By Ed Korczynski, Sr. Technical Editor
Atomic-Layer Etch (ALE) technology from Lam Research Corp. is now in beta-site evaluations with IC fabrication (fab) customers pursuing next generation manufacturing capabilities. So said Dr. David Hemker, Lam’s senior vice president and chief technical officer, in an exclusive interview with Solid State Technology and SemiMD during this year’s SEMICON West trade-show in San Francisco. Hemker discussed the reasons why ALE is now under evaluation as a critically enabling technology for next generation IC manufacturing, and forecast widespread adoption in the industry by 2017.
As detailed in the feature article “Moving atomic layer etch from lab to fab” in last December’s issue of Solid State Technology, ALE can be plasma enhanced with minor modifications to a continuous plasma etch chamber. The lab aspects including the science behind the process were discussed in a TechXPOT during SEMICON West this year in a presentation titled “Plasma Etch in the Era of Atomic Scale Fidelity” by Lam’s Thorsten Lill based on work done in collaboration with KU Leuven and imec. In that presentation, Lill reminded the attendees that the process has been explored in labs under a wide variety of names: ALET, atomistic etching, digital etch, layer-by-layer etch, PALE, PE-ALE, single layer etch, and thin layer etching.
ALE can be seen as a logic counter-part to atomic-layer deposition (ALD), with the commonality that both processes become cost-effective when the amount of material being either added or removed are readily measured in atomic layers. It’s comforting that when the industry needs control to the atomic-level we are dealing with such tiny structures that ALD and ALE can provide acceptable throughputs. “By 2017, we see able 15% of the opportunity for us could be addressed by atomic processing,” projected Hemker.
However, ALE as promoted by Lam differs from ALD, because etch processes generally need directionality. “That’s where it diverges from ALD,” explained Hemker. “Using ions we get all the benefits of directionality and selectivity. Likewise, if we design the process correctly, we could theoretically have infinite selectivity with under layers.” Figure 1 shows a trench formed in single-crystal silicon using ALE, with vertical side-walls and a bottom surface smooth at the atomic scale. Such process capability is based on the pulsing of both energy and chemistry into the reaction chamber.
Fig. 1: (Left) Schematic cross-section of Atomic-Layer Etch (ALE) of silicon using a silicon-oxide top mask, (Middle) SEM cross-section of nominal 40-nm silicon trench, and (Right) TEM close-up of the silicon surface showing atomic-scale smoothness.
“We need to be able to pulse multiple things at the same time,” explained Hemker. “So we can absorb a reactant, and then switch over to a plasma. The breakthrough in this is being able to pulse everything correctly.” Labs have been doing this but on a timescale of minutes per atomic layer removed. Lam productized the principle to run on a time-scale of seconds on the 2300 Kiyo tool, which is the current leading-edge hardware for conductor etch from the company.
Pulsing of energy into a reaction chamber has been used in the company’s high aspect-ratio etch process for 3D NAND which runs on the 2300 Flex tool for dielectrics. In this process flow, vias through alternating layers of oxide and nitride in a stack must be etched at 40:1 aspect-ratio today, with 60:1 and even 100:1 aspect-ratio specifications from Samsung for device evaluations. “You see it coming in with pulsing the plasma, allowing us to get ions in and reactants out,” explained Hemker. So the ALE process can be seen as an extension of this pulsing plasma approach, with the extra sophistication of pulsing the chemical precursors into the chamber. “The trick is how to do it repeatably and reliably so that it’s production worthy,” reminded Hemker.
When the ALE precursor adsorbs as a single-layer on surfaces, the connection to the surface could be merely van der Waals forces, or depending upon the application could include some reaction with underlying atoms. “The process conditions have to tailored for flows and gases, but it does open up the possibility of using less expensive process gases. There’s no new gases needed,” declared Hemker. “The real message is not that this is just a new process, but this shares a common background with ALD in pulsing things and having sophisticated enough control of the process.”
Such commonalities would seemingly extend to some chamber hardware and the vacuum and effluent abatement systems, such that it would be very straightforward to cluster single-wafer processing chambers for ALD with ALE with plasma pre-treat and possibly even with annealing. Such a cluster would allow for sophisticated “dep/etch” recipes to be developed for atomic-scale device fabrication.
Fig. 2: Commonality in the need for ALD and ALE process technologies when IC device dimensions scale to atomic levels.
Figure 2 shows the comparison between ALD and ALE processes for a trench structure, and why both are needed when device geometries reach atomic-scales. When trench aspect ratios (AR) are ~1:1 continuous deposition and etch processes can be fairly easily developed to provide uniform results. However, as the AR increases, reaction byproducts tend to non-uniformly deposit on sidewalls and especially at the corners of structures. Eventually, the top of high AR trenches “pinch-off” to create an open in IC circuitry, even when slowing down continuous processes to allow more time for byproducts to escape reaction areas.
Lam expects ALE to be used on the leading-edge of IC manufacturing within a few years, with increasing applications as more critical layers in a device must be patterned to smaller than 22nm half-pitch. “It’s not that you can’t do some of these processes with continuous etch, but ALE really opens up the process window,” explained Hemker. Now is the time for ALE, since the minimum variability of continuous etching consumes more and more of the critical dimension with ever smaller feature sizes.
“If you look at ALD as the for-runner of this, it was first adopted for capacitor deposition in a batch process, then it migrated to single-wafer for high-k metal-gate formation where greater control was needed,” reminded Hemker. “It was used but somewhat niche, and now we’re seeing traction on ALD for many more applications such as quadruple-patterning. The spacers themselves have to be perfectly conformal, because any thickness variation will be a CD variation and it compounds with quadruple patterning.”
Control of pattern fidelity at the atomic-scale will be needed as the commercial IC fab industry integrates new materials for improved device functionalities. ALE and other technologies that can control processing of individual atomic layers should be used to pattern ICs for the indefinite future.