Method for controlling the quality of a composite structure comprising a thin c-sic layer, and composite structure

Abstract

The invention relates to a method for controlling the quality of a composite structure comprising a thin layer made of single-crystal silicon carbide placed on a carrier substrate made of polycrystalline silicon carbide, the method comprising: a) inspection of a free surface (10a) of the thin layer (10) using a technique coupling visible-light confocal microscopy and photoluminescence imaging, making it possible to detect defects, called secondary defects, b) preliminary identification of the secondary defects, by similarity, on the basis of their visible-light image, by virtue of an image recognition algorithm trained on various types of defects such as holes, bubbles, scratches, and defects of crystalline origin; at the end of step b), each secondary defect is associated with one identified type of defect, with a certain level of similarity, c) final classification of at least certain secondary defects by application of the following first conditions: - if the level of similarity associated with a secondary defect is greater than a high level, said secondary defect is definitively classified in the identified type of defect, - if the level of similarity associated with the secondary defect is between a low level and the high level, the photoluminescence image of said defect is analysed; if the secondary defect is associated with a labelled type of PL defect, said secondary defect is definitively classified in the identified type of defect, - in all other cases, the secondary defect is definitively classified as not a defect.

Description

Method for controlling the quality of a composite structure comprising a thin c-SiC layer, and composite structure FIELD OF THE INVENTION

[0001] The present invention relates to the field of semiconductor materials. It relates in particular to a method for controlling the quality of a composite structure comprising a thin layer of monocrystalline silicon carbide arranged on a support substrate of polycrystalline silicon carbide.

[0002] TECHNOLOGICAL BACKGROUND OF THE INVENTION

[0003] Silicon carbide is a particularly interesting material for the manufacture of power devices, radio frequencies or even devices operating at very high temperatures.

[0004] The quality of monocrystalline silicon carbide (c-SiC) substrates has improved significantly over the last ten years, accompanied by increasingly precise knowledge and detection of the different types of crystalline defects likely to be present in this material. The JEITA ("Japan Electronics and Information Technology Industries Association") standard also includes four documents relating to the zoology of defects on / in a layer grown by homoepitaxy on a 4H-SiC substrate (EDR 4712 / 100) and to non-destructive procedures for optical inspection of these defects (EDR 4712 / 200, / 300, / 400). In particular, a procedure is indicated for evaluating and referencing defects by combining optical inspection and photoluminescence imaging.

[0005] Commercially available equipment, such as the SICA88 from Lasertec, allows the combination of visible light and photoluminescence confocal Nomarski prism microscopy techniques (as described by [1] T.Kimoto et al., “Fundamentals of Silicon Carbide Technology Growth Characterization Devices and Applications”, p.126, or [2] D. Baierhofer et al., Materials Science in Semiconductor processing 140 (2022) 106414) and is commonly used to inspect c-SiC substrates before or after epitaxy.

[0006] As an example, [3] Das et al (“Statistical analysis of killer and non-killer defects in SiC and the impact of device performance”, Material Science Forum, ISSN 1662-9752, Vol.1004, pp 458-463 (2020) Trans Tech Publications Ltd) proposes a statistical analysis of “killer” defects for devices grown on a homoepitaxial c-SiC layer, and shows optical and photoluminescence images of typical c-SiC defects.

[0007] Although rapidly developing, high-quality c-SiC substrates remain expensive and difficult to source in large sizes. It is therefore advantageous to use layer transfer solutions to develop composite structures comprising a thin monocrystalline SiC layer (from a high-quality c-SiC donor substrate) on a lower-cost support substrate, for example polycrystalline SiC (p-SiC), which can also offer advantages in terms of electrical conductivity. A well-known thin-film transfer solution is the Smart Cut process. TM, based on light ion implantation and direct bonding of a c-SiC donor substrate to a support substrate at a bonding interface. The implantation creates a buried fragile plane along which separation occurs, leading to the transfer of a thin c-SiC layer onto the support substrate to form the composite structure, and allowing the recovery and recycling of the rest of the donor substrate to potentially perform one or more other layer transfers. Epitaxy can then be performed on the thin layer of the composite structure, followed by the development of electronic devices.

[0008] For the implementation of a layer transfer process to be economically viable, it is important to know how to control the quality of the donor substrates, so as to avoid transferring a layer that would be out of specification, or which would give rise to an epitaxial layer out of specification, in terms of "killer" defects. It is therefore essential to detect but also to accurately classify the crystalline defects (so-called primary defects) present on the donor substrates, to downgrade those, among these donor substrates, which will not allow the fabrication of a thin film of the required quality. The zoology of c-SiC defects is relatively well known, and numerous studies tend to define the type and size of defects, present in the c-SiC epitaxial layer, which would be killer for the components; criteria for classifying primary defects, with a view to a c-SiC thin film transfer, can therefore be established.

[0009] Furthermore, the detection and recognition of defects (so-called secondary defects) on and / or in the thin film (from a donor substrate) of a composite structure whose support substrate is made of polycrystalline SiC (p-SiC) are complex because the p-SiC grains are visible under said thin film. An effective thin film quality control method is therefore required, on the one hand, to avoid continuing the epitaxy steps if secondary defects in the thin film are likely to generate a density of killer defects, out of specification, in the homoepitaxial layer; on the other hand, it is important that the control method is capable of distinguishing the defects in the thin film from potential "false defects" (or measurement noise) associated with underlying grains of the p-SiC substrate.

[0010] In view of the growing use of layer transfer techniques in the manufacturing chain of electronic devices on c-SiC, there is therefore a strong need to define control processes allowing reliable inspection of composite structures.

[0011] SUBJECT OF THE INVENTION

[0012] The present invention addresses the stated problem. The invention relates to a method for controlling the quality of a composite structure comprising a thin layer (from a monocrystalline silicon carbide donor substrate) arranged on a polycrystalline silicon carbide support substrate, said control method allowing reliable classification of secondary defects present on and / or in the thin layer.

[0013] BRIEF DESCRIPTION OF THE INVENTION

[0014] The invention relates to a method for controlling the quality of a composite structure comprising a thin layer of monocrystalline silicon carbide arranged on a support substrate of polycrystalline silicon carbide, the method comprising the following steps:

[0015] a) inspection of a free surface of the thin layer by a technique coupling confocal microscopy in visible light and photoluminescence imaging, making it possible to detect defects, called secondary defects, present on and / or in the thin layer, and to form, for each defect, a visible light image and a photoluminescence image,

[0016] b) the preliminary identification of secondary defects, by similarity, on the basis of their image in visible light, using an image recognition algorithm, trained on different types of defects likely to be present on and / or in a thin layer, such as holes, bubbles, scratches, defects of crystalline origin; at the end of step b), each secondary defect being associated with an identified type of defect, with a certain level of similarity,

[0017] (c) the final classification of at least some secondary defects by applying the following first conditions:

[0018] - if the level of similarity associated with a secondary defect is greater than a high level, said secondary defect is definitively classified in the identified type of defect,

[0019] - if the similarity level associated with the secondary defect is between a low level and a high level, the photoluminescence image of said defect is analyzed by an image recognition algorithm trained on different labeled types of defects; if the secondary defect is associated with a labeled type of defect, said secondary defect is definitively classified in the identified type of defect,

[0020] - in other cases, the secondary defect is definitively classified as not being a defect.

[0021] According to other advantageous and non-limiting characteristics of the invention, taken alone or in any technically feasible combination: the secondary defects to which the first conditions apply are those associated, at the end of step b), with holes or defects of crystalline origin; the high level, called the first high level, and / or the low level, called the first low level, used in step c) for a secondary defect associated with a hole in step b), are different from the high level, called the second high level, and / or the low level, called the second low level, used in step c) for a secondary defect associated with a defect of crystalline origin in step b);the final classification step c) is also based on the application of the following second conditions:if a secondary defect is identified as a bubble-type defect in step b) with a similarity level higher than a third level, said secondary defect is definitively classified as a bubble,if a secondary defect is identified as a bubble-type defect in step b) with a similarity level lower than or equal to the third level, said secondary defect is definitively classified as not being a defect;the final classification step c) is also based on the application of the following third conditions:if a secondary defect is identified as a scratch-type defect in step b) with a similarity level higher than a fourth level, said secondary defect is definitively classified as a scratch,if a secondary defect is identified as a scratch-type defect in step b) with a similarity level lower than or equal to the fourth level, said secondary defect is definitively classified as not being a defect;each secondary defect detected in step a) has a size greater than 1μm, 5μm, or even 10μm;step c) leads to a grading of the composite structure and to a physical sorting making it possible to separate downgraded composite structures and composite structures meeting the specifications;a composite structure is downgraded when it has a density of defects of crystalline origin greater than 0.25 defects / cm; 2and / or an overall density of bubbles and holes greater than 0.2 defects / cm 2 .

[0022] The invention also relates to a composite structure formed from a thin layer of monocrystalline silicon carbide arranged on a support substrate of polycrystalline silicon carbide, said composite structure having a density of defects of crystalline origin less than or equal to 0.25 defects / cm 2 and an overall density of bubbles and holes less than or equal to 0.2 defects / cm 2 . BRIEF DESCRIPTION OF THE FIGURES

[0023] Other characteristics and advantages of the invention will emerge from the detailed description of the invention which follows with reference to the appended figures in which:

[0024] It presents a donor substrate, a composite structure without and with epitaxial layer;

[0025]

[0026] Laet illustrate different types of defects with their final classification by applying the particular conditions of the control method according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0027] In the description to follow, we will conventionally call “primary defects” the defects present on and / or in a donor substrate 1, “secondary defects” the defects present on and / or in the thin layer 10 of a composite structure 100, and “tertiary defects” the defects present on and / or in the raw epitaxial layer 150 on the thin layer of a composite structure 100.

[0028] In a main plane (x,y), the donor substrate 1 and the composite structure 100, with or without epitaxial layer 150, are preferably in the form of circular wafers with a diameter of 100mm, 150mm, 200mm, or even more. They could nevertheless be in any other form allowing their subsequent processing for the manufacture of components. The thickness of the substrates and structures extends along the z axis on the. It is the front faces 1a, 10a, 150a of the substrate 1 and of the structure 100 which are inspected and likely to contain the aforementioned defects.

[0029] Before starting the manufacture of a composite structure 100 which comprises a thin layer of monocrystalline silicon carbide (c-SiC) transferred onto a support substrate of polycrystalline silicon carbide (p-SiC), it is important to carry out a quality control of the donor substrate 1 (in c-SiC, for example of the 4H-SiC type) from which the thin layer will be produced.

[0030] An inspection of the front face 1a of the donor substrate 1 by a photoluminescence imaging technique as described in references [1] and [2] cited in the introduction, is preferably carried out to detect and classify defects, called primary defects thereafter.

[0031] Primary defects are made visible in photoluminescence imaging due to local variations in the intensity of a photoluminescence (PL) signal emitted by the front face 1a, following excitation by an incident beam. Local variations in the intensity of the PL signal reflect modified properties of the material (stresses, roughness, flatness, etc.), corresponding to the primary defect. The latter can appear on the generated PL image, in the form of white or black spot(s) depending on its characteristics.

[0032] This inspection step can be carried out on known equipment, for example SICA88 (Lasertec company) or "Photoluminescence scanner" (Intego company) or "MiPlato SiC" (EtaMax company). These equipments traditionally use the combination of an optical microscopy image and the photoluminescence image of the same defect to assign dimensional (size, surface) and intensity (contrast) criteria from the measured optical microscope and photoluminescence signals; they also assign a typology criterion (predefined classes of defects) to each of the primary defects, based on a learning algorithm. The predefined classes typically correspond to defects of the micro-hole type ("micro-pipe" in English), stacking fault (SF for "stacking fault"), depression ("pit"), bump ("bump"), scratch ("scratch"), particle.

[0033] After inspection and potentially downgrading of certain donor substrates 1 whose density and / or typology of defects would be incompatible with the transfer of a viable thin layer, the composite structure 100 can be manufactured. Any thin layer transfer technique can be implemented, in particular the Smart Cut process. This process, based on light ion implantation and direct assembly by molecular adhesion, will not be described in detail here because it is well known. To obtain the composite structure 100, it is implemented with a donor substrate 1 made of c-SiC and a support substrate 20 made of p-SiC.

[0034] Typically, the thickness of the thin layer 10 is between 20nm and 1000nm, and its free surface has a roughness less than or equal to 0.5nm RMS, or even 0.1nm RMS (measurement by atomic force microscopy AFM, on scans for example of 10x10μm 2 at 30x30μm 2). The support substrate 20 may have a thickness of between approximately 50 μm and several hundred micrometers, for example between 200 μm and 700 μm. Optionally, the composite structure 100 may comprise an intermediate layer, interposed between the thin layer 10 and the support substrate 20. This intermediate layer may be made of a dielectric, semiconductor or metallic material (such as for example silicon oxide, silicon, silicon carbide, tungsten, titanium, etc.). A bonding interface is present between the thin layer 10 and the support substrate 20, or adjacent to the intermediate layer.

[0035] The present invention relates to a method for controlling the quality of a composite structure 100 comprising a thin layer 10 of c-SiC arranged on a support substrate 20 of p-SiC ().

[0036] The control method comprises a step a) of inspecting a free surface 10a of the thin layer 10 by a technique coupling confocal microscopy in visible light (differential interference contrast – DIC) and photoluminescence imaging, to detect defects, called secondary defects, present on and / or in the thin layer 10. The detected defects typically have a size (or an equivalent diameter) greater than or equal to 1μm, greater than or equal to 5μm, or even greater than or equal to 10μm.

[0037] Differential interference contrast (also called Nomarski) is a lighting technique that allows tiny variations in the topography of a surface to be discerned by exploiting the interference of light waves.

[0038] Photoluminescence imaging is based on local variations in the intensity of a photoluminescence (PL) signal emitted by the free surface 10a, following excitation by an incident electromagnetic beam. In SICA88 equipment, which will be favored in the remainder of this description, the incident excitation beam has a wavelength of 313nm and the emitted photoluminescence signal is collected in a wavelength range from 700nm to 1000nm. The local variations in the intensity of the PL signal reflect modified properties of the material (stresses, roughness, flatness, etc.), corresponding to the secondary defect. The latter may appear on the generated PL image, in the form of white or black spot(s) depending on its characteristics.

[0039] In step a), a visible light image – hereinafter called a DIC image – and a photoluminescence image – hereinafter called a PL image – are formed for each detected secondary defect.

[0040] Secondary defects may be induced by primary defects or by particles or other surface contaminations that have not been completely eliminated before the assembly of the donor substrate 1 and the support substrate 20. They may therefore be defects of crystalline origin, bonding defects such as holes (local absence of thin layer 10) or bubbles (defect at the bonding interface 40, at which the thin layer 10 is not bonded and forms a blister), or even surface defects such as scratches or layer chips (“flakes”) present on the front face 10a of the thin layer.

[0041] The control method then comprises a step b) corresponding to the preliminary identification of secondary defects, by similarity, on the basis of their DIC image.

[0042] The aforementioned equipment (and in particular the SICA88) are capable of identifying secondary defects on the basis of images captured by visible light microscopy, using an image recognition algorithm. The algorithm is fed and trained with verified images of defects likely to be present on and / or in a thin layer 10, such as holes, bubbles, scratches, chips, defects of crystalline origin: it can thus learn to recognize each of these defects and calculate a level of similarity between the DIC image of the detected defect and its database, taking into account criteria of size, shape, color, contrast, etc.

[0043] At the end of step b), each secondary defect is associated with an identified type of defect (for example, hole, bubble, scratch, chip or defect of crystalline origin), with a certain level of similarity (between 0 and 1).

[0044] Due to the presence of a polycrystalline support substrate 20 under the thin layer 10, the preliminary identification carried out automatically by the equipment is not sufficiently reliable to decide on the quality of the composite structure 100. Indeed, this identification can be impacted by the underlying grains and the level of similarity attributed to the secondary defect does not in itself make it possible to determine whether said defect is indeed of the identified type or whether it is a “false defect”, for example linked to the p-SiC grains.

[0045] The control method therefore includes a step c) of final classification of secondary defects by application of specific conditions, set out below.

[0046] First conditions apply in particular to the secondary defects identified in step b) as being holes or defects of crystalline origin: if the level of similarity associated with a secondary defect is greater than a high level, said secondary defect is definitively classified in the identified type of defect; if the level of similarity associated with the secondary defect is between a low level and the high level, the PL image of said defect is analyzed by an image recognition algorithm trained on different labeled types of defects.Two cases can then arise: if the secondary defect is associated with a labeled type of defect, said secondary defect is definitively classified in the identified type of defect; if the secondary defect is not recognized as being a labeled type of defect, said secondary defect is definitively classified as not being a defect (i.e. considered as a “false defect”); finally, if the level of similarity associated with the secondary defect is lower than the low level, said secondary defect is also definitively classified as not being a defect.

[0047] A “false defect” is a defect detected and associated with a given type (with a certain level of similarity) in step b), but which does not actually correspond to said type: it can either correspond to another type of defect in the thin layer, to a measurement artifact or to the image of a grain in the underlying support substrate 20.

[0048] The PL image recognition algorithm is fed and trained with verified PL images of defects likely to be present on and / or in a thin layer 10: it can thus learn to recognize each of these defects and to classify them into predefined categories (i.e., associate them with labeled defect types), taking into account criteria of size, shape, contrast, etc. A labeled defect type therefore corresponds to a true defect, the signature of which on the PL image is translated by particular criteria (such as those mentioned above).

[0049] In condition (ii), when the PL image associated with the analyzed secondary defect corresponds to one of the labeled types of defects, this confirms that we are dealing with a “true defect” and the secondary defect is thus classified in the type identified in step b); on the other hand, when the PL image does not correspond to any of the labeled types of defects and / or the PL image does not show any particular contrast (no PL signal), the secondary defect is considered to be a “false defect”, although it has been associated with an identified type of defect in step b).

[0050] Note that the high level (called first high level) and / or the low level (called first low level) used in step c) to judge the level of similarity of the DIC image, for a secondary defect associated with a hole, may be different from the high level (called second high level) and / or the low level (called second low level) used for a secondary defect associated with a defect of crystalline origin.

[0051] Second conditions may apply to secondary defects identified in step b) as bubbles: if the similarity level associated with a secondary defect identified as a bubble is greater than a third level, said secondary defect is definitively classified as a bubble, if the similarity level is less than or equal to the third level, said secondary defect is definitively classified as not being a defect.

[0052] Third conditions may apply to secondary defects identified in step b) as being scratches: if the similarity level associated with a secondary defect identified as being a scratch is greater than a fourth level, said secondary defect is definitively classified as a scratch, if the similarity level is less than or equal to the fourth level, said secondary defect is definitively classified as not being a defect.

[0053] This final classification step c) makes it possible to reliably categorize the critical defects present on and / or in the thin layer 10, and avoids taking into account “false defects”, in particular linked to the detection of the underlying grains of the support substrate 20.

[0054] As an example, the control method is applied to a composite structure 100 comprising a thin layer of c-SiC 600 nm thick, arranged on a p-SiC support substrate 150 mm in diameter. The equipment used is the SICA88. In the examples of la and la, the first low level and the first high level of similarity, used in step c) to judge the level of similarity associated with the secondary defects identified as holes are respectively defined at 0.3 and 0.85. The second low level and the second high level of similarity, used in step c) to judge the level of similarity associated with the secondary defects identified as defects of crystalline origin are respectively defined at 0.3 and 0.85. The third level used in step c) to judge the level of similarity associated with the secondary defects identified as bubbles is defined at 0.85.Finally, the fourth level used in step c) to judge the level of similarity associated with secondary defects identified as scratches is set at 0.97.

[0055] Figures 2a and 2b illustrate hole-type defects, crystalline defects, bubbles and scratches detected on and / or in the thin layer 10:

[0056] [1] DIC image: Similarity level higher than the first high level => Defect classified as hole;

[0057] [2] DIC image: Similarity level between the first high level and the first low level => Defect classified as a hole because labeled defect detected on the PL image;

[0058] [3] DIC image: Similarity level higher than the second high level => Defect classified as a defect of crystalline origin;

[0059] [4] DIC image: Similarity level between the second high level and the second low level => Defect classified as a defect of crystalline origin because labeled defect detected on the PL image;

[0060] [5] DIC Image: Similarity level lower than the second low level => Secondary defect classified as not being a defect of crystalline origin;

[0061] [6] DIC Image: Similarity level higher than the third level => Defect classified in bubble;

[0062] [7] DIC Image: Similarity level higher than the fourth level => Defect classified as scratch.

[0063] At the end of step c) of the control method, it is possible to assign a grade to the composite structures 100 controlled. In particular, a higher quality grade can be assigned to the structures 100 having an overall density of bubbles and holes less than or equal to 0.2 defects / cm 2. Preferably, such a grade will be attributed to composite structures 100 having an overall density of bubbles, holes and scratches less than or equal to 0.2 defects / cm 2 . Even more preferably, the higher quality grade can be attributed to structures 100 having an overall density of defects, excluding defects of crystalline origin, less than or equal to 0.2 defects / cm 2 .

[0064] The acceptable density of defects of crystalline origin is defined according to the quality level of the donor substrate 1 from which the thin layer 10 is derived. Advantageously, a density of defects of crystalline origin less than or equal to 0.25 / cm 2 for a structure 100 gives it a higher quality grade.

[0065] Conversely, a lower grade, leading to the downgrading of a composite structure 100 at the end of step c), may be defined for an overall density of defects greater than the thresholds previously stated.

[0066] The present invention therefore also relates to a composite structure 100 formed from a thin layer 10 of monocrystalline silicon carbide arranged on a support substrate 20 of polycrystalline silicon carbide. Said composite structure, finally inspected with the control method according to the invention, has: a density of defects of crystalline origin less than or equal to 0.25 defects / cm 2 , less than or equal to 0.2 defects / cm 2 , or even less than or equal to 0.18 defects / cm 2 , or even less than or equal to 0.15 defects / cm 2 , and an overall density of bubbles and holes (and possibly scratches) less than or equal to 0.2 defects / cm 2 , less than or equal to 0.18 defects / cm 2, less than or equal to 0.15 defects / cm 2 , or even less than or equal to 0.12 defects / cm 2 , or even less than or equal to 0.1 defects / cm 2 .

[0067] After grading, physical sorting of the 100 composite structures allows separation of the downgraded 100 structures from the 100 structures meeting the specifications.

[0068] In a subsequent manufacturing step, SiC epitaxial growth can be carried out on the thin layer 10 of the composite structures 100 in specification, in order to form the epitaxial layer 150 on and in which the devices will be developed (). The epitaxial growth step can be carried out according to the techniques known in the state of the art.

[0069] The method for controlling the quality of the composite structures 100, before the epitaxy step, makes it possible to ensure that the density of killer tertiary defects, linked to the quality of the thin layer 10, will be controlled, or even lower than a given specification, because this method allows a reliable classification of the defects present on and / or in the thin layer 10, without however carrying out over-quality by downgrading expensive composite structures 100, on the basis of “false defects” induced in particular by an underlying polycrystalline substrate.

[0070] Of course, the invention is not limited to the embodiments and examples described, and variant embodiments may be made without departing from the scope of the invention as defined by the claims.

Claims

Method for controlling the quality of a composite structure (100) comprising a thin layer (10) of monocrystalline silicon carbide arranged on a support substrate (20) of polycrystalline silicon carbide, the method comprising the following steps: a) inspection of a free surface (10a) of the thin layer (10) by a technique coupling confocal microscopy in visible light and photoluminescence imaging, making it possible to detect defects, called secondary defects, present on and / or in the thin layer (10), and to form, for each defect, a visible light image and a photoluminescence image, b) preliminary identification of the secondary defects, by similarity, on the basis of their visible light image, using an image recognition algorithm, trained on different types of defects likely to be present on and / or in a thin layer (10), such as holes, bubbles, scratches, defects of crystalline origin;at the end of step b), each secondary defect being associated with an identified type of defect, with a certain level of similarity,c) the final classification of at least some secondary defects by applying the following first conditions:- if the level of similarity associated with a secondary defect is greater than a high level, said secondary defect is definitively classified in the identified type of defect,- if the level of similarity associated with the secondary defect is between a low level and the high level, the photoluminescence image of said defect is analyzed by an image recognition algorithm trained on different labeled types of defects; if the secondary defect is associated with a labeled type of defect, said secondary defect is definitively classified in the identified type of defect,- in other cases, the secondary defect is definitively classified as not being a defect.; Method for controlling the quality of a composite structure (100) according to claim 1, in which the secondary defects to which the first conditions apply are those associated, at the end of step b), with holes or defects of crystalline origin. Method for controlling the quality of a composite structure (100) according to claim 2, in which the high level, called first high level, and / or the low level, called first low level, used in step c) for a secondary defect associated with a hole in step b), are different from the high level, called second high level, and / or the low level, called second low level, used in step c) for a secondary defect associated with a defect of crystalline origin in step b). Method for controlling the quality of a composite structure (100) according to one of claims 1 to 3, wherein the final classification step c) is also based on the application of the following second conditions: - if a secondary defect is identified as a bubble-type defect in step b) with a similarity level greater than a third level, said secondary defect is definitively classified as a bubble, - if a secondary defect is identified as a bubble-type defect in step b) with a similarity level less than or equal to the third level, said secondary defect is definitively classified as not being a defect. Method for controlling the quality of a composite structure (100) according to one of claims 1 to 4, wherein the final classification step c) is also based on the application of the following third conditions: - if a secondary defect is identified as a scratch-type defect in step b) with a similarity level greater than a fourth level, said secondary defect is definitively classified as a scratch, - if a secondary defect is identified as a scratch-type defect in step b) with a similarity level less than or equal to the fourth level, said secondary defect is definitively classified as not being a defect. Method for controlling the quality of a composite structure (100) according to one of claims 1 to 5, in which each secondary defect detected in step a) has a size greater than 1 μm, 5 μm, or even 10 μm. Method for controlling the quality of a composite structure (100) according to one of claims 1 to 6, in which step c) leads to a grading of the composite structure (100) and to a physical sorting making it possible to separate downgraded composite structures and composite structures meeting the specifications. A method of controlling the quality of a composite structure (100) according to claim 7, wherein a composite structure (100) is downgraded when it has a density of defects of crystalline origin greater than 0.25 defects / cm 2 and / or an overall density of bubbles and holes greater than 0.2 defects / cm 2 . Composite structure (100) formed from a thin layer (10) of monocrystalline silicon carbide arranged on a support substrate (20) of polycrystalline silicon carbide, said composite structure (100) having a density of defects of crystalline origin less than or equal to 0.25 defects / cm 2and an overall density of bubbles and holes less than or equal to 0.2 defects / cm 2 .

Images