Ion beam optimization to reduce EUV mask blank defects
PATRICK KEARNEY and FRANK GOODWIN, SEMATECH, Albany, NY
A likely cause of many of the defects seen on EUV mask blanks is the ion beam missing the target of the mask blank deposition system. This analysis provides a possible solution.
Extreme ultraviolet lithography (EUVL) is currently scheduled for high volume manufacturing starting in 2015. One of the top remaining hurdles to its implementation is the defect level of mask blanks. SEMATECH is working to accelerate progress in defect reduction by partnering with mask blank suppliers to improve the mask deposition technology at the Mask Blank Development Center (MBDC) in Albany, NY.
EUVL masks are composed of a substrate coated with a reflective molybdenum/silicon multilayer. The multilayer consists of 80 alternating layers of 4nm silicon and 3nm molybdenum. EUVL mask blanks are produced by ion beam sputtering in Veeco Nexus tools . Figure 1 shows the inside of one of our mask blank deposition tools. From left to right are the target turret, the ion source, and the mask holder. The ion beam leaves the ion gun and ideally strikes a small spot (~4" diameter) on the 12" target. The ions that strike the target knock off atoms of target material, which are then launched away from the target. Some of these atoms are caught on the mask substrate, forming a film. When a layer has been deposited, the ion beam is turned off and the alternate target is moved into the ion beam position. The ion beam is turned on, and the subsequent layer is deposited. Layer thicknesses are controlled by controlling the beam voltage and current and by adjusting the deposition time of each layer. This deposition technique can meet the stringent uniformity and reflectivity specifications of EUVL masks and has the lowest proven defect level of any deposition technique for such masks.
FIGURE 1. Veeco Nexus ion beam deposition tool used in this work.
Tool shield interaction
One source of mask blank defects added during deposition is the interaction of the ion beam with the tool shields. Figure 2 shows a stainless steel defect embedded inside the multilayer that we believe came from the shields. The defect was imaged using high angle annular dark field (HAADF) imaging, and the chemical analysis was performed using energy dispersive X-ray mapping by a scanning transmission electron microscope (STEM-EDS). The defect, which clearly contains iron and chrome, reached the substrate during the deposition of the multilayer. The shields are stainless steel that have been texturized by alumina grit blasting. They are subsequently cleaned to remove the residual alumina. This allows the surface to handle thick coatings without flaking, but leaves the surface fragile under ion irradiation. Ions that hit the shield will sputter away the shield material and can eventually liberate particles of the shield material. Understanding why the ion beam strikes these shields is crucial to being able to reduce this defect source.
FIGURE 2. STEM HAADF cross sections and STEM-EDX compositional analysis of an embedded stainless steel defect.
Ion beam formation
Ion source operation involves setting many input parameters including the beam voltage that determines the ion energy, the suppressor voltage that influences beam divergence and electron backstreaming into the source, the RF power that supplies the plasma, the flow and type of gas that forms the ions, and the chamber pumping speed. Complex limits on the input parameters determine when the beam is stable enough to use for production. These input parameters also interact to influence the divergence of the beam. The theory of ion beam formation for these sources is based on the work of Kaufmann et al. , which predicts a beam that is well confined on the target for normal beam parameters. Because experiments have shown evidence of etching outside the target, we decided to further investigate the beam and this etching.
A simple way to measure the beam profile was to attach oxide-coated silicon wafers to the target and etch them with the ion beam. As the oxide layer is etched away, the wafer changes color. Qualitatively, the beam erosion profile can be estimated by observing the color pattern on the etched wafer; quantitatively, data can be extracted from analysis of photographs of the wafers. Charging of the oxide wafer was prevented by over-neutralizing the ion beam. The influence of the ion source parameters on the beam profile was studied with the goal of confining more of the beam to the target. The beam focus was found to improve as the suppressor voltage was lowered, the beam voltage raised, and the RF power adjusted to maintain the beam current at an optimum level that depended on both the gas type and voltages that were used. Importantly, the behavior of the optimum beam current with ion mass and the improvement in beam focus with beam energy were different from what the theory predicted, which would have led us to operate the ion source with a wider beam than necessary. Figure 3 compares images of the etch patterns under previous ion source conditions and under the optimum conditions found in this work. On the left is an image of a wafer etched at our initial operating point (600V beam voltage and 500V suppressor voltage). On the right is an image of a wafer etched at our optimized operating point (1500V beam voltage and 200V suppressor voltage). In both cases, the time was set to etch approximately 50 microns of material from the wafer at the point of maximum etch. The outer teal ring on each sample represents a contour of a constant etch rate of approximately 0.25% of the maximum etch intensity. The figure indicates that for the initial operating point this contour extends beyond the edge of the target, while under the optimized condition the contour is contained on the target. The optimized beam parameters should result in less beam missing the target, leading to fewer shield defects on the mask.
FIGURE 3. Target beam profiles for initial and optimized source parameters showing much better beam containment on the target when parameters are optimized.
Finding the source of defects
To obtain more information about the mechanisms behind the etching, we measured where in the chamber the etching species originated using a pinhole camera with an oxide wafer as the recording medium (Fig. 4). Initial measurements with an Impedans Semion  retarding field analyzer failed to show any ions striking the shields in the area of the etch. This suggests that the etching species are neutrals, not ions. The obvious sources for neutral atoms with enough energy to etch the shields are charge exchange collisions either in the ion beam or inside the source. A high energy ion comes close to a neutral atom with a thermal velocity. An electron from the atom crosses over to the high energy ion, resulting in an ion and a high energy neutral. Cross sections for charge exchange collisions can be larger than for physical collisions between ions and atoms.
FIGURE 4. Pinhole camera image of ion source, ion beam and target and an optical image of the same locations.
We are currently improving the model of the ion source/ion beam in collaboration with Tech-X Corporation . The plan is to use their VORPAL software to model the plasma within each beamlet of the ion beam, including gas scattering and charge exchange collisions. The model should be able to more realistically predict the ion beam profile throughout the chamber as the source parameters are varied. We are also using the pinhole camera technique to quantify the amount of etch from both the ion source and ion beam as the beam parameters are varied. This data should help us keep the ion beam on the target and reduce ion beam-induced shield defects.
1. Veeco Instruments, Plainview, NY, USA. www.veeco.com
2. Kaufman, H., R., Cuomo, J., J., Harper, J. M., E., "Technology and applications of broad-beam ion sources used in sputtering. Part 1. Ion source technology," J. Vac. Sci. Technol., 21(3), 725-736 (1982).
3. Impedans, Dublin, Ireland, www.impedans.com
4. Tech-X Corporation, Boulder, CO, USA, www.txcorp.com
Patrick Kearney is an EUV Technical Expert and FRANK GOODWIN is the program manager for SEMATECH's EUV Mask Blank Defect Reduction Program and Mask Blank Development Center., Albany, NY. e-mail: Patrick.Kearney@sematech.org.
Solid State Technology, Volume 55, Issue 4, May 2012