On the dose-rate threshold of beam damage in TEM

Abstract

Electron beam damage in a CaF₂–Al₂O₃–SiO₂ glass is investigated using time-dependent Ca L₂₃ and O K-edge electron energy-loss spectroscopy. It appears that there is a threshold dose rate, below which the damage involving the formation of O defects may not be detected, at any total dose. This suggests that the threshold phenomenon of dose rate may result from the competition of damage and recovery processes. The accumulation of damage can only occur when the damage rate is higher than the recovery rate. For surface sputtering process in TEM, the recovery rate is negligible. Therefore, there is no threshold dose rate for surface sputtering.

Introduction

Radiation damage remains an important obstacle to extend applications of (scanning) transmission electron microscopy (TEM/STEM). Aberration correction allows the STEM objective aperture (condenser aperture in TEM nanodiffraction mode) to be enlarged so that the electron probe on the specimen may have a very high current density, e.g.>10⁶ A/cm² [1]. This value is about 10⁴–10⁵ times larger than the current density used in forming conventional HREM images, and about 10⁶–10⁷ larger than that used for bright-field diffraction contrast imaging. Can materials survive under these conditions?

The various phenomena associated with electron-beam damage in TEM/STEM have been extensively studied in recent decades [for a review, see [2], [3]]. In brief, electron-beam damage in specimens is mainly caused by the following three mechanisms. One is due to knock-on interaction through elastic scattering, in which the incident electrons transfer kinetic energy and momentum to atoms. If the kinetic energy acquired by an atom is higher than its displacement threshold energy (E_d) or surface binding energy (E_s), the atom may be displaced from its site to an interstitial or vacancy, forming a Frenkel pair in the bulk, or sputtered away from surfaces into vacuum. In TEM/STEM, the surface sputtering process usually dominates the former one, because E_s is usually much smaller than E_d. Next is radiolysis, which is due to ionization process through inelastic scattering. Some of the criteria required for radiolysis in TEM/STEM are: the excitation needs to be localized for a time long enough for the atom to respond mechanically, and the energy acquired by the excited atom must be convertible into momentum, resulting in atomic displacements [2]. Therefore, beam damage due to radiolysis process may occur in electronically insulating materials. It is suggested that radiolysis is responsible for the formation of F- and H-centers in alkali halides and for amorphization of crystalline SiO₂ and silicates [4], [5]. The third important mechanism is electrostatic charging of materials induced by the incident electron beam. Unlike SEM, charging in TEM/STEM is mainly caused by the ejection of secondary and Auger electrons into vacuum [6]. At high current density, the charge balance cannot be restored quickly enough by the environment, such as a Cu specimen supporter, for electronically insulating materials, and therefore a positive surface potential develops in the illuminated area. According to Cazaux [6], the estimated potential can be as high as 76 eV for a typical STEM probe with a diameter of 1 nm and current of 0.4 nA (i.e. an electron current density ∼0.4 nA/nm²), thus the maximum radial component of electric field at the edge of probe can be higher than 10¹⁰ V/m. This value is much larger than the breakdown voltages of most dielectric materials. Therefore, the positive potential induced by the incident electron may cause a lateral migration of cations and anions, drawing anions into the irradiated area and expelling cations [7], [8]. As for radiolysis, electrostatic charging only occurs in insulating materials. In many cases, in fact, the last two mechanisms cannot be distinguished, especially in the case of a highly intense and focused electron probe.

To avoid electron-beam damage, or to utilize the electron probe for direct-write lithography in TEM/STEM, it is crucially important to understand the damage thresholds of various mechanisms. The energy threshold for knock-on damage is well understood [9], [10]. It can be calculated reasonably accurately as long as E_d or E_s is known [11]. The energy threshold effect in ionization (radiolysis) damage has also been studied. Although there is still debate on whether the damage is dominated by the ionizations of inner shells [12], [13], [14] (which dump large amounts of energy, but less frequently) or valence exciton excitations [4] (which dump smaller amounts but more frequently), the energy threshold in radiolysis damage is generally unimportant since the kinetic energy of incident electrons in TEM/STEM (e.g.>100 keV) is much higher than these thresholds.

Radiation damage is thought to depend on the energy absorbed by the target and its mass. The measure of the amount of radiation, the Gray, is thus defined as absorption of one joule of ionizing radiation by one kilogram of matter (1 Gy=1 J/kg). In TEM/STEM, however, almost all the incident electrons pass through thin specimen. (Only a negligible portion of the electrons can be scattered laterally, and eventually absorbed by specimen after multiple scattering.) The energy deposited in specimens through inelastic scattering is only a small portion of total energy carried by the incident beam. Therefore, it is more convenient to use the number of incident electrons during an exposure (the fluence in C/cm² or e/nm²), as the “electron dose”, to represent the strength of irradiation in TEM/STEM. In the literature, this “dose” is defined as the product of electron current density (dose rate) and illumination (or exposure) time, and we will use this definition of dose in this paper. Generally, it has been considered that there is a “dose threshold”, also known as the “critical dose” [15] or “characteristic dose” [16], for each beam-sensitive material, below which beam damage is negligible. The well-known low-dose technique commonly used for biological materials is based on this idea [17]. For a given total dose there are two ways to achieve a low-dose condition in TEM/STEM: either by lowering the electron current density (dose rate) or by shortening the exposure time. The former is widely used in the low-dose technique, but the disadvantage is its low signal-to-noise ratio (SNR) and poor resolution. The latter may increase the SNR, but the very short acquisition time may induce artifacts due to the finite detector response time [18]. Differences in image quality for the same dose but different exposure times (dose rate effects) are known as “reciprocity failure”.

Although electron-beam damage has usually been measured using total dose in the electron microscopy literature [19], dose rate effects have also been noticed in a few cases [5], [20]. A threshold for dose rate was observed in nanofabrication and hole-drilling studies on oxides and fluorides, which occur only above some threshold current density [21], [22], [23]. Here we report a threshold phenomenon for electron current density (dose rate) observed during electron-beam damage in a silicate glass. It was discovered that electron-beam induced damage in this CaF₂–Al₂O₃–SiO₂ glass was dependent on the beam current density (dose rate). Damage could not be detected if the current density (dose rate) was lower than the threshold value, for any total dose. The basis of this threshold dose rate effect is also discussed.