Abstract
A method of integrating a first substrate having a first surface with a first insulating material and a first contact structure with a second substrate having a second surface with a second insulating material and a second contact structure. The first insulating material is directly bonded to the second insulating material. A portion of the first substrate is removed to leave a remaining portion. A third substrate having a coefficient of thermal expansion (CTE) substantially the same as a CTE of the first substrate is bonded to the remaining portion. The bonded substrates are heated to facilitate electrical contact between the first and second contact structures. The third substrate is removed after heating to provided a bonded structure with reliable electrical contacts.
Claims
1. (canceled)
2. A method comprising: providing a first substrate having a first coefficient of thermal expansion, the first substrate having a first bonding surface; providing a second substrate having a second coefficient of thermal expansion that is different than the first coefficient of thermal expansion, the second substrate having a second bonding surface; directly bonding the first bonding surface of the first substrate to the second bonding surface of the second substrate; thinning the second substrate; bonding the second substrate to a third substrate; heating the first, second, and third substrates after directly bonding the first bonding surface to the second bonding surface; and removing the third substrate from the second substrate.
3. The method of claim 2, wherein the thinning of the second substrate occurs after the directly bonding the first and second bonding surfaces.
4. The method of claim 2, wherein the directly bonding comprises directly bonding together the first and second substrates at room temperature.
5. The method of claim 4, wherein the heating comprises bringing the first, second, and third substrates to a temperature of at least 150 C.
6. The method of claim 5, wherein the heating comprises bringing the first, second, and third substrates to a temperature in a range of 75 C. to 150 C.
7. The method of claim 2, wherein the first substrate is formed substantially of a first semiconductor material and the second substrate is formed substantially of a second semiconductor material different from the first semiconductor material.
8. The method of claim 2, wherein: the first substrate comprises a first dielectric layer that defines the first bonding surface, and first metal contact structures are disposed in the first dielectric layer; the second substrate comprises a second dielectric layer that defines the second bonding surface, and second metal contact structures are disposed in the second dielectric layer; and the heating causes at least one of the first metal contact structures to expand and metallically bond with one of the second metal contact structures.
9. The method of claim 2, wherein the directly bonding comprises directly bonding first metal contact structures of the first substrate to second metal contact structures of the second substrate without an adhesive.
10. The method of claim 9, wherein the directly bonding comprises directly bonding a first nonconductive region of the first substrate to a second nonconductive region of the second substrate without an adhesive.
11. The method of claim 2, wherein the bonding the second substrate to the third substrate comprises directly bonding the second substrate to the third substrate without an adhesive.
12. The method of claim 2, further comprising heating the first, second, and third substrates while the third substrate is bonded to the second substrate.
13. The method of claim 2, wherein removing the third substrate from the second substrate is performed after the heating.
14. A method comprising: directly bonding a first bonding surface of a first substrate to a second bonding surface of a second substrate, the first substrate having a first coefficient of thermal expansion that is different from a second coefficient of thermal expansion of the second substrate; thinning the second substrate; and bonding the second substrate to a third substrate.
15. The method of claim 14, further comprising removing the third substrate from the second substrate after said heating.
16. The method of claim 14, wherein bonding the second substrate to the third substrate comprises directly bonding the second substrate to the third substrate without an adhesive.
17. The method of claim 14, wherein the directly bonding comprises directly bonding first metal contact structures of the first substrate to second metal contact structures of the second substrate without an adhesive.
18. The method of claim 17, wherein the directly bonding comprises directly bonding a first nonconductive region of the first substrate to a second nonconductive region of the second substrate without an adhesive.
19. The method of claim 18, further comprising heating the first, second, and third substrates.
20. The method of claim 19, wherein the heating causes at least one of the first metal contact structures to expand and metallically bond with one of the second metal contact structures.
21. The method of claim 19, wherein the heating comprises bringing the first, second, and third substrates to a temperature of at least 150 C.
22. The method of claim 21, wherein the heating comprises bringing the first, second, and third substrates to a temperature in a range of 75 C. to 150 C.
23. The method of claim 14, wherein the first substrate is formed substantially of silicon and the second substrate is formed substantially of a material different than silicon.
Description
BRIEF DESCRIPTION OF THE DRAWINGS
[0012] A more complete appreciation of the present invention and many attendant advantages thereof will be readily obtained as the same becomes better understood by reference to the following detailed description when considered in connection with the accompanying drawings, wherein:
[0013] FIG. 1 is a diagram of two wafers with a major portion and a direct metal bond portion.
[0014] FIG. 2 is a diagram of two wafers with a major portion and a direct metal bond portion aligned and placed together forming a bond interface.
[0015] FIG. 3 is a diagram of two wafers with a major portion and a direct metal bond portion aligned and placed together forming a bond interface with a substantial portion of the major portion of one of the two wafers removed resulting in a thinned portion of a direct metal bonded pair.
[0016] FIG. 4 is a diagram of a stiffening wafer attached to the thinned portion of a direct metal bonded pair.
[0017] FIG. 5 is a diagram of a thinned portion of a direct metal bonded pair after removal of a stiffening wafer.
[0018] FIG. 6 is a diagram of a second embodiment including filled vias.
[0019] FIG. 7 is a diagram of two wafers bonded each having filled vias.
[0020] FIG. 8 is a diagram of showing the structure of FIG. 7 with the vias exposed.
[0021] FIG. 9 shows the attachment of an additional substrate to the structure of FIG. 8.
[0022] FIGS. 10A-10C are detailed views of the metal bonding region.
[0023] FIG. 11 is a diagram of a wafer bonded with a clamp.
[0024] FIG. 12 is another diagram of a wafer bonded with a clamp.
[0025] FIG. 13 is a diagram of bonding using a flexible container.
[0026] FIG. 14 is a diagram of applying pressure to the flexible container of FIG. 13.
[0027] FIG. 15 is a diagram of bonding using a flexible container.
[0028] FIG. 16 is a diagram of applying pressure to the flexible container of FIG. 15.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0029] Referring now to the drawings, in particular FIG. 1, a first embodiment of the method according to the invention will be described. It is noted here that the drawings are not drawn to scale but are drawn to illustrate the concepts of the invention.
[0030] Two wafers, 1 and 2 are prepared for bonding. The wafers are of different material, and have different CTE. Wafer 2 includes a major portion 6 and a direct metal bond portion 5. Direct metal bond portion 5 has a surface with insulating and metal portions. The insulating portions are preferably an oxide or nitride, and more preferably a silicon oxide or silicon nitride. The portion 5 is shown in more detail in FIGS. 10A-10C. Metal pads 20 may be below, flush or above the oxide or nitride material 21 depending upon the process conditions and desired configuration. In the case where chemo-mechanical polishing is used on the surface, the metal pads can be dished and have a surface below the surface of the oxide or nitride material or the oxide or nitride material can be dished and have a surface below the surface of the metal pads.
[0031] Major portion 6 can include substrate, device, and interconnect portions that are, for example, found in industry standard manufactured semiconductor wafers, such as CMOS wafers that typically are manufactured with a copper or aluminum back-end-of-line process. Wafer 1 includes a major portion 3 and a direct metal bond portion 4. Major portion 3 can include substrate, device, and contact portions that are, for example, found in industry standard gallium nitride-based hetero-epitaxial device structures grown on sapphire (GaN/sapphire) that have contacts formed to the hetero-epitaxial material.
[0032] Wafer 1 and wafer 2 are direct metal bonded as described in application Ser. Nos. 09/505,283, 10/359,608 and 11/201,321, as shown in FIG. 2. If the surfaces of metal portions are below the surfaces of insulating portions, only the insulating portions may be in direct contact at interface 7 after the wafers are first placed into contact. The bonded wafers may then be heated to increase the bond energy between the bonded insulating portions, but not at too high a temperature for CTE induced strain to break the bond between the insulating portions or break the bonded wafers. The optimum temperature to increase bond energy will depend on the CTE difference and thickness of the wafers bonded. For example, when bonding a GaN/sapphire structure of approximate range 500-1000 micron sapphire thickness to silicon CMOS of approximate range 500-750 micron thickness, a temperature in the range of 75150 C. may be preferable to achieve a bond energy of greater than 1 J/m.sup.2 and preferably greater than 2 J/m.sup.2. Higher temperatures are possible if a thinner material is used or materials with a lower CTE difference are used to facilitate achieving even higher bond energies greater 2.5 J/m.sup.2. Although this temperature range can be sufficient to achieve a very high bond energy, it may not be sufficient to form 3D interconnections depending on the relative height of the metal and insulating portions and the type of metal used. For example, if copper is used, a temperature range of 150-250 C. may be required if the copper is 0-10 nm below a silicon oxide insulating portion. Alternatively, if nickel is used, a 250-350 C. temperature range may be required to make 3D interconnections if the nickel is 0-10 nm below a silicon oxide insulating portion. The lower temperature range requirement of the copper compared to the nickel is an example of where the type of metal can affect the temperature range wherein the higher expansion coefficient of copper (17 ppm/ C.) compared to that of nickel (13 ppm/ C.) results in more expansion at a given temperature resulting in a lower temperature range for a given difference in height between the metal and insulating portions of the bond surface Higher temperatures may thus be required to facilitate electrical interconnections while higher temperatures may not be possible with this configuration of bonded wafers due to CTE induced strain that would break the bond between the insulating portions or break the bonded wafers.
[0033] Major portion 3 is then thinned as shown in FIG. 3 to form thinned layer 8 having a thickness typically in the range of 1-10 microns. The thickness of layer 8 may be outside this range depending upon the application and materials. For example, bonded material combinations with a low CTE mismatch of <2 ppm/ C. may allow a thicker layer 8 in the range of 10-100 microns and applications requiring the transfer of layers less than one micron may use a thinner layer 8 of 0.10-1.0 microns. Thinning may include one or a combination of backgrinding, polishing, etching, or laser liftoff. For example, if wafer 2 is a GaN/sapphire structure, laser liftoff can be used to remove the sapphire resulting in a very thin GaN device layer with metal contact portions. The thinned layer 8 allows heating to a higher temperature without breaking the bond between the insulating portions or breaking the bonded wafers due to increased compliance or elasticity. The allowed increased temperature depends on the materials and the reduced thickness of layer 8. For example, temperatures in excess of 350 C., for example 350 C. to 400 C., can be enabled by this thinning for bonded material that have a high CTE mismatch of 2-5 ppm/ C. and a layer 8 thickness of 2-20 to microns. This increased temperature range enabled by reduced layer 8 thickness may be suitable to enable 3D interconnections or for other processing, for example oxide deposition or annealing. It is not necessary to use the full extent of this increased temperature range for other processing. For example, other processing below the maximum increased temperature range that is higher than that allowed prior to increasing the temperature range is possible.
[0034] In some cases, layer 8 may be too thin to provide adequate stiffness to produce adequate compression between metal portions at the surface of wafers 1 and 2 to form reliable 3D interconnections if wafers are heated to facilitate electrical interconnections. For example, if layer 8 is in the range of 1 to 10 microns thick, with an upper portion of this layer, for example 0.2 to 2.0 microns, comprising a heterogeneous combination of insulative and conductive bonding material, considerable stress normal to the bond interface in the vicinity of the interface between the insulating and conductive bonding material can be generated at low temperatures, for example less than 300 C., due to the CTE difference between insulative and conductive bonding material. This normal stress can distort the thin layer, resulting in less compressive force between metal portions and preventing electrical interconnections across the bond interface. This distortion results from a CTE mismatch induced extrusion of the conductive bonding material relative to the insulating bonding material at the thinned surface that is not constrained by the thinned layer due to the reduced stiffness of the thinned layer compared to that without partial or total removal of the substrate.
[0035] This reduced stiffness can be compensated by bonding a third wafer 9 to thinned major portion 8 to reduce or prevent the distortion of layer 8 and enable adequate compression between metal portions at the surface of wafers 1 and 2 to form 3D interconnections with heating after the bonding of third wafer 9 as shown in FIG. 4. The minimum thickness of third wafer 9 required can be determined experimentally, however, this thickness will typically be less than a standard wafer thickness, for example 50-100 microns for 100-300 mm diameter wafers with standard thickness of about 0.5-0.8 mm as the stiffness is increased substantially with relatively small thickness being dependent on the cube of the thickness of third wafer 9. A reduced thickness of third wafer 9 may be obtained by thinning third wafer 9 before or after attaching. A wafer 9 thickness larger than the minimum thickness, for example a standard wafer thickness, may also be used.
[0036] The attachment of third wafer 9 can be with a variety of methods, for example with a direct bond, as described in application Ser. No. 09/505,283, or a clamp 15 as shown in FIG. 11. If with a direct bond, the attachment may include the addition of bonding layers on wafer 9 and/or thinned portion 8. If with a clamp, clamp 15 is shown in FIG. 11 with the external pressure applied to both sides of wafer stack represented by the arrows. This can be done by applying external pressure from both sides as shown in FIG. 11 or by applying external pressure from one side with opposing side restrained by a chuck 16 as shown in FIG. 12. Third wafer 9 preferably has a CTE comparable to wafer 2 to prevent excessive stress during subsequent heating to form 3D interconnections. For example, if wafer 1 is GaN/sapphire, and wafer 2 is silicon CMOS, third wafer 9 can be silicon. The workable range of the difference in CTE depends on the materials, their area and their thicknesses. For example, when bonding 200 mm diameter silicon wafers and using silicon oxide and copper as insulating and conductive bond materials, respectively, it is preferable to have a difference in CTE less than 0.5 ppm/ C. When working with larger wafers, for example 300 mm wafers, it is preferable to have a smaller difference in CTE less than 0.3 ppm/ C. and when working with smaller wafers, for example 200 mm wafers, it is possible to have a larger difference in CTE less than 1.0 ppm/ C. When working with bond materials with a smaller difference in CTE, for example silicon oxide and nickel, it is preferable to have wafers with a smaller difference in CTE to allow for relatively more heating.
[0037] A flexible clamping arrangement as shown in FIGS. 13-16, can be used to assist in holding the bonded wafers together during heating while in addition accommodating bowing of the wafers which will naturally occur during heating due to the CTE difference of the wafers. This method can be applied to the first bond of two wafers of FIG. 2 with different CTEs (shown with flexible clamp methods in FIGS. 13 and 14), and also to the 3 layer stack of FIG. 4 with two thick CTE matched materials bonded to the thinner layer of different CTE material sandwiched between the two thicker layers (shown with flexible clamp methods in FIGS. 15 and 16). The flexible clamping arrangement consists of 2 layers of flexible material, for example a silicone rubber sheet, enclosing the bonded materials, and sealed at the edges forming a sealed envelope 16. The material used must be able to withstand the temperature that is to be applied to the bonded layers. Pressure is applied either by evacuating the space within the flexible material envelope, thus applying atmospheric pressure evenly around the outside of the structure (FIGS. 13, 15), and/or by putting the flexible envelope and its contents into a high pressure chamber 17, and applying the desired pressure evenly around the envelope and its contents to compress the layers together (FIGS. 14, 16). In both cases, heat is then applied to the envelope and its contents to strengthen the bond(s) while bowing of the bonded stack is allowed to accommodate CTE mismatch and prevent breaking of the layers.
[0038] The bonded stack of wafers 1, 2, and 3 shown in FIG. 4 can then be heated to higher temperatures than previously to form 3D interconnections if interconnections have not already been made due to a lack of temperature. For example, 300-350 C. may be required if wafer 1 is GaN/sapphire, wafer 2 is silicon CMOS, wafer 3 is silicon, and the metal is nickel. This ability to heat to a higher temperature has been enabled by replacing a major portion of wafer 2 that is CTE mis-matched to wafer 1 with wafer 9 that is CTE matched to wafer 1. These higher temperatures are now possible to facilitate electrical interconnections with higher temperatures due to this reduction in CTE mis-match. These higher temperatures were previously not possible due to CTE induced strain that would break the bond between the insulating portions or break the bonded wafers.
[0039] After heating, wafer 3 can be removed as shown in FIG. 5, to leave portion 10. Portion 10 may be essentially thinned portion 8 or may be thicker, for example if a bonding layer portion of wafer 9 is not removed or may be thinner, for example if a bonding layer portion of thinned portion 8 is removed. Removal can be with a variety of methods, for example, one or a combination of backgrinding, chemo-mechanical polishing, or etching. Such techniques may be used when wafer 3 is silicon. Wafer 3 may also not be removed, for example if wafer 3 has a useful function, for example as part of a packaging function. A second embodiment of the method according to the invention will now be described. Either one or both of wafer 11 and wafer 12 may contain a via or vias 13 filled with metal that extend through all, most, or a portion of either one or both of wafer 11 and wafer 12, respectively as shown in FIGS. 6 and 7 for filled vias 13 and 14 that extend through about half of wafer 11 and wafer 12, respectively, and at a larger pitch than the conductive material at the bond interface. The vias may be electrically connected to the conductive material at the bond interface. If both wafer 11 and wafer 12 have filled vias, the filled vias may be opposed during the alignment and placement of wafer 1 and wafer 2 together as shown for filled vias 13 and 14, respectively. Vias 13 and 14 may also be not opposed.
[0040] After wafer 12 is thinned, vias may be exposed as shown in FIG. 8 or may have a very thin residual portion of wafer 1 preventing exposure. For example, the residual wafer 12 thickness may be less than 100 microns. Heating of the structure in FIG. 8 or a similar structure with residual portion on filled vias will result in additional vertical and horizontal stress than described previously for the metal bonding due to the expansion coefficient difference between the filled vias surrounding material. The lack of a wafer 3 to inhibit relaxation of this stress can distort the direct metal bond interface and prevent proper bonding. The attachment or bonding of a wafer 15 as shown in FIG. 9 can inhibit relaxation of this stress and mitigate distortion of the metal bond interface and promote 3D electrical interconnections across the metal bond interface. As in embodiment 1, this wafer 15 is preferably made from material with a CTE match to that of wafer 11.
[0041] Numerous modifications and variations of the present invention are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims, the invention may be practiced otherwise than as specifically described herein.
