Welding assembly for power semiconductor module packages and method for manufacturing the same

Abstract

The present invention relates to a welded assembly for a power semiconductor module package, comprising a first member whose outermost layer on the first side is a non-copper metal layer, a second member having a first copper layer on the first side, and connecting copper layers that connect to the non-copper metal layer and the first copper layer on both sides, respectively, to form a first interface and a second interface, thereby integrally fixing the first member and the second member, wherein the average crystal grain size of the connecting copper layer is 35 nm to 400 nm. The present invention further provides a welded assembly for a semiconductor chip package relating to multiple members, and a method for manufacturing a welded assembly for a power semiconductor module package.

Description

[Technical Field]

[0001] The present invention relates to a welded assembly for a power semiconductor module package and a method for manufacturing the same, the welded assembly being particularly applicable to power semiconductor modules, and especially to power semiconductor modules used in inverter modules for new energy vehicles. [Background technology]

[0002] New energy vehicles (NEVs) can achieve advantages such as zero emissions by using clean energy, and are expected to gradually replace gasoline-powered vehicles in the future, becoming the most market-leading vehicle type. One of the core components of a NEV is the electrical control system, and among these, the inverter module plays the role of converting high-voltage, high-current DC and AC power to high-frequency power. By acting as a power relay between the onboard battery and the drive system, the performance of the inverter module directly affects the performance of the NEV's powertrain system.

[0003] The performance and safety of inverter modules in new energy vehicles are strongly influenced by the packaging. The core of the packaging lies in achieving conductivity and thermal conductivity between key components such as a conductive copper thin film, a semiconductor chip (covered on the back with one or more metal layers), a copper-clad substrate, and a heat sink, through welding. This mainly concerns three cases: 1) welding the outermost metal layer on the back of the semiconductor chip (generally a gold layer, silver layer, etc.) to the copper layer of the copper-clad substrate with a different low-melting-point metal solder; 2) connecting the copper layer of the copper-clad substrate to the surface aluminum layer of an aluminum heat sink with a different low-melting-point solder or a thermally conductive silicone rubber sheet; and 3) welding the outermost metal layer on the front of the semiconductor chip (generally an aluminum layer) to the surface copper layer of the conductive copper thin film with a different low-melting-point metal solder. The most widely used solder industrially is low-melting-point tin-based solder. Tin-based solder has a melting point between 200 and 250°C and possesses good welding performance (good wettability, low surface tension, etc.) and usability. A common welding method involves heating the solder to a temperature slightly above its melting point, for example, 250°C, for 1 to 10 minutes in a vacuum or inert gas protected environment. [Overview of the project] [Problems that the invention aims to solve]

[0004] However, welded assemblies obtained using tin-based solder have several drawbacks. On the one hand, tin is highly reactive at low temperatures and reacts with any metal such as gold, silver, aluminum, titanium, and nickel, resulting in the formation of intermetallic compounds at the weld interface between them. These intermetallic compounds grow rapidly at the interface at the inverter module's operating temperature, forming a large, micron-order layer of intermetallic compounds (e.g., Ag3Sn, AuSn4, etc., specifically due to the metals that react with tin). This reduces the electrical conductivity of the weld interface and makes it prone to brittle cracking. This not only reduces the power output of the inverter module but also critically impacts its lifespan, increasing safety risks during vehicle use. On the other hand, when welding a chip to a copper-clad substrate with tin-based solder, the resulting welded assembly contains large voids. These voids significantly reduce the electrical and thermal conductivity of the welded assembly, decreasing the module's power output and heat dissipation efficiency, preventing the module from achieving the desired performance. Furthermore, these voids become stress concentration points when subjected to mechanical impact or in thermal cycling environments, making them prone to microcrack formation and reducing shear strength. This leads to premature fracture and shortens the module's service life. At least these problems inherent in conventional welded assemblies limit the power output of conventional inverter modules and prevent reliable operation at high temperatures.

[0005] Therefore, there is a strong need for welded assemblies that can be serviced stably at high temperatures over long periods, thereby improving the long-term serviceability of devices containing such welded assemblies, by at least mitigating or eliminating the above-mentioned problems caused by welding dissimilar metals, namely the degradation of performance (e.g., shear strength, conductivity) due to voids and brittle intermetallic compounds. [Means for solving the problem]

[0006] This invention was made in view of the above-mentioned conventional problems.

[0007] One of the objectives of the present invention is The outermost layer on the first side is a first member of a non-copper metal layer, A second member having a first copper layer on the first side, The material includes connecting copper layers that integrally fix the first member and the second member by connecting to the non-copper metal layer and the first copper layer on both sides, respectively, to form a first interface and a second interface, The objective is to provide a welded assembly for a power semiconductor module package, wherein the average crystal grain size of the connecting copper layer is 35 nm to 400 nm.

[0008] The average crystal grain size of the connecting copper layer is preferably 35 nm to 380 nm, more preferably 45 nm to 380 nm, even more preferably 60 nm to 220 nm, and still more preferably 45 nm to 140 nm. Preferably, the average thickness of the connecting copper layer is 70 nm to 5 μm, more preferably 90 nm to 5 μm, and even more preferably 500 nm to 5 μm, and the average thickness of the connecting copper layer is at least 1 times its average crystal grain size.

[0009] Preferably, the width is set to a position 1 μm away from the connecting copper layer from the first and second interfaces, respectively, and the length is set to any 10 μm along the longitudinal direction of the first and second interfaces, within which the average porosity of the welded assembly is ≤3% and the pore size is ≤200 nm. More preferably, the average porosity of the welded assembly is ≤1%.

[0010] Preferably, in the welded assembly, 1) the first member is a semiconductor chip and the second member is a copper-clad substrate, or 2) the first member is a semiconductor chip and the second member is a conductive copper thin film or copper foil, or 3) the first member is a heat sink and the second member is a copper-clad substrate.

[0011] Preferably, the non-copper metal layer is a gold layer, and the average crystal grain size of the connecting copper layer is 35 nm to 140 nm.

[0012] Preferably, the non-copper metal layer is a silver layer, and the average grain size of the connecting copper layer is 100 nm to 400 nm. This is because, at the same temperature, the diffusion coefficient of silver in copper is closer to the diffusion coefficient of copper in silver, so the requirement for the average grain size needed to weaken the Kirkendall effect is lower than that for a gold layer.

[0013] Preferably, the first copper layer consists of the initial copper layer of the copper-clad substrate or conductive copper thin film, and its average crystal grain size is ≥ 500 nm.

[0014] Furthermore, preferably, the welding assembly is A second copper layer provided on the second side of the second member, The outermost layer on the first side is a non-copper metal layer, which is a third member. The present invention further includes another connecting copper layer that connects to the non-copper metal layer and the second copper layer of the third member on both sides, thereby integrally fixing the second member and the third member together. The average crystal grain size of the aforementioned other connecting copper layer is 35 nm to 400 nm.

[0015] It should be understood that the aforementioned welded assembly may also include further members and corresponding connecting copper layers having an average grain size of 35 nm to 400 nm for welding them.

[0016] Preferably, the shear strength of the welded assembly is 30-40 MPa, more preferably 35-40 MPa, and the resistivity of the connecting copper layer is ≤ 5 μΩ·cm.

[0017] Preferably, the chip according to the present invention includes a functional chip whose main material is silicon, silicon carbide, or gallium nitride, and which has one or more metal layers on its front and back surfaces, with the outermost metal layer being a non-copper metal layer. Preferably, the outermost metal layer may be a gold layer, an aluminum layer, a silver layer, a nickel layer, or a titanium layer. Preferably, the chip may be an insulated gate bipolar transistor (IGBT) chip, a metal oxide semiconductor field-effect transistor (MOSFET) chip, a bipolar transistor (BJT) chip, or a fast recovery diode (FRD) chip.

[0018] Preferably, the copper-clad substrate according to the present invention may be any substrate whose surface is a copper layer, and includes, but is not limited to, a direct bond copper ceramic substrate (DBC), an active metal brazed ceramic substrate (AMB), or a copper-clad metal heat dissipation structure.

[0019] Preferably, the heat sink according to the present invention has a heat dissipation structure in which the surface is an aluminum layer.

[0020] Another objective of the present invention is, A semiconductor chip having a non-copper metal layer on the outermost layer of the first and second sides, A copper-clad substrate having a first copper layer and a second copper layer on the first side and the second side, respectively, A heat sink with an aluminum layer on the first side, A conductive copper thin film having a third copper layer provided on the first side of the copper-clad substrate, A first connecting copper layer is provided on both sides, connecting to the non-copper metal layer and the third copper layer on the first side of the semiconductor chip, respectively, to integrally fix the semiconductor chip and the conductive copper thin film. A second connecting copper layer is provided on both sides, connecting to the non-copper metal layer on the second side of the semiconductor chip and the first copper layer, thereby integrally fixing the semiconductor chip and the copper-clad substrate. It includes a third connecting copper layer on each side, which connects to the second copper layer and the aluminum layer respectively, and integrally fixes the copper-clad substrate and the heat sink. The objective is to provide a welded assembly for a power semiconductor module package, wherein the average crystal grain size of the first, second, and third connecting copper layers is 35 nm to 400 nm.

[0021] Preferably, the average grain size of the first and second connecting copper layers is 35 nm to 380 nm, more preferably 45 nm to 380 nm, even more preferably 60 nm to 220 nm, and even more preferably 45 nm to 140 nm. The average grain size of the third connecting copper layer is preferably 35 nm to 380 nm, and more preferably 62 nm to 250 nm. Preferably, the non-copper metal layer of the semiconductor chip is a gold layer, an aluminum layer, a silver layer, a nickel layer, or a titanium layer. Preferably, the average porosity of the welded assembly is ≤3%, more preferably ≤1%, and the pore size is ≤200 nm.

[0022] A further object of the present invention is to provide a method for manufacturing a welded assembly for a power semiconductor module package, comprising the following steps. a) Prepare the first and second members to be welded, clean and dry them, and ensure that the first side of the second member has an initial copper layer, and the outermost layer on the first side of the first member is a non-copper metal layer, such as a gold layer, aluminum layer, silver layer, nickel layer, or titanium layer. b) A copper pre-plating layer is introduced to the surface of the initial copper layer of the cleaned second member, or to the outermost layer on the first side of the cleaned first member, and the copper pre-plating layer is cleaned to remove the oxide layer on its surface, wherein the average crystal grain size of the copper pre-plating layer is 20 nm to 360 nm. c) The copper pre-plating layer of the second member or the first member is brought into close contact with the non-copper metal layer of the first member or the initial copper layer of the second member to form a laminated structure. d) The welded assembly of the present invention is obtained by welding the first member and the second member by diffusion welding.

[0023] Preferably, the average crystal grain size of the copper pre-plated layer is 25 nm to 360 nm, more preferably 30 nm to 300 nm, and the average thickness of the copper pre-plated layer is at least twice the average crystal grain size. Even more preferably, the average thickness of the copper pre-plated layer is 100 nm to 5 μm, and even more preferably 500 nm to 5 μm, which is at least three times the average crystal grain size.

[0024] Preferably, the first component is a semiconductor chip, and its non-copper metal layer is a gold layer, aluminum layer, silver layer, nickel layer, or titanium layer. Preferably, the first component is a heat sink, its surface is an aluminum layer, and the average crystal grain size of the copper pre-plated layer is 30 nm to 200 nm, and the average thickness of the copper pre-plated layer is 100 nm to 3 μm, which is at least three times the average crystal grain size.

[0025] Preferably, in step d, the diffusion welding is carried out as follows: In a helium gas protected environment, pressure is applied from the top and bottom surfaces of the laminated structure at a temperature of 250 to 350°C, with a pressure of 15 to 40 MPa and a holding time of 10 to 30 minutes. More preferably, the holding temperature is 250 to 300°C, or the pressure is 20 to 30 MPa, or the holding time is 10 to 20 minutes.

[0026] Preferably, the copper pre-plated layer is formed by methods such as electroplating, vapor deposition, sputtering, or ion plating, and a copper pre-plated layer with relatively uniform thickness and grain size is provided.

[0027] Preferably, in step b, the copper pre-plating layer is realized by electrochemical deposition. The electrolyte solution used contains 200-250 g / L of anhydrous copper sulfate, 10-50 ml / L of sulfuric acid, and 0.01-0.05 g / L of thiourea, with a solution temperature of 30°C, and more preferably, the current density for electrochemical deposition is 0.05-0.15 A / cm². 2 The deposition time is 60 to 300 seconds. Preferably, the electrolyte solution contains 200 to 250 g / L of anhydrous copper sulfate, 10 to 50 ml / L of sulfuric acid, and 0.005 to 0.01 g / L of thiourea, the solution temperature is 30°C, and more preferably, the current density for electrochemical deposition is 0.01 to 0.04 A / cm². 2 The deposition time is 250-350 seconds. [Effects of the Invention]

[0028] The present invention provides a welded assembly with improved performance through simple operation. Specifically, the present invention involves applying a dense copper pre-plating layer for welding two members and controlling its crystal grain size to obtain a connecting copper layer with a desired crystal grain size, thereby resulting in a welded assembly with low porosity and resistivity. Moreover, because copper has low reactivity and small lattice parameters, the reaction between copper and dissimilar metals is extremely slow, resulting in very few intermetallic compounds in the welded assembly. Consequently, the welded assembly obtained by the present invention has low porosity and resistivity, almost no intermetallic compounds, and possesses excellent electrical conductivity, thermal conductivity, and high shear strength.

[0029] Furthermore, during service, the high grain boundary volume fraction of the connecting copper layer in the welded assembly provides rapid diffusion channels for dissimilar metals, resulting in a diffusion rate of dissimilar metal atoms being between 1 / 6 and 6 times, preferably 0.5 to 2 times, the diffusion rate of copper atoms. This nearly equalizes the amount of atoms entering and leaving the connecting copper layer, fundamentally reducing or avoiding the formation of Kirkendall voids. Moreover, this welded assembly has extremely low levels of intermetallic compounds. As a result of these two factors, even when operating in high-temperature environments for extended periods, the welded assembly does not experience significant formation and growth of Kirkendall voids or intermetallic compounds, thus improving the long-term service stability of the welded assembly without any apparent impact on its mechanical performance.

[0030] Compared to conventional technologies, the welding assembly for connecting dissimilar metals in the power semiconductor module package according to the present invention offers superior performance and stable service performance, solving the problems of low mechanical strength, high resistivity, and poor heat dissipation caused by the formation of Kirkendall voids and intermetallic compounds during servicing of dissimilar metal welding assemblies.

[0031] Those skilled in the art will understand that any range or value within the above intervals is applicable to the present invention. For example, the average crystal grain size of the connecting copper layer being 35 nm to 400 nm means that it can take any range or any specific numerical value within that range. For example, the average grain size is within one of the following ranges: 35nm~320nm, 40nm~300nm, 50nm~300nm, 50nm~280nm, 50nm~250nm, 50nm~200nm, 42nm~200nm, 45nm~150nm, 40nm~150nm, 35nm~100nm, 50nm~150nm, or 50nm~100nm. Alternatively, the average grain size is one of the following values: 55nm, 68nm, 75nm, 85nm, 90nm, 110nm, 140nm, 170nm, 220nm, 260nm, 275nm, 300nm, 320nm, 340nm, or 368nm. [Brief explanation of the drawing]

[0032] [Figure 1] A diagram illustrating an exemplary scenario involving a welded assembly according to the present invention. [Figure 2] This figure shows a structured illumination microscope (SIM) image of a cross-section of a copper-clad ceramic substrate after the copper pre-plating layer has been deposited according to the first embodiment of the present invention. [Figure 3] This figure shows an atomic force microscope (AFM) image of the surface morphology of the copper pre-plated layer on a copper-clad ceramic substrate after the copper pre-plated layer has been deposited according to the first embodiment of the present invention. [Figure 4] This figure shows a scanning electron microscope (SEM) image of a cross-section of a welded assembly according to the first embodiment of the present invention. [Figure 5] Figure 4 shows a partially magnified SEM image. [Figure 6] This figure shows a transmission electron microscope (TEM) image of a cross-section of a welded assembly according to the first embodiment of the present invention. [Figure 7] Figure showing an SEM image of the shear surface of a welded assembly according to the first embodiment of the present invention. [Figure 8] This figure shows an SEM image of a cross-section of a welded assembly according to the first embodiment of the present invention after being heated at 250°C for 165 hours. [Figure 9] This figure shows a partially magnified TEM image of the cross-section of a welded assembly according to the first embodiment of the present invention after being kept warm at 250°C for 165 hours. [Figure 10] This figure shows an SEM image of the shear surface of a welded assembly according to the first embodiment of the present invention after being kept at 250°C for 165 hours. [Figure 11] Figure showing an SEM image of a cross-section of a welded assembly according to a second embodiment of the present invention. [Figure 12] Figure showing a TEM image of a cross-section of a welded assembly according to a second embodiment of the present invention. [Figure 13] This figure shows an SEM image of a cross-section of a welded assembly according to a second embodiment of the present invention after being kept at 250°C for 165 hours. [Figure 14]This figure shows a partially magnified TEM image of the cross-section of a welded assembly according to a second embodiment of the present invention after being kept at 250°C for 165 hours. [Figure 15] Figure showing an SEM image of a cross-section of a welded assembly according to a third embodiment of the present invention. [Figure 16] Figure showing a TEM image of a cross-section of a welded assembly according to a third embodiment of the present invention. [Figure 17] This figure shows an SEM image of a cross-section of a welded assembly according to a third embodiment of the present invention after being kept at 250°C for 165 hours. [Figure 18] This figure shows a partially magnified TEM image of the cross-section of a welded assembly according to a third embodiment of the present invention after being kept at 250°C for 165 hours. [Figure 19] This figure shows a welded assembly of a silicon-based IGBT chip and a copper-clad ceramic substrate, welded with tin-based solder paste SAC305 in the first comparative example. [Figure 20] Figure showing an SEM image of the cross-section of the welded assembly in the first comparative example after being incubated at 200°C for 72 hours. [Figure 21] Figure showing an SEM image of the shear surface of the welded assembly in the first comparative example after being incubated at 200°C for 72 hours. [Figure 22] This figure shows a SIM image of the cross-section of a copper-clad ceramic substrate after the copper pre-plating layer in the second comparative example has been annealed. [Figure 23] This figure shows an AFM image of the surface morphology of the copper pre-plated layer of the copper-clad ceramic substrate after annealing in the second comparative example. [Figure 24] Figure showing an SEM image of the cross-section of the welded assembly in the second comparative example. [Figure 25] This figure shows an SEM image of the cross-section of the welded assembly in the second comparative example after being kept at a high temperature of 250°C for 165 hours. [Modes for carrying out the invention]

[0033] The present invention will be described below with reference to specific examples. Several embodiments are described in the embodiments and drawings for carrying out the invention, and these embodiments are intended to illustrate what is to be disclosed. One or more exemplary embodiments are shown below, but it should be understood that this disclosure is not limited to the exemplary embodiments, drawings, or technology shown below, and can be modified within the scope of all appended claims and equivalents. It should be understood that any numbering in the disclosed features (e.g., 1st, 2nd, etc.) and directional terms used with the disclosed features (e.g., front, back, top, bottom, etc.) are relative terms indicating exemplary relationships between related features or used to distinguish related features. The term “example” is intended to represent an example, illustration, or descriptive term when used herein. Unless expressly stated, all ranges include extreme values. Those skilled in the art will understand that the data and various parameters described in the examples are illustrative and do not limit the invention.

[0034] The inventors have discovered that tin-based welded assemblies have a defect in low strength. The main reasons for this are as follows: On the one hand, tin metal is highly reactive and reacts with multiple components of the surface to be welded during the packaging process to form brittle intermetallic compounds, such as AuSn4 when reacting with Au, Ag3Sn when reacting with Ag, Cu3Sn and Cu6Sn5 when reacting with Cu, and Ni3Sn when reacting with Ni. These intermetallic compounds generally require low energy to form and grow, so they can be formed and grown rapidly even in low-temperature environments, forming a weak interface that is weak and prone to fracture when subjected to force. As a result, the device is prone to premature failure during service. On the other hand, tin atoms and dissimilar metal atoms diffuse with each other at the weld interface. When the diffusion rates of two types of atoms differ significantly, atoms diffuse rapidly near the weld interface on the side with the faster diffusion rate. However, the vacancies left behind by the diffused atoms cannot be filled in a timely manner, resulting in the formation of a large number of Kirkendall voids. These voids grow as the service time increases, forming stress concentration zones. When subjected to force, the welded assembly becomes extremely susceptible to cracking along these voids, creating an additional mechanically weak layer, and the strength of the welded assembly gradually decreases during service. Moreover, a large number of voids reduce the electrical and thermal conductivity of the welded assembly, which is detrimental to improving the operating power of the device.

[0035] Similarly, the Kirkendall effect also exists in the interdiffusion between the coarse-grained copper layer and the non-copper metal layer. In this invention, copper with a grain size greater than 400 nm is defined as coarse-grained copper, while copper with grains of 400 nm or less is defined as fine-grained copper. For example, at 250°C, the diffusion rate D of copper atoms in the gold layer Cu / Au is 2.2 × 10 -18 m 2 The diffusion rate of gold atoms in the coarse-grained copper layer is D / s. Au / Cu is 2.4 × 10 -21 m 2 / s, where the former is approximately 1000 times that of the latter. Therefore, for a welded assembly composed of a coarse copper layer and a non-copper metal layer, during the welding process for forming the assembly and subsequent use at high temperatures, the diffusion of dissimilar metal atoms (e.g., gold atoms) is slow, and thus the voids left by the rapidly diffused copper atoms in the coarse copper layer cannot be compensated. As a result, a large amount of Kirkendall voids are formed in the coarse copper layer near the welding interface between the non-copper metal layer and the coarse copper layer, which not only becomes a weak point in a device including the welded assembly, but also reduces the electrical conductivity and thermal conductivity of the welded assembly, and adversely affects the use performance of the device.

[0036] In order to avoid or at least reduce the influence of Kirkendall voids on the performance of the welded assembly, the inventors of the present invention propose that by increasing the diffusion rate of dissimilar metal atoms in the copper layer in contact therewith to be close to the diffusion rate of copper atoms, the diffusion rates of different atoms entering and exiting the copper layer can be substantially balanced, thereby avoiding or reducing the formation of voids. The inventors of the present invention have recognized that the diffusion rate of dissimilar metals can be adjusted by controlling the grain boundary volume fraction. This is because the atomic arrangement at grain boundaries is loose, and compared with the energy barrier that atoms need to overcome for moving inside the orderly arranged crystal grains, the energy barrier that atoms need to overcome to break the arrangement at grain boundaries is lower, resulting in a higher diffusion rate, which allows grain boundaries to provide a fast channel for atomic diffusion. Therefore, the present invention proposes to use the characteristic of fast atomic diffusion at grain boundaries to solve the above problem.

[0037] Specifically, for a copper layer with small crystal grain size, since the diffusion length of the dissimilar metal is larger than the crystal grain size, the Harrison A-type diffusion model is satisfied, that is, grain boundary diffusion and lattice diffusion occur simultaneously. In this case, the effective diffusion rate D considering two diffusion modes eff can be calculated by the following formula (1).

[0038] [Numerical formula]

[0039] Here, f gb This is the grain boundary volume fraction, D gb This is the grain boundary diffusion coefficient, D l This is the lattice diffusion coefficient. D gb , D l Factors such as the grain boundary volume fraction f can be considered fixed values ​​at the same temperature for the same system (e.g., Cu-Au system), so gb By adjusting this, the diffusion rate in the copper layer in contact with the dissimilar metal (specifically, the surface metal layer for welding the first member) in the system (for example, the connecting copper layer in the welding assembly) can be controlled to be at least roughly balanced with the diffusion rate of copper atoms. In this application, roughly balanced diffusion rates mean that the diffusion rate of the dissimilar metal atoms is between 1 / 6 and 6 times the diffusion rate of copper atoms, and more preferably between 0.5 and 2 times. f gb This is influenced by the size of the crystal grains and can be calculated using the following equation (2).

[0040]

number

[0041] Here, Δ is the grain boundary thickness, which is set to 0.5 nm, and d is the crystal grain size.

[0042] According to this formula, the grain boundary volume fraction can be controlled by controlling the size of the crystal grains. Third-generation SiC semiconductor chip high-power devices will have an operating temperature of 200-300°C in the future, and the higher the operating temperature, the more susceptible they are to the Kirkendall effect. Therefore, in this invention, the temperature control range is set to 200-300°C. Within this temperature range, by controlling the size of the crystal grains of the connecting copper layer in contact with the dissimilar metal to 400 nm or less (i.e., the copper layer is a fine-grained copper layer), a grain boundary volume fraction of at least 0.375% can be obtained, and this grain boundary volume fraction increases as the size of the crystal grains decreases. For example, when the crystal grain size is 35 nm, the grain boundary volume fraction reaches approximately 4.3%, and the diffusion coefficient of the dissimilar metal in the copper layer and the diffusion coefficient of copper in the dissimilar metal are 1 / 6D Cu / X ≤D X / Cu-eff ≤6D Cu / X The conditions will be met, and eventually, D Cu / X D is the diffusion coefficient of Cu atoms in metal X, and D X / Cu-eff This is the effective diffusion coefficient of metal X atoms in the connecting copper layer. By reducing the difference between the diffusion coefficients of dissimilar metal atoms and the diffusion coefficient of copper atoms, the formation and growth of Kirkendall voids during the service life can be reduced, and even suppressed.

[0043] If the average crystal grain size of the copper layer is >400 nm, the grain boundary volume fraction of the copper layer will be <0.375%, and sufficient grain boundaries that act as high-speed channels to promote rapid diffusion of dissimilar metals will not be secured. In other words, the difference in diffusion coefficients between dissimilar metal atoms and copper atoms will still be large, and therefore the problem caused by Kirkendall voids cannot be clearly mitigated. Furthermore, while reducing the crystal grain size (increasing the grain boundary volume fraction) contributes to further improving the diffusion rate of dissimilar metal atoms in the copper layer, if the crystal grains are too fine, the diffusion rate of dissimilar metal atoms may excessively exceed the diffusion coefficient of copper atoms, and there is still a risk of causing the Kirkendall effect. For this reason, the present invention requires that the average crystal grain size of the connecting copper layer be 35 nm or larger, that is, the average crystal grain size of the connecting copper layer between the two welded members be 35 nm to 400 nm.

[0044] Below, using the Cu-X system in which the dissimilar metals X are Au, Ag, Al, Ti, and Ni respectively, we will use the above formula (1-2) to determine the D of the preferred crystal grain size. X / Cu-eff Calculate D gb , D l These are D X / Cu-gb , D X / Cu-l The values ​​are as follows. Table 1 lists the preferred crystal grain size of the connecting copper layer in these Cu-X systems and the diffusion coefficient of dissimilar metal X atoms in the connecting copper layer calculated therefrom, which are 0.5D. Cu / X ≤D X / Cu-eff ≤2D Cu / X Satisfy D Cu / X D is the diffusion coefficient of Cu atoms in metal X, and X / Cu This is the diffusion coefficient of metal X atoms in the connecting copper layer, and D X / Cu-gb This is the grain boundary diffusion coefficient of metal X atoms in the connected copper layer, and D X / Cu-l This is the lattice diffusion coefficient of metal X atoms in the connected copper layer, and D X / Cu-eff This is the effective diffusion coefficient of metal X atoms in the connecting copper layer, 1* D X / Cu-eff This is the effective diffusion coefficient of metal X atoms in a connected copper layer with an average grain size of 35 nm. 2* D X / Cu-eff This is the effective diffusion coefficient of metal X atoms in a connected copper layer with an average grain size of 400 nm. 1* D X / Cu-eff and 2* D X / Cu-eff It is 1 / 6D Cu / X ≤D X / Cu-eff ≤6D Cu / X It satisfies the condition.

[0045] Table 1 shows the preferred crystal grain size and diffusion coefficient at 250°C for the connecting copper layer used in the Cu-X system.

[0046] [Table 1]

[0047] Taking Cu / Au as an example, the diffusion rate of Cu in Au at 250°C is D.Cu / Au 2.2 × 10 -18 m 2 s -1 Therefore, the diffusion rate of Au in Cu is D. Au / Cu 2.4 × 10 -21 m 2 s -1 This is significantly lower than the former. When the average crystal grain size of the connecting copper layer is controlled to 35 nm, the grain boundary volume fraction is f gb = 3 × 0.5 × 10 -9 / (35×10 -9 ) = 0.0429. Substituting this into equation (1), we get the effective diffusion rate D of Au in Cu. Au / Cu-eff =(0.0429 × 1.0 × 10 -16 ) + 5.0 × 10 -24 = 4.3 × 10 -18 m 2 s -1 D Cu / Au This becomes approximately twice as much. Correspondingly, if the average crystal grain size of the connecting copper layer is 140 nm, then according to equation (1), D at this time is Au / Cu-eff is 1.1 × 10 -18 m 2 s -1 D Cu / Au This becomes 0.5 times. Furthermore, if the average crystal grain size of the connecting copper layer is 400 nm, then according to equation (1), D at this time Au / Cu-eff is 3.8 × 10 -19 m 2 s -1 D Cu / Au This is approximately 0.173 times. Therefore, when the average crystal grain size of the connecting copper layer is 35 nm to 400 nm at 250°C, the effective diffusion rate of Au in Cu is D Au / Cu-eff It is 1 / 6D Cu / Au ≤D Au / Cu-eff ≤6D Cu / Au By satisfying this condition, the diffusion rates of Cu and Au become roughly balanced, and the Kirkendall effect can be significantly reduced. Furthermore, when the average crystal grain size of the connecting copper layer is 35 nm to 140 nm, the effective diffusion rate of Au in Cu D Au / Cu-eff is 0.5D Cu / Au ≤D Au / Cu-eff ≤2D Cu / AuBy satisfying these conditions and further reducing the difference in diffusion rates between Cu and Au, the Kirkendall effect can be further weakened, resulting in a welded assembly with superior performance. Similarly, for other dissimilar metals X, the difference in diffusion rates between dissimilar metal X and Cu can be effectively reduced by setting the average grain size of the connecting copper layer to 35 nm to 400 nm. Furthermore, the difference in diffusion rates between dissimilar metal X and Cu can be further reduced by using preferred grain sizes corresponding to various metals. Although only exemplary data at 250°C is shown, the preferred average grain size of the connecting copper layer is applicable to the expected temperature range of 200 to 300°C.

[0048] Furthermore, in order to obtain a desired grain boundary volume fraction, i.e., a desired crystal grain size, of the joining copper layer (i.e., the fine-grained copper layer) in the assembly, the manufacturing method of the present invention proposes controlling the crystal grain size of the copper pre-plating layer for welding so that the crystal grain size of the connecting copper layer is 35 nm to 400 nm. In this case, both during the formation period of the welded assembly and during the service period of the obtained welded assembly, the small crystal grain size of both the copper pre-plating layer and the connecting copper layer is made to roughly approximate the diffusion rate of copper in the dissimilar metals. By improving the diffusion rate in the copper pre-plating layer and the connecting copper layer of the dissimilar metals being welded, the formation and growth of Kirkendall voids during welding and during the service period is avoided or at least reduced, thereby avoiding or at least mitigating the performance degradation of the welded assembly due to the Kirkendall effect, and improving the performance of the welded assembly. Moreover, due to the stable properties of copper itself, the formation of brittle intermetallic compounds during the welding process is also avoided, thus avoiding the influence of intermetallic compounds on the performance of the obtained welded assembly.

[0049] In this invention, porosity is defined as the ratio of the sampling area to the total area of ​​pores. The method for measuring the porosity of a welded assembly involves sampling the cross-section of the obtained welded assembly within a range where the width is between positions 1 μm away from the connecting copper layer from the first interface and the second interface, respectively (i.e., the connecting copper layer is included), and the length is any 10 μm along the longitudinal direction of the first / second interface. The total area of ​​all visible pores within the sampling range at each position is measured using the image processing software ImageJ. Based on the measurement results, the porosity at each sampling position is calculated, and the average value is taken. As schematically shown in Figure 4, the field of view has a width between positions 1 μm away from the connecting copper layer from the interface between the gold layer and the connecting copper layer and the interface between the connecting copper layer and the coarse-grained copper layer, respectively, and a length of any 10 μm along the longitudinal direction of the interface.

[0050] In this invention, testing the resistivity of the connecting copper layer involves simulating a welding scenario. A copper pre-plating layer of the same thickness as the actual welding is deposited on the surface of the first copper layer of a silicon wafer having a 1 μm first copper layer. A non-copper metal layer of the same thickness as the outermost non-copper metal layer of the component to be welded is sputtered directly onto the copper pre-plating layer, and the welding assembly obtained after welding is simulated by heating and pressurizing. Subsequently, the thin-film resistivity is measured on the surface of the welding assembly using the four-probe method, with the measurement depth including the non-copper metal layer, the connecting copper layer, and a small amount of the first copper layer. This resistivity can reflect the influence of Kirkendall voids on the welding assembly.

[0051] In this invention, the shear fracture test is performed by the following method: The bottom of the second member of the welded assembly is fixed to the test platform with resin, and a movable mechanical member is pressed against the side of the first member of the welded assembly, applying a shear force parallel to the sample at a moving speed of 20 μm / min. Five samples are tested for each type of product, and the average value is taken.

[0052] The superior performance of the welding assembly of the present invention will be explained below using specific examples. It should be understood that the following examples are merely specific examples of the welding assembly of the present invention and are not intended to limit the invention in any way.

[0053] Figure 1 schematically illustrates a composite welding assembly according to the present invention, illustrating various application scenarios to which the welding assembly of the present invention can be applied, for example, a welding assembly consisting of a tip and a conductive copper thin film. It should be understood that the multiple scenarios shown may exist individually or in combination and are described illustratively in each embodiment.

[0054] First embodiment A method for manufacturing a welded assembly according to the first embodiment of the present invention is as follows: Step 1.1 Prepare the copper-clad ceramic substrate and silicon-based IGBT chip to be welded. The surface of the copper-clad ceramic substrate is covered with an initial copper layer approximately 2.4 μm thick, consisting of coarse grains with an average crystal grain size of 500 nm or larger (see Figure 2). The crystal grain size is measured based on the maximum dimensions of each crystal grain shown in the figure. The metal layers on the back of the silicon-based IGBT chip are, from inside to outside, a titanium layer, a nickel layer, and a gold layer. Step 1.2 To remove surface impurities and oxide layers, the surface of the initial copper layer of the copper-clad ceramic substrate is sequentially cleaned using an ultrasonic cleaning method with acetone, anhydrous ethanol, and dilute hydrochloric acid (volume fraction HCl:H2O=1:1), and then washed with deionized water and dried. Step 1.3 The surface of the gold layer of the silicon-based IGBT tip is sequentially cleaned using acetone, anhydrous ethanol and isopropanol (IPA) by ultrasonic cleaning, and then washed with deionized water and dried. Step 1.4 A copper pre-plating layer is deposited on the surface of the initial copper layer in a 10 mm × 10 mm area by electrochemical deposition. The parameters for electrochemical deposition are as follows: DC plating with a current density of 0.05~0.15 A / cm². 2The electrolyte solution contains 200-250 g / L of anhydrous copper sulfate, 10-50 ml / L of sulfuric acid, and 0.01-0.05 g / L of thiourea, with a deposition time of 60 seconds and a solution temperature of 30°C. Figure 2 shows a SIM image of a cross-section of a copper-clad ceramic substrate after the deposition of the copper pre-plating layer, where the average thickness of the deposited copper pre-plating layer is approximately 100 nm. In actual production, the thickness of the connecting copper layer can be appropriately adjusted as needed. To protect the deposited copper pre-plating layer when creating the cross-section, a Pt protective layer is additionally deposited on the copper pre-plating layer before cutting. Figure 3 shows an AFM image of the surface morphology of the copper pre-plating layer on a copper-clad ceramic substrate after the deposition of the copper pre-plating layer, where the average grain size of the copper pre-plating layer is approximately 30 nm, which is significantly smaller than the grain size of the underlying initial copper layer. Preferably, the copper pre-plating layer may also be deposited on the gold layer of a silicon-based IGBT chip. Preferably, in order to avoid the influence of specific grain orientations in the single layer of crystal grains on the atomic diffusion rate, the thickness of the copper pre-plating layer is at least twice, more preferably three times, the average grain size. Step 1.5 To remove the surface oxide layer, the surface of the copper pre-plated layer of the copper-clad ceramic substrate is cleaned with dilute hydrochloric acid (volume fraction HCl:H2O=1:1), Step 1.6 The gold layer of the silicon-based IGBT chip and the copper pre-plated layer of the copper-clad ceramic substrate are brought into close contact to form a laminated structure. Step 1.7 In a helium gas protected environment, pressurize the laminated structure from the top and bottom surfaces at a temperature of 280°C, with a pressure of 20 MPa and an incubation time of 10 minutes. Step 1.8 The bonded silicon-based IGBT chip and copper-clad ceramic substrate are air-cooled to room temperature to obtain a welded assembly.

[0055] To better illustrate the details of the assembly, the obtained welded assembly is cut. Figure 4 shows an SEM image of the cross-section of a welded assembly obtained by the method described above according to a first embodiment of the present invention, and Figure 5 is a magnified view of a portion of Figure 4. As shown in Figures 4 and 5, the obtained welded assembly includes a silicon-based IGBT chip, a copper-clad ceramic substrate, and a connecting copper layer interposed between them to fix them together. The connecting copper layer is a fine-grained copper layer, which is bonded on both sides to the outermost gold layer of the chip and the coarse-grained copper layer (i.e., the first copper layer) of the copper-clad substrate, respectively, forming a first interface and a second interface. In Figure 4, the circled areas indicate the locations of pores, and all pores are 200 nm or less (the maximum measurable dimension of each pore shown). Figure 5 also supports this point, where the white dashed line is the interface between the gold layer and the fine-grained copper layer detected by software recognition, and the black dashed line is the interface between the coarse-grained copper layer and the fine-grained copper layer detected by software recognition. As shown in Figure 5, several voids exist on the fine-grained copper layer side near the interface between the gold layer and the fine-grained copper layer. The voids are measured below to calculate the porosity.

[0056] Figure 6 shows a TEM image of a cross-section of a welded assembly according to the first embodiment of the present invention, where the average grain size of the fine-grained copper layer is approximately 70 nm and the thickness is approximately 100 nm. Compared to the copper pre-plating layer before forming the welded assembly, the grain size of the connecting copper layer is larger, but the thickness remains essentially unchanged. This is because, during the joining process, the crystal grains of the copper pre-plating layer were incorporated into each other and grew. This incorporation occurs as grain boundaries gradually move, that is, the grain boundary of one crystal grain progresses to other surrounding crystal grains and incorporates them. Therefore, the connecting copper layer of the resulting welded assembly can inherit the thickness of the copper pre-plating layer if the crystal grains grow. Because nanocrystals are highly active, they grow very easily through such incorporation at welding temperatures. However, when the crystal grain size exceeds 100 nm, the activity of the nanocrystals decreases exponentially, and growth slows down (see Examples 2 and 3 described later).

[0057] According to equation (1), here, Dgb , D l respectively represent D at the current temperature Au / Cu-gb and D Au / Cu-l . At 200°C, D Au / Cu-gb =7.3×10 -19 m 2 s -1 , and D Au / Cu-l =1.0×10 -24 m 2 s -1 . When the crystal grain size is 70 nm, the effective diffusion rate D of gold atoms in the fine-grained copper layer at 200°C Au / Cu-eff reaches 1.6×10 -20 m 2 s -1 , which is closer to the diffusion rate of copper atoms in the gold layer (7.3×10 Au / Cu-eff ′=3.2×10 -23 m 2 s -1 with a crystal grain size exceeding 500 nm) than the effective diffusion rate D of coarse-grained copper -21 m 2 s -1 ). At 300°C, D Au / Cu-gb =4.8×10 -15 m 2 s -1 , and D Au / Cu-l =9.0×10 -24 m 2 s -1 . The calculated D Au / Cu-eff reaches 1.0×10 -16 m 2 s -1 , which is closer to the diffusion rate of copper atoms in the gold layer (6.0×10 Au / Cu-eff ′=9.4×10 -20 m 2 s -1 than D of coarse-grained copper -17 m 2 s -1 ). Therefore, the welded assembly having the above connecting copper layer can avoid or at least reduce the influence of the Kirkendall effect on its performance during the subsequent service period of 200 to 300°C. This is supported by the results of the following high-temperature durability test.

[0058] To calculate the porosity, samples are taken at five locations on the cross-section of the obtained welded assembly (one location is schematically shown in Figure 4) within the following field of view: The width is set between the first interface between the gold layer and the connecting copper layer and the second interface between the connecting copper layer and the coarse-grained copper layer, both 1 μm away from the connecting copper layer (i.e., the connecting copper layer is already included); and the length is set to any 10 μm along the length direction of the first / second interface. In this example, the sampling range is approximately 22 μm. 2 The total porosity at each location was measured, and the specific measurement results are shown in Table 2. According to Table 2, the average porosity in this example is approximately 0.79 ± 0.09%. In addition, the maximum measured porosity at each sampling location is 200 nm or less.

[0059] Furthermore, the resistivity of the connecting copper layer is measured. Specifically, a welding scenario is simulated in which a copper pre-plating layer with a thickness of 100 nm and a crystal grain size of 30 nm is deposited on the surface of a silicon wafer having a first copper layer of 1 μm thickness, a 100 nm gold layer is sputtered directly onto the copper pre-plating layer, and the welding assembly obtained after welding is simulated by heating and pressurizing as described above. The resistivity of the thin film is measured using the four-probe method and it is found that the average resistivity of the connecting copper layer is 3.0 ± 1.4 μΩ·cm. This low resistivity corresponds to a low porosity and is advantageous for realizing high-power operation of the welding assembly.

[0060] Porosity according to the first embodiment of the present invention (sampling area is approximately 22 μm²) 2 The results are shown in Table 2.

[0061] [Table 2]

[0062] A shear fracture test was performed on the welded assembly of this embodiment. The shear strengths of the five samples were 35.8 MPa, 36.6 MPa, 38.2 MPa, 34.7 MPa, and 36.2 MPa, respectively, with an average shear strength of 36.3 ± 1.2 MPa. Figure 7 shows an SEM image of the shear surface of the welded assembly of this embodiment, which exhibits ductile fracture.

[0063] Furthermore, in order to verify the service stability of the welded assembly of the present invention, a high-temperature durability test is performed on the welded assembly of this embodiment. The test parameters are as follows. The sample is kept at 250°C for 165 hours in a vacuum environment, and after the keeping is complete, the sample is air-cooled. Then, as described above, porosity measurement, resistivity measurement, and shear fracture test are performed on the sample that has undergone the high-temperature durability test.

[0064] Figures 8 and 9 show images of the cross-section of a welded assembly according to the first embodiment of the present invention after a high-temperature durability test. No obvious Kirkendall voids or intermetallic compound formation were observed. After the high-temperature durability test, the average porosity of the welded assembly remained almost unchanged at 0.82 ± 0.15%, and the average resistivity slightly improved to 3.5 ± 1.2 μΩ·cm. For the shear fracture test, the shear strengths of the five samples after the high-temperature durability test were 32.7 MPa, 35.2 MPa, 33.8 MPa, 34.5 MPa, and 31.3 MPa, respectively. The average shear strength was 33.5 ± 1.3 MPa, which is only about 7.7% lower than the shear strength before the high-temperature durability test, demonstrating excellent shear resistance. As shown in Figure 10, the shear surface of the welded assembly after the high-temperature durability test still exhibits ductile fracture, and no oxide concentration is observed.

[0065] As can be seen from the above, the welded assembly obtained by the first embodiment of the present invention inherently has a porosity of 1% or less, a resistivity of approximately 3.0 ± 1.4 μΩ·cm for the connecting copper layer, and a shear strength of 36.3 ± 1.2 MPa. Moreover, even after high-temperature durability testing, the welded assembly still has low porosity and resistivity and high shear strength, ensuring long-term reliability in high-power operation of the welded assembly. This is due to the fact that the welded assembly has very few Kirkendall voids, or if present, they are very small in size. As a result, this embodiment can avoid or at least reduce the impact of Kirkendall voids on the performance of the welded assembly.

[0066] Second example The method for manufacturing a welded assembly according to the second embodiment of the present invention is the same as the method of the first embodiment, except for the following differences. 1) The components to be welded are a copper-clad ceramic substrate and an aluminum heat sink, and the surface layer of the aluminum heat sink is an aluminum layer. 2) The current density in electrochemical deposition is 0.01 to 0.04 A / cm², the electrolyte solution contains 200 to 250 g / L of anhydrous copper sulfate, 10 to 50 ml / L of sulfuric acid, and 0.005 to 0.01 g / L of thiourea, the deposition time is 250 seconds, and the solution temperature is 30°C. The average thickness of the deposited copper pre-plated layer is approximately 700 nm, and the average grain size is approximately 200 nm (the measurement method is the same as in the first example, and details are omitted). This average thickness can avoid the generation of single-layer grains during welding and during use of the resulting welded assembly, thereby reducing the influence of specific grain orientations on atomic diffusion rates. Similarly, the copper pre-plated layer may be deposited on the aluminum layer of an aluminum heat sink. 3) In the welding step, pressurization is applied from the top and bottom surfaces of the laminated structure at a temperature of 280°C in a helium gas protected environment, with a pressure of 20 MPa and a holding time of 15 minutes.

[0067] Figure 11 shows an SEM image of a cross-section of a welded assembly of the second embodiment obtained by the method described above, which includes an aluminum heat sink, a copper-clad ceramic substrate, and a connecting copper layer that integrally fixes them together. The connecting copper layer is a fine-grained copper layer, which is bonded to the surface aluminum layer of the aluminum heat sink and the coarse-grained copper layer of the copper-clad substrate on both sides, respectively, forming a first interface and a second interface, correspondingly shown by the white dashed line and black dashed line in Figure 11.

[0068] Similar to the first embodiment, corresponding measurements and tests were performed on the welded assembly of the second embodiment. Figure 12 shows a TEM image of the cross-section of the welded assembly according to the second embodiment of the present invention, where the black dashed lines indicate some crystal grains in the connecting copper layer. Due to heating during welding, the average crystal grain size of the connecting copper layer increases to approximately 220 nm, but the thickness is maintained at approximately 700 nm. The average porosity of the welded assembly of the second embodiment is 0.76 ± 0.05%, the average resistivity of the connecting copper layer is 3.8 ± 0.8 μΩ·cm, and the shear strength is 35.8 ± 1.1 MPa. Figures 13 and 14 show SEM and TEM images of the cross-section of the welded assembly according to the second embodiment of the present invention after a high-temperature durability test, respectively, where no obvious Kirkendall voids were observed, nor was the formation of intermetallic compounds. After the high-temperature durability test, the average porosity was 0.83 ± 0.10%, the average resistivity of the connecting copper layer was 4.3 ± 1.0 μΩ·cm, and the shear strength was 34.0 ± 2.2 MPa. The shear surface of the welded assembly in the second embodiment exhibited ductile fracture characteristics both before and after the high-temperature durability test.

[0069] Similarly, the welded assembly of the second embodiment also has low porosity and resistivity, and high shear strength, ensuring long-term reliability in high-power operation of the welded assembly.

[0070] Third embodiment A method for manufacturing a welded assembly according to a third embodiment of the present invention is the same as the method of the first embodiment, except for the following differences. 1) The components to be welded are a conductive copper thin film and a silicon carbide MOSFET chip. Of these, the conductive copper thin film is a 20 μm thick coarse-grained copper thin film consisting of coarse grains with an average crystal grain size of 500 nm or larger, and the metal layers on the front surface of the silicon carbide-based MOSFET chip are, from inside to outside, a titanium layer, a nickel layer, and a silver layer. 2) A copper pre-plating layer is deposited on a conductive copper thin film, where the current density during electrochemical deposition is 0.005~0.015 A / cm². 2 The deposition time was 350 seconds, and the electrolyte solution contained 200-250 g / L of anhydrous copper sulfate, 10-50 ml / L of sulfuric acid, and 0.005-0.01 g / L of thiourea. The average thickness of the deposited copper pre-plating layer was approximately 1 μm, and the average grain size was approximately 360 nm. Similarly, the copper pre-plating layer may be deposited on the gold layer of a silicon-based IGBT chip. 3) In the welding step, pressurization is applied from the top and bottom surfaces of the laminated structure at a temperature of 280°C in a helium gas protected environment, with a pressure of 15 MPa and a holding time of 20 minutes.

[0071] Figure 15 shows an SEM image of a cross-section of a welded assembly of the third embodiment obtained by the method described above, which includes a silicon carbide MOSFET chip, a conductive copper thin film, and a connecting copper layer that integrally fixes them together. The connecting copper layer is a fine-grained copper layer, which is bonded on both sides to the surface silver layer of the silicon carbide MOSFET chip and the coarse-grained copper layer of the conductive copper thin film, respectively, forming a first interface and a second interface, correspondingly shown by the white dashed line and black dashed line in Figure 15.

[0072] Similar to the first embodiment, corresponding measurements and tests were performed on the welded assembly of the third embodiment. Figure 16 shows a TEM image of the cross-section of the welded assembly according to the third embodiment of the present invention, where the average grain size of the connecting copper layer is approximately 380 nm. The average porosity of the welded assembly of the third embodiment is 0.82 ± 0.11%, the average resistivity of the connecting copper layer is 3.1 ± 1.0 μΩ·cm, and the shear strength is 33.4 ± 1.3 MPa. Figures 17 and 18 both show SEM images of the cross-section of the welded assembly according to the third embodiment of the present invention after the high-temperature durability test, where no obvious Kirkendall voids were observed, nor was the formation of intermetallic compounds. After the high-temperature durability test, the average porosity of the welded assembly was 1.39 ± 0.09%, the average resistivity of the connecting copper layer was 6.2 ± 1.1 μΩ·cm, and the shear strength was 32.8 ± 1.2 MPa. The shear surface of the welded assembly in the third embodiment exhibits ductile fracture characteristics both before and after the high-temperature durability test.

[0073] Similarly, the welded assembly of the third embodiment also has low porosity and resistivity, and high shear strength, ensuring long-term reliability in high-power operation of the welded assembly.

[0074] As can be seen from these examples, the welded assemblies of the present invention inherently possess excellent performance (high electrical and thermal conductivity, and high shear strength), and furthermore, maintain excellent performance even after prolonged use in high-temperature environments, ensuring stability and reliability in the long-term servicing of devices including the welded assemblies. The reasons for this are as follows: Firstly, the welded assemblies of the present invention inherently have very few defects such as Kirkendall voids, so the impact of these defects on subsequent servicing performance is small. Secondly, the fine crystalline grains of the connecting copper layer in the welded assembly improve the diffusion rate of dissimilar metal atoms in the connecting copper layer and weaken the Kirkendall effect due to the difference in diffusion rates, thereby suppressing or at least mitigating the formation and growth of Kirkendall voids, and thereby improving the stability of the welded assembly during servicing. Thirdly, due to the stable properties of copper itself, the formation of brittle intermetallic compounds is avoided during welding and the subsequent servicing period, and the influence of intermetallic compounds on the performance of the resulting welded assembly is avoided.

[0075] First comparative example A silicon-based IGBT chip and a copper-clad ceramic substrate are welded using SAC305, a tin-based solder paste commonly used in the industrial sector. The specific steps are as follows: The silicon-based IGBT chip and the copper-clad ceramic substrate are cleaned to remove impurities from their surfaces and then dried. The metal layers on the back of the silicon-based IGBT chip consist of a titanium layer, a nickel layer, and a gold layer from the inside out, while the copper-clad ceramic substrate has an initial copper layer. A tin-based solder paste is printed onto the surface of the initial copper layer of the copper-clad ceramic substrate using tape to create a 10mm x 10mm area with a thickness of approximately 30μm. Next, the gold layer of the chip and the solder paste on the copper-clad ceramic substrate are brought into surface contact and stacked to form a chip-tin-based solder paste-copper-clad ceramic substrate layered structure. After that, the above layered structure is held in a vacuum environment at a temperature of 250°C for 2 minutes, and finally air-cooled to room temperature.

[0076] As shown in Figure 19, the welded assembly obtained by the above method includes a silicon-based IGBT chip, a copper-clad ceramic substrate, and a joint between the two. The joint consists of, from top to bottom, layers of intermetallic compound 1, tin solder, and intermetallic compound 2. The layers of intermetallic compound 1 and intermetallic compound 2 in the joint are bonded to the outermost gold layer of the chip and the coarse-grained copper layer of the copper-clad substrate, respectively, forming a first interface and a second interface, as shown by the white dashed lines in Figure 19. These white dashed lines are the interfaces between each layer detected by software recognition. This edge detection allows for the calculation of the area of ​​the formed intermetallic compound in the cross-sectional view, and the average thickness of the intermetallic compound is calculated by the ratio of this area to the straight-line length of the interface. Five locations are sampled in the cross-section of the same sample. The average thickness of the intermetallic compound 1 layer in the calculated joint was 7.2 μm, and the average thickness of the intermetallic compound 2 layer was 5.3 μm. The porosity sampling range was set to a length of 10 μm and a width from a position 1 μm above the first interface between the gold layer and the intermetallic compound 1 layer on the chip surface to a position 1 μm below the second interface between the intermetallic compound 2 layer and the coarse-grained copper layer. This sampling range included at least the intermetallic compound 1 layer, the tin solder layer, and the intermetallic compound 2 layer, all of which are regions where Kirkendall pores accumulate. The welded assembly in this comparative example had an average porosity of 4.4 ± 1.5% and a resistivity of 140 ± 2 μΩ·cm.

[0077] Porosity according to the first comparative example of the present invention (sampling area approximately 300 μm²) 2 ) are shown in Table 3.

[0078] [Table 3]

[0079] Similarly, shear fracture tests were performed on the welded assemblies of this comparative example. The shear strengths of the five samples were 22.1 MPa, 24.5 MPa, 23.8 MPa, 22.9 MPa, and 26.2 MPa, respectively, with an average shear strength of 23.9 ± 0.6 MPa, which is clearly lower than the shear strengths of each example of this application (approximately 33-36 MPa).

[0080] Furthermore, the welding unit of the comparative example was also subjected to high-temperature durability testing. The test parameters were as follows: The sample was kept at 200°C for 72 hours in a vacuum environment, and after the heating period was complete, the sample was air-cooled.

[0081] After the high-temperature durability test, cross-sections were sampled from five different locations on the same sample. As shown in Figure 20, a large number of Kirkendall voids were generated at the interface (porosity of 10.8 ± 2.1%), which adversely affects the shear performance of the welded assembly. The measured results showed that the average thickness of the intermetallic compound 1 layer was 7.8 μm, and the average thickness of the intermetallic compound 2 layer was 9.5 μm, an increase of approximately 79.2% compared to before the high-temperature durability test. This is because the energy required to form the Cu / Sn metal compound is lower than that of the Au / Sn compound, and the thickness of the intermetallic compound 2 layer, which is closer to the coarse-grained copper layer, grew faster. The increase in the thickness of the intermetallic compound means that when a welded assembly obtained using tin-based solder paste is used at high temperatures, the welded assembly will clearly change, and this is accompanied by the generation and growth of Kirkendall voids, which adversely affects the usability and durability of the welded assembly.

[0082] The results of the shear fracture test support this adverse effect. The shear strengths of the five samples after the high-temperature durability test were 18.0 MPa, 15.6 MPa, 23.8 MPa, 22.3 MPa, and 20.6 MPa, respectively, with an average shear strength of 20.3 ± 1.3 MPa. This represents a decrease of approximately 15% compared to the shear strength before the high-temperature durability test, which is far greater than the decrease in shear strength after the high-temperature durability test of the welded assemblies in each embodiment of the present invention. Furthermore, as shown in Figure 21, the welded assembly of this comparative example exhibits a rock-candy-like fracture surface in the shear plane after the high-temperature durability test, which is a clear characteristic of brittle fracture.

[0083] The welded assembly in this comparative example contains not only an intermetallic compound layer but also a porosity of 5% or more, which unfavorably reduces the shear strength of the resulting welded assembly. Furthermore, these intermetallic compounds and pores further form and grow during the service life of the welded assembly, unfavorably affecting the electrical conductivity, thermal conductivity, service stability, and durability of the welded assembly.

[0084] Second comparative example The method for manufacturing the welded assembly of the second comparative example is the same as the method of the first embodiment, with the following differences. 1) The copper-clad ceramic substrate has an initial copper layer with an average crystal grain size of 500 nm or more and a thickness of approximately 1 μm. 2) After depositing a copper pre-plating layer by electrochemical deposition, it is annealed to grow crystal grains. The parameters for electrochemical deposition are as follows: DC plating with a current density of 0.005-0.015 A / cm². 2 The electrolyte solution consisted of 200-250 g / L of anhydrous copper sulfate and 10-50 ml / L of sulfuric acid, with a deposition time of 400 seconds and a solution temperature of 30°C. Figure 22 shows a SIM image of a cross-section of a copper-clad ceramic substrate after the copper pre-plating layer has been annealed, where the average thickness of the copper pre-plating layer is approximately 1 μm. The annealed surface of the copper-clad ceramic substrate is shown in Figure 23, where the average grain size of the copper pre-plating layer is approximately 500 nm. 3) In the welding step, pressurization is applied from the top and bottom surfaces of the laminated structure at a temperature of 400°C in a helium gas protective environment, with a pressure of 20 MPa and a holding time of 30 minutes.

[0085] Figure 24 shows an SEM image of a cross-section of a welded assembly of the second comparative example obtained by the method described above. The obtained welded assembly includes a silicon-based IGBT chip, a copper-clad ceramic substrate, and a coarse-grained copper layer that integrally fixes them together. The coarse-grained copper layer is bonded to the outermost gold layer of the chip and the initial copper layer of the copper-clad substrate on both sides, respectively. The grain size of the coarse-grained copper layer and the initial copper layer of the copper-clad substrate are similar, and migration occurred through the grain boundaries at high temperatures, making it impossible for the software to directly distinguish between the coarse-grained copper layer and the initial copper layer of the copper-clad ceramic substrate. As mentioned above, the thickness of the pre-plating layer does not change fundamentally during the welding process. Therefore, in Figures 24 and 25, the boundary of the coarse-grained copper layer, theoretically formed by the copper pre-plating layer, is shown by a black dashed line, 1 μm below the interface between the gold layer and the coarse-grained copper layer, which is shown by a white dashed line. This welded assembly does not have a connecting copper layer with fine grains.

[0086] Similar to the first embodiment, corresponding measurements and tests were performed on the welded assembly of the second comparative example. For porosity measurement, the sampling range was set to a length of any 10 μm along the length direction of the interface, and a width from a position 1 μm above the interface between the gold layer and the coarse-grained copper layer (i.e., white dashed line) to a position 1 μm below the interface between the coarse-grained copper layer formed by the copper pre-plating layer and the initial copper layer of the copper-clad ceramic substrate (i.e., black dashed line), resulting in a measurement area of ​​approximately 30 μm. 2 The porosity of the welded assembly in the second comparative example is 0.95±0.20%, the average resistivity of the coarse-grained copper layer is 4.0±0.9 μΩ·cm, and the shear strength is 33.8±1.0 MPa.

[0087] Figure 25 shows an SEM image of a cross-section of the welded assembly of the second comparative example after being incubated at 250°C for 165 hours. Within the coarse-grained copper layer, a large number of Kirkendall voids are present near the interface between the gold and copper layers. As can be seen from the measurements, after the high-temperature durability test, the single pore size reaches a maximum of approximately 460 nm. Porosity and resistivity increase significantly, to approximately 8.53 ± 0.12% and 58 ± 2 μΩ·cm, respectively, and the shear strength drops sharply to 22.3 ± 1.8 MPa. Furthermore, the shear surface of the welded unit of the second comparative example changes from ductile fracture to brittle fracture after the high-temperature durability test.

[0088] Table 4 shows the measurement data for each example and comparative example.

[0089] Table 4 shows the porosity, resistivity, and shear strength (all average values) for each example and comparative example.

[0090] [Table 4]

[0091] Compared to the tin-based solder paste of the first comparative example, the welded assemblies of the three embodiments using the copper pre-plating layer exhibited lower porosity and resistivity, and higher shear strength, both when unused and after high-temperature durability testing, and were superior to the welded assemblies obtained using tin-based solder. The reasons for this are as follows: On the one hand, the copper pre-plating layer itself is denser than the tin-based solder paste, effectively removing voids between particles in the solder paste. Also, the small crystal grain size of the connecting copper layer results in nearly identical interdiffusion rates between dissimilar metals (aluminum, gold, silver, nickel, etc.) and copper, significantly reducing the formation and growth of Kirkendall voids, thereby mitigating adverse effects on the long-term stability and reliability of the welded assembly. On the other hand, the stable properties of copper itself prevent the inclusion of brittle intermetallic compounds in the welded assembly, and the occurrence of the latter during the service life is also suppressed, thereby eliminating adverse effects from intermetallic compounds.

[0092] In the case of the second comparative example, the welded assembly may have performance equivalent to that of the embodiment of the present invention when unused, but the durability test results were poor, and good operational stability and reliability could not be achieved. This shows that even if a dense copper pre-plating layer is used, a welded assembly with good service performance cannot necessarily be obtained.

[0093] By comparing the second comparative example with each embodiment of the present invention, it was found that achieving good service performance requires a specific grain size in the connecting copper layer of the welded assembly. As shown in the second comparative example, when the grain size in the joint of the welded assembly is greater than 400 nm, both the service stability and electrical conductivity of the welded assembly are insufficient. In contrast, as verified in each embodiment, when the average grain size of the connecting copper layer is 35 nm to 400 nm, the small grain size provides rapid diffusion channels for dissimilar metals, resulting in a diffusion rate of 1 / 6D for both dissimilar metal atoms and copper atoms. Cu / X ≤D X / Cu-eff ≤6D Cu / X This achieves the following: The Kirkendall effect is effectively weakened during the service life of the welded assembly, i.e., the generation and growth of Kirkendall voids can be significantly reduced, ensuring the stability and electrical conductivity of the welded assembly during subsequent service life. Accordingly, the present invention requires welding a first member and a second member using a copper pre-plating layer to ensure the stability and durability of the welded assembly in a high-temperature service environment, and also requires that the average crystal grain size of the connecting copper layer obtained from the copper pre-plating layer be 35 nm to 400 nm.

[0094] Furthermore, the results from each example show that within this range, the smaller the crystal grain size, the better the suppression effect on Kirkendall pores. For example, when the crystal grain size is 70 nm to 220 nm, the porosity after high-temperature heating remains below 1%, the resistivity is below 5 μΩ·cm, and the shear performance is higher than 33 MPa, compared to the larger crystal grain size of 380 nm in the third example, demonstrating further performance improvement. Also, Table 1 shows that different metals have different preferred average crystal grain sizes due to differences in diffusion coefficients. Therefore, a more preferred average crystal grain size can be selected according to actual usage conditions to further reduce the difference in diffusion rates between dissimilar metals and achieve better performance.

[0095] The present invention further includes a welded assembly for a semiconductor chip package relating to multiple components. For example, the welded assembly of the present invention includes a welded assembly formed by at least two of the first, second, and third embodiments. For example, as shown in Figure 1, a welded assembly according to another embodiment of the present invention is: A semiconductor chip having non-copper metal layers on the outermost layers of the first and second sides, A copper-clad substrate having a first copper layer and a second copper layer on the first side and the second side, respectively, An aluminum heatsink with an aluminum layer on the first side, A conductive copper thin film having a third copper layer on the first side, A first connecting copper layer is provided on both sides, connecting to the non-copper metal layer and the third copper layer on the first side of the semiconductor chip, respectively, to integrally fix the semiconductor chip and the conductive copper thin film. A second connecting copper layer is provided on both sides, connecting to the non-copper metal layer on the second side of the semiconductor chip and the first copper layer, thereby integrally fixing the semiconductor chip and the copper-clad substrate. It includes a third connecting copper layer on each side, which connects to the second copper layer and the aluminum layer, respectively, to integrally fix the copper-clad substrate and the aluminum heat sink. The average crystal grain size of the first, second, and third connecting copper layers is 35 nm to 400 nm.

[0096] Preferably, the first, second, and third connecting copper layers can be selected with a preferred grain size depending on the metal actually being connected, and their thickness is at least 1x the grain size.

[0097] The above examples and experimental data are for illustrative purposes only. Those skilled in the art will see that the present invention is not limited to these examples, and that various modifications can be made without departing the scope of protection of the invention. Furthermore, the steps in any of the above methods do not necessarily have to be performed in the order shown in the drawings, and one or more of the illustrated steps may be performed essentially simultaneously or in combination with additional steps.

Claims

1. A welded assembly for a power semiconductor module package, The outermost layer on the first side is a first member of a non-copper metal layer, A second member having a first copper layer on the first side, The material includes connecting copper layers that integrally fix the first member and the second member by connecting to the non-copper metal layer and the first copper layer on both sides, respectively, to form a first interface and a second interface. The average crystal grain size of the connecting copper layer is 35 nm to 400 nm. Welded assembly.

2. The connecting copper layer has an average crystal grain size of 35 nm to 380 nm and an average thickness of 70 nm to 5 μm, which is at least one times the average crystal grain size. The welding assembly according to claim 1.

3. Within a range where the width is between the first interface and the second interface, each 1 μm away from the connecting copper layer, and the length is any 10 μm along the longitudinal direction of the first interface, the average porosity of the welded assembly is ≤3%, and the pore size is ≤200 nm. The welding assembly according to claim 1.

4. Within a range where the width is between the first interface and the second interface, each 1 μm away from the connecting copper layer, and the length is any 10 μm along the longitudinal direction of the first interface, the average porosity of the welded assembly is ≤ 1%, and the pore size is ≤ 200 nm. The welding assembly according to claim 1.

5. The first component is a semiconductor chip, the second component is a copper-clad substrate, or the first component is a semiconductor chip, the second component is a conductive copper thin film or copper foil, the average crystal grain size of the first copper layer is ≥ 500 nm, the semiconductor chip is a functional chip mainly made of silicon, silicon carbide, or gallium nitride, with one or more metal layers provided on the first and second sides of the chip, the outermost metal layer being a non-copper metal layer, and the non-copper metal layer being a gold layer, aluminum layer, silver layer, nickel layer, or titanium layer. The welding assembly according to any one of claims 1 to 4.

6. The non-copper metal layer is a gold layer, and the average crystal grain size of the connecting copper layer is 35 nm to 140 nm. The welding assembly according to claim 5.

7. The non-copper metal layer is a silver layer, and the average crystal grain size of the connecting copper layer is 100 nm to 380 nm. The welding assembly according to claim 5.

8. The average crystal grain size of the connecting copper layer is 70 nm to 220 nm. The welding assembly according to claim 1.

9. The first component is a heat sink, its non-copper metal layer is an aluminum layer, and the second component is a copper-clad substrate, wherein the average crystal grain size of the first copper layer is ≥ 500 nm. The welding assembly according to any one of claims 1 to 4.

10. The welded assembly has a shear strength of 30 to 40 MPa and / or a resistivity of the connecting copper layer of ≤ 5 μΩ·cm. The welding assembly according to any one of claims 1 to 4.

11. A second copper layer provided on the second side of the second member, A third member having a non-copper metal layer as the outermost layer on the first side, The present invention further includes another connecting copper layer that integrally fixes the second member and the third member by connecting them to the non-copper metal layer and the second copper layer of the third member on both sides, respectively. The average crystal grain size of the aforementioned other connecting copper layer is 35 nm to 400 nm. The welding assembly according to any one of claims 1 to 4.

12. After being kept warm at 250°C for 165 hours, the shear strength of the welded assembly is 30-35 MPa, the resistivity of the connecting copper layer is ≤7 μΩ·cm, and / or the average porosity of the welded assembly is ≤3%, and the pore size is ≤200 nm. The welding assembly according to any one of claims 1 to 4.

13. A welded assembly for a power semiconductor module package, A semiconductor chip having a non-copper metal layer on the outermost layer of the first and second sides, A copper-clad substrate having a first copper layer and a second copper layer on the first and second sides, respectively, A heat sink having an aluminum layer on the first side, A conductive copper thin film having a third copper layer on the first side, A first connecting copper layer is provided on both sides, connecting to the non-copper metal layer on the first side of the semiconductor chip and the third copper layer, respectively, to integrally fix the semiconductor chip and the conductive copper thin film. A second connecting copper layer is provided on both sides, connecting to the non-copper metal layer on the second side of the semiconductor chip and the first copper layer, thereby integrally fixing the semiconductor chip and the copper-clad substrate. It includes a third connecting copper layer on each side, which connects to the second copper layer and the aluminum layer respectively, and integrally fixes the copper-clad substrate and the heat sink. The average crystal grain size of the first, second, and third connecting copper layers is 35 nm to 400 nm. Welded assembly.

14. The first, second, and third connecting copper layers have an average crystal grain size of 70 nm to 380 nm, and their average thickness is 70 nm to 5 μm, which is at least one times the average crystal grain size. The welding assembly according to claim 13.

15. The non-copper metal layer of the semiconductor chip is a gold layer, an aluminum layer, a silver layer, a nickel layer, or a titanium layer, and / or the average porosity of the welded assembly is ≤3%, more preferably the average porosity of the welded assembly is ≤1%, and the pore size is ≤200 nm. The welding assembly according to claim 13 or 14.

16. A method for manufacturing a welded assembly, Step a, a step in which a first member and a second member to be welded are prepared, cleaned, and dried, wherein the second member has an initial copper layer on its first side, and the outermost layer on the first side of the first member is a non-copper metal layer. Step b: A step of introducing a copper pre-plating layer to the surface of the initial copper layer of the cleaned second member, or to the outermost layer on the first side of the cleaned first member, and cleaning to remove the oxide layer on the surface of the copper pre-plating layer, wherein the average crystal grain size of the copper pre-plating layer is 20 nm to 360 nm. Step c involves bringing the copper pre-plated layer of the second member or the first member into close contact with the non-copper metal layer of the first member or the initial copper layer of the second member to form a laminated structure, Step d involves welding the first member and the second member by diffusion welding, including, method.

17. The average crystal grain size of the copper pre-plated layer is 25 nm to 360 nm. The method according to claim 16.

18. The first component is a semiconductor chip, and its non-copper metal layer is a gold layer, an aluminum layer, a silver layer, a nickel layer, or a titanium layer, or the average crystal grain size of the copper pre-plated layer is 30 nm to 200 nm. The method according to claim 16 or 17.

19. In step d, diffusion welding is performed in a helium gas protected environment at a temperature of 250-350°C, with pressure applied from the top and bottom surfaces of the laminated structure, respectively, at a pressure of 15-40 MPa, and a holding time of 10-30 minutes. The method according to claim 16 or 17.

20. The copper pre-plated layer is formed by electroplating, vapor deposition, sputtering, or ion plating. The method according to claim 16 or 17.

Images