November 3, 2011 -- Authors from Avantor Performance Materials and SSMC Inc. compare the use of a semi-aqueous post-etch ash residue remover with an industry-standard hydroxyl-amine (HA)-based residue removal chemistry. The semi-aqueous-based product was shown to reduce cost of ownership for manufacturing of low- and high-voltage logic devices. The use of this newer generation engineered product was also shown to increase yield when used to replace the existing residue removal chemistry in fabrication of 45nm logic devices.

SSMC decided to test BAKER ALEG-380 from Avantor Performance Materials, Inc. (formerly Mallinckrodt Baker, Inc.) to assess whether changing the process of record (POR) chemistry for post-etch metal and oxide residue removal could be accomplished without impacting yield, device quality, and with the desired cost savings.

The product being tested (BAKER ALEG-380) is an engineered blend of organic solvents and semi-aqueous components suitable for bulk photoresist removal and post-etch ash residue and sidewall polymer removal. Designed to provide broad latitude in terms of processing times and temperatures, it is 100% water soluble and contains no HA or fluoride elements.

Initial tests demonstrated improved yield

One of the significant process challenges associated with using HA-based post-etch residue removal products is the chemistry’s relatively limited process temperature range. The maximum recommended process temperature for the HA-based product in use by the manufacturer was 75°C for 20 minutes. Process engineers, however, require flexibility to use higher bath temperatures to enable better removal of excess photoresist and substrate residue that can result from certain metal and oxide etch processes.

A wider bath process temperature range gives engineers a better process window in case of process drift prior to the cleaning step. In certain cases, etch processes prior to cleaning can generate excess polymer residues, requiring higher bath temperatures to ensure complete residue removal. Currently, the operating temperature recommended for the new product being tested is 70°C; however, it can be increased to 85°C if needed without damaging the integrity of critical substrates.

Figure 1. SEM images of an HA-based chemistry and one that uses ALEG-380

As shown in the scanning electron microscope (SEM) images (Fig. 1), HA-based chemistries when used at higher temperatures will cause excessive etch pitting of metal layers, such as the etch results of the HA product used for 20 minutes at 85°C. In contrast, the new semi-aqueous product that was being tested is HA-free. It can be used at the higher temperatures to complete post-etch ash residue removal without etching barrier layers or sidewall structures -- including the ability to clean difficult-to-remove residues at the higher temperature.

In early 2008, the semiconductor manufacturer and chemistry supplier teamed to test the semi-aqueous chemistry on one of the eight lines at the company’s foundry in Singapore. Test wafers were sent through the standard manufacturing steps.

Eight wafers were tested using the new product; five test wafers were cleaned with the HA-based product that was used by the foundry at that time. All wafers were subjected to residue removal baths at 70°C for 20 minutes. Wafer Area Tests (also referred to as full wafer testing or full loop testing) were conducted to assess electrical performance on metal lines and vias.

Table 1. Comparison of electrical performance on metal lines and vias.

Significant improvements in yield were discovered using the semi-aqueous chemistry on the test wafers. As shown in the table (Table 1), wafers cleaned using the HA-based chemistry showed an 88.24% yield, which was in line with yields the foundry experienced on its production lines . For the wafers treated with the new semi-aqueous product, metal line yields increased to 90.71%, while via yields were raised to 88.63%.

Following the foundry’s standard testing protocol, two of the eight wafers treated with the new residue removal chemistry were removed at minus 10% run time (18 minutes), four followed the standard process (20 minutes) and two were removed at plus 10% run time (22 minutes). The yield results for these ranged from 90.71% to 93.81%.

The foundry reported two results from these tests: All test runs of the new semi-aqueous product generated improved yield compared to the existing HA-based product, and that yield improvement was significant: two to three percent consistently under test conditions. This equates to several million dollars in sellable die from a fifty-two thousand wafer start-per-month factory. The team decided to further investigate the process basis for this yield improvement.

It was theorized that the improved yield from the semi-aqueous residue removal chemistry was related to lower etch rates compared to HA-based removal chemistries. Etch rates for a variety of substrate materials and chemistry temperature ranges were investigated and compared.

Table 2. Etch rates for a variety of substrate materials and chemistry temperature ranges.

Etch rates for AlCu, titanium, tungsten and silicon oxide were measured in angstroms per minute (Å/m) on blanket films and coupon tests (Table 2). The most significant differences were in the titanium etch rates for the HA-based product compared with the new product being tested. At 65 °C, the HA-based product showed an etch rate of 34.2Å/m; at 75°C, the rate increased to 92.4Å/m. In a standard 20 minute residue removal bath, this theoretically could cause a loss of close to 1000Å of barrier metal thickness.

By comparison, the semi-aqueous product shows a titanium etch rate of less than 0.1Å/m at 85°C. Also, the new product displayed minimal etch rates at 85°C, thus allowing it to be used at significantly higher temperatures to remove tough residues.

Titanium Nitride (TiN) is used as a barrier metal in copper-based chips to prevent diffusion of the copper into surrounding materials while maintaining an electrical connection. In 45nm nodes and below, maintaining the integrity of the TiN layer is crucial to yield and device performance.

It should also be noted that exposing tungsten to hydroxyl-amine molecules can cause autocatalytic reactions which results in complete voids in via holes. In addition, hydroxyl-amine compatibility on titanium is not well-designed at high operating temperatures. At greater than 75°C, the titanium etch rate increases significantly for HA-based chemistries. It is not advisable to implement the HA-based chemistries at greater than 75°C due to incompatibility with titanium material.

Since it was established that the etch rates for the new residue removal chemistry being tested were much lower on both metal lines (Al and Cu) and on substrate materials, SEM studies were conducted on wafers to characterize various structures and clarify their condition.

Figure 2. SEM images of cross-sections of both wafer center and wafer edge metal via structures show no defects in the via tungsten plugs after treatment with the tested product.

Further SEM studies were conducted on wafers treated with the new chemistry at 70°C for 60 minutes -- three times longer than standard processing intervals. Cross-sections of both wafer center and wafer edge metal via structures show no defects in the via tungsten plugs -- no voids, black lines or evidence of corrosion or degradation that could cause electrical performance issues such as changes in resistivity or noise (Fig. 2). This demonstrated that the semi-aqueous removal chemistry has sufficient margin in processing time to prevent over-etch of the substrates.

Figure 3. SEM images of wafers treated with the semi-aqueous chemistry at 70°C for twenty minutes show no residual polymer or sidewall etching on either the top TiN or Al substrate.

SEM imaging was also used to characterize the wider performance window of the product being tested. These images show metal lines on wafers treated with the product at 70°C for 20 minutes (Fig. 3). This analysis showed no residual polymer particles, no sidewall etch or etching on either the top TiN or Al substrate.

Nearly identical results were characterized in similar metal lines on wafers treated with the new chemistry at 85°C for 20 minutes. The semi-aqueous chemistry offers process engineers the option of increasing residue removal chemistry temperatures to achieve desired cleaning performance, rather than repeating the wafer bath step, which would be required using the conventional 70°C HA-based processes.

We concluded that the two to three percent improved yields exhibited on the wafers treated with the tested chemistry could be related to the fact that the semi-aqueous product does not contain hydroxyl-amine, has a very low etch rate on metals such as TiN, tungsten and oxide substrates, thus reducing the incidence of metal line and via defects.

Production-level data confirms test yield results

The yield improvements demonstrated by the full wafer tests provided sufficient data to justify testing the new chemistry under full production conditions. The goal of the test production run was to confirm compatibility of the test product with foundry’s tool set and processes, confirm that it offered comparable post-etch ash residue removal performance with the existing HA-based product, and to assess whether the new chemistry’s yield advantage observed in the test wafers could be reproduced under production conditions.

Basic module characterization was performed over a 25-day run, and two critical dimensions were assessed: Offline particle performance and etch rate after processing.

The semi-aqueous chemistry offered a lower total defect count and variation performance profile, with offline particles of 0.2µm or larger ranging from 1.0 to 4.0 particles detected at different measurement times, compared to the HA-based POR chemistry.

In the second comparison, aluminum etch rates for both chemistries were assessed during the same period as the offline particle comparison. At the start of each eight-hour shift, test aluminum wafers were passed through the post-etch ash residue removal baths for 20 minutes at 70°C. After removal and rinse of the test wafers, the Al etch rate was measured in Å/m.

The etch rate performance of the semi-aqueous chemistry was comparable with the HA-based chemistry, exhibiting acceptably low etch levels. This demonstrated that the tested product is fully compatible with AlCu substrates, making it a technically feasible alternative to the more commonly used residue removal product.

The foundry also assessed cleaning performance for both products. After post-metal etch, and prior to ash residue removal on test wafers, SEM studies were conducted of both metal lines and vias.

Both chemistries show comparable performance: virtually complete removal of top polymer residue, no sidewall polymer residue or sidewall etching.

Detailed SEM studies demonstrate that the HA-based chemistry product generated pitting in via substrate layers. These side-by-side comparisons (See Fig. 1) show the impact of using the HA-based product for 20 minutes at 85°C on TiN/SiO metal stacks. The outcome shows that HA-based cleaning results in less-complete polymer removal and pitting of the metal substrate. The production wafers run using the tested product on the same structure under the same time/temperature conditions exhibited no pitting and complete residue removal.

Cleaning performance of vias was also compared. SEM images were taken of cross-sections of test wafer vias filled with tungsten after post-etch ash residue removal by the HA-based chemistry product and the semi-aqueous residue remover, demonstrating comparable residue removal by both products.

The foundry also compared electrical performance results to assess yield impact of using the semi aqueous chemistry to replace the HA-based product. The devices being fabricated on the pilot line consisted of five metal layers, necessitating multi-pass residue removal after each metal etch step. WAT results for metal line and via electrical performance characteristics showed comparable performance to the HA-based chemistry for via resistance in VIA 1 through VIA 5 and M1 through M5 metal layers.

Based on the results of the pilot production run with the tested product, it was concluded that the semi-aqueous chemistry removed residues as well as the HA-based product, yet improved line yields. The tested product demonstrated comparable particle removal, substrate etch and electrical performance while improving process yields.

Conclusion

Based on the results of the pilot production run, and the reduced cost compared to the HA-based product, SSMC determined that it could achieve a two to three point improvement in yield, and a corresponding 25-30% cost reduction in post-etch ash residue removal process using the tested product.

As a result, SSMC changed its process of record and replaced the HA-based product with BAKER ALEG-380 from Avantor on all of its production lines.

Acknowledgments
Avantor and BAKER ALEG are trademarks of Avantor Performance Materials.

Nik Mustapha is a senior applications engineer at Avantor Performance Materials, 222 Red School Lane, Phillipsburg, NJ 08865 USA; ph.: 1-855-AVANTOR; nik.mustapha@avantormaterials.com; www.avantormaterials.com

Subscribe to Solid State Technology