These teachings relate generally to semiconductor technology and, more specifically, relate to the formation and application of diffusion barriers used in interconnect technology, and elsewhere, to separate materials which are in close proximity and which may undergo temperature cycling.
A diffusion barrier is a thin layer of material, which could be a metal, a metal nitride, or an oxide, that separates two different materials that would interact chemically either at room temperature or at elevated temperatures. In the field of interconnect technology, one common application of a diffusion barrier is for the encapsulation of copper metal lines and vias which form elements of wiring structures. Because of the tendency of copper to react rapidly and detrimentally with silicon and insulators, use of a diffusion barrier is necessary. Without a diffusion barrier, copper would diffuse out from the metal lines, degrading the properties of the inter-layer dielectric (ILD), the silicon junction, and many of the other materials used in the fabrication of the circuit. The diffusion barrier additionally prevents contaminants from diffusing into the copper metal features. These contaminants include oxygen, water, other metals, solvents, etc., all of which have the potential to cause significant changes to the copper metal and decrease its electrical conductivity.
Conventional interconnect technology solutions have included a thin layer of tantalum nitride (TaN), which is used, often in conjunction with a thin layer of tantalum (Ta), as the diffusion barrier to limit diffusion both into and out from copper circuit features. The typical thicknesses of these layers are in the range of 70-150 angstroms (7-15 nm). The failure temperature of this material can, in some cases, be measured and it has been found that the minimum failure temperature for this barrier generally exceeds 700-730 degrees centigrade.
Diffusion barriers fail due to transport of atoms across the barrier. This transport occurs most easily along grain boundaries, which are present in all polycrystalline material. The activation energy for movement of impurity atoms along the grain boundary is typically comparable to surface diffusion, rather than bulk diffusion, and this is typically one half or less of the bulk diffusion activation energy. Since these are thermally activated processes, grain boundary diffusion occurs many times faster than bulk diffusion at any given temperature. Hence, it is generally considered that the primary failure mechanism for a diffusion barrier is by grain boundary diffusion.
Conventional diffusion barriers are generally divided into sacrificial barriers and stuffed barriers. Sacrificial barriers are thermodynamically unstable. Sacrificial barriers react with interconnection material or substrate material to prevent the diffusion of the material. The sacrificial barrier is consumed according to the reaction so that it loses its function as a diffusion barrier when it has been completely consumed. This finite lifetime is the major limitation of sacrificial barriers. Stuffed barriers prevent diffusion by filling other materials into grain boundaries that are paths for diffusion.
Attempts have been made to develop amorphous diffusion barriers. The addition of silicon (Si) to tantalum nitride (TaN) results in an amorphous material, which functions well as a diffusion barrier due to its thermal stability and absence of grain boundaries. Other approaches have used a multilayer stack of materials. These can be dissimilar materials, or they can be the same material which is deposited in thin layers, perhaps followed by air exposure to oxidize the surface. These stacks of films are composed of crystallized films, each having many grain boundaries. However, due to the discontinuous deposition process, the grain boundaries from each layer do not align with that of the next, and this slows diffusion considerably (reference in this regard may be had to Clarke, “Method of Depositing Materials on a Wafer to Eliminate the Effect of Cracks in the Deposition”, U.S. Pat. No. 6,086,947).
Neither of these solutions is totally acceptable from a manufacturing process point of view. A ternary compound, tantalum silicon nitride (TaSiN), is difficult to fabricate reliably with satisfactory chemical control. And the multilayer, discontinuous grain boundaries approach does not scale well to the very thin thicknesses required for barriers in future interconnect generations, where the total diffusion barrier thickness is preferably no more than a few nanometers.
The foregoing and other problems are overcome, and other advantages are realized, in accordance with the currently preferred embodiments of these teachings.
Accordingly, a method is provided for the formation of very thin, multilayer diffusion barriers composed of even thinner sub-layers, where the sub-layers are only a few atoms thick. An aspect of this invention when forming these layers is to use the interfaces between each of the sub-layers to inhibit the formation of a crystalline lattice in each sub-layer. A strong bond between each of the sub-layers perturbs the regular crystalline structure of the sub-layer, as long as the sub-layer remains very thin. Since the surface energies dominate the bulk binding energies, the sub-layer remains disordered and in a substantially amorphous state (i.e., essentially free of a regular crystalline structure). The lack of formation of a lattice within each sub-layer results in no grain boundary formation, and hence, no pathways for inter diffusion of metals through the barrier as long as the multilayer remains amorphous.
The foregoing and other aspects of these teachings are made more evident in the following Detailed Description of the Preferred Embodiments, when read in conjunction with the attached Drawing Figures, wherein:
Referring to
An enlarged view of this last interface is shown in
Conventional diffusion barrier materials include crystalline materials such as titanium nitride (TiN), and tantalum nitride (TaN). Copper is able to diffuse through the grain boundaries 215 in these materials 230, 235, as the crystalline materials form the grain boundaries 215 where two adjacent crystalline structures meet. Copper atoms drift through the grain boundaries 215 and diffuse into the underlying structures 230, 235.
A diffusion barrier 315 in accordance with this invention is shown in
Each layer preferably has a thickness in a range of about two atoms to about fifteen atoms (0.4 to 4.5 nm), and more preferably has a thickness in a range of about two atoms to about ten atoms (0.4 to 3.0 nm), and even more preferably has a thickness in a range of about two atoms to about five atoms (0.4 to 1.5 nm).
To overcome the problems associated with grain boundaries mentioned above, the diffusion barrier layers 350, 360 in the present invention are substantially amorphous. Since the diffusion barrier layers 350, 360 are not crystalline, there are no grain boundaries to extend through the layer 315 for the copper (or other chemical species) to drift or diffuse through.
For purposes of illustration, substrate 300 is shown as a simple planar structure. The method of the present invention may be applied to more complicated structures and is not limited to the illustrative example. For example, multilayer interconnects and vias could be formed using the present method as well as simple contact openings that provide electrical contact to a device.
Referring again to
The diffusion barrier sub-layer material 350, 360 is preferably deposited by physical vapor deposition (PVD) (sputtering and/or evaporation). Atomic layer deposition (ALD) can also be used to deposit the sub-layers 350, 360. Atomic layer deposition (also known as atomic layer epitaxy) is a process for depositing thin layers of solid materials from two or more vapor precursors. Advantages of atomic layer deposition include low impurities content, low processing temperatures, ultra thin film deposition and excellent thickness uniformity over large substrate areas. Chemical vapor deposition (CVD) could also be used to deposit the sub-layers 350, 360.
As illustrated by
In
One aspect to forming this class of materials is to choose the correct components for the individual sub-layers. There are at least two major considerations. First, the two materials chosen are preferably immiscible, or have at most only a very minor level (<1%) of solubility. Materials which fit this criterion can be identified by their binary phase diagram, which should show no stable, solid states for compounds with more than a 1% alloy composition. If there is any significant solubility of the two materials chosen, then upon heating they will form a composite alloy and lose the desired multilayer structure.
Second, the two materials selected preferably exhibit good mutual adhesion. This adhesion is necessary to both hold the stack of materials 475(a) . . . (n), generally 475, together as well as to provide the strongly bonded interface 470 which can dominate the bulk material of the film to inhibit crystal lattice formation.
There are a number of examples of materials that can be used to make multilayer atomic laminates that will result in alternating layers of material that do not form lattices, and hence do not form grain boundaries. The choice of the materials used is based on (1) the chemical and metallurgical nature of each layer with the other layers, and (2) the potential for a chemical interaction to occur between the layers chosen and the layers above and below the diffusion barrier film. It is preferred that these considerations also include any changes which might occur at elevated temperature.
In terms of selecting two materials to form the atomic laminates, it is important that these two materials form a strong adhesive bond between the layers. The strength of this bond is important such that the interface between the two materials can effectively dominate the few atomic layers between the bond, inhibiting lattice formation. In addition, it is important that the materials chosen for these alternating layers be stable with increasing temperature and that the two materials have very low solubility with each other at elevated temperature. An example of two materials which fit these criteria are copper (Cu) and tantalum (Ta). The adhesive bond is strong in this case, and there is essentially no mutual solubility of Cu in Ta or Ta in Cu as there are no intermediate compounds in the Cu-Ta binary system. An alternative embodiment may use Ta and TaN, or W and TaN, as these materials also fulfill the criteria of good adhesion and low solubility. However, there are other potential pairs of materials that fit these criteria and that can be used in further embodiments of the invention. For example, the combination of titanium (Ti) and most rare-earth materials such as gadolinium (Gd) would be appropriate, as are materials sets such as, but not limited to, scandium (Sc)-tungsten (W), chromium (Cr)-yttrium (Y), and copper (Cu)-chromium (Cr).
Depending on the application for the diffusion barrier 315, certain materials may be inappropriate because they may either have poor adhesion to the layers above or below the diffusion barrier atomic laminate structure or one of the materials may interact adversely with the adjacent layers. An example of this is the above-mentioned pair of materials Ta and Cu, if used for a diffusion barrier in an interconnect structure composed of Cu circuit elements on a dielectric. While Ta and Cu have good properties from an atomic laminate point-of-view, using Cu in the atomic laminate diffusion barrier may allow Cu interactions with the interlayer dielectric, which the diffusion barrier 315 is intended to prevent. In this case, the Cu in the diffusion barrier laminate may be replaced with scandium (Sc), yttrium (Y), lanthanum (La), tantalum nitride (TaN), or tungsten nitride (WN), or other materials that satisfy the above criteria. It should also be noted that there is no limitation on the use of metals or alloys in the diffusion barrier laminate structure. Depending on the application, layers of nitrides (TaN, TiN, WN, etc) may also be used, either in combination with metal layers or other nitrides. In addition, oxide layers may also be appropriate in some atomic laminate structures, so long as they fit the criteria described above.
It should thus be appreciated that the foregoing examples of materials and thicknesses are not limiting, for example the total thickness of the barrier layer 315 will be a function of the selected constituent elements and/or molecules (e.g. nitride and/or oxides).
However, all such and similar modifications of the teachings of this invention will still fall within the scope of this invention, for example, the invention can be practiced using a variety of materials, and deposition techniques. As is clear from the illustrative examples, alternative embodiments are possible within the scope of the present invention. Further, while the method described herein is provided with a certain degree of specificity, the present invention could be implemented with either greater or lesser specificity, depending on the needs of the user. Other variations of the method within the scope of the present invention will occur to those of ordinary skill in the art. As such, the foregoing description should be considered as merely illustrative of the principles of the present invention, and not in limitation thereof, as this invention is defined by the claims which follow.