Demonstration of electronic design automation flow for massively parallel e-beam lithography

Abstract

For proximity effect correction in 5 keV e-beam lithography, three elementary building blocks exist: dose modulation, geometry (size) modulation, and background dose addition. Combinations of these three methods are quantitatively compared in terms of throughput impact and process window (PW). In addition, overexposure in combination with negative bias results in PW enhancement at the cost of throughput. In proximity effect correction by over exposure (PEC-OE), the entire layout is set to fixed dose and geometry sizes are adjusted. In PEC-dose to size (DTS) both dose and geometry sizes are locally optimized. In PEC-background (BG), a background is added to correct the long-range part of the point spread function. In single e-beam tools (Gaussian or Shaped-beam), throughput heavily depends on the number of shots. In raster scan tools such as MAPPER Lithography's FLX 1200 (MATRIX platform) this is not the case and instead of pattern density, the maximum local dose on the wafer is limiting throughput. The smallest considered half-pitch is 28 nm, which may be considered the 14-nm node for Metal-1 and the 10-nm node for the Via-1 layer, achieved in a single exposure with e-beam lithography. For typical 28-nm-hp Metal-1 layouts, it was shown that dose latitudes (size of process window) of around 10% are realizable with available PEC methods. For 28-nm-hp Via-1 layouts this is even higher at 14% and up. When the layouts do not reach the highest densities (up to 10∶1 in this study), PEC-BG and PEC-OE provide the capability to trade throughput for dose latitude. At the highest densities, PEC-DTS is required for proximity correction, as this method adjusts both geometry edges and doses and will reduce the dose at the densest areas. For 28-nm-hp lines critical dimension (CD), hole&dot (CD) and line ends (edge placement error), the data path errors are typically 0.9, 1.0 and 0.7 nm (3σ) and below, respectively. There is not a clear data path performance difference between the investigated PEC methods. After the simulations, the methods were successfully validated in exposures on a MAPPER pre-alpha tool. A 28-nm half pitch Metal-1 and Via-1 layouts show good performance in resist that coincide with the simulation result. Exposures of soft-edge stitched layouts show that beam-to-beam position errors up to AE7 nm specified for FLX 1200 show no noticeable impact on CD. The research leading to these results has been performed in the frame of the industrial collaborative consortium IMAGINE. Downloaded From: http://nanolithography.spiedigitallibrary.org/ on 05/15/2015 Terms of Use: http://spiedl.org/terms

Alternative stitching method for massively parallel e-beam lithography

“…It was demonstrated previously that by a combination of over exposure of a pattern with negative bias, throughput can be exchanged for Exposure Latitude (EL) [3]. The key observation for the alternative stitching method proposed below is that the stitching region is by its definition passed by two beams and without cost in throughput, a dose can be deposited that is 2x higher than outside the stitching region.…”

“…[3]) a Monte Carlo (MC) approach is taken, as on top of the Process Window parameters exposure dose and tool spot size (the e-beam equivalent of focus) the beam to beam position error is varied. For each simulation site the CDU, expressed as |µ|+3σ deviation from target CD, is calculated over 1000 MC runs.…”

“…In this study, a Double-Gaussian Point Spread Function [4] is assumed with FW50 α =26nm, η=0.35 and 2.5%/nm Exposure Latitude (EL) of 32 nm half-pitch Dense Lines and Spaces (uDLS), similar to [3], at a tool spot size variation range of ±2nm. The latter originates from B2B differences in spot size as well as defocus.…”

Comparison between e-beam direct write and immersion lithography for 20nm node

“…In this paper we express the widths of the Gaussians in FW50 (the diameter that contains half the current) and use, close to the values in [3]: FW50 α = 26.1 nm, FW50 β = 487 nm and η = 0.35 at a constant resist threshold. The writing spot, FW50 α , consists of the quadratic addition of the nominal tool spot size of 23 nm and a resist contribution of 12.3 nm.…”

“…The Point Spread Function of the MAPPER tool is of double Gaussian form [3][4]. The first Gaussian comprises the principal writing spot: the contribution of the incident beam, forward scattering of the electrons in resist and acid diffusion in resist.…”