Composite material, heat dissipation substrate, and semiconductor device

Abstract

The present invention comprises a heat dissipation substrate which is in a plate shape and is formed of a composite material including metal and silicon carbide. The surface of the heat dissipation substrate on one side is partitioned into a plurality of regions in a grid shape. Each of a length B1 of a lengthwise side and a length L1 of a widthwise side of each of the plurality of regions is 10 mm-25 mm. The plurality of regions 2 in the grid shape occupy 70% or more of the total area of the surface of the heat dissipation substrate 1 on the one side. Each point of the plurality of regions in the grid shape has heat conductivity of 140 W / (m·K) or higher. A variation coefficient of the heat conductivity in the plurality of regions in the grid shape is less than 3.0%. The variation coefficient is expressed by standard deviation / average×100 (%).

Description

Composite material, heat dissipation substrate and semiconductor device

[0001] The present disclosure relates to a magnesium-based composite material in which magnesium or a magnesium alloy is combined with silicon carbide, a heat dissipation substrate formed from this composite material, and a semiconductor device including this heat dissipation substrate. This application claims priority to Japanese Patent Application No. 2022-133332, filed on August 24, 2022. The entire contents of this Japanese patent application are incorporated herein by reference.

[0002] Conventionally, composite materials have been disclosed in, for example, JP 2006-299304 A (Patent Document 1), JP 2010-090436 A (Patent Document 2), JP 2010-106362 A (Patent Document 3), JP 2018-105297 A (Patent Document 4), JP 2011-073055 A (Patent Document 5), JP 2016-007634 A (Patent Document 6), and JP 2011-73950 A (Patent Document 7).

[0003] Japanese Patent Application Laid-Open No. 2006-299304 Japanese Patent Application Laid-Open No. 2010-090436 Japanese Patent Application Laid-Open No. 2010-106362 Republished No. 2018-105297 Japanese Patent Application Laid-Open No. 2011-073055 Japanese Patent Application Laid-Open No. 2016-007634 Japanese Patent Application Laid-Open No. 2011-073950

[0004] "Control and Measurement Evaluation Methods of Thermal Conductivity and Thermal Diffusivity / Science & Technology Publication" Chapter 3, Section 1, Standards for Thermal Conductivity and Thermal Diffusivity, by Megumi Akoshima, National Institute of Advanced Industrial Science and Technology

[0005] The composite material of the present disclosure has a plate-shaped heat dissipation substrate made of a composite material containing magnesium or a magnesium alloy and silicon carbide, one side surface of the heat dissipation substrate is divided into multiple lattice-like regions, each of which has vertical and horizontal side lengths of 10 mm or more and 25 mm or less, the multiple lattice-like regions occupying 70% or more of the total area of ​​one side surface of the heat dissipation substrate, the thermal conductivity at each point of the multiple lattice-like regions being 140 W / (m·K) or more, and the coefficient of variation of the thermal conductivity within the multiple lattice-like regions being less than 3.0%, where the coefficient of variation is expressed as standard deviation / average value×100(%).

[0006] Fig. 1 shows a heat dissipation substrate 1 and a sample 4 cut out from a region surrounded by a plurality of lattice-shaped regions 2 and four intersections 3, which are measurement sites. Fig. 2 is a plan view of heat dissipation substrates 1 of sample numbers 8 and 9 and semiconductor elements 11 to 16 bonded thereon. Fig. 3 is a cross-sectional view of a semiconductor device 10 having a heat dissipation substrate 1, taken along the line indicated by the arrow III-III in Fig. 2.

[0007] [Problem to be Solved by the Present Disclosure] Conventional composite materials have a problem in that the thermal conductivity varies greatly.

[0008] [Description of Embodiments of the Present Disclosure] First, embodiments of the present disclosure will be listed and described.

[0009] (1) Scattering of the characteristics of heat dissipation substrates When multiple test pieces are extracted from each part of a heat dissipation substrate, which is a composite material (generally a rectangular flat plate with a thickness of 4 to 5 mm), and the thermal conductivity is measured, scattering of the thermal conductivity is found.

[0010] Heat dissipation substrates used in semiconductor devices are used to efficiently transfer and transmit the heat generated by semiconductor elements. The numerical value that determines the performance of a heat dissipation substrate is its thermal conductivity (unit: W / (m·K)), which is an index of heat transfer. For materials with the same components, volume, and mass, the higher the thermal conductivity value, the easier it is to transfer heat. The thermal conductivity of a heat dissipation substrate is measured using a small test piece cut out from the heat dissipation substrate.

[0011] When measuring the thermal conductivity of a plate-shaped heat dissipation substrate, a representative value of thermal conductivity may be obtained by measuring a test piece cut from an arbitrarily selected location, such as a representative central location or a representative peripheral location. Alternatively, test pieces may be cut from several predetermined locations and measured, and the average value may be used as the representative value. For example, values ​​may be obtained from a test piece cut from the central region and several peripheral locations, totaling two to six test pieces, and the average may be used as the representative value. This method is used to express the thermal conductivity of an individual heat dissipation substrate.

[0012] However, conventional composite materials made of metal and ceramics, and heat dissipation substrates using such materials, have the problem that even if the composition, such as the volume ratio and porosity of the composite metal and silicon carbide, of a single heat dissipation substrate is uniform, or even if the mechanical properties, such as strength and elastic modulus, are uniform, when the thermal conductivity measured at multiple locations is statistically compared, there is a large dispersion of properties, such as variation and distribution within the plate.

[0013] This characteristic scattering of thermal conductivity exists, and a heat dissipation substrate with a large rate of scattering poses a problem when used in a semiconductor device.

[0014] Thermal conductivity is expressed as in Equation 1. λ = α Cp ρ Equation 1 λ: Thermal conductivity [W / (m K)] α: Thermal diffusivity (m ２ / s) Cp: specific heat capacity [J / (kg K)] ρ: bulk density of sample (kg / m ３ ) Here, the value obtained by multiplying the specific heat capacity and density of a solid corresponds to "A; heat capacity per unit volume (volumetric specific heat)", and can be seen to be as shown in Equation 2.

[0015] A=Cp・ρ [J / (m ３ ·K)] Equation 2 When the overall composition and compounding ratio of the composite material is constant, the A value is a nearly constant value and can be expressed as follows:

[0016] λ=α·A Equation 3 Therefore, from equation 3, thermal conductivity can be seen as the value obtained by multiplying thermal diffusivity by a constant, and scattering of thermal conductivity is caused by scattering of thermal diffusivity.

[0017] The inventors conducted experiments and research over a wide area range to measure the thermal diffusivity of heat dissipation substrates, and discovered that the scattering of the thermal conductivity of the heat dissipation substrate is caused by the scattering of the thermal diffusivity. They then discovered a manufacturing method to suppress this scattering, and discovered a heat dissipation substrate made of a composite material with good heat dissipation properties.

[0018] A large number of measurement points were extracted from a wide area, and the coefficient of variation calculated from the standard deviation and average value was used as an index for comparing the degree of scattering of all values, and the values ​​were made dimensionless and statistically analyzed.

[0019] If verification is performed using a range that represents the difference between the maximum and minimum values, or the difference between the measured value and the average value, it is not possible to accurately compare the degree of scattering when comparing products with materials that have significantly different thermal conductivities.

[0020] The coefficient of variation is expressed as standard deviation / average value x 100 (%). The coefficient of variation of the volumetric specific heat A obtained by multiplying the density ρ and the specific heat capacity Cp is sufficiently smaller than the thermal diffusivity α, which indicates that there is little scattering between the density ρ and the specific heat capacity Cp, i.e., there is little variation.

[0021] In a dense composite material with a porosity of less than 3% by volume, small density dispersion is a necessary and sufficient element for the shape-stable heat dissipation component of Patent Document 4. In other words, a heat dissipation substrate with a stable shape can be provided by using a composite material with a dense and uniform structure and density.

[0022] (2) Manufacturing method of metal-ceramic composite materials Composite materials of metal and ceramic are used for heat dissipation substrates for semiconductor devices called power modules. Examples include composites of silicon carbide with metals such as pure aluminum, aluminum alloys, pure magnesium, magnesium alloys, copper, and copper alloys. This "metal-ceramic composite material" is manufactured by the so-called "infiltration method," in which a metal that has been heated to a temperature above its melting point is poured into an aggregate of ceramic particles, and then solidified and molded. It is also called the metal melting method or the impregnation method.

[0023] The infiltration method is selected from pressureless infiltration methods such as gravity casting or spontaneous infiltration, die casting, low-pressure casting, and pressure infiltration methods such as melt forging. These infiltration methods generally have the following characteristics:

[0024] (2-1) Infiltration process This is a method in which high-temperature molten metal is infiltrated into the gaps between ceramic aggregates, so it is impossible for the entire ceramic to be immersed in metal instantaneously (or at high speed). There are entrance points through which the molten metal preferentially enters, and the molten metal infiltrates and moves inside, spreading throughout the entire ceramic, creating a so-called flow of molten metal, which is an unavoidable manufacturing condition.

[0025] (2-2) Internal behavior during infiltration When molten metal infiltrates non-metallic ceramics, it incorporates impurities present on the ceramic surface into the molten metal.

[0026] These impurities are present on the surface of ceramics (including, for example, oxide films). They may also be intentionally placed in advance to form an alloy layer between the impurities and the molten metal. For example, an oxide film is formed as an impurity by heating in the atmosphere. Or, impurities may be intentionally added as an infiltrant (infiltration aid) to improve the contact wettability between the molten metal and non-metal.

[0027] These impurities are taken in and dissolved when they come into contact with the molten metal, and those that remain as solid matter even after the molten metal cools and solidifies are not selected.

[0028] When molten metal is infiltrated, some reactive gas may be generated, and if it remains and solidifies, it will cause defects (blowholes). These defects do not necessarily occur all over the area, but may occur locally due to the flow of the molten metal.

[0029] The pressure infiltration method can suppress defects by applying pressure even when the wettability between metal and ceramic is poor.

[0030] (2-3) Solidification process During the solidification process of molten metal, the volume of the metal shrinks, which makes it easy for localized voids, so-called shrinkage cavities, to occur. In particular, because there is a time lag in how the metal cools and solidifies, shrinkage cavities occur in the solidification areas that cool and solidify last.

[0031] Generally, when molten metal in a container solidifies, it cools and solidifies from the periphery, and the center, which has the highest temperature, cools last, so shrinkage cavities are usually more likely to occur in the center of the container.

[0032] To prevent this, temperature is directional during solidification to prevent shrinkage cavities from remaining inside the product.

[0033] For example, a temperature difference can be created between the top and bottom or the periphery of the container to allow for preferential solidification on one side. For example, the container can be exposed to a low-temperature atmosphere or brought into contact with a low-temperature object to allow for preferential solidification. Another example is to keep the area where cooling is to be delayed warm.

[0034] However, if the cooling is too fast, gas will be entrapped due to solidification shrinkage, so it is essential to control the cooling rate.

[0035] This directional cooling method can also be used to reduce blowholes caused by generated gases, as described above.

[0036] Furthermore, during solidification, the impurities that have been incorporated into the molten metal may precipitate during the solidification process.

[0037] These precipitates do not occur uniformly throughout the material, but occur unevenly depending on conditions such as the temperature during solidification.

[0038] (3) Main Causes of Scattering in Characteristics Due to Manufacturing Methods As explained above, when combining metal and ceramics using the infiltration method, linear flow occurs in the movement and solidification of the molten metal, which causes scattering of the thermal conductivity. In particular, when impurities are dissolved in the molten metal, local component separation occurs, and when the molten metal solidifies, crystals precipitate non-uniformly. Also, there are microscopic defects and internal structure distributions that cannot be detected with a microscope, such as those that exist in the gaps between the metal and ceramic after solidification. These are the causes of scattering of the thermal diffusivity.

[0039] Minute defects are voids so small that they do not affect density and cannot be detected by non-destructive testing such as ultrasonic testing or X-ray inspection. Naturally, defects large enough to be detected as obvious defects by these tests will not only reduce thermal conductivity but also other properties such as strength, making them unsuitable for use as a heat dissipation substrate.

[0040] It should be noted that non-metallic ceramic particles do not react with metals, resulting in particle destruction or damage.It has been found that composite materials made by infiltration have uneven contact between the metal and non-metal, causing scattering of thermal diffusivity.

[0041] (4) Reference comparison with other manufacturing methods For comparison, composite materials made by sintering powder, so-called sintering method (e.g., A.L.M.T.'s sintered Al-SiC), are made by pressing raw material powder into a wide area of ​​a plate shape and then sintering at a temperature lower than the melting point of the metal, so variations are unlikely to occur. However, in order to obtain a higher thermal conductivity, it is necessary to make a large-area product, for example, 50 cm ２ For areas above this, a high-pressure molding device is required, making it unsuitable for industrial mass production.

[0042] In power modules equipped with semiconductors that operate at high temperatures, large variations in characteristics such as distribution and variation in thermal conductivity create resistance to the flow of heat from the semiconductors.

[0043] (5) Effect of the Present Disclosure Since a power module typically includes circuits using multiple semiconductor elements arranged side by side, a local decrease in thermal conductivity reduces heat dissipation, shortening the lifespan of the semiconductors.

[0044] Patent Document 4 states that this stabilizes the shape and prevents a decrease in thermal resistance. Power modules used in automobiles, industrial equipment, power generation, railways, communication equipment, etc. are equipped with semiconductor elements (also called semiconductor devices) such as IGBTs (Insulated Gate Bipolar Transistors) and MOSFETs (Metal Oxide Semiconductor Field Effect Transistors), and are used for power conversion and equipment control to control current and voltage.

[0045] The inside of a power module is a structure in which circuit boards made up of multiple semiconductor elements are mounted side by side, and as the semiconductor elements repeatedly turn on and off power at high speed, heat is generated corresponding to the resistance components in the circuit.The heat dissipation substrate is required to efficiently transfer and transmit the heat generated by the semiconductor elements, and to be heat-resistant so that it does not deform due to the thermal history during the assembly of the power module or the thermal cycle during use.

[0046] Furthermore, in recent years, due to the trend toward higher voltages, larger currents, and smaller sizes, there has been a demand for the ability to release heat generated from semiconductor elements, that is, higher heat dissipation performance.

[0047] When designing the thermal performance of a power module, the flow of heat from high to low temperatures is simulated using the thermal conductivity of the semiconductor and other components, and then actual measurements are taken to calculate and verify the structure function. If there are areas where the thermal conductivity within the plate varies significantly or drops significantly, the thermal design must take this into account, and a high safety factor must be designed.

[0048] Power modules are designed to operate at a temperature below a set junction temperature of the semiconductor element. If the element exceeds the junction temperature, a high current will flow, causing the temperature to rise and resulting in thermal runaway, which will destroy the device.

[0049] As temperatures rise, phenomena such as peeling of joints due to temperature changes, deterioration of insulation performance, and fatigue failure of materials occur, resulting in a shorter lifespan. For this reason, power modules require a thermal management design that can dissipate heat more efficiently.

[0050] The performance required for a heat dissipation substrate naturally requires a material with higher thermal conductivity, and it is also required that the material have lower thermal resistance.

[0051] The heat dissipation member of the present disclosure, which has little scattering of characteristics such as distribution and variation in thermal conductivity, enables a thermal management design with reduced thermal resistance.

[0052] A power module is made by arranging multiple semiconductor elements, which are heat-generating elements, on a heat dissipation substrate and joining them with solder.

[0053] In this case, the heat generated by the adjacent semiconductor elements may interfere with each other, resulting in an overall temperature higher than when a single semiconductor element is used. This is called thermal interference.

[0054] If a semiconductor element is placed close to a location in the heat dissipation substrate where the thermal conductivity is highly dispersed, the thermal resistance will increase further, causing thermal interference and making it more likely that the temperature will rise.

[0055] Such thermal resistance is lower for a plate with low thermal conductivity scattering than for a conventional plate with high thermal conductivity scattering.

[0056] For example, when comparing sample (a): average 170 W / (m·K) with sample (b): average 210 W / (m·K), if both have an equally wide range of variation in scattering, the difference in thermal resistance between sample (a) and sample (b) will be difficult to see. However, if the range of variation in the thermal conductivity of both is reduced to suppress scattering, sample (b) has an even smaller thermal resistance with an average of 210 W / (m·K), and sample (b) has better heat dissipation properties than sample (a).

[0057] The thermal conductivity of a heat dissipation substrate made of a composite material is measured by cutting out a sample piece, which will be described later, and performing a so-called destructive test.

[0058] Therefore, when using a heat dissipation substrate with a large characteristic dispersion of thermal conductivity, designers of semiconductor devices such as power devices must design in advance with a margin for heat dissipation characteristics from the heat generating element in order to avoid this.

[0059] Alternatively, it may become necessary to design with a high safety factor in mind, such as by increasing the distance between heat-generating semiconductors.

[0060] In the case of a heat dissipation substrate with small characteristic scattering of thermal conductivity, heat generation due to mutual interference can be suppressed, close arrangement design can be easily performed, and reliability can be ensured.

[0061] Furthermore, when a material with higher thermal conductivity is used, the effect of reducing thermal resistance is significantly achieved by using a heat dissipation substrate with less characteristic scattering.

[0062] This can be explained for the same reason as in the comparison of the plates of sample (a): average 170 W / (m·K) and sample (b): average 210 W / (m·K). Thermal resistance is defined as follows:

[0063] Thermal resistance Rth = thickness L / (cross-sectional area × thermal conductivity), L: distance through which heat flows (6) Measurement of thermal conductivity For materials with high thermal conductivity, the thermal conductivity is measured by the laser flash method (LF), among various methods available.

[0064] The thermal conductivity of the composite material of the present disclosure was measured by the following method and calculated using Equation 1 or Equation 3.

[0065] The thermal diffusivity α was measured by the laser flash method (LF) in accordance with JIS R 1611 (2010) and ASTM E 1461.

[0066] The specific heat capacity Cp was measured by differential scanning calorimetry (DSC).

[0067] The bulk density ρ was calculated by dividing the mass of the sample by the external volume. From these, the thermal conductivity λ was calculated.

[0068] The principle of measuring the thermal diffusivity α by the laser flash (LF) method is to apply pulsed light to a sample cut out to a certain area to heat it, and then measure and calculate the thermal change, which is a so-called unsteady method.

[0069] In contrast, the steady-state method is a method in which measurements are made by applying a heat flux, and is not suitable for materials with high thermal conductivity such as heat dissipation members used in semiconductor devices.

[0070] The specific heat capacity Cp can be measured by differential scanning calorimetry or adiabatic calorimetry in addition to the DSC method. Alternatively, literature values ​​may be used.

[0071] The density ρ is determined by the bulk density calculated from the mass and external volume, or by the apparent density calculated by the Archimedes method using underwater weighing.

[0072] This is because measurement errors when measuring these do not affect the characteristic scattering in the composite materials of the present disclosure having a porosity of less than 3% by volume.

[0073] Measurements were performed on test pieces cut out to a diameter of 10 mm and a thickness of 1 to 5 mm. Thermal diffusivity was measured using an LFA447 manufactured by NETSZH.

[0074] The specific heat capacity was measured using a DSC3500 manufactured by NETSZH. (6-2) Measurement method to reduce measurement errors The thermal diffusivity is measured by the laser flash method, in which the surface of a flat sample is instantaneously heated with a spatially uniform pulsed light, and the temperature rise on the back surface of the sample is observed.

[0075] Measurement errors with the laser flash method are mainly due to errors in the surface temperature during measurement, and are caused by the surface condition (roughness), thickness, and parallelism of the test piece, so measurements and comparisons are made using test pieces that are standardized in these respects.

[0076] Generally, the average value is calculated from multiple measurements, but to eliminate measurement error factors, the average value of three or more measurements on the same sample is used, and if the difference between the maximum and minimum values ​​is 8% or more of the average value (a range of ±4% or more from the average value), the measurement value is not adopted. In such cases, factors such as sample error (surface condition, thickness, parallelism) and changes or deterioration of the measuring instrument are eliminated, corrected, and calibrated, and the value measured again using the same method is adopted.

[0077] (7) Improvements Made by the Present Disclosure This was achieved by improving the affinity (or wettability) between the ceramic particles and the molten metal during the infiltration and solidification processes, thereby stabilizing the solid solution segregation of impurities in the molten metal.

[0078] (8) Product composition (8-1) Metal Pure magnesium and magnesium alloys can be used.

[0079] In the case of pure metal, the fewer impurities the better, and the magnesium content is preferably 99.5 mass % or more, more preferably 99.9 mass % or more.

[0080] Elements are added to magnesium alloys to impart corrosion resistance, strength, and safety against combustion, and examples of the added elements include one or more elements selected from Al, Si, Ca, Zn, Li, Cu, Ti, Y, and Zr.

[0081] (8-2) Regarding ceramics, silicon carbide is applicable.

[0082] The purity of silicon carbide is preferably 95 mass % or more of silicon carbide components, more preferably 99 mass % or more.

[0083] The silicon carbide powder is classified by sieving with a standard mesh size based on JIS Z8801, and the particle size range is preferably 1 to 330 μm.

[0084] More preferably, the particle size is in the range of 5 to 150 μm, which is the smallest particle size range that is easy to handle, taking into consideration the possibility of powder scattering that could cause environmental and health problems, and the largest particle size range that is easy to process, taking into consideration factors such as processing resistance and particle shedding during subsequent cutting and polishing processes.

[0085] Fine particles and ultra-fine powders less than 1 μm in size are particularly difficult to handle. Wire-cut electric discharge machines, which will be described later, generally use electrode wires with a wire diameter of 0.2 to 0.3 mm. Therefore, if the particle diameter is significantly larger than the wire diameter, cutting resistance will occur and the material will not be able to be cut.

[0086] For the heat dissipation substrate of the present disclosure with little characteristic scattering of thermal conductivity, the particle size of silicon carbide does not matter and does not need to be specified.

[0087] The average value of the thermal conductivity of the entire heat dissipation substrate is determined by the volume fraction of the metal and silicon carbide.

[0088] That is, since the metal fills the gaps between the particles of the silicon carbide powder aggregate, a plate with the desired thermal conductivity can be obtained by determining the void ratio of the silicon carbide aggregate.

[0089] That is, to determine the void ratio of the aggregate, powder of a single particle size is not necessarily required, but silicon carbide powder having a particle size distribution, or a mixture of silicon carbide powders having a plurality of particle size distributions, may be used.

[0090] For example, by mixing two or more types of non-metallic powders with different average particle sizes, the void ratio becomes smaller and a plate with higher thermal conductivity can be produced.

[0091] However, industrially good quality can be maintained by manufacturing, managing, and manufacturing powders having a particle size distribution based on the average particle size, which is within the range sieved using a standard sieve specified in JIS Z8801.

[0092] (9) Example of a method for producing the magnesium-silicon carbide composite material of the present disclosure (9-1) Step of preparing silicon carbide The raw material silicon carbide powder is the silicon carbide described in item (8-2) “Regarding ceramics.” This silicon carbide powder is subjected to plasma treatment.

[0093] Plasma is broadly classified into high-temperature plasma (also called thermal plasma) and low-temperature plasma. High-temperature plasma is an ionized gas state at a temperature of several thousand degrees Celsius or higher. The plasma treatment disclosed herein uses low-temperature plasma. This plasma treatment can modify the surface of a material and remove organic matter using plasma-converted gas. The affinity (also called wettability) between silicon carbide particles and molten metals such as magnesium is improved, and solid solution segregation of impurities in the molten metal is suppressed.

[0094] Plasma treatment is generally carried out using oxygen, as well as fluorine, hydrogen, nitrogen, helium, etc. Any of these methods can be applied to the present disclosure, but oxygen plasma is most effective for silicon carbide particles.

[0095] The treatment method used is either vacuum plasma or atmospheric pressure plasma. The vacuum plasma method treats by introducing a small amount of radical gas under vacuum.

[0096] Furthermore, a similar effect can be obtained by irradiating a radicalized gas with atmospheric pressure plasma in an environment that is not reduced in pressure.

[0097] The difference between them is that vacuum plasma has limitations on the capacity to maintain the vacuum, making it difficult to put into production. Atmospheric pressure plasma does not have these limitations, but does require exhaust equipment to remove the ozone that is generated. Despite these differences, the effects are similar.

[0098] In plasma processing, the raw material powder is placed in a container made of glass, ceramics, metal, or the like that can withstand temperatures of around 100°C, and there are no particular restrictions on the shape or size of the container.

[0099] The plasma treatment is carried out by placing the object to be treated in a plasma treatment device, reducing the pressure, introducing a small amount of oxygen gas under a reduced pressure of 100 Pa or less, and applying high frequency power.

[0100] The plasma treatment can be performed using commercially available plasma treatment equipment from various companies. In this disclosure, the equipment used in this study is a plasma reactor manufactured by Yamato Scientific.

[0101] With the chamber of the device depressurized by a vacuum pump, oxygen gas is pumped from a cylinder at a rate of 50 to 100 cm per minute. ３ A power of 200 W is applied to the electrodes, and the treatment can be carried out for 10 minutes.

[0102] (9-2) Adjustment of Impurities Products, powders, binders, etc. of the following methods can be uniformly dispersed in the non-metallic aggregate after plasma treatment as "impurities." These impurities do not necessarily have to be present.

[0103] For example, there is a method of forming an oxide on the surface of silicon carbide powder, in which the silicon carbide powder is heated to 700 to 1100° C. in the atmosphere.

[0104] One method is to mix powdered oxides, such as aluminum oxide and silicon oxide, into the non-metallic powder.

[0105] One method is to mix silicon carbide powder with a binder and mold it into a molded body (also called a preform) in advance, or a combination of these methods may be used.

[0106] These impurities may be oxide films present in the silicon carbide powder or oxides intentionally added as infiltration aids to improve wettability.

[0107] (9-3) Step of placing non-metallic powder or a compact solidified from non-metallic powder in a mold in advance To place silicon carbide powder in a mold, there is a method of applying mechanical energy to the powder, such as vibrations such as shock waves or sine waves, or impacts such as tapping using a drop, to force it in. Alternatively, there is a method of placing the powder in the mold and then performing press molding. Either of these methods can be selected.

[0108] The method for solidifying the powder of the molded body can be selected from among pressure molding, slurry casting, slip casting, doctor blade, and the like.

[0109] The method for solidifying the molded body may involve molding silicon carbide powder with a binder, solidifying it, and then heating and firing it as necessary to make it possible to maintain its shape, and then placing it in a mold. For example, there is a method in which silicon carbide powder is mixed with liquid silicon oxide, molded, and then fired at 600°C or higher.

[0110] Silicon carbide powder may be heated and sintered at or above its recrystallization temperature and then placed in a molding die. For example, SiC powder may be mixed with liquid silicon oxide, molded, and then sintered at 2000°C or higher.

[0111] (9-4) Step of Composite Metal with Nonmetal: A metal is heated above its melting point and infiltrated into silicon carbide powder to obtain a composite. The manufacturing method is selected from among infiltration methods, such as gravity casting, spontaneous infiltration, die casting, low-pressure casting, and molten metal forging. As the metal, in addition to pure magnesium, known magnesium alloys such as AZ (Mg-Al-Zn alloy), AM (Mg-Al-Mn alloy), ZK (Mg-Zn-Zr alloy), and LA (Mg-Li alloy) can be used.

[0112] When the silicon carbide is impregnated with the molten metal, impurities present on the surface of the silicon carbide are taken into the molten metal.

[0113] For example, oxides of aluminum and silicon are taken into the molten metal and form a solid solution.

[0114] (9-5) Solidification process: Cooling is controlled, and the structure of precipitates is controlled to suppress shrinkage cavities. This is the same as the "solidification process" in (2-3) above.

[0115] (9-6) Processing process The composite material produced as described above is shaped using cutting, machining, grinding, electrical discharge machining, etc. The composite material is adjusted to the desired shape and thickness. A processing process is carried out to drill holes for fixing and to form a coating on the surface, such as plating, to produce a heat dissipation substrate.

[0116] The thickness of the heat dissipation substrate is, for example, 3 to 6 mm. Examples of sizes (length B x width L) include 190 mm x 140 mm, 130 mm x 140 mm, 100 mm x 140 mm, and 70 mm x 140 mm.

[0117] (10) Method for Measuring Thermal Conductivity (10-1) Measurement Areas Figure 1 shows a heat dissipation substrate 1, its measurement sites, and a sample 4 cut out from an area surrounded by multiple lattice-shaped areas 2 and four intersections 3. The surface of the heat dissipation substrate 1 is divided into multiple areas 2, each of which is within 25 mm in length and 25 mm in width, and is at least 10 mm in length and at least 10 mm in width.

[0118] A plurality of regions 2 are provided on the heat dissipation substrate 1. The plurality of regions 2 are imaginary regions on the heat dissipation substrate 1. A sample 4 is fabricated from an area surrounded by four intersections 3 of the plurality of regions 2.

[0119] The minimum range of the multiple regions 2 is determined so as to cut out a sample 4 having a diameter of 10 mm. The multiple regions 2 are necessarily larger than 10 mm.

[0120] The multiple regions 2 are set so that if there is a dissimilar material layer such as plating around the heat dissipation substrate 1, the dissimilar material layer does not come into contact with the φ10 mm sample 4. If there is a fastening hole or a part with an obviously different shape in part of the heat dissipation substrate 1 (for example, a fastening hole part made of only metal for reinforcement purposes), the multiple regions 2 will naturally be set to avoid this.

[0121] That is, only the magnesium-silicon carbide composite material portion is the target measurement portion, and other portions are avoided.

[0122] (10-2) Preparation of Sample 4 for Measurement Sample 4 for measurement is prepared by cutting out a disk shape with a diameter of 10 mm from a plurality of regions 2 of heat dissipation substrate 1. As for dimensional accuracy, a tolerance of ±0.10 mm is preferable for sample 4 having an outer diameter of approximately 10 mm in the present disclosure.

[0123] The preferred machining method for achieving this precision is either of the following: After rough cutting, the outer periphery is machined or ground on a lathe to form a disk with a diameter of 10 mm, or wire-cut electrical discharge machining is used to form a disk with a diameter of 10 mm.

[0124] Laser cutting and water jet cutting are not recommended because they result in poor dimensional and roundness accuracy. Punching with a die is likely to cause sagging, burrs, and cracks in the sample 4, so it is not recommended to use this method.

[0125] The measurement surfaces of the sample 4 (both sides of a circle with a diameter of 10 mm) were finished as smooth as possible into two parallel flat surfaces.

[0126] For example, grinding is performed with a diamond grindstone (abrasive grains, #120 to #320). The roughness of the measurement surface is preferably Ra≦15 μm and the parallelism is preferably ±0.05 mm.

[0127] The thickness of the sample 4 is set to 1 to 5 mm, and the φ10 mm samples 4 cut out from a single plate material are basically of a uniform thickness.

[0128] The individual variation in thickness of the φ10 mm sample 4 is within ±0.2 mm. The above processing is usually performed wet. This is because dry processing generates heat, which can cause problems with accuracy and thermal deformation. After processing, any sludge on the surface is removed and the sample is allowed to dry thoroughly. This is because dirt and moisture can interfere with accurate measurement.

[0129] Regarding the thickness and shape, if both surfaces of the heat dissipation substrate 1 are not flat, for example, if it has a curved surface such as a horseshoe shape or the side of a cylinder, a spherical or ellipsoidal surface, or a surface that is slightly curved in an irregular shape, it is better to measure the thickness and align the surfaces to be parallel after cutting out the plate.

[0130] Furthermore, when the thickness of the heat dissipation substrate 1 is not uniform, for example, when the thickness is intentionally increased in the central or peripheral portions of the plate, or when the plate has a shape like a convex or concave lens, it is better to measure the measurement samples 4 cut out in the same manner as above with uniform shapes.

[0131] If there is a metal additional layer formed by plating or spraying on the surface of the heat dissipation substrate 1, it is better to remove the additional layer from both surfaces of the measurement sample beforehand.

[0132] However, if this surface metal additional layer has a uniform layer thickness, interface adhesion, components, and composition overall, comparison is possible without removing it.

[0133] In this case, it would be even better if the influence of the interface resistance of the laminated layers could be predicted in advance and the measured value of the thermal conductivity could be corrected accordingly.

[0134] Supplementary Note 1: The composite member has a plate-shaped heat dissipation substrate 1 made of a composite material containing metal and silicon carbide. One side surface of the heat dissipation substrate 1 is divided into multiple lattice-shaped regions 2, and the vertical side length B1 and horizontal side length L1 of each of the multiple regions 2 are 10 mm or more and 25 mm or less. The multiple lattice-shaped regions 2 occupy 70% or more of the total area of ​​one side surface of the heat dissipation substrate 1. The thermal conductivity at each point of the multiple lattice-shaped regions 2 is 140 W / (m·K) or more. The coefficient of variation of the thermal conductivity within the multiple lattice-shaped regions 2 is less than 3.0%. The coefficient of variation is expressed as standard deviation / average value x 100 (%). If the thermal conductivity is less than 140 W / (m·K), the thermal conductivity is low and heat dissipation is insufficient. If the coefficient of variation is 3.0% or more, the difference in thermal conductivity between each portion becomes large and heat dissipation is insufficient.

[0135] Supplementary Note 2 In Supplementary Note 1, the metal includes at least one selected from the group consisting of pure magnesium and magnesium alloys.

[0136] Supplementary Note 3 In Supplementary Note 1 or 2, the non-metallic content is 40% by volume or more and 95% by volume or less, the porosity is less than 3% by volume, and the remainder is metal. If the non-metallic content is less than 40% by volume, there is a risk that the non-metallic component will be too small and the thermal expansion coefficient will be too large. If the non-metallic content exceeds 95% by volume, the proportion of non-metallic components will increase and the thermal conductivity will decrease. Note that "may" indicates that there is a slight possibility that this will happen, and does not mean that there is a high probability that this will happen.

[0137] Supplementary Note 4 In any one of Supplementary Notes 1 to 3, the thermal expansion coefficient at temperatures of 25° C. to 150° C. is 3.5 to 20 ppm / K. If the thermal expansion coefficient is outside this range, there is a risk that the difference in thermal expansion coefficient between the semiconductor element and the substrate will be large.

[0138] Supplementary Note 5: A heat dissipation substrate made of the composite material of any one of Supplementary Notes 1 to 4, wherein the coefficient of variation of the volumetric specific heat, calculated by multiplying the density and specific heat, at each point of the lattice-shaped regions is smaller than the coefficient of variation of the thermal conductivity. Because the coefficient of variation of the volumetric specific heat is smaller than the coefficient of variation of the thermal conductivity, the heat dissipation substrate of the present disclosure is a composite material with uniform physical properties and uniform mechanical properties such as strength, hardness, and toughness, and is also advantageous in terms of thermal deformation resistance. In other words, it has the advantage of being a heat dissipation substrate that combines uniform thermal properties with shape stability.

[0139] Supplementary Note 6: A semiconductor device according to the present disclosure includes the composite material according to any one of Supplementary Notes 1 to 5.

[0140] [Details of the embodiment of the present disclosure] (11) Examples (11-1) Preparation of Samples Samples with sample numbers 1 to 7 shown in Table 1 were prepared. The manufacturing method followed "(2) Manufacturing method of metal-ceramic composite material."

[0141]

[0142] Sample numbers 1 to 7 were divided into the following criteria: Metal: Two types, Mg and Mg alloy Powder pretreatment: Yes or no Form of impurities: Surface oxidation or silicon oxide added Plate size: Standard thickness of 5 mm, with 140 mm as the standard and other dimensions of 100 mm, 140 mm, and 190 mm Measurement points were extracted using a grid of 20 to 25 mm for each size.

[0143] The linear expansion coefficient was set to a range of 7 to 8 ppm / K. The lattice area ratio in Table 1 is the area ratio of the lattice-shaped regions 2 to the entire area of ​​one surface of the heat dissipation substrate 1.

[0144] (11-2) Measurement of Thermal Conductivity According to "(6) Measurement of Thermal Conductivity", the thermal conductivity of sample numbers 1 to 7 was measured and the coefficient of variation was calculated.

[0145]

[0146] A simulated power module configuration was fabricated using heat dissipation substrates fabricated by the same method as sample No. 1 and sample No. 5, and heat dissipation from the heat-generating semiconductor to the heat dissipation substrate was evaluated.

[0147] Heat dissipation substrates of sample numbers 8 and 9 shown in Table 3 were fabricated.

[0148]

[0149] Sample No. 8 was prepared under the same conditions as Sample No. 1. Sample No. 9 was prepared under the same conditions as Sample No. 5.

[0150] 2 is a plan view of the heat dissipation substrate 1 of sample numbers 8 and 9 and the semiconductor elements 11 to 16 bonded thereon. FIG. 3 is a cross-sectional view of the semiconductor device 10 having the heat dissipation substrate 1 taken along the line indicated by the arrow III-III in FIG.

[0151] Samples 8 and 9 are metal-silicon carbide composites, and are heat dissipation substrates 1 measuring approximately 140 mm x 190 mm x 5 mm thick. Semiconductor elements 11 to 16 were simulated and bonded to one surface of the heat dissipation substrate 1, and their behavior was measured.

[0152] An insulating substrate 41 was bonded to six locations on one surface of the heat dissipation substrate 1 with solder 51. Semiconductor elements 11 to 16 were bonded to the insulating substrate 41 with solder 31.

[0153] The semiconductor elements 11 to 16 are IGBT chips (insulated gate bipolar transistors).

[0154] Aluminum nitride was used as the insulating substrate 41. Copper circuits were bonded to both sides of the aluminum nitride, making the insulating substrate 41 a DCB (Direct Copper Bonding) substrate. The aluminum nitride had a thickness of approximately 0.6 mm and a diameter of 30 mm.

[0155] A cooler 21 was attached to the heat dissipation substrate 1 on the side opposite to the mounting surface of the semiconductor elements 11 to 16. The cooler 21 was fastened and positioned using holes around the periphery of the heat dissipation substrate 1 so that the entire surface of one side of the heat dissipation substrate 1 was in contact with the water-cooled cooler 21, which was maintained at 30°C.

[0156] The semiconductor elements 11 to 16 were powered and controlled at six locations so that their temperatures reached approximately 150°C. Once they reached a stable, steady state, the power was turned off to stop the heat generation, and the time it took for the heat from the semiconductor elements 11 to 16 to transfer to the heat dissipation substrate 1 and for the temperature to drop was measured. At this time, the heat dissipation substrate 1 was surrounded by a heat insulating material to eliminate the influence of the ambient temperature.

[0157] The temperature of the semiconductor elements 11 to 16 is measured by measuring the voltage transition while passing a weak current that generates negligible heat and converting it into temperature, utilizing the relationship between the input power and temperature of the semiconductor elements 11 to 16. Alternatively, similar measurements can be made using commercially available non-contact or contact temperature measurement devices.

[0158] The time taken for semiconductor elements 11 to 16 to cool from a steady state of 150° C. to a temperature drop of 80° C. was compared, and the results are shown in Table 3.

[0159] Samples 8 and 9 have the same maximum thermal conductivity, but sample 8 has a lower minimum value, and the average value is lower by about 10 W / (m·K).

[0160] As a result, it is clear that there is a delay and variation in the cooling time for sample number 8. This temperature difference is understood to be a difference in thermal resistance.

[0161] The 25 (5x5) multiple regions 2 (matrix) on the surface of sample numbers 2, 4, and 6 were numbered #1 to #5 in the horizontal direction (L direction). Numbers a to e were assigned in the vertical direction (B direction). The position of each of the 25 multiple regions 2 was identified by a combination of #1 to #5 and a to e. The data at each measurement point #1a to #5e for sample numbers 2, 4, and 6 are shown in detail in Tables 4 to 12.

[0162]

[0163]

[0164]

[0165]

[0166]

[0167]

[0168]

[0169]

[0170]

[0171] Thermal conductivity λ = thermal diffusivity α × specific heat capacity A per unit volume. Volumetric specific heat A per unit volume is expressed as specific heat capacity Cp × bulk density ρ.

[0172] These tables show that the cause of thermal conductivity scattering is due to scattering of thermal diffusion. The tables show that the coefficient of variation of the volumetric specific heat A, calculated by multiplying the density ρ and the specific heat capacity Cp, is sufficiently smaller than the thermal diffusivity α, which means that the scattering of the density ρ and the specific heat capacity Cp is small, i.e., there is little variation. The characteristic scattering of the thermal conductivity λ of the present disclosure can be suppressed by suppressing the scattering of the thermal diffusivity α.

[0173] The embodiments and examples disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is defined by the claims, not by the above description, and is intended to include all modifications within the meaning and scope of the claims.

[0174] 1 heat dissipation substrate, 2 multiple regions, 3 intersections, 10 semiconductor device, 11-16 semiconductor elements, 21 cooler, 31, 51 solder, 41 insulating substrate.

Claims

1. a plate-shaped heat dissipation substrate made of a composite material containing metal and silicon carbide; one side surface of the heat dissipation substrate is divided into a plurality of lattice-shaped regions, each of the plurality of regions having a length and a width of 10 mm or more and 25 mm or less, the plurality of lattice-shaped regions occupying 70% or more of a total area of ​​the one side surface of the heat dissipation substrate, and the thermal conductivity at each point of the plurality of lattice-shaped regions is 140 W / (m·K) or more; The coefficient of variation of thermal conductivity within the plurality of lattice-shaped regions is less than 3.0%, The coefficient of variation is expressed as standard deviation / average value×100(%), The composite material includes at least one metal selected from the group consisting of pure magnesium and magnesium alloys.

2. 2. The composite material according to claim 1, having a silicon carbide content of 40 volume % or more and 95 volume % or less, a porosity of less than 3 volume %, and the remainder being metal.

3. 3. The composite material according to claim 1 or 2, having a thermal expansion coefficient of 3.5 to 20 ppm / K at temperatures between 25°C and 150°C.

4. A composite material as described in claim 1 or 2, wherein the magnesium alloy has one or more elements selected from Al, Si, Ca, Zn, Li, Cu, Ti, Y, and Zr added to it.

5. A heat dissipation substrate made of the composite material according to claim 1 or 2, wherein at each point of the lattice-shaped regions, a coefficient of variation of volumetric specific heat obtained by multiplying density and specific heat is smaller than the coefficient of variation of thermal conductivity.

6. A semiconductor device comprising the composite material according to claim 1 or 2.