Synthesizing Low-K CVD Materials for Fab Use

Ravi K. Laxman, Neil H. Hendricks, Barry Arkles, Terry A. Tabler -- Semiconductor International, 11/1/2000

Several major OEMs and their customers are developing processes to fabricate low-k CVD films with dielectric constants as low as 2.5. The seemingly simple addition of methyl groups to silicon dioxide to form a film with lower dielectric constant, k, is actually not so simple when you consider the synthesis of the gas or liquid precursor. For instance, trimethylsilane (3MS), probably the most widely integrated low-k CVD candidate at the 0.13 µm node, requires a multi-step synthetic process. Its preparation is more complex than that of tetraethylorthosilicate (TEOS), the "baseline" CVD dielectric precursor chemical,^1,2 but it is more straightforward than the synthesis of low-k spin-on dielectrics, polymers that incorporate a much higher degree of molecular engineering. The following describes the various synthesis processes for several CVD low-k precursors, and shows how a novel delivery system can solve troubling chemical handling and transportation issues associated with 3MS delivery.

The OSG structure

When intermetal spacings in advanced IC interconnects become smaller than ~0.18 µm, low-k dielectric insulators are needed to minimize RC delays and power consumption. Many researchers have described motivations for integrating low-k intermetal dielectric (IMD) as a replacement for CVD-deposited SiO₂.^3-8 Among many low-k candidates, however, only a few materials have shown all the required properties needed for integration into high-volume manufacturing processes.⁹ Unfortunately, the most promising precursors, alkyl and alkoxy silanes, do not currently have large-scale electronic or other industrial uses. For this and other reasons, the chemical synthesis of the precursors used to deposit low-k films constitutes an important set of technologies that directly impact the development and manufacture of advanced ICs with copper damascene interconnects.

Several companies are developing low-k CVD films using a variety of carbon-containing precursors. The resulting organosilicate glass (OSG) films, also called carbon-doped oxides, have a composition of Si_wC_xO_yH_z. The implementation of CVD low-k technology can benefit from a study of the differences between these organosilicon CVD precursors (with their silicon-carbon bonds) and TEOS.

The organic groups in OSGs invariably take the form of tetravalent silicon with a wide range of alkyl and alkoxy substitutions. In these films, the silicon-oxygen network seen in glass is occasionally interrupted, in a more or less homogeneous fashion, by the presence of organic, typically methyl (CH₃) groups. Hydrido (H) substitution at silicon can also be present. The film's lower dielectric constant, k, is due to these changes to the SiO₂ network and the reduced density of the OSG film relative to SiO₂. In typical CVD low-k films, 10% to 25% of the silicon atoms are substituted with organic groups. A graphical representation (Figure) shows the differences in chemical bonding between amorphous SiO₂ and amorphous OSG (Si:C = 4:1). Incorporation of methyl or other groups has a significant effect on the key physical, thermal and chemical properties of the resulting film.⁹

Graphical representations of the difference in chemical bonding structure between SiO 2 (above) and organosilicate glass with a Si:C of 4:1 (below).

TEOS synthesis

TEOS is an important industrial chemical used as a binder for cement casting, in high-temperature zinc-rich paints, in fabrics, and in a wide range of industrial films and coatings such as plastic optics. In fact, electronics applications constitute less than 5% of the world's TEOS production. The industrial route to TEOS takes place in three steps by reacting: silica with graphite to make silicon metal; silicon metal with chlorine to make silicon tetrachloride; and SiCl₄ with ethanol to make TEOS¹⁰:

In device manufacturing, TEOS provides lower deposition and oxide formation temperatures as well as more conformal coatings than silane (SiH₄).^11,12 As a result, TEOS has replaced silane in many oxide CVD processes. In principal, the above reactions could be used to manufacture semiconductor-grade TEOS. However, the necessary removal of traces of chloride is problematic and expensive. Instead, chlorine-free TEOS¹³ is manufactured according to:

Alkyl silanes & siloxanes

OSG films require silicon-carbon bonds, of which no known examples exist in nature. The silicon-oxygen structure in TEOS, by contrast, is common in nature — more than 27% of Earth's crust is silicon, all reacted with oxygen.

OSG films can be prepared from both substituted silanes and siloxanes. As shown in the forthcoming reactions, the industry uses fundamentally different synthetic processes for low-k OSG precursors than those used for TEOS.

The substituted silanes are most commonly alkylated silicon derivatives (molecules) wherein one to four of the hydrogen atoms have been replaced with an organic group — typically, but not necessarily, a methyl (CH₃) group. As shown through the series of chemical reactions that follow, the ubiquitous presence of the methyl group in organosilicon chemistry can ultimately be traced to the availability of methanol from very large-scale petroleum processes.

Siloxanes are silicon derivatives that have Si-O-Si linkages. As is the case with alkylated silanes, the methyl group is the most common organic substituent in siloxanes. Hydrogen substitution on silicon (Si-H) is often present and plays an important role in both types of OSG precursors.

The chemistry of organically substituted silicon has been comprehensively reviewed.^1,2 The Table lists several key properties of representative OSG precursors.

The industry knows and has investigated all the alkylated (methylated) derivatives of silane (SiH₄) for use in CVD low-k film deposition. These include methylsilane (CH₃)SiH₃, dimethylsilane
(CH₃)₂SiH₂, trimethylsilane (CH₃)₃SiH and tetramethylsilane (CH₃)₄Si. While all of these methylated silanes are in principle useful in the deposition of CVD low-k films, trimethylsilane (3MS) and tetramethylsilane (4MS) are the derivatives that have received the most attention.

Trimethylsilane, 3MS

Trimethylsilane's most significant industrial use is in the semiconductor industry. Companies are using 3MS to deposit both low-k OSG films (k<3.0) and low-k barrier films (k~5) as a replacement for silicon nitride (k~7). Fabricators use a multi-step synthetic route based on dimethylchlorosilane (CH₃)₂Si(H)(Cl) to make 3MS. This route is problematic and rather expensive because dimethylchlorosilane is itself acquired as a very minor component resulting from the reaction of methylchloride with silicon:

where (CH₃)SiCl₃, (CH₃)₂SiCl₂ and (CH₃)₃SiCl are the primary products. Thus, the desired dimethylchlorosilane must be distilled from the reaction mixture using large, specialized distillation towers.

Then the trimethylsilane is prepared via the methylation of dimethylchlorosilane using the Grignard methylating reagent derived from methylchloride. The synthesis of methylchloride (which is also a reactant in reaction 5) starts with methanol (CH₃OH). Methanol forms the basis for most organically substituted silicon chemicals, and is synthesized from petroleum byproducts in large process chemical volumes. Methylchloride is converted to the required Grignard reagent through:

Finally, combination of the Grignard reagent with dimethylchlorosilane yields trimethylsilane:

Reaction 7 is carried out under quite high-dilution conditions, typically in tetrahydrofuran (THF), to optimize yield. This in turn adds cost and complexity. Finally, it is important to remove all traces of MgCl₂ and other sources of chloride for use in low-k semiconductor processes.

By analogy to the use of a chloride-free process in the synthesis of TEOS, one can use a second approach to 3MS that employs a chloride-free chemistry developed by Gelest. In the first step, Gelest treats hexamethyldisilazane (HMDS) with an alcohol to form an alkoxy derivative, also called a silicon ester:

Then they directly reduce the ester to 3MS according to:

This method features high yield and overall purity, and it results in a chloride-free 3MS product. Disadvantages include the relatively high cost of both the silicon ester and suitable reducing agents.

Tetramethylsilane, 4MS

Tetramethylsilane has historically been used as an internal standard in nuclear magnetic resonance (NMR) spectroscopy. Thus it had a minor application, largely as a research chemical, prior to its introduction as a CVD low-k precursor. In recent investigations into the environmental abatement of 4MS in the CVD low-k process, ATMI demonstrated that the chemical is more kinetically stable, and thus more resistant to decomposition, than other alkylated silanes. We can synthesize tetramethylsilane through the methylation of trimethylchlorosilane:

Because trimethychlorosilane is a product of reaction 5, one can prepare 4MS using a route similar to the traditional synthesis of 3MS. In both preparations, the fabricator must take great care to remove traces of chloride and other impurities.

Dimethyldimethoxysilane

Siloxanes, both acyclic and cyclic, can be used in the fabrication of OSG films. We show examples of both, starting with dimethyldimethoxysilane, (CH₃)₂Si(OCH₃)₂. Its use in CVD film fabrication has been described.¹⁴ Dimethyldimethoxysilane is prepared though the reaction of dimethyldichlorosilane (from reaction 5) with methanol:

As with other precursors, one must remove all traces of chloride from the dimethyldimethoxysilane. Unfortunately, highly toxic methanol is released when the CVD low-k precursor is used.

TMCTS

The cyclic 1,3,5,7-tetramethylcyclotetrasiloxane (TMCTS) may also be used as a low-k CVD precursor:^15-17

TMCTS is prepared through hydrolysis of methyldichlorosilane to firstly form a linear siloxane polymer that is end-capped with trimethylsilyl groups (derived from trimethylchlorosilane) according to:

The linear siloxane polymer is subsequently "backcracked" to form a combination of cyclic siloxanes, predominantly 6-, 8- and 10-membered rings:

The company then distills the product to separate the TMCTS from other products of the depolymerization.

OSG packaging, storage, delivery

As with most traditional CVD materials, the new low-k precursors are either gases (e.g., methylsilane and dimethylsilane) or liquids — including tetramethysilane (4MS) and alkoxy-substituted silanes and siloxanes. As such, the majority of these low-k precursors are amenable, with only slight modification, to the chemical packaging and CVD delivery equipment utilized with traditional CVD dielectric precursors. However, the physical properties of one important CVD low-k precursor, trimethylsilane, introduce significant challenges to chemical transport and delivery in the fab environment.

Trimethylsilane is a highly flammable, liquefied compressed gas with a relatively low vapor pressure of ~8 psig at 21°C, and a boiling point of 6.7°C. CVD low-k processes require flow rates as high as 2.0 slpm per chamber or greater. While personnel can transport 3MS through the fab as a liquid (like TEOS), adsorption of inert "push gases" would result. Conversely, the use of traditional, positive pressure, gas phase distribution methods creates conditions that support the liquefication (condensation) of 3MS in distribution tubing. Furthermore, 3MS's relatively low vapor pressure and high heat of vaporization (5.8 kcal/mole) readily cause the source vessel to drop below atmospheric pressure, generating flow perturbations. These fluctuations can cause mass flow controller errors and contribute to CVD tool downtime.

Since neither traditional liquid nor conventional gas delivery schemes appeared suitable for 3MS, ATMI developed an approach using a "sub-atmospheric fab-wide storage and delivery system."¹⁸ The delivery system incorporates an integral vacuum actuated cylinder¹⁹ for the fab-wide 3MS source. The system further uses an adsorbent-based, sub-atmospheric, point-of-use reservoir²⁰ for local storage and immediate availability of 3MS and a sub-atmospheric gas control system.²¹ The user avoids condensation problems by withdrawing the 3MS directly from the source vessel outlet as a sub-atmospheric vapor. By positioning the sub-atmospheric 3MS reservoir at or near the CVD tool, 3MS flow perturbations are eliminated. Finally, in addition to providing an uninterrupted supply of 3MS to multiple CVD tools, the 3MS delivery system increases safety by removing liquefied compressed gas cylinders and change-outs from the sub-fab. This dramatically reduces the capital and installation costs of the delivery system.

Conclusions

At the 0.13 µm device generation, leading-edge companies are beginning to incorporate low-k dielectric films in their copper damascene interconnect structures. From the standpoint of complexity, the synthesis, purification and handling of a low-k CVD precursor is more difficult to prepare than TEOS, but is considerably less complex than spin-on low-k polymeric films. These new CVD low-k films are being prepared using a wide range of organically substituted silicon-based chemicals. We showed the various synthesis techniques, noting the advantages and disadvantages associated with each route. To handle and deliver 3MS, ATMI developed a chemical storage and delivery system that helps move 3MS into the production fab environment. •

Ravi K. Laxman is global product manager for low-k products at ATMI Materials. He has worked as a technologist in the processing of low-k dielectric films at Novellus Systems Inc. (San Jose), and as a senior principal research chemist at Air Products and Chemicals Inc. (Allentown, Pa.), where he developed commercial silicon nitride and low-k dielectric CVD precursors. Laxman received his Ph.D. in chemistry from the University of Missouri (St. Louis).
1-408-577-1400
rlaxman@atmi.com

Neil H. Hendricks is the chief technologist of ATMI, which he joined in August 1999. Since 1991, he has focused his research on low-k dielectrics for IC intermetal dielectric applications. He received his Ph.D. in chemistry from Stanford University. Hendricks has over 40 publications and 15 U.S. patents in the areas of low-k dielectrics and other advanced materials.
1-408-577-1400
nhendricks@atmi.com

Barry Arkles is the president and founder of Gelest Inc., a manufacturer of organosilicon and organometallic chemicals. He co-founded and was president of Petrarch Systems (now Sivento). Arkles received his B.S. degree and Ph.D. from Temple University. He received the Leo Friend Award from the American Chemical Society. He has published more than 100 papers in the areas of polymer and organometallic materials science.
1-215-547-1015
info@gelest.com

Terry A. Tabler manages new product and market development efforts for ATMI Materials' gas operations and previously was VP and general manager of ATMI's sensor business, EcoSys, which acquired TeloSense in 1999. He is a 20-year veteran of the specialty gas industry, has served in various production management, sales and marketing positions, and was responsible for starting up Praxair's site gas management business. Tabler received his B.A. in chemistry from Wittenberg University (Springfield, Ohio).
1-203-794-1100
ttabler@atmi.com

---

REFERENCES

1. B. Arkles, Silicon Compounds (Silanes), Kirk-Othmer Encyclopedia of Chemical Technology , Vol. 22, Wiley, 1997, p. 38.

2. B. Arkles, Silicon Compounds (Ethers and Esters), Kirk-Othmer Encyclopedia of Chemical Technology , Vol. 22, Wiley, 1997, p. 69.

3. O.S. Nakagqwa et al., Proc. of the U.C. Berkeley Advanced Metalization and Interconnect Systems for ULSI Applications Conf. , Oct. 1995, Portland, Ore.

4. C.H. Ting and T.E. Seidel, MRS Proc. , Vol. 381, p. 3.

5. K.R. Carter et al., ibid, p. 197.

6. S.-P. Jeng et al., ibid., p. 197.

7. P.L. Pai and C.H. Ting, Proc. IEEE VMIC Conf. , 1989, p. 258.

8. J. Ida et al., Digest of Symp. on VLSI Technology , 1994, p. 59.

9. N. Hendricks, Proc. of the 6th International Dielectrics for ULSI Multilevel Interconnection Conf. (DUMIC) , Santa Clara, Calif., Feb. 2000, p. 17.

10. J. Von Ebelman, Analytical Chemistry , Vol. 57, 1946, p. 319.

11. R.M. Levin and K. Evans-Lutterodt, J. Vac. Science and Technology , Vol. B1, No.1, Jan-Mar 1983, p. 54.

12. R.S. Rosler, Solid State Technology , June 1991, p. 67.

13. G. Kreuzberg, A. Lenz, and W. Rogher (to Dynamit Nobel), U.S. Patent 4,113,761, Sept. 12, 1978.

14. N. Matsuki, U.S. Patent 935283, Aug. 11, 1999, to ASM Japan.

15. K. Gleason et al., U.S. Patent 6,045,877, April 2000.

16. Rose et al., U.S. Patent 6,068,884, May 2000.

17. A. Grill et al., MRS Spring Proceedings, 1999.

18. L.P. Wang and T. Tabler, patent pending.

19. U.S. Patents 6,101,816 and 6,089,027 and other patents pending.

20. U.S. Patent 6,027,547, U.S. Patent 5,985,008, and related patents and patents pending.

21. L.P. Wang, J. Dietz and T. Tabler, patent pending.

Talkback

We would love your feedback!

Post a comment

» VIEW ALL TALKBACK THREADS

Related Content

• Topics
• Author

Related Content

 

By This Author

There are no other articles written by this author.

SPONSORED LINKS