Addressing the challenges of spin-on low-k dielectric dispense management

Josh H. Golden, J. Eric Carrubba, and Jay Jung, Microbar

Optimizing dispense tool designs and SOD management will require cooperation between equipment manufacturers, materials suppliers, and end-users.

During the last five to eight years, the semiconductor industry, academia, and government labs have been engaged in an intense worldwide research effort in preparation for the integration of low-k dielectric materials in ICs at the 180-nm technology node, which was scheduled for first shipments in 1999. This effort was driven by the desire to uphold Moore's Law and the various revisions of the industry roadmap.^1–4 However, the transition from silicon dioxide films applied by chemical vapor deposition (CVD), which have a dielectric constant (k) value of 4.1 to 4.2, to inorganic, organic, and hybrid materials with k values of less than 3.0 has proven to be more difficult than predicted. Figure 1 illustrates the delayed adoption of low-k dielectrics based on the roadmaps.^4,5

Figure 1: The delayed implementation of low-k materials at each device generation. (Adapted from reference 5.)

As both research and delays continue, a debate has been raging over the choice of techniques for applying low-k materials. Both spin-on and CVD candidates present integration challenges because their properties differ from those of the benchmark, SiO^₂, which is hard, relatively inert, thermally stable, and easy to deposit or grow.^6–8

Table I compares the dielectric properties of SiO^₂ to those of organic spin-on dielectrics (SODs) and Si^_wC^_xO^_yH^_z CVD films. The organic SODs may present a greater integration challenge than do the hybrid Si^_wC^_xO^_yH^_z CVD films, which are more "silica-like." For example, SiO^₂ has a modulus of 72 GPa, while a leading polyarylene SOD exhibits a modulus of 2.7 GPa.⁸ This approximately 27-fold difference in modulus has a significant effect on the efficacy of chemical mechanical planarization (CMP) processes, which subject copper interconnect layers to significant shear forces.⁹

Property Material Silicon Dioxide Organic SODs CVD SiwCxOyHz Dielectric Constant (k) 3.9—4.5 2.4—2.85 2.5—3.2 Modulus (GPa) 72 2—3 4.2—10.8 Hardness (GPa) 8.7 ~0.25 0.23—1.41 Composition SiO2 Various: Poly(arylene ether) Poly(arylene) Poly(benzoxazole) 15—22% Si. 17—23% C, 18—30% O, 26—45%H

Table I: Comparison of some properties of silicon dioxide with those of SODs and carbon-doped siloxane film. (Adapted from reference 8.)

In addition to the modulus, heat dissipation is another important issue with low-k candidates, because their thermal conductivity is typically 20 to 30% lower than that of SiO^₂.^6,7,10 Poor heat dissipation can cause localized metal-line heating, which eventually leads to device failure. Thermally induced stresses within the low-k film may lead to delamination because of mismatches between the film's coefficient of thermal expansion (CTE) and the substrate CTE. For example, silicon dioxide's CTE is 0.5 ppm/°C, aluminum's CTE is 23.1 ppm/°C, and copper's CTE is 16.5 ppm/°C. Because the CTE of many organic dielectrics is >50 ppm/°C, CTE mismatches can lead to high tensile stresses following thermal cycles.⁷

A successful low-k candidate must display several critical material properties: chemical resistance to oxidation and moisture absorption during plasma ashing, stripping and cleaning, and CMP processes; thermal stability (no weight loss or shrinkage following repeated isothermal soaks at 400°C); and the ability to adhere to substrates, including liners and barriers, in order to withstand the shearing and delamination forces exerted by the CMP process.

A variety of options have recently emerged to address some of the challenges associated with using low-k materials in ICs. These include delaying the implementation of such materials as well as new integration schemes. Approaches include:

• Combining new circuit architectures with aluminum and low-k dielectrics to achieve lower resistance-capacitance (RC) delays.^2,3,5,11
• A fluorinated silica glass technique.
• An embedded approach that uses low-modulus and low-thermal-conductivity low-k materials only between the lines, where 80% of the reduced RC-delay benefit is realized. At the via level, SiO^₂ or a spin-on glass (SOG) technique may be used for added strength and improved thermal conductivity.^5,12
• Using a polyarylene thermoset at the lower five tight-pitch copper levels and fluorinated silica glass at the top three global wiring levels for strength and heat dissipation.^8,13
• Using relatively inert (k ~ 4.5) silicon carbide as a CMP barrier film and hard mask to protect a potentially reactive low-k film.⁶

While these possibilities raise many interesting issues, this article focuses on some of the benefits and concerns associated with the adoption of SODs. The need for dispensing equipment especially designed for this application and the need for SOD material management capabilities are particularly emphasized.

SODs versus Hybrid CVD Films

The April 2000 announcement by IBM that it will adopt Dow Chemical's SiLK resin as an insulator for its dual-damascene copper processes at the 130-nm technology node represented the first major breakthrough in the spin-on versus CVD debate.¹³ Since then, UMC in Hsinchu, Taiwan, and Altis Semiconductor (an IBM-Infineon joint venture) in Corbeil Essones, France, have announced that they will also use the 130-nm process technology developed with IBM, and it is likely that more companies will announce their adoption of SiLK sometime this year. However, other 2.2–3.1-k SODs are still being considered, including Honeywell's HOSP, a hybrid siloxane-organic polymer; Dow-Corning's XLK, a porous hydrogen silsesquioxane (HSQ); Schumacher's MesoELK, a porous silica product; and JSR's MSQ hybrid.

SODs' main competitors at the 130-nm technology node are hybrid Si^_wC^_xO^_yH^_z CVD films with dielectric constants in the 2.5–3.0 range. These films are more like SiO^₂ than the SODs are, and they have a lower density and lower polarizability because of their pendant organic methyl groups. Other advantages of the films include the good availability and low cost of the precursor materials, as well as their ability to be processed with existing CVD tool sets.^5,14,15 However, the Si^_wC^_xO^_yH^_z films are relatively soft compared to SiO^₂ (0.23–1.41 versus 8.7 GPa).⁸ It also appears that Si^_wC^_xO^_yH^_z hybrids with a dielectric constant of <2.5 display film properties that will be inadequate for the 100-nm technology node, although research efforts are under way by materials and CVD tool suppliers to address such issues.

On the other hand, SODs have certain intrinsic material advantages over low-k CVD films. For example, a SOD material typically has a well-defined chemical composition that undergoes a specific and well-characterized physicochemical change (i.e., cross-linking) during thermal cure. In contrast, the composition of Si^_wC^_xO^_yH^_z films may vary from recipe to recipe and from tool to tool, as a CVD process is developed and qualified.

In addition, because SOD compositions are created using established chemical synthesis and modification procedures, the preparation of porous materials with dielectric constants as low as 2.0 is readily achievable. A variety of approaches are being used by synthetic chemists in the preparation of both open- and closed-cell porous networks with optimized pore sizes and distributions. These include the use of porogens, such as a high-boiling-point solvent (Dow-Corning's XLK), or a thermally decomposable organic molecule, such as a dendrimer.¹⁶

Mesomorphous silicas represent another class of SODs. To create these materials, an organized porous silica or Si^_wC^_xO^_yH^_z network is formed by templating silicon-containing precursors with organic surfactants in a mixed-solvent system. Sol-gel or sol-gel-like ordered structures are then created in a prebake step, followed by the burning off of the templating organic molecules. In contrast to these porous-SOD technologies, the preparation of porous materials for plasma deposition is a greater challenge because it is more difficult to control pore size, achieve pore-size homogeneity, and create closed-cell architectures. Both CVD equipment suppliers and materials manufacturers are researching ways to meet this challenge.

Another advantage of SODs over films applied by CVD is that the sensitivity of SODs to moisture, oxygen, and heat is generally understood and somewhat predictable. Thus, necessary precautions can be integrated into handling and dispensing procedures without extensive experimentation. The ease with which the handling and dispensing of these materials can be optimized, and the relative simplicity of the spin-on process, help to hold down the end-user's cost of ownership (COO). In contrast to SODs, the optimization of Si^_wC^_xO^_yH^_z processes requires rigorous experimental designs in order to achieve the best combination of mass flow and multiple-component precursor ratios, as well as the necessary process tool parameters. Significant effort must therefore be expended to maintain the process window, the cost of which typically has not been included in COO studies.¹⁵ Another hidden cost is that incurred in the transference of the process from one site or tool to another, which can be difficult and time-consuming.

Proponents on both sides of the dielectric debate continue to argue the respective strengths and weaknesses of the various technologies, from materials properties to ease of use to COO.¹⁵ However, it appears that a combination of SOGs, SODs, and CVD films will be used in conjunction with new integration schemes at both the 180- and 130-nm technology nodes, and possibly at the 100-nm node as well.

SOD Dispense Tool Design Issues

The implementation of SODs will require that new cooperative relationships be developed between tool manufacturers, materials suppliers, and end-users. Formerly, when spin-on tools were used only for photoresist, there was a mostly closed loop between spin-on track manufacturers and dispense cabinet providers. Now, however, because of a variety of factors, including the existence of multiple SOD candidates with different chemistries and their high costs (>$1000/L), it has become critical for equipment manufacturers, materials suppliers, and end-users to work closely to ensure that the chosen SOD will be optimally dispensed onto the wafer.

There are significant differences between photoresist dispense and SOD dispense that can affect tool design and process parameters. After dispense, a photoresist undergoes a variety of processes, including prebake, exposure, development, and cure, after which it is stripped off the wafer. But because an SOD is integrated permanently into the device structure, careful attention must be paid to preventing contamination and maintaining the correct material thickness and composition during the dispense process. Failure to control particles and other contaminants results in the classic "fly stuck in amber" scenario, which can ultimately lead to device failures.

In addition, while most photoresists can be handled similarly, each SOD has a different set of handling criteria because of its unique chemistry and physical properties. Dispense tool manufacturers must understand these properties and those of the carrier solvent(s), applying this knowledge to the design of their hardware and software. Knowledge of the SODs' properties must drive the choice of tool component materials and the integration of safeguards against contamination. Table II lists some SODs and relevant information on their properties.

Property SOD Material Dielectric A Dielectric B Dielectric C Dielectric constant (k) 2.65 2.0–2.5 2.5 Type (solute) Polyarylene Porous HSQ Organic siloxane polymer Solvent(s) -Butyrolactone, cyclohexanone, mesitylene 2-Pentanone, toluene, tetradecane Propyl acetate Solute moisture sensitivity (25°C storage, 1 week) No No Yes Hygroscopic solvent(s) Yes Yes Yes Thermal sensitivity (25°C storage, 1 month) No Yes Yes Oxygen sensitivity (25°C storage, 1 week) No No No Particle formation potential (25°C storage, 1 week) No No Yes Flammable mixture Yes Yes Yes

Table II: Comparative data on SOD properties that are useful in the design and operation of dispense tools. (Data collected from materials suppliers' material safety data sheets and and personal communications.)

Tool manufacturers must also consider the properties of SOGs and their derivatives, which may be used as etch stops and barriers in conjunction with a low-k SOD.^6,11 Because the wetting characteristics of SOGs typically are optimized for gap-fill and planarization applications, these materials are prone to wicking and may form small inclusions in nooks and crannies in the dispense apparatus' fluid path. This phenomenon can lead to the formation of cross-linked silica-like particles upon solvent removal or evaporation. One solution to this problem is to use low-roughness-average materials for the dispense tool's fluid paths and to provide a solvent-saturated inert-gas environment within the tool.

Many polar carrier solvents are hygroscopic or may undergo acid or base hydrolysis under certain conditions. While gross solvent hydrolysis is unlikely in a well-engineered dispense apparatus, even small amounts of moisture can have a deleterious effect on the composition of SODs that contain metastable siloxane oligomers or silane or siloxane adhesion promoters. Therefore, careful attention to SOD bottle changes and the use of an inert-gas backfill to prevent oxygen or moisture infiltration is required.

Some SOD materials are thermally sensitive and thus require a temperature-controlled environment to prevent spontaneous polymerization, cross-linking, or condensation of oligomeric moieties, which can result in particle formation. Possible measures to prevent such reactions and to optimize dispense temperature include transferring the cold-preserved SOD from the container to a secondary, temperature-controlled reservoir and the inclusion of thermal controls (±1°C) along the dispense tool's fluid path. Temperature control is important not only for preventing particle formation, but also for helping maintain the fluid viscosity necessary for correct on-wafer film thickness.

In addition to temperature controls, an effective SOD dispensing system should also provide solvent-saturated backfill and automated flush-purge capabilities for preventing particle formation and deposition within the dispense apparatus. While solvent-saturated backfill ensures a wetted path for the SOD with a like solvent or solvent mixture, an automatic solvent flushing protocol cleans the system and then dries it with an inert gas. An automated flush-purge system flushes and cleans dispense lines between batch changes and helps perform routine maintenance. The ability to isolate and purge portions of the fluid path for cleaning and servicing in a multitrack dispense tool while processing continues uninterrupted will become increasingly important as SOD end-users ramp up capacity and move from using 1-L containers to bulk delivery systems.

Another challenge facing SOD dispense systems is the removal of particles and microbubbles, which is accomplished through one-pass filtration or filtration with recirculation, depending on the material's requirements. A study of particle and microbubble removal using filtration with recirculation was conducted at the Honeywell Electronic Materials Star Center (Sunnyvale, CA) using a PMS M65 particle detector from Particle Measuring Systems (Boulder, CO) integrated with the Low-k 2.0 SOD/SOG dispense system from Microbar (Sunnyvale, CA).

In the study, the particle-size detection limit was 0.065 µm, and the dispense tool included a 0.065-µm filter. Data were collected at 25.5°C at a sampling flow rate of 0.6 ml/min. Figure 2 shows the results of this study for an Si^_wC^_xO^_yH^_z hybrid SOD that is sensitive to both temperature and moisture and thus has the potential to form insoluble particles. As the figure indicates, acceptable particle reduction was achieved after approximately 3 hours, or 175 recirculation cycles. While it proved difficult to quantitatively differentiate microbubbles from particles because of their similar size and an inability to isolate each component, the study found that both particles and microbubbles were removed to acceptable levels without affecting film thickness or quality.

Figure 2: Data collected for an SiwCxOyHz SOD hybrid using a particle detector that was integrated with an SOD dispense system equipped for filtration with recirculation.

Filter selection can be critical to achieving good results. While the choice of filter material and type is a materials-compatibility issue, the choice of pore size can directly affect SOD composition and molecular weight. Many SODs consist of oligomers or prepolymers and/or polymeric materials that have a specific molecular-weight distribution and average thermodynamic radius. A filter whose pore size is too small has the potential to fractionate this molecular-weight distribution and thus remove a certain population of oligomers or polymers from the SOD mixture. This can adversely affect the material's on-wafer thickness, physical properties, and, possibly, device yield performance. Because the molecular-weight distribution may also be affected by shearing forces (polymer chain cleavage), careful fluid-path design and selection of transfer pumps and associated apparatus are necessary for some SOD materials as well.

Dispense tool manufacturers also must address environmental safety and health concerns, including the prevention of spills and leaks of reactive, noxious, and flammable spin-on materials. Double-containment piping, spill sensors, quick-release fittings, and proper ventilation of and safe access to the dispense cabinet are among the features that can be included. Multiple-track SOD dispense schemes with built-in safety features can also increase tool efficiency and process uptime. Safety features that facilitate bottle change will also prove beneficial as SOD use increases.

SOD Management

The tracking and management of SODs is another important reason why materials suppliers, dispense tool manufacturers, and their customers must develop cooperative relationships. Issues that affect an SOD's life span include supplier batch and materials information, the transport environment from the supplier to the end-user, and the dispense protocol and associated parameters. A combination of hardware and software strategies can be used to compile, manage, and share the required data. For example, SOD containers can include a wireless tracking device that operates in the RF spectrum and is located in a well or integrated in a dip tube. Activated at the SOD point of fill, the device logs such transport parameters as time, thermal history, shock, and tilt. Because some SODs are sensitive to temperature and shock, monitoring these parameters can pinpoint potential upsets that can lead to the spoilage of the material.

Batch-tracking capability is important for maintaining quality control, process stability, inventories, and orders. In a comprehensive SOD-tracking scheme, data collected at the point of fill, during shipment, at fab delivery, and before bottle installation in the dispense tool can be uploaded into a software package that allows analysis and the sharing of data among authorized users. Analytical features can include troubleshooting and go/no-go protocols to be implemented when an SOD shipment enters the fab receiving area and again during bottle changeover at the dispense tool. Wireless tracking can also be integrated with bar code schemes to enable analysis. Figure 3 illustrates a possible combination of hardware and software technologies that can serve as the anchors of an integrated system for the management of SODs.

Figure 3: Schematic showing how a central software chemical-management package can link customers, materials suppliers, and dispense tool manufacturers with data from monitoring devices and process tools.

One example of such a system, developed by Microbar, includes the company's software platform and wireless tracking device. The software is capable of collecting data from the manufacturer's tool sets and those of other equipment suppliers, and can transmit both data and analytical results to materials suppliers and end-users via interfaces and the Internet. The types of information provided by the platform include histograms of equipment uptime and performance, maintenance and service logs, security information to be used by process engineers to validate chemicals, and mass balances of chemical-distribution loops for environmental and operational expense measurements. The implementation of such a system should minimize equipment COO by promoting the efficient utilization of expensive SOD materials.

Conclusion

Although the debate over the optimal technique for applying low-k dielectrics continues, some semiconductor manufacturers have announced that they are adopting SODs. Such materials have several advantages over CVD films, including their well-defined chemical composition and physical characteristics. However, their adoption will require new cooperative relationships between equipment manufacturers, materials suppliers, and end-users. Tool designers, in particular, must understand the properties of the many available SOD materials to ensure that they are dispensed optimally. Dispense tools must be designed to enable the prevention and removal of contamination, and must have built-in safety features. SOD management schemes that link suppliers and customers will also be needed to minimize cost of ownership and facilitate the efficient use of the costly materials.

References

1. PV Zant, Microchip Fabrication: A Practical Guide to Semiconductor Processing, 4th ed. (New York: McGraw-Hill, 2000), 9.
2. TN Theis, "The Future of Interconnection Technology," IBM Journal of Materials Research and Development 44, no. 3 (2000): 379–390.
3. XW Lin and D Pramanik, "Future Interconnect Technologies and Copper Metallization," Solid State Technology 43, no. 10 (2000): 63–79.
4. The International Technology Roadmap for Semiconductors, (San Jose: Semiconductor Industry Association, 1999), 172–173.
5. L Peters, "Low-k Dielectrics: Will Spin-On or CVD Prevail?" Semiconductor International 23, no. 6 (2000): 108–124.
6. L Peters, "Solving the Integration Challenges of Low-k Dielectrics," Semiconductor International 22, no. 13 (1999): 56–64.
7. M Morgan et al., "Low Dielectric Constant Materials for ULSI Interconnects," Annual Review of Materials Science 30 (2000): 645–680.
8. JG Ryan et al., "Copper and Low-k Dielectric Integration Challenges," in Proceedings of the Low-k Dielectric Materials Technology Conference, Semicon West (San Jose: SEMI, 2000), A1–A5.
9. M Fury, "CMP Processing with Low-k Dielectrics," Solid State Technology 42, no. 7 (1999): 87–96.
10. J Wetzel, "An Evaluation of Low-k Materials for Interconnect Roadmap Requirements," in Proceedings of the Low-k Dielectric Materials Technology Conference, Semicon West (San Jose: SEMI, 2000).
11. T Hasegawa et al., "Copper Dual Damascene Interconnects with Low-k Dielectrics Using FLARE and an MSQ Hard Mask," in Proceedings of the Low-k Dielectric Materials Technology Conference, Semicon West (San Jose: SEMI, 2000): G1–G6.
12. H Kudo et al., "Copper Dual Damascene Interconnects with Very Low-k Dielectrics Targeting for 130-nm Node," in Proceedings of the 2000 IITC (San Francisco: IEEE, 2000), 270–272.
13. R Goldblatt et al., "A High Performance 0.13-µm Copper BEOL Technology with Low-k Dielectric," in Proceedings of the 2000 IITC (San Francisco: IEEE, 2000), 261–263.
14. R Laxman et al., "Synthesizing Low-k CVD Materials for Fab Use," Semiconductor International 23, no. 13 (2000): 95–102.
15. M Corbett and J Davis, "Tipping the Scales for Spin-On vs. CVD: Which Process Will Emerge as the Winner in Dielectric Polymer Technology?" Semiconductor Magazine 1, no. 9 (2000): 40–46.
16. C Hawker et al., "Supramolecular Approaches to Nanoscale Dielectric Foams for Advanced Microelectronic Devices," Materials Research Society Bulletin (April 2000): 54–58.

Josh H. Golden, PhD, is director of process technology at Microbar in Sunnyvale, CA. In that position, he is involved in the development of dispense solutions for low-k spin-on dielectrics and copper chemistries, and in the development of new technologies for the treatment of CMP wastewater. He was formerly a polymer chemist at Cytec Industries and lead chemist in the low-k dielectric materials program at Watkins-Johnson (now SVG). Golden has authored 20 papers and holds several patents in the areas of materials science and chemistry. He received a PhD in chemistry from Cornell University in Ithaca, NY. (Golden can be reached at 408/542-9069 or jgolden@microbar.com.)

J. Eric Carrubba is a project engineer at Microbar with more than 13 years of semiconductor equipment experience. Before joining the company, he was an engineering manager at the semi-gas equipment division of Matheson Tri-Gas. He received a BS in aeronautics from the engineering department of San Jose State University, in San Jose, CA. (Carrubba can be reached at 408/542-9061 or ecarr@microbar.com.)

Jay Jung is director of controls engineering at Microbar. Since 1988, he has worked in the pharmaceutical, paper, and semiconductor industries. He has expertise in bringing processing equipment from R&D to manufacturing and in implementing factorywide automation. He received a BS in electronic engineering from California Polytechnic State University in San Luis Obispo. (Jung can be reached at 408/542-9072 or jjung@microbar.com.)

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit