lines (spaces) in an opaque background were measured with filar and image-shearing (intensity splitting) eyepieces using bright-field transmitted illumination. Five measurements were made of the width of each of three lines and spaces using each eyepiece. The means and three-sigma values (three sample standard deviations) are listed in table 11. These data further confirm the previously reported result that with transmitted illumination the filar eyepiece indicates a wider width than the image-shearing eyepiece for clear spaces and a narrower width for opaque lines (NBS Spec. Publ. 400-19, pp. 38-41, 41-43).

objective lens had a magnification of 63x, a numerical aperture of 0.90, and a depth of field of 1.23 μm. The position of best focus was a subjective decision by the operator based on the quality of the image. The results obtained at seven focal positions, extending beyond the normal depth of field, are given in table 12. Within this range, the clear line measures maximum width at or near best focus. This result is also consistent with the previously reported model (NBS Spec. Publ. 400-19, pp. 38-41). (F. W. Rosberry#)

An experimental comparison was made of linewidth measurements using optical microscopes and operating conditions representative of those employed by the microelectrics industry for photomask inspection. Most photomask line-width measurements made in industry are done with conventional optical microscopes fitted with visual measuring eyepieces of either the filar or image-shearing types; to a limited degree, automatic TV-microscope systems are in use.

The majority of measurements reported here were made with a monocular research-quality microscope with separate filar and imageshearing eyepieces. During the course of the study, a binocular image-shearing microscope and an automatic TV-microscope system were made available on loan for short term use.

The monocular microscope was equipped with an objective lens, with a numerical aperture of 0.95 and a magnification of 80x, and either bright-field or dark-field illumination; its image-shearing eyepiece had an intensity beamThe binocusplitter as the shearing element. lar microscope was equipped with an objective

-1.5

-1.0

-0.5

#.

NBS Optics and Micrometrology Section, Mechanics Division.

lens, with a numerical aperture of 0.90 and a magnification of 63x, and bright-field illumination only; its image-shearing eyepiece had a polarization beamsplitter as the shearing element. The TV-microscope system was equipped with an objective lens, with a numerical aperture of 0.95 and a magnification of 80x, and bright-field illumination only. All three microscopes had both reflected and transmitted light capabilities. The three systems differed in their fundamental principles of operation and means of data output. The filar and image-shearing eyepieces require operator judgment for locating image edges in a line-width measurement; the TV system uses automatic image scanning and electronic thresholding. The filar eyepiece had a vernier scale readout which requires the calculation of line widths from arbitrary scale divisions; the binocular image-shearing system had direct digital readout as did the TV-microscope system.

Three filar eyepieces were available for use with the monocular microscope. One, with a 7x magnification and a movable, full-field pair crosshair configuration, had the best repeatability of the three and was used throughout the study. Preliminary repeatability tests for each of the optical microscope measurement systems resulted in variations from the mean of ten measurements of the width of a 5-um transparent line of 0.03 to 0.09 um (three sample standard deviations). The linearity of each of the microscope systems was checked for gross misbehavior by measuring four line spacings after calibration with the NBS line-standard interferometer [82]. Within the measurement repeatability and the calibration accuracy, no significant deviations from linearity were observed. Cosine error results when the axis of measurement is not perpendicular to the line whose width is to be measured. In a filar eyepiece, cosine error is minimized when the crosshair lies parallel to the line and travels perpendicular to it. In the shearing eyepiece, the direction of shear must also be perpendicular; similarly with a TV-microscope system. For filar and image-shearing eyepieces, the degree to which the line to be measured can be made to lie along the axis of measurement depends on the fraction of the field of view the line occupies and the least detectable increment of field of view between some part of the line and a reference mark on the viewer. A series of experiments showed that, provided the line fills at least 5 to 10 percent of the field, misalignment need not be a significant source of measurement error.

Detailed line-width measurements were made with each of the systems on transparent lines of 2-, 5-, and 10-um nominal width in an opaque field. Each microscope was calibrated in magnification using a line spacing previously measured to be 34.927 ± 0.007 μm with the NBS line-standard interferometer [82]. In addition, the TV-microscope system was adjusted to measure a line-to-space ratio of a value predetermined by an artifact supplied by the manufacturer.

The results of the measurements appear in table 13. Each of the entries in the table represents the mean of ten measurements taken one after another. To determine how much significance can be attributed to each entry, measurements on each of three systems were repeated ten times successively on nine occasions for measurements of the 5-μm line in bright-field transmitted illumination. Typical data for a series of ten measurements on a single occasion are summarized in table 14 which also shows the arithmetic means and sample standard deviations of the means for nine such series taken on different occasions.

A study was undertaken to delineate the tolerances, design criteria, and limitations associated with the measurement of line widths by analysis of the diffraction pattern. The instrument for performing this measurement is shown schematically in figure 43 [84]. A laser beam is shaped into a line by means of two cylindrical lenses, C1 and C2, and is focused on the plane, P, through an adjustable aperture slit A. The specimen to be measured is placed at the focus of the slit image, S, formed by a spherical 50x microscope objective lens with a numerical aperture of 0.85. This lens also serves to focus the effective source, P, on a diode array, D, located an appropriate distance behind the specimen.

In this arrangement, the diffraction pattern in the plane of the diode array is given by the convolution of the source image with the object transform. In the case of a gap of width w in an opaque background in the object plane (fig. 44a) with ideal transmission characteristics (fig. 44b), the object transform is given by wsin("fw)/(πfw) (fig. 44c). The intensity distribution of this diffraction pattern (fig. 44d) is measured and digitized (fig. 44e), and a discrete Fourier transform is computed (fig. 44f). The intercept of this autocorrelation function on the horizontal axis occurs a distance w from the origin. When the instrument is used to measure opaque lines, it is necessary to consider the image of the aperture slit which surrounds the line in the object plane. In this case, provided that the line is reasonably well centered in the image of the slit, the autocorrelation function has a region of positive slope to

the right of the vertical axis which, when extrapolated back, intercepts the horizontal axis a distance equal to the width of the line from the vertical axis [85].

Experiments were carried out to determine the sensitivity of the system to focus, aperture size, source intensity, the nature of the object measured (line or gap), and the width of the object. The measurement process was found to be sensitive to the centering of the line in the aperture, the aperture width relative to the line width, and focus and location of the aperture. In addition, a theoretical analysis showed that if the optical density of the opaque regions is less than 3.0 the positions of the intercept can be significantly different from that expected on the basis of ideal considerations. Variations in the intensity of the illuminating laser do not appear to affect the measurement.

Better linearity and repeatability were achieved for measurement of opaque lines in a clear field than for measurement of clear gaps (or slits) in an opaque field. This appears to be due in part to the different criteria used for centering the aperture on the object plane; the criterion used in the case of opaque lines is both more sensitive and more repeatable.

Analysis of the optical system shows that the system response depends on the uniformity and phase of the aperture illumination and on the numerical aperture and quality of the objective lens as well as the characteristics of the line or gap being measured. Anomalies which have been noted in the autocorrelation functions suggest that additional analysis and instrumental improvements are required to

Figure 43.