Plain talk on low-k dielectrics
06/01/2003
Why can't the industry make a solid film with k = 2.2, a Young's modulus (E) as high as 10GPa, and leakage current as low as 1x10-10A/cm2 in combination? The answer is: Three simple golden rules for making good low-k materials have been violated.
When it comes to material science, true knowledge is mostly based on a few very simple principles. First, a low-k material must not contain unpaired electrons. Any unpaired electron will be subject to polarization under an electric field and will form hydrogen bonding with water. All current low-k materials violate this simple rule, and so they all adsorb water and have high leakage current.
A second principle: plasma cannot directly be used for making any designed materials. Fundamentally, plasma reactions are not selective for breaking specific chemical bonds because plasma generates many chemically different reactive species (cation, radical, anion, etc.); thus, it cannot be used to make any "designed" organic polymers, or good low-k materials with a high yield.
A 1992 published paper showed the results of some of the plasma polymerized siloxanes similar to many of the current SiCOH type materials such as Black Diamond or Coral [1]. It is now known that plasma polymerized polysiloxanes networks leave too many dangling chain ends that are bad for dielectric loss under high frequency in the microwave region. Additionally, some of these films tend to have high leakage current compared to FAR2.2 (Fig. 1).
Furthermore, doping carbon and hydrogen by plasma will fall short of achieving a good low-k film with k <2.7 and a Young's modulus >4GPa. Too many C and H atoms are needed to be included into the siloxane networks to achieve k = 2.7 (from 4.0 of SiO2); therefore the cross-linkage density of the resultant SiO-networks will be too low for sustaining a sufficiently high Young's modulus.
 Figure 1. Leakage current as a function of the electric field for the FAR2.2 film. |
null
The third golden rule is: "spin-on" low-k polymers are not sufficiently rigid. The fundamental science to make high crystalline polymers form CDP (crystallization during polymerization) was presented in 1978 [2]. At that time, it was predicted that CDP would be useful for making thin films for applications in semiconductors. CDP still remains the best method to make "designed" low-k polymers with high crystallinity and a high Young's modulus.
Since 1985, when polyimidesiloxanes (k = 2.7) were first made at General Electric (polyimidesiloxanes, k = 2.7), nearly 20 years were spent (by Chung Lee) making low-k material from a spin-on process, before realizing it is not the preferred method for making the best low-k polymers for IC fabrication [3, 4].
 |
null
When dissolved in a solution, all polymers, in general — except for liquid crystal polymers — will have polymer chains presented as random coils. After spin-on and removal of the solvent, these polymer chains will be in the amorphous state. Following the entropy principle (second law of thermodynamics), there are very few ways for a polymer to form a highly crystalline state from its molten or solution state. For this reason, a polymer with very high cohesive energy can be designed, but it cannot have a high cohesive energy density because polymer chains are loosely packed in the amorphous state. Only when most of the polymer chains are in proximity to each other, which occurs in the crystalline state, can a polymer material approach its full potential of high cohesive energy density and high Young's modulus. The importance of cohesive energy density is explained by considering molecular design.
Molecular design for low-k
The fundamental principles useful for predicting low-k properties are based on the concept of a measurable physical property of polymers — the heat capacity, or a derivable cohesive energy density (CED) of polymers. It was demonstrated in 1988 that the glass transition temperature, Tg, of all semi-rigid, thermally stable polymers can be calculated using the group contribution of heat capacity jump, ΔCP, and group contribution of the cohesive energy from their repeating units [5].
Modern materials science already provides ways for predicting almost all critically important properties of low-k materials such as dielectric constant, k, decomposition temperature, Td, Young's (E) and shear (G) modulus (for CMP), glass transition temperature, Tg, and melting temperature, Tm. Proper design of useful polymer structures can be done to meet requirements for IC fabrication before the low-k polymers are actually synthesized. This approach will eliminate the trial and error approach to making so many polymers and going through very costly integration processes.
The atomic contribution for C, F, and H in various bonding states (i.e., Sp3C-X, sp2C-X and sp3C α -X) for most low-k materials has been derived. The physical properties listed above have been calculated for many polymers and some have been verified. The table on p. 000 summarizes calculated results for some amorphous polymers; there are many useful polymers yet to be made. Using a CDP process, the Young's modulus of these crystalline polymers can be as high as 10–20GPa [6].
A database can derive the effects by replacing a Sp3C-H bond with a Sp3C-F bond on the properties such as k, E, G, Td, Tg, and Tm for polymers and explains why all C-F bonds are not equal, addressing concerns about fluorine instability in fluorinated low-k materials. The molecular structure can result in a very stable fluorine bond or, on the other hand, result in F-doped α-carbon type material with poor thermal stability.
 Figure 2. Thermal decomposition scan (TDS) of the FAR2.2 film. |
null
Figure 2 shows the thermal decomposition scan (TDS) results for the thermal stability of the C-F bonds in DSI's low-k, FAR 2.2 (k = 2.2) solid film that, depending on its crystallinity, has a Young's modulus ranging from 6–12GPa.
Impact of CTE on yield
The importance of achieving CED for the designed low-k polymers cannot be stressed enough. Large values of CED provide large values of Tg, Tm, G, and Young's modulus, and a low coefficient of thermal expansion (CTE). It is undesirable to lower the density in order to achieve low-k because this action results in reduction of cohesive energy density as well as the value of Young's modulus.
A large Young's modulus is favorable for using a thinner cap layer during CMP in the current copper dual damascene process, so it can result in a lower effective dielectric constant, compared to a low-k film that has a small Young's modulus value but an identical dielectric constant.
In addition, more than 95% of failures in multichip modules (MCM) manufacturing are thermo-mechanical in nature, namely, CTE mismatch. Despite the so-called "low CTE-polyamide"(having an xy-CTE ranging from 3.6–6ppm/°C) used in the MCM, its z-CTE can be as high as 40–80ppm/°C [7]. The CTE mismatch
problem becomes more pronounced when the low-k /Cu layers in a chip metallization are increased beyond three levels. In fact, increasing the low-k /Cu layers to six has been shown to dramatically lower the MCM yield during manufacturing steps.
CTE mismatch can also wreak havoc with yield in IC chips made from current low-k (k = 2.7) materials that have very high CTE relative to copper. As a testimonial, a device manufacturer used a commercially available k = 2.7 dielectric film in conjunction with a copper dual damascene process. The manufacturer encountered a drastic yield loss. After a thorough investigation, the failure mode was attributed to the CTE mismatch. The CTE for Si, copper, and underfill for this flip-chip package are 2.5, 17, and 25–30ppm/°C, respectively. Ceramic and epoxy substrates for this package are 5.8 and 17ppm/°C, respectively.
Conclusion
To design a practical low-k dielectric film applicable to IC manufacturing, the three fundamental principles, or "golden rules," should not be violated. Lack of attention to CTE mismatch will result in a premature IC failure, which will diminish the packaging yield. Some of the commercially available low-k (k = 2.7) materials have CTE values ranging from 50–65ppm/°C; this range will not be suitable for reliable IC packaging applications.
The recent introductions of the low stress, noncontact CMP may temporarily qualify some low-k and high CTE films, including low-k porous films, during the integration phase. Many of these films will ultimately fail the packaging reliability test.
Low-k solid films with high Young's modulus and low CTE (e.g., FAR2.2 with a k = 2.2, E ranging from 6–12GPa, and a CTE from 14-20ppm/°C) provide a much needed solution to the manufacturing of next generation semiconductor devices.
Chung J. Lee, Atul Kumar, Abe Ghanbari Dielectric Systems, Fremont, California
Acknowledgments
The authors wish to thank Dr. Takuya Fukuda from the ASET organization in Yokohama Research Center, Japan, for his valuable insight and suggestions and for his contribution in analysis and testing of the FAR2.2 film.
Black Diamond is a registered trademark of Applied Materials Inc. Coral is a registered trademark of Novellus Inc. FAR2.2 is a trademark of Dielectric Systems Inc.
References
1. C.J. Lee, et al., "Elastic Polysiloxanes Thin Films Obtained from Plasma Polymerization," 6th SAMPE Electronic Conference, 1992.
2. C.J. Lee, "Transport Polymerization of Gaseous Intermediates and Polymer Crystal Growth," J. Macromol Sci.-Rev. Macromol Chem., C16 (1), pp. 79–127, 1977–1978.
3. C.J. Lee, "Novel Soluble Silicone-Imide Copolymers," 32nd Int'l SAMPE Symposium, p. 576, 1987.
4. C.J. Lee, "Polyimidesiloxanes: Chemistries and Applications," Chapter 5 in C.P. Wong, ed., Polymers for Electronic and Photonic Applications, Academic Press, 1992.
5. C.J. Lee, "Polyimides Polyquinolines and Polyquinoxalines: Tg-Structure Relationships," Rev. Macromol. Chem. Phys., C29 (4), pp. 431–560, 1989.
6. C.J. Lee, et al., DSI patent applications Nos. 09,795,217, 10/028,198, and 10/029,373.
7. D. Petila, L. Debra, C.J. Lee, "Effects of Processing Conditions on the Thermal Properties of PIM Thin Film for HD I/P," Proc. MRS Fall Meeting, 1990.
For more information, contact Chung J. Lee, cofounder and president of Dielectric Systems Inc., 45500 Northport Loop West, Fremont, CA 94538; ph 510/979-0900, fax 510/979-1237, [email protected].