Bonded body, ceramic circuit board, and semiconductor device

Abstract

A bonded body according to the present embodiment is provided with a ceramic substrate, a copper plate, and a bonding layer. The bonding layer is disposed on at least one surface of the ceramic substrate and bonds the ceramic substrate and the copper plate. The bonding layer includes a Ti reaction layer and a plurality of first alloys. The Ti reaction layer contains titanium nitride or titanium oxide as a main component. The plurality of first alloys are located between the Ti reaction layer and the copper plate. Each of the plurality of first alloys contains at least one selected from a Cu-Sn alloy and a Cu-In alloy. The plurality of first alloys have mutually differing Sn concentrations or In concentrations. The amount of warping can be reduced by using the present embodiment. Furthermore, the speed of heating and the speed of cooling in a joining process can be increased. In the present embodiment, a silicon nitride substrate is suitable as the ceramic substrate.

Description

[Technical Field]

[0001] The embodiments described below relate to a bonded body, a ceramic circuit board, and a semiconductor device. [Background technology]

[0002] A bonded body of a ceramic substrate and a copper plate is used as a circuit board on which a semiconductor element or the like is mounted. International Publication No. 2018 / 021472 (Patent Document 1) discloses a ceramic copper circuit board in which a ceramic substrate and a copper plate are bonded. In Patent Document 1, a brazing material containing Ag, Cu, Ti, etc. is used for the bonding layer. In Patent Document 1, a 1×10 -3 The bonding is carried out under a pressure of 0.1 Pa or less. Such a 1×10 -3 A pressure below 100 Pa is called a vacuum. The joining method using Ti is called the active metal joining method. Because Ti is an active metal, it is easily nitridized or oxidized. In the active metal joining method, joining is performed in a vacuum to prevent Ti from nitriding or oxidizing before joining. To join in a vacuum, the inside of the joining device must be evacuated. Because the temperature is raised after creating a vacuum, only batch processing was possible in the thermal joining process. In the thermal joining process, there are four steps per batch: evacuation, heating, joining, and cooling. Each batch takes more than 24 hours. This makes it unsuitable for mass production. Meanwhile, International Publication No. 2018 / 199060 (Patent Document 2) discloses a thermal bonding process performed in a continuous furnace. In the example of Patent Document 2, bonding is performed with a furnace length of 3 m and a conveying speed of 10 cm / min. Including holding times during the process, thermal bonding can be completed in about 2 hours. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2018 / 021472 [Patent Document 2] International Publication No. 2018 / 199060 Summary of the Invention [Problem to be solved by the invention]

[0004] Patent Documents 1 and 2 use an Ag-Cu-Ti brazing filler metal containing 40% or more by mass of Ag. Ag and Cu are components that form a eutectic. The solidification temperature of AgCu eutectic is approximately 780°C. The bonding temperature in active metal bonding is 800°C to 950°C. Since the solidification temperature of AgCu eutectic is approximately 780°C, solidification begins during the bonding process. In the bonding process, the material is maintained at the bonding temperature and then cooled to room temperature. If the solidification temperature of the compound present in the bonding layer is high, thermal stress increases during the cooling process. The increased thermal stress leads to an increased amount of warpage of the bonded body. In order to address this problem, the present invention provides a bonded body in which a compound with a low solidification temperature or phase change temperature is present in the bonding layer. [Means for solving the problem]

[0005] The bonded structure according to the embodiment includes a ceramic substrate, a copper plate, and a bonding layer. The bonding layer is disposed on at least one surface of the ceramic substrate and bonds the ceramic substrate to the copper plate. The bonding layer includes a Ti reaction layer and a plurality of first alloys. The Ti reaction layer contains titanium nitride or titanium oxide as a main component. The plurality of first alloys are located between the Ti reaction layer and the copper plate. Each of the plurality of first alloys contains one or more alloys selected from a Cu-Sn alloy and a Cu-In alloy. The plurality of first alloys have different Sn or In concentrations. [Brief explanation of the drawings]

[0006] [Figure 1] FIG. 1 is a schematic view showing an example of a bonded body according to an embodiment. [Figure 2] FIG. 2 is a schematic cross-sectional view showing an example of a bonding layer of the bonded structure according to the embodiment. [Figure 3]A ternary phase diagram showing an example of the composition of a CuSnTi alloy. [Figure 4] 1 is a schematic view showing an example of a ceramic circuit substrate according to an embodiment. [Figure 5] 1 is a schematic diagram showing an example of a semiconductor device according to an embodiment; DETAILED DESCRIPTION OF THE INVENTION

[0007] The bonded structure according to the embodiment includes a ceramic substrate, a copper plate, and a bonding layer. The bonding layer is disposed on at least one surface of the ceramic substrate and bonds the ceramic substrate to the copper plate. The bonding layer includes a Ti reaction layer and a plurality of first alloys. The Ti reaction layer contains titanium nitride or titanium oxide as a main component. The plurality of first alloys are located between the Ti reaction layer and the copper plate. Each of the plurality of first alloys contains one or more alloys selected from a Cu-Sn alloy and a Cu-In alloy. The plurality of first alloys have different Sn or In concentrations. FIG. 1 is a schematic diagram illustrating an example of a bonded structure according to an embodiment. FIG. 2 is a schematic cross-sectional view illustrating an example of a bonding layer of a bonded structure according to an embodiment. In FIGS. 1 and 2, 1 denotes a bonded structure, 2 denotes a ceramic substrate, 3 denotes a copper plate, 4 denotes a bonding layer, 5 denotes a Ti reaction layer, 6 denotes a first alloy, 7 denotes a second alloy, and 8 denotes titanium silicide particles. In FIG. 2, the first alloy 6 is indicated by a dotted line. The second alloy 7 is indicated by dotted shading. Titanium silicide particles 8 are indicated by black fill. In the bonded structure 1 illustrated in FIG. 1, copper plates 3 are provided on both sides of a ceramic substrate 2 via bonding layers 4. The length and width of the ceramic substrate 2 are the same as the length and width of the copper plates 3. The bonded structure according to an embodiment is not limited to this configuration. The bonded structure may have a structure in which the copper plate 3 is provided on only one side of the ceramic substrate 2. The length and width of the ceramic substrate 2 may be different from the length and width of the copper plate 3.

[0008] Examples of the ceramic substrate 2 include a silicon nitride substrate, an aluminum nitride substrate, an aluminum oxide substrate, and an alu-zirconium substrate. The alu-zirconium substrate is a ceramic substrate made of a mixture of aluminum oxide and zirconium oxide. The thickness of the ceramic substrate 2 is preferably 0.1 mm or more and 1 mm or less. If the substrate thickness is less than 0.1 mm, the strength may decrease. Furthermore, if the substrate thickness is greater than 1 mm, the ceramic substrate may become a thermal resistor, which may reduce the heat dissipation performance of the bonded body. The three-point bending strength of the silicon nitride substrate is preferably 600 MPa or more. The thermal conductivity is preferably 80 W / m·K or more. Increasing the strength of the silicon nitride substrate allows the substrate thickness to be reduced. For this reason, the three-point bending strength of the silicon nitride substrate is preferably 600 MPa or more, and more preferably 700 MPa or more. The thickness of the silicon nitride substrate can be reduced to 0.40 mm or less, and even 0.30 mm or less. The three-point bending strength of an aluminum nitride substrate is approximately 300 to 450 MPa. On the other hand, the thermal conductivity of an aluminum nitride substrate is 160 W / m·K or more. Because the strength of an aluminum nitride substrate is low, the substrate thickness is preferably 0.60 mm or more. The three-point bending strength of aluminum oxide substrates is approximately 300 to 450 MPa, but aluminum oxide substrates are inexpensive.Also, the three-point bending strength of Al2O3 substrates is high at approximately 550 MPa, but their thermal conductivity is approximately 30 to 50 W / m·K. The ceramic substrate 2 is preferably either a silicon nitride substrate or an aluminum nitride substrate. Silicon nitride substrates and aluminum nitride substrates are nitride ceramic substrates. Nitride ceramics react with a Ti-containing active metal brazing material to form a reaction layer mainly composed of titanium nitride. Furthermore, oxide ceramics react with a Ti-containing active metal brazing material to form a reaction layer mainly composed of titanium oxide. Examples of oxide ceramics include aluminum oxide substrates and Aldrich substrates. The Ti reaction layer 5 is a layer mainly composed of titanium nitride or a layer mainly composed of titanium oxide. The layer mainly composed of titanium nitride or a layer mainly composed of titanium oxide is formed by a reaction between the ceramic substrate 2 and the Ti in the active metal brazing material.

[0009] Copper plates 3 are preferably disposed on both sides of the ceramic substrate 2. By bonding copper plates to both sides, warping of the bonded body can be suppressed. A pure copper plate or a copper alloy plate can be used as the copper plate 3. The copper plate 3 is preferably oxygen-free copper. As specified in JIS-H-3100 (ISO1337, etc.), the copper purity of oxygen-free copper is 99.96 mass% or more. Oxygen-free copper is a type of pure copper. The copper plate 3 is used as a circuit part or a heat sink. By increasing the thickness of the copper plate 3, the current-carrying capacity and heat dissipation performance can be improved. For this reason, the thickness of the copper plate 3 is preferably 0.6 mm or more, and more preferably 0.8 mm or more.

[0010] The bonding layer 4 includes a Ti reaction layer 5 whose main component is titanium nitride or titanium oxide. The Ti reaction layer 5 is formed when a so-called active metal bonding method is used. When the ceramic substrate 2 is a nitride-based ceramic, the main component of the Ti reaction layer 5 is titanium nitride (TiN). The titanium nitride, which is the main component of the Ti reaction layer 5, may contain a compound having an atomic ratio of titanium to nitrogen other than 1:1. When the ceramic substrate 2 is an oxide-based ceramic, the main component of the Ti reaction layer 5 is titanium oxide (TiO2). The titanium oxide, which is the main component of the Ti reaction layer 5, may contain a compound having an atomic ratio of titanium to oxygen other than 1:2, such as TiO2, TiO, or Ti2O3. The Ti reaction layer 5 refers to a region having a Ti content of 50 mass % or more formed on the surface of the ceramic substrate 2. The Ti reaction layer 5 is preferably present in an area of ​​40% to 100% per 200 μm unit length on the surface of the ceramic substrate 2. "40% or more per 200 μm unit length" refers to a state in which, when a cross section of the bonding layer 4 is observed, the Ti reaction layer 5 is formed over a total of 40 μm or more of the 200 μm on the surface of the ceramic substrate 2.

[0011] The bonding layer 4 includes a plurality of first alloys 6 located between the Ti reaction layer 5 and the copper plate 3. The composition ratios of the plurality of first alloys 6 are different from one another. Specifically, the plurality of first alloys 6 have different Sn or In concentrations. Each of the plurality of first alloys contains one or more selected from Cu-Sn alloys and Cu-In alloys. It is not necessary for all Cu-Sn alloys or Cu-In alloys present in the bonding layer 4 to have different Sn or In concentrations. It is sufficient that two or more first alloys 6 having different Sn or In concentrations are present in the bonding layer 4. The following mainly describes an example in which the first alloy 6 is a Cu-Sn compound. In the following example, the Cu-Sn compound can be replaced with a Cu-In compound. The term "different Sn concentrations" means that the Sn concentrations differ by 2 atomic % or more when the total of the Cu content and the Sn content is 100 atomic %, and the term "different In concentrations" means that the In concentrations differ by 2 atomic % or more when the total of the Cu content and the In content is 100 atomic %. The atomic ratio of Cu to Sn can be measured by point analysis using energy dispersive X-ray spectroscopy (EDX). EDX is sometimes called EDS. Here, SEM-EDX is used as EDX. Field emission SEM (FE-SEM) may also be used as SEM. The FE-SEM used is a JEOL JSM-7200F or a device with equivalent performance, and the EDX used is a JEOL EX-74600U4L2Q or a device with equivalent performance. The cross section of the bonding layer 4 is observed using FE-SEM. The cross section is a plane parallel to the thickness direction. The thickness direction is perpendicular to the surface of the ceramic substrate 2 and parallel to the direction connecting the ceramic substrate 2 and the copper plate 3. The measurement conditions for the FE-SEM are set to an acceleration voltage of 15 kV and a magnification of 3000 times. The field of view area is set to the thickness of the bonding layer x 40 μm in the width direction. The width direction is parallel to the surface of the ceramic substrate 2 and perpendicular to the thickness direction. The EDX measurement conditions are set to 50 scans and a dwell time of 0.2 ms. Dwell time is the measurement speed per pixel. When performing EDX area analysis, the capture pixel count is set to 256 horizontal x 198 vertical pixels, the detection count is set to 3700-4100 cps (counts per second), and the quantitative map is set to 5 x 5 bits / point. Point analysis is an analysis using the results of a single detection count (one location). Area analysis is an analysis using the results of multiple detection count measurements. Area analysis is also sometimes called surface analysis. EDX analysis can be performed after first examining the elements that make up the bonding layer and identifying the specific elements. For example, if Cu (copper), Sn (tin), Ti (titanium), Si (silicon), O (oxygen), and N (nitrogen) are detected as elements that make up the bonding layer, these elements are set as specific elements. The area ratio can be determined by mapping the specific elements using the results of area analysis. If the thickness of the bonding layer is 30 μm, the field of view area is 30 μm in the thickness direction and 40 μm in the width direction. For SEM-EDX mapping, use the function attached to the SEM-EDX. If the SEM-EDX does not have a mapping function, use image software for mapping. The area ratio is calculated by EDX area analysis of three regions and using the average value. The area of ​​one region (field of view area) is the bonding layer thickness x 40 μm in the width direction. The measurement field of view is aligned with the bonding layer 4. Any three adjacent regions are analyzed from the cross section of the bonding layer 4. The bonding layer 4 ranges from the boundary between the ceramic substrate 2 and the bonding layer 4 to the boundary between the bonding layer 4 and the copper plate 3. The boundary between the ceramic substrate 2 and the bonding layer 4 is the bonding interface between the surface of the ceramic substrate 2 and the Ti reaction layer 5. For example, when a silicon nitride substrate is used, the bonding interface between the silicon nitride substrate and the titanium nitride layer is the boundary between the ceramic substrate 2 and the bonding layer 4. For example, the dashed line BL1 shown in FIG. 2 is the boundary between the ceramic substrate 2 and the bonding layer 4. The boundary between the bonding layer 4 and the copper plate 3 is defined based on the Ti content. Near the boundary between the bonding layer 4 and the copper plate 3, the Ti content decreases from the bonding layer 4 toward the surface of the copper plate 3. The boundary is defined based on the location where a region with a Ti content of 1 atomic % or less is formed continuously for 50 μm in the width direction. If there are multiple such locations, the boundary between the bonding layer 4 and the copper plate 3 is defined based on the location closest to the bonding layer 4 (ceramic substrate 2). In Figure 2, the solid line SL between the bonding layer 4 and the copper plate 3 indicates a set of points where the Ti content is 1 atomic %. As an example, near the solid line SL, in parts P1 and P2, a region with a Ti content of 1 atomic % or less is formed continuously for 50 μm in the width direction. Part P1 is located closer to the ceramic substrate 2 than part P2. Therefore, the dashed line BL2, based on part P1, is defined as the boundary between the bonding layer 4 and the copper plate 3. EDX point analysis is used to measure the amount of Ti at the boundary between the bonding layer 4 and the copper plate 3. Point analysis is performed at 5 μm intervals to determine the amount of Ti. Furthermore, the region where the amount of Ti is 1 atomic % or less also includes a region where the amount of Ti is 0 atomic % (below the detection limit). The first alloy 6 refers to an alloy in which the sum of Cu and the first element is 50 atomic % or more when the sum of Cu, the first element, and Ti is 100 atomic %. The first element is one or two selected from Sn and In. For example, in a Cu-Sn alloy, the sum of Cu and Sn is 50 atomic % or more when the sum of Cu, Sn, and Ti is 100 atomic %. In a Cu-In alloy, the sum of Cu and In is 50 atomic % or more when the sum of Cu, In, and Ti is 100 atomic %. The composition of the first alloy 6 is determined by EDX point analysis. The second alloy 7 refers to an alloy in which the sum of Ti and the first element is 50 atomic % or more when the sum of Cu, the first element, and Ti is 100 atomic %. For example, in the case of a Ti-Sn alloy, the sum of Ti and Sn is 50 atomic % or more when the sum of Cu, Sn, and Ti is 100 atomic %. In the case of a Ti-In alloy, the sum of Ti and In is 50 atomic % or more when the sum of Cu, In, and Ti is 100 atomic %. The composition of the second alloy 7 is determined by EDX point analysis. The Cu-Sn alloy may be a mixture of Cu and Sn, or may contain a compound of Cu and Sn. The compound may include an intermetallic compound. The Cu-Sn alloy may contain Ti mixed therein or in a solid solution state. The first alloy 6 includes an alloy having a melting point of 400°C or higher and 600°C or lower. The melting point of the first alloy 6 can be determined from the freezing point of the alloy. The freezing point refers to the temperature at which a liquid becomes solid. In other words, it refers to the temperature at which the brazing filler metal melts, becomes liquid, and then cools to become solid. A simple method for determining the freezing point of the compound may be to use a phase diagram. Alternatively, the bonding layer 4 may be cut out from the bonded body 1 and gradually heated to measure the melting temperature.

[0012] The melting point of the first alloy 6 can be measured from the DSC curve. The DSC curve is the result of applying heat to a sample using a differential scanning calorimeter (DSC) to measure whether or not an endothermic or exothermic reaction occurs. When an endothermic or exothermic reaction occurs, a peak appears on the DSC curve. A negative peak indicates the occurrence of an endothermic reaction. A positive peak indicates the occurrence of an exothermic reaction. An endothermic reaction indicates the melting or decomposition of the sample. An exothermic reaction indicates the reaction of the constituent elements of the sample to form a compound (including an alloy) or solidify. The larger the peak, the greater the heat of reaction. Here, a negative peak is called an endothermic peak, and a positive peak is called an exothermic peak. The apex of the peak is called the peak top. The difference between the maximum and minimum points of the peak is called the peak height.

[0013] For example, since an endothermic peak is a peak in the negative direction, the DSC curve goes down and then up near the endothermic peak. Near the endothermic peak, the DSC curve goes from maximum to minimum to maximum. This minimum point (the point where it is lowest in the negative direction) is the peak top. Furthermore, of the maximum points on the low-temperature and high-temperature sides of the minimum point, the larger value is considered to be the maximum point (the point where it is highest in the positive direction). The peak height is the value obtained by subtracting the minimum point from the maximum point. Since the exothermic peak is a peak in the positive direction, the DSC curve rises and falls near the exothermic peak. Near the exothermic peak, the DSC curve goes through a minimum, maximum, and then minimum. This maximum (the point where the temperature rises most in the positive direction) is the peak top. Furthermore, of the minimum points on the low-temperature and high-temperature sides of the maximum, the smaller value is the minimum (the point where the temperature drops most in the negative direction). The peak height is the value obtained by subtracting the minimum from the maximum. Note that the end point of an endothermic peak (negative peak) may appear to be an exothermic peak (positive peak). Here, this end point is also counted as an exothermic peak. Alternatively, a baseline may be drawn to determine the height of each peak. In addition, on the DSC curve, a change of 0.02 mW / mg or more on the vertical axis is counted as a peak. In other words, a change of less than 0.02 mW / mg is not counted as a peak.

[0014] In the DSC curve during the cooling process, an exothermic peak is preferably detected in the range of more than 600°C to less than 900°C. The exothermic peak within this range is primarily due to the solidification reaction. Furthermore, it is also preferable for there to be an exothermic peak within the range of 400°C to 600°C. As described above, an exothermic reaction indicates that the brazing filler metal components react with each other to form a compound (including an alloy) or solidification occurs. In other words, an exothermic peak within the range of 400°C to 600°C indicates that a compound is formed or a solidification reaction occurs within that temperature range. Solidification is a reaction that involves a phase change from a liquid to a solid. Furthermore, the formation of a compound may also involve a phase change from a solid to a solid with a different crystalline structure. By causing such phase changes to occur in multiple stages, stress during cooling can be alleviated. With the brazing filler metal composition described below, the phase change to form the compound is primarily responsible for the exothermic peak. In other words, the presence of the first alloy 6 can cause an exothermic peak to occur within the range of 400°C to 600°C on the DSC curve during the cooling process. The presence of exothermic peaks in the ranges of 400°C to 600°C and 600°C to 900°C allows for multi-stage reactions involving phase changes. Here, the temperature at which a phase change from liquid to solid occurs is called the freezing point. The temperature at which a phase change to a solid with a different crystalline structure occurs is called the phase change point. When multiple alloys (including compounds) with different composition ratios exist, the exothermic peaks of those alloys may overlap. In other words, even if only one exothermic peak is detected, multiple alloys (including compounds) with different composition ratios may exist. The freezing point is the temperature at which a material changes from a liquid phase to a solid phase. The phase change point is the temperature at which a material changes into a solid with a different crystalline structure. Solidification or phase change within the range of 400°C to 600°C can lower the temperature at which thermal stress occurs. The freezing point of a conventional AgCu eutectic compound is approximately 780°C. During the cooling process, a compound solidified at 780°C generates thermal stress before returning to room temperature. The existence of first alloy 6, whose freezing point or phase change point is within the range of 400°C to 600°C, can reduce the thermal stress occurring until the material returns to room temperature after solidification or phase change at 400°C to 600°C. The multiple first alloys 6 have different composition ratios, and therefore their freezing points or phase change points are also different. Multiple first alloys 6 with different freezing points or phase change points exist within the range of 400°C to 600°C. This allows the solidification or phase change of the bonding layer to occur in multiple stages, further reducing thermal stress. Thermal stress occurs when the brazing filler metal melts, turns to a liquid phase, solidifies, and cools to room temperature. It is particularly important to reduce the thermal stress from solidification to cooling to room temperature. By allowing the solidification or phase change of the first alloy 6 to occur in multiple stages, the thermal stress from solidification to cooling to room temperature can be reduced. If the freezing point or phase change point of the first alloy 6 is lower than 400°C, the freezing point or phase change point of the bonding layer 4 may be lowered, potentially resulting in poor temperature cycle test (TCT) characteristics. If the freezing point or phase change point of the first alloy 6 is higher than 600°C, thermal stress may be increased. Therefore, the solidification point or phase change point of the first alloy 6 is preferably 400° C. or higher and 600° C. or lower, and more preferably 450° C. or higher and 550° C. or lower. Even if the exothermic peaks between 400° C. and 600° C. or lower on the DSC curve in the cooling step overlap, the presence of multiple first alloys 6 with different composition ratios still provides the effect of solidification or phase change occurring in multiple stages.

[0015] Furthermore, the first alloy 6 preferably contains one or more selected from Cu-Sn compounds and Cu-In compounds. In compounds, constituent elements are bonded to each other at the atomic level. Examples of compounds include intermetallic compounds. Examples of alloys include solid solutions and eutectics. Compounds are distinguished from solid solutions and eutectics. The phase structure can be specified by the crystal structure and composition. For example, in a CuSn binary alloy, if the Sn content is 9 atomic % or less, the alloy is mainly a solid solution. If the Sn content is 10 atomic % or more, the alloy is more likely to form intermetallic compounds. It is preferable that the Sn content of at least some of the multiple first alloys 6 is in the range of 3 atomic % to 30 atomic %. When the Sn content is in the range of 3 atomic % to 30 atomic %, it is easy to control the freezing point or phase change point. If the Sn content is less than 3 atomic %, the freezing point or phase change point will not be sufficiently lowered. On the other hand, if the Sn content exceeds 30 atomic %, the freezing point or phase change point may be lowered too much. For this reason, the Sn content is preferably 3 atomic % to 30 atomic %. Furthermore, a first alloy 6 having an Sn content of 3 atomic % to 30 atomic % will easily form a Cu-Sn compound. When EDX point analysis is performed on a Cu-Sn alloy region where Cu is found to be high in EDX area analysis, it is preferable that 80% or more of the region is a Cu-Sn alloy with an Sn content of 3 atomic % or more and 30 atomic % or less. EDX point analysis is performed 10 or more times. A Cu-high region refers to a region where Cu is found to be higher in atomic % than in atomic % Ti by EDX area analysis.

[0016] The first alloy 6 may also contain 2 atomic % or less of metal components other than Cu and Sn. The metal components other than Cu and Sn are metal components or impurities constituting the brazing filler metal. Examples of components of the brazing filler metal other than Cu and Sn include titanium (Ti) and carbon (C), as described below. For example, the freezing point of a TiSn alloy exceeds 600°C. If the amount of Ti in a Cu-Sn alloy increases, the freezing point or phase change point may not be within the range of 400°C to 600°C. For this reason, it is preferable that the amount of Ti in at least some of the multiple first alloys 6 is 2 atomic % or less, or even 1 atomic % or less. EDX point analysis is used to measure the amount of Ti in the Cu-Sn compound. Examples of components other than metal components include oxygen and nitrogen. The first alloy 6 may also contain 10 atomic % or less of components other than Cu, Sn, and metal components. When Ag is contained in the brazing filler metal, an AgCu eutectic is easily formed. The melting point of AgCu eutectic is approximately 780°C. For this reason, it is preferable that the brazing filler metal does not contain Ag. In other words, it is preferable that the bonding layer 4 does not contain Ag. Ag-Cu-Sn-Ti, which is mainly composed of Ag, is sometimes used as the brazing filler metal. When Ag is the main component, AgCu or AgSn is formed. This reduces the proportion of CuSn alloy contained in the bonding layer 4. This reduces the possibility of forming a Cu-Sn alloy with a different composition. From this point of view, it is also preferable that the bonding layer 4 does not contain Ag.

[0017] Preferably, some of the multiple first alloys 6 are Cu-Sn alloys with an Sn content of 3 atomic % or more and 9 atomic % or less, and another part of the multiple first alloys 6 are Cu-Sn alloys with an Sn content of 10 atomic % or more and 30 atomic % or less. Here, a Cu-Sn alloy with an Sn content of 3 atomic % or more and 9 atomic % or less is referred to as a first Cu-Sn alloy, and a Cu-Sn alloy with an Sn content of 10 atomic % or more and 30 atomic % or less is referred to as a second Cu-Sn alloy. The first Cu-Sn alloy is preferably a Cu-Sn compound. The second Cu-Sn alloy is preferably a Cu-Sn compound, because compounds are more likely to undergo multi-stage solidification or phase change. The bonding layer 4 preferably contains a second alloy 7. The second alloy 7 contains one or two selected from a Ti—Sn alloy and a Ti—In alloy. The Ti—Sn alloy or the Ti—In alloy may contain an intermetallic compound. For example, the intermetallic compound of the Ti—Sn alloy may be one or more selected from SnTi3, Sn3Ti5, and Sn5Ti6. In the average value measured at three locations on the bonding layer 4, the total area of ​​the multiple first alloys 6 is preferably larger than the total area of ​​the second alloys 7. That is, the total area of ​​the Cu-Sn alloy or Cu-In alloy is preferably larger than the total area of ​​the Ti-Sn alloy or Ti-In alloy. The field area is set to the thickness of the bonding layer x 40 μm in the width direction. Measurement of one field area may be performed multiple times. Furthermore, when measuring multiple times, the areas are set so that they do not overlap with each other. Furthermore, at least a part of the Cu-Sn alloy and the Ti-Sn alloy detected by EDX point analysis of the bonding layer 4 are preferably within the region surrounded by (97,3,0), (60,40,0), (2,40,58), and (39,3,58) in the Cu, Sn, Ti ternary phase diagram. Here, this region is referred to as the first composition region. Figure 3 is a ternary phase diagram of Cu, Sn, and Ti, where the concentration of each element is shown in atomic percent. First, image data is obtained by element mapping using the SEM-EDX area analysis function. Using the mapping function, EDX point analysis is performed on the areas identified as the first alloy 6 and the second alloy 7. Element mapping allows the difference in concentration of a specific element to be displayed by brightness. Areas with high concentration have high brightness, while areas with low concentration have low brightness. For example, by specifying Ti as the specific element, the Ti concentration gradient can be visualized. This function allows the first alloy 6 and the second alloy 7 present in the field of view to be distinguished. That is, by specifying the first element as the specific element and visualizing the concentration of this element, the first alloy 6 present in the field of view can be distinguished. Similarly, by specifying Ti as the specific element and visualizing the concentration of this element, the second alloy 7 present in the field of view can be distinguished. When EDX point analysis is performed on the areas identified as the first alloy 6 and the second alloy 7, it is preferable that the composition falls within the first composition range. While alloys outside the first composition range are acceptable, maintaining the alloy composition within the first composition range can improve properties. More preferably, when EDX point analysis is performed, the composition of at least a portion of the Cu-Sn alloy is within the region surrounded by (97,3,0), (77,23,0), (52,23,25), and (72,3,25) in the Cu, Sn, Ti ternary phase diagram. This region is referred to as the second composition region. The bonding layer 4 may contain a first alloy 6 having a composition outside the second composition region. However, if the amount of the first alloy 6 having a composition in the second composition region is large, the properties can be further improved. Furthermore, when the Cu-Sn alloy is observed by elemental mapping based on SEM-EDX area analysis and subjected to EDX point analysis, it is preferable that 80% or more of the analyzed points have a composition within the range of the second composition region. In the EDX point analysis, 10 or more points spaced 3 μm or more apart from each other are analyzed. The Ti-Sn alloy detected by EDX point analysis of the bonding layer 4 is preferably within the region bounded by (41,23,36), (24,40,36), (2,40,58), and (19,23,58) in the Cu, Sn, Ti ternary phase diagram. This region is referred to as the third composition region. When the bonding layer 4 contains multiple Ti-Sn alloys, it is sufficient that at least some of the compositions of the multiple Ti-Sn alloys fall within the third composition region. On the other hand, a larger number of Ti-Sn alloys with compositions within the third composition region can further improve the properties. Furthermore, when the Ti-Sn alloy is observed by elemental mapping based on SEM-EDX area analysis and subjected to EDX point analysis, it is preferable that 80% or more of the analyzed points have a composition within the third composition region. In the EDX point analysis, 10 or more points spaced 3 μm or more apart from each other are analyzed.

[0018] The Ti reaction layer 5 preferably contains titanium nitride particles with an average particle size of 50 nm or less. Furthermore, it is preferable that Cu and the first element are present at the grain boundaries of the titanium nitride particles. The size of the titanium nitride particles in the Ti reaction layer 5 is analyzed using a scanning transmission electron microscope (STEM). An enlarged photograph of the cross section of the bonding layer is obtained using the STEM. The particle size is determined as the distance between the two furthest points on the outer edges of the titanium nitride particles in the enlarged photograph. The particle sizes of 30 random particles are measured. The average value of these measurements is determined as the average particle size. A small average particle size of 50 nm or less makes it easier for Cu and the first element to penetrate the grain boundaries. The presence of Cu and the first element at the grain boundaries of the titanium nitride particles allows the solidification or phase change of the bonding layer 4 to occur in multiple stages.

[0019] The Cu content in the Ti reaction layer 5 is preferably in the range of 0.5 atomic % or more and 5 atomic % or less. When the Cu content in the Ti reaction layer 5 is in this range, thermal stress can be reduced without reducing the bonding strength. When the Cu content in the Ti reaction layer 5 is less than 0.5 atomic %, the Cu content is too small and the effect of reducing thermal stress may not be sufficiently obtained. When the Cu content is greater than 5 atomic %, the strength of the Ti reaction layer 5 may be insufficient, resulting in a reduction in bonding strength. The amount of the first element in the Ti reaction layer 5 is preferably 0.05 atomic % or more and 2 atomic % or less. When the amount of the first element in the Ti reaction layer 5 is within this range, thermal stress can be reduced without reducing the bonding strength. Furthermore, Cu and Sn in the Ti reaction layer 5 are easily alloyed. When Cu and Sn in the Ti reaction layer 5 become an alloy (preferably a compound), the occurrence of thermal stress can be suppressed. The amounts of Cu and the first element in the Ti reaction layer 5 refer to the amounts present at the grain boundaries of titanium nitride particles. The titanium nitride in the Ti reaction layer 5 may contain both TiN and Ti2N. The presence of both TiN and Ti2N allows the layer containing titanium nitride as the main component to solidify in multiple stages. The thickness of the Ti reaction layer 5 is preferably 1 μm or less. More preferably, the thickness of the Ti reaction layer 5 is 0.3 μm or less. By thinning the Ti reaction layer 5, it is possible to make it easier for Cu and the first element to be present at the grain boundaries. The thickness of the Ti reaction layer 5 can be measured using a transmission electron microscope (TEM). The thickness of the Ti reaction layer 5 is measured based on the interface between the Ti reaction layer 5 and the ceramic substrate. TEM-EDX is used for composition analysis of the Ti reaction layer 5.

[0020] In the average value obtained by EDX area analysis of three regions of the bonding layer 4, the total area of ​​the first alloy 6 is preferably 50% to 95% of the area of ​​the three regions. The field area (size of one region) is set to the bonding layer thickness x 40 μm in the width direction. When the total area ratio of the Cu-Sn alloy is within the range of 50% to 95%, the thermal stress of the bonding layer 4 can be reduced. When the total area ratio of the Cu-Sn alloy is less than 50%, the proportion of the Cu-Sn alloy is low, which may result in insufficient effect of reducing thermal stress. Furthermore, when the total area ratio of the Cu-Sn alloy is greater than 95%, the freezing point or phase change point of the bonding layer 4 may be too low. A lowering of the freezing point or phase change point of the bonding layer 4 may degrade the TCT characteristics of the bonded body 1. Therefore, the total area ratio of the Cu-Sn alloy is preferably 50% to 95%, more preferably 70% to 90%. Furthermore, it is preferable that the total area of ​​the Cu—Sn alloy is within the range of 50% to 95% no matter which of the three regions in the bonding layer 4 is measured. In the average value obtained by EDX area analysis of the three regions of the bonding layer 4, the total area of ​​the Ti-Sn alloy is preferably within a range of 5% to 30% of the area of ​​the three regions. The total area of ​​the titanium silicide is preferably within a range of 0.5% to 15% of the area of ​​the three regions. Furthermore, the area ratio of the other regions is preferably within a range of 10% or less. Examples of other material structures include a Ti reaction layer and carbides. To measure the area ratio, the area analysis function of the SEM-EDX is used to obtain image data by elemental mapping. The mapping function attached to the area analysis function of the SEM-EDX is used. If the SEM-EDX does not have a mapping function, an electron probe microanalyzer (EPMA) can also be used. Image data obtained by SEM-EDX can also be processed using image analysis software. Image-J or equivalent software can be used as image analysis software. The field of view area is set to the bonding layer thickness x 40 μm in the width direction. The measurement conditions for area analysis are as described above. The element map obtained by area analysis is used to calculate the area ratio. For the element map, the pinning size is set to 5 x 5 bits / point. When using image analysis software such as Image-J, binarization processing is performed. The concept behind the binarization process is as follows. First, let Sx be the area of ​​the element x you want to detect. This can be calculated by Sx = ((number of pixels in the extracted area after binarization of element x) / (total number of pixels in the image)) x 100. The element x you want to detect can be changed to Ti, Si, N, or C, and the area can be calculated for each. For example, the area ratio of the Ti-Sn alloy can be calculated as follows. As mentioned above, in the second composition region of the Cu-Sn alloy (first alloy 6), the amount of Ti is between 0 atomic % and 25 atomic %. Therefore, the region where the amount of Ti is 25 atomic % or less is excluded. The element map is displayed with a brightness according to the Ti concentration. Using this function, the region where the amount of Ti is 25 atomic % or less can be excluded. The image of the element map excluding the region where the amount of Ti is 25 atomic % or less is binarized, and the area of ​​the Ti element and the area of ​​the Si element are measured. The area of ​​the Ti-Sn alloy is calculated as S. Sn-Ti , the area of ​​Ti element is S Ti , the area of ​​Si element is S Si , the area of ​​N elements is S N , the area of ​​the C element is S C The area of ​​the Ti-Sn alloy is S Sn-Ti =S Ti -(S Si +S N +S C) can be obtained. The Si present in the bonding layer is titanium silicide. The N (nitrogen) present in the bonding layer is titanium nitride. If the boundary between the silicon nitride substrate surface and the bonding layer is wavy, silicon nitride may be detected. Also, the C (carbon) present in the bonding layer is either elemental carbon or titanium carbide. S Ti , S Si , S N , S C Alternatively, the image obtained by mapping only Ti, Si, N, and C may be binarized and calculated using these images. The area ratio of the Cu-Sn alloy is S Cu-Sn Let's say S Cu-Sn The Ti content ranges from 0 atomic % to 25 atomic %. Si , S N , S C It can be found by subtracting The Ti element may be contained in both Cu-Sn alloys and Ti-Sn alloys. Cu-Sn alloys and Ti-Sn alloys can be distinguished based on a Ti content of 25 atomic %. The bonding layer 4 may contain carbon. The carbon present in the bonding layer 4 may be elemental carbon or carbide. The addition of carbon can control the fluidity of the bonding brazing material. This makes it easier to control the distribution of the Cu-Sn alloy in the bonding layer. Materials other than the Cu-Sn alloy present in the bonding layer 4 include titanium nitride (TiN), a Ti-Sn alloy, and titanium silicide. Furthermore, when carbon is contained, titanium carbide (TiC) may also be present. Furthermore, the components constituting the bonding layer 4 may exist as simple metals. Note that alloys and simple metals can be distinguished from TEM electron diffraction images.

[0021] The bonding layer 4 preferably contains titanium silicide particles 8 having a major axis of 0.5 μm or more and 6 μm or less. When the field area of ​​the bonding layer 4 = bonding layer thickness × width direction 200 μm, the total area of ​​the titanium silicide particles 8 having a major axis of 0.5 μm or more and 6 μm or less is preferably 1% or more and 15% or less of the field area. SEM photographs are used to measure the long diameter of the titanium silicide particles 8. The long diameter is defined as the distance between the two furthest points on the outer edge of the titanium silicide particle 8 in an enlarged photograph of the cross section of the bonding layer 4. The titanium silicide particles 8 present in the bonding layer 4 can be identified using the mapping function included in the area analysis function of the SEM-EDX. Specifically, Ti and Si are designated as specific elements, and the presence of these elements is visualized. The area where both Ti and Si are present corresponds to the titanium silicide particle 8. If titanium silicide particles 8 are present, it is preferable that 80% to 100% of the titanium silicide particles 8 are present within 1 μm of the Ti reaction layer 5. When Ti that did not form the Ti reaction layer 5 or the Ti-Sn alloy reacts with the silicon nitride substrate, titanium silicide particles are formed. The presence of titanium silicide particles 8 near the Ti reaction layer 5 makes it easier for Cu and the first element to penetrate into the grain boundaries of the titanium nitride particles. If the major axis of the titanium silicide particles 8 exceeds 6 μm or the area ratio exceeds 15%, the formation of the Ti reaction layer 5 may be insufficient.

[0022] The thickness of the bonding layer 4 is preferably 5 μm or more and 60 μm or less. Within this range, it is easy to control the area ratio of the Cu—Sn alloy. The thickness of the bonding layer 4 is the distance from the boundary between the ceramic substrate 2 and the Ti reaction layer 5 to the boundary between the bonding layer 4 and the copper plate 3. The concentration (atomic %) of the first element at the boundary between the bonding layer 4 and the copper plate 3 is defined as a first concentration C1. The concentration (atomic %) of the first element in a region in the copper plate 3 that is 10 μm away from the boundary between the bonding layer 4 and the copper plate 3 is defined as a second concentration C2. The ratio C2 / C1 of the second concentration C2 to the first concentration C1 is preferably 0.60 or less. Furthermore, the concentration (atomic %) of the first element in a region in the copper plate 3 that is 20 μm away from the boundary between the bonding layer 4 and the copper plate 3 is defined as a third concentration C3. The ratio C3 / C1 of the third concentration C3 to the first concentration C1 is preferably 0.40 or less. When a pure copper plate is used as the copper plate 3, the second concentration C2 and the third concentration C3 indicate the amount of the first element diffused into the copper plate 3. When Sn is used in the bonding layer 4, the concentration of Sn is used as the concentration of the first element. When In is used in the bonding layer 4, the concentration of In is used as the concentration of the first element. When both Sn and In are used in the bonding layer 4, the sum of the concentrations of Sn and In is used as the concentration of the first element. For example, when the bonding layer 4 contains a Cu—Sn alloy, the first concentration is C1 Sn , the second concentration is C2 Sn , the third concentration is C3 Sn It is written as C2 Sn / C1 Sn ≦0.60 or C3 Sn / C1 Sn The fact that the value is ≦0.40 indicates that the diffusion of Sn into the copper plate is suppressed. By having the first alloy 6 with a different composition ratio present in the bonding layer 4, the amount of diffusion of the first element into the copper plate 3 can be reduced. The diffusion distance of the first element from the boundary is preferably 50 μm or less. The diffusion region refers to the distance from the boundary to the point where the Sn content is 0 mass % (below the detection limit). Furthermore, as will be described later, if a bonding method using a continuous furnace is used, the first concentration can be reduced to 8 atomic % or less. The second concentration C2 and the third concentration C3 in the copper plate 3 are analyzed by EDX. Line analysis is used for the analysis. By reducing the amount of Sn diffused into the copper plate 3, the heat resistance of the copper plate 3 can be improved. For example, when Sn diffuses into the copper plate 3, a CuSn alloy is formed in the copper plate. The CuSn alloy has a lower melting point than the copper plate. Therefore, if the CuSn alloy in the copper plate increases, the copper plate becomes more susceptible to deformation due to heat. This may result in a decrease in the TCT characteristics of the joined body. While the explanation here is based on Sn, the same applies to In. The above configuration reduces the amount of voids in the bonding layer. By reducing the amount of Sn (or In) diffused into the copper plate, the volumetric void amount in the bonding layer can be reduced to 0% or more and 3% or less. Ultrasonic testing (SAT) is used to calculate the amount of voids in the bonding layer. A flaw detection probe is placed in contact with the surface of the bonded body 1, and ultrasonic waves are transmitted toward the bonding layer 4. The voids present in the bonding layer can be determined from the reception results of the reflected waves. Furthermore, the volumetric void amount can be calculated from the reception results. Reducing the amount of voids in the bonding layer improves the bonding strength of the copper plate. The bonding strength is measured by a peel test. The bonding strength of the copper plate can be increased to 15 kN / mm or more, and even 25 kN / mm or more.

[0023] The bonded body as described above is suitable for use in ceramic circuit boards. FIG. 4 is a schematic diagram showing an example of a ceramic circuit substrate according to an embodiment. In FIG. 4, 10 denotes a ceramic circuit substrate, 11 denotes a circuit portion, and 12 denotes a heat sink. The ceramic circuit substrate 10 shown in FIG. 4 is produced by processing the copper plate 3 on the front side of the bonded body 1 into the circuit portion 11 and processing the copper plate 3 on the back side into the heat sink 12. The copper plate 3 having a circuit pattern is referred to as the circuit portion 11. In FIG. 4, two circuit portions 11 are provided. Note that the ceramic circuit substrate 10 according to the embodiment is not limited to this structure. The number and size of the circuit portions 11 can be appropriately changed as needed. Also, in FIG. 4, the copper plate 3 on the back side is referred to as the heat sink 12, but a circuit structure may be provided on both copper plates 3. If needed, a sloped shape may be provided on the side of the circuit portion 11 or the side of the heat sink 12. If needed, the bonding layer 4 between the ceramic substrate 2 and the circuit portion 11 may have a protruding portion extending beyond the side edge of the circuit portion 11. The bonding layer 4 between the ceramic substrate 2 and the heat sink 12 may have a protruding portion that protrudes beyond the side edge of the heat sink 12. An etching process is preferably used to impart the circuit structure and the inclined shape to the side surface of the copper plate.

[0024] The ceramic circuit board according to the embodiment is suitable for use in a semiconductor device. Fig. 5 is a schematic diagram showing an example of a semiconductor device according to an embodiment. In Fig. 5, 20 is a semiconductor device, 21 is a semiconductor element, and 22 is a lead frame. In the semiconductor device 20 shown in Fig. 5, the semiconductor element 21 is mounted on one of two circuit sections 11. A lead frame 22 is connected to the other of the two circuit sections 11. Note that the semiconductor device 20 according to the embodiment is not limited to this structure. The number and size of the circuit sections 11, the number and size of the semiconductor elements 21, etc. can be changed as needed.

[0025] Next, a method for manufacturing the bonded body according to the embodiment will be described. As long as the bonded body according to the embodiment has the above-described configuration, the manufacturing method is not limited. Below, an example of a method for obtaining the bonded body with a high yield will be described. First, prepare a ceramic substrate 2. Examples of the ceramic substrate 2 include a silicon nitride substrate, an aluminum nitride substrate, an aluminum oxide substrate, and an alu-zirconium substrate. The alu-zirconium substrate is a substrate formed by mixing aluminum oxide and zirconium oxide. A pure copper plate or a copper alloy plate can be used as the copper plate 3. The copper plate is preferably oxygen-free copper. As specified in JIS-H-3100, oxygen-free copper is copper with a purity of 99.96 mass% or more. Next, a brazing filler metal is prepared. The brazing filler metal preferably contains 50 atomic % or more of Cu, 5 atomic % to 40 atomic % of Ti, 2 atomic % to 35 atomic % of the first element, and 0 atomic % to 12 atomic % of carbon. These ratios are based on the total of Cu, Ti, the first element, and C being 100 atomic %. Ti may be added as titanium hydride (TiH2). It is effective not to add Ag to the brazing filler metal.

[0026] The characteristics of the brazing filler metal used in manufacturing the bonded body are preferably examined in advance by DSC. The DSC curve of the brazing filler metal is measured by setting the temperature rise rate in the heating step to 20°C / min and the temperature drop rate in the cooling step to 20°C / min. For DSC, a NETZSCH TGA-DSC simultaneous thermal analyzer STA449-F3-Jupiter or a device with equivalent performance is used. Measurements are performed in an Ar (argon) flow by dropping an appropriate amount of brazing filler metal into an alumina container. Measurements are performed with the lid of the alumina container closed. By measuring in an Ar atmosphere, it is necessary to prevent the brazing filler metal from reacting with the atmosphere. The amount dropped (mg) is measured in advance using a balance. A sample of 15 mg or more is used. In the temperature-raising step, the sample was heated from room temperature to 950°C at a rate of 20°C / min. This was followed by a holding step in which the sample was held at 950°C for 20 minutes. In the temperature-lowering step, the sample was cooled from 950°C to room temperature at a rate of 20°C / min. The temperature profile used to determine the DSC curve consisted of a temperature-raising step, a holding step, and a temperature-lowering step. Hereinafter, the DSC curve obtained in the temperature-lowering step at a rate of 20°C / min will also be referred to as the DSC curve in the temperature-lowering step. Furthermore, the peak was defined as the point where the heat flow differed by 0.02 mW / mg or more. If a first alloy 6 having a freezing point or phase change point between 400°C and 600°C is present, an exothermic peak is detected in the range of 400°C to 600°C on the DSC curve in the cooling step. If multiple exothermic peaks are detected between 400°C and 600°C on the DSC curve in the cooling step, the temperature at which the largest peak appears is taken as the freezing point or phase change point of the bonding layer 4.

[0027] Average particle size D of Cu powder used as raw material for brazing material 50 The average particle size D of the Ti powder or TiH2 powder is preferably 12.0 μm or less, and more preferably 10.0 μm or less. 50 The average particle size D of the Sn powder or In powder is preferably 6.0 μm or less, and more preferably 4.0 μm or less. 50 The average particle size D of the powder C is preferably 16.0 μm or less, and more preferably 14.0 μm or less. 50 The average particle size D of the Ag powder is preferably 6.0 μm or less, and more preferably 4.0 μm or less. 50 The particle size is preferably 3.0 μm or less, and more preferably 2.0 μm or less. By controlling the particle size of the powder, the reaction of each powder can be made uniform. Average particle size of Cu powder D 50is the average particle size D of the Sn or In powder 50 As described above, Cu and the first element form the base material of the brazing filler metal composition. The first element has a lower melting point than Cu. By increasing the particle size of the Sn powder or In powder, the reaction between the Cu powder and the powder of the first element can be made uniform. This allows the ceramic substrate 2 and the copper plate 3 to be sufficiently bonded together even if the temperature rise rate or temperature fall rate is increased.

[0028] The process of joining a ceramic substrate and a copper plate is carried out using the above-described brazing filler metal. The brazing filler metal is mixed with an organic substance to prepare a brazing paste. The brazing paste is applied to the surface of the ceramic substrate 2 (or copper plate 3) to form a brazing paste layer. The copper plate 3 (or ceramic substrate 2) is placed on the brazing paste layer. The thermal bonding process is carried out in a nitrogen atmosphere. A nitrogen atmosphere is an atmosphere containing 70 vol% or more of nitrogen. The nitrogen atmosphere preferably contains 70 vol% or more of nitrogen, and more preferably 85 vol% or more and 100 vol% or less. Components other than nitrogen in the nitrogen atmosphere include an inert gas or air. Thermal bonding in a nitrogen atmosphere can be carried out under various conditions, such as normal pressure, reduced pressure, or increased pressure.

[0029] In the temperature-raising step of the thermal bonding process, the temperature-raising rate is set to 30°C / min or more. There is no particular upper limit to the temperature-raising rate, but it is preferably 100°C / min or less. If the temperature-raising rate is too fast and exceeds 100°C / min, the bondability may be reduced. Therefore, the temperature-raising rate is preferably 30°C / min or more and 100°C / min or less, and more preferably 40°C / min or more and 70°C / min or less. The bonding temperature is set to 750°C or higher. The bonding temperature is preferably 1000°C or lower. If the bonding temperature is higher than 1000°C, the bonding temperature approaches the melting point of copper (1085°C). As a result, the copper plate may be deformed. For this reason, the bonding temperature is preferably 750°C or higher and 1000°C or lower, and more preferably 800°C or higher and 950°C or lower. In the heat bonding process, the bonding temperature should be maintained for 10 minutes or more. The time maintained at the bonding temperature is called the heating holding time. The heating holding time is preferably 10 minutes or more and 100 minutes or less. If the heating holding time is less than 10 minutes, there may not be enough time for the brazing material to melt and solidify. If the heating holding time is longer than 100 minutes, it may be difficult to control the distribution of the alloy in the bonding layer 4. After the heating and holding time is over, a temperature-lowering process is performed. This process involves cooling from the bonding temperature to room temperature. The temperature-lowering rate is set to 30°C / min or more. In conventional active metal bonding methods, the temperature-lowering rate is approximately 5°C / min. By using a brazing filler metal having the DSC curve described above, bonding is possible even if the temperature-lowering rate is increased. The upper limit of the temperature-lowering rate is not particularly limited, but is preferably 100°C / min or less. If the rate is faster than 100°C / min, the bondability may be reduced. For this reason, the temperature-lowering rate is preferably 30°C / min or more and 100°C / min or less, and more preferably 30°C / min or more and 70°C / min or less. Furthermore, a continuous furnace is preferably used to increase the heating and cooling rates. A continuous furnace can perform heat treatment in a nitrogen atmosphere. Since a vacuum is not required, the heating and cooling rates can be increased.

[0030] By increasing the temperature drop rate, it is possible to control the distribution of the Cu-Sn alloy in the bonding layer 4. This is because by quickly lowering the temperature after the bonding brazing material has melted, it is possible to form an alloy that corresponds to the solidification point or phase change point. Furthermore, by allowing Ti that does not form the Ti reaction layer 5 to exist, it is possible to form Ti-Sn alloy and titanium silicide particles. Rapid temperature drop can suppress unwanted diffusion into the copper plate. The difference between the heating rate and the temperature drop rate is preferably 20°C / min or less. In other words, it is preferable to satisfy |heating rate - temperature drop rate|≦20°C / min. In the brazing material layer, heating causes a melting reaction and a solidification reaction. These reactions occur in the heating process and the temperature drop process. By reducing the difference between the heating rate and the temperature drop rate, the stress generated by the reaction can be made uniform. This can suppress defects such as warping. If necessary, the heat bonding step may be carried out while a weight is placed on the bonded body 1. The bonding process described above can produce the bonded body 1. Furthermore, since the thermal stress in the bonding layer 4 can be reduced, warping of the bonded body 1 can be reduced. The resulting bonded body 1 is etched to form a ceramic circuit substrate 10. If necessary, the bonded body 1 may be scribed to form multiple pieces.

[0031] (Example) (Examples 1 to 8, Comparative Example 1) The brazing filler metals were prepared as shown in Tables 1 and 2. Table 1 shows the particle sizes of the raw material powders, and Table 2 shows the composition ratios.

[0032] [Table 1]

[0033] [Table 2]

[0034] The brazing filler metal components of the examples and comparative examples were mixed with an organic binder to prepare brazing filler metal pastes, and the DSC curves of each brazing filler metal paste were measured using a differential scanning calorimeter. The DSC used was a NETZSCH TGA-DSC simultaneous thermal analyzer STA449-F3-Jupiter. Measurements were performed in an Ar flow with an appropriate amount of brazing filler metal dropped into an alumina container. The temperature program was set to a heating rate of 20°C / min, a 20-minute hold at 950°C, and a 20°C / min cooling rate. The temperature at which an exothermic peak was detected in the DSC curve during the cooling step was measured. Table 3 shows the temperature at which the largest exothermic peak appeared between 400°C and 600°C in the DSC curve during the cooling step.

[0035] [Table 3]

[0036] As can be seen from Table 3, for the brazing filler metals according to the examples, the exothermic peak temperatures in the temperature-lowering process of the DSC curves were in the range of 400°C or more and 600°C or less. In contrast, for Comparative Example 1, the exothermic peak temperatures exceeded 600°C. No exothermic peak was detected within the range of 400°C or more and 600°C or less. Furthermore, for the brazing filler metals according to the examples, an exothermic peak was also detected within the range of more than 600°C and 900°C or less. Next, silicon nitride substrates were prepared as ceramic substrates. Silicon nitride substrates with a thermal conductivity of 90 W / m K, a three-point bending strength of 600 MPa, and dimensions of 300 mm x 200 mm x 0.32 mm were designated as silicon nitride substrate 1. Silicon nitride substrates with a thermal conductivity of 85 W / m K, a three-point bending strength of 650 MPa, and dimensions of 300 mm x 200 mm x 0.25 mm were designated as silicon nitride substrate 2. Oxygen-free copper measuring 300 mm long x 200 mm wide x 0.5 mm thick was used as copper plate 1. Oxygen-free copper measuring 300 mm long x 200 mm wide x 0.8 mm thick was used as copper plate 2. Brazing paste was applied to both sides of the ceramic substrate in a thickness of 30 μm, and a copper plate was placed on each side. Next, a thermal bonding process was carried out. The bonding atmosphere was uniformly a nitrogen atmosphere. The bonding conditions were 850°C to 920°C for 30 to 50 minutes, and a continuous furnace was used. The temperature rise rate and temperature fall rate were set as shown in Table 4. The difference between the temperature rise rate and temperature fall rate was calculated by |Temperature rise rate - Temperature fall rate|.

[0037] [Table 4]

[0038] Bonded bodies were manufactured using the above process. The cross-sections of the bonded layers of the obtained bonded bodies were observed to check for the presence or absence of Cu-Sn alloys with different composition ratios, the average grain size of titanium nitride in the Ti reaction layer, and the amounts of Cu and Sn in the Ti reaction layer. Furthermore, when Cu-Sn alloys with different composition ratios were present, the presence or absence of a first Cu-Sn alloy with an Sn content of 3 atomic % to 9 atomic % and a second Cu-Sn alloy with an Sn content of 10 atomic % to 30 atomic % were checked. The area ratios of the Cu-Sn alloy and the Ti-Sn alloy were also investigated. The major axis and area ratio of titanium silicide were also investigated. The area ratios were calculated by binarizing the EDX elemental map using image software. The detailed method is as described above. The results are shown in Tables 5 to 8.

[0039] [Table 5]

[0040] [Table 6]

[0041] [Table 7]

[0042] [Table 8]

[0043] As can be seen from Tables 5-8, the joints of the examples met the desirable conditions. The Cu-Sn alloy was the main component of the joint layer. It was also observed that the first Cu-Sn alloy and the second Cu-Sn alloy each formed an intermetallic compound. EDX point analysis was performed on 10 randomly selected points in the Cu-Sn alloy region. EDX point analysis was also performed on 10 randomly selected points in the Ti-Sn alloy region. The results are shown in the ternary phase diagram in Figure 3, and all of the results were within the first composition region. Furthermore, the EDX point analysis of the Cu-Sn alloy showed that the composition of 80% to 100% of the Cu-Sn alloy was within the second composition region. The EDX point analysis of 80% to 100% of the Ti-Sn alloy was also performed on the third composition region. The "Other" area ratio in Table 8 is the area where the Ti reaction layer, carbide, or metal component was dissolved and separated as a single element. The thickness of the Ti reaction layer in the joined body according to the example was 1 μm or less. In addition, both TiN and Ti 2 N were observed in the Ti reaction layer. In contrast, no Cu-Sn alloy with a different composition ratio was observed in Comparative Example 1. Although a small amount of Cu-Sn alloy was observed in Comparative Example 1, no Cu-Sn alloy with a different composition was observed. This is thought to be because the bonding layer in Comparative Example 1 was mainly composed of AgCu eutectic or AgSn. Next, 100 bonded bodies were produced for each of the examples and comparative examples, and the bond strength and warpage of each were measured. The bonding strength was measured by a peel test. Specifically, a peel test sample was prepared for each example and comparative example. The sample was prepared by bonding a rectangular copper plate to a ceramic substrate. At this time, the copper plate was bonded so that one end of the copper plate protruded from the ceramic substrate. The protruding copper plate was pulled vertically to measure the peel strength. The amount of warpage was measured on the long side of the bonded piece. Bonded pieces with 90 or more sheets and a warpage of 0.3 mm or less were rated "best." Bonded pieces with 70 to 89 sheets and a warpage of 0.3 mm or less were rated "good." Bonded pieces with 31 or more sheets and a warpage of more than 0.3 mm were rated "poor." The results are shown in Table 9.

[0044] [Table 9]

[0045] As can be seen from Table 9, the joining strength of the joined bodies according to the example was equivalent to that of the joined body according to Comparative Example 1, which used an Ag-containing brazing filler metal. Even when the size of the joined body was increased to 200 mm or more, the amount of warpage was suppressed. Even when the heating rate and the cooling rate were increased, the amount of warpage was suppressed. This demonstrates that the joined bodies are suitable for mass production. In contrast, in Comparative Example 1, many joined bodies had warpage exceeding 0.3 mm. This is because a Cu-Sn alloy with a different composition ratio was not formed.

[0046] The amount of Sn in the copper plate of the bonded bodies of the examples and comparative examples was investigated. The amount of Sn was measured by analyzing the cross section of the bonded body using EDX. The boundary between the bonding layer and the copper plate was defined as described above. The amount of Sn at the boundary between the bonding layer and the copper plate was defined as Sn-1, the amount of Sn at a location 10 μm away from the boundary as Sn-2, and the amount of Sn at a location 20 μm away from the boundary as Sn-3. Sn-2 / Sn-1 and Sn-3 / Sn-1 were calculated. We also checked whether the farthest point from the boundary in the area where Sn was detected was within 50 μm. Cases where it was within 50 μm were marked "Good", and cases where it was over 50 μm were marked "Poor". The presence or absence of voids in the bonded layer was also investigated. The presence or absence of voids was checked using ultrasonic flaw detection (SAT). A void volume ratio of 0% to 0.3% was marked with a "◎", 0.4% to 1% was marked with a "〇", 2% to 3% was marked with a "△", and 4% or more was marked with an "×". The results are shown in Table 10.

[0047] [Table 10]

[0048] As can be seen from Tables 9 and 10, in the bonded bodies according to the examples, it was found that the diffusion of Sn into the copper plate was suppressed. It was also found that the void area was small. In particular, in Examples 6 to 8, in which the void area was reduced, the bonding strength was high at 25 kN / mm or more. In contrast, in the comparative example, the amount of Sn diffused into the copper plate was large, and the void area ratio was also large.

[0049] Although several embodiments of the present invention have been described above, these embodiments are presented by way of example only and are not intended to limit the scope of the invention. These novel embodiments can be embodied in various other forms, and various omissions, substitutions, modifications, etc. can be made without departing from the spirit of the invention. Modifications of these embodiments are included within the scope and spirit of the invention, as well as within the scope of the invention and its equivalents as set forth in the claims. Furthermore, the above-described embodiments can be implemented in combination with each other. [Explanation of symbols]

[0050] 1...Zygote 2...Ceramic substrate 3...Copper plate 4...Joining layer 5...Ti reaction layer 6...Cu-Sn alloy or Cu-In alloy 7...Ti-Sn alloy or Ti-In alloy 8...Titanium silicide particles 10...Ceramic circuit board 11...Circuit section 12...Heat sink

Claims

1. a ceramic substrate; Copper plate and a bonding layer disposed on at least one surface of the ceramic substrate and bonding the ceramic substrate and the copper plate; The bonding layer is a Ti reaction layer containing titanium nitride or titanium oxide as a main component; a plurality of first alloys located between the Ti reaction layer and the copper plate, each containing one or more selected from a Cu—Sn alloy and a Cu—In alloy, and having different Sn concentrations or In concentrations; A zygote comprising:

2. The bonding layer further includes a second alloy located between the Ti reaction layer and the copper plate, the second alloy contains one or more selected from a Ti—Sn alloy and a Ti—In alloy; 2. The bonded body according to claim 1, wherein the total area of ​​the plurality of first alloys is larger than the total area of ​​the second alloys in an average value obtained by analyzing three regions each having a dimension of the thickness of the bonding layer times 40 μm in the width direction.

3. At least some of the first alloys are Cu—Sn alloys; the second alloy is a Ti—Sn alloy; 3. The bonded body of claim 2, wherein the at least some of the first alloys and the second alloy detected by EDX point analysis of the bonding layer are within the region surrounded by (97, 3, 0), (60, 40, 0), (2, 40, 58), and (39, 3, 58) in a Cu, Sn, Ti ternary phase diagram.

4. At least some of the first alloys are Cu—Sn alloys; 4. The joined body according to claim 1, wherein at least a portion of the plurality of first alloys detected by EDX point analysis of the joining layer are within a region surrounded by (97, 3, 0), (77, 23, 0), (52, 23, 25), and (72, 3, 25) in a ternary phase diagram of Cu, Sn, and Ti.

5. the second alloy is a Ti—Sn alloy; 4. The joined body according to claim 2, wherein the second alloy detected by EDX point analysis of the joining layer is within the region surrounded by (41, 23, 36), (24, 40, 36), (2, 40, 58), and (19, 23, 58) in a ternary phase diagram of Cu, Sn, and Ti.

6. 6. The joined body according to claim 1, wherein the Ti content in at least one of the plurality of first alloys is 1 atomic % or less.

7. In some of the plurality of first alloys, the Sn content is 3 atomic % or more and 9 atomic % or less, 7. The joined body according to claim 1, wherein the Sn content in another part of the plurality of first alloys is 10 atomic % or more and 30 atomic % or less.

8. 8. The joined body according to claim 1, wherein the solidification point or phase change point of one or more of the plurality of first alloys is 400°C or higher and 600°C or lower.

9. the Ti reaction layer contains titanium nitride particles having an average particle size of 50 nm or less, 9. The joined body according to claim 1, wherein a first element containing Cu and one or two selected from Sn and In is present at the grain boundaries of the titanium nitride particles.

10. 10. The bonded body according to claim 1, wherein the total area of ​​the plurality of first alloys is 50% or more and 95% or less of the area of ​​the three regions, in an average value obtained by analyzing three regions each having a thickness of the bonding layer and a width of 40 μm.

11. 11. The bonded body according to claim 1, wherein the bonding layer contains titanium silicide particles having a major axis of 0.5 μm or more and 6 μm or less.

12. the bonding layer includes a plurality of titanium silicide particles having a major axis of 0.5 μm or more and 6 μm or less, 12. The bonded body according to claim 1, wherein an area ratio of the plurality of titanium silicide particles in a region of the bonding layer having a thickness and a width of 200 μm is 1% or more and 15% or less.

13. 13. The bonded body according to claim 1, wherein the bonding layer further contains carbon.

14. 14. The bonded body according to claim 1, wherein the thickness of the Ti reaction layer is 1 μm or less.

15. a first element including one or two selected from Sn and In is present at the boundary between the bonding layer and the copper plate; 15. The joined body according to claim 1, wherein a ratio C2 / C1 of a first concentration C1 of the first element at the boundary to a second concentration C2 of the first element in a region in the copper plate 10 μm away from the boundary is 0.60 or less.

16. A ceramic circuit board comprising the bonded body according to any one of claims 1 to 15.

17. The ceramic circuit substrate according to claim 16; a semiconductor element mounted on the ceramic circuit board; A semiconductor device comprising:

Images