2.7.4 Ion-Beam

Besides electron-beams also ion-beams can be used for lithographic applications [37]. Ion-beams can either expose the resist through a mask using broad beams (1-2 cm²), or serially by writing with a finely focused beam (0.1 nm in diameter). Because of their larger mass in comparison to electrons, ions transfer their energies more efficiently to the resist and scatter much less. The total spread including forward and back scattering of the ``stiffer'' ion-beams is typically less than 10 nm. Ion-beams also produce secondary electrons with much lower energies and hence shorter range than electron-beams. They thus require only about 1-10% of the electron dose needed to expose the resist. The resist can be an ordinary PMMA resist that absorbs most of the ions during exposure. Radiation damage to sensitive, underlying structures is therefore minimized and smaller than in electron-beam and X-ray lithography [38]. Consequently, very high resolution and sensitivity can be achieved with minor proximity and heating effects. The equivalent De Broglie wavelength is in the range of 0.28-0.9 picometer depending on the chosen element and charge state, e.g., H⁺ protons,^e as well as the accelerating voltage, e.g., 10-100 keV. Diffraction effects in mask exposure tools are therefore negligible and printing of lines down to 65 nm has recently been reported [39]. The depth of focus is limited by telecentricity requirements in the ion optics and lies typically above 10 $\mu$m. This extremely large depth of focus makes ion-beams especially attractive for highly nonplanar lithographic applications such as the fabrication of TFT displays. In today's semiconductor manufacturing focused ion-beams are mainly used in mask and circuit repair, as the exposure is still too slow due to the limited beam intensity.

In 1997 a joint European program^f was founded to develop a full-field ion projection lithography (IPL) tool with projected prototype mask-making by 1999 and commercial availability by 2001 [40]. The efforts are headed by Siemens AG, Munich, Germany. The major project partners are Ion Microfabrication Systems, Vienna, Austria; ASM Lithography, Veldhoven, The Netherlands; and Leica Lithography Systems, Cambridge, United Kingdom. The tool operates with 4x demagnification and is designed to meet the lithographic demands of sub-0.13 $\mu$m design rules, while maintaining a high throughput on 300 mm-wafers. Projected numbers are 75 wafers/hour, a minimal resolution of 100 nm, and 40 nm level-to-level overlay at exposure field up to 22 mm x 22 mm. Special features of the planned IPL tool are a stencil mask illumination system, a patented ``pattern-lock'' system for a real-time beam stabilization, and an off-axis optical wafer alignment system. The key technical challenges associated in realizing the IPL tool are the development of high quality stencil masks, control of pattern distortion and blur due to aberrations in the ion optics, and the effects of ion space charge.

For all post-optical lithographies the masks are complex and expensive. In IPL stencil masks, i.e., a thin silicon membrane typically 2 $\mu$m thick and with precisely patterned holes etched through it, are used since the energy of the incident ions is about 10 keV so that all materials are highly absorbing. The silicon membrane is fabricated by etching away the central portion of a single crystalline silicon wafer up to an etch-stop junction. Around the perimeter a thicker rim is left to act as a frame. The membrane and the rim are attached to a heavier frame of polycrystalline silicon for compatible thermal expansion and good conductivity. Conventional ion implantation controls the stress in the membrane [41]. The holes are defined by electron-beam lithography and reactive ion etching, whereby an oxide is used as a mask [42]. One obvious problem with stencil masks is that certain geometries are not allowed, e.g., an annulus-like pattern. This limitation implies that--except for simple sparse masks such as vias, active areas or contacts--a set of two complementary masks will be needed to expose one level. Another crucial mask consideration is membrane in-plane distortion that is caused by stress relief due to the etched hole patterns or due to membrane heating by the incident ions. For example, for 100 nm minimum features the error budget is only 10 nm for mask membrane distortions [40]. The stress relief distortion can be ``local'' or ``global'' depending on whether the shape of a single hole is distorted or the feature position is altered due to combined effects of all holes in the membrane. Local distortion can be moderated by segmentation of larger features on complementary masks, whereas global distortion is reduced by using stress relieving perforations around the membrane circumference. Membrane materials other than silicon may be preferable. For example, diamond has an elastic constant that is about six times larger and thermal expansion coefficient that is two times smaller. Mask heating and cooling is an extremely serious concern in IPL. The mask membrane can only lose heat by radiation and by conduction to the frame that is held at room temperature. Without corrective action the temperature of the center of the mask membrane would be expected to rise 6-10^oC above the perimeter depending on the mask emissivity [43]. The resulting nonuniform radial displacements are of the order of 100 nm and thus unacceptable. Introducing a blackened radiative-cooling cylinder in the vicinity of the mask and cooling it about 40^oC below room temperature, a uniformity better than $\pm$ 1^oC can be achieved [40]. This reduces the radial distortions to a tolerable level. The cooling cylinder makes the temperature nearly constant so that there is little lateral heat flow. This has the big advantage that the cutting of the holes will have little effect on the temperature profile. The research on mask-making technology for IC production performed so far indicates that the perceived problems of mask distortion due to stress relief as well as to mask heating can be solved.

The planned IPL system is also designed to minimize distortion and blur in the projected pattern that can arise from geometric and chromatic aberrations in the optics as well as from space charge effects. Low-order geometrical aberrations must be compensated in the ion optics. Cylindrical electrostatic lenses are only able to correct positive aberrations, whereas the combination of the stencil mask acting as a plane electrode and the nearest cylindrical lens introduces a negative spherical aberration. This combination allows the ion-optical column to compensate third-order geometrical aberrations. The remaining higher-order aberrations lead to pattern displacements on the wafer, but the wafer axial location can be optimized so that these displacements are less than 10 nm on a 20 mm x 20 mm exposure field. Chromatic aberrations couple energy spread in the ion-beam into image blur at the wafer plane. The corresponding critical dimension budget limits the axial energy spread tolerable at the ion source to less than 2 eV. Space charge interactions also introduce energy spread in the ion column [44]. Such space charge effects can be separated into global and stochastic contributions. The global space charge contributions act like a thick defocusing lens due to the average interaction with all the ions in the beam. As such, it can be canceled by additional focusing elements or, alternatively, neutralized by injecting a low-energy electron cloud in the field-free region near the crossover. On the other hand, the random close collisions of ion pairs in the beam cause the stochastic space-charge effects. These collisions are most numerous near the crossover, where the beam density becomes very high. The random, stochastic nature of these collisions means that they cannot be canceled by additional lenses or external neutralization. But since the effect of the collisions also depends on the ``correlation time'' during which the ion pairs are in close proximity, the projector can be designed for high energy and steep trajectories at the crossover to reduce the collision effect [45]. A substantially enabling technology for IPL is the ``pattern-lock'' system [46]. Hereby, the stencil mask includes a number of apertures outside the printed area to create ion alignment beamlets that traverse the ion optics on trajectories parallel to the patterned ion-beams. Scanning systems measure the displacement of the alignment beam with respect to targets in a reference plate. These measured displacements are used to set correcting voltages for the ion optics in real time as part of a feedback stabilization system. The correction elements can thereby actively adjust the beam position, magnification, rotation, and trapezoid distortion while the beam is being printed.

These combined efforts make IPL a potential candidate for post-optical lithography as it will be required for the printing of sub-100 nm minimal feature sizes [3].

Footnotes

... wavelength^d

... protons,^e

The rest mass of a proton is m₀ = 1.6725  10^-27 kg. Hence, the De Broglie wavelength^d of a proton is approximated by $\lambda$ $\cong$ 28.7/$\sqrt{V}$, with $\lambda$ in pm and V in V.

... program^f

The project is part of the MEDEA (Micro-Electronics Development for European Applications) program. For further information see the WWW at http://www.eureka.be/.