Principles of Dual-Beam Low-Energy Electron Microscopy

Marian Mankos*,1; Vassil Spasov*; Eric Munro†    * KLA-Tencor, 160 Rio Robles, San Jose, CA 94301, USA
† MEBS Ltd., 14 Cornwall Gardens, London SW7 4AN, UK
1 Current address: Electron Optica, Palo Alto, CA, USA

1 INTRODUCTION

The continuing trend toward smaller features in the semiconductor industry poses a formidable problem for scanning electron beam tools because of their relatively low throughput. The throughput of an electron beam tool is determined by the time required to deliver the electron dose needed to provide a useful signal with sufficient signal-to-noise ratio, so it is proportional to the maximum total electron beam current. However, the large current required to deliver the necessary throughput results in increased electron-electron (e-e) interactions, which blur the image and result in loss of resolution. As the features and pixel sizes become smaller, the beam current must be reduced to maintain the resolution, while the number of pixels to be examined on a wafer increases, resulting in inspection times that exceed practical limits.

One possible approach to circumvent this problem is to replace the serial acquisition process of scanning electron microscopes (SEMs) with a parallel scheme, where all the image pixels of interest on the surface are acquired in parallel on a scintillating screen and further processed on a computer. A low-energy electron microscope (LEEM; Telieps and Bauer, 1985; Tromp and Reuter, 1993) optimized for high throughput (i.e., large beam currents and field sizes) is ideally suited for this application. In a LEEM, a flood beam illuminates the sample with electrons with energies ranging from a few hundred electron volts to near zero electron volts, depending on the substrate bias. The fact that electrons reflect back and travel along the path of the incident beam poses a major challenge in the design of an electron microscope. Since independent control of the illumination and projection is required, the optical axis is split by a magnetic sector field, a nonradially symmetric optical element. This requires a departure from the traditional design with a straight optical axis, resulting in a more complex optical design. We have adopted a design with a straight gun-to-screen axis, which significantly eases column alignment. The four subsystems of the electron-optical column—the magnetic prism array, the illumination, objective, and projective optics—are shown in Figure 1. The illuminating electrons are emitted from the surface of a flat cathode and accelerated to their final beam energy, forming a crossover inside the electron gun. The cathode temperature and extraction field determine the total beam emitted from the gun. The adjacent condensor lenses form a zoom lens that maintains a focused image of the gun crossover at the illumination shape aperture and allows the current illuminating the wafer to be varied and therefore determines the number of electrons/pixel reaching the detector. An additional set of lenses is used to vary the current density at the wafer and therefore determines the size of the illuminated area. The magnetic prism array deflects the electron beam from the illumination optics into the objective optics. Below the magnetic prism array, the electron-optical components of the objective optics are common to the illumination and projection optics. The immersion cathode objective lens decelerates the electrons before they reach the substrate and illuminates the wafer surface with a nearly uniform beam. The electrostatic part of the objective lens creates an electric field of ~5 kV/mm at the substrate surface. In the opposite direction (i.e., upward from the substrate), the objective lens simultaneously forms a magnified image of the substrate. As the electrons reenter the prism array, they are deflected into the projection optics. The magnetic prism array is followed by a diffraction lens, which forms an image of the objective lens back focal plane in the pupil aperture plane and simultaneously forms a magnified image of the wafer in the object plane of the projection zoom optics. The projection zoom section is followed by the final magnifying element of the projection optics, the final projector lens. The electron image formed at the scintillating screen is then viewed by a charge-coupled device camera and further processed on a computer.

When the wafer is biased positively with respect to the electron source, the electrons scatter at or near the surface and either reflect back from the sample, undergoing low-energy electron diffraction, or generate secondary electrons, provided the bias and therefore kinetic energy of the illuminating electrons is large enough (few tens to hundreds of electron volts). When the substrate is biased slightly negative (~1 V) with respect to the electron source, the illuminating electrons are reflected above the surface without hitting the surface. This imaging mode is also known as mirror electron microscopy (MEM). When the substrate is illuminated by a source of ultraviolet (UV) or shorter-wavelength light, photoelectrons are emitted, resulting in the well-known photoelectron emission microscopy mode.

However, when a conventional LEEM instrument is used to image substrates with a variety of insulating materials at the surface, the imbalance of the electron flux results in charging effects that can significantly reduce the imaging quality. In this review, we show how the need for imaging of insulating surfaces at high throughput affects the electron-optical design of a LEEM and present experimental results from several types of substrates used in the semiconductor industry.

2 DUAL-BEAM APPROACH

The dual-beam approach is driven by the difficulties encountered when electron microscopes are used to image insulating surfaces. The imbalance between the arriving and outgoing flux of electrons causes the surface to charge up, resulting in increased blur and distortions. On a homogeneous insulator surface, the charging can be suppressed by operating at a landing energy resulting in a net secondary yield of 1. However, this approach restricts the landing energy and typically does not work when different insulating materials are present on the surface. We have developed a dual-beam approach that mitigates the charging effect when either two electron beams with different landing energies or an electron and photon beam are used for imaging (Adler and Marcus, 2005; Mankos et al., 2007; Veneklasen and Adler, 2003).

The basic principle of dual-beam charge control is shown in Figure 2. When an insulating substrate is illuminated with a single electron beam, the surface charges either negatively (i.e., in mirror mode when the landing energy is very low, and electrons are absorbed) or positively (electron yield >1, landing energy greater than a few hundreds of electron volts). In the case of UV photon illumination, the surface charges positively as electrons are emitted. However, when two beams with opposite charging characteristics—that is, a mirror beam and either a higher-energy electron beam or UV photon beam—are superimposed on the substrate, charging effects can be neutralized.

A more detailed description of this charge balance for the case of two illuminating electron beams is given in Figure 3. The energy spectrum of the illuminating electrons approaching the substrate surface and signal electrons leaving the surface is shown in Figure 3a. Typically the electron energy of illumination has a Maxwell–Boltzmann distribution peaked at 0.25 eV with a full width half maximum of approximately 0.5 eV. The first beam is partially mirrored and its high-energy tail is absorbed, which charges the surface negatively. The second beam, frequently referred to as the charge control beam, strikes the wafer with energies of typically a few hundred electron volts, which results in a total (secondary and backscattered) yield σ (=δ+η) larger than 1 that charges the surface positively. The portion of the mirror beam current Im that is absorbed equals αIm, and the second charge control beam current equals Icc, so the condition for charge equilibrium then can be written as

Im=(σ-1)Icc.

This state of charge control is a dynamic quasi-equilibrium, and the surface is at a potential that is slightly (<1 eV) more negative than the cathode potential (0 V), depending on the fraction of absorbed mirror electrons. This scenario is demonstrated in Figure 3a. When the charge control beam current (or total yield σ) slightly increases to Icc1, the surface begins to charge...