by Katherine Derbyshire, contributing editor, Solid State Technology
Nov. 26, 2008 – Though fabs and their suppliers are making enormous investments in state-of-the-art masks and exposure systems, lithography is only part of the story. The ultimate goal is pattern transfer — replicating the mask pattern on the wafer itself to complete a layer of the growing circuit. Typically, a patterned photoresist layer protects the wafer during an etch or ion implantation step, then is removed to make way for the next layer.
For accurate pattern transfer, the photoresist must both capture the image defined by the exposure system and preserve that image through the subsequent transfer process. These two requirements come into conflict as feature sizes shrink and aspect ratios increase. As lithographers extend 193nm lithography into the deep sub-wavelength, hyper-NA regime, the depth-of-focus of the aerial image goes down. The process window can be interpreted as a measure of the maximum resist thickness — if the resist is thicker than the zone of best focus, then non-uniform exposure result. Yet a thin resist layer might not survive etching of a high aspect ratio wafer feature. In memory devices, the problem is particularly severe for trench etches in the 4xnm and smaller generations. For logic, gate etches become especially challenging for the 32nm and smaller generations. In both cases, ashable hardmasks are emerging as a possible solution.
As Novellus SVP Tim Archer explained in a recent interview, an ashable hardmask (AHM) is typically an amorphous carbon layer, deposited by PECVD from a CxHy precursor, and removable by standard oxygen plasma-based resist ashing. The choice of precursor defines such key properties as etch selectivity and mechanical hardness. Etch selectivity is important because a more highly selective layer can be thinner and still provide adequate etch protection, eroding less over a given etching time. Thinner layers often can be deposited and removed more quickly as well.
Mechanical properties are especially important because the hardmask must maintain its shape as the film’s stress distribution changes during etching. Narrow lines in a thick mask layer can topple over, unable to support their own weight. If the film is not stiff enough, it can relax in the direction of the patterned lines (parallel to the wafer plane), creating wavy feature edges (see figure). As Archer explained, the film precursor also defines the deposition rate, and therefore the cost of the process. Novellus’s AHM systems use acetylene, but propylene, toluene, methane, and other hydrocarbons can also be used.
Post-etch line bending in an ashable hardmask layer. (Image courtesy Novellus).
Good step coverage is also important for a successful AHM. Deposition of the AHM is followed by the anti-reflective layer (ARL), then the actual photoresist. Uniform ARL coverage is hard to achieve over topography, so the AHM must smooth out whatever steps are already present on the wafer. Good ARL coverage is especially important in case resist rework is needed. If the resist can be removed cleanly, then the wafer can be reworked by simply applying fresh resist. If the resist removal step punches through the ARL to the AHM layer, however, rework is no longer possible and the wafer must be scrapped.
Novellus expects the number of AHM layers to continue to climb as feature sizes shrink, with the PECVD AHM market segment likely to reach $300M by 2012. The company’s Vector Extreme AHM system integrates post-deposition edge bevel film removal, eliminating the need for a separate removal step, and achieves throughput in excess of 150 wafers/hour. –K.D.