GB2099742A

GB2099742A - Bonding metals to non-metals

Info

Publication number: GB2099742A
Application number: GB8117314A
Authority: GB
Original assignee: Philips Electronic and Associated Industries Ltd
Current assignee: Philips Electronics UK Ltd
Priority date: 1981-06-05
Filing date: 1981-06-05
Publication date: 1982-12-15
Also published as: GB2099742B

Abstract

A ceramic substrate (1) is directly bonded between two copper members (2a, 2b) by placing the substrate between the copper members and heating in an appropriate atmosphere. To avoid blistering, channels (3a, 3b) and/or holes (6) are provided in the substrate or the copper members to allow gas to escape from between the components during bonding. <IMAGE>

Description

SPECIFICATION Bonding metals to non-metallic substrates This invention relates to a method of bonding a metallic member to a non-metallic substrate and further relates to an assembly of a metallic member bonded to a non-metallic substrate by such a method.

The bonding of metallic members to nonmetallic substrates is used in, for example, the semiconductor industry and more particularly in the manufacture of power semiconductor devices where an electrical conductor in the form of a metallic member is bonded to an electrically insulating substrate of a package in which the semiconductor device is mounted. In the case of power semiconductor devices it is often important that the substrate also provides a good thermal heat path to facilitate heat transfer to the atmosphere or to an external heat sink. The material of the substrate should, therefore, be a good thermal conductor. A ceramic material such as, for example, alumina or beryllia may be chosen for the material of the substrate because it is a good electrical insulator and a good thermal conductor.

Various methods of bonding metallic members to non-metallic substrates are known in which an intermediary layer is used in order to obtain a strong bond. However, such an intermediary layer is undesirable in many applications as it presents certain disadvantages.

For example, the presence of an intermediary layer generally hinders heat transfer from the metallic member to the substrate. Furthermore, in high frequency applications the current flow occurs in a thin skin at the boundary between the metal and the substrate. In such applications an intermediary layer is undesirable because the material of such a layer is generally more resistive than the metallic member. An additional disadvantage is that intermediary layers tend to be susceptible to corrosion which may lead to premature failure of the semiconductor device.

It is known that these disadvantages can be overcome by bonding a metallic member to a nonmetallic substrate so that the bond between them is less than a few tens of nanometres thick.

Such a method of bonding metallic members to non-metallic substrates is described in U.K.

Patent Specification No. GB 2,059,323. In particular that Patent Specification describes a method suitable for bonding a metallic member to a ceramic substrate. In that case a layer of metal is deposited on a surface of the substrate, after which the metal layer is oxidized throughout its thickness to form a metal oxide layer. The metallic member is then placed on the substrate so that the oxide layer is in direct contact with the metallic member. A heating step is carried out to form between the substrate and the metallic member a eutectic comprising the material of the metallic member and the oxide layer, after which the assembly is cooled with the member bonded to the substrate.

Unfortunately it has been found that this known method, as well as other known methods of bonding metallic members to non-metallic substrates, suffer from the drawback that voids (which are, in essence, bubbles or localized pockets of gas) can form in the eutectic during bonding. This phenomenon, which may be observed as blisters on the surfaces of the bonded elements, is problematic because it not only weakens the bond but also it can seriously hinder heat dissipation from the metallic member to the substrate. In fact, the problem of voids becomes more acute as the area of the elements to be bonded increases. This is unfortunate because, although it is more economical to bond large-area elements together and then to divide the bonded assemly into smaller units of the desired size, the cost-saving advantage is very much eroded by the reduced yield of usefully bonded units.

The problem of voids has troubled other workers in this art. U.S. Patent No. 3,911,533 relates to a method of bonding a metallic member to a ceramic substrate by curving the metal and allowing it to unroll as it becomes more ductile at elevated temperatures. Although this method goes some way to reducing voids it is relatively cumbersome and difficult to carry out effectively.

Also, it is difficult to position the copper member accurately. In addition, this method is restricted to the bonding of a single metallic member to a single substrate at any one time.

According to the present invention a method of bonding an element in the form of a metallic member to another element in the form of a nonmetallic substrate, including the steps of placing the metallic member substantially flat on the substrate, heating the elements to form therebetween a eutectic comprising the material of the metallic member, and cooling the elements to bond one to the other, is characterized in that at least one of the elements is provided with means which, when the metallic member is placed on the substrate, defines a passage for gas to escape from between said elements.

By providing gas escape means any bubbles or voids which would otherwise form between the elements during bonding tend to be avoided. Thus the present invention provides a method which is not only more straightforward than the previous attempt at solving the problem of voids as mentioned above, but which also can increase the yield of usefully bonded elements.

The gas escape means may be a hole or a channel in the metallic member or the substrate. If the assembly of bonded elements is to be severed to form smaller units, the severing operation is facilitated when at least a part of the gas escape means coincides with the line along which the assembly is to be severed.

A method in accordance with the invention is particularly suitable for bonding simultaneously a sandwich-like assembly of a non-metallic substrate between two metallic members, or a metallic member between two non-metallic substrates. In both of these cases the two outermost elements act to prevent bowing of the assembly due to any mismatch in thermal expansion of the different elements. Preferably the two outermost members are made of the same material, but they may be made of different materials with similar degrees of thermal expansion. When bonding such a sandwich-type assembly the same or different gas escape means may be provided at the opposite surfaces of the middle element.To assist later severing of the bonded assembly, however, it is preferable that, in the case where the middle element is the nonmetallic substrate, channels are provided at two opposite major surfaces of the substrate and that the channels at one surface are in registration with those at the opposite surface.

An embodiment of the invention will now be described with reference to the accompanying drawing in which the single figure is an isometric view of three elements to be bonded by a method in accordance with the invention.

It is to be noted that for the sake of clarity the various features of the Figure are not in proportion to each other.

The following method is concerned with bonding simultaneously an element in the form of a ceramic substrate 1 between two other elements in the form of copper members 2a, 2b.

The elements may all be of approximately the same dimensions, for example they may be square having lateral dimensions of 5 cm. and having a thickness of 0.25 mm. Because the outer elements are both copper members 2a, 2b and because they both have the same dimensions the difference in thermal expansion of copper and the ceramic does not cause bowing of the assembled elements when they are heated during bonding.

The ceramic substrate 1 may be made of, for example, alumina. Channels 3a which are mutually orthogonal are formed at the major surface 4 of substrate 1 and similar channels 3b are formed in registration with channels 3a at the opposite major surface 5. In this case the channels divide the surfaces of the substrate into four equal areas. To allow the efficient escape of gas from between the elements during bonding, it is desirable that the channels are at least 10 micrometres deep. However, the channel depth is preferably between 15 and 20 per cent of the thickness of the substrate, particularly for substrate thicknesses of between 0.25 and 0.63 mm. This range of depths reflects the following compromise. On the one hand, the channel is sufficiently deep that it facilitates cracking of the substrate when the bonded assembly is later divided into smaller units.On the other hand, the channel is sufficiently shallow that the bonded assembly can be transported and handled without it cracking apart. In the present case, where the thickness of the alumina substrate 1 is 0.25 mm. the channel depth may be, for example, 45 micrometres and the width may be, for example, 0.02 to 0.1 mm. The channels may be cut using a Nd :YAG or CO2 laser. It is noted here that the pulse frequency must be sufficiently high to form a continuous channel as opposed to a series of discrete indentations in the substrate.

After forming the channels a thin copper layer is deposited on the surfaces 4 and 5 of alumina substrate 1, for example by evaporation or sputtering. The copper layers are, for example, only 60 nanometres thick and, for the sake of clarity, they are not shown in the drawing. The copper layers are then oxidized throughout their thickness by heating the coated substrate in air at 4000C until all the copper is converted to cupric oxide. With the thickness of copper mentioned above, a time of about 30 minutes is adequate.

The alumina substrate 1 is then placed flat between two copper members 2a, 2b so that the cupric oxide layers formed on the opposite major surfaces 4, 5 of the substrate 1 are in direct contact with the copper members 2a, 2b. This assembly is then placed in a furnace and heated to a temperature of, for example, 1 ,0700C for about 5 to 10 minutes. The temperature must remain below the melting point of the metallic member, in the case of copper 1 ,083 or, as otherwise the metallic member loses its structural integrity and liquid drops are formed. In practice the Applicants have used copper which has not been freed of oxygen, such as commercially available C101/102, from the copper members 2a, 2b.With this type of copper successful bonding has been achieved when the heating step is carried out in an atmosphere in which the amount of oxygen ranges between 0.5 parts per million (p.p.m.) and 100 p.p.m. However, when a purer form of copper is used it appears that strong bonding is achieved by heating in an atmosphere containing a greater concentration of oxygen. Thus, in general it may be said that the limits of the range of the amount of oxygen in the atmosphere during this heating step are determined by the purity of the metal used. Under the conditions specified above the cupric oxide dissociates to cuprous oxide and the cuprous oxide and the copper form a eutectic which melts at 1 ,0650C. The molten eutectic forms a uniformly thick layer which wets the surfaces of the alumina substrate 1 and the copper members 2a, 2b.It may be desirable to provide some pressure on the system by applying a weight in the form of, for example a 200 g. tile on top of the assembly. The assembly is then cooled and the copper members 2a, 2b are strongly bonded to the alumina substrate 1. Bond strengths of, for example, 40 to 50 MPa can thus be obtained. The channels 3a and 3b define a passage for gas to escape from between the copper members and the substrate so that bubbles or voids do not form between the substrate 1 and the copper members 2a, 2b.

If desired, before dividing the assembly into smaller units, one of the copper members may be shaped using conventional masking and etching techniques to form conductor patterns. During this etching treatment the channels at the surface of the substrate facing the copper member being shaped can be exposed to facilitate division of the assembly. By shaping after bonding the copper member can be positioned to a relatively high degree of accuracy. Alternatively one, or for that matter both, of the copper members may be shaped before bonding.

The bonded assembly can now be divided into four equally sized smaller units by severing it along lines which coincide with the channels 3a and 3b. In fact the channels are sufficiently deep that the smaller units can be formed simply by cracking the bonded assembly.

It is not necessary to form channels only where the assembly is later to be severed. On the contrary, the channels can be formed elsewhere and it has been found that such channels have no significant influence on the thermal efficiency of the assembly. In fact, when bonding larger area elements, the channels along which the substrate is to be cracked may be so far apart that it is preferable to provide additional channels to further reduce the risk of void formation.

As indicated previously, instead of (or as well as) forming channels in the substrate, holes 6 may be formed in one or both of the copper members, as shown in member 2b in the Figure.

Alternatively channels can be formed at the facing surfaces of the copper members 2a, 2b. As with the channels in the substrate these holes or channels may or may not coincide with the line along which the assembly is severed in the case where smaller units are to be formed. So far, the method has been described in terms of bonding a sandwich-type assembly comprising a ceramic substrate between two copper members. Clearly the method can also be used for bonding a metallic member between two ceramic substrates.

Moreover, a method in accordance with the invention can, of course, be used for bonding a single ceramic substrate to a single metallic member. In this case the gas escape means may be a channel in the inwardly-directed surface of either element, a series of holes in either element, or a combination of these.

It will be clear to a person skilled in the art that other metallic members may be bonded to other non-metallic substrates using a method in accordance with the invention. For example, the material of the substrate may be another ceramic such as beryllia and other metals for the metallic member may be, for example iron, chromium or silver. Furthermore, the substrate need not be ceramic, but may be, for example a single crystal substrate.

Claims

1. A method of bonding an element in the form of a metallic member to another element in the form of a non-metallic substrate, including the steps of placing the metallic member substantially flat on the substrate, heating the elements to form therebetween a eutectic comprising the material of the metallic member, and cooling the elements to bond one to the other, characterized in that at least one of the elements is provided with means which, when the metallic member is placed on the substrate, defines a passage for gas to escape from between said elements.

2. A method as claimed in Claim 1, in which the gas escape means is at least one hole in the metallic member.

3. A method as claimed in Claim 1, in which the gas escape means is at least one hole in the nonmetallic substrate.

4. A method as claimed in Claim 1, in which the gas escape means is at least one channel in the metallic member.

5. A method as claimed in Claim 1, in which the element in the form of a non-metallic substrate is bonded between two other elements in the form of metallic members, in which method the substrate is placed between the two metallic members, the elements are heated to form therebetween a eutectic comprising the material of the metallic members, the elements are cooled to bond the metallic members to the substrate, and at least one of the elements is provided with means which, when the substrate is placed between the members, defines a passage for gas to escape from between the elements.

6. A method as claimed in Claim 1 or Claim 5, in which the gas escape means comprises at least one channel in the substrate.

7. A method as claimed in Claim 5 or Claim 6 when appendant from Claim 5, in which the gas escape means is constituted by at least one channel at one major surface of the substrate and at least one channel at the opposite major surface of the substrate, the channels at each of said surfaces being formed in registration with each other.

8. A method as claimed in Claim 6 or Claim 7, in which the channel depth is at least 10 micrometres.

9. A method as claimed in any of Claims 6 to 8, in which the channel depth is at most 20% of the substrate thickness.

10. A method as claimed in Claim 9, in which the channel depth is at least 15% of the substrate thickness.

11. A method as claimed in any of Claims 6 to 10, in which the channel in the substrate is formed using a laser.

12. A method as claimed in any of the preceding Claims, in which the bonded elements are severed along a line which coincides with at least part of the gas escape means.

13. An assembly of a metallic member bonded to a non-metallic substrate by a method claimed in any of the preceding Claims.

14. A method of bonding a metallic member to a non-metallic substrate substantially as herein described with reference to the accompanying drawing.