The present invention relates to a method for fabricating integrated circuit devices.
Copper is preferred in forming metal interconnects in ultra-large scale integrated ULSI semiconductor devices. This is because, as semiconductor device geometries continue to scale down below 0.25 μm, and approach 0.13 μm feature sizes, the metal interconnect lines which carry current between devices on a chip begin to dominate the overall circuit speed. In order to enhance interconnect speed and reliability, the semiconductor industry is moving away from blanket deposition and etch of aluminum (Al) based metallization towards single-damascene and dual-damascene interconnect structures with copper (Cu) based metallizations.
Copper is a lower resistivity metal than aluminum, which results in lower RC interconnect delay. Copper has also been shown to have superior electromigration characteristics over aluminum, but is more difficult to process, primarily because it is more difficult to etch, and it acts as a deep level trap in silicon (Si) based devices. The preferred way to process copper interconnects is to etch a line trench, via hole, or a contact hole into a dielectric material, deposit the interconnect metallization to fill the trench or hole, and then polish the metal back to remove any metal from the surface of the substrate or wafer. The resulting metal-filled trenches and holes form the electrical interconnect. Forming an interconnect structure by filling a trench or hole with metal is known as a damascene process. If a trench an underlying via hole are filled simultaneously, it is known as a dual-damascene process.
The process of forming conventional single or dual-damascene structures is shown in
Although copper has desirable functional properties that make its use advantageous for forming interconnect and via structures, some reliability issues remain. For example, copper interconnects can suffer from the problem of hillock formation due to stresses resulting from the various high temperature process steps required to form semiconductor devices. Hillocks are spike-like projections that can erupt in response to compressive stresses generated in metal films due to differences in thermal expansion coefficients between adjacent materials in the semiconductor structure. These compressive stresses can cause hillocks to form and consequently protrude from the copper surface, as shown in
It is believed that two major factors affect the formation of hillocks: (1) non-uniform Cu oxide formation on the surface of the copper structure, and (2) high temperatures associated with subsequent chemical vapor deposition (CVD).
Non-uniform Cu native oxide is formed when the Cu film is exposed to the manufacturing atmosphere subsequent to polarization (CMP). Since the oxide formation process is uncontrolled, the oxide layer is typically uneven, and as such it can cause unbalanced expansion of Cu grains when the structure is subjected to high temperatures (e.g. during subsequent CVD). The resulting difference in expansion rates across the uneven copper oxide layer can induces hillocks.
A typical high-temperature process step that can trigger hillock formation is the deposition of an intermetal dielectric (IMD) layer 190 over the structure after the dual damascene interconnect such as 180 in
To ameliorate the formation of hillocks during the CVD process, current techniques utilize a high temperature annealing step just prior to CVD of the insulating layer. Thus, after the via and trench openings 150, 170 are filled with metal, and CMP has been performed to remove excess Cu material from the top of the semiconductor structure, an annealing step is carried out at a temperature of about 400° C. and is held for about 20 seconds. This annealing step is intended to relax the internal grain structure of the Cu, and to release internal stresses in the Cu generated during Cu plating and the CMP processes.
As noted, these protruding hillocks can adversely affect the ability of overlying dielectric layers to adequately passivate the top surface of the underlying copper structure. If hillocks are large enough, they can protrude through the dielectric layer and result in short-circuit types of defects either immediately or over time, which irrevocably damage the integrated circuit. Thus, it would be advantageous to provide a method for reducing or eliminating such hillocks.
These and other features and advantages of the present invention will be more fully disclosed in, or rendered obvious by, the following detailed description. A method is disclosed comprising the steps of: (a) depositing a copper layer on a substrate, (b) planarizing the copper layer; (c) annealing the copper layer; and (d) applying a dielectric layer over the copper layer; wherein the annealing step inhibits the formation of hillocks in the copper layer during the step of applying the dielectric layer.
A method is also disclosed comprising the steps of: (a) depositing a copper layer on a substrate; (b) planarizing the copper layer; (c) annealing the copper layer while exposing the copper layer to a reductive gas environment comprising hydrogen; and (d) applying a dielectric layer over the copper layer using CVD; wherein the annealing step substantially removes copper oxides from the surface of the copper layer, thereby preventing formation of surface discontinuities in the copper layer during the step of applying the dielectric layer.
These and other features and advantages of the present invention will be more fully disclosed in, or rendered obvious by, the following detailed description of the preferred embodiment of the invention, which is to be considered together with the accompanying drawings wherein like numbers refer to like parts, and further wherein:
a-c show the prior art process of forming a copper semiconductor structure in which copper hillocks are formed;
According to an embodiment of the present invention, disclosed herein is a low temperature process for inhibiting the formation of hillocks in copper semiconductor structures when the structures are exposed to the temperatures involved in semiconductor processing.
As shown in
An opening 4 is then formed in the upper insulating layer 3 using, for example, a dual damascene technique using reactive ion etching (RIE) with CHF3 as an etchant, although any suitable dry or wet etching technique known in the art may be used to form the desired opening. In the illustrated embodiment (
After the opening 4 is formed, a copper layer 5a is plated or otherwise deposited onto the upper insulating layer 3, filling the opening 4 and blanketing the upper surface of the insulation layer 3. The copper layer 5a is then planarized using any appropriate technique, such as CMP, so that the copper layer is removed from the top surface of the upper insulating layer 3 but remains within the opening 4, forming a copper interconnect structure 5b.
As illustrated in
The inventors have found that annealing the copper film at temperatures greater than 300° C. increases hillock formation, while annealing at temperatures less than 200° C. results in non-uniform reduction of copper oxide on the copper film surface, which likewise results in aggravated hillock formation.
The annealing temperature should be held substantially within the selected temperature range for a period of from about 150 seconds to about 250 seconds. In one preferred embodiment, the time period is about 180 seconds. This hold time is important to ensure sufficient reduction in the native oxide film from the top surface of the copper material, and to ensure adequate relaxation of the copper grain structure. Greater hold times would unnecessarily extend the wafer processing sequence, and could also aggravate hillock formation.
Furthermore, the reducing or forming gas atmosphere preferably is a hydrogen (H2) gas atmosphere diluted with nitrogen (N2) gas. This gas atmosphere can act to remove native oxides on the Cu film surface through a reduction reaction between the H2 and the Copper oxides. This reduction may be represented by the equation CuO+H2→Cu+H2O. In one example, the reducing or forming gas atmosphere comprises about 5 parts per hundred of H2 gas and about 95 parts per hundred of N2 gas (i.e. a 1/20 ratio of reducing to forming gas). Alternatively, a ratio of reducing to forming gas of from about 1/10 to about 1/25 could be used. Regardless of the ratio, it is important that a sufficient quantity of reductive gas be provided throughout the annealing step, in order to remove native oxides, since it is believed that non-uniform native copper oxide formation on the copper surface would hinder the “relaxation” of the copper grains and would thus aggravate hillock formation.
Subsequent to the annealing, the structure can be transferred to a cooling station (at room temperature) and subjected to a flow of N2 gas for about 120 seconds, or until the structure reaches room temperature.
Thus the disclosed procedure allows for shorter annealing times as compared to present methods, and also requires less input energy since the annealing temperature is lower than present methods. Even with the reduction in annealing time and temperature, hillock formation is significantly reduced as compared to methods of device formation in which no pre-CVD annealing procedure is employed. Thus, the disclosed method provides for more efficient device processing compared to prior methods, which can lead to a reduction in overall processing costs.
It is believed that the combination of the temperature and the reducing or forming gas atmosphere inhibits the formation of copper oxide (CuO) on the upper surface of the copper film 5b. Since copper oxide film formation is a contributor to subsequent hillock formation, reducing or eliminating the copper oxide film will reduce the formation of hillocks during subsequent processing. This low-temperature annealing eliminates the need for performing the current high-temperature annealing step that is performed prior to subsequent dielectric CVD.
Although copper oxides will reform on the copper surface once it is removed from the reducing atmosphere, it would be preferable to maintain the copper surface in the reducing atmosphere between the annealing and CVD steps. However, this is not required, and the inventors have found that simply minimizing the time between the annealing step and the subsequent CVD step is sufficient to substantially reduce hillock formation during the CVD step, even where a slight oxide layer is formed subsequent to annealing. This is believed to be due to the substantial relaxation of the Cu grains achieved by the low temperature annealing process.
Table 1 represents hillock counts measured on copper test samples, where the copper samples were subjected to a pre-CVD annealing process under a 95/5 N2/H2 blanket for a period of about 180 seconds. As can be seen from the representative data in Table 1, low temperature annealing of copper samples at between 200° C. and 260° C. resulted in significantly reduced hillock counts, as compared to annealing temperatures of greater than 300° C. and 180° C. less. For the purposes of this test, hillocks were measured using an AIT inspection system manufactured by KLA Tencor, which is capable of scanning and measuring defects on a wafer/film surface. “Saturated” means that the results of the AIT scan were so high as to be beyond the limit that the AIT tool
While the foregoing invention has been described with reference to the above embodiments, various modifications and changes can be made without departing from the spirit of the invention. Accordingly, all such modifications and changes are considered to be within the scope and range of equivalents of the appended claims.
Number | Name | Date | Kind |
---|---|---|---|
6248665 | Bao et al. | Jun 2001 | B1 |
6348410 | Ngo et al. | Feb 2002 | B1 |
6368948 | Ngo et al. | Apr 2002 | B1 |
6391754 | Paranjpe | May 2002 | B1 |
6391777 | Chen et al. | May 2002 | B1 |
6500754 | Erb et al. | Dec 2002 | B1 |
6515373 | Barth | Feb 2003 | B2 |
6518183 | Chang et al. | Feb 2003 | B1 |
6562712 | Liu et al. | May 2003 | B2 |
6670274 | Liu et al. | Dec 2003 | B1 |
6713407 | Cheng et al. | Mar 2004 | B1 |
6734101 | Bao et al. | May 2004 | B1 |
6806184 | Chen et al. | Oct 2004 | B2 |
6943451 | Whitehair et al. | Sep 2005 | B2 |
7074721 | Wang | Jul 2006 | B2 |
20040003873 | Chen et al. | Jan 2004 | A1 |
20040067643 | Liu et al. | Apr 2004 | A1 |
20040087147 | Lavric et al. | May 2004 | A1 |
20040092100 | Cheng et al. | May 2004 | A1 |
20040110373 | Liu et al. | Jun 2004 | A1 |
20040198055 | Wang | Oct 2004 | A1 |
Number | Date | Country | |
---|---|---|---|
20060252258 A1 | Nov 2006 | US |