Pattern Generators for Reflective Electron-Beam Lithography (REBL)

Allen M. Carroll1    Taba Research, San Jose, California
1 Corresponding author: email address: allen.carroll@sbcglobal.net

1 Introduction

As is well known, the principal application of electron beam lithography (EBL) in the semiconductor industry is in integrated-circuit (IC) mask fabrication. Relentless advances in circuit nodes bring ever-tightening requirements on pattern fidelity, control and consistency of critical feature dimensions, large-scale accuracy of pattern element placement, and mask-to-mask overlay. These requirements have been successfully met by gradual improvement of e-beam mask-maker technology since the introduction of the first commercial e-beam mask-making tools in the 1970s.

The use of EBL as a mass-production technique has always been inhibited by the relatively low throughput of electron-beam tools. Even so, a novel architectural concept arises from time to time that offers the prospect of applying ideas of massive parallelism to electron-beam systems, and thus to overcome the throughput barrier. Reflective electron-beam lithography (REBL) is such a concept (Hess, Mankos, & Adler, 2010; Petric et al., 2009, 2010, 2011; McCord, Kojima, Petric, Brodie, & Sun, 2010; Freed et al., 2011; Gubiotti et al., 2013).

The inspiration for REBL evolved from an investigation into the use of a low-energy electron microscope (LEEM)–like electron optical system to detect defects in conventionally fabricated masks and integrated circuit wafers. But whereas a defect detection system would typically magnify the beam reflected from a sample (in mask or wafer form) and analyze the reflected and magnified beam, the lithography tool produces an electron pattern and demagnifies it in order to expose circuit features that are a few tens of nanometers in size on a resist-coated wafer. REBL produces the patterned beam by means of a close-packed array of micrometer-scale switchable electron reflectors that locally modulate the illumination beam so that the reflected beam carries a pixilated image of the integrated circuit pattern to be printed. This array of switchable reflectors, called the digital pattern generator (DPG), is the principal subject of this chapter.

In the section “REBL and DPG Basics,” the basic concepts of REBL and of the DPG devices that enable it are reviewed. The next section, “Lenslet Analysis,” examines the behavior of the switchable electron reflectors in more detail. In the section “REBL Patterning Analysis and Column Optics,” brief comments on the characteristics of REBL e-beam columns are provided and the means for directly observing DPG and column performance are discussed. Then the section “Charge Draining” discusses this very important subject. The design of a DPG for printing on a moving wafer is discussed in the section “DPG2 Design and Realization,” and the next section, “Example of Lithographic Results,” shows several sample lithographic results obtained using such a device. The section “Innovations in DPG3” points the way toward higher-performance DPGs, and the section “Summary” presents a brief summary.

Electron-optical constraints and Coulomb interaction of electrons conspire to limit the throughput of a single REBL DPG and projection-optics column to about 1–2 wafers per hour for advanced-node lithographic applications. Thus, the massive parallelism and relatively high beam currents enabled by the reflective patterning device do not completely solve the throughput problem. Today's wafer steppers, the workhorses of mass-production IC lithography, have throughputs in the range of 100–200 wafers per hour. Boosting the 1–2 wafers per hour available from a single-DPG and projection optics system to these levels requires more tiers of parallelism, but exploring how that might be done would move far afield from the subject of this chapter, which is the DPG itself, so this discussion will not go into detail about that.

2 REBL and DPG Basics

The individual electron reflector element of the REBL DPG is a column of axially aligned, ring-shaped electrodes 1–2 μm in diameter (Grella, Freed, & McCord, 2012; Grella et al., 2013). It is typically fabricated by etching a hole in a stack of alternating metallic and dielectric materials (Vereecke et al., 2011). This cylindrical arrangement, called a lenslet, is open at one end for electrons to enter and depart and closed at the other end by the switching electrode. Judging from practical experience, it also seems necessary to coat the interior of the stack with a high-resistance material that can drain charge that may accumulate on the walls. Figure 1 is a cross section of an array of such lenslets. The stack materials in this case are TiN (metal) and SiO2 (dielectric). This particular micrograph is of a fabrication test structure (and does not include a charge-drain coating).

When the lenslet array is illuminated with a low-energy flood electron beam, and with appropriate potentials applied (at most a few tens of volts), the upper electrodes within each lenslet focus the incoming low-energy beamlets, and then the switching electrode applies the modulation. Suitable potentials on the switching electrode can cause incoming electrons to be absorbed, paraxially reflected, or deflected in such a way that they can be intercepted by apertures in the projection optics.

Referring to Figure 1, if the illumination is considered to be coming from above, the electrodes can be named “Top,” “Upper,” “Middle,” “Lower” and “Bottom”; and their corresponding potentials as “VTOP,” “VUPPER,” etc. In an array of such structures, all the “Top” electrodes are fabricated as a common plane, and the same goes for “Upper,” “Middle,” etc. However, to make a device that can support completely general lithography, the switching (“Bottom”) electrodes would be electrically isolated from one another so that lenslets can be turned on or off individually. Note that the structure shown in Figure 1 has all the electrodes as common planes: these lenslets cannot be individually controlled.

The illuminating electron beam is obtained from a 50–100 kV electron gun with a thermionic cathode. To separate the illuminating beam from the reflected beam so that the gun can be outside the path of the reflected beam, a Wien (E × B) filter is introduced into the illumination system. The illumination beam is bent through a small angle in the Wien filter to provide normal illumination at the DPG surface (the underside of the topmost element shown in Figure 2), while the reflected beam follows a straight-through path. The DPG is maintained nearly at cathode potential, so the volume immediately above the Top electrode (i.e., below the DPG in the configuration of Figure 2, as the lenslet openings face downward) is a decelerating space for illumination electrons approaching the DPG, but an accelerating space for the reflected beam. The numerical aperture of the illumination, determined by an aperture in the electron gun, is chosen to match the characteristics of the downstream demagnification optics and to minimize the likelihood of electrons landing on lenslet walls.

For simple types of lithographic patterning, such as line-space arrays for demonstrating resolution, lenslets can be ganged into rows or columns or other predetermined patterns by wiring their switching electrodes together. The wires would be connected directly to the contact pads of the DPG chip; and switching is then controlled by external circuits. Such devices are simple and robust, and because switching is controlled externally, very high switching potentials can be applied. Power dissipation in such devices is tiny, and they can be mounted completely in a vacuum fairly easily,...