Ion Projection Lithography Research Articles

Ion Projection Lithography for Nano Patterning

AbstractAs a result of continuous improvement of the resist process, the experimental ion projector in the Fraunhofer Institute in Berlin (manufactured by Ion Microfabrication Systems, IMS, Vienna) has been able to print 75 nm lines and spaces into 180 nm thick standard DUV resist UV II HS without pattern collapse. A new wafer flow process for more reliable open stencil mask making was developed by IMS –Chips, Stuttgart (Germany), based on SOT wafers. Resistless direct surface modification by He and Xe ions has been tested on metallic and magnetic films in the Berlin projector. This method opens up a new possibility for the production of patterned media for future magnetic storage disks.

Sub-0.1 µm Patterning Characteristics of Inorganic Thin Films by Focused-Ion-Beam Lithography

Sub-0.1 µm patterning characteristics of low-pressure chemical-vapor-deposited Si3N4 and normally(0°) and obliquely(60° and 80°) deposited Se75Ge25 inorganic thin films have been investigated using low-energy Ga+ focused-ion-beam (FIB) lithography. These thin films act as a negative type resist for CF4-reactive-ion-etching (RIE) development. In the case of 0°-deposited Se75Ge25 and Si3N4 resists exposed by 30 keV Ga+ FIB and developed by CF4 RIE under a gas pressure of 150–200 mTorr, the image-contrasts (γ) are reasonably high, about 4.4 and 3.5, respectively. Their sensitivities (S), nevertheless, are relatively low in comparison with that of conventional polymer resist. Therefore, this problem should be solved, especially in order to employ these films as resists of ion-projection lithography. In the case of an 80°-deposited Se75Ge25 resist, columnar structures with an a ngle of approximately 65° are formed which disappear after annealing at the glass-transition temperature. The γ of this resist is evaluated to be about 4.0, similar to that of 0°-deposited Se75Ge25, but the S is enhanced to be about 7.0×1014 ions/cm2, which corresponds to below half that of 0°-deposited Se75Ge25. With an exposure of 30 keV FIB with a beam diameter of 0.1 µm and just above the threshold dose, negative-type fine patterns with linewidths of about 0.06–0.09 µm are fabricated successfully.

Ion projection lithography: Status of the MEDEA project and United States/European cooperation

Structure and targets of the European MEDEA project on ion projection lithography as well as related U.S./European cooperation are explained. By assuming 10 μm virtual source size and 1 eV (full width half maximum) energy spread calculations for a multielectrode electrostatic ion–optical system (1.25 m between ion source and stencil mask, ≈1.8 m between mask and wafer) we realize the possibility of 100 nm resolution (line and space) over an exposure field of 22×22 mm2 even when using the MONTEC model for calculating the stochastic blur and when running 3.3 μA He+ ion beam current through the ion–optical column, thus more than twice exceeding target specifications. Thus, for 100 nm resolution and 50% pattern density the raw throughput is ≈12 cm2/s corresponding to >75 WPH (pattern within 80% of 300 mm wafer area).

P-n junction-based wafer flow process for stencil mask fabrication

The development of stencil masks is considered to be critical to the success of the new ion projection lithography technology. We present here a p-n junction wafer flow process where all fabrication steps are realized on a bulk Si wafer except the final trench etching through the 2–4-μm-thick Si membrane. Stencil masks were produced in a conventional complementary metal-oxide-semiconductor 150 mm wafer line, using an e-beam direct writing tool for patterning. The resist patterns were transferred by standard reactive ion etching (RIE) into a stress-controlled SiON hard mask layer. Subsequent to depositing an Al metal layer for contact to the n-doped wafer surface, the membrane was realized by a wet chemical etch which implemented well established reverse biased p-n junction etch stop techniques. Then, openings through the Si membrane were etched by RIE or inductively coupled plasma etching. Finally, the remaining hard mask layer was removed in BHF. The realized Si membrane diameter was 120 mm with a stencil pattern field of 60 mm×60 mm. Results from LMS-IPRO placement measurements are in agreement with the simulation of the stencil mask fabrication process using finite element methods.

Trajectory displacement effect in particle projection lithography systems: Modifications to the extended two-particle theory and Monte Carlo simulation technique

The extended two-particle model for statistical interactions in particle beams, developed by Jansen [Coulomb Interactions in Particle Beams (Academic, Boston, 1990)], has been refined to improve the accuracy of the predictions of the trajectory displacement effect in particle beam projection systems. The original theory was developed for probe forming systems, such as electron and ion scanning microscopes and Gaussian or shaped beam lithography systems. Fit functions are used within the theory to express part of the numerical output into explicit analytical prescriptions. These functions were found to become inaccurate for the relatively wide beams typically used in the more recently developed projection type lithography systems. New fit functions are presented which extend the applicability of the theory to the wide beams and doublet configurations used in projection systems. The Monte Carlo program MONTEC, used to verify the results of the analytical theory, has been modified as well to account for the first order space charge magnification effect. This effect could be ignored for the relatively small spots of Gaussian and shaped beam systems, but would yield a significant overestimation of the trajectory displacement effect—assumed to be identical to the remaining blur after refocusing—for the wide images used in projection type of systems. The refined analytical theory and the modified MONTEC program have been used to evaluate the impact of statistical interactions on the performance of the SCALPEL electron projection system and a hypothetical ion projection lithography system, representing a simplified model of the IMS ALG-1000 (He+) system. The analytical predictions are in good agreement with the Monte Carlo results. An estimate of the total system resolution, determined by the combined effect of statistical interactions and geometrical aberrations, indicates that maximum attainable beam current for a 0.18 μm design rule is about 25 μA for the SCALPEL and 0.2–0.3 μA for the ion projection system, leading to an ∼10× higher throughput for the former taking the difference in resist sensitivities for electrons and ions into account.

Economic and technical case for ion projection lithography

For more than a decade, technical arguments have been used to predict the demise of optical lithography. Recently economic arguments are being used, dramatically increasing the likelihood of the predictions coming true. Presently four technologies are competing to replace optical lithography. This article will describe the current state-of-the-art in ion projection lithography, and contrast its advantages and disadvantages with respect to the other nonoptical technologies. An argument will be made that ion projection lithography, uniquely, offers the possibility of reducing the dramatically escalating costs of lithography, resulting in overall lower integrated circuit production costs. This potential for reduced production costs is the prime motivator to consider ion projection lithography as a production lithography candidate.

A review of ion projection lithography

Although optical lithography has been extended to far smaller dimensions than was predicted 15 years ago, there are definite physical barriers to extending it to the minimum dimensions of 70 nm that are projected to be required 15 years from now. Both focused, point electron beams and ion beams have been used to write dimensions in resist well below 20 nm, albeit at speeds far too slow for production lithography. Projection systems, which employ a mask and, in effect, produce a large array of beams, can provide both small minimum dimensions and high throughput. Ions are particularly well suited for this because they suffer little or no scattering in the resist, the linewidth is not a strong function of dose (good process latitude), and the resist sensitivity is relatively independent to resist thickness or ion energy. IMS in Vienna, Austria has built two generations of ion projection lithography systems which have demonstrated many of the features needed for high throughput lithography. In these systems a stencil mask is irradiated with a uniform beam of light ions, H+, H2+, or He+, and the transmitted pattern is demagnified (by 10× to 3×) and focused on a resist covered wafer at energies in the 70–150 keV range. So far, minimum dimensions down to 70 nm line-space pairs have been demonstrated, drift has been eliminated with a “pattern lock” servo system, and field distortion of less than 0.15 μm over 8×8 mm has been measured in agreement with calculations. Based on these achievements a new generation ion lithography machine has been designed which uses 3× demagnification and will expose a 20×20 mm field at 0.12 μm minimum dimensions with less than 10 nm of distortion introduced by the ion optics. Global and stochastic space charge effects have been modeled and, in some cases, measured with the existing machines. Stochastic space charge effects will not cause unacceptable blur in the new design if the total ion current is kept below 3 μA. At this total current the time to expose one chip is still of order 0.5 s so that the calculated throughput is about 70 wafers (200 mm) per hour. Ion sources with low energy spread (∼2 eV) have been developed and will provide uniform illumination of the mask. Stencil mask fabrication on membranes of 2.5-μm-thick silicon has been developed. Distortion of the pattern cut in the membrane due to stress relief has been modeled, and with proper mask design can be kept below 20 nm (6.7 nm on the wafer). According to calculations and measurements mask distortion due to ion beam heating can be reduced to a negligible level if a radiation cooling cylinder is used. As a result of the building and evaluation of the existing machines and the design of the next generation, significant progress has been made in ion projection lithography which we will review in this article. The next step in the development of ion projection lithography is being conducted by the MEDEA program in Europe, which will develop a full field processing tool.

Lifetime enhancement of a multicusp ion source for lithography

The acceleration system designed by Ion Microfabrication System (IMS) for ion projection lithography (IPL) generates a divergent beam of which only a small portion is employed in the process. Computer simulation shows that by optimizing the accelerator design, the source can be operated with lower discharge power, resulting in longer lifetime and lower energy spread. RF induction discharge operation provides long ion source lifetime with energy spread comparable to that of the filament discharge multicusp source when RF modulation in the extraction voltage is eliminated through proper shielding. New antenna designs have been developed which can further extend the lifetime of the RF-driven ion source.

Radio frequency driven multicusp sources (invited)

The radio frequency (rf)-driven multicusp source was originally developed for use in the superconducting super collider injector. The source can routinely provide 30 mA of H− beam at 0.1% duty factor. By adding a minute quantity of cesium to the discharge, H− beam current in excess of 100 mA and e/H∼1 has been achieved. The rf-driven H∼ source is being further developed for 6% duty factor operation to be used in the national spallation neutron source. Application of the rf-driven multicusp source has been extended to radioactive ion beam production, ion projection lithography, and compact neutron tubes.

Ion projection lithography: Next generation technology?

It is generally agreed that optical lithography is the method of choice for production of integrated circuits having minimum dimensions of down to 200 nm. By applying special enhancement techniques and complex resist processes the limit of optical lithography can be shifted down to about 150 nm in production using an exposure wavelength of 193 nm. However, this 193 nm equipment will be a one generation tool, so that new concepts are necessary for exposing structures smaller than 150 nm. The situation for this time (after about 2003) is still unclear. As lithography tools for after the year 2003 are not yet defined, there is an urgent need for preparative work for closing this technology gap. One of the most promising candidates for reaching 100 nm and below in production is ion projection lithography. In this article, the advantages and risks of this technique are compared to other possible candidates like x-ray, e-beam, and extreme ultraviolet. Insoluble production problems are identified. The feasibility of the relevant mask technologies, the availability of resist techniques are throughput and cost competitiveness are discussed.

Stochastic Coulomb interactions in ion projection lithography systems with aberration-broadened crossover

The allowable current in ion beam projection lithography is limited by stochastic Coulomb interactions. Theories for the dependence of these interactions on the parameters of the system, i.e., length, beam energy, print size, crossover diameter, etc. are thus important to guide the design of future machines. All existing theories on blur from stochastic Coulomb interaction assume a homogeneous current distribution in the crossover, where a large part of the blur is produced. However, in actual lithography machines, the crossover may be substantially broadened by the spherical aberration of the ion lenses, thus giving a very inhomogeneous current distribution in the crossover. In most analytical theories, the stochastic blur is independent of the radius of the homogeneously field crossover rc for very small crossovers, and relates to rc as (1/rc)1/2 or (1/rc)1/3 for larger ones. If aberration broadening could do the same, the effect of broadening the crossover from rc=1.5 to 50 μm would decrease the blur by a factor as much as 6, without the negative effect of increasing the curvature-of-field blur. We find physical arguments that such a large factor is impossible. Monte Carlo simulations first of all show that the blur reduction is larger near the edges of the image field than in the center, near the optical axis. For the simulated model system, the reduction on axis can be increased to a factor 2 at very large aberration coefficients.

Analysis of stencil mask distortion in ion projection lithography

Accurate feature placement on the wafer in ion projection lithography requires that the distortion due to stress relief produced by the pattern of holes in the mask be controlled. We have used the finite element method provided by the ANSYS software package to calculate the two dimensional mask distortion due to non-symmetrical patterns of openings in stencil masks and to analyze a method for controlling distortion. The ion projection lithography mask is a circular silicon membrane 120 mm in diameter 2.5 μm thick rigidly held around the perimeter. The central 60 × 60 mm square area will contain the pattern to be printed, demagnified by 3x. We take the intrinsic stress of the membrane to be 10 MPa, the Young's modulus of the membrane to be E = 1.5 × 1011 Pa and the Poisson ratio ν = 0.17. We have explored the x-y distortion produced by various simple geometries. For example, consider an asymmetric grid of square holes which is 22% open on one half and 10% open on the other half. High distortion in this case will occur along the diameter, between the two halves. The boundary between the two halves (along the diameter) is bowed by 182 nm.A method for reducing the distortion is to cut a pair perforation rings along the perimeter. The perforation rings will have the effect that the central part of the membrane will have the stress relieved and it will be held in effect by “springs” at a constant, low stress level. We have calculated that the bow of the diameter can be reduced to 18 nm by a suitable geometry of the perforations. The perforation ring is an effective means of reducing distortion to well below an acceptable level. The uniform shrinkage of the pattern due to stress relief is simply a small change in magnification which is automatically corrected by the beam lock in the ion optical column.

Study of energy broadening of high-brightness ion beams from a surface plasma Penning source and its relevance in ion projection lithography

The energy spread ΔE (Full Width at Half Maximum of the energy distribution) of negative hydrogen ion (H−) beams has been measured for angular beam intensities (IΩ) over a range of about 40 to 100 mA/sr. An IΩ23 dependence of the energy spread is observed. This trend supports the mechanism of energy broadening due to potential energy relaxation. The observed energy spread of ≲ 3 eV at IΩ of about 40 mA/sr projects a favorable scenario toward the demand of a future production-line ion projection lithography system. The maximum value of IΩΔE2 = 4 mA/sr eV2 suggests the possibility of H− beams for ion microscopy and surface analysis applications also. The H− ion source is currently operated in pulsed mode (maximum duty of 1%-- 1 ms, 10 Hz); cw operation can be achieved by improving the cooling system of the electrodes.

Axial energy spread measurements of an accelerated positive ion beam

A multicusp ion source has been designed for use in ion projection lithography. Longitudinal energy spreads of the extracted positive hydrogen ion beam have been studied using a retarding field energy analyzer. It has been found that the filament-discharge multicusp ion source can deliver a beam with an energy spread less than 3 eV which is required for the ALG-1000 machine. The multicusp ion source can also deliver the current required for the application.

High-brightness ion source for ion projection lithography

A Penning-type surface plasma source has been developed with the goal to achieve the beam requirements for ion projection lithography (IPL). The present source, with simple, forced air cooling runs up to a duty factor of 1% (pulse length of 1 ms and repetition rate of 10 Hz); a cw Penning source is being planned for IPL. H− emission current density of more than 2 A/cm2 has been obtained with negligibly low noise. The perpendicular temperature is about 0.6 eV for emission current density of about 1 A/cm2—this yields normalized beam brightness of more than 1012 A/(mrad)2. The energy spread full width at half-maximum (FWHM) for a 5 kV beam with angular current density of 33 mA/sr is about 3.3 eV (close to instrumental resolution). The energy spread increases with beam intensity. In preliminary experiments on positive ion extraction, noiseless H+ beams with emission current density of about 150 mA/cm2 have been achieved.

Development of ion sources for ion projection lithography

Multicusp ion sources are capable of generating ion beams with low axial energy spread as required by the ion projection lithography (IPL). Longitudinal ion energy spread has been studied in two different types of plasma discharge: the filament discharge ion source characterized by its low axial energy spread, and the rf-driven ion source characterized by its long source lifetime. For He+ ions, longitudinal ion energy spreads of 1–2 eV were measured for a filament discharge multicusp ion source which is within the IPL source requirements. Ion beams with larger axial energy spread (∼7 eV) were observed in the rf-driven source. A double-chamber ion source has been designed which combines the advantages of low axial energy spread of the filament discharge ion source with the long lifetime of the rf-driven source. The energy spread of the double chamber source is approximately 2 eV.

Surface plasma source to generate high-brightness H− beams for ion projection lithographya)

A surface plasma source with Penning-type electrodes is developed to generate H− beams for ion projection lithography applications. The source presently runs in pulsed mode with a pulse width of about 1 ms and repetition rate of 10 Hz. The discharge and the extracted H− beam are maintained in stable, noiseless condition; gas pressure plays a critical role here. The maximum H− beam current density at the emission surface is about 1.7 A/cm2. Preliminary measurements suggest that the normalized brightness of the core beam is ∼1.7×1013 A/(m rad)2 and its perpendicular temperature is about 0.2 eV.

Surface plasma source to generate high-brightness H− beams for ion projection lithography (abstract)a)

A surface plasma source with Penning-type electrodes is developed to generate H− beams for ion projection lithography applications. The source presently runs in pulsed mode with a pulse width of about 1 ms and repetition rate of 10 Hz. The discharge and the extracted H− beam are maintained in stable, noiseless condition; gas pressure plays a critical role here. The maximum H− beam current density at the emission surface is about 1.7 A/cm2. Preliminary measurements suggest that the normalized brightness of the core beam is ∼1.7×1013 A/(m rad)2 and its perpendicular temperature is about 0.2 eV.

Global and stochastic space-charge effects in ion beam lithography

Ion projection lithography (IPL) is a candidate technology for meeting expected circuit design rules (less than 0.18 μm) for future generations of semiconductor devices. The Advanced Lithography Group (ALG), a consortium of industrial and U. S. government laboratories and universities, is developing IPL technology with funding from the Advanced Research Projects Agency (ARPA) and the state of Maryland. A prototype IPL device, the ALG-1000, is being developed to demonstrate the capability of IPL to meet future requirements for pattern overlay. To realize IPL technology requires control of space charge effects in the ion optical column. Due to the length of the IPL system (several meters), the precision of the calculation to predict distortion at the wafer plane becomes difficult to perform. Including the effects of the space charge is even more difficult. Both global and stochastic space-charge phenomena occur. This article presents a system of computer models that allows simulations of both global and stochastic space-charge effects. In particular, the models will be used on the ion beam projector in the ALG-1000 device. The calculations are carried out using a self-consistent equilibrium ray tracing code, where the applied fields are calculated from the actual lens column geometry. For global space charge, the model also includes optimization of the lens electrode voltages to minimize pattern distortion at the wafer plane.

Control of temperature gradients and distortion of ion projection lithography masks

In ion projection lithography a stencil mask is back ‘‘illuminated’’ by a broad beam of ions, and its demagnified image (∼3×) is projected onto a resist covered substrate. The stencil mask that will be used in the next generation systems consists of a 120-mm-diam 2.5-μm-thick single crystal silicon membrane etched in a 150-mm-diam wafer whose edge acts as part of a supporting frame. The main sources of distortion are thought to be stress relief due to the cutting of holes and the thermal expansion due to mask heating by the incident ions. In this paper we calculate the temperature distribution in the mask for the expected heat input by the beam, the resulting mask distortion, and the effect of optimized radiation cooling. The central 100-mm-diam area of an unpatterned membrane is assumed to be irradiated with an energy input of 3.3 mW/cm. The edge of the membrane at the frame is assumed to be at room temperature and held rigidly. Heat is lost by radiation and by conduction. An effective method of reducing temperature gradient in the membrane is to place a cooled cylinder above the mask surrounding the ion beam. The expected temperature profiles for various sample emissivities as well as the corresponding distortion (radial displacement of the mask membrane) are computed with and without the cooled cylinder. These computations are done for various geometries and cylinder temperatures. For example, in a membrane of emissivity 0.55 with a cylinder, 6.5 cm high 13 cm in diameter, placed 3.2 cm above the mask and cooled to 273 K, the temperature is uniform around 300 K to better than ±1.5 K and the expected mask distortion is then less than ±10 nm. Since the ion beam will be incident on the mask continuously, the cooling cylinder will be effective in reducing thermal distortion to well below the permitted limit.