Process for joining two substrates by molecular adhesion and structure obtained by such a process

Abstract

This invention relates to a method for molecularly bonding two substrates, each substrate having a principal facet, and at least one of the substrates having a dielectric surface layer on one side of its principal facet. The method includes (a) bringing the principal facests of the two substrates into contact, and then (b) initiating and propagating a bonding wave between the principal facests of the two substrates to bond the two principal faces together. According to the invention, the method includes a step of preparing the dielectric surface layer prior to the contact step, with the aim of introducing a sulfur dose greater than 3.0 E13 at / cm^2 into this layer. The invention also relates to a structure obtained by this method.

Description

Technical Field

[0001] This invention is particularly applicable to fields such as microelectronics, microsystems, and photonics. Prior Technology

[0002] Molecular adhesion is a method of joining two objects in which completely clean, flat, and smooth principal surfaces or faces of the two objects come into close contact to promote the formation of molecular bonds, such as van der Waals bonds or covalent bonds. Therefore, the joining of the two objects can be achieved without the use of adhesives. These bonds can be strengthened by heat treatment of the joined parts.

[0003] This method is particularly useful for producing substrates for microelectronics, microsystems, photonics, and other fields. In these fields, the substrate is typically a circular wafer material that can be single-crystal, polycrystalline, or amorphous, and can be a conductor, semiconductor, or insulator.

[0004] In these fields, to facilitate substrate bonding, a dielectric layer, typically silicon oxide, is usually formed by depositing it on one and / or the other of two substrates, followed by a subsequent grinding and polishing step as needed, or by simple oxidation when the substrate is formed of or contains silicon. The primary surfaces are also typically prepared to be clean (especially in terms of particles and contaminants) and smooth (to provide a surface roughness typically below 0.5 nm, as the root mean square value over a 5 μm by 5 μm field of view). Sometimes, it is also conceivable to make at least one of these surfaces reactive by applying plasma (typically oxygen or nitrogen). The aim of this surface preparation is to obtain maximum bonding energy immediately after the bonding step and before any thermal annealing.

[0005] After fabrication, one substrate is placed on top of another on the main surfaces. Applying pressure to one of the two substrates causes intimate point contact between the main surfaces, which propagates throughout the main surfaces in the form of a "bonding wave".

[0006] Patents, technologies, and scientific literature contain a wealth of documents outlining the principles and applications of this bonding method. See in particular Rieutord et al.'s publication, "Dynamics of a bonding front," Physical Review Letters, vol. 94, pp. 236101, 2005.

[0007] A common problem encountered in the development of molecular adhesion methods is the appearance of peripheral defects (often called "picots") at the bonding interface and the point where the bonding wave intersects with the substrate edge. When the wave is initiated at the substrate edge, these peripheral defects are generally distributed in the peripheral region of the edge completely opposite to the wave initiation point. When the wave is initiated in the central region of one of the substrates, these peripheral defects may be arranged across the entire profile. These defects cause the bonding interface between the two substrates to trap gas and / or water. When defects occur, the two substrates are not in complete contact, and the local adhesion energy is very low or even non-existent. Although these defects exist, they are not necessarily visible immediately after the bonding wave propagates, but may appear during the thermal annealing of the joint formed by the two substrates being joined together.

[0008] The velocity of the bonding wave has been identified as a factor influencing the number of defects, and reducing this velocity can typically reduce or even eliminate the number of defects. Various techniques for controlling this velocity have been proposed, particularly by treating the surfaces of one and / or another substrate, especially by heating (document WO2007060145). Another method proposed to reduce the velocity of the bonding wave is to inject gas (which may be hot gas, if necessary) in the direction of the bonding wave initiation point during bonding. However, while this treatment effectively reduces the bonding velocity, it can also easily reduce the adhesion energy, which may be detrimental to some applications. On the other hand, document WO2013160841 envisions maintaining the substrate in a gaseous environment (e.g., helium, neon, or hydrogen) during bonding wave propagation, where the gas has a negative Joule-Thomson coefficient at the temperature and pressure of the environment.

[0009] While solutions based on prior knowledge are effective, they still have significant limitations. Summary of the Invention

[0010] The object of this invention is to provide a method for molecular adhesion bonding that, unlike prior art, can potentially prevent or at least limit the occurrence of "ring-hole" defects. More specifically, the object of this invention is to provide a method for molecular adhesion bonding that provides high bonding energy and has far fewer "ring-hole" defects than prior art, and is not limited by prior art.

[0011] To achieve this objective, the present invention provides a method for molecularly bonding two substrates, each substrate having a principal facet, and at least one of the two substrates having a dielectric surface layer on its principal facet. The method includes the steps of: a) bringing the principal facests of the two substrates into contact, and then; b) initiating and propagating a bonding wave between the principal facests of the two substrates to bond the two principal faces together.

[0012] The method is noteworthy in that it includes a step of preparing a dielectric surface layer prior to the contact step, with the aim of introducing a sulfur dose greater than 3.0 E13 at / cm^2 into this dielectric surface layer.

[0013] Surprisingly, introducing this dose of sulfur into the surface layer prevents or at least limits the occurrence of annular defects, even when the two substrates are bonded with high bonding energy.

[0014] Other advantageous and non-limiting features of the invention can be carried out alone or in any technically feasible combination: The amount of sulfur introduced was 3.5 E13 at / cm^2; The preparation of the dielectric surface layer also aims to introduce a fluorine dosage greater than 4.0 E14 at / cm^2 into this dielectric surface layer; The preparation steps include activating the dielectric surface layer with a plasma containing sulfur gas; The preparation steps include activating the dielectric surface layer with a plasma containing sulfur hexafluoride; Plasma also includes oxygen or nitrogen; The activation time for the surface layer is between 15 seconds and 2 minutes; Throughout the activation process, the dielectric surface layer is exposed to oxygen or nitrogen plasma, and a controlled amount of sulfur hexafluoride is mixed with oxygen or nitrogen during a specified period of the activation time; Before and after the specified period, there are periods of time during which the dielectric surface layer is exposed to plasma formed by an activating gas of oxygen or nitrogen; The preparation steps include preparing the chamber of the plasma activation device with a sulfur-containing gas plasma, and once the chamber is ready, treating the dielectric surface layer in the chamber with an oxygen- or nitrogen-containing plasma; The preparation steps include preparing the chamber of the plasma activation device with a plasma containing sulfur hexafluoride, and once the chamber is ready, treating the dielectric surface layer in the chamber with an oxygen- or nitrogen-containing plasma; The joining method involves heat treatment of the two substrates to be joined together; Both substrates have a dielectric surface layer of either silicon oxide or silicon nitride on their main surface side, and the fabrication steps are performed on each dielectric surface layer; The dielectric surface layer is composed of silicon oxide; The silicon oxide in the dielectric surface layer is bonded to nitrogen, preferably with a nitrogen / oxygen ratio of less than 0.5; The dielectric surface layer is composed of silicon nitride; The bonding method includes a step of cleaning one of the main surfaces of at least one of the two substrates between the preparation step and the contact step.

[0015] According to another aspect, the present invention provides a structure comprising a supporting substrate, a dielectric embedded layer disposed on and in contact with the supporting substrate, and a working layer disposed on and in contact with the dielectric embedded layer.

[0016] According to the present invention, the dielectric buried layer contains a sulfur dose greater than 3.0 E13 at / cm^2.

[0017] Other advantageous and non-limiting features of the invention can be carried out alone or in any technically feasible combination: Sulfur dosage greater than 3.5 E13 at / cm^2; The dielectric embedding layer contains a fluorine dosage greater than 4.0 E14 at / cm^2; The dielectric embedding layer is composed of silicon oxide; The silicon oxide bonded to nitrogen in the dielectric embedded layer is preferably bonded at a nitrogen / oxygen ratio of less than 0.5; The dielectric embedding layer is composed of silicon nitride. Simple Explanation of the Diagram

[0018] Other features and advantages of the present invention will become more apparent in the description of embodiments thereof, which is made with reference to the accompanying drawings, wherein: Figures 1a and 1b illustrate two examples of joining methods according to the present invention; Figure 2 illustrates a structure that can be obtained by the method of the present invention. Implementation

[0019] This invention relates to a method for bonding two substrates by molecular adhesion.

[0020] The two substrates can be of any type: insulating (such as quartz, glass, or piezoelectric materials), conductive (metallic), or semiconductor (such as silicon, gallium nitride, silicon carbide), and can take any available form: single-crystal, polycrystalline, or amorphous. The two substrates can each combine different types of materials. For example, one substrate can be formed from a piezoelectric layer, which can be transferred directly or through an intermediate layer (e.g., an adhesive) on a silicon-supported substrate. The two substrates can be of the same or different types. In particular, they can have different coefficients of thermal expansion, which limits the temperature at which the joint can be exposed, thereby increasing the bonding energy between the two substrates. The substrates can be in the form of circular wafers with standard diameters of 150, 200, or 300 mm, as is commonly used in the microelectronics field, but the invention is not limited to bonding two circular or standardized wafers.

[0021] Regardless of the nature and shape of the two substrates, these substrates are designed to be joined together through their "primary surfaces".

[0022] To facilitate this bonding, at least one of the two substrates has a dielectric layer, such as a silicon oxide layer or a silicon nitride layer, on one side of its main surface. This dielectric surface layer thus constitutes the bonding surface. The dielectric surface layer can have any thickness, typically between a few nanometers and a few micrometers, for example, between 1 nanometer and 5 micrometers. In some cases, each substrate may have its own dielectric surface layer.

[0023] The term "silicon oxide layer" refers to a thin layer primarily composed of silicon oxide. Other chemical elements may also be incorporated in trace amounts or at higher concentrations, as will be explained below. Nitrogen may be incorporated into this layer (to form a SiON layer), for example, at a nitrogen / oxygen ratio of less than 0.5. The proportion of nitrogen in the layer can be characterized by measuring its refractive index, which ranges from 1.44 for SiO₂ to 2.35 for Si₃N₄ at a wavelength of 620 nm. Therefore, a dielectric surface layer can be formed from SiO₁ ...

[0024] The dielectric surface layer can be formed using any suitable technique. When the substrate contains silicon, a silicon oxide dielectric surface layer can be formed by dry or wet oxidation. More generally, this dielectric surface layer can be formed by deposition. For example, this can involve LPCVD (low-pressure chemical vapor deposition) or PECVD (plasma-enhanced chemical vapor deposition) techniques. It can also involve a centrifugal deposition technique called "spin-on-glass". The deposited dielectric layer can be polished to reduce its roughness and make it compatible with bonding methods; its roughness is typically below 0.7 nm, which is the root mean square value of a 5-micron square field of view.

[0025] More generally, during the preparatory steps of preparing the two substrates, the exposed surfaces of the main faces can be prepared to make them clean (especially regarding particles and contaminants) and smooth (to provide a surface roughness lower than the values ​​mentioned above). For this purpose, the method according to the invention may include a pre-cleaning and / or polishing step of the substrates (as is well known) to condition the two substrates for subsequent bonding steps. The cleaning step may in particular include brushing the main faces while pouring a liquid (such as deionized water) onto these surfaces.

[0026] As described in the introduction above, these bonding steps include: a) bringing the principal surfaces of the two substrates into contact, and then; b) initiating and propagating a bonding wave between the principal surfaces of the two substrates to bond the principal surfaces together.

[0027] For illustrative purposes only, these bonding steps can be performed in an automated bonding apparatus. In this case, the first substrate is manipulated by the arm of the apparatus handling system to position it on a flat support, exposing its principal surface. Then, the second substrate is manipulated by the arm of the handling system to contact the first substrate on one side of its principal surface. An actuator is moved to apply a moderate pressure of several Newtons to the exposed surface of the second substrate, thereby establishing close contact between the first and second substrates and initiating and propagating the bonding wave. This pressure can be applied anywhere, particularly at the periphery or center of the exposed surface of the second substrate.

[0028] In all cases, regardless of the method of initiating the bonding wave, the bonding wave propagates from the starting position to the periphery of the joint to ensure close contact between the two substrates across their entire primary surface. The wave propagation is not instantaneous but can last for several seconds, depending on the substrate size.

[0029] A plasma activation process for preparing dielectric surface layers is also typically provided, usually using oxygen or nitrogen plasma. This preparation step may involve placing the substrate in a chamber of a plasma activation apparatus, such as on a flat support disposed within the chamber, to expose the surface layer to a plasma source. An oxygen or nitrogen stream is introduced into the plasma source of the apparatus at a controlled flow rate, and then a plasma of this gas is generated (e.g., through a variable magnetic or electric field). The plasma species are directed to the dielectric surface layer to activate the surface of the substrate. These operations can be performed at atmospheric pressure or below. This activation can typically increase the bonding energy between the two substrates immediately after bonding or after the application of a strengthening heat treatment.

[0030] The applicant has conducted preliminary experiments to examine the effect of plasma activation on the bonding quality of molecular adhesion obtained from the silicon oxide surface layer, as described above. Specifically, these preliminary experiments attempted to evaluate the benefits of plasma activation when combined with sulfur hexafluoride (SF6). The results are summarized in the table below.

[0031] In these preliminary experiments, the substrate consisted of pairs of 150 mm lithium tantalate and silicon wafers, each with a 30 nm and a 650 nm (nitrogen-free) silicon oxide surface layer deposited by PECVD. Before bonding using the molecular adhesion method described above, each surface layer underwent a 30-second plasma activation time at 150 W. After a 60-minute intensified heat treatment at 180°C, the bonding energy was measured using the double cantilever beam method. All experiments in the table were performed in a CCP (capacitively coupled plasma) plasma apparatus, with the exception of Experiment 10, which was performed in an ICP (inductively coupled plasma) apparatus.

[0032] [Control Experiment] Experiment 1 in the table constitutes a control experiment corresponding to the activation of the surface layer solely with oxygen plasma. During this activation (and during all experiments listed in the table), the oxygen flow rate was set at 75 SCCM. The bonding energy was measured at 1905 mJ / m², the bonding wave propagated for 6 seconds, and numerous annular defects appeared in the joint. SIMS and TXRF analyses of the silicon oxide surface layer revealed a fluorine doping dose of 3.0 E+13 at / cm² and a sulfur doping dose of 2.4 E+12 at / cm², corresponding to trace amounts of these elements.

[0033] [experiment] [2] [to] [6] In these experiments, a controlled amount of SF6 was introduced during oxygen-based plasma activation, and this controlled amount was increased in experiments 2 through 6. The controlled amount of SF6 was obtained by adjusting the flow rate and duration. This introduction was preferred at the beginning of plasma activation or in the middle of activation, as the latter is more stable and the phenomenon is better controlled. It was observed that significant sulfur and fluorine dosages were incorporated into the surface layer during the activation process conducted in experiments 2 through 6. In experiment 5, the controlled amount of SF6 was introduced in two sequences. The first sequence introduced 3 SCCM of SF6 for 3 seconds, followed by a flow pause, and then the next sequence introduced 5 SCCM of SF6 for 5 seconds.

[0034] It should be remembered that the unit of measurement SCCM (standard cubic centimeters per minute) is the physical unit (in cm 3 / min) for the mass flow rate of a gas at a certain density under standard temperature and pressure conditions.

[0035] [experiment] [7] [to]

[10] In these experiments, during the preparatory step, a plasma device source activated at 2000 W was supplied with a mixed gas containing approximately 20% SF6 (10 SCCM) in 80% oxygen (50 SCCM), and no substrate was placed in this device. The purpose of this preparatory step was to prepare the chamber of the plasma activation device, with the inner surface of this chamber coated with sulfur- and fluorine-containing species. Experiments were then conducted in this prepared chamber using a silicon oxide surface layer of a conventional plasma-activated substrate based solely on oxygen. During oxygen plasma activation, the sulfur- and / or fluorine-containing species coated on the inner surface of the chamber were peeled off from these surfaces and incorporated into the plasma throughout the activation process.

[0036] Compared to Control Experiment 1, all other experiments observed resulted in a significant increase in the measured bonding energy. The effect of fluorine has been reported elsewhere and appears to be due to the incorporation of fluorine into the surface layer at a dose greater than 4.0 E14 at / cm^2.

[0037] Even more surprisingly, these results show that the velocity of the bonding wave is affected by the sulfur dosage introduced into the dielectric surface layer: the higher the dosage, the greater the velocity reduction. This effect was obtained without significantly affecting the bonding energy, and in all cases, the bonding energy remained higher than that in the control experiment. Therefore, it appears that the velocity of the bonding wave can be controlled by adjusting the sulfur dosage introduced into the dielectric surface layer.

[0038] If the sulfur dosage is sufficient, i.e. greater than 3.0 E13 at / cm^2 and preferably greater than 3.5 E13 at / cm^2, then the wave velocity is low enough that the annular defects no longer exist to a significant degree, as shown in Experiments 4, 5, 6, and 10.

[0039] Other experiments can reproduce these results for dielectric surface layers, except for the silicon oxide (SiO2) layer used in the experiments listed in the table. Specifically, similar results can be obtained when the dielectric surface layer is formed of silicon oxide with a nitrogen / oxygen ratio strictly greater than 0, and when this dielectric surface layer is formed of silicon nitride.

[0040] The applicant has proposed a method for molecular adhesion bonding based on these experiments and observations, which can prevent or at least limit the occurrence of ring-hole type bonding defects. Furthermore, this benefit can be obtained without affecting the bonding energy of the bonded joint, or even with a significantly improved bonding energy.

[0041] In the sequence consisting of the aforementioned steps of contacting the main surfaces of the two substrates and initiating and propagating the bonding wave between the two substrates, the method according to the invention therefore provides a step of preparing a dielectric surface layer prior to the contact step, with the aim of introducing a sulfur dose greater than 3.0 E13 at / cm^2 into this dielectric surface layer. It is advantageous when this dose is greater than 3.5 E13 at / cm^2. This dose can be measured by TXRF or by SIMS, as in the experiments reported above.

[0042] It is more advantageous if the dose corresponds to a thickness of no more than 10 nm (or even no more than 7.5 nm) immediately below the main surface of the substrate, since the bonding properties are affected and can only reach this thickness at most.

[0043] To benefit from improved bonding energy while limiting the occurrence of toroidal defects, an introduction of a fluorine dosage of 4.0 E14 at / cm² into the dielectric surface layer is also advantageous. This dosage corresponds to a thickness of no more than 10 nm (or even no more than 7.5 nm) immediately below the substrate's main surface, a thickness that is limited due to the influence of bonding properties. This dosage can be measured using TXRF or SIMS.

[0044] One approach that allows for the introduction of sulfur at a target dose may include activating the dielectric surface layer plasma with a sulfur-containing gas, such as sulfur hexafluoride. This approach is schematically illustrated in Figure 1a.

[0045] A sulfur-containing gas (e.g., sulfur hexafluoride) can be mixed with an activating gas (composed of, for example, oxygen or nitrogen). The dosage of this mixture is configured to reach or exceed the target dosage of sulfur (and, as needed, the target dosage of fluorine) at the end of the plasma activation step.

[0046] The duration of plasma activation is typically between 15 seconds and two minutes. More specifically, the activation duration is the time from the moment the plasma ignites in the device source until it extinguishes. During this period, the dielectric surface layer is exposed to the plasma formed by an activating gas of oxygen or nitrogen.

[0047] Sulfur-containing gases can be introduced at a controlled flow rate throughout the entire period or only for a portion of the period, to mix with oxygen or nitrogen within a specified period of activation. This specified period preferably excludes the instant of plasma ignition, as ignition would cause sulfur- and fluorine-containing species present in the plasma to be introduced into the surface layer in a non-reproducible manner, making it impossible to achieve perfect control over the introduced dosage and potentially detrimental to good process control, especially when the target dose of sulfur (and the target dose of fluorine, if used) is close to the limit dose. Therefore, it is advantageous to expose the dielectric surface layer to plasma formed by an activating gas of oxygen or nitrogen for a period before and after the specified period (during which the dielectric surface layer is exposed to the plasma formed by the sulfur-containing gas).

[0048] Referring to Figure 1a, after the step of preparing the dielectric surface layer to introduce a sulfur dose greater than 3.0 E13 at / cm^2 into the layer, there is a step of contacting the main surfaces of the two substrates, followed by a step of initiating and propagating a bonding wave between the main surfaces of the two substrates to bond the two surfaces together.

[0049] As described above, a step of cleaning the main surfaces of the substrate can be included between the preparation step and the contact step. This cleaning step may correspond to or include brushing the surface while pouring a liquid (such as demineralizing water) onto it. Advantageously, the present invention applies the cleaning step to the main surfaces of both substrates, and then the main surfaces are joined together.

[0050] According to another method, as shown in Figure 1b, the preparation steps include preparing a chamber for plasma activation using a sulfur-containing gas plasma, and once the chamber is ready, treating at least one dielectric surface layer of a substrate with an oxygen- or nitrogen-containing plasma in the chamber.

[0051] As shown in the figure, when there is no substrate to be prepared in the chamber, plasma activation is performed using a sulfur-containing gas plasma. The sulfur-containing gas can be sulfur hexafluoride mixed with an activating gas (such as oxygen or nitrogen). The purpose of this substep is to coat the inner surface of the chamber with sulfur-containing species (and, if necessary, fluorine-containing species), which are then peeled off from the inner surface of the chamber during plasma activation of the dielectric surface layer provided in the substep following the preparation step. Parameters (especially the flow rate of the sulfur-containing gas, the duration of the plasma, and its power) are adjusted to generate sufficient species and incorporate them into the surface layer during the substep following plasma activation.

[0052] Referring to Figure 1b, after the step of preparing the dielectric surface layer according to this method, a step of bringing the main surfaces of the two substrates into contact is also performed, followed by a step of initiating and propagating a bonding wave between the main surfaces of the two substrates to bond the two surfaces together. The cleaning step described in the preceding method can also be provided.

[0053] The dielectric surface layer is advantageously a silicon oxide layer, and nitrogen can be incorporated therein in any desired proportion, or it can be a silicon nitride layer.

[0054] Before, during, or at the end of the above-described bonding method, further steps may be applied to the substrate, especially steps in a method of creating a structure by molecularly bonding two substrates together.

[0055] In particular, it is conceivable to apply a strengthening heat treatment to the joint, and to reduce the thickness of at least one of the two substrates to form a single layer. This reduction in thickness can be achieved by thinning the substrate (e.g., by grinding) or by fracturing. In this case, it is conceivable to introduce a lightweight species into the substrate to be fractured, typically prior to the contact step of the joining method, in order to form a brittle fracture surface in the substrate where the thickness needs to be reduced.

[0056] In this case, the bonding method is part of a layer transfer method, which at the end of which (refer to Figure 2) yields a structure 1, which includes a support substrate 2, a dielectric embedded layer 3, which is advantageously a silicon oxide or silicon nitride layer disposed on and in contact with the support substrate 2, and a working layer 4 disposed on and in contact with the dielectric embedded layer 3.

[0057] The embedded layer has a bonding interface, which may correspond to its interface with the supporting substrate 2, its interface with the working layer 4, or be bonded to its own thickness, depending on whether the method uses a single dielectric surface layer disposed on one of the two substrates, or two dielectric surface layers disposed on each of the main surfaces of the two substrates.

[0058] The dielectric buried layer 3 contains a sulfur dosage greater than 3.0 E13 at / cm^2, advantageously greater than 3.5 E13 at / cm^2. It may have a fluorine dosage greater than 4.0 E14 at / cm^2. As previously mentioned, when the dielectric buried layer is composed of silicon oxide, it may be doped with other species, especially nitrogen. It is advantageous when these dosages correspond to a thickness of no more than 10 nm on one side and / or the other side of the bonding interface.

[0059] The present invention is not limited to the embodiments described herein, and variations may be made to its application without departing from the scope of the invention as defined by the claims.

[0060] Although this specification uses sulfur hexafluoride as an example of a sulfur-containing (and fluorine-containing) gas, the invention is by no means limited to using only this gas. It is conceivable to use any sulfur-containing gas instead, such as sulfonyl fluoride (SO2F2) or tetrafluorothionyl sulfite (SO2F4).

[0061] The preparation steps according to the present invention can be applied to a single dielectric surface layer formed on a substrate, even if both substrates include a dielectric surface layer. Similarly, when each substrate has a dielectric surface layer, the preparation steps of the present invention can be applied to each substrate.

[0062] Finally, according to an undescribed alternative, sulfur-containing and fluorine-containing species (if present) may be incorporated into the dielectric surface layer during its formation (e.g., during the oxidation or nitriding of the silicon-containing substrate, or during its deposition).

[0063] 1: Structure 2: Supporting base material 3: Dielectric Embedded Layer 4: Working layer

[0064]

Claims

1. A method for molecularly bonding two substrates, each substrate having a principal facet, at least one of the two substrates having a dielectric surface layer on its principal facet, the method comprising the steps of: (a) bringing the principal facests of the two substrates into contact, and then; (b) initiating and propagating a bonding wave between the principal facests of the two substrates to bond the principal facests together; the method is characterized in that it includes, prior to the contact step, a step of preparing the dielectric surface layer with the aim of introducing a sulfur dose greater than 3.0 E13 at / cm^2 into the dielectric surface layer.

2. The method of claim 1, wherein the sulfur dosage is greater than 3.5 E13 at / cm^2.

3. The method of claim 1 or 2, wherein the step of preparing the dielectric surface layer also preferably introduces a fluorine dose greater than 4.0 E14 at / cm^2 into the dielectric surface layer.

4. The method of claim 1 or 2, wherein the preparation step includes activating the dielectric surface layer with a sulfur-containing gaseous plasma, such as a sulfur hexafluoride-containing plasma.

5. The method of claim 4, wherein the plasma also includes oxygen or nitrogen.

6. The method of claim 5, wherein the activation time of the dielectric surface layer is between 15 seconds and 2 minutes.

7. The method of claim 6, wherein throughout the activation process, the dielectric surface layer is exposed to an oxygen or nitrogen plasma, and a controlled amount of sulfur hexafluoride is mixed with the oxygen or nitrogen during a specified period of the activation time.

8. The method of claim 7, wherein before and after the specified period, the dielectric surface layer is exposed to plasma formed by an activating gas of oxygen or nitrogen for a period of time.

9. The method of claim 1 or 2, wherein the preparation step includes preparing a chamber for plasma activation with a sulfur-containing gas plasma, and once the chamber is ready, treating the dielectric surface layer in the chamber with an oxygen- or nitrogen-containing plasma.

10. The method of claim 1, which includes heat treatment of the two substrates joined together.

11. The method of claim 1, wherein the two substrates have a dielectric surface layer of silicon oxide or silicon nitride on their principal side, and the preparation step is performed on each dielectric surface layer.

12. The method of claim 1, wherein the dielectric surface layer is composed of silicon oxide.

13. The method of claim 1, comprising a step of cleaning the principal surface of at least one of the two substrates between the preparation step and the contact step.

14. A structure (1) obtained by a method of molecularly bonding two substrates, comprising a supporting substrate (2), a dielectric buried layer (3) disposed on and in contact with the supporting substrate (2), and a working layer (4) disposed on and in contact with the dielectric buried layer (3), characterized in that the dielectric buried layer (3) contains a sulfur dosage greater than 3.0 E13 at / cm^2.

15. The structure (1) of claim 14, wherein the sulfur dosage is greater than 3.5 E13 at / cm^2.

16. The structure (1) of claim 15, wherein the dielectric buried layer (3) contains a fluorine dose greater than 4.0 E14 at / cm^2.

17. The structure (1) of any one of claims 14 to 16, wherein the dielectric buried layer is composed of silicon oxide.

Images