transmitted or if the line is opaque on a transparent background, the sign of the error is reversed.

This may also

It should be stressed that these results are for the case of wide lines where the two edges may be treated independently. It should also be noted that in practice higher numerical aperture objectives are frequently used in line-width measurements although they are not used to full capacity due to a lower numerical aperture of the condenser. At best they are illuminated non-uniformly which reduces their resolving power. be true of the 0.65 numerical aperture considered here. Many other factors which affect line-width measurement by narrowing or broadening the line image were not considered. These factors include the properties of the photomask materials, electronic clipping and edge enhancement, and the variability of the human eye. The method of calibration of the measuring instrument also was not considered.

230

μπ

(D. Nyyssonent)

Figure 25. Interim chromium-on-glass linewidth measurement artifact with transparent lines on an opaque background. (All dimensions are nominal, ±0.25 μm.)

A series of measurements was made on the dimensions A, B, and C on the artifact illustrated in figure 25. These measurements were made to obtain additional data to assist in developing a basis for comparison of measurements made with microscopes equipped with different eyepieces. All of these measurements were made using a narrow-pass filter with the band center at 486 nm. The center-tocenter spacing between two lines (dimension C) was measured on the NBS line-standard interferometer [48] as 34.924 ± 0.0025 μm. The uncertainty in this measurement is the 3sigma value and represents the range in which 99 percent of the mean values of any set of measurements of this dimension will fall when made with the NBS line-standard interferometer. The results of the measurements made with various microscope eyepiece combinations are listed in table 7; the uncertainties represent the sample standard deviation obtained for each set of measurements. The measurements were normalized to the mean value of dimension C. The standard deviation associated with dimension C represents calibration uncertainty.

Table 7

Reflected Light, Bright Field

Reflected Light, Dark Field

Differences between values measured with filar and image shearing eyepieces for the brightfield, transmitted-light case agree in sign with the predictions of the calculations summarized in table 6 for matched numerical apertures. The differences for the brightfield, reflected-light case are opposite in sign as predicted from the discussion of section 7.1. Caution, however, should be used in comparing the differences in table 7 with the errors predicted in table 6. model used to predict the errors assumed no interaction between the two edge profiles, an assumption probably not valid for 5-um wide lines.

The

(F. W. Rosberry* and D. B. Novotny)

A polarizing interferometer with a 1 nm displacement resolution was constructed to make in-situ measurements of the movement of a scanning electron microscope (SEM) stage without interference from subtle changes in the optical path length caused by flexing of the vacuum system walls. A photograph of this interferometer mounted on an SEM stage is shown in figure 26; the configuration of its components is also shown schematically in the figure. The principles of operation of this interferometer are similar to those previously described (NBS Spec. Publ. 400-17, p. 36). The relative sizes of the interferometer and other components shown in the

photograph may be estimated by comparing them with the diameter of the specimen, S, which is 12 mm. The interferometer consists of one beam-splitting polarizing cube, A; one corner cube, B; three quarter-wave retardation plates, C (the one at the bottom of the interferometer is not shown); and a reference mirror cemented to one of the retardation plates, C2. The path of the laser beam is shown by a dashed line. The beam enters the SEM chamber through a side port and is directed to the interferometer and to the stage mirror, M, by the system of prisms, P

1'

This circuitous beam path was necessary to minimize the blockage of secondary electrons by the interferometer and optical components. Failure to minimize this blockage was shown to result in an unacceptably low signal-tonoise ratio. The positions of the interferometer and its components were determined by placing objects simulating their sizes into the SEM chamber in various possible alignment configurations. All of the configurations which lent themselves to easy optical alignments blocked 80 to 90 percent of the emitted secondary electrons which would normally be detected. The illustrated configuration results in a decrease of only 20 percent in the normal signal-to-noise ratio. (A. W. Hartman*)

NBS Optics and Micrometrology Section, Mechanics Division.

A computer program was developed to solve for the error in measuring the sheet resistance of a van der Pauw structure designed with orthogonal boundaries and finite size contacts. This program uses a nine-point finite difference approximation to Laplace's equation and a successive over-relaxation solution [50]. This solution is used to set up the indefinite admittance matrix for a four terminal network which is solved for the van der Pauw voltage [51]. Checks for residue oscillation are made and the over-relaxation factor adjusted accordingly. The program starts with a coarse grid which is refined in steps until the desired accuracy is obtained. At present this program can operate on several geometries ranging from a square to a cross, including all orthogonal "pinwheels" having rotational symmetry.

metric, orthogonal quadrate crosses in figure 30. These figures can also be used to obtain the true sheet resistance from

R_(true)

S

This analysis demonstrates that many novel van der Pauw designs are possible. For example, a cross constructed with minimum process stripe widths (D/S = A/S = 1.0) is found to have negligible error. (J. M. David)

8.2. Charge-Coupled Device Test Pattern

This study was undertaken to investigate the applicability of the charge-coupled device (CCD) as a test structure for use in semiconductor process control. The study was concluded with additional correlation of parameters as measured from various CCD structures and those measured using more conventional structures such as MOS capacitors, MOS transistors, and gated diodes.

Additional measurements were made of oxide thickness, dopant density, flat band voltage, and mobile charge density on two wafers using both the 32-bit and 128-bit circular CCDs (operated as capacitors) and various MOS capacitors (NBS Spec. Publ. 400-17, pp. 28-31). The results are summarized in table 8.

Additional measurements of transistor parameters were also made on a wafer using several MOS field effect transistors and the 32bit and 128-bit circular CCDs operated as transistors (NBS Spec. Publs. 400-12, pp. 23, 25, and 400-17, pp. 30-31). The results of these measurements are summarized in table 9. Surface state densities were determined on gated diodes by measuring the leakage current as a function of gate voltage. When this is done the fraction of the leakage current due to the surface states can be extracted. This fraction, I is given by gen, s'

where q is the electronic charge, A is the area of the gate, n is the intrinsic cari rier density, and s is the surface recombination velocity which is given by

о