Process for transferring a thin layer onto a carrier substrate

Abstract

This invention relates to a method for transferring a thin layer onto a carrier substrate, comprising: providing a bonding assembly including a donor substrate and the carrier substrate, the two substrates being bonded together by direct bonding along a bonding interface on their respective front sides; the bonding assembly having an unbonded localized region within the bonding interface; the donor substrate further including an embedded weakening plane; and, after thermal activation to allow microcracks to grow in the embedded weakening plane, separation proceeding along the embedded weakening plane starting from the unbonded localized region, the separation resulting in the transfer of a thin layer from the donor substrate to the carrier substrate. Notably, the unbonded localized region is generated solely from a deliberately formed roughened region on at least one front side of both the donor and carrier substrates prior to bonding, the roughened region having no topological structure and having a predetermined roughness ranging between 0.5 nm RMS and 60.0 nm RMS.

Description

[Technical Field]

[0001] This invention relates to the fields of microelectronics and semiconductors. In particular, this invention relates to a method for transferring a thin layer onto a carrier substrate based on Smart Cut™ technology, the thin layer having an improved surface roughness after separation. This transfer method is particularly useful for fabricating an SOI structure. [Previous Technology]

[0002] Smart Cut™ technology is well known for fabricating SOI (Silicon-on-Insulator) structures, but it is more commonly used for transferring thin layers. This technology is based on the formation of an embedded weakening plane in a donor substrate by implanting light ions into the substrate; this embedded weakening plane defines the thin layer to be transferred with the front side of the donor substrate. The donor substrate and carrier substrate are then bonded together at their respective front sides to form a bonded assembly. Bonding is advantageously achieved through direct bonding, through molecular adhesion, i.e., without the use of adhesive materials: a bonded interface is thus established between the two bonded substrates. Thermal activation causes microcracks to grow in the embedded weakening plane, leading to spontaneous separation along the plane, thereby transferring the thin layer onto the carrier substrate (forming a stacked structure). The remaining portion of the donor substrate can be reused for subsequent layer transfers. After separation, the stacked structure is typically finished to restore the crystalline quality and surface roughness of the transferred thin layer. These finishing steps may in particular involve oxidation or smoothing heat treatment (in an inert or reducing atmosphere), cleaning and / or chemical etching and / or chemical mechanical polishing steps, as is known to those skilled in the art. Finally, various tools for inspecting the final structure can monitor the entire surface of the thin layer.

[0003] When the separation in the embedded weakened plane is spontaneous, considerable changes in the surface roughness of the transferred thin layer are observed at both high frequencies (micro-roughness) and low frequencies (ripples, localized areas of high roughness, marbling, etc.). These changes can be observed and measured specifically using the aforementioned inspection tools when monitoring the thin layer in the final structure.

[0004] It should be remembered that the surface roughness of the finished thin layer can be imaged using an image obtained using a KLA-Tencor Surfscan™ inspection tool (Figure 1). Roughness grades also have latent patterns (marbling, M, dense zones, DZ, etc.), which can be measured or revealed using haze measurement methods. Haze corresponds to the intensity of light scattered by the thin layer surface. The haze signal varies linearly with the square of the RMS surface roughness (root mean square roughness) in the spatial frequency range of 0.1 μm⁻¹ to 10 μm⁻¹. More complete information on this technique for inspecting and evaluating the roughness of large surface areas can be found in F. Holsteyns's article "Seeing through the haze" (Yield Management Solution, Spring 2004, pp. 50-54).

[0005] The images in Figure 1 show the surface roughness of two thin layers that were transferred from the two bonding components and processed in exactly the same way until finishing. In image (A), a residual roughness periphery region called the "dense zone (DZ)" is observed (especially at the top of the image); image (B) has no dense zone (DZ) at all. A more pronounced marble pattern (M) is also visible in image (A). In addition, there are significant differences in the average and maximum roughness (expressed as "haze" in ppm) between the two images (A) and (B). Figure 1 illustrates the variation in the final quality and roughness of the thin layers, which is mainly due to the variation in surface roughness (high frequency and low frequency) after separation.

[0006] To improve the quality of the thin layers in the final stacked structure, it is therefore important to reduce the surface roughness of these layers after transfer (regardless of spatial frequency).

[0007] According to US2010 / 330779, an unbonded local zone can be formed at the bonding interface, bordering a bonded zone, to form a trigger for separation, thereby limiting the roughness of the silicon thin layer transferred to a glass substrate. This unbonded local zone is obtained by creating a topological structure corresponding to a cavity and / or a dome (or peak) on the surface to be bonded on the donor substrate (silicon) or the carrier substrate (glass). Cavities and / or peaks on the order of 2 to 3 micrometers are created on the glass substrate to induce an unbonded zone of tens of square millimeters. [Summary of the Invention]

[0008] This invention proposes a transfer method that utilizes a specific fracture initiation point to obtain a thin layer with improved surface roughness after separation, thereby achieving excellent surface quality after the finishing steps of the stacked structure. This method is particularly advantageous for the fabrication of SOI structures.

[0009] The present invention relates to a method for transferring a thin layer onto a carrier substrate, comprising: providing a bonding assembly including a donor substrate and the carrier substrate, the two substrates being bonded together by direct bonding along a bonding interface on their respective front sides, the bonding assembly having an unbonded local area within the bonding interface, the donor substrate further including an embedded weakening plane, and after microcracks are grown in the embedded weakening plane by thermal activation, separation is performed starting from the unbonded local area along the embedded weakening plane, the separation causing a thin layer to transfer from the donor substrate to the carrier substrate.

[0010] It is noteworthy that the unbonded local region is generated only by a rough region deliberately formed on at least one front side of the donor substrate and the carrier substrate before the donor substrate and the carrier substrate are bonded. The rough region has no topological structure and has a predetermined roughness ranging from 0.5 nm RMS (excluding) to 60.0 nm RMS (excluding).

[0011] According to some advantageous features of the invention, it can be implemented alone or in any feasible combination: - The rough area is generated by at least one laser bombardment on the front side of the donor substrate and / or the carrier substrate, the laser bombardment only melting the surface layer of the material constituting the donor substrate and / or the carrier substrate within the first 1 to 30 nanometers; - The pulse duration of the laser bombardment is between 1 ns and 1000 ns, preferably between 10 ns and 500 ns; - The laser bombardment is performed with a laser with a wavelength of 308 nm and an energy density between 1.8 J / cm² and 2.5 J / cm²; - The laser bombardment irradiates a circular surface area with a diameter of less than 200 micrometers; - The rough area is formed by using multiple laser bombardments, for example, 2 to 15 laser bombardments, with adjacent circular surface areas irradiated by successive laser bombardments being tangent to each other or having an overlap percentage of 1% to 95%; - The front surface with the roughened region is made of monocrystalline silicon, and the predetermined roughness ranges between 0.5 nm RMS and 4.0 nm RMS, preferably between 1.0 nm RMS and 2.5 nm RMS; - The front surface with the roughened region is made of polycrystalline silicon, and the predetermined roughness ranges between 0.5 nm RMS and 5.0 nm RMS, preferably between 2.0 nm RMS and 5.0 nm RMS; - The front surface with the roughened region is made of silicon oxide, and the predetermined roughness ranges between 1.0 nm RMS and 6.0 nm RMS.Between 0 nm RMS; - The roughening region is formed on the front side of the donor substrate before the embedded weakening plane is formed in the donor substrate; - The donor substrate is formed of a first material on its front side, and a preliminary roughening region is formed on the surface of the first material, and after forming the preliminary roughening region and before the embedded weakening plane is formed, the first material of the donor substrate is subjected to a thermal oxidation treatment to form an insulating layer of the carrier substrate to be bonded to the bonding assembly, the insulating layer including the roughening region perpendicularly aligned with the preliminary roughening region on its free surface; - The carrier substrate is formed of a first material on its front side, and a preliminary roughening region is formed on the surface of the first material, and after forming the preliminary roughening region, the first material of the carrier substrate is subjected to a thermal oxidation treatment to form an insulating layer of the donor substrate to be bonded to the bonding assembly, the insulating layer including the roughening region perpendicularly aligned with the preliminary roughening region on its free surface; - The unbonded region has a shape in the plane of the bonding interface, the radius of curvature of at least a portion of its outline being smaller than the radius of a circular bonding defect of the same area; - The unbonded region has at least one lateral dimension of less than 300 micrometers in the plane of the bonding interface; - The unbonded region is located in the central region of the bonding assembly in the plane of the bonding interface; - The embedded weakening plane is formed by implanting a light element such as hydrogen or helium, or a combination of both, into the donor substrate; - The donor substrate and / or the carrier substrate have an insulating layer at least on their respective front sides, forming an embedded insulating layer in the bonding assembly adjacent to the bonding interface; - The thin layer from the donor substrate is made of monocrystalline silicon, and the carrier substrate comprises monocrystalline silicon to form a stacked structure of one type of SOI.

Implementation Method

[0015] This invention relates to a method for transferring a thin layer onto a carrier substrate to form a stacked structure. As stated in the preamble, such a stacked structure can be, in particular, of the SOI type, comprising a silicon-based surface thin layer, a silicon oxide intermediate insulating layer, and a silicon-based carrier substrate. The carrier substrate may, as needed, include other functional layers, such as a charge-trapping layer, for example, an SOI structure intended for radio frequency (RF) applications. However, the transfer method according to the invention is not limited to the fabrication of SOIs; it can also be applied to many other stacked structures in the fields of microelectronics, microsystems, and semiconductors.

[0016] The transfer method according to the present invention is based on Smart Cut™ technology. When separation in the embedded weakened plane is spontaneous, the fracture time (i.e., the time when separation ends during thermal fracture tempering) may differ among multiple bonded components treated in the same manner with the same tempering and furnace. The fracture time (FT) depends on various parameters related to the formation of the embedded weakened plane, fracture tempering, and the properties of the bonded components. The applicant has observed that for bonded components prepared in a similar manner and subjected to the same fracture tempering, the final stack structure resulting from separation occurring at short fracture times (sTF) (i.e., after transfer and finishing) has a lower high-frequency surface roughness (micro-roughness) in the thin layer than that of those occurring at longer fracture times (lFT), as shown in Figure 2. Furthermore, long fracture times result in very high roughness in localized areas at the edges of the thin layer after fracture (referred to as dense areas DZ), which is hardly or not at all observed with short fracture times. As shown in image (A) of Figure 1, even after finishing, the dense region reduces the roughness quality of the thin layer.

[0017] Therefore, the transfer method according to the present invention aims to initiate spontaneous separation in the embedded weakened plane in a desired manner (short fracture time) and repeatably (low dispersion fracture time between multiple similar bonded components) in order to substantially improve the surface roughness of the transferred thin layer.

[0018] To this end, the transfer method first includes providing a bonding assembly 100 comprising a donor substrate 1 and a carrier substrate 2, wherein the two substrates are bonded together by direct bonding along a bonding interface 3 at their respective front sides (1a, 2a) (Fig. 3a).

[0019] The donor substrate 1 is preferably in wafer form with a diameter of 100 mm, 150 mm, 200 mm, 300 mm, or 450 mm and a thickness typically between 300 μm and 1 mm. It has a front side 1a and a back side 1b. The surface roughness of the front side 1a is selected to be less than 1.0 nm RMS, and even more preferably less than 0.5 nm RMS (measured by atomic force microscopy (AFM), for example, within a 20 μm × 20 μm scan range). The donor substrate 1 can be made of silicon or any other semiconductor or insulating material, which may be capable of thin-layer transfer (e.g., SiC, GaN, III-V compounds, piezoelectric materials, etc.). It should also be noted that the donor substrate 1 may include at least one or more additional layers 12, such as an insulating layer, on its front side 1a. As illustrated in FIG3a, after bonding the donor substrate 1 and the carrier substrate 2, the additional layer 12 becomes an embedded intermediate layer in the bonding assembly 100.

[0020] The donor substrate 1 includes an embedded weakening plane 11 defining one of the thin layers 10 to be transferred. As is known from Smart Cut™ technology, this embedded weakening plane 11 can be formed by implanting a lightweight element such as hydrogen, helium, or a combination of both. The lightweight element is implanted into the donor substrate 1 at a given depth, consistent with the target thickness of the thin layer 10. These lightweight elements will form microcavities near the given depth, distributed in the form of a thin layer, generally parallel to the front side 1a of the donor substrate 1, i.e., parallel to the plane (x, y) in the diagram. For simplicity, this thin layer is referred to as the embedded weakening plane 11.

[0021] The implantation energy of the lightweight element is selected to reach a given depth. For example, hydrogen ions are implanted at energies between 10 keV and 210 keV, and doses between 5E16 / cm² and 1E17 / cm², to define a thin layer 10 with a thickness on the order of 100 nm to 1500 nm. It should be remembered that an additional layer may be deposited on the front side 1a of the donor substrate 1 prior to the ion implantation step. This additional layer may be composed of a material such as silicon oxide or silicon nitride. It may be retained for the next bonding step (and form all or part of the intermediate layer of the bonded assembly 100), or it may be removed.

[0022] The carrier substrate 2 preferably has a diameter of 100 mm, 150 mm, 200 mm, 300 mm, or 450 mm and a thickness typically between 300 μm and 1 mm in wafer form. It has a front side 2a and a back side 2b. The surface roughness of the front side 2a is selected to be less than 1.0 nm RMS, or even more preferably less than 0.5 nm RMS (measured by AFM, e.g., within a 20 μm × 20 μm scan range). The carrier substrate 2 can be made of silicon or any other semiconductor or insulating material that can be used for thin-layer transfer (e.g., SiC, GaN, III-V compounds, piezoelectric materials, insulating materials, etc.). It should also be noted that the carrier substrate 2 may include at least one or more additional layers, such as an insulating layer and / or a charge-trapping layer, on its front side 2a. After bonding the donor substrate 1 and the carrier substrate 2, this (or these) additional layers are embedded in the bonding assembly 100.

[0023] The bonding between the donor substrate 1 and the carrier substrate 2 is based on direct bonding via molecular attachment. It is well known that this type of bonding does not require an adhesive material because the bonding occurs at the atomic level between the bonding surfaces, forming a bonding interface 3. Several types of molecular attachment bonding exist, and their isothermal, pressure, atmospheric conditions, or treatments before surface contact can be noted to differ. Room temperature bonding (regardless of whether the surfaces to be bonded are pre-activated by plasma), atomic diffusion bonding (ADB), and surface activated bonding (SAB) are worth mentioning.

[0024] Before bringing the front surfaces 1a and 2a to be bonded into contact, the bonding steps may include a conventional sequence of chemical cleaning (e.g., RCA cleaning) and surface activation (e.g., with the aid of oxygen or nitrogen plasma) or other surface preparation (such as washing), which may promote the quality of the bonding interface 3 (low defect density, high adhesion energy).

[0025] The bonding component 100 according to the present invention has a unique feature (FIG. 3a) including an unbonded local region 31 inside the bonding interface 3. In other words, the unbonded local region 31 is in contact with the bonding interface 3, which is closed and exhibits molecular adhesion to the front surfaces 1a and 2a of the bonding substrates 1 and 2.

[0026] Before the donor substrate 1 and the carrier substrate 2 are bonded, the unbonded local region 31 is generated solely by a rough region 31a deliberately formed on at least one front surface 1a, 2a of the donor substrate 1 and the carrier substrate 2 (Fig. 3b). This rough region 31a has no topological structure or ripples at low frequencies, i.e., its conventional wavelength is greater than 100 nm. It has a predetermined high-frequency roughness, which is greater than the roughness of the front surfaces 1a, 2a surrounding the rough region 31a. The roughness of the front surfaces 1a, 2a can generally vary depending on the material properties of the surfaces 1a, 2a and the type of direct bonding used, but a bonding (closed) interface can still be obtained, while the predetermined roughness in the rough region 31a prevents local bonding between the two surfaces 1a, 2a.

[0027] The range of the predetermined roughness is strictly greater than 0.5 nm RMS and strictly less than 60.0 nm RMS, with wavelengths typically between 10 nm and 100 nm (corresponding to high-frequency micro-roughness). It is well known that RMS (root mean square) designation corresponds to a root mean square value of the roughness. The technique for measuring this micro-roughness is atomic force microscopy (AFM), with a scanning range from 10 × 10 μm² to 30 × 30 μm². It should be remembered that the rough region 31a does not have a topological structure corresponding to low-frequency ripples, but only micro-roughness within the aforementioned spatial frequency range. The maximum peak-to-valley (PV) range in the rough region 31a is typically between 5.0 nm and 300.0 nm, preferably between 5.0 nm and 60.0 nm.

[0028] In particular, in the example of FIG3b, rough region 31a has a roughness of 2.3 nm RMS and 25 nm PV, which was measured by AFM in a 10 μm x 10 μm scan range (right image in FIG3b).

[0029] According to a particular embodiment, particularly for fabricating SOI structures, the front surfaces 1a, 2a on which the rough region 31a is formed are made of single-crystal silicon, and the predetermined roughness preferably ranges between 0.5 nm RMS and 4.0 nm RMS (typically 5.0 nm PV to 40.0 nm PV). More preferably, the range is between 1.0 nm RMS and 2.5 nm RMS (typically 10.0 nm PV to 25.0 nm PV), or even between 1.5 nm RMS and 2.5 nm RMS.

[0030] According to another advantageous embodiment, the front surfaces 1a and 2a of the substrates 1 and 2 on which the roughened region 31a is formed are made of polycrystalline silicon, and the predetermined roughness ranges between 0.5 nm RMS and 5.0 nm RMS (typically 5.0 nm PV to 60.0 nm PV). More preferably, the range is between 2.0 nm RMS and 5.0 nm RMS (typically 20.0 nm PV to 60.0 nm PV). This embodiment is particularly advantageous for the fabrication of SOI structures suitable for RF applications, for which the carrier substrate 2 includes a charge-trapping layer made of polycrystalline silicon on its front surface 2a.

[0031] According to another embodiment, the front surfaces 1a and 2a on which the rough region 31a is formed are made of silicon oxide, and the predetermined roughness ranges between 1.0 nm RMS and 6.0 nm RMS.

[0032] It should be noted that when the rough area 31a is the front side 1a of the donor substrate 1, it is best to make the rough area 31a before the embedded weakening plane 11 is formed in the donor substrate 1, so as to limit the premature maturation of any damage or microcavities that constitute the rough area.

[0033] Alternatively, a preliminary roughening region may be formed on a first material forming the front side of the donor substrate 1; for example, the first material may be silicon. The first material of the donor substrate 1 is then subjected to a thermal oxidation treatment to form an insulating layer (e.g., silicon oxide). The thickness of the insulating layer may be, for example, less than or equal to 200 nm. The insulating layer (corresponding to the additional layer 12 illustrated in FIG. 3a) includes a roughening region 31a on its free surface (the front side 1a of the donor substrate 1) that is perpendicularly aligned with the preliminary roughening region generated in the first material. Specifically, the oxidation treatment at least partially preserves the high-frequency roughness of the first material in the preliminary roughening region: an attempt is made to form a micro-roughness in the preliminary roughening region that, after the oxidation treatment, produces a predetermined roughness of the roughening region 31a within the aforementioned range. After the embedded weakened plane 11 is formed in the donor substrate 1, the insulating layer 12 is bonded to the carrier substrate 2 in the bonding assembly 100, and the roughened region 31a generates the unbonded local area 31 within the bonding interface 3. In this case, the roughened region 31a on the front side 1a of the donor substrate 1 is not directly created by the "roughening" of the insulating layer 12, but is generated by a preliminary roughened region existing in the underlying material.

[0034] Although the formation of the initial roughened zone is described with reference to the donor substrate 1, it can of course also be applied to the carrier substrate 2.

[0035] The bonding component 100 is formed to include the unbonded local area 31 within its bonding interface 3. The transfer method of the present invention contemplates applying a thermal tempering to it, which will cause spontaneous separation along the embedded weakening plane 11. This separation results in the transfer of the thin layer 10 from the donor substrate 1 to the carrier substrate 2 to form a stacked structure 110 (FIG. 4). The untransferred area 31b of the thin layer 10 is perpendicularly aligned with the position of the unbonded local area 31. Furthermore, the remaining portion 1' of the donor substrate is also obtained.

[0036] After microcracks are grown in the embedded weakening plane 11 through thermal activation, the unbonded local region 31 acts as a fracture initiation point in the embedded weakening plane 11 and at or near the location perpendicularly aligned with or near the unbonded local region 31. This fracture initiation occurs in a desired manner relative to the bonded assembly 100 which does not contain the unbonded local region 31: this provides a short fracture time, resulting in low surface roughness on the surface 10a of the thin layer 10 after transfer. The example detailed below shows a significant improvement in surface roughness after transfer due to the shorter fracture time achieved through the unbonded local region 31 serving as a fracture initiation point (Fig. 6, stacked structure 110).

[0037] The unbonded local region 31 of the present invention differs from the prior art in that it is only due to the presence of a rough region 31a on the front surfaces 1a, 2a of one of the donor substrate 1 and the carrier substrate 2. This unbonded local region does not involve low-frequency topological structures (bumps, holes, cavities, particles). The internal volume of the unbonded local region 31 thus formed is very small; in fact, within this internal volume, especially the accumulation of various gases obtained from the monolayer evaporation of water present on the bonding surfaces 1a, 2a or from diffusion from the light elements outside the front surface 1a of the donor substrate 1, rapid pressurization conducive to fracture initiation is allowed once the maturity of the microcracks in the embedded weakening plane 11 is sufficient.

[0038] The unbonded local area 31 may be located at different positions in the plane of the bonding interface 3: particularly in the central area of ​​the bonding assembly 100, or in the peripheral area, or in the middle area between these two extremes. In the peripheral area, the unbonded local area 31 is preferably at least 1 mm away from the (unbonded) peripheral ring of the bonding assembly 100. It should be remembered that the unbonded ring (visible in FIG. 8b) is related to the edge rounding and chamfer present at the edges of the substrates 1 and 2 to be bonded.

[0039] The applicant has observed that the unbonded local area 31 located in the center or the central area of ​​the bonded component 100 (as shown in Figure 3b) not only reduces the final roughness ("haze") of the thin layer 10 after transfer, but also significantly reduces or even eliminates the dense area DZ on the entire surface of the thin layer, which corresponds to the local peripheral area with extremely high roughness (Figure 6).

[0040] The unbonded local region 31 may have various shapes in the plane of the bonding interface 3 (i.e., in the (x, y) plane). Some examples are illustrated in Figure 7.

[0041] Advantageously, the unbonded local region 31 has a shape in the (x, y) plane, the radius of curvature of at least a portion of its profile being smaller than the radius of a circular bonding defect of the same area (Fig. 7(b), (c), (d)). The applicant has demonstrated that having a small radius of curvature, a straight portion, or singularities (e.g., cuspidal points) in the profile can impart excellent fracture initiation effects to the unbonded local region 31: thus, short and not very dispersed fracture times can be obtained among a plurality of bonded components, resulting in low, uniform, and reproducible surface roughness of the thin layer 10.

[0042] Figures 8a and 8b illustrate optical microscope images of untransferred regions 31b (generated from unbonded local regions 31) on the stacked structure 110 obtained by the transfer method of the present invention. These untransferred regions 31b have a specific shape outline as described above, in contrast to the effect of increasing unbonded local regions 31 as a fracture initiation point.

[0043] In order to avoid adversely affecting the transfer integrity of the thin layer 10, the unbonded local region 31 advantageously has at least one lateral dimension of less than 300 micrometers in the (x, y) plane. This is typically achieved by forming a rough region 31a on one of the front surfaces 1a, 2a of the substrates 1, 2, which has a relevant lateral dimension of less than or equal to 200 micrometers.

[0044] The roughened area 31a can be produced by various techniques, especially by wet or dry chemical etching after the front surfaces 1a and 2a in question have been protected except for the areas to be roughened.

[0045] However, it is advantageous to form the rough area by minimizing its contact with the rest of the front surfaces 1a and 2a, thereby reducing the risk of contamination or deterioration (scratches, etc.) of the surfaces to be joined.

[0046] A particularly advantageous technique for forming the roughened region 31a is the use of laser bombardment, which is capable of producing superficial melting of the frontal materials 1a and 2a only in the region to be roughened. This superficial melting typically occurs a few nanometers before the frontal materials 1a and 2a, between 1 nm and 30 nm: therefore, it is within a melting limit regime, rather than a completely melted state where material is removed or hollowed out. The laser bombardment used here does not produce any topological structure in the aforementioned RMS and PV ranges; it does not produce cavities or bumps, but only high-frequency roughness.

[0047] Therefore, the laser bombardment is preferably carried out with a very short pulse, typically between 1 ns and 1000 ns, and more advantageously between 10 ns and 500 ns. The wavelength of the laser used can be selected between 100 nm and 550 nm, preferably between 250 nm and 400 nm.

[0048] For example, when the material is silicon, the laser bombardment can be performed using a blaze with a wavelength of 308 nm and energy densities between 1.8 J / cm² and 2.5 J / cm². Figure 5 illustrates the curves relating the roughness of a single-crystal silicon surface (RMS and PV) irradiated with laser (wavelength 308 nm, pulse duration 160 ns) to the energy density of the laser bombardment. Importantly, it needs to be within a melting limit range (1.9 ± 0.1 J / cm² in this example) in order to produce the desired range of high-frequency roughness. This state corresponds to "discontinuous" melting of the irradiated surface; that is, molten and unmolten regions are induced on the surface, which is beneficial for forming the desired high-frequency roughness. Too low an energy density (< 1.8 J / cm 2) will not achieve melting and will produce little or no roughness; too high an energy density (> 2 J / cm 2) will cause the entire irradiated surface to melt "uniformly" and recrystallize uniformly, while its roughness will be too limited.

[0049] It should be noted that when the material of the front surfaces 1a and 2a (which will be subjected to laser bombardment) is silicon oxide and the underlying material is silicon, the laser bombardment penetrates the silicon oxide layer and causes the silicon to melt, which will result in wrinkles on the oxide surface. These wrinkles create the desired high-frequency roughness in the irradiated areas on the front surfaces 1a and 2a.

[0050] As previously described regarding the preferred maximum lateral dimension of the unbonded local area 31, the laser bombardment irradiates a generally circular surface area with a diameter advantageously less than 200 micrometers. The roughened area 31a produced after the laser bombardment is approximately the size of the irradiated surface area.

[0051] In order to produce the favorable shape of the unbonded local region 31, the rough region 31a can be produced by a plurality of laser bombardments, such as two (Fig. 7(b)), three, four (Fig. 7(c)), five, six (Fig. 7(d)), or even fifteen laser bombardments. In order to produce a single unbonded local region 31 with a specific shape rather than a plurality of circular regions 31, the surfaces irradiated by successive laser bombardments must be at least tangent; they may also have a specific percentage of overlap (between 1% and 95%, or between 20% and 70%). As illustrated in Figs. 7(b), (c), and (d), it is therefore possible to form a profile of the unbonded local region 31 that is favorable for fracture initiation, because its local areas have singularities or low radii of curvature, with an overlap of 50%, which may form a capsule-shaped rough region 31a (Fig. 7(e)): a portion of the profile of the thus produced unbonded local region 31 has a straight section, which makes it effective for initiating fracture.

[0052] The shapes of the rough region 31a and the unbonded local region 31 described herein are not exhaustive, and any other geometric or non-geometric shapes may be conceived.

[0053] Example of implementation: According to the first example, a transfer method is performed to produce an FDSOI (fully depleted SOI) structure, namely a thin surface layer with a small thickness and a thin embedded insulating layer with a small thickness.

[0054] The donor substrate 1 is a single-crystal silicon wafer with a diameter of 300 mm, and contains an insulating layer 12 of silicon oxide with a thickness of 35 nm on its front side 1a. The buried weakening plane 11 is formed by implanting helium and hydrogen ions with energies of 40 keV and 25 keV, respectively, and doses of 1E16 / cm2 and 1E16 / cm2, respectively.

[0055] The substrate 2 is a single-crystal silicon wafer with a diameter of 300 mm.

[0056] Laser bombardment was performed on the center of the carrier substrate 2, on its front side 2a, to cause surface melting / recrystallization only within the melting limit. The laser conditions were as follows: wavelength 308 nm, laser pulse duration 160 ns, energy density 1.9 J / cm², and a quasi-circular irradiation surface area with a radius of 65 μm. Therefore, laser bombardment formed a rough region on the irradiated surface with an RMS roughness of 2.3 nm ± 0.5 nm, which was measured by AFM within a 10 x 10 μm² scan range (see, for example, the AFM image in Figure 3b).

[0057] Then conventional surface preparation (cleaning operation, plasma activation) of the two substrates 1 and 2 is performed to prepare for bonding through molecular attachment.

[0058] The bonding system is achieved by bringing the front surfaces 1a and 2a of the donor substrate 1 and the carrier substrate 2 into direct contact, making it possible to obtain a bonded assembly 100. The bonded assembly 100 includes an unbonded local region 31, which is vertically aligned with the rough region 31a within the bonding interface 3 (at the center of the bonded assembly 100). The unbonded local region 31 has a diameter on the order of 130 to 250 μm.

[0059] Fracture tempering (capable of collectively processing multiple bonded components 100) performed in a horizontal furnace is carried out between 200°C and 400°C.

[0060] Under isothermal tempering, the unbonded local area 31 can initiate fracture along the embedded weakening plane 11 in a short time, that is, the time is about 25% of the average fracture time without the unbonded local area 31. When multiple bonded components 100 are processed collectively, the unbonded local area 31 can also obtain a less dispersed fracture time.

[0061] The SOI structure 110 obtained after separation has an untransferred region 31b that replaces the unbonded local region 31 of the bonding component 100, and the size of the unbonded local region 31 is approximately the same.

[0062] Compared to an SOI structure 110' obtained from a conventionally bonded assembly without unbonded local regions 31, the surface quality of the transferred thin layer 10 of the present invention is improved. This is particularly evident after applying finishing steps (mainly thermal oxidation and smoothing treatment) to correct the transferred thin layer 10 and smooth its surface: unlike some SOI structures 110' obtained by conventional methods (without unbonded local regions 31), the SOI structure 110 obtained by the method of the present invention does not exhibit rough and dense regions around the thin layer 10.

[0063] According to the second example, a transfer method is performed to produce an RFSOI (SOI for radio frequency applications) structure, that is, having a surface thin layer, an insulating layer and a charge trapping layer on the carrier substrate 2.

[0064] The donor substrate 1 is a single-crystal silicon wafer with a diameter of 300 mm, and contains an insulating layer 12 of silicon oxide with a thickness of 200 nm on its front side 1a. The buried weakening plane 11 is formed by co-implanting helium and hydrogen ions with energies of 35 keV and 50 keV, respectively, and doses of 1.2E16 / cm2 and 1.1E16 / cm2, respectively.

[0065] The carrier substrate 2 is a single-crystal silicon wafer with a diameter of 300 mm. A charge-capturing layer of polycrystalline silicon with a thickness of approximately 1 μm is located on the front side 2a of the carrier substrate 2.

[0066] Laser bombardment is performed on a peripheral area of ​​the carrier substrate 2, on its front surface 2a (and thus on polycrystalline silicon), so as to cause only surface melting / recrystallization. In particular, the laser bombardment is performed at a distance of 3 mm to 10 mm, for example 5 mm, from the edge of the carrier substrate 2.

[0067] The laser conditions are as follows: wavelength 308 nm, laser pulse duration 160 ns, energy density 1.9 J / cm², and a quasi-circular irradiation surface area with a radius of 65 μm. Therefore, laser bombardment results in the formation of a rough region on the irradiated surface, with an RMS roughness of 2.7 nm ± 0.5 nm, measured by AFM within a 10 x 10 μm² scan range.

[0068] Then, conventional surface preparation (cleaning operation, plasma activation) of the two substrates 1 and 2 is performed to prepare for bonding through molecular attachment.

[0069] The bonding system is achieved by bringing the front surfaces 1a and 2a of the donor substrate 1 and the carrier substrate 2 into direct contact, thereby obtaining a bonding assembly 100. The bonding assembly 100 includes an unbonded local region 31, which is perpendicularly aligned with the rough region 31a within the bonding interface 3 (at the edge of the bonding assembly 100). The unbonded local region 31 has a diameter on the order of 130 to 250 μm.

[0070] Fracture tempering (capable of collectively processing multiple bonded components 100) performed in a horizontal furnace is carried out between 200°C and 550°C.

[0071] The unbonded local area 31 can initiate fracture along the embedded weakening plane 11 in a short time, which is about 30% of the average fracture time without the unbonded local area 31. When multiple bonded components 100 are processed collectively, the unbonded local area 31 can also achieve a less dispersed fracture time.

[0072] The SOI structure 110 obtained after separation has an untransferred region that replaces one of the unbonded local regions 31 of the bonding component 100, and its size is approximately the same as that of the unbonded local region 31, typically less than 250 μm.

[0073] Compared to an SOI structure 110' obtained from a conventional bonded assembly without unbonded local regions 31, the surface quality of the transferred thin layer 10 of the present invention is improved. This is particularly evident after applying finishing steps (mainly thermal oxidation and smoothing treatment) to correct the transferred thin layer 10 and smooth its surface: unlike some SOI structures 110' obtained by conventional methods (without unbonded local regions 31), the SOI structure 110 of FIG. 6 obtained by the method of the present invention does not have rough dense regions around the thin layer 10; and exhibits a lower overall roughness level on the entire surface of the thin layer 10 of the SOI structure 110.

[0074] According to the third example, a transfer method is performed to produce an SOI structure having a surface thin layer, an insulating layer and a carrier substrate 2.

[0075] The donor substrate 1 is a single-crystal silicon wafer with a diameter of 300 mm.

[0076] A laser bombardment is performed on the front side 1a (and thus on silicon) of the central or peripheral region of the donor substrate 1 to cause only surface melting / recrystallization. The laser conditions are as follows: wavelength 308 nm, laser pulse duration 160 ns, energy density 1.9 J / cm², and a quasi-circular irradiation surface area with a radius of 65 μm. Thus, a laser bombardment results in the formation of a preliminary roughened region on the irradiated surface with an RMS roughness of 2.3 nm ± 0.5 nm, which was measured by AFM within a 10 x 10 μm² scan range.

[0077] After standard cleaning, the donor substrate 1 is typically thermally oxidized in a temperature range of 900°C – 1050°C. At the end of this oxidation, the donor substrate 1 contains an insulating layer 12 of silicon oxide with a thickness of 100 nm on its front side 1a. A roughened region 31a, perpendicularly aligned with the initial roughened region, exists on the free surface 1a of the insulating layer 12. The roughened region 31a has approximately the same RMS roughness as the initial roughened region, i.e., approximately 2.2 nm ± 0.5 nm (10 x 10 μm 2AFM scan range).

[0078] The embedded weakening plane 11 is formed by implanting helium and hydrogen ions through the insulating layer 12 in the donor substrate 1 with energies of 35 keV and 50 keV, and doses of 1.2E16 / cm2 and 1.1E16 / cm2, respectively.

[0079] The substrate 2 is a single-crystal silicon wafer with a diameter of 300 mm.

[0080] Then, conventional surface preparation (cleaning operation, plasma activation) of the two substrates 1 and 2 is performed to prepare for bonding through molecular attachment.

[0081] The bonding system is achieved by directly contacting the front surfaces 1a and 2a of the donor substrate 1 and the carrier substrate 2, thereby obtaining a bonding assembly 100. The bonding assembly 100 includes an unbonded local region 31, which is perpendicularly aligned with the rough region 31a within the bonding interface 3. The unbonded local region 31 has a diameter on the order of 100 μm.

[0082] Fracture tempering (capable of collectively processing multiple bonded components 100) performed in a horizontal furnace is carried out between 200°C and 550°C.

[0083] The unbonded local area 31 can initiate fracture along the embedded weakening plane 11 in a short time, which is about 30% of the average fracture time without the unbonded local area 31. When multiple bonded components 100 are processed collectively, the unbonded local area 31 can also achieve a less dispersed fracture time.

[0084] The SOI structure 110 obtained after separation has an untransferred region that replaces one of the unbonded local regions 31 of the bonding component 100, and its size is approximately the same as that of the unbonded local region 31, typically less than 200 μm.

[0085] Compared to an SOI structure 110' obtained from a conventional bonded assembly without unbonded local regions 31, the quality of the surface 10a of the thin layer 10 after transfer according to the present invention is improved: it has no rough dense region DZ around the thin layer 10 and exhibits a lower overall roughness level on the entire surface of the thin layer 10.

[0086] According to the fourth example, a transfer method is performed to produce an RFSOI (SOI for radio frequency applications) structure, that is, the carrier substrate 2 has a surface thin layer, an insulating layer and a charge trapping layer.

[0087] The donor substrate 1 is a single-crystal silicon wafer with a diameter of 200 mm, and contains an insulating layer 12 of silicon oxide with a thickness of 400 nm on its front side 1a. The buried weakening plane 11 is formed by implanting hydrogen ions with an energy of 50 keV and a dose of 6E16 / cm2.

[0088] The substrate 2 is a single-crystal silicon wafer with a diameter of 200 mm. A polycrystalline silicon charge-capturing layer with a thickness of approximately 2 μm is located on the front side 2a of the substrate 2.

[0089] A plurality of consecutive laser bombardments are performed on the peripheral region of the carrier substrate 2, on its front surface 2a (and thus on polycrystalline silicon), so as to cause only surface melting / recrystallization. In particular, the laser bombardment is performed at a distance of 3 mm to 10 mm, for example 5 mm, from the edge of the carrier substrate 2.

[0090] The laser conditions were as follows: each laser bombardment wavelength was 308 nm, the laser pulse duration was 160 ns, the energy density was 2 J / cm², and the irradiated surface area had a quasi-circular radius of 65 μm. Eleven consecutive aligned laser bombardments resulted in approximately 10% overlap between adjacent irradiated surfaces, leading to the formation of a rough region 31a with a shape similar to one of the shapes shown in Figure 7(d). The RMS roughness of this rough region was approximately 3 nm ± 0.5 nm, as measured by AFM within a 10 x 10 μm² scan range.

[0091] Then, conventional surface preparation (cleaning operation, plasma activation) of the two substrates 1 and 2 is performed to prepare for bonding through molecular attachment.

[0092] The bonding system is achieved by directly contacting the front surfaces 1a and 2a of the donor substrate 1 and the carrier substrate 2, thereby obtaining a bonding assembly 100. The bonding assembly 100 includes an unbonded local region 31 that is vertically aligned with the rough area 31a within the bonding interface 3. The unbonded local region 31 has a length on the order of 2 mm and a width on the order of 100 to 250 μm in the plane ((x,y) plane) of the bonding interface.

[0093] The fracture tempering system carried out in a horizontal furnace is applied between 200°C and 550°C.

[0094] The unbonded local area 31 can initiate fracture along the embedded weakening plane 11 in a short time, which is about 30% of the average fracture time without the unbonded local area 31. When multiple bonded components 100 are processed collectively, the unbonded local area 31 can also achieve a less dispersed fracture time.

[0095] The SOI structure 110 obtained after separation has an untransferred region 31b that replaces the unbonded local region 31 of the bonding component 100, and its size is approximately the same as that of the unbonded local region 31 (Fig. 8b).

[0096] Compared to the SOI structure obtained by conventional bonding components without unbonded local regions 31, the quality of the surface 10a of the thin layer 10 after transfer in this invention is further improved.

[0097] It should be noted that the unbonded local area 31 having the specific shape described in the fourth example can also provide the above advantages when applied to the bonding components 100 described in the first, second and third examples.

[0098] Of course, the present invention is not limited to the embodiments described in this specification, and variations of the embodiments may be implemented without departing from the scope of the present invention as defined in the claims. [Simplified Explanation of the Diagram]

[0012] Other features and advantages of the present invention will become apparent from the following detailed description with reference to the accompanying drawings, in which: FIG1 illustrates two images representing the surface roughness of two transferred thin layers, according to a conventional method, the two thin layers being from two bonded components that have been treated in the same manner up to finishing; these two images were obtained using a Surfscan™ inspection tool; FIG2 illustrates a graph of the surface roughness of the thin layer as a function of fracture time, according to a conventional method, for a plurality of bonded components (different from the bonded components mentioned with reference to FIG1) and treated in the same manner up to finishing; FIG3a illustrates a bonded component as an intermediate step in the transfer process according to the present invention; FIG3b illustrates an example of a donor or carrier substrate including a roughened region intentionally formed on its front side by the transfer method according to the present invention, the roughened region having a predetermined roughness; FIG4 illustrates a layer obtained by the transfer method according to the present invention. A stacked structure and the remainder of the donor substrate; Figure 5 illustrates the RMS (root mean square) roughness and PV (peak-to-valley) variation of a silicon surface irradiated by a laser, which is a function of laser energy density; Figure 6 illustrates two images representing the surface roughness (after finishing) of two transferred thin layers used in the first and second RFSOI structures, the first (110) obtained by a transfer method according to the present invention, and the second (110') obtained by a conventional transfer method; both images were obtained using the Surfscan™ inspection tool; Figure 7 illustrates examples of different shapes of rough regions and associated unbonded local regions used in the transfer method according to the present invention; Figures 8a and 8b illustrate images of an untransferred region corresponding to an unbonded local region of a bonded component; the untransferred region is observed on a stacked structure obtained by a transfer method according to the present invention.

[0013] Some diagrams are presented in a schematic and simplified manner and are not drawn to scale for clarity. In particular, the layer thickness along the z-axis is not proportional to the lateral dimensions along the x-axis and y-axis.

[0014] In the drawings and descriptions, the same symbols may be used for elements of the same nature.

Claims

1. A method for transferring a thin layer (10) onto a carrier substrate (2), comprising: A bonding assembly (100) is provided, comprising a donor substrate (1) and a carrier substrate (2), the two substrates being joined by direct bonding along a bonding interface (3) on their respective front sides (1a, 2a). The bonding assembly (100) has an unbonded local area (31) inside the bonding interface (3). The donor substrate (1) further comprises an embedded weakening plane (11). After microcracks are grown in the embedded weakening plane (11) by thermal activation, separation occurs starting from the unbonded local area (31) along the embedded weakening plane (11). This separation results in a thin layer (10) migrating from the donor substrate (1) to the carrier substrate (2). The method is characterized in that the unbonded local region (31) is generated only by a rough region (31a) deliberately made on at least one front side (1a, 2a) of the donor substrate (1) and the carrier substrate (2) before the donor substrate (1) and the carrier substrate (2) are bonded. The rough region (31a) has no topological structure and has a predetermined roughness in the range of 0.5 nm RMS and 60.0 nm RMS.

2. The method of claim 1, wherein the rough region (31a) is generated by at least one laser bombardment of the front surface (1a, 2a) of the donor substrate (1) and / or the carrier substrate (2), said laser bombardment only melting the surface layer of the material constituting the donor substrate (1) and / or the carrier substrate (2) within the first 1 to 30 nanometers.

3. The method of claim 2, wherein the pulse duration of the laser bombardment is between 1 ns and 1000 ns, preferably between 10 ns and 500 ns.

4. The method of claim 3, wherein the laser bombardment is performed with a laser with a wavelength of 308 nm and an energy density between 1.8 J / cm² and 2.5 J / cm².

5. The method of any one of claims 2 to 4, wherein a laser bombardment irradiates a circular surface area with a diameter of less than 200 micrometers.

6. The method of claim 5, wherein the rough area (31a) is generated by using a plurality of laser bombardments, such as 2 to 15 laser bombardments, wherein adjacent circular surface areas irradiated by successive laser bombardments are tangent to each other or have an overlap percentage of 1% to 95%.

7. As in request item 1, where: The front surface (1a, 2a) formed on the rough region (31a) is made of monocrystalline silicon, and the predetermined roughness ranges between 0.5 nm RMS and 4.0 nm RMS, preferably between 1.0 nm RMS and 2.5 nm RMS; or the front surface (1a, 2a) formed on the rough region (31a) is made of polycrystalline silicon, and the predetermined roughness ranges between 0.5 nm RMS and 5.0 nm RMS, preferably between 2.0 nm RMS and 5.0 nm RMS; or the front surface (1a, 2a) formed on the rough region (31a) is made of silicon oxide, and the predetermined roughness ranges between 1.0 nm RMS and 6.0 nm RMS.

8. The method of claim 1, wherein the rough area (31a) is formed on the front side (1a) of the donor substrate (1) before the embedded weakening plane (11) is formed in the donor substrate (1).

9. As in request item 8, wherein: The donor substrate (1) is formed of a first material on its front side (1a), and a preliminary roughening region is formed on the surface of the first material. After the preliminary roughening region is formed and before the embedded weakening plane (11) is formed, the first material of the donor substrate (1) is subjected to a thermal oxidation treatment to form an insulating layer of the carrier substrate (2) to be bonded to the bonding assembly (100). The insulating layer includes the roughening region (31a) perpendicularly aligned with the preliminary roughening region on its free surface.

10. As in request item 1, wherein: The carrier substrate (2) is formed of a first material on its front side (2a), and a preliminary roughening region is formed on the surface of the first material. After the preliminary roughening region is formed, the first material of the carrier substrate (2) is subjected to a thermal oxidation treatment to form an insulating layer of the donor substrate (1) that will be bonded to the bonding assembly (100). The insulating layer includes a roughening region (31a) that is perpendicularly aligned with the preliminary roughening region on its free surface.

11. The method of claim 1, wherein the unbonded local area (31) has a shape in the plane of the bonding interface (3) and the radius of curvature of at least a portion of its profile is smaller than the radius of a circular bonding defect of the same area.

12. The method of claim 1, wherein the unbonded local region (31) has at least one lateral dimension of less than 300 micrometers in the plane of the bonding interface (3).

13. The method of claim 1, wherein the unbonded local area (31) is located in the central area of ​​the bonding component (100) in the plane of the bonding interface (3).

14. The method of claim 1, wherein the thin layer (10) from the donor substrate (1) is made of monocrystalline silicon and the carrier substrate (2) contains monocrystalline silicon in order to form a stacked structure (110) of one type of SOI.

Images