METHOD FOR PRODUCING A CONTINUOUS NITRIDE LAYER

Description

TECHNICAL FIELD

[0001] The present invention relates to the fabrication of continuous layers made of a nitride (N) preferably obtained from at least one of gallium (Ga), indium (In), and aluminum (Al). The invention finds application, for example, in the field of optoelectronic devices comprising a plurality of micrometer-sized light-emitting diodes (LEDs), generally called micro-LEDs. In this field, a particularly advantageous use of the invention concerns the fabrication of small screens generally called micro-displays. STATE OF THE ART

[0002] Many microelectronic and optoelectronic applications require the fabrication of nitride layers using at least one of gallium, indium, or aluminum. Specific applications include, for example, the fabrication of micro-LEDs.

[0003] A nitride layer is classically obtained by epitaxial growth from a crystalline layer covering a plate.

[0004] A major challenge is minimizing the defect density in the nitride layer obtained by epitaxy. Indeed, the performance of microelectronic or optoelectronic devices made from these nitride layers is highly sensitive to the density of structural defects such as dislocations.

[0005] These dislocations originate from the difference in lattice parameter between the epitaxial layer and the substrate, as well as the coalescence of small grains that are formed at the beginning of growth; these grains are slightly disoriented with respect to each other and they join together, forming structural defects at the coalescence joint, including dislocations, which can then traverse the entire epitaxial structure.

[0006] The most direct way to solve these problems is to use substrates of the same nature as the layers to be epitaxially treated (homo-substrates). However, these substrates are either not commercially available or are still only small in size and very expensive, which makes it impossible to cut substrates of sufficient size for the intended industrial applications.

[0007] The solutions currently considered for industrial applications are therefore primarily based on the use of heterosubstrates in combination with methods known as "lateral overgrowth" or ELOG, an acronym for "epitaxial lateral overgrowth." This method, based on the use of a mask to block dislocations, reduces their density. However, these dislocations are distributed non-uniformly, which can pose a problem during device fabrication.

[0008] Another solution is to grow material by epitaxy onto pre-existing patches of that material: this is the so-called pendeo-epitaxy process which makes it possible to avoid growth on the mask. In However, conventional pendeo-epitaxy solutions do not eliminate, or even significantly reduce, the appearance of defects generated by the coalescence of adjacent germs.

[0009] Patent application WO2019122461 describes an illustrated solution to Figures 1A to 1D : a. A first step consists of providing a stack comprising a creep layer 200, a crystalline layer 300, and a nitride layer 500 or priming layer 500, typically a gallium nitride (GaN) layer. A buffer layer 400 may also be present between the creep layer and the nitride layer 500. This step is illustrated in Figure 1Ab. A second step consists of forming 1000A1-1000A4, 1000B1-1000B4 pads by etching the crystalline layer 300, possibly the buffer layer 400, and at least a portion of the creep layer 200 ( figure 1B ). Each plot thus comprises a creep segment 220A1-220A4, 220B1-220B4 originating from creep layer 200 and a crystalline segment 300A1-300A4, 300B1-300B4 originating from crystalline layer 300. c. Grow by epitaxy crystallites 510A1-510A4, 510B1-510B4 on plots 1000A1-1000A4, 1000B1-1000B4 until their coalescence ( Figures 1C and 1D ). The layer thus formed can continue its growth by thickening.

[0010] During their epitaxial growth, the crystallites formed at the top of the pads coalesce to form a continuous layer destined to form an optoelectronic or electronic device. During epitaxy, the portion of the pad formed by the creep segment reaches (or exceeds) its glass transition temperature or a temperature very close to it. Under mechanical stress, this portion of the pad can thus deform ( figure 1EThus, when two crystallites come into contact and coalesce, the mechanical stresses generated by this contact are transferred to the pads and therefore to the creep segments supporting them. These segments deform, thereby absorbing some of the mechanical stresses. This mobility of the pads, due to the presence of the creep segments, allows them to accommodate relative misorientations of the crystallites, whether in the plane in which the substrate mainly extends ("twist") or out of plane ("tilt").

[0011] However, it appears that this method does not allow for obtaining satisfactory defect densities. As illustrated in Figures 2A and 2B , continuous layers obtained by such a process exhibit in places very high orientation angles of the material at the base of the nitride layer, reflecting the existence of structural defects which are detrimental to the manufacture of devices from these continuous layers.

[0012] Therefore, there is a need to limit or even eliminate the drawbacks of known solutions. The present invention aims, in particular, to provide a solution to meet this need.

[0013] An objective of the present invention is to propose a solution for obtaining a nitride (N) layer obtained from at least one of gallium (Ga), indium (In) and aluminium (Al), exhibiting a significantly reduced defect density.

[0014] Another prior art is known from US 2022 / 0251730 A1 and US 2003 / 0030068 A1. SUMMARY

[0015] To achieve this objective, according to one embodiment, a process is provided for obtaining a layer made at least in part of a nitride (N) preferably obtained from at least one of gallium (Ga), indium (In) and aluminium (Al), comprising the following steps: a. provide a stack comprising at least one set of pads extending from a substrate, each pad comprising at least: i. a first segment, called the creep segment, formed in an amorphous material having a glass transition temperature Tglass transition, ii. a second, crystalline segment, called the crystalline segment, overlying the creep segment, b. grow by epitaxy a crystallite on at least some of said pads and continue the epitaxial growth of the crystallites until coalescence of the crystallites carried by the pads of the set of pads, so as to form a nitride layer,

[0016] The process is characterized in that the plots of said set of plots are distributed on the substrate in such a way that the relative arrangement of the plots of the set of plots is such that during the epitaxy of the crystallites, the progressive coalescence of the crystallites always takes place between on the one hand a set, called a coalesced set, comprising a crystallite or a plurality of coalesced crystallites and on the other hand at least one crystallite, called an isolated crystallite, which has not already coalesced with other crystallites.

[0017] Thus, at a point of coalescence, the progressive coalescence of crystallites always occurs between, on the one hand, a crystallite or a plurality of coalesced crystallites, and on the other hand, a single crystallite, which can also be described as an isolated crystallite since it has not previously coalesced with one or more other crystallites. The creep segment underlying this isolated crystallite possesses all the necessary mobility to compensate for any misalignments and misorientations between the isolated crystallite and the crystallite or the plurality of coalesced crystallites with which it coalesces. This prevents the formation of structural defects within the crystallites, and therefore within the continuous layer, during successive coalescences, ultimately resulting in a continuous layer of excellent quality.

[0018] Electronic and optoelectronic devices, for example LEDs or vertical devices such as vertical transistors or lateral components such as HEMT transistors, exhibiting very good performance can thus be manufactured from these layers exhibiting low densities of structural defects. BRIEF DESCRIPTION OF THE FIGURES

[0019] The aims, objects, features and advantages of the invention will become clearer from the detailed description of an embodiment thereof, which is illustrated by the following accompanying drawings in which: THE figures 1A to 1E illustrate steps in a process for forming a continuous layer according to the prior art. figure 2A is a photograph obtained by scanning electron microscopy (SEM) which illustrates continuous layers obtained by the process illustrated in figures 1A to 1E . There figure 2Bis an image obtained by dark-field X-ray microscopy showing the orientation variations of the crystallites obtained by the process illustrated in figures 1A to 1E . THE figures 3A to 3H illustrate steps in a process according to a non-limiting example of the process according to the present invention. figure 3A illustrates an example of stacking from which an example of a method according to the invention can be implemented. figure 3B illustrates the stacking of the figure 3A on which a primer layer is formed. The figure 3C illustrates the result of a step consisting of forming a set of blocks from the stacking of the figure 3A or that of the figure 3B . There 3D figure This illustrates a phase of epitaxial growth of crystallites on the tops of the bumps, this growth phase not yet being complete. figures 3E to 3H illustrate the successive coalescence of crystallites to form a continuous layer. figures 4A to 4Fillustrate examples of block arrangements enabling the fabrication of continuous layers with a low defect density. Figures 4A and 4B illustrate an example in which the plots are arranged in a logarithmic spiral. figure 4C This illustrates an example in which some of the plots are arranged along several parallel alignment axes. figure 4D This illustrates an example in which the plots are arranged along an alignment axis. figures 4E And 4F illustrate an example in which the plots are arranged along five intersecting alignment axes. Figures 5A and 5B represent the coalescence of two continuous layers themselves previously formed by the coalescence of crystallites. figure 6A is a SEM image of four assemblies, each consisting of three silicon pads (creep segment) and GaN (crystalline segment overlying the creep segment), all supported by a silicon-on-insulator (SOI) substrate. figure 6B is a SEM image of the sets illustrated on the figure 6A after the coalescence of the plots within the same set. The figure 7 This illustrates results obtained by X-ray diffraction analysis, showing the orientation of GaN layers obtained by coalescence of the pads. Figures 8A and 8B illustrate results obtained by X-ray diffraction analysis showing the orientation of silicon segments after coalescence of the pads.

[0020] The drawings are provided by way of example and are not limiting to the invention. They constitute schematic representations of principle intended to facilitate understanding of the invention and are not necessarily to scale with practical applications. In particular, the relative thicknesses of the various layers, sections, crystallites, and continuous layers are not representative of reality. DETAILED DESCRIPTION

[0021] Before beginning a detailed review of embodiments of the invention, optional features which may possibly be used in combination or alternatively are stated below: According to an advantageous embodiment, the pads carrying the crystallites forming the nitride layer are substantially arranged, in projection in a horizontal plane in which mainly the upper face of the substrate extends, in a logarithmic spiral.

[0022] According to one embodiment, the pads carrying the crystallites forming the continuous nitride layer are, in projection into a horizontal plane in which mainly the upper face of the substrate extends, substantially aligned along at least one axis called alignment.

[0023] According to one embodiment, the plots arranged on the same alignment axis are spaced at increasing distances by traversing the alignment axis in a given direction from a given point, called the initial point.

[0024] According to one example, the pads bearing the crystallites forming the continuous layer are separated by a distance varying logarithmically from one to the next, said distance being measured along their alignment axis.

[0025] According to one example, the pads bearing the crystallites forming the continuous nitride layer are substantially aligned, in projection into a horizontal plane in which mainly the upper face of the substrate extends, along intersecting axes, the intersecting axes verifying a rotational symmetry with each other around a central point in projection into the horizontal plane.

[0026] In one example, the posts are aligned along five intersecting axes.

[0027] According to one example, epitaxial growth is carried out at an epitaxial temperature Tepitaxis, such that Tepitaxis > k1 x Tglass transition, with k1 ≥ 0.8.

[0028] According to one example, k 1 ≥ 1, and preferably k 1 ≥ 1.5.

[0029] Advantageously, the set of studs includes at least 3 studs.

[0030] According to one embodiment, the stacking comprises a plurality of sets of studs on the substrate, a continuous nitride layer being formed from each set of studs, the epitaxial growth of the crystallites being stopped before the crystallites belonging to two distinct sets of studs coalesce, so that the continuous layers formed from each set of studs are spaced apart from each other.

[0031] In this embodiment, where several assemblies allow the formation of multiple continuous and disjoint layers on the same plate, these different continuous layers are typically called "vignettes." Each of these vignettes can then be used in the implementation of an electronic or optoelectronic arrangement, such as a LED or micro-LED or a vertical device such as a vertical transistor. The characteristics described below with reference to a set of pads, its crystallites and the continuous layer formed from this set of pads apply mutatis mutandis to other sets of pads, the crystallites they carry and the continuous layers formed by coalescence of these crystallites.

[0032] According to one example, each plot has a top face in which the epitaxial growth of crystallites takes place at least in part and preferably only from said top face.

[0033] According to one example, the process further includes a training step in the continuous layer of a device, the device being taken from among an LED and a transistor, for example a vertical transistor, for example a HEMT type transistor.

[0034] According to one embodiment, the studs of said set of studs are distributed on the substrate so that the relative arrangement of the studs of the set of studs is such that during the epitaxy of the crystallites, at least at certain times the progressive coalescence of the crystallites takes place between on the one hand said coalesced set and on the other hand several isolated crystallites.

[0035] Alternatively, the plots of said set of plots are distributed on the substrate so that the relative arrangement of the plots of the set of plots is such that during the epitaxy of the crystallites, the progressive coalescence of the crystallites always takes place between, on the one hand, said coalesced set and, on the other hand, a single isolated crystallite.

[0036] According to an example, each plot has a cross-section whose maximum dimension d plot is between 10 and 500 nm (10 -9 < meters), the maximum dimension d plot being measured in a plane parallel to a horizontal plane in which the upper face of the substrate extends mainly, preferably 20 nm ≤ d plot ≤ 200 nm and preferably 50 nm ≤ d plot ≤ 100 nm.

[0037] According to a preferred example, the continuous layer has a cross-section whose maximum dimension d layer is between 0.5 and 10 µm (10 -6< meters), the maximum dimension d layer being measured in a plane parallel to a horizontal plane in which the upper face of the substrate extends mainly, preferably 0.8 µm ≤ d layer ≤ 3 µm and preferably 1 µm ≤ d layer ≤ 2 µm.

[0038] Preferably, the creep layer is made of a material chosen from: a. a silicon oxide SixOy, x and y being integers, and preferably the creep layer is in SiO2, b. a glass, c. a borosilicate glass, d. a borophosphosilicate glass (BPSG).

[0039] Advantageously, T epitaxy ≤ k 2 x T melting min , T melting min being the lowest melting temperature among the melting temperatures of the segments forming the pads, with k 2 ≤ 0.9 and preferably k 2 ≤ 0.8.

[0040] According to an advantageous embodiment, the pads comprise at least one buffer layer overlying the crystalline layer, and made of a material different from that of the continuous nitride layer; preferably, the continuous nitride layer is made of gallium nitride (GaN) and the buffer layer is made of aluminum nitride (Ain). This prevents alteration phenomena of the GaN layer by reaction with the silicon layer, phenomena usually referred to as "melt-back etching".

[0041] According to one example, the plots include, prior to the epitaxial growth stage of continuous nitride layers, at least one priming layer, surmounting said buffer layer and made of gallium nitride (GaN).

[0042] According to an advantageous example, providing said stacking includes providing an elaborate silicon-on-insulator (SOI) substrate comprising a base substrate successively surmounted by an oxide layer forming said creep layer and a semiconducting layer forming said crystalline layer.

[0043] According to an advantageous example, the creep section has a height e 220 such that e 220 ≥ 0.1xd plot, A plot being the diameter of the plot or more generally the edge-to-edge distance of the plot taken, at the level of the creep section and in a direction parallel to a horizontal plane in which mainly extends a top face of the substrate, preferably e 220 ≥ 1xd plot.

[0044] According to one example, the pad formation step includes etching the crystalline layer and etching only a portion of the creep layer so as to retain a portion of the creep layer between the pads.

[0045] In the following description, the terms crystals and crystallites will be considered equivalent.

[0046] It is specified that, within the context of the present invention, the terms "on," "overhangs," "covers," or "underlying," or their equivalents, do not mean "in contact with." Thus, for example, "the deposition of a first layer on a second layer" does not necessarily mean that the two layers are directly in contact with each other, but rather that the first layer at least partially covers the second layer, either by being directly in contact with it or by being separated from it by at least one other layer or at least one other element, including air. Similarly, "a stud overhanging a first layer" does not mean that the stud is necessarily in contact with that first layer, but rather that the stud is either in contact with that first layer or in contact with one or more layers arranged between the first layer and the stud.

[0047] The formation stages of the different layers and regions are understood in a broad sense: they can be carried out in several sub-stages which are not necessarily strictly successive.

[0048] In the following description, thickness or height is measured perpendicular to the principal faces of the different layers. In the figures, thickness or height is measured vertically or along the z-axis of the orthogonal coordinate system illustrated in figures 1A to 1E , 3A to 3H And 5A and 5B .

[0049] Similarly, when we indicate that an element is located opposite another element, this means that these two elements are both located on the same line perpendicular to the main plane of the substrate, or on the same vertically oriented line (z-axis) in the figures.

[0050] A substrate, layer, or device is defined as being "based" on a material. M, an Msubstrate, a layer, a device comprising this material M only or this material M and possibly other materials, for example alloying elements, impurities or dopant elements.

[0051] The terms "approximately," "about," and "on the order of" mean "within 10%" or, when referring to angular orientation, "within 10°." Thus, a direction approximately normal to a plane means a direction at an angle of 90±10° to the plane.

[0052] The term "micro-LED" refers to an LED in which at least one dimension, measured in a plane parallel to the principal plane in which the substrate supporting the micro-LED extends (i.e., the XY plane of the orthogonal coordinate system referenced in the figures), is micrometric, that is, strictly less than 1 mm (10⁻³ meters) and preferably less than or equal to 100 µm (1 µm = 10⁻⁶ meters). In the context of the invention, the micro-LEDs, when projected onto a principal extension plane parallel to the main faces of the micro-LEDs, i.e., parallel to a top face of the substrate, have maximum dimensions of micrometric size in the plane. Preferably, these maximum dimensions are less than a few hundred micrometers.

[0053] In the present invention, "vertical devices" means devices in which the load carriers move within the thickness of the layer, that is, in a direction perpendicular to the planes in which the principal faces of this layer extend.

[0054] In the present invention, "HEMT" (High Electron Mobility Transistor) refers to field-effect transistors with high electron mobility, sometimes also called heterostructure field-effect transistors. Such a transistor comprises two superimposed semiconductor layers with different band gaps that form a quantum well at their interface. Electrons are confined within this quantum well to form a two-dimensional electron gas. For reasons of high-voltage and high-temperature resistance, the materials of these transistors are chosen to exhibit a wide band gap.

[0055] An example of a method for forming a continuous layer according to the invention will now be described with reference to figures 3A to 3H .

[0056] It should be noted that, in one example, the process described below allows the formation of a single continuous layer on a substrate or on a wafer. In another example, the process described below allows the formation of multiple continuous layers on the same substrate or on the same wafer. In this latter case, each layer is separate from the others and can be called a vignette. Each vignette can then be used to fabricate a device such as an LED or a micro-LED, or any other vertical nitride-based device such as a HEMT transistor.

[0057] As illustrated in figure 3A , we provide a stack comprising at least one substrate 100, successively surmounted by a creep layer 200 and a crystalline layer 300. Thus the creep layer 200 is disposed between the substrate 100 and the crystalline layer 300.

[0058] According to one embodiment, the substrate 100 is silicon-based, amorphous or crystalline. It ensures the mechanical strength of the stack.

[0059] The crystalline layer 300 has a lower face 302 facing the creep layer 200 and an upper face 301 whose function is to serve as a base layer for the epitaxial growth of the nitride in the continuous layer 550A that is ultimately desired. For example, the desired final layer is a gallium nitride (GaN) layer. In one embodiment, the crystalline layer 300 is based on single-crystal silicon. Alternatively, the crystalline layer 300 can be based on SiC or Al₂O₃.

[0060] Preferably, the creep layer 200 is made of a viscous material. The creep layer 200 has a glass transition temperature. It therefore exhibits a glass transition and the behavior of glass transition materials. Like all materials with a glass transition temperature, the creep layer 200, under the effect of a temperature increase, deforms without breaking and does not return to its initial position after a temperature decrease. In contrast, the crystalline layer 300 does not naturally exhibit a glass transition. The crystalline layer 300 deforms, then disintegrates and can break. Consequently, the creep layer 200 and the crystalline layer 300 are different. The creep layer 200 is not crystalline.

[0061] The creep layer 200 is made of an amorphous material such as an oxide, preferably a silicon oxide SixOy, such as SiO2. The role of this layer will be explained later in the description.

[0062] Advantageously, but not limitingly, this stack comprising the substrate 100, the creep layer 200, and the crystalline layer 300 constitutes a semiconductor-on-insulator (SOI) substrate, preferably silicon-on-insulator (SOI). In this case, the creep layer 200 is formed by the buried oxide layer (BOX) of the SOI substrate.

[0063] According to an example of an advantageous implementation illustrated in figure 3AA buffer layer 400 is deposited epitaxially onto the upper face 301 of the crystalline layer 300. When the desired continuous layer 550A is made of GaN and the crystalline layer 300 is silicon-based, this buffer layer 400 is typically aluminum nitride (AIN). This avoids the phenomenon known as "melt-back etching," generated by the very high reactivity between silicon and gallium at typical epitaxial temperatures (1000 / 1100°C), which leads to significant degradation of the continuous GaN layer 550A.

[0064] Typically, the thickness of the AIN layer is between 10 and 200 nanometers (10⁻⁹ meters).

[0065] As illustrated in figure 3BAlternatively, a priming layer 500 can be deposited epitaxially on the upper face 401 of the buffer layer 400. The function of this priming layer 500 is to facilitate the resumption of growth of the 510A1-510A5 crystallites during subsequent steps. In this case, the epitaxial growth of the 510A1-510A5 crystallites, illustrated in Figure 500, occurs at least partially from one of the upper faces of the priming layer 500. 3D figure This 500 primer layer is preferably made of the same material as the 550A continuous layer that is ultimately desired. Typically, when the 550A continuous layer material is gallium nitride (GaN), the 500 primer layer is also made of GaN. This 500 primer layer typically has a thickness between 50 and 200 nanometers.

[0066] For the sake of brevity and clarity, only five 1000A1-1000A5 pads are shown in the figures to support a continuous 550A layer. Naturally, a continuous 550A layer can be formed on a larger number of pads. As will be described later, the number of pads and their period will be adjusted according to the desired size of the micro-LED.

[0067] Note that the 400 buffer and 500 priming layers are optional. Thus, according to embodiments not illustrated in figures 3A to 3H , we can plan only the 400 buffer layer or only the 500 priming layer, or neither of these two layers 400 and 500.

[0068] As illustrated in figure 3C , then 1000A1-1000A5 pads are formed from the stack. These pads are obtained by etching the stack into the creep layer 200, at least part of the etching extending within the creep layer 200.

[0069] To form the blocks by etching, numerous etching techniques known to those skilled in the art can be used. These include classic lithography techniques, such as photolithography, which involves creating a mask, for example in resin, and then transferring the patterns from the mask into the stack. Electron beam lithography (e-beam) or nanoscale printing techniques can also be employed.

[0070] These 1000A1-1000A5 pads are small and can be described as nano-pads. Typically, the maximum dimension of the pad cross-section, taken in a plane parallel to the horizontal XY plane of the orthogonal XYZ coordinate system or to the plane of the upper face 101 of the substrate 100, and referenced as d pad in figure 3CThe maximum dimension, dplot, is between a few tens and a few hundred nanometers. More precisely, dplot is between 10 and 500 nanometers, and preferably between 20 and 150 nm, and preferably between 50 and 100 nm, for example, on the order of 50 nm or 100 nm. If the pads have circular cross-sections, this maximum dimension, dplot, corresponds to the diameter of the pads. If the pads have hexagonal cross-sections, this maximum dimension, dplot, corresponds to the diagonal or the diameter of the circle passing through the corners of the hexagon. If these pads have rectangular or square cross-sections, this maximum dimension, dplot, corresponds to the longest diagonal. It is understood that the cross-section of the pads may have a shape other than circular, hexagonal, rectangular, or square.

[0071] Within the framework of the present invention, the 1000A1-1000A5 pads are distributed in a particular way in projection in the XY plane: their arrangement is such that, as will be illustrated later, the coalescences occurring successively during the implementation of the process involve, on the one hand, a single crystallite 510A1 (during the first coalescence which will be described later with reference to the figure 3E ) or a plurality of already coalesced crystallites 510A1-510A4 (during subsequent coalescences, figures 3F to 3H), and on the other hand, at least one isolated 510A2-510A5 crystallite. In other words, the arrangement of the 1000A1-1000A5 plots prevents any coalescence between, on the one hand, a first plurality of coalesced crystallites and, on the other hand, a second plurality of coalesced crystallites. To achieve this, the 1000A1-1000A5 plots are not all equidistant from their nearest neighbors. The arrangement of the 1000A1-1000A5 plots must take into account the increase in the projected dimensions in the XY plane of the 510A1-510A5 crystallites as they grow by epitaxy. Examples of the 1000A1-1000A5 plot arrangement will be given later with reference to figures 4A to 4F .

[0072] As illustrated in figure 3CThe pads are etched through the entire 500 initiator layer, the entire 400 buffer layer (when present), and the entire 300 crystalline layer. Preferably, only a 220 portion of the 200 creep layer is etched. This embodiment has the advantage of preventing the nitride of the 550Ase continuous layer from growing on the 220 creep segments during epitaxy. This epitaxial selectivity is particularly relevant when the 550A nitride continuous layer being grown epitaxially is GaN and the creep segments are SiO2. Conversely, if the creep layer 200 is etched through its entire thickness using the same materials, then during epitaxy, the nitride of the continuous layer 550A develops from the upper face 101 of the substrate 100, which is usually made of silicon. This situation is obviously undesirable.

[0073] Furthermore, it has been observed that preserving an unetched portion 210 of the creep layer 200 facilitates the creep of the section 220, particularly when the crystallites are twist-oriented, i.e., in principal planes of extension of the continuous layer 550A that one wishes to obtain. These principal planes of extension of the continuous layer 550A are parallel to the horizontal XY plane of the XYZ coordinate system.

[0074] Preferably, the engraved thickness e 220, which forms the height of the creep section 220, is equal to half the thickness of the creep layer 200. This allows for very good reorientation of the crystallites during the formation of grain boundaries.

[0075] There 3D figure illustrates the formation of 510A1-510A5 crystallites by epitaxial growth from the 500 priming layer (or from the top face 301 of the 300 crystalline layer when the 400 and 500 layers are absent).

[0076] As illustrated on this 3D figure The 1000A1-1000A5 pads each support a 510A1-510A5 crystallite carried by a stack of 400A1-400A5, 300A1-300A5, and 220A1-220A5 segments. The segments extend along the principal direction of pad extension, i.e., vertically (Z) on the figures 3A to 3H .

[0077] The segments form circular cylinders if the cross-section of the studs is predominantly circular. If the cross-section of studs 1000A1-1000A5 is polygonal, for example hexagonal, the segments then form cylinders with a hexagonal cross-section. Preferably, the segments are solid. The cross-section of the studs is taken parallel to the horizontal XY plane, that is, parallel to the planes in which the creep layer 220 and the crystalline layer 300 extend predominantly.

[0078] Regardless of the embodiment chosen, i.e., with or without a 500 initiator layer and with or without a 400 buffer layer, the epitaxial growth of the 510A1-510A5 crystallites occurs at least partially or entirely from the upper face 1001A1-1001A5 of the 1000A1-1000A5 block. Thus, this upper face 1001A1-1001A5 is formed either by the 300A1-300A5 crystal segment, or by the 500A1-500A5 segment formed by the 500 initiator layer, or by the 400A1-400A5 segment formed by the 400 buffer layer. This allows, in particular, for the rapid production of 510A1-510A5 crystallites of significant thickness.

[0079] It will be noted that the upper faces 401, 501 of the buffer layer 400 and the priming layer 500, that is to say the faces turned with regard to the continuous layer 550A which we wish to grow, have Gallium (Ga) type polarities, and not nitrogen (N), which considerably facilitates the obtaining of a high quality continuous 550A layer of epitaxial nitride.

[0080] The growth of crystallites 510A1-510A5 continues and extends laterally, particularly along planes parallel to the xy plane. Crystallites 510A1-510A5 grow until those borne by the two closest pads, referenced 1000A1 and 1000A2 on the figures 3C to 3H and separated by a distance P 1, coalescent ( figure 3EThis coalescence occurs with few or no dislocations within the crystallites. It allows the formation of a continuous 500A layer from the 510A1 and 510A2 crystallites. This continuous 500A layer constitutes a coalesced unit.

[0081] The growth of the crystallites involved in this first coalescence 510A1, 510A2, as well as the other crystallites 510A3-510A5, continues until the continuous layer 500A formed by the crystallites 510A1 and 510A2 coalesces with the crystallite 510A3 carried by the nearest 1000A3 block. The 510A3 crystallite is thus added to the coalesced set. It should be noted that, due to the distribution of the 1000A1-1000A5 blocks on the substrate, from the beginning of its growth ( 3D figure ) and up to this stage of the process ( figure 3F), the 510A3 crystallite, being coalesced with the continuous 550A layer, has never coalesced with another crystallite. Its 1000A3 plot therefore exhibits all the necessary mobility to accommodate misalignment in the XY plane or out of plane with the continuous 550A layer formed by the first two coalesced crystallites 510A1, 510A2. On the figure 3F A deformation of the 220A3 creep section according to its tilt angle is illustrated. A deformation of this 220A3 creep section according to its twist angle can occur alternately or simultaneously. The properties of this 220A3 creep section allow for a significant reduction of defects at the interface between the 550A continuous layer and the 510A3 crystallite. Explanations regarding the benefits of this precaution are provided later.

[0082] The continuous layer 500A is then formed by the crystallites 510A1, 510A2 and 510A3.

[0083] The growth of the continuous layer 500A continues, as does that of the as-yet-uncoaled crystallites 510A4, 510A5. A new coalescence occurs when the continuous layer 500A comes into contact with the crystallite 510A4 ( figure 3G ). The creep section 220A4 of the latter deforms in such a way that the entire 1000A4 block deforms and thus accommodates any misalignment and misorientation of the continuous layer 500A with the crystallite 510A4. The crystallites 510A4 and 510A5 are thus added to the coalesced whole.

[0084] Finally, the figure 3H illustrates a coalescence allowing the 510A5 crystallite to be integrated into the continuous 500A layer (i.e. into the coalesced whole).

[0085] As can clearly be seen from the figures, the continuous layer 550A extends at the end of the process between several 1000A1-1000A5 pads and forms a continuous layer.

[0086] It is understood that the figures 3A to 3Hillustrate the example of a continuous 500A layer formed by adding a single 510 crystallite at each coalescence step, but it is perfectly conceivable that several 510A1-510A5 crystallites could coalesce simultaneously with the continuous 500A layer, provided that these are not themselves already coalesced with each other or with one or more other crystallites. Such an example will be described further with reference to the figure 4F .

[0087] THE figures 3A to 3HThese figures are not intended to provide a realistic visual representation of the deformation of the creep segments. They aim to provide a schematic representation to facilitate understanding of the principle of crystallite coalescence without dislocation at the coalescence boundaries. Specifically, the deformations of the creep segments are not shown in all planes for the sake of brevity. Furthermore, the deformations of the patches have been exaggerated for illustrative purposes. The upper surface of the continuous layer obtained at the end of the process has a much less rough, even smooth, appearance than depicted.

[0088] The growth of the 510A1-510A5 crystallites does not extend downwards. Furthermore, this growth is selective in that it does not occur on the creep layer 200, which is typically made of an oxide. InIn this sense, the growth of 510A1-510A5 crystallites occurs according to the principle of pendeo-epitaxy.

[0089] It should be noted that it is particularly advantageous to etch the 1000A1-1000A5 pads after the formation by epitaxy of the buffer layer 400 and the priming layer 500 (when these layers are present). In Indeed, if one of these layers 400, 500 were deposited after etching, it would form at least in part between the pads 1000A1-1000A5 on the upper face of the creep layer 200. In the case where the epitaxial nitride is GaN, and the creep layer 200 is SiO2, then, at the temperature of the epitaxial deposition, the epitaxial growth of the continuous nitride layer 550A would not occur selectively but would, on the contrary, also take place between the pads 1000A1-1000A5, which is naturally not desirable.

[0090] In a particularly advantageous way, the epitaxy temperature T at which the epitaxy is carried out is greater than or on the order of the glass transition temperature T of the creep layer 200. Thus, during the epitaxy, the creep sections 220A1-220A5 are brought to a temperature which allows them to deform.

[0091] Therefore, if the 510A11-510A12 crystallites carried by two adjacent 1000A1-1000A2 patches are misoriented with respect to each other, during the coalescence of these two crystallites, the joint formed at their interface, usually called a grain boundary or coalescence boundary, will form without dislocations to compensate for these misorientations. The deformation of the creep segments 220 thus allows these misorientations to be corrected and a continuous 550A layer to be obtained with little or no dislocations at the coalescence boundaries.

[0092] The maximum dimension of a continuous layer measured parallel to the xy plane is denoted by dlayer. Thus, dlayer corresponds to the maximum dimension of a projection of the continuous layer onto a plane parallel to the xy plane. Preferably, 0.8 µm ≤ dlayer ≤ 3 µm and preferably 1 µm ≤ dlayer ≤ 2 µm. dlayer depends on the rate and duration of epitaxial growth, as well as the number and size of the spots.

[0093] The distances (for example, referenced as P1, P2, P3, P4... in the figures) separating two pads whose overlying crystallites are successively coalesced to the continuous layer are typically greater than 300 nm, and for example, between 300 nm and 10 µm, or between 300 nm and 5 µm. These distances can be chosen, in particular, according to the size of the component to be manufactured from the continuous layer formed at the end of the process. The increase in the distance between pads as one considers pads with increasingly higher refractive indices is typically more rapid for a small component than for a large component. Indeed, for a large component, the pads could then be very far apart.However, achieving coalescence between crystallites from widely separated pads can be complex, particularly because the continuous layer is quite thick at later stages of the process. It should be noted, however, that a significant thickness of the continuous layer may be desirable for the fabrication of certain devices, for example, in power electronics. Therefore, coalescence between crystallites and the continuous layer is possible when the nitride layer thickness is substantial (potentially up to several tens of micrometers).

[0094] The process for producing the continuous 550A layer can be stopped at the end of the figure 3H Alternatively, this process can be continued to form a micro-LED from the continuous 550A layer.

[0095] According to a non-limiting embodiment not shown, quantum wells are fabricated within the 550A continuous layer. This embodiment advantageously allows for the direct fabrication of a micro-LED with a size corresponding to the initial size of the continuous layer. Those skilled in the art can implement known prior art solutions to fabricate quantum wells within the 550A continuous layer. Thus, once the 510A1-510A5 crystallites have coalesced, the same growth conditions for the wells are adopted as in conventional two-dimensional growth.

[0096] The smallest possible size for micro-LEDs depends on the ultimate resolution of the chosen structuring methods: for example, for arrays of dots fabricated by nanoimprinting, dot sizes of 50 nm and spacing between dots of 100 to 150 nm are achieved. This means that continuous layer dimensions of 1 to 2 µm are obtained. This is therefore on the order of the pixel sizes sought for high-resolution micro-displays. Explanation of the principle of coalescence without dislocation or with few dislocations within continuous layers

[0097] As mentioned above with reference to figures 3E to 3H , the formation of the continuous 500A layer occurs through the successive coalescences of isolated 510A1-510A5 crystallites.

[0098] As illustrated in Figures 5A and 5B The coalescence of already coalesced crystallite assemblages induces defects at the coalescence joints.

[0099] THE Figures 5A and 5BThey illustrate the example of two continuous layers 500A', 500B', each obtained by the coalescence of two crystallites (510A1' and 510A2' on the one hand, 510B1' and 510B2' on the other), each supported by a pad (1000A1' and 1000A2' on the one hand, 1000B1' and 1000B2' on the other). The crystal planes of these two layers exhibit an angular orientation defect identified by the angle β.

[0100] One might expect that, as in the case of the coalescence of two isolated crystallites, this orientation defect would be compensated by the properties of the creep segments of each of the particles. However, it has been observed that no compensation occurs, or at least does so to an insufficient extent to satisfactorily limit the coalescence defects.

[0101] This can be explained by the following factors. During the initial coalescence of a crystallite with a neighboring crystallite, the underlying creep segment deforms according to the relative misorientations of these two initial crystallites. After this initial coalescence, the creep segment is therefore mechanically constrained in a direction dependent on these misorientations.

[0102] If a new coalescence occurs, two cases can then be distinguished: a. The coalescence of the continuous layer occurs with an isolated crystallite (a crystallite that has not yet coalesced with one or more other crystallites is called an isolated or single crystallite): the creep segment underlying this isolated crystallite exhibits considerable mechanical freedom and can, on its own, compensate for the misorientations between the crystal lattices of the continuous layer and the isolated crystallite. It is understood that deformation of the creep segments underlying the continuous layer is also possible, but to a much lesser extent than the deformation of the creep segment underlying the isolated crystallite. b. The coalescence of the continuous layer occurs with another continuous layer, also obtained by the coalescence of crystallites carried by several pads (case illustrated in Figure 1). figure 5BThe various creep segments are then excessively constrained due to their mechanical dependence on the continuous layer and other creep segments. Their presence is insufficient to compensate for the misalignments between the two continuous layers, and these coalesce, causing numerous defects within the crystallites, particularly at the coalescence boundary. It is understood that the creep segments underlying the two continuous layers can deform, but to an extent too small to compensate for the misalignments of the continuous layers on their own.

[0103] In other words, the creep segments of the 1000A1', 1000A2', 1000B1', 1000B2' blocks supporting the crystallites involved 510A1', 510A2', 510B1', 510B2' in a coalescence between already coalesced crystallites exhibit much lower mobility than when each of them supports an isolated crystallite. It is this difference in mobility that explains why the principle of coalescence of raw crystallites by pendeo-epitaxy does not apply satisfactorily to a network of blocks with arbitrary relative arrangements.

[0104] THE Figures 5A and 5B illustrate the fact that an out-of-plane misalignment of continuous layers 500A', 500B' is not compensated by the presence of creep segments, but it is understood that the same observation can be made for a misalignment in the xy plane.

[0105] As explained above, it has been observed that if an isolated crystallite is involved in coalescence, its creep segment alone compensates for any misalignments with the continuous layer. Therefore, the invention proposes to arrange the pads on the substrate surface in such a way that only this situation is encountered during the progressive formation of the continuous layer, and that the coalescence of two continuous layers, each originating from several crystallites, is avoided. Thus, according to the invention, at a point of coalescence, the progressive coalescence of the 510A1-510A5 crystallites always occurs between, on the one hand, a single 510A1 crystallite or a plurality of coalesced 510A1-510A4 crystallites, and on the other hand, an isolated 510A2-510A5 crystallite. This prevents the formation of structural defects within the crystallites and therefore within the continuous layer during successive coalescences, ultimately resulting in a continuous layer of excellent quality.

[0106] As previously stated, this does not preclude the possibility of multiple coalescences occurring simultaneously within the entire cluster. In this case, a plurality of crystallites that have already coalesced can coalesce simultaneously with a plurality of crystallites that are isolated from one another, that is, that have not previously coalesced with any other crystallite.

[0107] It is understood that any two crystallites forming part of the continuous layer created by successive coalescences are considered coalesced with each other, whether or not they were in direct contact with each other during the coalescence of one of them into the continuous layer. Generally, any crystallite in the continuous layer is considered coalesced with all the other crystallites forming the continuous layer, from the moment of its integration by coalescence into the continuous layer. The continuous layer forms, in effect, a coalesced whole within which the boundaries between crystallites cannot be clearly defined, particularly as the epitaxy progresses. Thus, the progressive coalescence of the crystallites, which is controlled during the process according to the invention, is understood as the successive addition of one or more crystallites to the continuous layer.In particular, it is entirely conceivable that during the process according to the invention, two crystallites forming part of the continuous layer may come into contact and coalesce again, this time directly. These coalescences can be referred to as internal coalescences. Internal coalescences do not affect the progressive coalescence of the crystallites, since no new crystallite is added to the continuous layer during an internal coalescence. As will be discussed later, precautions regarding the arrangement of the pads can be taken to optimize these internal coalescences.

[0108] It should be noted that a variation in growth rate can occur during the process due to unavoidable variations in the process itself (for example, due to precursor injection, temperature or pressure variations, etc.). Therefore, advantageously, it is anticipated that varying the distance between pellets from a center of symmetry of the coalescence will compensate for this variation in coalescence rate for the same set of pellets. In this way, even under the effect of a variation in the coalescence rate, the principle of continuous layer formation through successive coalescences of isolated crystallites is verified. Examples of plot arrangement allowing coalescence without dislocation or with few dislocations within continuous layers

[0109] Various examples of pad arrangement on substrate 100 to obtain a continuous layer with a low defect density will now be described with reference to figures 4A to 4F .

[0110] It is understood that the arrangement of the plots is understood as their distribution in space in projection in the xy plane, and particularly in projection on the upper face 101 of the substrate 100.

[0111] The distance between two neighboring plots is measured from one edge of one of these two plots to one edge of the other of these two plots, projected onto the xy plane, as illustrated in the figure 3C .

[0112] A first example of the arrangement of the 1000A1-1000A5 pads on substrate 100 is illustrated in Figures 4A and 4BIn this example, plots 1000A1-1000A5 are arranged in a logarithmic spiral. Such a spiral is a curve whose polar equation is of the form r = ab θ < , where a and b are positive real numbers and r is the radius measured between the center of the spiral and the point on the curve forming an angle θ with an arbitrary origin axis in the horizontal XY plane. Advantageously, therefore, plots 1000A1-1000A5 are arranged on a theoretical curve of this shape. The distribution of plots 1000A1-1000A5 on this theoretical curve is chosen based on the level of control of the processes involved, particularly the epitaxy process, and their variability. The better the control, the more the spacing between two consecutive plots on the curve can increase slightly as one moves along the curve from the origin.

[0113] There figure 4A illustrates a theoretical arrangement following such a model, while the figure 4B is a photograph MEB in view of top of plots arranged according to this embodiment.

[0114] This embodiment is particularly advantageous because it allows for a high density of spots on the substrate surface, leading to rapid crystallite coalescence. Indeed, with identical epitaxial growth parameters, the greater the density of spots, the faster the coalescence rate. The density of spots on the substrate surface is defined, projected onto the horizontal XY plane, as the ratio between the area occupied by the spots on the upper face 101 of the substrate 100 and the area of ​​this upper face 101 of the substrate 100.

[0115] According to another embodiment illustrated in the figure 4DThe plots are aligned along a single axis, called the alignment axis. A direction of alignment is defined for this alignment axis, as well as an initial point positioned on this same axis. A first plot, 1000A1, is located on this alignment axis, for example, at the initial point. The other plots are also located on this alignment axis, and the distance between two consecutive plots, 1000A1-1000A6, increases as one moves along the alignment axis in the direction of alignment. Denoting P1, P2, P3, P4, and P5 as the distances separating, respectively, the first 1000A1 and second 1000A2 plots, the second 1000A2 and third 1000A3 plots, the third 1000A3 and fourth 1000A4 plots, the fourth 1000A4 and fifth 1000A5 plots, and finally the fifth 1000A5 and sixth 1000A6 plots, the order of the plots being given according to the direction of alignment, we thus have: P1 <P 2 <P 3 <P 4 <p 5 . en notant i l'indice du plot sur l'axe d'alignement, avec allant de 1 à n, n étant le nombre plots disposés selon on a préférence, pour i≥2, p>1.1*P i-1 . Advantageously, the distance between plots evolves, on the alignment axis and according to the direction of alignment, according to a geometric sequence: for i≥1, P i =aq i< , a and q being positive real numbers, with preferably 2 ≤ q and / or q ≥ 1.1.

[0116] This embodiment can be described as an "online" arrangement.

[0117] According to another embodiment illustrated in the figure 4C , some of the plots are arranged along alignment axes as defined previously with reference to the figure 4D These alignment axes are parallel to each other. In order to allow for good initiation of coalescence at the beginning of the process, it is also planned that other plots do not necessarily have to be aligned on one of the alignment axes.

[0118] The studs can, for example, be arranged along axes parallel to each other and perpendicular to the previously mentioned alignment axes. These axes are called secondary alignment axes. Studs arranged along the same secondary alignment axis are preferably spaced at a constant interval. Furthermore, considering all secondary alignment axes parallel to all the alignment axes and in the same direction as these axes, the interval between studs aligned along the same secondary alignment axis increases. In this example, care is taken to ensure that the distance between two successive alignment axes (which can also be seen as the interval between studs along the secondary alignment axes) is always greater than the interval between studs aligned along the alignment axes. This is the example illustrated in the figure 4C .

[0119] According to an embodiment illustrated in figures 4E And 4F The plots are arranged along a plurality of intersecting alignment axes, for example five as shown. A central point can be defined, at which a plot may or may not be located, corresponding to the point of intersection of these alignment axes. The plots are then distributed along branches, each originating from the central point.

[0120] Advantageously, and as is highlighted in the figure 4F The plots are distributed on concentric circles.

[0121] On each of the branches, the plots can be arranged as described for a single alignment axis, the initial point of the alignment axis as defined in the previous embodiment then being the point of intersection of the alignment axes defined here.

[0122] This embodiment can be described as a "star" arrangement.

[0123] There figure 4F This figure highlights the order of crystallite coalescence in an example of this embodiment. It is a top view of a set of patches at different stages of epitaxial crystallite growth. The patches are represented by the black dots. The various concentric disks around each patch correspond to different stages of growth of the crystallite supported by that patch. More precisely, a disk corresponds to a stage of growth at which a coalescence (or several simultaneous coalescences) occurs. This figure takes into account internal coalescences as defined above.

[0124] In this example, a first plot 1000A1 is located at the central point of the plot set, and five second plots 1000A2 are located on a first circle centered on the central point of the plot set. An initial growth phase occurs and ends with the coalescence of the first plot 1000A1 with the five second plots 1000A2 (these five coalescences occurring approximately simultaneously). This coalescence forms a continuous layer.

[0125] It should be noted that during the first growth phase, the distance between the second 1000A2 pads and the third 1000A3 pads (and, more generally, any other higher pad) is sufficiently large that their respective crystallites do not coalesce at this stage of the process (i.e., P2 > P1). This ensures that the crystallites supported by the second 1000A2 pads are indeed isolated crystallites during their coalescence with the first plot 1000A1.

[0126] During a second growth phase, all the crystallites continue to grow until the continuous layer coalesces with the crystallites supported by the third 1000A3 pads (here, five in number). It should be noted again that during this second growth phase, the distance between the third 1000A3 pads and the fourth 1000A4 pads (and, more generally, any other higher pad) is sufficiently large that their respective crystallites do not coalesce at this stage of the process (i.e., P3 > P2). This ensures that the crystallites supported by the third 1000A3 pads remain isolated crystallites upon coalescence with the continuous layer.

[0127] The same observations can then be made for the coalescence of the third plots 1000A3 and the fourth plots 1000A4 respectively during a third and a fourth phase of growth.

[0128] It is understood that the different phases of epitaxial growth typically occur continuously and are represented and described separately here for explanatory purposes only. In particular, the concentric circles are merely schematic representations of the growth progression at different points in the process and have no physical reality at the end of the process.

[0129] The various examples of 1000A1-1000A5 block arrangement on substrate 100 described above resulted in very high-quality continuous layers, significantly reducing defect density compared to layers obtained without careful block arrangement. In particular, the rotational symmetry inherent in the block distribution around the first 1000A1 block limits or even eliminates potential defects caused by internal coalescence within the continuous layer. It also simplifies block fabrication.

[0130] It is understood, however, that any arrangement of the pads allowing the formation of a continuous layer solely by adding isolated crystallites to a cluster of crystallites (or possibly, at the beginning of the process, to another isolated crystallite) will yield equally satisfactory results. For example, removing the central pad is a viable option.

[0131] THE Figures 6A, 6B , 7 , 8A and 8B These are experimental results obtained by implementing the process according to the invention as described above. figure 6A is a top view of four assemblies E1, E2, E3, E4, each containing three pads 1000A1, 1000A2, 1000A3, each featuring a silicon creep section surmounted by a GaN crystalline section. These pads are supported by an SOI substrate. Within an assembly E1, E2, E3, E4, the pads are arranged along an alignment axis, as in the embodiment described above and schematically represented in the figure. figure 4D Epitaxial growth was performed on these patches until the three patches in each set coalesced. The result is presented in figure 6B . Each set of plots E1, E2, E3 E4 made it possible to form a continuous 550A layer of GaN.

[0132] THE figures 7 , 8A and 8B are results obtained by X-ray diffraction analysis of the ensemble.

[0133] There figure 7 This shows the orientation of the continuous GaN layers obtained after coalescence. It can be observed that the E3 and E4 assemblies, in particular, allowed the formation of continuous layers with a very uniform orientation.

[0134] THE Figures 8A and 8B The diagrams show the orientation of the silicon segments in the pads after coalescence. It can be observed that the silicon segments in assembly E1 maintained a very similar orientation between the beginning and end of the process. The silicon segments in assemblies E3 and E4, however, were oriented differently. The segments thus deformed, resulting in the formation of a highly aligned continuous layer of GaN.

[0135] These figures thus illustrate the fact that the alignment of continuous GaN layers by deformation of creep sections during the process works as expected.

[0136] The non-limiting example described above refers to the use of the nitride layer to form an LED or micro-LED. As previously stated, this example is not exhaustive. The described process could equally well be used to fabricate other nitride-based devices, for example, vertical devices such as HEMT transistors.

[0137] The invention is not limited to the embodiments described above and extends to all embodiments covered by the claims.

Claims

1. Method for obtaining a layer, at least partially made of a nitride (N) obtained preferably from at least one from among gallium (Ga), indium (In) and aluminium (Al), the method comprising the following steps: • providing a stack comprising at least one assembly of pads (1000A1-1000B4) extending from a substrate (100), each pad (1000A1-1000B4) comprising at least: i. one first section, called creep section (220A1- 220A5), formed in an amorphous material having a vitreous transition temperature Tvitreous transition, ii. one second section, crystalline, called crystalline section (300A1-300A5), surmounting the creep section (200A1-200A5), • epitaxially growing a crystallite (510A1-510A5) on at least some of said pads (1000A1-1000A5) and continuing the epitaxial growth of the crystallites (510A1-510A5) until coalescence of the crystallites (510A1-510A5) carried by the pads (1000A1-1000A5) of the assembly of pads, so as to form a continuous nitride layer (550A), the method being characterised in that the pads (1000A1-1000A5) of said assembly of pads are distributed over the substrate (100) such that the relative arrangement of the pads of the assembly of pads is such that during the epitaxy of the crystallites (510A1-510A5), the progressive coalescence of the crystallites (510A1-510A5) is always done between, on the one hand, an assembly, called coalesced assembly, comprising a crystallite (510A1) or a plurality of coalesced crystallites (510A1-510A4) and, on the other hand, at least one crystallite, called isolated crystallite (510A2-510A5), not having already coalesced with other crystallites.

2. Method according to the preceding claim, wherein the pads (1000A1-1000A5) carrying the crystallites forming the continuous nitride layer (550A) are disposed, projecting into a horizontal plane (XY) into which the upper face (101) of the substrate (100) extends, along a logarithmic spiral.

3. Method according to claim 1, wherein the pads (1000A1-1000A5) carrying the crystallites forming the continuous nitride layer (550A) are, projecting into a horizontal plane (XY) into which the upper face (101) of the substrate (100) extends, aligned along at least one so-called axis of alignment.

4. Method according to the preceding claim, wherein the pads (1000A1-1000A5) disposed on one same axis of alignment are spaced apart by an increasing distance by travelling the axis of alignment along a given direction from a given point, called initial point.

5. Method according to any one of the two preceding claims, wherein the pads (1000A1-1000A5) carrying the crystallites forming the continuous layer (550A) are separated by a distance varying logarithmically step by step, said distance being measured along their axis of alignment.

6. Method according to claim 1, wherein the pads (1000A1-1000A5) carrying the crystallites forming the continuous nitride layer (550A) are aligned, projecting into a horizontal plane (XY) into which the upper face (101) of the substrate (100) extends, along concurrent axes, the concurrent axes verifying between them, a symmetry of rotation around a central point, projecting into the horizontal plane (XY).

7. Method according to the preceding claim, wherein the pads (1000A1-1000A5) are aligned along five concurrent axes.

8. Method according to any one of the preceding claims, wherein the epitaxial growth is done at a temperature Tepitaxy, such that Tepitaxy > k1 x Tvitreous transition, with k1 ≥ 0.8.

9. Method according to the preceding claim, wherein k1 ≥ 1, and preferably k1 ≥ 1.5.

10. Method according to any one of the preceding claims, wherein the assembly of pads comprises at least 3 pads.

11. Method according to any one of the preceding claims, wherein the stack comprises a plurality of assemblies of pads on the substrate, a continuous nitride layer being formed from each assembly of pads, the epitaxial growth of crystallites being interrupted before the crystallites belonging to two distinct assemblies of pads come into coalescence, such that the continuous layers formed from each assembly of pads are distant from one another.

12. Method according to any one of the preceding claims, wherein each pad (1000A1-1000A5) has an upper face (1001A1) and wherein the epitaxial growth of the crystallites (510A1-510A5) is done at least partially and preferably only from said upper face (1001A1).

13. Method according to any one of the preceding claims, further comprising a formation step of a device in the continuous layer (550A), the device being taken from among an LED and a transistor, for example, a vertical transistor, for example, an HEMT-type transistor.

14. Method according to any one of the preceding claims, wherein the pads (1000A1-1000A5) of said assembly of pads are distributed over the substrate (100) such that the relative arrangement of the pads of the assembly of pads is such that during the epitaxy of the crystallites (510A1-510A5), at least at certain instants, the progressive coalescence of the crystallites (510A1-510A5) is done between, on the one hand, said coalesced assembly and, on the other hand, several isolated crystallites (510A2-510A5).