JP2002121639A - Heat radiation substrate, and high-power high-frequency transistor package using it - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a heat radiation substrate which is composed of a tungsten/ copper and/or molybdenum/copper composite material and in which the increase of residual thermal stress resultant from the difference in a thermal expansion coefficient between the peripheral members of the substrate and the substrate itself is minimized while maintaining high thermal conductivity and also to provide a high-power high-frequency transistor package having high practical reliability and manufactured by the use of the heat radiation substrate. SOLUTION: The heat radiation substrate is composed of a tungsten/copper and/or molybdenum/copper composite material containing 25-45 mass% copper and having <=250 GPa Young's modulus. The high-power high-frequency transistor package can be manufactured by using the heat radiation substrate.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-frequency transistor package for various electric and electronic devices, and more particularly to a heat-radiating substrate used for a high-power high-frequency transistor package for electric and electronic devices of a communication base station, and a heat-radiating substrate using the same. The present invention relates to a high-power high-frequency transistor package.

[0002]

2. Description of the Related Art In recent years, market demands for high-speed operation and high integration of semiconductor devices (in the present invention, various devices using semiconductor elements are collectively referred to as the same hereinafter) are rapidly increasing. At the same time, the heat radiation substrate for mounting the semiconductor element in the device has been required to further improve the thermal conductivity in order to more efficiently release the heat generated from the element. At the same time, in order to further reduce the thermal strain between the semiconductor element and the peripheral member disposed adjacent to the heat dissipation substrate, it has been required to further reduce the difference in the thermal expansion coefficient between them. In particular,
The thermal expansion coefficients of silicon (hereinafter also referred to as Si) and gallium arsenide (hereinafter also referred to as GaAs) usually used as semiconductor elements are 4.2 × 10 ⁻⁶ /.
℃, about 6.5 × 10 ^-6 / ° C., it has been desired that the thermal expansion coefficient of the substrate is close to these.

[0003] In particular, with the recent rapid expansion of information communication networks such as mobile phones, the amount of processing information and the processing speed of various devices and devices of the base station are rapidly increasing. Accordingly, the degree of integration of the semiconductor elements mounted on the high-power high-frequency transistors built therein has been increasing day by day.

Conventionally, a package used for such a high-power high-frequency transistor is, for example, a copper-tungsten type (hereinafter also referred to as Cu-W type) or a copper-molybdenum type (hereinafter also referred to as Cu-Mo type). The composite material has been widely used as the heat dissipation substrate. FIG. 1 schematically shows a structural example of this type of package. In the figure, a is a bipolar type (Bi Polar).
Here's what. Among packages expected to increase in the future, b is an LDM on which a Si semiconductor element is mounted.
OS type (Laterally Diffused Met)
al Oxide Silicon) or GaAs
MSFET type (Metal S
emicductor Eield Effect T
tranistor). In the figure, 1
The heat dissipation substrate made of a composite material of Cu-W or Cu-Mo, 2 is for example, alumina (hereinafter Al ₂ O ₃ both describe.), Aluminum nitride (also written as follows AlN.), Also written as beryllia (hereinafter BeO ) Is a substrate made of ceramics having high thermal conductivity and electrical insulation. Note alumina, aluminum nitride, the thermal expansion coefficient of the beryllia is sequentially with 10 ^-6 / ° C. Unit 6.5,4.
The thermal conductivity is about 20, 170, 280 in W / m · K unit in order. 3 is Si or G
A semiconductor element (semiconductor integrated circuit portion) 4 made of a semiconductor material such as aAs is a metal member made of a high thermal conductive metal such as Kovar (a Fe-Ni-Co alloy, trade name).

Japanese Patent Publication No. 4-65544 discloses this type of package. According to the description of the publication, the use of a composite material having a copper content of 30% by mass or less for the heat dissipation substrate does not damage the ceramics as a peripheral member of the package, and the copper content is 25% or less. It is said that there is no practical problem even if this is directly connected to a peripheral member made of ceramics by using a brazing material by using a material having a mass% or less. However, subsequent studies by the present inventors have shown that the copper content is 25%.
When it is less than mass%, the rigidity of the substrate itself increases,
In particular, it has been confirmed that, in a package having a large amount of heat radiation, unless a considerably thick brazing material layer or a stress relaxation layer is interposed between these connection portions, it may not be able to withstand a cooling / heating cycle in practical use. From these facts, it can be seen that merely controlling the amount of copper in the material does not necessarily provide a highly reliable heat dissipation substrate.

According to Tables 1 and 2 of Japanese Patent Publication No. 4-65544, the thermal expansion coefficient of a composite material having a copper content of 30% by mass or less is 8.3 × in the case of Cu—W. 10
⁻⁶ / ° C. or less, and 8.5 × 10 ⁻⁶ / ° C. or less in the case of Cu—Mo. The thermal conductivity is 309 W / Cu-W.
m · K (0.74 cal / cm · sec · ° C) or less, C
In the case of u-Mo, 293 W / m · K (0.70 cal / c
m · sec · ° C) or less. In addition, the copper content is 25
The thermal expansion coefficient in the case of not more than mass% is in the case of Cu-W.
5 × 10 ⁻⁶ / ° C. or less, 7.8 × 10 ^{−6 for} Cu—Mo
/ ° C or less. The thermal conductivity is 2 for Cu-W.
90W / m · K (0.70 cal / cm · sec · ° C)
Hereinafter, 280 W / m · K (0.67 c
al / cm · sec · ° C.) or less.

[0007]

However, as described above,
Recently, as the degree of integration of semiconductor elements of this type of package has rapidly increased, the size of the elements and the amount of heat generation have rapidly increased. For this reason, the heat dissipation board has been required to (1) have a thermal conductivity exceeding the above level. In addition, the heat dissipation board has been required to (2) suppress an increase in residual thermal stress generated between the heat dissipation board and a peripheral member disposed adjacent to the heat dissipation board. Under such demands, when using a Cu-W-based or Cu-Mo-based composite material for the heat dissipation substrate, considering an extension of the conventional technology described in Japanese Patent Publication No. 4-65544, In order to meet the requirement (1), it is necessary to increase the amount of copper. However, in such a case, when the peripheral members are made of a material having a thermal expansion coefficient as small as that of the semiconductor element, the difference in thermal expansion coefficient between them and the heat dissipation board increases, and as a result, The thermal stress generated at the interface between the two increases. Therefore, the result is far from the request of the above (2).

In such a case, a considerably thick layer having a low Young's modulus of, for example, 1 mm to several mm is provided at a connection portion between the heat radiating substrate and the peripheral member, or a material having a thermal expansion coefficient intermediate between them. Means for absorbing and relaxing the thermal stress generated at the interface between the two layers by interposing a layer made of. However, this method is troublesome and increases production costs due to increased material costs. In addition, the thinning of packages is rapidly progressing, and this is going against the trend.

The effect of relaxing the thermal stress at the connection interface is improved as the Young's modulus of the heat radiation substrate itself is smaller. From this point of view, Cu-W and Cu-Mo have high Young's modulus W and M.
Since o is the main component, it is desirable that these amounts be as small as possible. Note that these Young's moduli are such that W is 40
7 GPa and Mo are about 332 GPa, and that of Cu is about 130 GPa. However, on the other hand, when the amount of W or Mo is reduced, the coefficient of thermal expansion increases. Therefore, it is difficult for conventional Cu-W or Cu-Mo to simultaneously satisfy these two contradictory requirements.

The object of the present invention is to overcome the above-mentioned problems without increasing the thickness of the connection layer between the heat-radiating substrate and the peripheral member, and to provide an inexpensive and simultaneously satisfying the above requirements (1) and (2). An object of the present invention is to provide a heat-dissipating substrate made of a highly reliable Cu-W-based or Cu-Mo-based composite material, and to provide a high-power high-frequency transistor package utilizing its features.

[0011]

A heat dissipation board according to the present invention for solving the above-mentioned problems contains a copper-tungsten type and / or a copper-molybdenum type having a copper content of 25 to 45% by mass and a Young's modulus of 250 GPa or less. This is a heat dissipation board made of a composite material. Further, the heat dissipation board of the present invention includes, in addition to the above-described configuration, one in which the content of the iron group metal is controlled to 0.2% by mass or less.

The present invention also includes a high-power high-frequency transistor package using the above-described heat dissipation substrate, and a semiconductor device using the same.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION The heat radiating substrate of the present invention comprises
Copper-tungsten and / or copper containing 45% by mass
It is made of a molybdenum-based composite material. Here, when the amount of copper is within the above range and the balance is tungsten, the amount of copper in this material corresponds to 42 to 64% by volume. When the balance is molybdenum, it corresponds to 27 to 48%.

Here, by simply applying the compound rule in this composition range and estimating the theoretical value of the Young's modulus using the aforementioned values of the Young's modulus of Cu, W and Mo, for example, the copper content is 25 to 45 mass%. % Cu-W, about 230 GPa
From about 290 GPa, and similarly from about 250 GPa to about 280 GPa for Cu-Mo in the same composition range. However, the Young's modulus in the corresponding composition range of the conventional material manufactured by the infiltration method (a method of manufacturing a porous body of W or Mo and infiltrating Cu in the pores) is W or Mo. Since the particles are connected to each other to form a skeleton, they are usually higher than the above theoretical values, and about 240 G
From Pa (Cu 45 mass%) to 300 GPa (Cu 25 mass%), about 250 GPa (Cu 45 mass%) to 290 GPa (Cu 25 mass%) for Cu-Mo. On the other hand, in a conventional material produced by a sintering method (a method of performing liquid phase sintering of a molded body of a mixture composed of Cu and W or a corresponding composition of Cu and Mo at a temperature equal to or higher than the melting point of Cu), Since the Young's modulus in the region is a structure like a mixture of W or Mo particles and Cu, it is usually on the level of the above theoretical value. That is, when the amount of copper is 25 to 45% by mass,
In W, about 230 GPa (Cu 45 mass%) to about 29 GPa
0 GPa (Cu 25% by mass), similarly Cu in the same composition range
In the case of -Mo, the range is from about 250 GPa (Cu 45 mass%) to about 280 GPa (Cu 25 mass%).

On the other hand, the heat radiation material of the present invention having the above composition has a Young's modulus of 250 GPa or less when the copper content is in the range of 25 to 45% by mass. Note that the Young's modulus decreases with an increase in the amount of copper according to the compound rule, and is remarkably smaller within the above-mentioned range of the amount of copper as compared with the conventional one having the same component composition. For example, if the amount of copper is 45% by mass, the Young's modulus is 210 GPa or less. In the composite material of the present invention,
When the amount of copper is less than 25% by mass, the thermal conductivity of Cu-W is less than 290 W / m · K, and that of Cu-Mo is 280 W / m.
-Since it is less than K, if this is used for a heat-dissipating substrate of a high-power high-frequency transistor, the heat-dissipating ability is not sufficient. As a result, even if the Young's modulus is low as described above, the substrate itself may be warped in practical use, or may be easily damaged near a connection interface with a peripheral member, thereby shortening the life of the package. On the other hand, when the amount of copper exceeds 45% by mass, the coefficient of thermal expansion becomes C
Since u-W exceeds 9.5 × 10 ^-6 / ° C.
Since -Mo exceeds 10.0 × 10 ⁻⁶ / ° C., even if the Young's modulus is low as described above, damage due to thermal stress near the connection interface with the peripheral member is likely to occur due to thermal cycling during practical use. . In either case, there is a high possibility that reliability will be impaired.

The microstructure of the heat radiation substrate of the present invention is different from that of the conventional similar material even if the composition is the same, that is, the hard particles (W or Mo) and the amount ratio of Cu in the material are the same. Conventional Cu-W or Cu-M obtained by the infiltration method
In the o-based material, hard particles are continuously connected to each other to form a three-dimensional network (skeleton). However, in the material of the present invention obtained by the infiltration method,
In the state where the connection of the hard particles is partially cut or the contact area of the connection of the same particles is smaller,
Forming a three-dimensional network. That is, in the case of the present invention, the connection of the hard particles is more imperfect than that of the conventional one having the same composition. As a result, when comparing materials having the same composition, the one of the present invention has a smaller Young's modulus than the conventional one.

Unlike Cu-W or Cu-Mo materials obtained by sintering or casting, hard particles or aggregates of the same particles are contained in a Cu matrix, unlike those obtained by infiltration. The tissue is distributed in the form of islands.
The same applies to the present invention, but the three-dimensional arrangement of hard particles or aggregates of the same particles in the material becomes more sparse compared to the material obtained by the conventional sintering method, even with the same composition. I have. That is, the average distance between the hard particles or the aggregates of the hard particles is large. As a result, when comparing materials having the same composition, the one of the present invention has a smaller Young's modulus than the conventional one. The material obtained by the casting method is
As compared with those obtained by the sintering method or the infiltration method, vacancies are easily generated, so that they are inferior in thermal conductivity. In addition, segregation of hard particles is likely to occur during the casting process, and it is difficult to obtain a homogeneous product.

As a preferable material composition of the present invention for realizing the above composite structure, there is a composition in which the content of iron group metals (iron, nickel and cobalt) is suppressed to 0.2% by mass or less. Preferably 0.05 to 0.1% by mass
It is. In a Cu-W-based or Cu-Mo-based composite material, in order to increase the wettability between a conventionally melted matrix metal (a metal having Cu as a main component) and hard particles, an iron group metal is usually added in an amount of 0%. In most cases, it is added in an amount of 0.5% by mass or more. In the present invention, by reducing the content thereof to an amount smaller than this, while ensuring the wettability between the matrix metal and the hard particles, a structure in which the connection between the hard particles is partially cut, or the same particles It is possible to easily obtain a structure in which the area of the connection portion between them is reduced. If the amount exceeds 0.2% by mass, it is difficult to realize the amount, and the thermal conductivity of the material is liable to decrease. If the distribution of the iron group metal is extremely fine and uniform, the decrease in the thermal conductivity of the material can be reduced even with the same content.

Next, a method for manufacturing a heat dissipation substrate (hereinafter, also simply referred to as a substrate) of the present invention will be described. The substrate of the present invention is manufactured by a conventional manufacturing method, that is, a method according to the aforementioned infiltration method, sintering method and casting method. Tungsten, molybdenum and copper powders as starting materials may be commercially available, but the tungsten and molybdenum powders constituting the hard particles are compared in the particle size distribution width (difference between the maximum particle size and the minimum particle size). It is better to use a narrow object. For example, the ratio of the distribution width to the average particle size may be controlled in a range of about 30 to 50%. In the infiltration method, when preparing a composite material, a powder containing hard particles as a main component is molded in advance to form a molded body, or further sintered to form a porous intermediate. At that time, when the particle size distribution of the hard particles is adjusted as described above, it is easy to partially cut the connection between the hard particles in these intermediates, or to reduce the area of the connection between the particles. Become. The same applies to the case where the composite material of the present invention in which the average distance between the hard particles or the particle aggregates is increased by the sintering method.

When an iron group metal powder is added, it is preferable to use a powder having an average particle size as small as possible. Preferably it is 5 μm or less, more preferably 1 μm or less. The form to be added may be any form other than metal as long as it is a compound that converts to metal at a temperature of 1000 ° C. or lower. For example, by adding it in the form of an organometallic compound or an inorganic salt of an iron group element, it can be easily finely distributed (1 μm or less) and very uniformly distributed. The amount to be added is desirably 0.2% by mass or less for the above-described reason, and is particularly preferably 0.05 to 0.1% by mass.
1% by mass is desirable. When the iron group metal or a compound thereof is added and mixed, it is desirable to use a mixing vessel made of the same iron group metal. This is because, due to wear of the container, extremely fine particles of the iron group metal are uniformly mixed together with the mixing.

In the case of the infiltration method, first, a powder mainly containing tungsten and / or molybdenum is prepared.
In this case, if necessary, an iron group metal powder or a powder of a compound which converts to the same metal at a high temperature is added after compounding in an amount such that the amount finally falls into the above range in terms of a simple substance of the iron group metal. You may. In the case of the sintering method, first, a powder mainly containing copper is added to a powder mainly containing tungsten and / or molybdenum and mixed.

Next, the powder prepared in the form of granules, slurry, or clay is formed into a predetermined shape to obtain a formed body. The molding method may be any molding method commonly used in powder metallurgy, such as dry molding, extrusion molding, doctor blade molding, and injection molding. However, the molding density at that time is desirably as high as possible in order to secure the handling strength of the molded body and to suppress the deformation due to shrinkage during firing. For this reason, in any of the infiltration method and the sintering method, an organic binder may be added and granulated into granules in order to impart excellent moldability as necessary. Thereby, the filling property of the powder at the time of molding can be enhanced, and the molding density can be easily increased. Particularly in the case of dry molding, it is desirable to granulate the powder in advance.

The compact to which the organic binder is added is calcined in a non-oxidizing atmosphere (vacuum, reduced pressure or non-oxidizing gas) to remove the binder. In the case of the infiltration method, if necessary, the temperature is further increased in the same atmosphere to adjust the porosity of the molded body. Thereafter, in the same atmosphere, preferably in a hydrogen atmosphere, the temperature of the metal containing copper as a main component is raised to a temperature equal to or higher than the melting point of the metal, so that the melt of the metal is infiltrated and filled in the pores. If the porosity after firing from the volume and the molding density of the organic binder in the molded body can be designed in advance,
The molded body contains copper as a main component, and a metal lump in an amount corresponding to the volume thereof is brought into contact with the infiltration material, and after the binder is removed, the temperature is continuously raised to a temperature equal to or higher than the melting point of copper. To infiltration.

On the other hand, in the case of the sintering method, after removing the binder, in the same atmosphere, preferably in a hydrogen atmosphere,
By raising the temperature to a temperature equal to or higher than the melting point of a metal containing copper as a main component, liquid phase sintering with a melt of the metal is performed. Also in this case, both the heating steps of removing the binder and sintering may be performed separately, or may be performed by a continuous program.

When a material having a relatively large copper content of 33 to 45% by mass is prepared by the infiltration method, a porosity sufficient for maintaining the shape of the molded body is secured (usually less than 50% by volume). If necessary, a powder containing 10% by mass or less of copper as a main component and a metal containing metal to be infiltrated into the pores may be added in advance to the initial powder (premixed infiltration method). In either case of the infiltration method or the sintering method, the surface of the molded body or the intermediate body that is not in contact with the infiltrant (metal mainly composed of copper) may be dissolved out of the surface of the molten metal. In order to prevent oozing, a thin layer of a ceramic or the like (elution inhibitor) that does not wet the molten metal may be provided. This reduces the formation area of the elution and oozing parts,
A net-shaped material can be easily obtained.

The cross-sectional structure of the bipolar high-power high-frequency transistor package using the composite material of the present invention thus obtained is shown in FIG.
Shown in Note that the reference numerals in FIG. When the package of the present invention is compared with the conventional package in the cross section of the connection part, the two are the same between the heat radiating substrate and the semiconductor element and other members arranged adjacent thereto. The structure of the connection layer between the member and the heat dissipation board is significantly different. In other words, by using the material having a low Young's modulus of the present invention for the heat dissipation board, if the amount of copper on the board increases, and even if the difference in thermal expansion coefficient between adjacent members, particularly members made of ceramics, becomes considerably large,
A direct connection by a brazing material layer and a connection by a thinner brazing material layer, which were not realized by the conventional device, can be realized.

Further, in the case of using the conventional material for a particularly large package, it is necessary to additionally interpose a thick thermal stress relieving layer at a connection portion with an adjacent member. When such a layer is provided, it takes time to dispose the layer, and the thermal resistance of the connection layer increases, which is not good from the viewpoint of heat radiation. However, by using the substrate of the present invention, the connection can be made even if the relaxation layer is not provided or the thickness is reduced. For larger packages, to ensure the reliability of the connection layer, the board is made of a material with a small amount of copper and a small coefficient of thermal expansion. Was arranged to cope with an increase in the amount of heat radiation. However, by using the substrate of the present invention, it is not necessary to additionally dispose such a heat radiating plate as long as a necessary heat radiation amount can be secured. As described above, when the substrate of the present invention is used, the structure of the connection portion of the package is significantly improved.

[0028]

EXAMPLES Example 1 (Sintering Method) Powders of tungsten, molybdenum, electrolytic copper and iron each having a purity of 99.5% or more were prepared. The total content of iron group elements in the tungsten and molybdenum powders was 0.05% by mass or less. The average particle diameters of the tungsten, molybdenum, electrolytic copper and iron powders were 8, 9, 10 and 0.8, respectively, in μm units. The distribution width of the particle size of the tungsten and molybdenum powders (the difference between the maximum particle size and the minimum particle size) was 3 μm. The content of the iron group element in the electrolytic copper powder was 0.01% by mass or less in total thereof.

These raw material powders were weighed at the ratio described in the column of "composition" in Table 1, and paraffin as an organic binder was added to the powder at a ratio of 1 to 100 by mass. The amount of iron added depends on the main components (Cu,
Table 1 shows the mass ratio of the total amount of W and / or Mo) to 100. Thereafter, these were put into a pot made of the same stainless steel lining together with ethanol and stainless steel balls, and mixed with the same ball mill for the time shown in Table 1.

The contents of the main component and the iron group metal elements (Fe, Ni and Co) in the powder of the mixed slurry can be determined by heating the slurry to 400 ° C. under reduced pressure to remove the binder and volatile components, (Metal mixture consisting essentially of Cu, W and / or Mo), which was confirmed by chemical analysis. The results are shown in the "mixture composition" column of Table 1. The amount of iron group metal in this column is a mass ratio to the whole powder particles, and the main component mass ratio is Cu
Value / amount of hard particles.

The solid content was acid-treated to extract only W and Mo particles, which were observed with a scanning electron microscope to confirm the average particle diameter and the distribution width of the particle diameter. After the acid-treated powder was dispersed on a sample stage with ethanol, a 3000-fold rectangular field photograph was taken, and the total length and number of particles cut by two diagonal lines in the rectangular field were confirmed. The value obtained by dividing the former by the latter was defined as the average particle size of the hard particles. The difference between the maximum particle length and the minimum particle length at that time was defined as the value of the particle size distribution width of the hard particles.

[0032]

[Table 1]

The slurry body was granulated by a spray drier to obtain a granular powder. Then, it was formed by a dry press to obtain a disk-shaped formed body having a diameter of 25 mm and a thickness of 5 mm. These compacts are placed in a vacuum furnace, and
After heating to 00 ° C. to remove paraffin and volatile components, the sample was placed on an alumina refractory plate and sintered in a hydrogen furnace at a maximum temperature of 1200 ° C. for 2 hours in a hydrogen stream. In addition, in order to prevent the copper melted during the liquid phase sintering of copper from seeping out around the compact and obtain a net-shaped sintered body, before placing it on the same refractory board, A thin layer of carbon powder (elution inhibitor) was formed on the entire surface except for the circular surface.

Each sintered body was blast-finished to remove the layer of the elution preventive agent on the surface, and was easily ground to finish.
It was made into a disk shape of 3 mm and thickness of 4 mm. Thereafter, the amounts of Cu, W, Mo and the iron group metal in each sample were confirmed by chemical analysis. The surface of the ground specimen was observed with a scanning electron microscope in the same manner as in the case of the above-mentioned mixture, and the average particle size of the hard particles and the distribution width of the particle size were confirmed. As a result, the mass ratio of the main component, that is, the mass ratio of Cu to hard particles (W and / or Mo particles), and the amount (% by mass) of the iron group metal were at the levels confirmed in the state of the mixture described above. Was almost the same as Also, the average particle size of the hard particles and the distribution width of the particle size were almost the same as the levels confirmed in the state of the mixture. Further, the thermal conductivity, the coefficient of thermal expansion, and the Young's modulus were separately confirmed, and the results are shown in the “sintered body” column of Table 2. Although not listed in Table 2,
The relative densities of the sintered bodies (the values obtained by dividing the measured densities measured by the underwater method by the theoretical densities calculated from the component amounts) were all 99% or more. The thermal conductivity was determined by the laser flash method, the coefficient of thermal expansion was determined by the differential transformer method using a separately prepared cylindrical specimen, and the Young's modulus was measured by applying a separately prepared tensile test piece to a tensile tester.・ Draw a strain curve and calculate from this.

Sample No. 25 was powdered of iron stearate having an average particle size of 0.1 μm instead of iron powder (changed to iron particles having an average particle size of 0.1 μm or less during the above-mentioned sintering process). The sample was prepared in the same procedure as in Sample No. 4 except that Sample No. 4 was used. Samples 26 to 28 had substantially the same average particle size (8 μm) in place of the tungsten powder.
m), and was prepared in the same procedure as Sample No. 4 except that tungsten powder having a different particle size distribution width was used. In this case, the distribution width of the tungsten particles in the mixture is changed as shown in Table 1. Sample No. 2
Sample No. 9 was prepared in the same procedure as Sample No. 4 except that tungsten powder having an average particle size of 4 μm and a particle size distribution width of 1 μm and a small particle size was used instead of the above tungsten powder. is there. In this case, as shown in Table 1, the average particle size and distribution width of the tungsten particles in the mixture are considerably small.

[0036]

[Table 2]

From the above results, (1) the Cu-W-based and Cu-Mo-based composite materials of the present invention obtained by the sintering method are as follows:
The copper content is 25% by mass, and the Young's modulus is 250 GPa or less. This is the same as the usual level of the conventional material having the same composition as described above, ie, 290 GPa of Cu-W type and 2% of Cu-Mo type.
It can be seen that it is significantly lower than 80 GPa. Cu-W
The same can be said for the Mo-based composite material. It can also be seen that this level of Young's modulus is significantly lower than that of the conventional one, as well as the estimated value for the same composition from the compound rule described above, when the amount of copper is in the range of 25-45% by mass. In particular, it can be seen that the effect is enhanced by suppressing the content of (2) the iron group metal to 0.2% by mass or less.

Example 2 (Infiltration method) The same powders of tungsten, molybdenum, electrolytic copper and iron as in Example 1 were prepared. These raw material powders were weighed at the ratio described in the column of "composition" in Table 3, and paraffin as an organic binder was added to the powder at a ratio of 2 to 100 by mass. Table 3 shows in advance the mixed powders of Sample Nos. 5 to 9 and 18 in which the final volume of copper after compounding (the pore volume of the porous body before infiltration) is expected to be 50% by volume or more. A predetermined amount of electrolytic copper powder was blended (pre-blended). The amount of iron added is shown in Table 3 as a mass ratio to the total amount 100 of the main components (Cu, W and / or Mo). Thereafter, mixing was carried out for each of the times shown in Table 3 using the same ball mill as in Example 1. The main components and the iron group element metals (Fe, Ni and C) in the powder of the slurry after mixing.
The content of o), the average particle size of the hard particles, and the particle size distribution width were confirmed in the same manner as in Example 1.

The slurry body was granulated by a spray drier to obtain a granular powder. Then, it was formed by a dry press to obtain a disk-shaped formed body having a diameter of 25 mm and a thickness of 5 mm. These compacts are placed in a vacuum furnace, and
The mixture was heated to 00 ° C. to remove paraffin and volatile components to form a porous body (skeleton). In addition, 5
The molding density of each sample was changed by changing the molding pressure in the range of 0 to 500 MPa, and the porosity of the porosity level described in Table 3 was adjusted. The porosity is calculated based on the theoretical density ρ _t (g / cm ³ unit, tungsten,
The densities of molybdenum and copper are calculated in g / cm ³ as 19.3, 10.2 and 8.9, respectively),
The measured volume V calculated from the measured diameter and thickness of the porous body
(Cm ³ ) and the measured mass W (g) of the porous body,
It was calculated by the calculation formula (1−W / ρ _t V) × 100 (%).

[0040]

[Table 3]

The mass of copper sufficient to fill all the pores of these porous bodies was calculated, and a copper plate of the same material as the electrolytic copper powder of Example 1 having a slightly larger mass than that was prepared. Next, these copper plates were placed on an alumina-based refractory plate, and a porous specimen having a corresponding amount of porosity was placed thereon, and was held at a maximum temperature of 1200 ° C. for 2 hours in a hydrogen stream in a hydrogen furnace to obtain a porous plate. The molten copper was infiltrated into the pores of the body. In addition, in order to prevent molten copper from seeping out around the molded body and obtain a net-shaped infiltrated body, before placing the porous sample on the copper plate,
A thin layer of carbon powder (anti-elution agent) was formed on the entire surface other than the one circular surface in contact with the copper plate.

Each infiltrated body was finished in the same manner as in Example 1 to form a disk having a diameter of 23 mm and a thickness of 4 mm.
Thereafter, in the same manner as in Example 1, the mass ratio between Cu and hard particles, the amount of iron group metal, thermal conductivity, thermal expansion coefficient, and Young's modulus in each sample were confirmed. Table 4 shows the results.
In the “Sintered Body” column. Although not listed in Table 4, the relative densities of the sintered bodies were all 99.5% or more. As a result, the mass ratio of the main component, that is, the mass ratio of Cu to hard particles (W and / or Mo particles), and the amount (% by mass) of the iron group metal are determined by the level confirmed in the state of the mixture described above. It was almost the same. Also, the average particle size of the hard particles and the distribution width of the particle size were almost the same as the levels confirmed in the state of the mixture.

Sample No. 25 was prepared by the same procedure as Sample No. 4 except that iron stearate powder having an average particle size of 0.1 μm was used instead of iron powder. Samples Nos. 26 to 28 had substantially the same average particle size (8 μm) in place of the above tungsten powder, and were the same as Sample No. 4 except that a tungsten powder having a particle size distribution width of 10 μm was used. It was produced by the procedure. Sample No. 29 has an average particle size of 4 μm and a particle size distribution width of 1 μm instead of the tungsten powder.
Except that a tungsten powder having a small particle size was used. In this case, as shown in Table 4, both the average particle size and the distribution width of the tungsten particles in the mixture are considerably small.

[0044]

[Table 4]

From the above results, (1) the Cu-W-based and Cu-Mo-based composite materials of the present invention obtained by the infiltration method
The copper content is 25% by mass, and the Young's modulus is 250 GPa or less. This is 300 GPa of Cu-W type and 2 levels of Cu-Mo type which are the usual levels of the same material of the same composition as the conventional one described above.
It can be seen that it is significantly lower than 90 GPa. Cu-W
The same can be said for the Mo-based composite material. It can also be seen that the level of this Young's modulus is significantly lower than that of the conventional one as well as the estimated value for the same composition from the above-mentioned compound rule when the amount of copper is in the range of 25-45% by mass. In particular, it can be seen that the effect is enhanced by suppressing the content of (2) the iron group metal to 0.2% by mass or less.

Example 3 (Casting Method) The same tungsten, molybdenum, and iron powders as in Example 1 and a copper plate having substantially the same purity and material as the electrolytic copper powder of Example 1 were prepared. Tungsten, molybdenum, copper and iron powders were weighed at the mass ratios described in the composition column of Table 5 and mixed for 2 hours using the same ball mill apparatus as in Example 1, and b to f and h was obtained. However, a and g in Table 5 are the above-mentioned tungsten and molybdenum powders not subjected to the above-mentioned mixing treatment. The amount of the iron group metal in the mixture, the average particle size of the hard particles, and the particle size distribution width were confirmed by the same procedure as in Example 1, and the results are shown in Table 5.

[0047]

[Table 5]

The powders (a) to (h) in Table 5 and the electrolytic copper plate were compared with the main component mass ratio in Table 6 (note that the numerator of the mass ratio in the table is the mass ratio of the electrolytic copper plate, and the denominator is the mass ratio of the hard particles). Was weighed. Thereafter, the weighed electrolytic copper plate was first put into a graphite crucible having an inner size of 32 mm in width and length and 5 mm in depth, and this was heated at 1200 ° C. in nitrogen to form a molten metal. Then, while stirring the molten metal, powder of the weighed hard particles was put into the molten metal, and the molten metal was gradually cooled to produce respective casts having the main component mass ratios shown in Table 6. Thereafter, the cast body was cut to obtain a specimen of the cast body having a width of 30 mm and a thickness of 3 mm. Specimens other than this shape were also prepared, and various properties of the cast body described in Table 6 were confirmed in the same procedure as in Example 1.
Although not shown in Table 6, the relative density of the sintered body is
It was in the range of 98-98.5%.

[0049]

[Table 6]

From the above results, (1) the Cu-W-based and Cu-Mo-based composite materials of the present invention obtained by the casting method are:
The copper content is 25% by mass, and the Young's modulus is 250 GPa or less. This is 300 GPa of Cu-W type and 2 levels of Cu-Mo type which are the usual levels of the same material of the same composition as the conventional one described above.
It can be seen that it is significantly lower than 90 GPa. Cu-W
The same can be said for the Mo-based composite material. It can also be seen that the level of this Young's modulus is significantly lower than that of the conventional one as well as the estimated value for the same composition from the above-mentioned compound rule when the amount of copper is in the range of 25-45% by mass. In particular, it can be seen that the effect is enhanced by suppressing the content of (2) the iron group metal to 0.2% by mass or less.

Example 4 (Rolling method) A sample having a width of 15 mm and a thickness of 3 m having the same composition as that of each of the samples 4 obtained in each of the examples 1 to 3 was obtained.
A material made of a composite material having a length of 200 mm and a length of 200 mm was prepared. After preheating these materials to 900 ° C. in nitrogen, they are immediately rolled with a roll to have a width of 20 mm and a thickness of 1.
A hoop-shaped material having a length of 5 mm and a length of approximately 300 mm was obtained.
From these materials, the width and length are 20 mm and the thickness is 1.
After cutting out a test piece having a shape of 5 mm, the structure of the material and the same items as in Example 1 were evaluated. Table 7 shows the results. The “aspect ratio” in Table 7 is the average value of the hard particles distributed in the composite material, and the confirmation procedure is as follows. Using a photograph of a rectangular visual field taken in the same procedure as the particle size observation of the hard particles in Example 1, the length D of the largest dimension portion of all the hard particles included in the whole visual field in the same visual field
The maximum and the length Dmin of the minimum dimension portion were measured, and the sum was calculated by dividing the sum of the aspect ratios Dmax / Dmin of the individual particles by the number of the measured particles. Further, as a result of confirming the aspect ratio of the hard particles in the starting material as the starting material by the same procedure, all were almost 1.

[0052]

[Table 7]

From the above results, although the material after rolling has a directional structure, the thermal conductivity and the coefficient of thermal expansion show some directionality, but the coefficient of thermal expansion is almost the same as before rolling. It can be seen that the high thermal conductivity is maintained and the Young's modulus is further reduced by about 1 to 2%.

Example 5 The materials prepared in the same manner as in the materials of Examples 1 to 4 described in the column of “Heat sink” in Table 9 were used as heat dissipation boards, and the three types shown in Table 8 were used. Thus, test pieces for a high-frequency transistor package using members made of various materials and external sizes were manufactured. The structure of the three types of packages manufactured is shown in FIG. 3 for the LMDMOS type.
FIG. 4 schematically shows the bipolar type and FIG. 5 shows the MSFET type. In each figure, the upper figure is a view of the package from the top,
The bottom is a view cut along the AA 'section. The alumina used for the ceramic member 2 and the Si and Ga for the semiconductor element 3 were used.
The thermal conductivity of the metal member 4 made of As and Kovar is W /
They are 20, 140, 54, and 17 in the order of m · K, and the thermal expansion coefficients are 6.5, 4 in the order of 10 ⁻⁶ / ° C.
2, 6.5 and 5.3.

[0055]

[Table 8]

The connection between the members of the package was performed as follows. Heat radiation board of the present invention and ceramic (alumina)
The thickness between the substrate and the substrate is 9 through the metallized layers.
By forming a 0 μm Ag-Cu brazing layer at 800 to 900 ° C., a 200 μm thick Ag—Sn brazing layer is formed between the ceramic (alumina) base material and the metal member (Kovar) via the metallized layers. 350-40 layers
By forming at 0 ° C., the semiconductor device (Si
Alternatively, between the GaAs) and the ceramic (alumina) base material, A
Each connection was made by forming a u-Si solder layer at 400 to 500 ° C. According to the above procedure, ten packages were produced for each type of the heat dissipation board of the present invention.

The package program assembled as described above was selected three by three for each package type, and each cycle was held at -60 ° C for 15 minutes, and then held at 150 ° C for 15 minutes. A cooling test was repeated 3000 times. Table 9 shows test results of LDMOS type package specimens as a typical example. Although not described in Table 9, the same test was performed on the bipolar type shown in FIG. 4 and the MSFET type shown in FIG. 5, and the results were almost the same as those of the LDMOS type. In this test, the amount of warpage of the package specimen was confirmed initially and every 500 cycles thereafter. Even if one of the three specimens has a warpage of 0.5 μm, which is a practical problem
/ Mm or more (in the present invention, this is referred to as a critical value of warpage), the test of the sample number was terminated. The three specimens of the censored sample numbers were checked for the final warpage amount shown in Table 9 at that time, and the output ratio E / E ₀ at that time and before the test (however, the initial output was E ₀ , And the output at that time is E).
In Table 9, the evaluation results of the above-mentioned sustainability of the package performance in the thermal cycle are indicated by × every 500 cycles. However, ○ and × indicate that the warpage is less than the critical value at that time,
Indicates that it is above the critical value. The numbers in the “heat sink” column in Table 9 are the numbers of the heat sink materials used. The first numerical value is the example number, and the latter numerical value is the sample number. For example, in the case of Sample 2 in Table 9, the heat sink material used is Sample 3 of Example 1. The numerical value in the column of “Result of sustainability evaluation” indicates the sustaining cycle in “kilocycle” unit.

The amount of warpage of the test piece was confirmed as follows. As schematically shown in FIG. 6, the heat radiation substrate 1 is placed on the surface plate 5 so that the back side (the side opposite to the surface on which the semiconductor element is mounted) is facing upward, and while maintaining the same distance from the surface of the surface plate, , The laser light source 6 is scanned in two diagonal directions of the same square on the back side, and the maximum distance Lmax and the minimum distance L
After confirming min, the difference Lmax−Lmin (μm) is calculated, and this difference is divided by the scanning distance (mm) in the diagonal direction to obtain the amount of warpage (μm / mm). In addition, the above output ratio E
/ E ₀ was confirmed by preparing an output measuring jig provided with a simple operation circuit in advance and connecting a package and a test piece thereto.

The magnitude of the amount of warpage depends on the magnitude of the residual stress at the interface between the ceramic substrate and the heat dissipation board in this type of package. It is affected by the conductivity and the Young's modulus of the heat dissipation board. In particular, the greater the thermal conductivity of the heat dissipation substrate and the smaller its Young's modulus, the greater the effect of alleviating the stress caused by the difference in the coefficient of thermal expansion. Therefore, if the thermal conductivity of the heat radiating substrate is low, heat stays at the connection interface, and if the Young's modulus is large, the relaxation of thermal stress in the cooling / heating cycle becomes insufficient, so that the warpage is further promoted. This reduces the output of the transistor. Further, if the brittle ceramic substrate side is damaged, the operation becomes inoperable.

[0060]

[Table 9]

[Table 10]

As is evident from the results in Table 9, in the package specimen using the heat radiation board of the present invention, even if the difference in the coefficient of thermal expansion from the ceramic substrate is relatively large, the warpage reaches the critical value. It can be seen that the number of cycles is markedly increased, and that it has a long service life and high practical reliability. This is because the heat radiating substrate of the present invention has a significantly lower Young's modulus than conventional materials of the same component composition. When this package was incorporated into various semiconductor devices, it was found that the package was sufficiently applicable to a high-output semiconductor device, which was impossible with a conventional heat radiation board of the same material.

[0062]

As described in detail above, the Cu-W of the present invention
, Heat-dissipating substrates made of Cu-Mo-based and Cu-W-Mo-based composite materials have a significantly smaller Young's modulus than that of the conventional composition having the same component composition. When used as, it has a long life and high practical reliability. Therefore, a high-power high-frequency transistor package with high practical reliability and a semiconductor device using the package can be easily provided.

[Brief description of the drawings]

FIG. 1 is a diagram schematically illustrating a structural example of a high-frequency transistor package.

FIG. 2 is a diagram schematically illustrating a cross-sectional structure of the present invention and a conventional high-frequency transistor package.

FIG. 3 is a diagram schematically illustrating the structure of an LDMOS high-frequency transistor package according to an embodiment of the present invention.

FIG. 4 is a diagram schematically illustrating a structure of a bipolar high-frequency transistor package according to an example of the present invention.

FIG. 5 is a diagram schematically illustrating the structure of an MSFET-type high-frequency transistor package according to an example of the present invention.

FIG. 6 is a diagram illustrating a method for measuring the amount of warpage of a package according to the present invention.

[Explanation of symbols]

 1, heat dissipation substrate 2, ceramic substrate 3, semiconductor element 4, metal member

Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat II (Reference) H01L 23/373 H01L 23/36 M

Claims (4)

[Claims] (1) containing 25 to 45% by mass of copper and having a Young's modulus of 2
A heat dissipation substrate made of a copper-tungsten-based and / or copper-molybdenum-based composite material of 50 GPa or less. 2. The heat dissipation board according to claim 1, wherein the content of the iron group metal is 0.2% by mass or less. 3. A high-power high-frequency transistor package using the heat dissipation substrate according to claim 1. 4. A semiconductor device using the package according to claim 3.