TW201637152A - Heat sink substrate - Google Patents

Abstract

A rectangular heat dissipation substrate formed by alternately stacking a Cu layer and a metal A layer made of a metal A, wherein the Cu layer and the metal A layer are a total of 5 or 9 layers, and the heat dissipation substrate of the metal layer A has a longitudinal direction The ratio L/t of the average crystal grain length L to the average crystal grain length t in the thickness direction is 1.5 to 10, and the average crystal grain length L in the longitudinal direction and the average crystal grain in the width direction of the heat dissipation substrate of the metal A layer. The ratio d of the length d is from 1.2 to 7, and the above features provide a heat-dissipating substrate which can improve long-term reliability.

Description

Heat sink substrate Technical field

The invention relates to a heat dissipating substrate, in particular to an electric device The heat sink substrate of the force module.

Background technique

Power modules such as electric vehicles, hybrid electric vehicles, and wind power generation are used as components for power control. The power module is bonded to an insulating substrate made of ceramics and a heat-dissipating substrate formed of metal, and the semiconductor device is brazed through the bonding material, in particular, an LSI, an IC, a power transistor, or the like that operates by a large power. A semiconductor device that operates with a large power generates heat when it is used.

The heat dissipation substrate needs to diffuse and dissipate heat generated by the semiconductor devices more efficiently. However, as described above, since the power module is a bonded body composed of different kinds of materials, internal stress is generated not only at the time of manufacture but also due to temperature change during use. Due to this internal stress, there is a problem that the heat dissipation substrate is deformed. Therefore, the heat dissipation substrate preferably has high mechanical strength and high thermal conductivity.

On the other hand, for example, Patent Document 1 discloses a coating material in which a Cu layer, a Mo layer, and a Cu layer are sequentially laminated as a heat dissipation base composed of a three-layer structure. board. By changing the Mo volume ratio in the three-layered cladding material in the range of 20% to 99.6%, the thermal conductivity and the thermal expansion coefficient can be controlled, and the thermal conductivity and the specific Cu ratio higher than that of the Mo monomer can be obtained. Small thermal expansion coefficient.

Further, Patent Document 2 discloses a relationship between a thermal expansion coefficient and a Cu volume ratio of a three-layer structure cladding material in which a Cu layer, a Mo layer, and a Cu layer are sequentially laminated. In the case of the cladding material of this structure, when the Mo layer is one layer, in order to have, for example, a thermal expansion coefficient of 12 × 10 ^-6 /K or less, the amount of Mo used having a low thermal conductivity must be 20% or more of the total mass. Therefore, the thermal conductivity of the cladding material in the thickness direction is about 230 W/(m.K).

Further, Patent Document 3 discloses a cladding material in which five or more layers of a Cu layer and a Mo layer are alternately laminated. In this case, by laminating 5 or more layers, a cladding material having a smaller thermal expansion coefficient and a higher thermal conductivity can be obtained.

Prior technical literature Patent literature

Patent Document 1: Japanese Laid-Open Patent Publication No. 2-102551

Patent Document 2: Japanese Laid-Open Patent Publication No. Hei 6-268115

Patent Document 3: Japanese Laid-Open Patent Publication No. 2007-115731

Summary of invention

However, there is still a need for a more reliable heat dissipation substrate that can cope with the demand for power module power. Particularly noticeable is the lamination of the layers between the Cu layer and the Mo layer, and after mounting the semiconductor device bonded to the heat dissipation substrate. Long-term reliability. Since a semiconductor device such as Si or SiC having a different thermal expansion coefficient and a copper electrode on a ceramic substrate are bonded by a heat dissipation substrate, thermal stress is applied during a thermal cycle test, and thus defects such as cracks and voids are formed between the layers of the heat dissipation substrate. A problem of cracks occurring in a semiconductor device. Due to such defects, there is a problem that the bonding strength is lowered and the heat conduction is lowered, and the semiconductor device is defective.

Therefore, an object of the present invention is to provide a heat dissipation substrate which can further improve long-term reliability.

The heat dissipation substrate of the present invention is a rectangular heat dissipation substrate in which a Cu layer and a metal A layer made of a metal A are alternately laminated, wherein the Cu layer and the metal A layer are stacked in a stack of 5 or 9 layers, and the metal layer A The ratio (L/t) of the average crystal grain length L in the longitudinal direction of the heat dissipation substrate to the average crystal grain length t in the thickness direction is 1.5 to 10.0, and the average crystal grain length L in the longitudinal direction and the aforementioned metal A layer The ratio (L/d) of the average crystal grain length d in the width direction of the heat dissipation substrate is 1.2 to 7.0.

According to the present invention, since the average aspect ratio (L/t) and (L/d) of the crystal grain length of the heat dissipating substrate are large, cracks can be suppressed from occurring between the layers, and cracks can be suppressed from occurring between the wafers, so that the long-term increase can be further improved. reliability.

10‧‧‧heated substrate

12A, 12B‧‧‧Cu layer

14‧‧‧Metal A layer

18‧‧‧ crystal grain

d‧‧‧Average length in the width direction

L‧‧‧Average length in the long-side direction

t‧‧‧Average length in the thickness direction

X‧‧‧Longside direction

Y‧‧‧Width direction

Z‧‧‧ Thickness direction

Simple description of the schema

Fig. 1 is a longitudinal sectional view showing the structure of a heat dissipation substrate of the embodiment.

Fig. 2 is a perspective view showing the structure of the heat dissipation substrate of the embodiment.

Fig. 3 is a schematic cross-sectional view showing the metal layer A in the longitudinal direction of the heat dissipation substrate of the embodiment.

Fig. 4 is a schematic cross-sectional view showing the metal layer A in the width direction of the heat dissipation substrate of the embodiment.

Form for implementing the invention

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

1. Implementation

(1) Overall structure

As shown in Fig. 1, the heat dissipation substrate 10 of the present embodiment is formed by alternately stacking 5 or 9 layers of Cu layers 12A and 12B and a metal A layer 14 made of metal A. In the case of this figure, the Cu layers 12A, 12B and the metal A layer 14 have a total of five layers. Metal A can use Mo or W of a low thermal expansion material. As shown in FIG. 2, the heat dissipation substrate 10 is formed in a rectangular plate shape. In the figure, the longitudinal direction is X, the width direction is Y, and the thickness direction is Z. The heat-dissipating substrate 10 is bonded to a semiconductor device (hereinafter, also referred to as a wafer) using a die attach material as a bonding material on one surface, for example, and is joined by a hard solder material on the other side surface. Insulating substrate.

(crystal size)

As shown in FIG. 3, the metal A layer 14 has an average length L of the longitudinal direction X of the heat dissipation substrate 10 of the crystal grains 18 of the metal A larger than the average length t of the thickness direction Z, that is, the crystal grain 18 of the metal A layer 14. The average aspect ratio (L/t) of the long side/plate thickness is large. Therefore, the metal A layer 14 can simultaneously suppress the adjacent Cu layers 12A, 12B by heating. Thermal expansion and heat shrinkage due to cooling. Specifically, the metal A layer is elastically deformed in the longitudinal direction X due to the large average length of the longitudinal direction X. Accordingly, it is assumed that the metal A layer 14 is elastically deformed in accordance with thermal expansion of Cu due to heating, and the adhesion between the Cu layers 12A and 12B and the metal A layer 14 (hereinafter also referred to as "interlayer") can be maintained. Further, the metal A layer 14 can maintain the adhesion between the layers of the Cu layers 12A, 12B and the metal A layer 14 by controlling the heat shrinkage of the Cu layers 12A, 12B during cooling, and suppress the occurrence of cracks between the layers.

The metal A has a high effect of suppressing thermal deformation of the Cu layers 12A and 12B due to a low coefficient of thermal expansion. On the other hand, since the thermal expansion coefficient of the metal A is greatly different from that of Cu, the difference in thermal expansion rate causes a crack between the layers. In the case of the present embodiment, by making the average aspect ratio (L/t) of the long side/thickness of the crystal grains 18 of the metal A layer 14 1.5 or more, the heat in the interlayer and the Cu layers 12A and 12B can be alleviated. stress. By setting the average aspect ratio (L/t) of the long side/thickness of the crystal grains 18 of the metal A layer 14 to 1.5 or more, in the case of using a high temperature solder as a die attaching material, the condition is strict thermal cycle. Cracks between the layers are suppressed, so high reliability can be obtained. When the average aspect ratio (L/t) of the long side/thickness of the crystal grains 18 of the metal A layer 14 is less than 1.5, the thermal expansion of the heating generated by the Cu layers 12A and 12B and the heat shrinkage of the cooling cannot be sufficiently suppressed. Therefore, it is impossible to suppress the crack growth between the layers. Further, when the average aspect ratio (L/t) of the long side/thickness of the crystal grains 18 of the metal A layer 14 exceeds 10.0, the effect of suppressing crack growth between the layers is saturated. The average length L of the longitudinal direction X of the crystal grains 18 of the metal A layer 14 and the average length t in the thickness direction are backscattered by electron beams in the cross section of the metal A layer 14 including the longitudinal direction X and the thickness direction Z. Observing by EBSD: Electron Backscatter Diffraction, The value of the average size of the cross section in the thickness direction of the crystal grains 18 was calculated by analyzing the software.

Further, as shown in FIG. 4, the average aspect ratio (L/t) of the long side/thickness of the crystal grains 18 of the metal A layer 14 is 1.5 to 10.0, and the heat dissipation substrate 10 of the crystal grains 18 of the metal A layer 14 is provided. The average length L of the longitudinal direction X is larger than the average length d of the width direction Y, that is, the average aspect ratio (L/d) of the long side/width of the crystal grains 18 of the metal A layer 14 is large, so that the heat dissipation substrate 10 can be suppressed. The overall thermal deformation is to reduce the damage to the wafer mounted on the heat dissipation substrate 10. This is considered because by making the average aspect ratio (L/d) of the long side/width large, deformation due to thermal expansion and thermal contraction generated by the adjacent Cu layers 12A, 12B can be reduced, and metal A in the long-side direction X can be reduced. The thermal strain of the layer 14 can further increase the strength and ductility.

That is, the heat dissipation substrate 10 has an average aspect ratio (L/t) of the long side/thickness of the crystal grains 18 of the metal A layer 14 of 1.5 to 10.0, and the long side of the crystal grain 18 of the metal A layer 14/ The average aspect ratio (L/d) of the width is 1.2 to 7.0, and the metal A layer 14 can effectively suppress thermal deformation caused by thermal expansion and thermal contraction of the Cu layers 12A, 12B. Therefore, the heat dissipation substrate 10 can suppress cracks between the layers and reduce the damage of the wafer, and can further improve the adhesion between the wafer and the heat dissipation substrate 10, thereby further improving long-term reliability.

By making the average aspect ratio (L/d) of the long side/width of the crystal grains 18 of the metal A layer 14 1.2 or more, it is possible to reduce the warpage of the heat dissipation substrate 10, etc. under the thermal cycle conditions where the temperature difference is severe. The wafer is cracked, so high reliability can be obtained. When the average aspect ratio (L/d) of the long side/width of the crystal grains 18 of the metal A layer 14 is less than 1.2, heating and cooling cannot be sufficiently suppressed. At the time, the Cu layers 12A and 12B are thermally deformed, so that cracks in the wafer cannot be suppressed. In addition, when (L/d) exceeds 7, the yield when processing and cutting to produce the heat-dissipating substrate 10 is lowered, and there is a problem that productivity is lowered. The average length d of the width direction Y of the crystal grains 18 of the metal A layer 14 and the average length d of the thickness direction Z are observed by EBSD in the cross section of the metal A layer 14 including the longitudinal direction and the thickness direction, and by The analysis software calculates the value of the average size of the cross section of the crystal grains 18 in the width direction.

The average aspect ratio (L/t) of the long side/thickness of the crystal grain 18 of the metal A layer 14 as described above is 1.5 to 10.0, and the average aspect ratio of the long side/width of the crystal grain 18 of the metal A layer 14 is as follows. The effect of (L/d) of 1.2 to 7.0 can be observed by repeating 500 times of the -40 ° C / 175 ° C thermal cycle after the current reliability test - 40 ° C / 200 ° C thermal cycle test to observe the interlaminar cracks and repeat The wafer crack was observed after 500 cycles of -40 ° C / 225 ° C heat cycle test, and was evaluated as an example of the test conditions of the heat cycle.

Preferably, by making the average aspect ratio (L/t) of the long side/thickness of the crystal grains 18 of the metal A layer 14 2.0 or more, even a longer thermal cycle can improve reliability, and even if it is repeated After 500 cycles of -40 ° C / 225 ° C thermal cycle test, the effect of suppressing interlayer cracks can also be obtained.

More preferably, by making the average aspect ratio (L/t) of the long side/thickness of the crystal grains 18 of the metal A layer 14 3.0 or more, the thermal cycle is performed under strict conditions such as assuming high temperature mounting of a SiC wafer or the like. The effect of highly inhibiting interlayer cracks can also be obtained. The effect when used in the high-temperature mounting aspect as described above can be evaluated by a test which repeats 1000 times -40 ° C / 225 ° C as a heat cycle test condition.

Further, preferably, if the average aspect ratio (L/d) of the long side/width of the crystal grains 18 of the metal A layer 14 is in the range of 2.0 to 7.0, even if 1000 times of heat of -40 ° C / 225 ° C is repeated. After the cycle test, the effect of highly inhibiting the cracking of the wafer can also be obtained.

(Cu layer thickness)

If the ratio (H ₁ /H ₂ ) of the thickness H _{1 of the} Cu layer 12B disposed on the surface to the thickness H ₂ of the Cu layer 12A disposed at the center is in the range of 0.2 to 0.9, the wafer is brazed to In the reliability test after mounting on the heat dissipation substrate 10, damage in the voids in the brazing material and cracks in the wafer can be reduced. By making the thickness H ₁ of the aforementioned Cu layer 12B on the surface side thin, the elongation and contraction of Cu generated during heating and cooling can be alleviated to reduce the stress applied to the hard solder material and the wafer.

That is, the average aspect ratio (L/d) of the long side/width of the crystal grains 18 of the metal A layer 14 is 1.2 or more, and the ratio (H ₁ /H ₂ ) of the thickness H ₂ of the Cu layer is further made In the range of 0.2 to 0.9, it is possible to obtain an effect of highly suppressing the void of the hard-weld material or the crack generated by the wafer in the reliability test of the wafer connected to the heat-dissipating substrate 10. By the suitability of H ₁ /H ₂ , it is particularly confirmed that voids generated by forces other than cracks are reduced in the brazing material. Even after repeating the heat cycle test of -40 ° C / 225 ° C for 500 times as a specific example, the effect of suppressing generation of voids in the brazing material can be obtained.

The heat dissipation substrate 10 has a ratio (H ₁ /H ₃ ) of the thickness H _{1 of the} Cu layer 12B disposed on the surface and the thickness H ₃ of the metal layer 14 connected to the Cu layer 12B of 2.0 to 8.0. The warpage caused by thermal deformation of the entire heat dissipation substrate 10 during high-temperature heating of the wafer and the monomer not mounted on the insulating substrate is suppressed. Since the warpage can be suppressed, as a result, in the step of heating at a high temperature of about 800 ° C in order to bond the wafer and the heat dissipation substrate 10 by the brazing material, it is possible to suppress generation of a gap between the hard solder material and the heat dissipation substrate 10, whereby The adhesion can be further improved to improve reliability. Even if the heat radiating substrate 10 is cooled by a normal temperature to 800 ° C and kept for 5 minutes as a specific example, the warpage of the heat radiating substrate can be reduced, and as a result, the effect of improving the adhesion to the wafer can be obtained.

A metal layer by crystal grains 14 of the longitudinal / thickness of 18 average aspect ratio (L / t) of 1.5 or more, and that the surface of the Cu layer thicknesses H ₁ and 12B of the central thickness of the Cu layer 12A of the H ₂ The ratio (H ₁ /H ₂ ) is 0.2 to 0.9, and the ratio (H ₁ /H ₃ ) of the thickness H _{1 of} the surface Cu layer 12B and the thickness H ₃ of the metal A layer 14 connected to the Cu layer 12B is further 2.0. Up to 8.0, the effect of suppressing the warpage of the heat dissipation substrate 10 can be further improved. This is because if the ratio (H ₁ /H ₃ ) is 2.0 or more, the thermal conductivity can be improved and the local concentration of the thermal strain can be alleviated. Further, when the ratio (H ₁ /H ₃ ) exceeds 8.0, the thermal deformation of the Cu layer 12B on the surface is large, and the damage to the wafer is large.

(alloy layer)

An alloy layer made of Cu and metal A is formed between the Cu layers 12A and 12B and the metal A layer 14. The alloy layer contains a region of 1 to 10 at% of Cu and metal A. Here, the reason why the concentration of the alloy layer is 1 to 10 at% is that the strength of the alloy is high in the concentration range, and the effect of improving the adhesion between the Cu layers 12A and 12B and the metal A layer 14 can be expected.

The position where the concentration of Cu and metal A in the center of the interlayer is 50 at% each is called a joint interface. The boundary between the alloy layer and the Cu layer 12A, 12B or the metal A layer 14 It is a position where the concentration changes by less than 1 at%. That is, the concentration of the metal A decreases from the bonding interface toward the Cu layers 12A and 12B, and the position at which the metal A concentration starts to be less than 1 at% is the boundary between the alloy layer and the Cu layers 12A and 12B. Similarly, the concentration of Cu decreases from the bonding interface toward the metal A layer 14, and the position at which the Cu concentration starts to be less than 1 at% is the boundary between the alloy layer and the metal A layer 14. Even if the concentration change at the side of the Cu layer 12A, 12B or the metal A layer 14 is 1 at% or more from the position where the concentration starts less than 1 at%, the above position is not changed as the alloy layer and the Cu layer 12A, 12B or the metal A layer. The boundary of 14.

Since the alloy layer is low in concentration and thin, the composition analysis is preferably an energy dispersive X-ray spectrometry (EDS) of a longitudinal section of a transmission electron microscope (TEM). This method is highly suitable for composition analysis of low concentrations of several at% in fine areas of several nm.

Actually, the EDS device attached to the TEM device was used, and the bonding interface was subjected to dot analysis in the thickness direction of the Cu layers 12A and 12B and the metal A layer 14 in the middle. The analysis is performed in the range of the thickness of the Cu layers 12A, 12B and the metal A layer 14 in the thickness direction of the bonding interface, in the range of about 500 to 1000 nm. Since the concentration sharply changes in the narrow range of 0 to 30 nm in the thickness direction of the Cu layers 12A, 12B and the metal A layer 14 from the bonding interface, the dot analysis is analyzed in detail at intervals of 2 nm, and the concentration changes little in the region exceeding 30 nm. The point analysis was performed at intervals of 10 nm.

The number of measurements per sample is preferably three or more. If the number of measurements is three or more, the reproducibility can be confirmed. In the case of the present embodiment, it is preferable to measure between two different layers in the four layers.

In the case of the present embodiment, the thickness of the alloy layer is 30 to 400 nm in total. Here, the "total thickness of the alloy layer" is formed between one of the plurality of layers of the heat dissipation substrate 10, and the total of the alloy layers on both sides is formed by the joint interface. By using an alloy layer containing Cu and metal A formed between the Cu layers 12A, 12B and the metal A layer 14, the thickness of the alloy layer containing Cu and metal A each having a concentration ranging from 1 to 10 at% is 30 to a total of 30 In the range of 400 nm, the adhesion between the layers can be improved. Thereby, it is possible to withstand the temperature rise and fall of the thermal cycle test or the like, the shear stress caused by the difference in thermal expansion between Cu and the metal A, and the occurrence of voids between the layers can be suppressed, so that long-term reliability can be improved. The voids between the layers are a cause of lowering the joint strength, and there is a problem that the long-term reliability is lowered. If the thickness of the alloy layer is less than 30 nm, the effect of suppressing the generation of voids is reduced. If the thickness of the alloy layer exceeds 400 nm, the thermal resistance of the alloy layer is large, so that heat conduction is lowered, and in addition, voids are generated due to the action of Kinknendall when forming an alloy layer between layers (hereinafter also referred to as Kogenda Hole) problem. Here, the KGunda action is a phenomenon in which the interface moves when two different metals are closely bonded and heated. The effect of the alloy layer as described above can be evaluated by repeating the test of 500 times - 40 ° C / 200 ° C as a thermal cycle test condition.

The alloy layer is preferably formed in a region of 80% or more between the layers. The adhesion between the Cu layers 12A, 12B and the metal A layer 14 can be further improved by forming an alloy layer in an area of 80%.

In the alloy layer, the thickness of the low concentration layer in which the concentration of Cu and the metal A is 1 to 3 at% each is preferably 20 to 300 nm. By making the thickness of the low concentration layer in the range of 20 to 300 nm, even for a long time or multiple times between layers The application of shear stress also suppresses the generation of voids, thereby improving long-term reliability in a thermal cycle environment. Regarding the effect of the low-concentration layer, it was confirmed that when the same temperature difference is repeated for a long time, an effect higher than when the temperature difference is large can be obtained. This is considered to be because the low concentration layer has the effect of relaxing the residual stress. The effect of the low-concentration layer as described above can be evaluated by repeating the test of 1000 times - 40 ° C / 200 ° C as a thermal cycle test condition.

Although the high-concentration layer having a concentration of 3 to 10% higher than the concentration of the low-concentration layer has an effect of improving adhesion, the concentration gradient of the high-concentration layer is large in a narrow region, and therefore it is difficult to increase heat only by using the high-concentration layer. Long-term reliability in a cyclic environment. Further, in order to form a high-concentration layer in a large area, there is a problem that a long-time high-temperature heat treatment is required to reduce productivity.

The thickness of the aforementioned Cu layers 12A, 12B and the aforementioned metal A layer 14 is preferably 0.5 to 2 mm in total. When the total thickness is within the above range, the overall heat conduction and thermal expansion can be controlled, and the performance can be stably achieved in a practical thermal cycle environment or a TCT (Temperature Cycle Testing) test or the like.

(2) Manufacturing method

Next, the heat dissipation substrate 10 of the present embodiment can be manufactured by superimposing a Cu plate and a metal A plate on each other and performing hot press processing. Hereinafter, a method of forming the metal A layer 14 having the crystal grains 18 of a desired size and a method of forming the alloy layer will be described in order.

(metal layer A)

Controlling the average crystal grain length of the long side direction X of the metal A of the metal A layer 14 The method of the L, the average crystal grain length t in the thickness direction, and the average crystal grain length d in the width direction includes a method of controlling in the manufacturing step of the metal A plate unit, and bonding the Cu plate and the metal A plate. The method of control in the step. The length L, t, and d are controlled by the rolling and heat treatment in the production of the metal A plate unit before joining, and the length L and d are effective in the step of cutting the original plate of the bonded Cu plate and the metal A plate. of. Since L and d are determined by the direction in the plane of the heat dissipation substrate 10, they can be adjusted by cutting the original plate into small pieces. Hereinafter, a method of controlling the metal L-plate lengths L and t before joining will be described.

In order to make the average aspect ratio (L/t) of the long side/thickness of the crystal grain 18 of the metal A layer 14 in the range of 1.5 to 10.0, the rolling workability of the metal A plate is increased, and at the same time, it is carried out during processing. The intermediate heat treatment temperature is lower than the recrystallization temperature. In order to increase the average aspect ratio (L/t) of the long side/plate thickness in the range of 1.5 to 10.0, it is preferred to carry out the intermediate heat treatment twice or more, and the intermediate heat treatment temperature is lower than the recrystallization temperature. On the other hand, by reducing the degree of processing of the rolling and making the intermediate heat treatment temperature higher than the recrystallization temperature, the average aspect ratio (L/t) of the long side/thickness can be less than 1.5. Therefore, the lengths L and t can be easily controlled by performing rolling after performing the intermediate heat treatment before cooling the sample.

In order to make the average aspect ratio (L/d) of the long side/width of the crystal grains 18 of the metal A layer 14 in the range of 1.2 to 7.0, it is controlled to pass the primary reduction ratio when rolling the sheet composed of the metal A, The rolling speed is such that the crystal grains 18 in the rolling direction are greatly grown, and it is effective to cut the rolling direction so as to be parallel to the longitudinal direction of the heat dissipation plate. That is, by making the rolling reduction in the range of 30 to 80% and controlling the rolling speed in the range of 0.3 to 10 m/min, the knot is made The growth of the crystal grains 18 coincides with the rolling direction and can be easily adjusted (L/d) in the range of 1.2 to 7.0. On the other hand, (L/d) may be less than 1.2 by reducing the rolling reduction ratio of the rolling or cutting the rolling direction so as to be the width direction of the heat dissipation plate. Further, (L/d) may exceed 7.0 by increasing the rolling reduction ratio of the rolling, increasing the number of passes, and lowering the intermediate heat treatment temperature.

Hereinafter, in the case where Mo is used as the metal A, the intermediate heat treatment is exemplified as a manufacturing step in which a 5 mm thick Mo plate is rolled to a predetermined thickness, and as an example of specific production conditions, that is, rolling is performed by rolling one pass. The shrinkage ratio is in the range of 30 to 70%, and the rolling speed is in the range of 0.5 to 3 m/min, and the average aspect ratio (L/d) of the long side/width can be adjusted. In terms of heat treatment, it may be divided into a heat treatment once in the range of 1 to 1.5 mm in thickness (first intermediate heat treatment), and a heat treatment in which the thickness is 0.1 to 0.3 mm or more (second intermediate heat treatment) ). Adjusting the average aspect ratio (L/t) of the long side/plate thickness to the range of 1.5 to 4, and adjusting the effective heat treatment conditions in the range of (L/d) to 1.2 to 7.0 is performed once the first intermediate heat treatment, 2 or more second intermediate heat treatments, and the temperature range is 700 to 1200 °C. The effective heat treatment conditions for the average aspect ratio (L/t) of the long side/plate thickness in the range of 4 to 10.0 are in the first intermediate heat treatment of 600 to 800 ° C or more, and the second intermediate heat treatment is 500 to 800 ° C. It is carried out twice or more in the temperature range. The original plate is cut into small pieces in a direction in which the rolling direction is parallel to the longitudinal direction of the heat-dissipating substrate 10, whereby a heat-dissipating substrate 10 having a predetermined size can be obtained.

The production conditions are not limited to those described above, and a desired average aspect ratio (L/t) or (L/d) can be obtained by adapting conditions such as temperature, length L, t, and heat treatment temperature.

(alloy layer)

As a method of forming an alloy layer between layers in the present embodiment, a method of forming an alloy layer between layers by controlling a temperature history and a pressure history when a Cu plate and a metal A plate are overlapped and performing hot press processing is controlled, and mass production is high. Industrial is also simple. Further, as another method, a thin Cu may be plated on the metal A layer 14 side before bonding, and then heat treatment may be performed to form a part of the alloy layer of Cu and metal A in advance.

In order to form an alloy layer containing Cu and metal A in a concentration range of 1 to 10 at% and a total thickness in the range of 30 nm to 400 nm between layers, it is necessary to appropriately condition the temperature history and the pressure history. One of the processing conditions is preferably to increase the rate of temperature rise in the low temperature region and to decrease the rate of temperature rise in the high temperature region. For example, it is effective to utilize 2-stage control at a boundary of about 500 °C. Specifically, the first temperature increase rate to 500 ° C is 30 to 80 ° C / min, and the second temperature increase rate at 500 ° C or higher to the final heating temperature is 20 to 80 ° C / min. Alloy layer. The final heating temperature here is in the temperature range of 850 to 1050 ° C and is maintained in this temperature range for 20 to 50 minutes.

Here, the temperature at which the temperature rise rate is changed is preferably in the range of 400 to 600 °C. This is because the temperature range of 400 to 600 ° C is close to the recrystallization temperature of Cu, and is softened by recrystallization, thereby promoting deformation and diffusion at the interface during pressurization. In addition, in addition to the above two-stage control, three-stage control can also be used, but mass production management is complicated.

Here, if the first temperature increase rate to 500 ° C is too slow, the alloy layer is formed into a granular shape and has a problem in quality. In addition, if the first heating rate is too fast, I am afraid that K-Dakong will be generated due to the difference in the diffusion speed of Cu and metal A. hole. When the second temperature increase rate of 500 ° C or more is too slow, the thickness of the alloy layer is not uniform. Further, when the second temperature increase rate is too fast, temperature unevenness occurs in the furnace, and the alloy layer concentration distribution is not uniform due to the location. Without being limited to the above-described temperature conditions, it is possible to industrially form a desired alloy layer as desired by recognizing the temperature history.

Further, the second temperature increase rate is preferably 10 ° C/min or more slower than the first temperature increase rate. When the difference between the second temperature increase rate and the first temperature increase rate is less than 10 ° C / min, it may be difficult to control the growth of the low-concentration alloy layer and the thickness thereof.

It is preferable to adjust the pressure history in conjunction with the above temperature history, and to increase the pressurizing pressure in the low temperature region and to lower the pressurizing pressure in the high temperature region. For example, it is effective to utilize 2-stage control at a boundary of about 500 °C. In order to obtain metal joining in a low temperature region, it is conceivable to promote interfacial deformation during pressurization, and excessively interfacial deformation is performed by softening Cu in a high temperature region, so as to have an effect of suppressing discontinuity of the alloy layer. The pressing pressure in the low temperature region is preferably in the range of 1.2 to 2 times the pressing pressure in the high temperature region. Specifically, the pressurization pressure to 500 ° C is in the range of 36 to 260 kgf / cm ² , and the pressurization pressure of 500 ° C or more is in the range of 30 to 130 kgf / cm ² , whereby it is easy to form a desired Alloy layer. Here, when the pressurization pressure is less than the lower limit value, the metal joining is insufficient and the metal layer is discontinuous. When the pressure exceeds the upper limit, cracks or the like are formed in the brittle metal layer A, and it is difficult to form the alloy layer stably. If the high temperature region exceeds the upper limit, the thickness of the Cu layers 12A and 12B is not uniform. Etc.

Further, the alloy layer can be formed at 80% between the layers by the above temperature history and pressurization history. On the other hand, in order to make the thickness of the low concentration layer It is effective to reduce pores, cracks, and the like in the alloy layer, and to appropriately change the temperature at the time of cooling and the pressurization pressure. It is preferable to lower the cooling rate in a high temperature region and increase the cooling rate in a low temperature region. In addition, it is preferred to also reduce the pressurization pressure in stages. Thereby, the cooling rate can be reduced in the high temperature region to alleviate the strain caused by the difference in thermal expansion to suppress the occurrence of cracks in the thin alloy layer. In addition, work efficiency can be improved by increasing the cooling rate in the low temperature region. Further, if the pressurization pressure is rapidly reduced at a high temperature, the low-concentration layer may be deformed by the action of thermal stress, and the thickness of the low-concentration layer may become uneven.

As a specific example of the temperature change and the pressurizing pressure at the time of cooling, in the case of cooling by the heating temperature of 850 to 1050 ° C, the cooling rate to 700 ° C is 10 to 30 ° C / min and the pressurizing pressure is 60 to 200 kgf / cm. Within the range of ² , the cooling rate below 700 ° C is 40 to 80 ° C / min and the pressurizing pressure is also useful for mass production in the range of 30 to 130 kgf / cm ² . It is preferred to have at least a cooling rate of 10 ° C / min and a pressurization pressure of 30 kgf / cm ² . Here, the change temperature is about 700 ° C, which is effective in improving interlayer adhesion. Although the details are not known, it is considered that the relationship between the strength, the ductility, and the internal diffusion behavior of the low concentration layer formed at the interface is defined by the vicinity of the temperature.

With respect to each raw material, the purity of Cu is preferably 99.3% or more from the viewpoint of thermal conductivity, and oxygen-free copper, refined copper, or the like can be used. As the Mo and W of the metal A, a commercially available raw material having a purity of 99.3% or more can be used. Further, in the use of the heat-dissipating substrate 10 requiring high strength, Cu, Mo, W or the like containing 5% or less of an additive element may be used.

2. Examples

(1) Sample

According to the procedure described in the above "Manufacturing Method", a heat-dissipating substrate having a 5-layer structure and a 9-layer structure was prepared as a sample. First, a Cu plate and a Mo plate or a W plate of a predetermined thickness are prepared. Next, a washing treatment for improving the adhesion at the joint interface is performed. In order to remove the oxide film on the Mo plate and the W plate, the aluminum plate is washed with hot water of about 50 ° C, and then the Cu plate is pickled by dilute sulfuric acid or the like. After washing, it is washed with water and dried. Finally, the Cu board and the Mo board or W were laminated and joined by hot press working to form the heat dissipating substrate (original board) of the examples and the comparative examples. The sample was cut out from the original plate by electric discharge machining so that the sample size was 20 mm in the longitudinal direction and 10 mm in the width direction. The prepared sample specifications are shown in Table 1.

(2) Evaluation

For the heat dissipation substrates of the examples and the comparative examples, the lengths of the Mo layer and the W layer in the thickness direction were observed by EBSD (Ultra 55 manufactured by Zeiss), and the lengths L and t of Mo and W were observed, and the average particle diameter was calculated by analyzing the software. . Further, concentration analysis of the joint interface was performed using a TEM (JEM-2100F, manufactured by JEOL Ltd.) and an EDS device (manufactured by JEOL Ltd., JED-2300T). The concentration analysis is performed in a manner in which the bonding interface is interposed in the vertical direction in the range of about 200 to 500 nm on both sides of the Cu layer and the Mo layer. The interval for performing the concentration analysis was basically performed at intervals of 10 nm. Further, in order to make the boundary of the alloy layer clear, the concentration region corresponding to about 1 to 10 at% of the low concentration region was analyzed in detail at intervals of 2 nm. The measurement of each sample was performed by EDS line analysis with more than 3 lines. Preferably, at two of the majority of the Cu layer and the Mo layer Different layers are measured on top.

The evaluation of long-term reliability is the implementation of the TCT test. The sample used was a sample in which a Cu electrode of an alumina DCB (Direct Copper Bond) substrate was bonded by a Ni alloy hard solder on one side of the heat dissipation substrate by Ag hard soldering of the Si wafer. The TCT test uses two test conditions with different heating temperatures. The TCT test conditions are the thermal cycling conditions under which the TCT test conditions (2) are stricter than the TCT test conditions (1). The temperature range is from -40 to +200 °C in the test conditions (1), and -40 to +225 °C in the test conditions (2).

The TCT test is carried out 500 times or 1000 times in the test condition (1) (range of -40 to +200 ° C), and is carried out 500 times under the test condition (2) (-40 to +225 ° C range). Or 1000 temperature rises and falls. After the TCT test, the cross section of the heat-dissipating substrate was observed and evaluated. The number of samples is two each. For cross-sectional observation, cross-section cutting and mechanical polishing of the heat-dissipating substrate are performed. The cross-sectional observation of the heat-dissipating substrate was carried out by selecting three different layers among the plurality of layers, and observing the voids and cracks between the layers of about 2 mm by SEM. According to the size difference, the pore size is less than 10 μm (mainly in the shape of a dot), and the crack is more than 10 μm (mainly linear).

The number of holes in the interlayer observation of the heat-dissipating substrate is investigated. If the number is 5 or less per 1 mm, the long-term reliability of the TCT test is good. Therefore, it is marked as ○ in Table 2, and in the range of 6 to 20. In the meantime, it is judged that there is no problem in practical use, but the quality is expected to be improved, and it is indicated by Δ mark in Table 2, and if it is 21 or more, there is a problem that reliability is lowered. Therefore, it is indicated as × mark in Table 2.

The number of interlayer cracks in the heat-dissipating substrate is investigated, if the number When the number is 5 or less per 1 mm, the long-term reliability of the TCT test is good. Therefore, it is indicated as ○ mark in Table 2, and if it is in the range of 6 to 15, it is judged that there is no problem in practical use, but the quality is expected to be improved. In Table 2, the symbol is denoted by Δ. If it is 16 or more, the reliability is lowered. Therefore, it is indicated by × in Table 2.

When the number of voids of 5 μm or more is investigated by cross-section observation, the long-term reliability of the TCT test is good when the number is 10 or less per 1 mm, so it is marked as ○ mark in Table 2. If it is in the range of 11 to 20, it is judged that there is no problem in practical use, but the quality is expected to be improved, and it is indicated as Δ mark in Table 2, and if it is 21 or more, the reliability is lowered, so in Table 2, Recorded as × mark.

Further, the heat-dissipating substrate which was not mounted was used as a sample, and the warpage caused by the thermal deformation of the heat-dissipating substrate after the high-temperature heating was evaluated. The number of samples is two each. The high-temperature heating condition was heated from room temperature to 800 ° C, and then kept at 800 ° C for 5 minutes, and then cooled in the atmosphere (test condition (3)). The warpage was evaluated by using a laser microscope (manufactured by Lasertec Co., Ltd., H1200) to measure the length E from the center to the end in the diagonal direction, and the height difference G between the center portion and the end portion, and G/E. As the warpage ratio. When the warpage ratio is less than 0.03%, the thermal deformation is small and is good. Therefore, it is marked as ○ mark in Table 2, and if it is in the range of 0.03 to 0.1%, it is judged that there is no practical problem, but the quality is expected to be improved. In the case of the Δ mark, if it exceeds 0.1%, there is a problem that the adhesion is lowered due to warpage, and therefore, it is denoted by X mark in Table 2.

(3) Results

As can be seen from Tables 1 and 2, Examples 1 to 24 have crystal grain lengths due to the metal A layer. The average aspect ratio (L/t) of the edge/plate thickness is 1.5 to 10.0, and the average aspect ratio (L/d) of the crystal grain length/width of the metal A layer is 1.2 to 7.0, and the TCT test condition (1) ( The result of the interlaminar crack observation after the installation of the cycle condition of 500 times was good, and the result of the wafer crack observation after the mounting of the TCT test condition (2) (circular condition 500 times) was good. On the other hand, in Comparative Examples 1 to 9, the average aspect ratio (L/t) of the crystal grain long side/sheet thickness of the metal A layer, or the average aspect ratio (L/d) of the crystal grain long side/width of the metal A layer. ) is less than the lower limit of the above range or exceeds the upper limit, so the observation result of the interlaminar crack after the installation of the TCT test condition (1) (circular condition 500 times) is ×, and after the installation of the TCT test condition (2) (circular condition 500 times) The wafer crack observation result is ×.

Further, in Examples 1, 2, 5, 6, 8, 9, 11, 12, and 14 to 23, since the above (L/d) was 2.0 or more, the TCT test condition (2) (circulation condition 1000 times) was installed. The wafer crack observation was good.

In Examples 1, 2, 4 to 12, and 14 to 24, the average aspect ratio (L/t) of the crystal grain long side/thickness of the metal A layer was 2.0 or more, and the TCT test condition (2) (circulation condition 500 times) The interlaminar crack observation after the installation was good.

Examples 1, 2, 5, 6, 8 to 12, 15 to 17, 19 to 23 have an average aspect ratio (L/t) of crystal grain length/sheet thickness of the metal A layer of 3.0 or more, and TCT test conditions ( 2) The results of observation of interlaminar cracks after installation (1000 cycles) were good.

8,10,11,19 Examples 1 to 21 and 24 due to the surface of the Cu layer and the thickness H ₁ of the central Cu layer thickness ratio H ₂ (H ₁ / H ₂₎ from 0.2 to 0.9, TCT test conditions (2 The results of the void observation in the hard-weld material after the installation (500 cycles) were good.

Examples 1, 2, 4 to 8, 10 to 14, 16, 19 to 24 have a ratio (H ₁ /H ₃ ) of the Cu layer thickness H _{1 of the} surface and the metal layer _thickness H ₃ connected to the Cu layer. From 2.0 to 8.0, the monomer warpage evaluation result of the test condition (3) (high temperature heating at 800 ° C) was good.

In Examples 1 to 4, 6 to 10, and 12 to 24, since the thickness of the alloy layer was 30 to _{400 nm in} total, the observation of the interlaminar pores after the installation of the TCT test condition (1) (cycle condition 500 times) was good.

Examples 2 to 4, 7 to 10, 12 to 20, and 22 to 24 have a low concentration layer thickness of 20 to 300 nm, and the TCT test condition (1) (circulation condition 1000 times) is good after the installation of the interlayer voids. .

3. Modifications

The present invention is not limited to the above embodiment, and can be appropriately modified within the scope of the gist of the invention.

10‧‧‧heated substrate

12A, 12B‧‧‧Cu layer

14‧‧‧Metal A layer

Claims (8)

A heat dissipating substrate is a rectangular heat dissipating substrate formed by alternately laminating a Cu layer and a metal A layer made of a metal A, wherein the Cu layer and the metal A layer are a total of 5 or 9 layers; and the metal A layer is The ratio (L/t) of the average crystal grain length L in the longitudinal direction of the heat dissipation substrate to the average crystal grain length t in the thickness direction is 1.5 to 10.0; and the average crystal grain length L in the longitudinal direction and the aforementioned metal layer A are The ratio (L/d) of the average crystal grain length d in the width direction of the heat dissipation substrate is 1.2 to 7.0. The heat dissipation substrate of a request, wherein the configuration on the surface of the Cu layer thicknesses H ₁ and H the thickness of the metal layer A is connected to the ratio of the Cu-layer (H ₁ / H _{3) 3} is 2.0 to 8.0. The requested item of the heat-dissipating substrate 1 or 2, wherein the Cu-layer is disposed on the surface of the thickness H ₁ and is arranged at the center of the thickness of the Cu layer of H ₂ (H ₁ / H ₂₎ from 0.1 to 0.9. The heat dissipation substrate according to claim 1 or 2, wherein an alloy layer composed of Cu and the metal A is formed between the Cu layer and the layer of the metal A layer; and the alloy layer contains 1 to 10 at% of the foregoing Cu or the foregoing Metal A, and the total thickness is 30 to 400 nm. The heat dissipation substrate of claim 4, wherein the alloy layer is formed in a region of 80% or more of the layer. The heat-dissipating substrate according to claim 4, wherein in the alloy layer, the concentration of the Cu or the metal A is 1 to 3 at%, and the low-concentration layer has a thickness of 20 to 300 nm. The heat-dissipating substrate of claim 1 or 2, wherein the thickness of the laminated Cu layer and the metal A layer is 0.5 to 2 mm in total. The heat dissipation substrate of claim 1 or 2, wherein the aforementioned metal A is Mo or W.