Air Gaps Are a Low-k Alternative

Let's face it. The air gap is a pretty crazy idea. Rather than filling the space between interconnecting wires with anything that would increase line-to-line capacitance and RC delay, just go straight for the best performing low-k dielectric of all: Air. More unusual things have happened.

Like effectively taking sandpaper to your device — later perfected into the state-of-the-art technology of CMP. And purposely engineering strain into the channel of transistors to increase performance. Balancing the stresses of tensile and compressive within nanometers in a device to separately optimize PMOS and NMOS transistors is both brilliant and, as it turns out, manufacturable. So when scaling gets really tough, engineers find alternatives.

The air gap may be the alternative to porous low-k dielectrics that the industry needs at this time. Porous dielectrics have too many problems — at least as many as their dense counterparts — and the overall benefit they deliver, represented as the effective k value, may be small given the integration, yield and reliability challenges they pose and the costs required to surmount them.

We published one of our first articles on air gaps in July 1999: "Gas Dome Dielectric System Provides Unity-k Dielectric ," by Thomas Wade of the University of South Florida. The article highlighted important requirements for a low-k dielectric, including process compatibility; good thermal properties; and mechanical, chemical and thermal stability. All of these issues were to become problems for low-k dielectrics. Advantages of air gaps include no intrinsic stress, no adhesion or planarization problems, no defect density issues, and, of course, low cost. The gas dome approach used a rigid oxide dome to support the overall device with SiO₂ at the topmost metal layers. Lower interconnects were separated only by gas, with some supported vertically by SiO₂ pillars.

I will admit that, in mid-1999, the gas dome approach was both radical and untested — the latter of which led to criticism. Wade's article aroused a few negative verbal and written responses regarding the publication of such a controversial paper on such an unknown (and some said ridiculous) technology. I recall being surprised at the voraciousness of some of these responses, so I decided to closely watch the developments of this alternative technology. While most people remained skeptical from year to year, I saw a small but steady number of articles on air gaps appear in the IEDM and other journals. Recently, I had discussions about the "air gap phenomena" with researchers at IMEC — people investigating manufacturing solutions for 45 nm and beyond. As Karen Maex, IMEC fellow, put it, "We had early successes with air gaps, which was much more than we could say about the porous dielectrics." She warned, however, that early successes do not always lead to later achievements, and emphasized the importance of testing hypotheses and models in silicon.

Some of the challenges of porous low-k materials include the need to seal the exposed dielectric after the vias and trenches have been etched, prior to barrier metal deposition. Without pore sealing, the barrier rapidly diffuses into pores, as does any chemical or gas the dielectric is exposed to during photoresist stripping and cleaning. Maex said the most promising pore sealing methods today are plasma-based, but the industry is far from having a production-ready solution. Engineers are working hard at increasing the mechanical stability of porous dielectrics. In narrow structures, the sealed low-k edge could consume a large portion of the real estate, making achievement of low-k most difficult where it is most needed.

So in this age when the D in R&D is emphasized more heavily than the R, it might behoove the industry to not discount seemingly ridiculous process alternatives simply because they initially appear unlikely. At one time, copper would not be brought into a fab. Optical lithography resolution was limited to 1 µm. Rapid thermal processors melted silicon wafers. The seemingly insurmountable has been mastered and made production-worthy. Perhaps air gaps are among these solutions.