The present disclosure relates generally to semiconductor device manufacturing techniques and, more particularly, to the formation of interconnect structures with selective local metal cap regions for improved electromigration behavior.
Integrated circuits are typically fabricated with multiple levels of patterned metallization lines, electrically separated from one another by interlayer dielectrics containing vias at selected locations to provide electrical connections between levels of the patterned metallization lines. As these integrated circuits are scaled to smaller dimensions in a continual effort to provide increased density and performance (e.g., by increasing device speed and providing greater circuit functionality within a given area chip), the interconnect linewidth dimension becomes increasingly narrow, which in turn renders them more susceptible to deleterious effects such as electromigration.
Electromigration is a term referring to the phenomenon of mass transport of metallic atoms (e.g., copper or aluminum) which make up the interconnect material, as a result of unidirectional or DC electrical current conduction therethrough. More specifically, the electron current collides with the diffusing metal atoms, thereby pushing them in the direction of current travel. Over an extended period of time, the accumulation of metal at the anode end of the interconnect material significantly increases the local mechanical stress in the system. This in turn may lead to delamination, cracking, and even metal extrusion from the metal wire, thereby causing an electrical short to adjacent interconnects. Electromigration becomes increasingly more significant in integrated circuit design, as relative current densities through metallization lines continue to increase as the linewidth dimensions shrink.
In an exemplary embodiment, a method of forming a wiring structure for an integrated circuit device includes forming one or more copper lines within an interlevel dielectric layer (ILD); masking selected regions of the one or more copper lines; selectively plating metal cap regions over exposed regions of the one or more copper lines; and forming a conformal insulator layer over the metal cap regions and uncapped regions of the one or more copper lines.
In another embodiment, a wiring structure for an integrated circuit device includes one or more copper lines formed within an interlevel dielectric layer (ILD); a plurality of metal cap regions formed over selected regions of the one or more copper lines; and a conformal insulator layer formed over the metal cap regions and uncapped regions of the one or more copper lines.
Referring to the exemplary drawings wherein like elements are numbered alike in the several Figures:
a) is a schematic diagram illustrating the direction of electromigration force and electron flow away from the cathode end of IC interconnect structure, leading to migration of atoms from the cathode end;
b) is another schematic diagram illustrating both electromigration force and a stress-induced back flow of atoms when diffusion barriers are located at both cathode and anode ends of an interconnect structure;
a) and 2(b) are Dual Damascene copper versions of the examples shown in
As indicated above, electromigration (EM) is a reliability failure mechanism for metal interconnects in which metal atoms migrate under the influence of the electric field and electron flow. For the case of aluminum (Al) and copper (Cu) interconnects, the electromigration mass flow is in the direction of electron flow. During electromigration, the electron wind applies a force that results in an atomic flux, J, given by the following equation:
where n is the density of atoms, ve is the drift velocity of migrating atoms, D is the effective diffusivity, k is Boltzmann's constant, T is the absolute temperature, j is the current density, ρ is the resistivity and eZ* is the effective ion charge.
However, in the presence of a diffusion barrier, atoms accumulate at the anode end and deplete the cathode end of the conductor, leading to a stress gradient and back diffusion of atoms (see, for example, I. A. Blech, J. Appl. Phys. 47, 1203 (1976)). The combination of electromigration and the stress-induced back flow of atoms gives rise to a net atomic flux, Jeff, given by the following equation at steady state:
where vb is the back flow velocity of atoms, Δσ is the difference in stress between the cathode and the anode ends, Ω is the atomic volume and L is the conductor length. As illustrated in
When the back stress gradient balances the electromigration force, mass transport is completely suppressed. This phenomenon is referred to as the electromigration threshold or the short-length effect, and occurs for sufficiently short interconnects and low current densities. The threshold condition is defined from the above relation for Jeff such that:
where (jL)th is referred to as the threshold length product. For jL values less than (jL)th, there is no electromigration failure in the interconnect structure. If j and L correspond exactly to the threshold condition, then the length of the interconnect corresponds to what is referred to as the critical length. The short-length effect has been observed in AlCu interconnects with W interlevel studs, as well as in Dual Damascene Cu interconnects with interlevel vias, wherein the Dual Damascene interconnects utilize liner materials such as tantalum (Ta), tantalum nitride (TaN), Ti, TiN, W, ruthenium (Ru), ruthenium nitride (RuN), and tungsten nitride (WN), for example as diffusion barriers.
a) and 2(b) are Dual Damascene Cu versions of the examples shown in
Regardless of the specific type of via and interconnect metal(s) used, one way to take advantage of the short-length effect is to simply design short interconnects since the allowed current density increases as the conductor length decreases. However, this approach has limitations since the design of “short” interconnects (by definition) requires more interlevel vias, which in turn may cause yield degradation as well as increases in resistance. Therefore, it would be beneficial to design a layout that can tolerate higher current densities without significantly reducing the conductor length.
It is also known that, for Dual Damascene Cu interconnects, the main diffusion path leading to electromigration failure is along the Cu/dielectric cap interface. Studies have shown that the electromigration lifetime is dramatically increased by depositing a metal cap layer on the Cu lines as the rate of Cu diffusion is much lower along the Cu/metal cap interface. Again, exemplary materials for the metal cap may include CoWP, Ta, Ru, Co or other materials. One of the concerns, however, with implementing a metal cap layer is the risk of depositing some metal cap material in dielectric regions between the Cu lines. Although the metal cap deposition is intended to be selective, some metal particles may actually be deposited between the Cu lines, leading to early time dependent dielectric breakdown (TDDB) failure.
One of the more popular electroless capping processes for BEOL semiconductor copper wiring (as described above) includes selective electroless deposition of a thin cap (e.g., 100 Å or less) of CoWP. There is, however, one disadvantage in using CoWP, since it has been found to increase line RC (resistance times capacitance), which can degrade circuit performance. There are two apparent reasons for RC increases. First, the CoWP process raises the resistance of the copper lines, as measured at final test, by apparently stabilizing the microstructure enough to prevent the significant resistance decrease that is normally observed by final test (and attributed to subsequent anneals as more wiring levels are added, which involve deposition processes at elevated temperatures).
Second, the CoWP process can also increase capacitance, by making the lines stand taller. The exact amount of capacitance increase will depend on exactly how the CoWP process is run. Typically, the process involves removal of a tiny amount of Cu (in addition to any native Cu oxide from the tops of the copper lines) by a wet clean before the actual cap deposition process begins, to ensure that an oxide free copper surface is exposed. If the thickness of the CoWP cap is greater than the thickness of Cu removed, then the metal line after capping will necessarily stand taller and therefore have greater capacitance. It has been observed for 32 nanometer (nm) technology that thin wires capped with CoWP are typically about 10% higher in RC at final test than thin wires not capped with CoWP.
Accordingly, embodiments disclosed herein provide methods and structures for depositing a metal cap layer at periodic intervals such that the distance between metal cap layers is equal to the critical length. Since the diffusivity is lower for the Cu/metal cap interface than for the Cu/dielectric cap interface, regions having a metal cap will act as a diffusion barrier. Such a design may be implemented to increase the allowed current density, and without the need for introducing additional interlevel vias. Notably, electromigration improvement may still occur if the distance between metal cap layers is greater than the critical length, as the presence of a metal cap layer will prevent electromigration failure from occurring in those regions. On the other hand, having the distance between metal cap layers equal to the critical length will also prevent electromigration failure from occurring in the regions without a metal cap layer.
Further embodiments disclosed herein involve lithographic masking prior to electroless metal capping, using a mask designed to allow capping of those areas where electromigration concerns override RC concerns, but not of those areas where the desire for low RC overrides electromigration concerns. Some specific examples of criteria for choosing these areas are described in further detail below. The mask design to be used may be varied from level to level. In general, embodiments of this type involve coating the wafer (after the complete Cu and liner CMP process is performed) with a bottom antireflective coating (BARC) that is on the order of a few hundred to one thousand Å in thickness, followed by a photoresist which is on the order of one thousand to a few thousand Å in thickness, exposing the resist through the mask, and developing the resist. The BARC is then opened in the areas where the resist has been removed by the developer.
Referring generally now to
As particularly shown in
Referring to
As further illustrated in
Another exemplary embodiment for forming intermittent metal cap regions, in addition to the dielectric cap-open approach described above includes the use of a sacrificial hardmask. In this regard,
Referring to
Among the advantages of the embodiments described above in forming metal cap layers at periodic intervals (e.g., cap open process, sacrificial hardmask process) include, in addition to electromigration improvement, a resulting structure that does not have residual metal cap material between the Cu lines at the Cu/dielectric cap interface. Also, TDDB concerns are alleviated even if some residual material is deposited between the nitrogen-doped silicon carbide layer and the conformal nitride layer, since this will not be the path for TDDB failure.
Referring now to
Still another possibility is to use a BARC 326 which is both photoactive and developable at the same time that the resist is developed, as shown in
Masks used in the exemplary processes depicted in conjunction with any of
Although the resolution of the exemplary wet processes described above would not be sufficient to resolve individual nested wires or individual nested metal shapes at minimum pitch for 32 nm nodes or below, a wet process with 110 nm resolution as referenced would allow groups of several wires (or larger “blocks” of circuitry) to be masked off to prevent CoWP deposition, or to receive CoWP while other areas do not receive it. One example would be to block CoWP deposition in SRAM cell areas, where most metal shapes are small, while allowing CoWP deposition, if desired, in SRAM support circuitry and other surrounding areas. The SRAM cell could therefore take advantage of the lowest possible metal RC to give the lowest possible switching delays.
As mentioned earlier, long lines are generally more prone to electromigration failure than short lines. If so desired, a mask could be designed such that long lines (longer than the resolution limit of the masking process) could receive CoWP along some but not all of their length, preferably in regions separated by diffusion length L, to achieve some degree of “compromise” between electromigration enhancement and RC minimization.
There are various methods for determining where in a design (by spatial position and by metal level) it would be most beneficial to deposit or not deposit CoWP, and of automating the design of the masks to be used prior to the CoWP capping process. Metal shapes sensitive or insensitive to RC changes can be understood in a number of ways. The schematic portion of a design gives a very clear understanding of the power (DC) distribution. Typically, these nets are marked with a net name that can be used in a post processing algorithm (dataprep) to distinguish ground nets and Vdd nets from all other nets. Once that understanding is clear, a mask layer can be created to open up the resist to deposit the CoWP in the desired places. A “resolution limit” (for example, 110 nm as discussed above) could be included as part of the mask generation process flow, such that areas (to be masked or not masked) that are smaller than this resolution limit could be automatically deleted from the mask design if so desired.
Additionally, modern VLSI designs have multiple voltage domains. Each of these domains must have independent net names for power rails, potentially enabling CoWP to be deposited or not deposited as a function of domain voltage if so desired. Another method of identifying nets is to examine the timing report. The output of the timing report typically contains wiring delay sensitivity information to help the designer understand how to improve/close timing from a physical layout perspective. Nets that are associated with a timing run that show a high sensitivity to wiring delay should not be capped with CoWP.
Still another method would allow the designer to place marker levels in their design. These marker levels would be coincident with metal layers where the designer wanted increased resistance to electromigration. Parasitic Extraction (PEX) could then be updated to provide the designers with accurate resistance/capacitance in the extracted netlist and used during formal design verification closure and timing closure. Design rules could be used, if needed, to govern how the marker shapes could be placed (for example, a rule for minimum space between adjacent marker shapes, which would depend on the resolution of the masking process).
Still another method would be to base the decision on the current densities in each metal shape during circuit operation. Current density can change along the length of a wire as the polygon changes, and as multiple nodes come and go. Current density analysis may be back annotated from the schematic analysis into physical layout (polygons) noting regions where high current densities exist. Furthermore, a single metal polygon 400 may have both DC and AC regions thereon (
While the disclosure has been described with reference to an exemplary embodiment or embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof without departing from the scope of the disclosure. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the disclosure without departing from the essential scope thereof. Therefore, it is intended that the disclosure not be limited to the particular embodiment disclosed as the best mode contemplated for carrying out this disclosure, but that the disclosure will include all embodiments falling within the scope of the appended claims.
Number | Name | Date | Kind |
---|---|---|---|
4962060 | Sliwa et al. | Oct 1990 | A |
5410490 | Yastrow | Apr 1995 | A |
5439731 | Li et al. | Aug 1995 | A |
5470788 | Biery et al. | Nov 1995 | A |
5737580 | Hathaway et al. | Apr 1998 | A |
5899707 | Sanchez et al. | May 1999 | A |
6245996 | Atakov et al. | Jun 2001 | B1 |
6251528 | Uzoh et al. | Jun 2001 | B1 |
6489684 | Chen et al. | Dec 2002 | B1 |
6597067 | Biery et al. | Jul 2003 | B1 |
6677647 | Dawson | Jan 2004 | B1 |
6678951 | Atakov et al. | Jan 2004 | B2 |
7016794 | Schultz | Mar 2006 | B2 |
7026714 | Cunningham | Apr 2006 | B2 |
7062850 | Atakov et al. | Jun 2006 | B2 |
7307344 | Minamihaba et al. | Dec 2007 | B2 |
7407605 | Restaino et al. | Aug 2008 | B2 |
7439173 | Greco et al. | Oct 2008 | B2 |
7451411 | Hau-Riege et al. | Nov 2008 | B2 |
7468320 | Hu et al. | Dec 2008 | B2 |
7471539 | Oates et al. | Dec 2008 | B2 |
7727892 | Morrow et al. | Jun 2010 | B2 |
8056039 | Chanda et al. | Nov 2011 | B2 |
8119531 | Arnold et al. | Feb 2012 | B1 |
8138604 | Yang et al. | Mar 2012 | B2 |
8211776 | Horak et al. | Jul 2012 | B2 |
20030116439 | Seo et al. | Jun 2003 | A1 |
20040198062 | Ye et al. | Oct 2004 | A1 |
20070063352 | Archer, III et al. | Mar 2007 | A1 |
20080026567 | Greco et al. | Jan 2008 | A1 |
20090053890 | Bonilla et al. | Feb 2009 | A1 |
20090294973 | Chanda et al. | Dec 2009 | A1 |
20100084767 | Yang et al. | Apr 2010 | A1 |
20100219512 | Tada et al. | Sep 2010 | A1 |
20110101523 | Hwang et al. | May 2011 | A1 |
20110175226 | Bonilla et al. | Jul 2011 | A1 |
20120104610 | Filippi et al. | May 2012 | A1 |
20120198403 | Arnold et al. | Aug 2012 | A1 |
Entry |
---|
I. A. Blech, “Electromigration in thin aluminum films on titanium nitride”, J. Appl. Phys., vol. 47, 1976, pp. 1203-1208. |
Douglas J. Guerrero, et al.,“Anti-reflective coating for multipatterning lithography,” Advances in Resist Materials and Processing Technology XXV, edited by Clifford L. Henderson, Proc. of SPIE vol. 6923, 69230X, 2008, pp. 1-7. |
L. T. Romankiw, “A path: from electroplating through lithographic masks in electronics to LIGA in MEMS,” Electrochimica Acta, vol. 42, Nos. 20-22, pp. 2985-3005, 1997. |
R.A. Wachnik, et al.,“Practical Benefits of the Electromigration Short-Length Effect, Including a New Design Rule Methodology and an Electromigration Resistant Power Grid with Enhanced Wireability,” Symposium on VLSI Technology Digest of Technical Papers, IEEE, pp. 1-2, 2000. |
V. Sukharev, et al., “A Model for Electromigration-Induced Degradation Mechanisms in Dual-Inlaid Copper Interconnects: Effect of Microstructure;” Journal of Applied Physics 102; 2007; pp. 053505-1-053505-14. |
Number | Date | Country | |
---|---|---|---|
20140203435 A1 | Jul 2014 | US |