Laser welding method

Abstract

The invention relates to a laser welding method comprising the following steps: 1) placing (62) a first face of a first semiconductor part directly against a second face of a second semiconductor part in order to form a joining plane; then 2) producing welding spots in the joining plane using laser pulses, wherein the reflection of a laser pulse by a defect in the joining plane is minimised: a) by emitting the laser pulse in a wavelength range in which this reflection is minimised; b) by modifying the thickness of the defect in order to obtain a thickness that minimises the reflection of the laser pulse; or c) by treating (122) the first and / or the second face in order to form therein cavities having different depths, which cavities are uniformly distributed over this face.

Description

[0001] Laser welding process

[0002] [1] The invention relates to a method for laser welding a first semiconductor part to a second semiconductor part. The invention also relates to laser welding stations for implementing this laser welding method.

[0003] [2] Laser welding is a technology that is widely used in industry today. However, until now, laser welding has not been used to weld two semiconductor materials together. However, it has already been shown in the following article that welding a part made of semiconductor material to another part made of semiconductor material is possible when a thin layer of gold is present at the joint plane between these two parts: Maxime Chambonneau et al: “Ultrafast laser welding of silicon”, Advanced Photonics Research, 26 / 02 / 2023.

[0004] [3] It has also been shown in the following article that welding a part of semiconductor material directly onto another part of semiconductor material was possible: Pol Sopeha et al: “Transmission Laser Welding of Similar and Dissimilar Semiconductors Materials”, Laser Photonics Reviews, 9 / 08 / 2022. Hereinafter, this article is referred to as “article Al”.

[0005] [4] Article Al reveals that there are physical limitations that prevent an efficient and reproducible process for laser welding a part made of semiconductor material directly onto another part made of semiconductor material. These physical limitations arise from the fact that there are defects in the joint plane between the two parts to be welded. These defects form Fabry-Pérot cavities that reflect the laser pulses and therefore prevent the formation of a weld point at the location of these defects. Such a problem appears when the materials to be welded are semiconductors because semiconductor materials have the particularity of exhibiting high reflection indices. Because of this problem, the adhesion force between two parts made of semiconductor material directly welded to each other is not as strong as expected.[5] The invention aims to overcome this drawback by proposing a method of laser welding two parts made of semiconductor material directly onto each other which makes it possible to obtain greater adhesion strength.

[0006] [6] The invention is set forth in the attached set of claims.

[0007] [7] The invention will be better understood on reading the description which follows, given solely as a non-limiting example and made with reference to the drawings in which:

[0008] - figure 1 is a schematic illustration, in perspective, of two pieces of semiconductor material,

[0009] - figure 2 is a schematic illustration, in perspective, of a stack of the parts of figure 1,

[0010] - figures 3, 5, 7, 9, 11 and 13 are schematic illustrations of different embodiments of a welding station for the parts of figures 1 and 2,

[0011] - Figures 4, 6, 8, 10, 12 and 14 are flowcharts of different laser welding processes using the welding stations of Figures 3, 5, 7, 9, 11 and 13.

[0012] [8] In this description, the terminology, conventions and definitions of the terms used in this text are introduced in a chapter I. Then, detailed examples of embodiments are described in a chapter II with reference to the figures. In a chapter III, variants of these embodiments are presented. Finally, the advantages of the different embodiments are specified in a chapter IV.

[0013] [9] Chapter I: Definitions, terminologies and conventions:

[0014]

[0010] In the figures, the same references are used to designate the same elements.

[0015]

[0011] In the remainder of this description, the characteristics and functions well known to those skilled in the art are not described in detail.

[0016]

[0012] The figures are oriented relative to an orthogonal XYZ coordinate system, where the X and Y directions are horizontal and the Z direction is vertical. Terms such as "above", "below", "top", "bottom", "upper", "lower" are defined relative to the Z direction.

[0017]

[0013] The symbol “*” denotes scalar multiplication.

[0014] The expression “an element made of a material A” or the expression “an element of material A” means that material A represents 90% or 95% or 99% of the mass of this element.

[0018]

[0015] The term "semiconductor" refers to a band gap material, i.e. a material having a valence band and a conduction band separated by an energy range, called the "band gap" where no electronic state exists (in the ideal case of a defect-free semiconductor) and where the Fermi level is located. The valence and conduction bands are close enough to this level to be populated by electrons or holes. For this, the width of the band gap is less than 5 eV and often less than 3 eV or 2.5 eV at a temperature of 0 K.

[0019]

[0016] The infrared range refers to the wavelength range, in a vacuum, which extends from 700 nm to 20 pm. This infrared range is subdivided into two sub-ranges, namely the near infrared sub-range and the mid-infrared sub-range. The near infrared sub-range extends from 700 nm to 2000 nm. The mid-infrared sub-range extends from 2 pm to 20 pm.

[0020]

[0017] The transparency domain of a part is the wavelength range for which the transmittance of this part is greater than 10% or 50% or 70% and, preferably, greater than 90% or 95%. This wavelength range is greater than a wavelength À ta , where the wavelength A ta is defined by the following relation: To ta =0.9*h*c / Dl, where

[0021] - h is Planck's constant (h="6.62607015*10-34 Js),

[0022] - c is the speed of light (c=2.99792458 *108 m / s), and

[0023] - DI is the band gap of the semiconductor material expressed in joules.

[0024] Thus, the transparency range of silicon is greater than 1150 nm for example.

[0025]

[0018] A “micrometric” element is an element whose largest dimension is less than 500 pm or 1 mm. A “sub-micrometric” element is an element whose largest dimension is less than 1 pm. Generally, the largest dimension of a micrometric or sub-micrometric element is greater than 0.2 pm.

[0026]

[0019] The largest dimension of an element is equal to the length of the parallelepiped of smallest volume which entirely contains this element.

[0027]

[0020] In this text, a laser pulse designates a single light pulse or a train of light pulses used to deposit energy at a welding point on which this laser pulse is focused. A train of light pulses is known by the English term "burst". In the case of a train of light pulses, the light pulses are sufficiently close to each other so that the material does not have time to relax between two light pulses so that the energy provided by each light pulse accumulates, for example, until it exceeds the melting threshold of the semiconductor materials to be welded. Typically, the frequency of the light pulses in a train of light pulses is greater than 100 kHz.

[0028]

[0021] The power spectrum of a laser pulse is bell-shaped or Gaussian. The width of a laser pulse refers to the width at half-maximum of this bell or Gaussian. This width is equal to Àsup - Àjnf , WHERE Àsup and Xinf are the upper and lower limits, respectively, at which the height of the power spectrum of the laser pulse is equal to half of its maximum height.

[0029]

[0022] The median wavelength of a laser pulse means the wavelength equal to (À sup + X in f) / 2. Thus, the laser pulse is centered on this median wavelength.

[0030]

[0023] In this text, the expression "a range of wavelengths in which the transmitted energy is sufficient to melt the joined faces of two semiconductor parts" designates a continuous range of wavelengths of the power spectrum of a laser pulse within which the power of the laser pulse is sufficient to melt the joined faces when this laser pulse is focused on the joint plane located between these two joined faces. This range of wavelengths is located within the limits À sup and To inf of the laser pulse.

[0031]

[0024] The median wavelength of a wavelength range is the wavelength located in the middle of this wavelength range.

[0032]

[0025] A “broad spectrum” laser pulse means a laser pulse whose width is greater than λ / 2, where λ is the median wavelength of this laser pulse. Typically, the width of a broad spectrum laser pulse is greater than 100 nm. A broad spectrum laser pulse may comprise several distinct and non-overlapping wavelength ranges in which the transmitted energy is sufficient to melt the adjoining faces of two semiconductor parts.

[0033]

[0026] A "broad spectrum" laser source is a laser source that emits broad spectrum laser pulses.

[0027] A monochromatic laser pulse is a laser pulse that is not a broad spectrum laser pulse. Thus, the width of a monochromatic laser pulse is less than 100 nm and, typically, less than 10 nm. Typically, a monochromatic laser pulse has at most a wavelength range in which the transmitted energy is sufficient to melt the adjoining faces of two semiconductor parts.

[0034]

[0028] A monochromatic laser source is a laser source that emits monochromatic pulses.

[0035]

[0029] Absorbance characterizes the ability of a face to absorb an incident optical signal at a given wavelength. Here, absorbance is measured as described in the following article: C. Wu et al.: “Near-unity below-band gap absorption by microstructured silicon”, Appl. Phys. Lett. 78 (13), pp. 1850-1852, 2001. Thus, absorbance A is defined by the following relationship: A = 1- R - T, where R is the reflectance and T is the transmittance of the face. Reflectance R and transmittance T are measured using a spectrophotometer.

[0036]

[0030] In this text, a defect designates a defect at a joint plane between two faces joined against each other. Such a defect is a cavity which traps a fluid whose refractive index is substantially different from the refractive indices of the semiconductor materials in which the joined faces are made. Typically, the difference between the refractive index of the fluid trapped in the cavity and the refractive index of the semiconductor materials is greater than 0.2 or 0.4. Under these conditions, as observed in article A1, such a defect forms a Fabry-Perot cavity capable of reflecting an incident laser pulse if the thickness of this defect is close to an integer multiple of λ / 2, where λ is the median wavelength of the incident laser pulse. Typically, the fluid trapped in a defect is air.

[0037]

[0031] Among the different monochromatic laser pulses that a defect is capable of reflecting, there is one whose median wavelength λ e is the smallest. The thickness of a defect is defined as being equal to À e / 2 in the case where the fluid trapped in this defect is air.

[0038]

[0032] Subsequently, two refractive indices are said to be “close” if the difference, in absolute value, between the real parts of these two refractive indices is less than 0.2.

[0033] Chapter: Examples of embodiments

[0039]

[0034] Figure 1 shows two parts 2 and 4 before they are welded together. These parts 2 and 4 are both made of semiconductor material. For illustration purposes, parts 2 and 4 are made of silicon. Each of these parts 2 and 4 has a contact face. The contact faces of parts 2 and 4 bear the reference numerals 6 and 8, respectively. Faces 6 and 8 are the faces of parts 2 and 4 intended to be assembled together by laser welding to form a stack. For this purpose, faces 6 and 8 are flat. Typically, the roughness Ra of faces 6 and 8, measured conventionally, is less than 1 μm and, preferably, less than 100 nm or 10 nm. Each of parts 2 and 4 may include components such as micrometric electromechanical systems and / or micrometric fluidic systems.Micrometric electromechanical systems are better known by the acronym MEMS (MicroElectroMechanical Systems). Typically, the stack of parts 2 and 4 forms part of a three-dimensional integrated circuit in which the different components that constitute it are distributed in different layers stacked on top of each other.

[0040]

[0035] Figure 2 represents the stack 10 obtained by joining the face 6 to the face 8. The stack 10 comprises a joining plane 12. The plane 12 is located at the interface between the faces 6 and 8 directly in mechanical contact with each other. In Figure 2, a trajectory 14, contained in the plane 12 is represented. This trajectory 14 is defined by a series of welding points to be produced, one after the other, in the plane 12. An example of trajectory 14 is visible, by transparency through the part 2. By way of illustration, the trajectory 14 extends from a starting point 16 to an arrival point 17. Between the points 16 and 17, the trajectory comprises tens or hundreds of successive welding points.Each welding point corresponds to a location in plane 12 where the contact faces 6 and 8 must be melted so that after this localized melting of faces 6 and 8, the materials of faces 6 and 8 interpenetrate to form, on cooling, a correct welding point which mechanically bonds, without any degree of freedom, parts 2 and 4. In this example, path 14 zigzags in plane 12 to form meanders.

[0041]

[0036] 11.1 - Appropriate selection of wavelengths:

[0037] Figure 3 shows a laser welding station 20 used to weld the parts 2 and 4 together. The station 20 comprises a laser pulse emitter 22. The duration of each laser pulse is typically greater than 1 ps or 60 ps and, generally, less than 100 ns or 100 ps or 1 ms. Below 1 ps it is difficult to control the propagation of the laser pulse through the semiconductor material and therefore to correctly focus this laser pulse on the joint plane. Above 1 ms, the energy of the laser pulses is generally insufficient to weld the semiconductor parts 2 and 4 together. In addition, the transmitter 22 is capable of transmitting, by passing only through the part 2, sufficient energy to melt the faces 6 and 8 at each welding point of the path 14 through the part 2.For example, in the case of silicon parts 2 and 4, the energy of a laser pulse must be able to achieve an energy density at the welding point greater than 5 kJ / cm. 3 to melt faces 6 and 8. The energy of a laser pulse to melt faces 6 and 8 is determined experimentally. Typically, the energy of a laser pulse is between 100J and 10J. The energy of a pulse is greater the longer its duration.

[0042]

[0038] More specifically, the transmitter 22 is capable of transmitting, using a laser pulse, this sufficient energy both in a first wavelength range and in a second distinct wavelength range. These first and second wavelength ranges are both located in the transparency range of the part 2. In the case where the part 2 is made of silicon, each of the first and second ranges is included in the near infrared range. In the near infrared range, the transmittance of the part 2 is greater than 10% and, typically, greater than 95%. The first and second wavelength ranges are distinct and do not overlap. These first and second ranges are centered on median wavelengths, respectively, λ and λ. The first and second wavelength ranges are constant and do not vary during the execution of the laser welding method using the station 20.

[0043]

[0039] The first and second wavelength ranges are chosen so that, for the majority of defects located at the plane 12, the energy transmitted in at least one of the first and second wavelength ranges causes the faces 6 and 8 to melt at the weld point to be made. For this, at least the energy transmitted in one of these two wavelength ranges must not be reflected by this defect. For this purpose, the difference between the median wavelengths λi and λ2 is far from an integer multiple of λi / 2 and from an integer multiple λ2 / 2. For this, the difference between the wavelengths λi and λ2 does not satisfy any of the following conditions:

[0044] - Condition 1): there exists an integer m such that + Ài / k, and

[0045] - Condition 2): there exists an integer p such that + À2 / k where k is an integer equal to ten or eight.

[0046]

[0040] In this example, the wavelengths λi and λ2 are such that the difference | λi - λ2| is between n*λi / 4 - λi / 16 and n*λi / 4 + λi / 16 or between n*λ2 / 4 - λ2 / 16 and n*λ2 / 4 - λ2 / 16, WHERE n is an odd integer. Preferably, the difference | λi - λ2| is less than λi / 2 or λ2 / 2. For example, here, the wavelength λ2 is equal to λi + λi / 4 or the wavelength λi is equal to λ2 - λ2 / 4. For illustration, the wavelength λi is equal to 1550 nm and the wavelength λ2 is equal to 1937.5 nm.

[0047]

[0041] In this first exemplary embodiment, the transmitter 22 is capable of emitting monochromatic laser pulses centered on the two distinct wavelengths λ and λ. Thus, in this first exemplary embodiment, the first and second wavelength ranges are located in distinct laser pulses and not within the same broad spectrum laser pulse.

[0048]

[0042] Furthermore, here, the emitter 22 is designed to simultaneously emit, on the stack 10, the laser pulses at wavelengths λ and λ2. For this purpose, for example, the emitter 22 comprises two laser sources 24 and 26. The laser sources 24 and 26 are laser sources capable of emitting the monochromatic laser pulses centered on the wavelengths, respectively, λ and λ2. For example, each of these laser sources 24 and 26 is identical to the laser source used and described in the article A1 except for the wavelength at which it emits is different. Thus, the sources 24 and 26 are, for example, erbium-doped fiber sources.

[0049]

[0043] Here, these laser sources 24 and 26 are arranged relative to each other so that the laser pulses from these two sources are focused on the same focal point 28. The point 28 is located in the joining plane 12.

[0050]

[0044] Station 20 also includes:

[0051] - a positioning table 30, - a device 32 for mechanically compressing the face 6 against the face 8, and

[0052] - a 34 control unit.

[0053]

[0045] The table 30 comprises a plate 40 and an electric actuator 42. The plate 40 extends mainly in a horizontal plane. The actuator 42 is capable of moving the plate 40 in the X, Y and Z directions.

[0054]

[0046] The device 32 presses the part 2 against the part 4 so as to reduce the thickness of the defects located at the joint plane 12. For this, the device 32 forces the part 2 against the part 4 so that the pressure of the face 6 against the face 8 is greater than 0.1 MPa and, preferably, greater than 0.5 MPa. For example, here the device 32 comprises an upper jaw 44 and a lower jaw 46. The upper jaw 44 rests only on the periphery of the upper face of the stack 10. Thus, the upper face of the stack 10 remains directly exposed to the laser pulses of the emitter 22. The lower jaw 46 rests at least on the periphery of the lower face of the stack 10 and, here, on the entire lower face of the stack 10.

[0055]

[0047] The device 32 also comprises a mechanism 48 for clamping the jaw 44 against the jaw 46 to achieve the desired pressure between the faces 6 and 8. For example, this mechanism 48 is a manual screw and nut mechanism or an electric mechanism comprising a controllable actuator which, when actuated, exerts a clamping force on the jaw 44 against the jaw 46.

[0056]

[0048] Here, moreover, the jaw 46 is fixed, without any degree of freedom, on the upper face of the plate 40. Thus, the device 32 also serves to fix the stack 10 on the upper face of the plate 40 so that when the plate 40 is moved, the stack 10 also moves.

[0057]

[0049] The control unit 34 is configured to control the transmitter 22 and the positioning table 30 to execute the welding method of FIG. 4. For this purpose, the unit 34 comprises a microprocessor 50 capable of executing instructions and a memory 52 comprising the instructions and data necessary for executing the method of FIG. 4. In particular, the memory 54 comprises the coordinates, for example in the XYZ reference frame, of each welding point to be produced.

[0058]

[0050] Figure 4 shows a process for laser welding the part 2 onto the part 4 using the welding station 20.

[0051] The process begins with a pre-adjustment phase 60. During this phase 60, the positioning of the sources 24 and 26 relative to the upper face of the stack 10 is adjusted so that the focal points of the laser pulses from these two sources coincide and are located on point 28. For example, for this, the procedure is as described in chapter 2.1 of article A1. It is emphasized here that when the gap | Ài - À2| is less than Δi / 2 or Δ2 / 2, the heights of the lenses of the sources 24 and 26, relative to the joining plane 12, for which the focal points are located in this plane 12 are identical or very close so that it is not necessary to introduce an offset between the height of the lens of the source 24 and the height of the lens of the source 26.

[0059]

[0052] Optionally, during phase 60, the power of the laser pulses from sources 24 and 26 is also adjusted to obtain the fusion of faces 6 and 8 at each welding point. This power is then constant.

[0060]

[0053] Once phase 60 is completed, during a step 62, the face 6 of the part 4 is attached to the face 8 of the part 6 to form the stack 10.

[0061]

[0054] The stack 10 is then arranged between the jaws 44 and 46 so that the part 2 is located above the part 6.

[0062]

[0055] Then, during a step 64, the device 32 is actuated to compress the face 6 against the face 8. From this moment, the stack 10 is also firmly held on the plate 40.

[0063]

[0056] During a welding step 66, the unit 34 automatically controls the table 30 and the transmitter 22 to transmit, at each welding point to be made in the joining plane 12, sufficient energy to melt the faces 6 and 8 at the location of this welding point to be made. For this, sufficient energy is transmitted through the part 2 and using the laser pulses from the sources 24 and 26 focused on this welding point to be made.

[0064]

[0057] For example, during step 66, the unit 34 controls the actuator 42 to position the focal point 28 on the next welding point to be made. For this, in this simplified example, the actuator moves the plate 40 only in a horizontal plane because the joining plane 12 is also horizontal.

[0065]

[0058] Then, once the focal point 28 is correctly positioned on the next weld point to be made, the unit 34 controls the sources 24 and 26 to simultaneously emit the laser pulses at wavelengths λ and λ2.

[0059] When there is no defect at the location of the weld point to be made, each of the laser pulses at wavelengths λ and λ2 provides sufficient energy, on its own, to cause the faces 6 and 8 to melt at the location of the weld point to be made.

[0066]

[0060] When there is a defect at the location of the weld point to be made whose thickness is close to or equal to an integer multiple of λi / 2, the laser pulse at wavelength λi is reflected and therefore does not provide sufficient energy to melt the face 8. On the other hand, in this case, the laser pulse at wavelength λ2 is distant from an integer multiple of λi / 2. Under these conditions, the laser pulse at wavelength λ2 is correctly absorbed by the face 8 so that it alone causes the faces 6 and 8 to melt and therefore a correct weld point to be obtained despite the presence of the defect.

[0067]

[0061] The same phenomenon occurs when there is a defect at the location of the weld point whose thickness is close to or equal to an integer multiple of λ2 / 2. In this case, it is the laser pulse at wavelength λ2 which is reflected and does not provide sufficient energy to melt the face 8. On the other hand, in this case, the laser pulse at wavelength λi is distant from an integer multiple of λ2 / 2. Under these conditions, the laser pulse at wavelength λi is correctly absorbed by the face 8 so that it alone allows the faces 6 and 8 to be melted and therefore a correct weld point to be obtained despite the presence of the defect.

[0068]

[0062] Thus, whatever the thickness of the defect, the method of FIG. 1 makes it possible to produce a correct welding point without it being necessary to have prior knowledge of the thickness of the defects present in the joint plane 12.

[0069]

[0063] Step 66 is repeated for each welding point of the trajectory 14 between points 16 and 17.

[0070]

[0064] Once the arrival point 17 is reached, step 66 stops and during a step 68, the stack 10 is removed from the welding station 20. For this, the jaws 44, 46 of the device 32 are first loosened and then the stack 10 is removed.

[0071]

[0065] Figure 5 shows a welding station 70. The welding station 70 is identical to the welding station 20 except that the emitter 22 is replaced by an emitter 72. The emitter 72 comprises a single monochromatic laser source 74 capable of transmitting sufficient energy to melt the faces 6 and 8 at each welding point of the path 14 through the part 2. The characteristics of the laser pulse emitted by the source 74 are for example identical to the characteristics of the laser pulses of the sources 24 and 26 except that this laser pulse is centered on a single median wavelength λ. The wavelength λ is in the transparency range of the part 2. However, in this embodiment, to prevent the laser pulse at the wavelength λ from being reflected by a defect, the wavelength λ is chosen to be at least three times or five times greater than the maximum thickness of a defect present in the joint plane 12.In practice, faces 6 and 8 have a roughness Ra of less than 300 nm and, typically, a roughness Ra of less than 10 nm or 1 nm. Under these conditions, when face 6 is pressed against face 8 with a pressure greater than 0.3 MPa or 0.5 MPa, almost all the defects have a thickness of less than 300 nm and, usually, less than 100 nm. Under these conditions, the wavelength λ is greater than or equal to 2 pm and, preferably, greater than 3 pm. For example, here, the wavelength λ is between 2 pm and 7 pm or between 3 pm and 6 pm.

[0072]

[0066] Figure 6 shows a method of laser welding the part 2 to the part 4 using the welding station 70. This method is identical to the method of Figure 4 except that step 66 is replaced by a welding step 76. Step 76 is identical to step 66 except that the emitter 72 is used instead of the emitter 22.

[0073]

[0067] Figure 7 shows a laser welding station 170. The welding station 170 is identical to the welding station 70 except that:

[0074] - the transmitter 72 is replaced by a transmitter 172 of monochromatic laser pulses, and

[0075] - the welding station 172 also includes a sensor 176 arranged below the stack 10.

[0076]

[0068] The transmitter 172 is identical to the transmitter 72 except that the laser pulses are laser pulses centered on an adjustable wavelength λ which is not necessarily greater than 2 pm. On the contrary, the wavelength λ may be less than 2 pm and, for example, equal to one of the wavelengths λ and λ described with reference to FIG. 3. The only constraint on the wavelength λ is that it must be located in the transparency domain of the part 2. For this purpose, the transmitter 172 comprises a tunable laser source 174.

[0069] The sensor 176 measures the transmittance of the laser pulse through the stack 10. The transmittance measured by the sensor 176 is transmitted to the control unit 34.

[0077]

[0070] Figure 8 shows a method of laser welding part 2 to part 4 using welding station 170. This method is identical to the method of Figure 6 except that step 76 is replaced by steps 180, 182 and 184.

[0078]

[0071] During step 180, the unit 34 controls the actuator 42 to move the stack 10 so that the focal point 28 is located at the location of the next welding point to be made.

[0079]

[0072] Then, during step 182, the unit 34 selects a wavelength λ for which the laser pulse is not reflected even if a defect is present at the location of the weld point to be made. For this purpose, the fact that the transmittance measured by the sensor 176 is maximum when the reflection of the laser pulse by the defect is minimal is exploited. Consequently, when the measured transmittance is maximum for a wavelength λ, this means that a laser pulse centered on this wavelength λ makes it possible to provide sufficient energy to melt the faces 6 and 8 and thus obtain a correct weld point despite the presence of the defect.

[0080]

[0073] During step 182, the unit 34 controls the laser source 174 to emit low-energy laser pulses at different wavelengths between a low threshold λmin and a high threshold λ max . Typically, the gap between the thresholds Àmin and À ma x is greater than A ma x / 2. In parallel, the transmittance of the stack 10 is measured, by the sensor 176, for each of the different wavelengths of the emitted laser pulses. For example, during step, the unit 34 varies, in regular steps AÀ, the wavelength À from Àmin to À max . In this case, the number of regular steps AÀ is preferably greater than or equal to five or ten. Then, the wavelength À chosen to carry out the next welding step is that for which the transmittance measured by the sensor 176 is maximum.

[0081]

[0074] The low-energy laser pulses used during step 182 are laser pulses whose energy is sufficiently low so that the non-linear energy absorption phenomena in the parts 2 and 4 do not occur or practically do not occur. In other words, the energy of these laser pulses is very little absorbed by the stack 10 and, moreover, this low energy absorption varies practically linearly as a function of the distance traveled inside the stack 10. In particular, the low-energy laser pulses do not cause the parts 2 and 4 to melt.

[0082]

[0075] Once the wavelength λ has been selected during step 182, the welding step 184 is executed. During step 184, the unit 34 controls the transmitter 172 to transmit, at the welding point to be made in the joining plane 12, sufficient energy to melt the faces 6 and 8 at the location of this welding point to be made. For this, sufficient energy is transmitted only by the laser pulse centered on this wavelength λ. During step 184, the wavelength λ remains constant.

[0083]

[0076] At the end of step 184, the method returns to step 180 if there is a next weld point to be made. Otherwise, the method continues with step 68.

[0084]

[0077] 11.2 - Adaptation of the thickness of the defect

[0085]

[0078] Figure 9 shows a laser welding station 80. The welding station 80 is identical to the welding station 70 except that:

[0086] - transmitter 72 is replaced by transmitter 82, and

[0087] - the compression device 32 is replaced by a compression device 84.

[0088]

[0079] The emitter 82 is identical to the emitter 72 except that the laser pulses are monochromatic laser pulses centered on a wavelength λ which is not necessarily greater than 2 pm. On the contrary, the wavelength λ may be less than 2 pm and, for example, equal to one of the wavelengths λ or λ described with reference to FIG. 3. The only constraint on the wavelength λ is that it must be located in the transparency domain of the part 2. In addition, the wavelength λ is not necessarily adjustable. For this purpose, the emitter 82 comprises a non-tunable monochromatic laser source 86 capable of emitting laser pulses at the wavelength λ.

[0089]

[0080] The compression device 84 is identical to the device 32 except that the clamping mechanism 48 is replaced by a clamping mechanism 88 which makes it possible to vary, in stages, the pressure exerted by the face 6 against the face 8. Varying the pressure of the face 6 against the face 8 varies the thickness of the defects. Here, the device 84 is capable of obtaining variations in the thickness of the defects greater than λ / 4 or λ / 2. For this purpose, the mechanism 88 makes it possible to vary the pressure between the faces 6 and 8 from a low threshold P min to a high threshold P ma x by regular steps AP. The step AP is equal to (P ma x-Pmin) / N, where N is a number of steps greater than or equal to five or ten. For example, Pmin is equal to 0.1 MPa or 0.3 MPa or 0.5 MPa and P ma x is equal to or greater than 10 MPa. The pressure P max is less than the pressure at which part 2 or part 4 breaks.

[0090]

[0081] Figure 10 shows a method of laser welding the part 2 onto the part 4 using the welding station 80. This method is identical to the method of Figure 6 except that steps 64 and 76 are replaced by a step 90 of moving the stack 10, a welding step 92 and a step 94 of modifying, in stages, the pressure of face 6 against face 8.

[0091]

[0082] During step 90, the unit 34 controls the device 88 so that the pressure exerted by the face 6 against the face 8 is equal to Pmin. During step 90, the unit 34 also controls the actuator 42 to move the stack 10 so that the focal point 28 is located at the location of the next welding point to be made.

[0092]

[0083] Then, during the welding step 92, the unit 34 automatically controls the transmitter 82 to transmit, at the welding point to be made in the plane 12, sufficient energy to melt the faces 6 and 8 at the location of this welding point to be made. For this, sufficient energy is transmitted through the part 2 and using a laser pulse from the source 74 centered on the wavelength λ. During step 92, the pressure exerted by the face 6 against the face 8 remains constant. This constant pressure is called “current pressure” and noted “Pc”.

[0093]

[0084] Finally, during step 94, the unit 34 controls the clamping mechanism 88 to move to the next pressure level, i.e. to a level where the pressure exerted by the face 6 against the face 8 is equal to Pc+AP. This pressure Pc+AP then becomes the new current pressure. The method then returns to step 92. During this new iteration of step 92, the current pressure has therefore been incremented by the step AP. Steps 92 and 94 are repeated in a loop as long as the current pressure has not reached or exceeded the threshold P max . When the current pressure reaches or exceeds the threshold P max , the method returns to step 90 to perform the next weld spot along path 14.

[0094]

[0085] In the previous embodiments, an appropriate wavelength λ was selected so that the laser pulse was not reflected by a defect. In this embodiment, it is the opposite, that is to say that it is the thickness of the defect which is adapted to a predefined wavelength λ. Indeed, by varying the pressure Pc in steps, the thickness of the defect also varies in steps over a range greater than λ / 2 or λ. Under these conditions, even if a defect is present at the location of the weld point to be produced, during one of the iterations of step 92, the pressure Pc is such that the thickness of this defect is not an integer multiple of λ / 2. Thus, during this iteration of step 92, the laser pulse is practically not reflected by this defect and a correct weld point is produced.

[0095]

[0086] Figure 11 shows a laser welding station 100. The welding station 100 is identical to the welding station 80 except that it additionally comprises a sensor 102 arranged under the stack 10. The sensor 102 measures the transmittance of the laser pulses through the stack 10. The transmittance measured by the sensor 102 is transmitted to the control unit 34. For example, the sensor 102 is identical to the sensor 176 of the welding station 170.

[0096]

[0087] Figure 12 shows a process for laser welding the part 2 onto the part 4 using the welding station 100. This process is identical to the process of Figure 10 except:

[0097] - that additional steps 104 and 106 are added between steps 90 and 92, and

[0098] - that step 94 is omitted.

[0099]

[0088] During step 104, the unit 34 selects a pressure Pc to be exerted between the faces 6 and 8 so that the laser pulse is not reflected by a defect present at the location of the welding point to be made. For this purpose, as during step 182 previously described, the fact that the transmittance measured by the sensor 102 is maximum when the reflection of the laser pulse by the defect is minimum is exploited.

[0100]

[0089] During step 104, the unit 34 controls the compression device 10 to emit low-energy laser pulses on the welding point to be produced for different pressure levels between the low threshold P mm and the high threshold P ma x. In parallel, the transmittance of the stack 10 is measured, by the sensor 102, for each of the pressure levels where a low-energy laser pulse is emitted. For example, step 104 is identical to the reiteration of steps 92 and 94 of the method of FIG. 10 except:

[0101] - that during step 92 it is a low energy laser pulse which is emitted at wavelength λ, and

[0102] - that the transmittance of the stack 10 is measured for each pressure level.

[0090] Then, the pressure Pc selected to carry out the next welding step is that for which the transmittance measured by the sensor 102 is maximum.

[0103]

[0091] Once the pressure Pc has been selected, during step 106, the unit 34 controls the compression device 84 to obtain this pressure Pc, selected during step 104, between the faces 6 and 8.

[0104]

[0092] After step 106, welding step 92 is performed. During step 92, the pressure Pc remains constant.

[0105]

[0093] At the end of step 92, the method returns to step 90 if there is a next weld point to be made. Otherwise, the method continues with step 68. Thus, in this welding method, welding step 92 is only executed once for each weld point to be made.

[0106]

[0094] 11.3 - Treatment of a face

[0107]

[0095] Figure 13 shows a laser welding station 120. The welding station 120 is identical to the welding station 100 except that the compression device 84 is replaced by the compression device 32.

[0108]

[0096] Figure 14 represents a method of laser welding of the part 2 on the part 4 using the welding station 120. This method is identical to the method of Figure 6 except that it additionally comprises a step 122 of pre-treatment of the face 6 or 8 and that the welding step 76 is replaced by a welding step 124.

[0109]

[0097] Welding step 124 is identical to step 76 except that emitter 82 is used instead of emitter 72.

[0110]

[0098] In the preceding embodiments, the faces 6 and 8 do not undergo any pre-treatment, except possibly cleaning, before being welded to each other. In the method of FIG. 14, on the contrary, at least one of the faces 6 and 8 undergoes a pre-treatment which makes it possible to eliminate the capacity of the majority of the defects to reflect the laser pulse at the wavelength λ. For this, during step 122 at least one of the faces 6 and 8 is structured, not to reduce its reflectance or to increase its absorbance, but mainly to create defects of different thicknesses uniformly distributed throughout the joint plane. In this embodiment, by "uniformly distributed" is meant the fact that each defect of thickness ei is mostly surrounded and, preferably only surrounded, by defects of thicknesses different from ei.Thus, during step 124, if the defects of thickness ei reflect the laser pulses at the wavelength λ and prevent the creation of a weld point at these defects of thickness ei, the defects which surround each of the defects of thickness ei do not reflect the laser pulses at the wavelength λ. Therefore, each defect of thickness ei is surrounded by correctly created weld points. Thanks to this, despite the fact that each of the laser pulses has the same wavelength λ, the welding of the face 6 on the face 8 is correct because the weld points are uniformly distributed in the joining plane.

[0111]

[0099] Preferably, this treatment of one of the faces 6 and 8 is carried out using the laser source 86, that is to say the same laser source as that which is used during the welding step 124. In addition, this treatment is carried out under the same working conditions as those implemented during the welding step 124. Thus, in particular, this treatment does not require working in an ultra-clean environment such as inside a clean room. On the other hand, the laser source typically works in ablation mode to structure the face 6 or 8 as described below.

[0112]

[0100] Subsequently, processing step 122 is described in the particular case where only face 8 is processed and face 6 is not processed. However, the teaching given in this particular case can also be used to process face 6.

[0113]

[0101] Here, this treatment of the face 8 is carried out using the welding station 120. However, during step 122, only the part 4 is fixed on the plate 40 using the compression device 32 so that the laser pulses from the source 86 directly reach the face 8 without having to pass through the part 2.

[0114]

[0102] In this embodiment, the processing step 122 comprises the structuring of the face 8. This structuring of the face 8 consists of creating on this face 8 several hundreds or thousands of micrometric or sub-micrometric cavities which are arranged, for example at regular intervals, over the entire face 8. Each cavity is a hole dug in the face 8. In a plane parallel to the face 8, the largest dimension of each of the cavities is, for example, between 0.2 μm and 20 μm. The interval between two consecutive cavities is generally between 1 μm and 500 μm or between 30 μm and 300 μm. The interval between a first and a second consecutive cavities is equal to the distance between the longitudinal axes of these two cavities. The longitudinal axis of a cavity is the axis which passes through the center of gravity of this cavity and which is perpendicular to the plane in which the face 8 mainly extends.

[0115]

[0103] For example, during step 122, the face 8 is structured by “hatching” in the following manner. During step 122, the actuator 42 is controlled so that the focal point 28 scans the face 8 a first time. During this first scan of the face 8, the focal point 28 moves at constant speed v along a plurality of parallel lines on the face 8. These parallel lines are spaced from each other by a constant distance DU greater than the diameter of the focal point 28. Preferably, the distance DU is two to five times greater than the diameter of the focal point 28. Typically, this distance DII is between 1 pm and 500 pm. Along each of these lines, laser pulses are emitted at regular intervals. Each laser pulse thus emitted digs, to a given depth £, the face 8 at the location of this pulse. Thus cavities are formed at each location where a laser pulse is received.Here, the speed v and / or the number r of laser pulses emitted per second are adjusted so that the excavated cavities touch each other, thus forming a continuous line of cavities.

[0116]

[0104] Once the first scan is completed, several additional scans of face 8 are performed. Each of these additional scans is identical to the first scan except that the parallel lines are oriented in a different direction than the direction of the parallel lines of the other scans. The spacing between the lines and the directions of the parallel lines during the different scans are chosen:

[0117] - so that some of the laser pulses are emitted at locations where there is already a cavity excavated during a previous scan, and

[0118] - so that, conversely, certain cavities dug by previous scans do not receive a new laser pulse during this scan.

[0119]

[0105] Furthermore, here, the spacing between the lines and the directions of the parallel lines are chosen so that certain locations of the face 8, uniformly distributed on this face 8, do not receive any laser pulse. These locations are therefore not hollowed out.

[0120]

[0106] By proceeding in this way, cavities of different depths uniformly distributed on the face 8 are created. The depth of each of these cavities is, as a first approximation, equal to n,*£, where n, is an integer greater than one and the index i is an identifier of a cavity depth. Here, the values ​​of the number n, vary between 1 and N ma x. Subsequently, it is considered that a location on face 8 on which no cavity has been excavated corresponds to a zero depth. With this convention, the number of different depths is equal to N max +1. The number N maxof scans is typically greater than or equal to four or five so that there are at least five different depths. The number N max is preferably chosen to be greater than or equal to eight or ten. In addition, advantageously, the number N max is chosen so that N max *£ is greater than λ / 7 and, preferably, greater than or equal to λ / 5 or λ / 4, where λ is the wavelength of the laser pulses used during the welding step 124.

[0121]

[0107] Chapter III: Variants:

[0122]

[0108] Variants of appropriate wavelength selection:

[0123]

[0109] Alternatively, the emitter 22 comprises a single tunable monochromatic laser source. In this case, this single laser source is controlled to emit, at each weld point, first a laser pulse at wavelength λi and then a laser pulse at wavelength λ2. The wavelengths λi and λ2 are those defined with reference to FIG. 3. In this case, the laser pulses at wavelengths λi and λ2 are not emitted simultaneously but emitted one after the other. Preferably, in this case, the difference between wavelengths λi and λ2 is less than λi / 2. Indeed, in this case, the laser pulses at wavelengths λi and λ2 are correctly focused on the joining plane 12 for the same position of the laser source relative to the stack 10.This avoids having to change the position of the laser source between the situation where it emits the laser pulse at wavelength λ and the situation where it emits the laser pulse at wavelength λ.

[0124]

[0110] The transmitter 22 may also comprise a laser source capable of simultaneously emitting the monochromatic laser pulses centered on the wavelengths λi and λ2. In this case also, preferably, the difference between the wavelengths λi and λ2 is less than λi / 2.

[0125]

[0111] Alternatively, the position of this laser source relative to the stack 10 is automatically modified as a function of the wavelength of the emitted laser pulse to maintain each laser pulse correctly focused on the joining plane 12. For this purpose, for example, the unit 34 controls the actuator 42 before a laser pulse at a new wavelength is emitted.

[0112] In another embodiment, the sources 24 and 26 are each replaced by a laser source capable of emitting broad spectrum laser pulses centered, respectively, on the wavelengths λ and λ2.

[0126]

[0113] In another variant, the emitter 22 comprises a single laser source capable of emitting only one broad spectrum laser pulse. In this case, the width of the spectrum of this laser pulse extends over both a first and second distinct range of wavelengths and the energy transmitted in each of these first and second ranges is sufficient, in itself, to melt the faces 6 and 8. These first and second ranges of wavelengths have the same characteristics as those described with reference to FIG. 3 in the particular case where the first and second ranges correspond to the spectra of a first and a second monochromatic laser pulse centered on the wavelengths, respectively, λi and λ2. In particular, the median wavelengths of the first and second ranges are chosen as described in the case of the wavelengths λi and λ2.In this case, during the welding step, at each point of the trajectory, only this broad spectrum laser pulse is emitted instead of emitting two monochromatic laser pulses centered on the wavelengths, respectively, λi and λ2.

[0127]

[0114] In another variant, the emitter 22 is configured to emit sufficient energy to cause the fusion of the faces 6 and 8 at each welding point to be produced, in more than two distinct wavelength ranges. For example, the emitter 22 is capable of emitting sufficient energy in three distinct wavelength ranges. For this, the emitter 22 comprises, for example, a third monochromatic laser source which transmits sufficient energy in a third distinct wavelength range centered on a median wavelength λ3. The wavelength λ3 is chosen such that the difference | λi - λ3| is different from an integer multiple of λi / 2 and an integer multiple of λ3 / 2.

[0128]

[0115] During step 182 of selecting the wavelength λ which makes it possible to obtain a correct welding point, the laser source used to measure the transmittance of the stack 10 for different possible wavelengths is not necessarily the same as the laser source 174 used during the welding step 184. For example, as a variant, during step 182, a tunable laser source dedicated to measuring the transmittance is used. For this purpose, for example, the transmitter 172 comprises two separate tunable laser sources, one of which is used only during step 182 and the other is used only during the welding step 184.

[0129]

[0116] Other offsets between the wavelengths λi and λ2 are possible to be practically sure that a defect at the weld plane cannot reflect the laser pulses at these two wavelengths λi and λ2. For example, the difference between the wavelengths λi and λ2 may be equal to an odd multiple of λi / 4 or λ2 / 4. For example, the wavelength λ2 is equal to λi+5*λi / 4. The difference between the wavelengths λi and λ2 is also not necessarily an odd multiple of λi / 4 or λ2 / 4. For example, alternatively, the difference between the wavelengths λi and λ2 is an odd multiple of λi / 3 or λ2 / 3. If the difference | λi - λ2 | is much greater than Δ2 / 2, then it may be necessary to introduce an offset between the height of the lens of the source 24 and the height of the lens of the source 26 to succeed in simultaneously focusing the laser pulses of these two sources on the joint plane 12.

[0130]

[0117] Different solutions are possible to avoid having to introduce an offset between the position of the laser source which emits at the wavelength λi and the position of the laser source which emits at the wavelength λ2. For example, as a variant, the welding station comprises an achromatic device for focusing the laser pulses at the wavelengths λi and λ2. Such an achromatic focusing device is for example a curved mirror. In this case, the position of the focal point does not vary as a function of the wavelength.

[0131]

[0118] In a particularly interesting variant, the embodiments of figures 3 and 5 are combined. For this, the laser source 26 of the welding station 20 is replaced by the laser source 74. In this case, the wavelength λ2 is greater than or equal to 2 pm and, preferably, greater than 3 pm.

[0132]

[0119] The appropriate selection of the laser pulse length can be combined with any other embodiments. For example, during the welding step, it is possible to simultaneously implement the appropriate selection of the laser pulse wavelength and the modification of the defect thickness. For this, for example, in any of the welding stations 80 and 100, the emitter 82 is replaced by the emitter 22 or 72.

[0133]

[0120] In the welding stations 20, 70, 120 and 170, the compression device 32 can be omitted in particular if the faces 6 and 8 are sufficiently flat so that it is not necessary to compress them against each other.

[0121] Similarly, the appropriate selection of the wavelength of the laser pulses can be combined with the step 122 of treating the face 6 and / or the face 8. For this, for example, the step 122 is introduced between the steps 60 and 62 of the methods of FIGS. 4 and 6.

[0134]

[0122] Variants of modifying the thickness of a defect:

[0135]

[0123] Alternatively, during step 94, the pressure is not modified in regular steps. In this case, the different pressure levels are not uniformly distributed between P min 6t P max -

[0136]

[0124] The step AP can be negative. In this case, during step 90, the device 84 is controlled so that the initial current pressure is equal to P max then, during each iteration of step 94, the current pressure is decreased until reaching the pressure P min -

[0137]

[0125] In another variant, the pressure varies continuously between P min and P max and the laser pulses are emitted, for example, at regular intervals while, during the same time, the pressure varies continuously from Pmin to Pmax. For example, the pressure varies linearly between Pmin and Pmax. Under these conditions, each pulse is applied at a time when the pressure exerted by face 6 on face 8 is different. Thus, this variant is equivalent to the case where the pressure is modified in successive steps.

[0138]

[0126] The modification of the thickness of the defect can be combined with step 122 of treatment of face 6 and / or face 8. For this, for example, step 122 is introduced between steps 60 and 62 of the methods of figures 10 and 12.

[0139]

[0127] Variants of the treatment of a face:

[0140]

[0128] Other embodiments of step 122 of structuring the face 6 and / or 8 are possible. For example, the power of the laser pulses is modified to obtain more or less deep cavities. Similarly, during each scan, the trajectory followed by the focal point can be more complex than simple parallel lines.

[0141]

[0129] In another variant of step 122, no location of the structured face is left free of cavities. In this case, the number of different depths is equal to N max -

[0142]

[0130] Alternatively, on the face which has not been subjected to structuring as described previously, it is possible to apply a treatment which reduces its reflectance at wavelength λ or which increases its absorbance at this wavelength λ. Subsequently, these treatments are described in the particular case of face 6. For example, for this, the face is treated as described in the following article except that it is the laser source 86 which is used to form micrometric peaks on the face: C. Wu et al.: “Near-unity below-band gap absorption by microstructured silicon”, Appl. Phys. Lett. 78 (13), pp. 1850-1852, 2001.

[0143]

[0131] Other treatments of the face 6 are possible to reduce its reflectance at the wavelength λ. For example, as a variant, the structuring of the face 6 involves the creation of a network of holes to obtain a progressive lowering of the refractive index of the part 2 as one approaches its face 6. At the level of the face 6, the real part of the refractive index at the wavelength λ is close to that of air, that is to say of the medium contained in the defect. Under these conditions, the defect no longer forms a Fabry Perot cavity capable of reflecting the laser pulses at the wavelength λ. It is emphasized that a reduction in the reflectance of a surface also generally causes an increase in its absorbance. However, depending on the treatment applied, the increase in the absorbance of the face 6 may be more or less significant.

[0144]

[0132] Another possible treatment of face 6 consists of increasing the porosity of part 2 as one approaches face 6 to obtain, at the level of face 6, a refractive index at wavelength λ close to that of air.

[0145]

[0133] The absorbance of face 6 can also be increased by applying other treatments to it. For example, the absorbance of face 6 is increased by applying a treatment that modifies the state of the matter at face 6. Such a modification of the state of the matter is, for example, an amorphization or a crystallization of the surface layer forming face 6. This other treatment can also include the generation of defects in the surface layer forming face 6 to increase its absorbance. For example, these defects are isolated atomic defects called “color centers” whose density makes it possible to obtain a variation greater than 0.01 of the real part or the imaginary part of the refractive index.

[0146]

[0134] In another variant, the structuring of the face 8 is not implemented. In this case, the treatments, described above, aimed at reducing the reflectance of the face 6 or increasing its absorbance are applied, in addition to or instead of, the face 8.

[0135] The laser pulse emitter used to treat the face 6 and / or 8 is not necessarily the same as that used during the welding step 124.

[0147]

[0136] In another variant, the face 6 and / or 8 is treated by a treatment method not using laser pulses. For example, the face 6 or 8 is structured during step 122 by implementing a photolithography method.

[0148]

[0137] Other variants:

[0149]

[0138] Whether during the processing step 122 or during the welding step, it is not necessarily the part 4 which is moved to move the focal point 28 relative to this part. Thus, as a variant, the part 4 is stationary and it is the focal point 28 which is moved along the face 8. For this, a scanner can be used. A scanner is a device which makes it possible to deflect the laser beam to place the focal point at a desired location. Such a scanner generally comprises a mirror, of the galvanometric mirror type, which moves the beam before the focusing optics. It is also possible either to move only the focal point 28 by moving, for example, the laser pulse focusing optics or by moving the laser pulse emitter. In this case, the actuator 42 can be omitted.

[0150]

[0139] Many variants of the trajectory 14 are possible. In particular, the trajectory 14 does not necessarily extend continuously from its starting point 16 to its arrival point 17. For example, the trajectory 14 may be formed solely by lines parallel to each other or solely by concentric circles.

[0151]

[0140] The joining plane 12 is not necessarily flat but can adopt any desirable conformation. For example, as a variant, the plane 12 is curved. In this case, the trajectory 14 is not flat but curved. In order for the focal point 28 to be able to follow such a curved trajectory 14 in space, during the welding step, the unit 34 also controls the actuator 42 to move, in addition, the stack 10 in the Z direction.

[0152]

[0141] During step 60, the adjustment of the position of the transmitter to focus the laser pulses on the point 28 can also be carried out as described in application WO2021255035A2.

[0153]

[0142] Before step 62 of joining face 6 to face 8, the method may comprise a step of polishing faces 6 and 8 so as to make them as smooth as possible. For example, at the end of this polishing step, preferably, the roughness Ra of each of faces 6 and 8 is less than 20 nm.

[0154]

[0143] Parts 2 and 4 may be made of other semiconductor materials such as, for example, Ge or GaAs. In addition, the laser welding methods described here also apply to the case where parts 2 and 4 are not made of the same semiconductor material.

[0155]

[0144] Parts 2 and 4 may contain components other than MEMS or fluidic microsystems. For example, as a variant, one or both parts 2 and 4 comprise, in addition to or instead of, optical and / or electronic components made, in part or in full, of silicon. Some of these components may be passive, such as silicon waveguides or electrical connections. Others of these components may be active, i.e. require a power supply to operate. For example, an active component may be a laser source, a modulator, a transistor or others. Typically, the stack of parts 2 and 4 forms part of a three-dimensional integrated circuit in which the different components that constitute it are distributed in different layers stacked on top of each other.

[0156]

[0145] Several of the variants described above can be combined in the same embodiment.

[0157]

[0146] Chapter IV: Advantages of the described embodiments:

[0158]

[0147] It has been observed that a defect reflects a laser pulse when it forms a Fabry-Perrot cavity whose thickness is close to an integer multiple of λ / 2, where λ is the median wavelength of this laser pulse. All the embodiments previously described make it possible to avoid the presence of such a Fabry-Perrot cavity. Therefore, when welding parts 2 and 4, the number of correctly produced welding points is much greater than if no precautions were taken to avoid the presence of these Fabry-Perrot cavities which reflect the laser pulse. As a result, the adhesion strength obtained after welding between parts 2 and 4 is much greater. In addition, this improvement in the welding process is obtained without it being necessary to add another material, such as a metal film, at the joint plane. It is also not necessary to have prior knowledge of the thickness of the defects.Finally, the laser welding processes described here Tl can be implemented under less restrictive working conditions. In particular, the laser welding processes described here can be implemented in ambient air and outside a clean room. The welding processes described here therefore remain simple to implement.

[0159]

[0148] More specifically, by transmitting the energy necessary to melt the faces 6 and 8, at a weld point to be made, both in a first and a second wavelength range and under conditions where the gap | Ài - À2| is different from an integer multiple of Ài / 2 and an integer multiple of À2 / 2, no defect can form a Fabry-Perrot cavity capable of effectively reflecting the laser pulses at both wavelengths Ài and À2. This ensures that, despite the presence of a defect, the weld point is correctly made. It is emphasized that, in this case, the increase in adhesion strength is obtained without the need to carry out prior treatments, such as polishing, to minimize the number of defects present in the joint plane.

[0160]

[0149] The fact that the gap | Ài - À2| is close to an odd integer multiple of Ài / 4 or of À2 / 4, maximizes the efficiency of the energy transfer to the welding point to be made in the case where there is a defect at this welding point to be made which is capable of reflecting the laser pulse either at the wavelength Ài or at the wavelength À2.

[0161]

[0150] The fact that the difference between the wavelengths λi and λ2 is less than λi / 2 simplifies the implementation of the welding process. Indeed, in this case, the positions of the laser sources 24 and 26 for which the laser pulses at the wavelengths λi and λ2 are correctly focused in the joint plane are the same or practically the same. It is therefore not necessary to determine the position of the laser source 24 a first time and then determine the position of the laser source 26 a second time. This is also particularly advantageous when it is the same laser source that emits the laser pulses at both the wavelength λi and the wavelength λ2.

[0162]

[0151] The fact that the wavelength of the laser pulse is much greater than the thickness of the defects makes it possible to obtain correct welding points even in the presence of defects and without using laser pulses at different wavelengths for this.

[0163]

[0152] Varying the thickness of the defects makes it possible to obtain a configuration where the energy transfer to the welding point to be made is carried out correctly even in the presence of a defect. Furthermore, in this case, it is not necessary to carry out an appropriate selection of the wavelength of the laser pulse.

[0164]

[0153] Measuring the transmittance of the laser pulse through the stack 10 using low energy laser pulses makes it possible to select the appropriate thickness of the defect at which the welding step is to be performed while consuming less energy than if the same laser pulse as that used in the welding step were used for this.

[0165]

[0154] Using the same laser pulse emitter during the processing step 122 and during the welding step simplifies the welding process.

[0155] The fact that the faces 6 and 8 to be welded each have a roughness Ra of less than 300 nm makes it possible to guarantee that the thickness of practically all the defects is less than 300 nm and therefore much less than the wavelength λ, greater than 2 pm, of the laser pulses used to weld these faces 6 and 8 together. Under these conditions, the number of correct welding points is greater, which increases the robustness of the weld obtained.

[0166]

[0156] The fact that the pressure of face 6 on face 8 is greater than 0.3 MPa makes it possible to further reduce the thickness of the defects present between these faces and therefore to further increase the number of correct welding points obtained when the wavelength of the laser pulses is greater than 2 pm.

Claims

Claims 1. Method for laser welding a first semiconductor part onto a second semiconductor part, comprising the following steps: 1) attaching (62) a first face of the first semiconductor part directly to a second face of the second semiconductor part to form a stack comprising the first and second semiconductor parts and a joining plane formed by the attachment of the first and second faces, then 2) at each welding point to be made in the joining plane, transmitting (66; 76; 184; 92; 124), only through the first part and using at least one laser pulse focused on this welding point to be made, the energy sufficient to melt the first and second faces at the location of this welding point to be made, the duration of each laser pulse being between 1 ps and 100 ps, ​​characterized in that, in the presence of a defect which forms a cavity at a welding point to be made, the reflection of the laser pulse by this defect is minimized: a) by emitting the laser pulse in a wavelength range in which this reflection is minimized by the implementation of at least one solution chosen from the group consisting of the following solutions: - transmitting (66) sufficient energy to melt the first and second faces at the welding point to be produced in a first and a second distinct wavelength range, the difference between the median wavelengths of these first and second ranges being different from an integer multiple of λi / 2 and an integer multiple of λ2 / 2, where λi and λ2 are the median wavelengths, respectively, of the first and second ranges, - the transmission (76) of sufficient energy to melt the first and second faces at the welding point to be produced in a wavelength range whose median wavelength is greater than 2 pm, and - measuring (182) the transmittance of the stack at the welding point to be produced in different wavelength ranges, then selecting (182), as a function of the transmittances measured for different wavelength ranges, a wavelength range which minimizes the reflection of the laser pulse at the welding point weld to be made, and, finally, the transmission (184) of sufficient energy to melt the first and second faces at this weld point to be made in the selected wavelength range, or b) by modifying the thickness of the defect to obtain a thickness which minimizes the reflection of the laser pulse by implementing a solution chosen from the group consisting of the following solutions: - the modification (94) of the thickness of the defect to successively obtain several different thicknesses and for each different thickness, the transmission (92) of sufficient energy to melt the first and second faces at the welding point to be produced in the same wavelength range, and - measuring (104) the transmittance of the stack at the weld point to be made in the same wavelength range for different thicknesses of the defect, then selecting (104), as a function of the transmittances measured for different thicknesses of the defect, a thickness which minimizes the reflection of the laser pulse at the weld point to be made in this wavelength range, and, finally, transmitting (92) sufficient energy to melt the first and second faces at this weld point to be made in this wavelength range and when the thickness of the defect is equal to the selected thickness, or c) by treating (122) the first or the second face to dig cavities of different depths uniformly distributed on this face, the number of different depths being greater than or equal to five and these different depths ranging from a minimum depth pmm to a maximum depth p max, where the width of the interval [p m in ; Pmax] is greater than λ / 5 and λ is the wavelength of the laser pulses used in step 2).

2. Method according to claim 1, in which, during step 2), at each welding point to be made in the joining plane: - a first laser source emits (66), only in the first wavelength range, laser pulses which transmit sufficient energy to melt the first and second faces, and - a second laser source, distinct from the first laser source, emits (66) simultaneously, only in the second wavelength range, laser pulses that transmit sufficient energy to melt the first and second faces.

3. Method according to any one of the preceding claims, in which the difference | Ài - À21 does not satisfy any of the following conditions: - Condition 1): there exists an integer m such that Ài / 10, and - Condition 2): there exists an integer p such that At 2 / 10.

4. Method according to claim 3, in which the difference | Ài - À2| is between n*Ài / 4 - Ài / 16 and n*Ài / 4 + Ài / 16 or between n*À2 / 4 - À2 / 16 and n*À2 / 4 - À2 / 16, where n is a positive and odd integer.

5. Method according to any one of the preceding claims, in which the difference between the wavelengths λi and λ2 is less than λi / 2.

6. Method according to any one of the preceding claims, in which: - the modification (94) of the thickness of the defect involves the modification, in successive stages, of the pressure exerted by the first face on the second face, then - the method comprises, for each pressure level reached and before moving on to the next pressure level, the transmission (92) of sufficient energy to melt the first and second faces at the welding point to be produced in the same wavelength range.

7. Method according to any one of the preceding claims, wherein the same laser pulse emitter is used in the processing step (122) and in the step 2) (124).

8. Method according to any one of the preceding claims, in which, during step 1), the roughness Ra of the first face and the second face which are directly attached to each other to form the stack, is less than 300 nm.

9. Method according to claim 8, wherein, during step 2), the pressure exerted by the first face against the second face is greater than 0.3 MPa.

10. Method according to any one of the preceding claims, in which obtaining the treated face comprises the application of a treatment chosen from the group consisting of: - a treatment which makes the face to be treated amorphous, - a treatment which crystallizes the face to be treated, and - a treatment which creates defects in the surface layer forming the face to be treated so as to increase the absorbency of the face to be treated.

11. Laser welding station for welding a first semiconductor part onto a second semiconductor part, this welding station comprising: - a laser pulse transmitter (22) configured to, at each welding point to be made in a joining plane between the first and second parts, transmit, through the first part and using at least one laser pulse focused on this welding point to be made, sufficient energy to melt first and second faces, respectively, of the first and second parts at the location of this welding point to be made, and - a unit (34) for controlling the laser pulse transmitter, characterized in that: - the laser pulse emitter (22) is capable, at each welding point to be produced in the joining plane, of transmitting sufficient energy to melt the first and second faces at each welding point to be produced in a first and a second distinct wavelength range, the difference between the median wavelengths of these first and second ranges being different from an integer multiple of λi / 2 and an integer multiple of λ2 / 2, where λi and λ2 are the median wavelengths, respectively, of the first and second ranges, and - the control unit (34) is configured to control the laser pulse transmitter so as to systematically transmit, at each welding point, sufficient energy to melt the first and second faces at this welding point to be produced in the first and second wavelength ranges.

12. Laser welding station for welding a first semiconductor part onto a second semiconductor part, this welding station comprising: - a laser pulse emitter (172) configured to, at each welding point to be made in a joining plane between the first and second parts, transmit, through the first part and using at least one laser pulse focused on this welding point to be made, sufficient energy to melt first and second faces, respectively, of the first and second parts at the location of this welding point to be made, and - a unit (34) for controlling the laser pulse transmitter, characterized in that: - the welding station also comprises a sensor (176) capable of measuring the transmittance of the stack, formed by the adhesion of a first face of the first semiconductor part directly to a second face of the second semiconductor part, at the level of the welding point to be produced in different wavelength ranges, and - the control unit (34) is configured to: - acquire the transmittances measured at the welding point, then - select, based on the transmittances measured for different wavelength ranges, a wavelength range which minimizes the reflection of the laser pulses at the welding point to be made, and, finally, - controlling the laser pulse transmitter to transmit sufficient energy to melt the first and second faces at this weld point to be made in the selected wavelength range.

13. Laser welding station for welding a first semiconductor part onto a second semiconductor part, this welding station comprising: - a laser pulse transmitter (82) configured to, at each welding point to be made in a joining plane between the first and second parts, transmit, through the first part and using at least one laser pulse focused on this welding point to be made, sufficient energy to melt first and second faces, respectively, of the first and second parts at the location of this welding point to be made, and - a unit (34) for controlling the laser pulse transmitter, characterized in that: - the welding station comprises a controllable device (84) capable of modifying the thickness of a defect which forms a cavity at the level of a welding point to be produced, and - the control unit (34) is configured to control the device capable of modifying the thickness of the defect to modify the thickness of the defect to successively obtain several different thicknesses and for each different thickness, control the laser pulse emitter, to transmit sufficient energy to melt the first and second faces at the welding point to be produced always in the same wavelength range.

14. Laser welding station for welding a first semiconductor part onto a second semiconductor part, this welding station comprising: - a laser pulse transmitter (82) configured to, at each welding point to be made in a joining plane between the first and second parts, transmit, through the first part and using at least one laser pulse focused on this welding point to be made, sufficient energy to melt first and second faces, respectively, of the first and second parts at the location of this welding point to be made, and - a unit (34) for controlling the laser pulse transmitter, characterized in that: - the welding station includes: - a controllable device (84) capable of modifying the thickness of a defect which forms a cavity at the level of a welding point to be produced, - a sensor (102) capable of measuring the transmittance of the stack formed by the joining of the first and second faces, at the level of the welding point to be produced, and - the control unit (34) is configured to: - control the device capable of modifying the thickness of the defect to obtain different thicknesses of this defect, - acquire the transmittances measured at the welding point for these different thicknesses and in the same wavelength range, then - select, based on the transmittances measured for these different thicknesses, a thickness which minimizes the reflection of the laser pulse at the welding point to be made in this wavelength range, then - control the device capable of modifying the thickness of the defect to obtain the selected thickness of the defect, then - controlling the laser pulse transmitter to transmit sufficient energy to melt the first and second faces at this weld point to be made in this wavelength range and when the thickness of the defect is equal to the selected thickness.

Images