11. DEVICE

11.1 Automated Scanning Low-Energy Electron Probe

The initial electron gun structure (NBS Spec. Publ. 400-12, p. 28) for the ASLEEP system was completed and mounted in the vacuum system for testing and alignment. The gun was placed in the circuit shown in figure 39 for beam current tests. The target was an aluminum grid comprising 60 um wide lines on 600 centers over silicon dioxide (NBS Spec. Publ. 400-17, p. 46) mounted on the specimen zanipulator. The 6-V battery was inserted into the circuit during initial testing to overcome patch effects at the target. The

focus coil was adjusted to a magnetic induction of 70 gauss (7 mT) and the target current maximized by adjusting the various potentials. The maximum current collected by the target was 1.7 nA, obtained with the heater current at 9.3 A, grid 1 at 30 V and grid 3 at 500 v.

The first electron gun developed a short between the high voltage and the cathode necessitating a modification of the design. The glass insulators used to shield the high voltage leads were fractured where they crossed the low voltage elements of the gun. Notching the grid and cathode holders to allow clearance for stronger tubular ceramic insulators corrected this problem. The original gun used three sapphire rods to hold the gun elements in place. The rods were notched and metal tabs were inserted in the notches and spotwelded to the gun elements. This scheme made alteration and repair of the gun difficult. The elements of the gun are now held in their proper relationship by hollow ceramic spacers through which a bolt can be run to hold the assembly under compression. The modified gun has been tried and found to be more successful.

The computer control system rasters the electron beam over the test grid. If the electrometer in the beam current test circuit (fig. 39) is replaced by a low noise amplifier, an image may be formed on a cathode ray tube (CRT) by using the amplified target current signal to control the CRT beam intensity. Figure 40 is a picture of one such image taken of the aluminum grid during preliminary system checkout. The narrow line at about one half the grid spacing near the edge of the picture is a scratch in the silicon dioxide surface. Figure 41 is an image of an array of MOS capacitors made of 30-mil (0.75-mm) diameter dots of aluminum on sili

con dioxide 100 nm thick. The capacitors were scanned with a beam of fixed energy (approximately 3.5 eV). The four white capacitor dots in the center of the picture are shorted to the substrate due to defects in the oxide. Similarly, two rows of shorted capacitors are located where the wafer was scribed and broken.

Additional software has been generated for presenting surface potential mapping information. The electron beam is scanned in the x-y plane, the retarding voltage is ramped to a preselected threshold current level, the surface potential is scanned over any selected area of the target, and the results are presented in a three dimensional display all under computer program control. In the presentation, the y-axis is somewhat elongated and the z-axis deflection of the scan line is proportional to the relative surface potential. Figure 42 shows a surface potential scan of an array of capacitors on silicon dioxide. Variations in height of the profiles are a direct measure of the dielectric uniformity assuming the metal work function to be constant. A high dot indicates a high capacitance. This result qualitatively demonstrates the ASLEEP as a non-destructive computerized surface diagnostic technique.

where is expressed in micrograms per square RG centimeter* and ER is expressed in kiloelectron volts, which was valid over the range 5 keV < Eg ≤ 25 keV. Second, he showed that the shape

of the depth-dose curve is practically invariant if the penetration distance is normalized to R and the energy is normalized to Eğ Everhart and Hoff [68] extended these general conclusions to solids and obtained a generalized depth-dose curve for solid materials by taking the steady-state electron beam induced current through the insulating layer of a metal-oxide-semiconductor structure as a measure of the energy dissipation in that layer. They found for structures of aluminum, silicon dioxide, and silicon the projected range,

ber in the range 10<Z< 15, the energy dissipation is given by

where I is the SEM beam current, is registered by the ammeter in figure 44b. The quantity N is therefore the current gain; N is plotted as a function of electron beam energy for several thicknesses of absorber above the silicon in figure 45 and as a function of absorber thickness for several beam energies in figure 46. The nomograph in figure 47 can be used to convert aluminum, silicon dioxide, or aluminum plus silicon dioxide thickness in micrometers to absorber thickness in micrograms per square centimeter.

As an example of the use of these graphs to make a quick estimate of the current gain to

The measure of length used here is the product of material density and thickness.

be expected when performing an EBIC examination, consider a 10-keV electron beam incident on a silicon structure with a 500-nm thick surface film of silicon dioxide. From figure 47, the total absorber thickness is 115 ug/cm2 and from figure 46 the current gain is 800.

An experimental measurement was made for comparison with the results of this calculation. The device chosen for study was a p-n diode similar to the one shown in figure 44. The diode was fabricated in a wafer of phosphorus-doped silicon with room temperature resistivity in the range 5 to 10 cm and a <111> surface orientation. The p-region was formed by diffusing boron to a depth of approximately 1.4 um; the diffusion profile was gaussian and the sheet resistance was about 180 /0. The thickness of the oxide layer above the junction was approximately 330 nm. The I-V characteristics of the diode were unchanged when measured after performing the EBIC measurements.

B

The diode was exposed to both 100 and 500 pA SEM electron beams at energies from 2.5 to 30 keV. Although the active region of the diode was 1.5 mm (0.06 in.) in diameter only a small portion of the central part (approximately 10-3 cm2) was scanned in order to eliminate edge effects. The electron beam induced current was measured with an ammeter capable of measuring currents from 1 nA to 100 μA connected across the p-n junction. The results are shown in figure 48 in which Ig/Ig (or N) is plotted as a function of beam energy. The solid line represents the calculation for an absorber thickness of 330 nm. It is apparent that there is reasonable agreement between the experimental and theoretical results. Discrepancies at higher beam energies may arise from failure to collect all the carriers produced while the disagreement below 5 keV may be attributed to the injection of energy (low energy x-rays, electron straggling, etc.) or charge from the absorber. Plotting the induced current normalized by the beam current produces a curve that is essentially independent of beam current, as expected. Similar results were cbtained for other similar diodes.

The simple calculation presented here seems to give an adequate first order prediction of the electron beam induced currents for simple device structures. However, the validity of the assumptions used should be assessed for each specific application.

(K. F. Galloway, K. O. Leedy, D. E. Sawyer, and W. J. Keery)

There have been several experimental determinations of ʼn using a variety of experimental techniques [71-75]. In the energy range of interest here (usually E≤ 30 keV), the fraction of electrons backscattered from aluminum or silicon is almost independent of the beam energy, Eg, as shown in figure 49. Data on the fractional mean energy backscattered are scarce [70,76,77]. _Figure 50 illustrates the variation of E./E, with Bck' B

beam energy for electrons backscattered from aluminum. The values given by Thomas [76] were measured at 138 deg with respect to the beam direction; the average value over all backscattering angles would be greater. values of f. for an aluminum specimen can be

B