High-brightness source of energetic He atoms

The helium ion microscope (HIM) has enabled the fabrication of a class of novel devices that rely on ∼30 keV He⁺ ion irradiation to directly modify substrates at a resolution far beyond the capability of resists. Examples include the first insulating tunnel junctions in YBa₂Cu₃O_7 − δ (Ref. 1) by radiation damage to a single-crystal substrate and ultrafine graphene nanoribbons by surface ablation.² Causer et al.³ have demonstrated chemical mixing in FePt₃ at 15 keV without disrupting the crystal lattice, a critical development for applying direct substrate modification to nanomagnetic device technology. Moreover, molecular dynamic simulations suggest that ultralow energy, heavy ions (e.g., ∼60 eV Xe⁺) may enable the formation of semiconducting graphene nanoroads⁴ by selectively depopulating F-atoms on a fluographene sheet.⁵ The critical fluence required for these applications, 10¹⁶–10¹⁷ particle/cm², is extremely high compared to resist-based approaches; this, coupled with a serial writing modality, severely limits the scale of circuits that can be fabricated by an HIM. Since beam current at the highest resolution (0.5 nm) is ∼1 pA, it would take ∼22 h (plus overhead) to write a 7 × 7 mm² chip with 10% pattern density and 1 × 10¹⁶ particle/cm² exposure fluence, and 220 h for 1 × 10¹⁷ particles/cm². Moreover, the field size, 30 × 30 μm² for subnanometer resolution, is extremely small due to a limited depth of field⁶ and field stitching capability is currently absent in HIM instruments. This paper describes a source of energetic He atoms for direct substrate modification through a stencil mask, a parallel exposure strategy with potentially much larger chip size and higher throughput than can be achieved by an HIM. Our goal is a source with the combination of virtual source size, angular flux density, and beam diameter required for system-level studies of these intriguing technologies.

We follow a classical approach^7,8 where He⁺ ions are extracted from a plasma source and neutralized by charge-transfer scattering in a differentially pumped, high pressure cell. A downstream electrostatic deflector removes unscattered ions from the mixed beam emanating from the cell. The absence of space-charge repulsion in the neutral beam implies that the critical lithographic parameters, penumbral blur, flux density, and field size, depend only on the source—its diameter, angular flux density, and divergence angle-irrespective of the atomic particle flux through the system. It is, therefore, straightforward to extract the source parameters at any point along the beamline and to use that data to explore the design space of optimized systems for the applications discussed above.

Figure 1 illustrates the source concept. He⁺ ions are extracted from a decapole multicusp ion source,⁹ focused by two-stage accelerating optics, and neutralized by charge-transfer scattering¹⁰ in a high pressure, He-filled cell. An electrostatic deflector removes unscattered ions from the beam. The ions are accelerated to 3.1 keV in the first stage and to the final energy (typically 10–50 keV) in the second. Space charge, acting as a negative lens,¹¹ prevents the ion beam from crossingover; instead, it converges to a waist of minimum cross section¹² and diverges downstream. The first stage uses a repurposed lens from a Coates and Welter Cwikscan microscope; the second uses a three-element lens from a Varian ion implanter. The intermediate energy was chosen empirically to optimize ion beam uniformity downstream of the beam waist.

Since charge-transfer scattering angles are extremely small (<0.2 mrad at 50 keV), atoms leave the point of neutralization on straight line trajectories that are essentially tangent to the, possibly curved, trajectories of the parent ions. These trajectories crossover, becoming the virtual source of the atom beam for downstream applications. The cell is positioned in the converging ion beam just downstream of the final electrode of the second stage lens. This minimizes space charge in the cell, which would cause the ion trajectories to curve, blurring the virtual atom source.

The charge-transfer cell is a 2.54 cm long cylinder with 3 mm circular differential pumping apertures. The flow of He gas to the cell is controlled by a needle valve. The pressure of the source chamber, measured by an ion gauge, was converted to cell pressure using the conductance of the apertures and the speed (1200 l/s) of the pumps.

The Faraday cup provides a near-absolute determination of flux density for singly charged ions. In lithography, they are used to determine the relationship between the thickness of resist remaining (TRR), after exposure and development, on ion fluence. Typically, this information is displayed as a Hurter and Driffield (H&D) curve,¹³ normalized TRR versus log₁₀(fluence), where fluence is the integrated flux density over the duration of the exposure. This curve provides the important resist parameters including critical dose and contrast.

Faraday cups have also been used to monitor neutral atom flux density.¹⁴ In the experiments reported here, atoms pass through an aperture in a grounded enclosure, traverse a cylindrical electrode and impinge on a collector plate.¹⁵ The collector is connected to the input, a virtual ground, of a Keithley 414A picoammeter. For He⁺ ions, the cylindrical electrode is negatively biased (−90 V) to return secondary electrons to the collector. For atoms, it is positively biased (+90 V) to attract secondary electrons. Then, collector current I_c is equal to the secondary electron current generated by atom impact on the collector; I_c = eAγf, where e, A, γ, and f, are, respectively, the elementary charge, the area of the cup aperture, the secondary electron yield of the collector, and the atom flux density. To quantify the atom flux requires an accurate determination of γ. While γ is well known for clean metals, its value for contaminated surfaces can be highly variable. The collector of the present cup, which has not been cleaned in over 25 years, is coated by a brown, carbonaceous film due to ion/atom beam polymerization of residual hydrocarbons in the vacuum systems (2.0–5.0 × 10⁻⁷ Torr base pressure). It is likely that this coating has been converted to a terminal state of maximum graphitic order¹⁶ by the radiation. A method for measuring γ is discussed in the paragraph below.

Resist exposure processes are the same for He⁺ ions and He atoms of the same energy because the charge state of the incident particles converges within a few atomic layers to a value that depends only on the resist material.^17,18 It follows that the fluence F required for a given value of normalized TRR is the same for ions and atoms. For neutral atoms, F = C/eAγ, where C is the secondary electron charge produced by neutral particle fluence F. Let T(log₁₀(F)) be normalized TRR as a function of log₁₀(F), the function represented by the H&D curve. For neutral particles, T(log₁₀(F)) = T(log₁₀(C/eAγ)) = T(log₁₀(C/eA) − log₁₀(γ)). Thus, the normalized TRR for neutral particles is T(log₁₀(C/eA)), shifted by log₁₀(γ).

Figure 2 shows the H&D curve for 50 keV He ions and normalized TRR versus log₁₀(C/eA) for 50 keV He atoms for a plasma deposited resist.¹⁹ Best alignment occurs for γ = 0.99. The curves for γ = 0.94 and 1.04, which bound both the ion and best fit atom curves, illustrate the sensitivity of the atom curves to the value of γ. Lines between the data points are meant only to guide the eye. Data for atom fluences below 1.5 × 10¹⁴ particles/cm² have been omitted due to systematic errors related to develop swelling of partially cross-linked films. We assumed for simplicity that γ is unity for the measurements of atom flux density and fluence reported in this paper.

The flux density f of the ion/atom beam was measured using a scanning Faraday cup, with a 3 × 0.100 mm² aperture, 7.0 m from the atom beam crossover. The angular flux density is given by $fΩ=f/Ω$⁠, where $Ω$ is the solid angle subtended by the cup aperture when viewed from the source.

The size and position of the crossover were determined by a pair of knife-edge experiments.²⁰ Briefly, a Faraday cup is first used to profile the penumbral shadow of a razor blade at position x, assumed to be downstream from the virtual source. A cumulative Gaussian distribution is fitted to the profile and the standard deviation $σp$ extracted. The diameter $2σv$ of the atom source is then given by $2σv=((x−x0)/g)2σp,$ where $x0$ is the unknown position of the source and g is the distance from the knife-edge to the Faraday cup. A second measurement, at a different knife-edge position, enables $2σv$ and $x0$ to be determined.

The source size versus pressure data were smoothed for the calculation of brightness using a least square fit to a first order response function [$σv=a⋅(1−e−b⋅P)+c$⁠, where a, b, c are the fitting parameters]. The standard deviations²¹ of these fits are 0.04 and 0.1 μm at 25 and 50 keV, respectively. The smoothed source size and the raw data for angular flux density were used to derive the brightness $β=fΩ/2πσv2$⁠.²² 

Figure 3 shows angular flux density versus Faraday cup position for 25 and 50 keV ion and atom beams. The charge-transfer cell pressure was 13 mTorr. The beam is wider for ions than for atoms due to space-charge repulsion near the beam waist. The charge-transfer efficiency, the ratio of total atom flux to total ion flux, is 85% at 25 keV and 65% at 50 keV, probably reflecting the difference in the total charge-transfer cross sections at these energies. The width of the uniform region in the center of the atom profiles is ∼2.5 cm for both energies; it is ∼2.7 cm, full-width-at-half-maximum. The angular flux density, source size $(2σ)$⁠, and brightness at 50 keV are 8.7 × 10¹⁷ particles/s sr, 124 μm, and 3.75 × 10²¹ particles/s sr cm², respectively. At 25 keV, the angular flux density and source size are 3.7 × 10¹⁷ particles/s sr and 88 μm, respectively.

Figure 4 shows (a) source size (2σ_v), (b) angular flux density f_Ω at beam center, and (c) brightness as functions of cell pressure P for 25 and 50 keV helium atom beams. Angular flux density increases with increasing pressure due to the increased neutralization probability within the cell; it saturates at high cup pressure due to the establishment of a steady state between the neutralization and reionization processes. The increase in source size with increasing pressure is probably due to direct scattering of He-on-He;¹⁰ the reason for the saturation remains unclear. Note that brightness, which is included for completeness, is not proportional to the beam energy, as would be expected for ion beams.

We assume for this model a 50 keV beam to minimize lateral scattering in the near-surface region of the substrate and a charge-transfer cell pressure of 13 mTorr to maximize charge-transfer efficiency. These are the same conditions used to obtain the data discussed in Sec. III A (beam profiles). Accordingly, penumbral blur $2σp$⁠, flux density f, and field size S for a short system with source-to-mask distance L and mask-to-substrate gap g are given, respectively, as follows:$2σp=(g/L)⋅2σv,$ $f=fΩ/L2,$ $S=S7⋅(L/L7)$⁠, where $2σv=124μm$⁠, $fΩ=8.7×1017particles/ssr$⁠, $S7=2.5cm$⁠, and $L7=7m$⁠. We choose a gap $g=5μm$ to minimize penumbral blur and limit diffraction. Though small, this gap can be easily maintained by clamping mica spacers between the mask and substrate—this is also an easy way to eliminate relative motion during exposure. Diffraction, which could become important in the low nanometer range of feature sizes with this gap, is poorly understood in this setting and will not be included in the model. Finally, exposure time $t=F/f$⁠, where F is the fluence required to modify the substrate.^1,2 Figure 5 shows flux density and penumbral blur as functions of source-to-mask distance L. The scale at the top of the figure shows field size corresponding to the length scale below the figure. This density is the lower bound of the range $1016−1017particles/cm2$⁠, discussed in the introduction for direct substrate modification. Under these conditions, L = 2.8 m yields penumbral blur, exposure time, and field size of 0.2 nm, 14 min, and 1 cm, respectively. The other end of the range, $1017particles/cm2$⁠, L = 1.4 m, yields corresponding values of 0.5 nm, 35 min, and 0.5 cm.

We have designed and characterized a 50 keV source of energetic He atoms for proximity lithography and explored its applicability to nanosystem prototyping based on resistless, direct substrate modification. We find that it can access the range of exposure densities up to 10¹⁷ particles/cm² with penumbral blur below 0.5 nm, field diameters above 0.5 cm, and exposure times below 35 min; for 10¹⁶ particles/cm², the corresponding values are 0.2 nm, 1 cm, and 14 min, respectively. While far from VLSI standards, throughput seems large enough to enable system-level studies of high-temperature superconducting and nanomagnetic device technologies. This paper addresses only the source issue for nanosystem prototyping by direct substrate modification. The fabrication of masks with the required resolution, line-edge roughness, and pattern density, as well as the potential impact of diffraction remains as a serious open issue. There is, however, a substantial body of knowledge on mask fabrication for integrated circuit manufacturing²³ that can serve as a starting point for nanoscale mask development.

The research was supported by the National Science Foundation (NSF) under Award Nos. DMI-0521523 and ECS-0404308, the Texas Center for Superconductivity at the University of Houston, and the Cullen Foundation. The authors are grateful to Hongjie Guo for his early contributions to this project and to Michael R. Young for many helpful discussions and outstanding machine shop support over many years. The authors would also like to thank John Notte and Ray Hill of Carl Zeiss Microscopy for useful discussions of field size in the helium ion microscope and Horst Rogalla (University of Colorado, Boulder) for sharing his insight into the lithography requirements of nanosystem prototyping.