Entwicklung einer verbesserten sondenstromstabilisierung für rasterelektronenmikroskope development of an improved probe current stabilizing unit for SEMs

Abstract

AbstractA probe current stabilizing unit is important in quantitative XRMA with long counting times. The uncorrected long‐term drift in a Cambridge Mk II A SEM is of the order of 40% during the first 3 h after switch‐on and is caused by an increase in the emission current by more than 20%, probably due to thermal expansion of the long metal tube containing the filament mounting which reduces the distance to the Wehnelt electrode, and a decrease in probe current due to a horizontal shift of the filament tip and a misalignment of the electron beam. Summarizing the literature the following stabilizing methods can be determined: A motor‐driven readjustment of the filament (Balk et al 1975), control of the Wehnelt bias (Cleaver and Smith 1972, 1973), regulation of the filament current (Chapman 1972, Close and Yarwood 1972, Wellenstein and Ensman 1973, Reed 1968, Nakagawa and Yanaka 1975, Baines et al 1975). Our first method of stabilizing the total emission current (Arndt et al 1978) is working well, but does not solve the problem of constant probe current. In the present experiments, we insulated the three spray apertures (Fig. 1) and recorded the emission current and the specimen current during the first 3 h (Fig. 2). The increasing current at aperture 1 confirms the effect (a) mentioned above. The approximately equal decrease in the current at aperture 3 and at the specimen, initiated experiments aimed at stabilizing the probe current using the current from spray aperture 3. However, this was not an appropriate method because overcompensation effects were observed. Therefore, we finally insulated the whole final aperture holder containing the three limiting apertures below the final lens. The voltage drop on a resistor of a few MΩ is amplified by an FET, and the difference voltage U compared to the compensation voltage is used to make a feed‐back loop to the power supply unit of condenser lens 1. Fig. 3 shows the dependence of the probe current on the current IC1 of this lens. The normal operating point is at IC1 = 0.55 A. We limit the current of this lens to between values of IC1 = 0.4–0.7A to avoid the cut‐off region below 0.4A. The compensation voltage can be changed by a potentiometer “zero” (Fig. 12) to calibrate IC1 to 0.55A with a current meter on the panel, because the electron emission of the filament does not only change with time but also from filament to filament. Fig. 4 shows the variations in the feedback voltage and IC1 with a low and high sensitivity of the feed‐back loop. Fig. 5 shows the dependence of the probe current on IC2‐ Increasing values of IC2 result in an equivalent decrease in IC1. In order to change the probe current using condenser lens 2 a compensation voltage is generated by a logarithmic amplification of a voltage drop at a resistor in series with the lens current meter. Fig. 6 shows the difference between the actual voltage obtained by this method and the voltage required for ideal compensation. This difference causes a small shift of IC1 if IC2 is changed (Fig. 7). The time constant was of the order of 4s to avoid oscillations, because the magnetic lenses are slow to react to fast current changes.The result obtained with the stabilizing unit is presented in Fig. 8. The preexisting decrease of 47% in 3 h was reduced to approximately 1.6% (for comparison see Fig. 2). Figs. 9 and 10 demonstrate how the stabilizing unit regulates the current using high and low sensitivity when the filament current is decreased from 3.0 to 2.1 A. The decrease in emission current is actually compensated for over a broad range. The overall efficiency of the stabilizing unit can be seen from Fig. 11.No defocusing could be seen at low and intermediate magnifications but a small shift of 1 μm was observed at M = 20000 × when the full range of stabilizing was used. It normally takes several hours to reach the limits of the stabilizing range and this unwanted shift can therefore be ignored.