilayer AlGaN structure is configured as four layers of Al stacked sequentially. X Ga 1-X N-structure, 0.1≤X≤0.9; for example, Al is sequentially distributed along the direction from the substrate to the high-resistivity layer. 0.80 Ga 0.20 N, Al 0.50 Ga 0.50 N, Al 0.20 Ga 0.80 N, Al 0.10 Ga 0.90 The nitrogen (N) and aluminum content decrease in that order.

[0052] For example, the multilayer AlGaN structure is configured as 5 layers of Al stacked sequentially. X Ga 1-X N-structure, 0.1≤X≤0.9; for example, Al is sequentially distributed along the direction from the substrate to the high-resistivity layer. 0.90 Ga 0.10 N, Al 0.80 Ga 0.20 N, Al 0.50 Ga 0.50 N, Al 0.20Ga 0.80 N, Al 0.10 Ga 0.90 The nitrogen (N) and aluminum content decrease in that order.

[0053] For example, the multilayer AlGaN structure is configured as three layers of Al stacked sequentially. X Ga 1-X N-structure, 0.1≤X≤0.9; for example, Al is sequentially distributed along the direction from the substrate to the high-resistivity layer. 0.80 Ga 0.20 N, Al 0.50 Ga 0.50 N, Al 0.20 Ga 0.80 The nitrogen (N) and aluminum content decrease in that order.

[0054] For example, the multilayer AlGaN structure is configured as two layers of Al stacked sequentially. X Ga 1-X N-structure, 0.1≤X≤0.9; for example, Al is sequentially distributed along the direction from the substrate to the high-resistivity layer. 0.80 Ga 0.20 N, Al 0.50 Ga 0.50 The nitrogen (N) and aluminum content decrease in that order.

[0055] In some embodiments, the superlattice structure is made of at least one of AlN / AlGaN, AlGaN / AlGaN, and AlGaN / GaN. Based on the gradual differences and good compatibility of the lattice constants between these materials, periodic interfaces are formed. The material combinations of AlN / AlGaN, AlGaN / AlGaN, and AlGaN / GaN can construct gradient lattice transitions. Stress can be released stepwise by means of a controllable stress field generated by interlayer lattice mismatch. Furthermore, the chemical stability of the material is adapted to the high-temperature environment of epitaxial growth, and it can stably participate in the growth process.

[0056] In some embodiments, the multilayer superlattice structure is configured as a y-layer superlattice structure stacked sequentially, where y≥2. By utilizing the interface effect between layers in the superlattice structure, stress can be dispersed and released at each interface through multilayer stacking, avoiding stress concentration at a single interface. At the same time, the periodic structure can effectively suppress the longitudinal propagation of defects such as dislocations.

[0057] In some embodiments, the aluminum content of the multilayer superlattice structure decreases along the direction from the substrate to the high-resistivity layer. This enhances the orderliness of stress release, reduces defects and cracks caused by lattice mismatch, and improves the crystal quality of the superlattice structure. The orderly decrease in aluminum content ensures a smoother performance transition between the multilayer superlattice and adjacent layers. In synergy with other structures, it further guarantees the thickness and compositional uniformity of the epitaxial structure, providing a more stable foundation for gallium nitride devices and chips, and helping to improve device performance and yield.

[0058] Understandably, the aluminum content in the multilayer AlGaN structure decreases along the direction from the substrate to the high-resistivity layer; and the aluminum content in the multilayer superlattice structure decreases along the direction from the substrate to the high-resistivity layer, where the aluminum content is the molar content of aluminum.

[0059] For example, the multilayer superlattice structure is configured as a four-layer superlattice structure stacked sequentially, with a first superlattice structure layer, a second superlattice structure layer, a third superlattice structure layer, and a fourth superlattice structure layer distributed sequentially along the direction from the substrate to the high-resistivity layer, with aluminum contents of 40-45%, 24-28%, 15-18%, and 6-9%, respectively, showing a decreasing trend in aluminum content.

[0060] For example, the multilayer superlattice structure is configured as a five-layer superlattice structure stacked sequentially, with the first superlattice structure layer, the second superlattice structure layer, the third superlattice structure layer, the fourth superlattice structure layer and the fifth superlattice structure layer distributed sequentially, and the aluminum contents are 40-45%, 33-36%, 24-28%, 15-18% and 6-9% respectively, with the aluminum content decreasing in that order.

[0061] For example, the multilayer superlattice structure is configured as a three-layer superlattice structure stacked sequentially, with a first superlattice structure layer, a second superlattice structure layer, and a third superlattice structure layer distributed sequentially, and aluminum contents of 40-45%, 15-18%, and 6-9%, respectively, with the aluminum content decreasing in that order.

[0062] For example, the multilayer superlattice structure is configured as two superlattice structures stacked sequentially, with a first superlattice structure layer and a second superlattice structure layer distributed sequentially, and aluminum contents of 33-36% and 15-18% respectively, with the aluminum content decreasing in that order.

[0063] In some embodiments, the buffer layer thickness is 2000nm~3500nm. Based on the transition effect of the buffer layer between the substrate and the high-resistivity layer, within this thickness range, the lattice mismatch stress between the substrate and the epitaxial layer can be gradually released through a gradient structure, while ensuring uniform heat conduction between layers. This adapts to the temperature field control requirements of the graphite disk and provides a stable substrate for the subsequent growth of high-resistivity layers. If the buffer layer thickness is <2000nm, the stress release is insufficient, which can easily lead to cracking of the upper structure. If the buffer layer thickness is >3500nm, it will increase the growth time and material consumption, and may cause thickness unevenness due to overgrowth, both of which will reduce the quality of the epitaxial wafer and the device yield.

[0064] In some embodiments, the substrate thickness is 0.7 mm to 1.2 mm. Based on the growth requirements of 8-inch epitaxial wafers, this thickness range can reduce substrate warpage, ensure good contact with the graphite disk, facilitate uniform temperature distribution, and help optimize the thickness and composition uniformity of the epitaxial structure. Furthermore, this thickness range ensures sufficient mechanical strength to support the subsequent stacking of epitaxial layers (nucleation layers, buffer layers, etc.) without warping. It also ensures stable contact with the graphite disk, guaranteeing a uniform heat conduction path and reducing local deformation caused by insufficient thickness or heat conduction hysteresis caused by excessive thickness. If the substrate thickness is <0.7 mm, the substrate's mechanical strength is insufficient, making it prone to warping due to epitaxial layer growth stress, exacerbating temperature imbalance. If the substrate thickness is >1.2 mm, the heat conduction efficiency decreases, making it difficult to respond to the temperature field control of the graphite disk, leading to local temperature deviations, disrupting uniformity, and ultimately affecting device yield.

[0065] In some embodiments, the nucleation layer material is AlN, with a thickness of 150 nm to 300 nm. AlN is chosen as the nucleation layer material because it has good lattice matching and compatibility with the substrate and subsequent epitaxial layers (such as buffer layers), serving as a stable basis for epitaxial growth. Setting the nucleation layer thickness to 150 nm to 300 nm ensures a continuous and dense nucleation layer, providing uniform nucleation sites for subsequent layers, while avoiding excessive thickness that could lead to stress accumulation. This ensures the uniformity and stability of subsequent epitaxial layer growth, helping to optimize the thickness and composition distribution of the epitaxial structure. If the nucleation layer thickness is <150 nm, the nucleation layer will be discontinuous and prone to defects; if the nucleation layer thickness is >300 nm, excessive stress will accumulate, leading to cracking of the epitaxial layer. Both of these conditions will disrupt uniformity and reduce device yield.

[0066] In some embodiments, the high-resistivity layer material is carbon-doped GaN with a thickness of 1000 nm to 1800 nm. Using carbon-doped GaN significantly improves the resistivity of GaN, meeting the high-resistivity requirement. The thickness of the high-resistivity layer, set to 1000 nm to 1800 nm, provides sufficient thickness to create effective isolation and support, adapting to the structural connection with the upper and lower layers (buffer layer, channel layer), while also ensuring charge isolation and preventing interference with the electrical performance of the channel layer. If the thickness of the high-resistivity layer is <1000 nm, the isolation effect is insufficient, easily leading to leakage; if the thickness of the high-resistivity layer is >1800 nm, the total epitaxial thickness is too large, resulting in a worse crack level.

[0067] In some embodiments, the channel layer material is intrinsic GaN, with a thickness of 150 nm to 500 nm. Using intrinsic GaN for the channel layer provides high electron mobility due to its lack of doping with other elements, making it a highly efficient channel for carrier transport. The channel layer thickness of 150 nm to 500 nm provides sufficient space to accommodate a two-dimensional electron gas, adapting to the structure of high-resistivity layers and insertion layers, while avoiding the decrease in carrier mobility due to excessive thickness or the limitation of carrier concentration due to excessive thinness. If the channel layer thickness is <150 nm, the carrier space is insufficient, limiting the current capacity; if the channel layer thickness is greater than 500 nm, it is prone to having more background carriers in the channel layer, leading to increased leakage current.

[0068] In some embodiments, the intercalation layer material is AlN, with a thickness of 0.5 nm to 1.5 nm. The AlN intercalation layer material exhibits lattice mismatch with the channel layer (intrinsic GaN) and the barrier layer (AlGaN), which can induce a strong polarization electric field and enhance the two-dimensional electron gas concentration. Setting the intercalation layer thickness to 0.5 nm to 1.5 nm allows for effective control of the polarization effect through an extremely thin monolayer structure, adapting to the interface characteristics with adjacent layers, while avoiding excessive thickness leading to increased interface defects or insufficient polarization due to insufficient thickness. If the intercalation layer thickness is <0.5 nm, the polarization effect is insufficient, resulting in low carrier concentration; if the intercalation layer thickness is >1.5 nm, interface defects are easily generated, hindering carrier transport. Both of these factors reduce the electrical performance and uniformity of the device, affecting yield.

[0069] Understandably, intrinsic GaN is pure GaN material without any other doping, consisting only of gallium (Ga) and nitrogen (N), and follows the GaN crystal structure (hexagonal wurtzite structure).

[0070] Carbon-doped GaN is formed by doping GaN materials with carbon (C). GaN itself is composed of gallium (Ga) and nitrogen (N), while carbon-doped GaN introduces carbon atoms as impurities to replace some of the nitrogen atoms.

[0071] In some embodiments, the barrier layer material is AlGaN, with a thickness of 10 nm to 20 nm. The use of AlGaN for the barrier layer is advantageous because its lattice mismatch and spontaneous polarization difference with the channel layer (intrinsic GaN) can generate a polarization electric field, promoting the generation of a high-concentration two-dimensional electron gas in the channel layer. The barrier layer thickness of 10 nm to 20 nm is designed to enhance the polarization effect and maintain sufficient carrier concentration, adapting to the interface characteristics of the insertion layer and the channel layer, while avoiding excessive stress due to excessive thickness or insufficient polarization due to insufficient thickness. If the barrier layer thickness is <10 nm, the polarization effect is insufficient, resulting in low carrier concentration; if the barrier layer thickness is greater than 20 nm, the compressive stress weakens, leading to a deterioration in the polarization effect.

[0072] In some embodiments, the conditions for forming a buffer layer by in-situ MOCVD are as follows: The reaction temperature was 1000–1040 °C, the reaction pressure was 50–60 mbar, the hydrogen flow rate was 80–120 slm, the ammonia flow rate was 3–10 slm, the trimethylgallium flow rate was 30–300 sccm, the trimethylaluminum flow rate was 300–800 sccm, the ethylene flow rate was 0–400 sccm, and the total growth time was 60–200 min. Under these conditions, the 1000–1040 °C temperature was suitable for AlGaN growth kinetics, the 50–60 mbar pressure ensured stable reaction gas concentrations, hydrogen and ammonia provided the reaction atmosphere and nitrogen source, the trimethylgallium and trimethylaluminum were used to precisely control the gallium-aluminum ratio, ethylene was used to adjust doping, and the 60–200 min total growth time ensured the buffer layer thickness met requirements. By using in-situ MOCVD combined with the above parameter settings, uniform growth of the buffer layer can be achieved, effectively releasing stress and forming a high-quality buffer layer. This provides a stable substrate for subsequent layer growth and improves the uniformity of the epitaxial structure.

[0073] In some embodiments, the conditions for forming a high-resistivity layer by in-situ MOCVD are as follows: The reaction temperature is 980–1020℃, the reaction pressure is 60–200 mbar, the hydrogen flow rate is 80–130 slm, the ammonia flow rate is 10–30 slm, the trimethylgallium flow rate is 250–400 sccm, the ethylene flow rate is 150–400 sccm, and the growth time is 20–50 min. Under these conditions, the reaction temperature of 980–1020℃ is suitable for carbon-doped GaN growth, the pressure of 60–200 mbar ensures stable participation of the reaction gases, hydrogen and ammonia provide the reaction atmosphere and nitrogen source, trimethylgallium serves as the gallium source, and ethylene serves as the carbon source for precise doping control, and the growth time of 20–50 min ensures that the high-resistivity layer thickness meets the requirements. By using in-situ MOCVD combined with the above parameter settings, uniform growth of the high-resistivity layer can be achieved, forming a high-quality high-resistivity layer that effectively performs its isolation function, stabilizes the electrical performance of the channel layer, and simultaneously ensures high resistivity characteristics.

[0074] In some embodiments, the conditions for forming the channel layer by in-situ MOCVD are as follows: The reaction temperature was 1030–1060 °C, the reaction pressure was 200–400 mbar, the hydrogen flow rate was 90–110 slm, the ammonia flow rate was 30–45 slm, the trimethylgallium flow rate was 100–300 sccm, and the growth time was 15–40 min. Under these conditions, the reaction temperature of 1030–1060 °C is suitable for the growth kinetics of intrinsic GaN, which is beneficial for forming high-quality crystals; the pressure of 200–400 mbar ensures sufficient participation of the reaction gases, reducing defects; hydrogen and ammonia provide a clean reaction atmosphere and nitrogen source; trimethylgallium allows for precise control of gallium content; and the growth time of 15–40 min ensures the thickness meets the requirements of 150 nm–500 nm. By using in-situ MOCVD combined with the above parameter settings, uniform growth of the channel layer is achieved, ensuring high electron mobility and forming a high-quality intrinsic GaN channel layer, providing an efficient channel for carrier transport and stabilizing the device's conductivity. If the temperature deviates from the above range, the crystal quality will decrease; if the pressure or gas flow rate deviates from the above range, the growth uniformity will be affected; if the time deviates from the above range, the thickness will not meet the standard.

[0075] In some embodiments, the conditions for forming the insertion layer by in-situ MOCVD are as follows: The reaction temperature was 1030~1060℃, the reaction pressure was 50~150mbar, the hydrogen flow rate was 80~110slm, the ammonia flow rate was 10~35slm, the trimethylaluminum flow rate was 40~120sccm, and the growth time was 0.5~2min.

[0076] In the above process conditions, a reaction temperature of 1030~1060℃ is suitable for the growth characteristics of AlN, which is conducive to the formation of dense crystals; a pressure of 50~150mbar ensures effective participation of the reaction gas, which is suitable for the uniform growth of ultra-thin layers (0.5nm~1.5nm); hydrogen and ammonia provide a clean reaction atmosphere and nitrogen source; trimethylaluminum precisely controls the aluminum content; and a growth time of 0.5~2min ensures that the thickness meets the ultra-thin requirements. By using the in-situ MOCVD method combined with the above parameter settings, the insertion layer grows uniformly, ensuring its polarization control function, forming a high-quality AlN insertion layer, enhancing the polarization effect with adjacent layers, and increasing the two-dimensional electron gas concentration. If the temperature deviates from the above range, the quality of AlN crystals will decrease; if the pressure or gas flow rate deviates from the above range, it will affect the uniformity of the layer thickness; and if the time deviates from the above range, the thickness will deviate.

[0077] In some embodiments, the conditions for forming a barrier layer are: The reaction temperature was 1030–1060 °C, the reaction pressure was 50–150 mbar, the hydrogen flow rate was 80–110 slm, the ammonia flow rate was 10–35 slm, the trimethylgallium flow rate was 30–60 sccm, and the trimethylaluminum flow rate was 40–120 sccm. The growth time was 2–5 min. Under these conditions, the reaction temperature of 1030–1060 °C is suitable for the growth characteristics of AlGaN material, ensuring crystal quality; the pressure of 50–150 mbar ensures stable reaction gas concentration, facilitating uniform growth; hydrogen and ammonia provide the reaction atmosphere and nitrogen source; the precise control of the gallium-aluminum ratio of trimethylgallium and trimethylaluminum ensures the formation of the required AlGaN composition; and the 2–5 min growth time guarantees a thickness of 10 nm–20 nm. By using in-situ MOCVD combined with the above parameter settings, uniform growth of the barrier layer is achieved, ensuring the polarization effect between it and the channel layer, forming a high-quality barrier layer, effectively inducing a high-concentration two-dimensional electron gas in the channel layer, and ensuring the device's conductivity. If the temperature deviates from the above range, it will lead to a decrease in crystal quality or abnormal composition; if the pressure or gas flow rate deviates from the above range, it will affect the growth uniformity or gallium-aluminum ratio; if the time deviates from the above range, it will cause the barrier thickness to deviate.

[0078] According to a second aspect of this application, an epitaxial structure is provided, which is prepared by the method described in the first aspect. Therefore, this epitaxial structure possesses all the beneficial effects of the epitaxial structure prepared by the method of the first aspect, which will not be elaborated further here. Preferably, the epitaxial structure is circular in shape; preferably, the size of the epitaxial structure is 6 inches, 8 inches, or 12 inches.

[0079] For example, such as Figure 2 As shown, the epitaxial structure of this application comprises, from bottom to top, a substrate 00, a nucleation layer 10, a buffer layer 20, a high-resistivity layer 30, a channel layer 40, an insertion layer 50, and a barrier layer 60.

[0080] In some embodiments, the epitaxial structure is 8 inches in size; for example, the epitaxial structure is an 8-inch epitaxial wafer.

[0081] According to a third aspect of this application, a semiconductor device is provided, comprising an epitaxial structure fabricated by the method of the first aspect, or an epitaxial structure as described in the second aspect. The semiconductor device may include, but is not limited to, power devices, radio frequency devices, etc. Power devices may be, for example, high electron mobility transistors (HEMTs); radio frequency devices may be, for example, radio frequency switches, radio frequency power amplifiers, low noise amplifiers, etc. Because this semiconductor device includes the aforementioned epitaxial structure, it also possesses the various beneficial effects brought about by the aforementioned epitaxial structure, such as stable electrical performance and high yield.

[0082] Example The following examples, using epitaxial wafers as a case study, illustrate the implementation of the epitaxial structure of this application in more detail through embodiments and comparative examples.

[0083] Test methods and equipment Thickness test: The total thickness and distribution of the epitaxial wafer were determined using an Etamax PLATO photoluminescence spectrometer.

[0084] For example, the epitaxial wafer to be tested is placed on the sample stage of the Etamax PLATO photoluminescence spectrometer, ensuring the wafer surface is clean and free of impurities, and that it makes good contact with the sample stage to avoid signal interference. Then, the instrument parameters are set, including the excitation wavelength, power, and detection range. The excitation light must be incident perpendicularly to the surface of the epitaxial wafer, covering different radial positions (such as the center, the central annular region, and the edge). The instrument irradiates the epitaxial wafer with a laser of a specific wavelength. The laser light is reflected at the interfaces of each epitaxial layer, and the reflected light is received by the spectrometer to form an interference signal. Based on the spacing and number of interference fringes, combined with the refractive index parameters of the epitaxial material, the total thickness at each test point is calculated. Multiple radially uniformly distributed test points (no fewer than 10) are tested, and the average total thickness and the deviation of each point from the average are statistically analyzed to characterize the uniformity of the thickness distribution.

[0085] Component and distribution testing: The aluminum composition and distribution of the barrier layer were detected using an Etamax PLATO photoluminescence spectrometer.

[0086] For example: The epitaxial wafer is fixed on the sample stage, and the image is focused on the barrier layer region. The excitation light intensity and spectral acquisition range of the instrument are adjusted so that the excitation light energy is sufficient to excite the barrier layer material to generate a characteristic photoluminescence signal. Changes in the aluminum composition will cause changes in the bandgap of the barrier layer, which in turn will shift the wavelength of the characteristic emission peak; the higher the aluminum content, the shorter the wavelength of the emission peak. By scanning different radial positions of the epitaxial wafer (corresponding to the thickness test points), the emission spectra of each point are collected, and the wavelength of the characteristic emission peak of the barrier layer is identified and recorded. Based on the calibration curve of wavelength versus aluminum composition (pre-calibrated using standard samples), the aluminum composition content of each test point is calculated. The composition data of multiple test points are statistically analyzed, and the average value and deviation are calculated to characterize the uniformity of composition distribution.

[0087] Example 1-1 The fabrication method of an eight-inch gallium nitride epitaxial wafer is achieved through the following specific process steps; Step S1, provide a gallium nitride epitaxial graphite disk 70, such as Figure 3As shown, with the graphite disk surface as a reference, the concave surface 701 located in the central region of the graphite disk has a recess of 30 μm and a spherical shape; the convex surface 702 of the graphite disk has a protrusion of 50 μm, and the convex surface is a ring-shaped ridge surrounding the concave surface; the graphite disk is made of graphite, and the surface SiC coating has a thickness of 100 nm. Step S2: An eight-inch single-crystal silicon substrate 00 with a thickness of 725 μm is provided on the graphite disk 70 of step S1. Step S3: On the eight-inch single-crystal silicon substrate 00 of step S2, a nucleation layer 10 is prepared by in-situ MOCVD. The nucleation layer material is AlN and the thickness is 200nm. Step S4: A buffer layer 20 is prepared on the nucleation layer 10 in step S3 using in-situ MOCVD. This buffer layer contains four AlGaN structures from bottom to top, with Al as the material. 0.80 Ga 0.20 N, Al 0.50 Ga 0.50 N, Al 0.20 Ga 0.80 N, Al 0.10 Ga 0.90 N. Wherein: Al 0.80 Ga 0.20 The nitrogen growth temperature was 1040℃, the reaction pressure was 50mbar, the hydrogen flow rate was 120slm, the ammonia flow rate was 5slm, the trimethylgallium flow rate was 40sccm, the trimethylaluminum flow rate was 420sccm, the growth time was 45min, and the thickness was 600nm. Al 0.50 Ga 0.50 The nitrogen growth temperature was 1030℃, the reaction pressure was 50mbar, the hydrogen flow rate was 115slm, the ammonia flow rate was 5slm, the trimethylgallium flow rate was 80sccm, the trimethylaluminum flow rate was 100sccm, the ethylene flow rate was 100sccm, the growth time was 45min, and the thickness was 800nm. Al 0.20 Ga 0.80 The nitrogen growth temperature was 1030℃, the reaction pressure was 50mbar, the hydrogen flow rate was 110slm, the ammonia flow rate was 8slm, the trimethylgallium flow rate was 220sccm, the trimethylaluminum flow rate was 120sccm, the ethylene flow rate was 150sccm, the growth time was 20min, and the thickness was 800nm. Al 0.10 Ga 0.90The nitrogen growth temperature was 1030℃, the reaction pressure was 50mbar, the hydrogen flow rate was 110slm, the ammonia flow rate was 10slm, the trimethylgallium flow rate was 300sccm, the trimethylaluminum flow rate was 300sccm, the ethylene flow rate was 250sccm, the growth time was 16min, and the thickness was 1000nm. Step S5: A high-resistivity layer 30 is prepared on the buffer layer 20 in step S4 by in-situ MOCVD. The reaction temperature is 1000℃, the reaction pressure is 100mbar, the hydrogen flow rate is 105slm, the ammonia flow rate is 10slm, the trimethylgallium flow rate is 350sccm, the ethylene flow rate is 300sccm, the growth time is 34min, and the thickness is 1700nm. Step S6: In step S5, a channel layer 40 is prepared on the high-resistivity layer 30 by in-situ MOCVD. The reaction temperature is 1050℃, the reaction pressure is 400mbar, the hydrogen flow rate is 100slm, the ammonia flow rate is 35slm, the trimethylgallium flow rate is 150sccm, the growth time is 15min, and the thickness is 300nm. Step S7: An insertion layer 50 is prepared on the channel layer 40 in step S6 by in-situ MOCVD. The reaction temperature is 1030℃, the reaction pressure is 100mbar, the hydrogen flow rate is 85slm, the ammonia flow rate is 30slm, the trimethylaluminum flow rate is 90sccm, the growth time is 45s, and the thickness is 1nm. Step S8: A barrier layer 60 is prepared on the insertion layer 50 in step S7 by in-situ MOCVD. The reaction temperature is 1030℃, the reaction pressure is 100mbar, the hydrogen flow rate is 85slm, the ammonia flow rate is 30slm, the trimethylgallium flow rate is 40sccm, the trimethylaluminum flow rate is 90sccm, the growth time is 3min, and the thickness is 15nm.

[0088] Examples 1-2 to 1-16 and Comparative Examples 1-1 to 1-3 Except for setting the protrusion of the convex surface of the graphite disk, the indentation of the concave surface of the graphite disk, and the structure of the buffer layer according to Table 1, the rest is the same as in Example 1-1.

[0089] Comparative Examples 2-1 to 2-5 Except for setting the corresponding parameters for steps S4, S5, S6, and S8 according to Table 2, the rest are the same as in Example 1-1; wherein step S8 of Comparative Example 2-5 is the same as step S8 of Comparative Example 2-1.

[0090] The epitaxial wafers prepared according to Tables 1 and 2 were subjected to testing according to the above-described testing methods and equipment. Thickness test and composition and distribution test results are summarized in Tables 1 and 2.

[0091] Table 1

[0092] Table 2

[0093] in Figure 4 The total thickness distribution curves of the epitaxial structures prepared in Examples 1-1 and 2-1 provided for specific embodiments of this application; Figure 5 The distribution curves of Al mass content in the barrier layer of the epitaxial structures prepared in Examples 1-1 and Comparative Examples 2-1, which are specific embodiments of this application, are shown.

[0094] Combination Figure 4 As shown in Table 2, the total thickness distribution in the radial direction of Examples 1-1 and Comparative Example 2-1 shows that the thickness of Comparative Example 2-1 exhibits a distinct M-shaped distribution in the radial direction, with an average thickness of 5.47 μm and a uniformity of 0.82%. Compared to Comparative Example 2-1, Example 1-1, employing the specific graphite disk structure and corresponding epitaxial wafer preparation method described in this application, significantly improved the non-uniform distribution of the epitaxial wafer thickness, ultimately achieving a total thickness uniformity of 0.18% with an average thickness of 5.48 μm.

[0095] Combination Figure 5 As shown in Table 2, the aluminum composition distribution of the barrier layer in the radial direction of Examples 1-1 and Comparative Example 2-1 shows that the aluminum composition of the barrier layer in Comparative Example 2-1 exhibits a distinct W-shaped distribution with an average composition of 23.04% and a uniformity of 1.58%. After adopting the scheme of this application, Example 1-1 significantly eliminated the W-shaped distribution characteristic of the composition, ultimately achieving an average composition of 23.07% and a composition uniformity of 0.48% in the barrier layer.

[0096] Table 1 shows that the average thickness of Examples 1-1 to 1-4 ranges from 5.21 to 5.51 μm, the thickness uniformity ranges from 0.18% to 0.27%, the average aluminum content ranges from 23.07% to 23.13%, and the composition uniformity ranges from 0.37% to 0.49%. Example 1-1 exhibits the best thickness uniformity. The results from Examples 1-1 to 1-4 demonstrate that the design of decreasing aluminum content in multilayer AlGaN structures balances the temperature field through a graphite disk structure with convex and concave surfaces, releasing stress. The 4-layer structure (Example 1-1) shows better coordination between stress gradient release and temperature field matching, resulting in better uniformity. When the number of layers decreases (Examples 1-2 and 1-3), stress release is insufficient, leading to a slight decrease in uniformity. While the 5-layer structure (Examples 1-4) increases the number of layers, the cumulative growth deviation makes the uniformity slightly inferior to 1-1, indicating that the 4-layer AlGaN structure and the graphite disk parameters have a better synergistic effect.

[0097] The thickness uniformity (0.18%) and composition uniformity (0.48%) of Example 1-1 are better, while the uniformity of Examples 1-5 to 1-8 is relatively worse (thickness 0.25%-0.28%, composition 0.55%-0.64%).

[0098] The thickness uniformity of Examples 1-9 to 1-12 is 0.21%-0.25%, and the composition uniformity is 0.44%-0.51%, both approaching the excellent level of Example 1-1. This indicates that both buffer layer structures achieve stress gradient release by decreasing aluminum content along the substrate to the high-resistivity layer, forming a synergistic effect with the temperature field control of the graphite disk (convex surface heating and concave surface temperature control). Multilayer AlGaN smoothly transitions stress through gradient composition, while superlattices disperse stress through periodic interfaces. Although the structures are different, both can effectively balance stress and temperature field, thus achieving good thickness and composition uniformity. This shows that both buffer layer designs can be adapted to the optimization effect of graphite disks.

[0099] The thickness uniformity of Example 1-1 is 0.18% and the composition uniformity is 0.48%, which is significantly better than that of Comparative Examples 1-1 to 1-14 (thickness 0.72%-1.04%, composition 1.17%-1.79%). Comparative Example 1-1 has a temperature field imbalance due to the graphite disk not involving convex and concave surfaces, Comparative Example 1-2 only has a convex surface, Comparative Example 1-3 only has a convex surface, or the convex surface protrusion of Example 1-13 is set beyond the range, and the concave surface indentation of Example 1-14 is set beyond the range. This makes it impossible to correct the "M-type" thickness and "W-type" composition distribution, resulting in poor composition uniformity and an increase in the numerical values ​​of thickness uniformity and composition uniformity. The uniformity of Examples 1-15 (convex protrusion of 10μm to less than 30μm) deteriorated to 1.21%, and the uniformity of Examples 1-16 (concave indentation of 5μm to less than 15μm) deteriorated to 0.91%, indicating that when the convex protrusion and concave indentation are not within the corresponding range, the component uniformity is poor.

[0100] Table 2 shows that Example 1-1 has the best uniformity in thickness (0.18%) and composition (0.48%), while Comparative Example 2-1 has the worst uniformity (0.66% thickness, 0.79% composition) due to simultaneous adjustment of multiple step parameters. Comparative Examples 2-2 to 2-5 only adjusted a single step, and their uniformity was between the two (0.19%-0.32% thickness, 0.51%-0.82% composition). This indicates that the process parameters of each layer in Example 1-1, in conjunction with the temperature field control of the graphite disk and the buffer layer structure, matched the growth requirements and ensured uniform growth of each layer. The deviation of multiple step parameters in Comparative Example 2-1 disrupted the balance between temperature field and growth conditions. The adjustment of a single step (Comparative Examples 2-2 to 2-5) had a limited impact, but still weakened the synergistic effect. Therefore, its uniformity was not as good as that of Example 1-1, indicating that the process parameters of each layer need to be matched with the design of the graphite disk and the buffer layer to achieve better uniformity.

[0101] It should be noted that, in this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, or article that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, or article.

[0102] The various embodiments in this specification are described in a related manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0103] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for preparing an epitaxial structure, characterized in that, include: One side of the substrate is placed on a graphite disk, and a nucleation layer, a buffer layer, a high-resistivity layer, a channel layer, an insertion layer, and a barrier layer are sequentially formed on the other side of the substrate. The graphite disk has a convex surface and a concave surface, the concave surface is located in the middle region of the graphite disk, and the convex surface is located on at least one side of the concave surface; The buffer layer includes at least one of a multilayer AlGaN structure and a multilayer superlattice structure.

2. The preparation method according to claim 1, characterized in that, The convex surface satisfies at least one of the following conditions: (1) The convex surface is annularly raised, and the inner ring of the annularly raised surface is located on the outer periphery of the concave surface; (2) The protrusion of the convex surface is 30μm~100μm.

3. The preparation method according to claim 1, characterized in that, The concave surface satisfies at least one of the following conditions: (1) The concave surface is spherical or ellipsoidal in shape; (2) The indentation of the concave surface is 15μm~60μm.

4. The preparation method according to claim 1, characterized in that, The concave surface satisfies at least one of the following conditions: (1) The graphite disk is made of graphite; (2) The surface of the graphite disk is provided with a SiC coating, and the thickness of the SiC coating is 50nm~150nm.

5. The preparation method according to claim 1, characterized in that, The buffer layer satisfies at least one of the following conditions: (1) The multilayer AlGaN structure is configured as at least two layers of Al stacked sequentially. X Ga 1-X N-structure, where 0.1 ≤ X ≤ 0.9; (2) Along the direction from the substrate to the high-resistivity layer, the aluminum content in the multilayer AlGaN structure decreases. (3) The multilayer superlattice structure is configured as a stacked y-layer superlattice structure, where y≥2; (4) The material of the superlattice structure is at least one of AlN / AlGaN, AlGaN / AlGaN, and AlGaN / GaN; (5) Along the direction from the substrate to the high-resistivity layer, the aluminum content of the multilayer superlattice structure decreases.

6. The preparation method according to claim 1, characterized in that, The prepared epitaxial structure satisfies at least one of the following conditions: (1) The thickness of the substrate is 0.7 mm to 1.2 mm; (2) The nucleation layer material is AlN, and the thickness is 150nm~300nm; (3) The thickness of the buffer layer is 2000nm~3500nm; (4) The high-resistivity layer material is carbon-doped GaN with a thickness of 1000nm~1800nm; (5) The channel layer material is intrinsic GaN with a thickness of 150nm~500nm; (6) The insertion layer material is AlN, and the thickness is 0.5nm~1.5nm; (7) The barrier layer material is AlGaN with a thickness of 10nm~20nm.

7. The preparation method according to claim 1, characterized in that, The preparation method satisfies at least one of the following conditions: (1) The conditions for forming the buffer layer are: The reaction temperature was 1000~1040℃, the reaction pressure was 50~60mbar, the hydrogen flow rate was 80~120slm, the ammonia flow rate was 3~10slm, the trimethylgallium flow rate was 30~300sccm, the trimethylaluminum flow rate was 300~800sccm, the ethylene flow rate was 0~400sccm, and the total growth time was 60~200min. (2) The conditions for forming the high-resistivity layer are: The reaction temperature was 980~1020℃, the reaction pressure was 60~200mbar, the hydrogen flow rate was 80~130slm, the ammonia flow rate was 10~30slm, the trimethylgallium flow rate was 250~400sccm, the ethylene flow rate was 150~400sccm, and the growth time was 20~50min. (3) The conditions for forming the channel layer are: The reaction temperature was 1030~1060℃, the reaction pressure was 200~400mbar, the hydrogen flow rate was 90~110slm, the ammonia flow rate was 30~45slm, the trimethylgallium flow rate was 100~300sccm, and the growth time was 15~40min. (4) The conditions for forming the insertion layer are: The reaction temperature was 1030~1060℃, the reaction pressure was 50~150mbar, the hydrogen flow rate was 80~110slm, the ammonia flow rate was 10~35slm, the trimethylaluminum flow rate was 40~120sccm, and the growth time was 0.5~2min. (5) The conditions for the formation of the barrier layer are: The reaction temperature was 1030~1060℃, the reaction pressure was 50~150mbar, the hydrogen flow rate was 80~110slm, the ammonia flow rate was 10~35slm, the trimethylgallium flow rate was 30~60sccm, the trimethylaluminum flow rate was 40~120sccm, and the growth time was 2~5min.

8. An epitaxial structure, characterized in that, The epitaxial structure obtained by the preparation method according to any one of claims 1 to 7.

9. The epitaxial structure according to claim 8, characterized in that, The epitaxial structure measures 8 inches.

10. A semiconductor device, characterized in that, The epitaxial structure includes the preparation method described in any one of claims 1 to 7, or the epitaxial structure described in any one of claims 8 to 9.

Images