1. Introduction

Surface quality assurance has been indispensable in various surface treatments and related applications, including semiconductors, displays, mobile phone components, high-quality optics, and many others [1–5]. In order to identify and quantify micrometer-sized surface defects in these applications, several attempts have been made using vision systems, scatterometers with light-emitting diodes or laser illumination, and topographical surface profiling methods including atomic force microscopy (AFM), and confocal microscopy [6–16]. In recent years, industries push the challenging demands on the surface inspection of larger targets with achieving higher precision levels, while maintaining high-throughputs. However, the microscopy based techniques have fundamental limitations in that the challenging demands owing to their intrinsically low measurement speeds and the vision based inspection system requires of high-cost cameras incorporating dense pixels with sophisticated imaging systems as the target size increasing with maintaining precision.

Among the aforementioned attempts, a notable approach is the use of a laser-based scatterometer comprising a circular-shaped waveguide for the response to the detection of small defects with high-throughputs in production lines [15]. However, the circular-shaped waveguide in this inspection system suffers not only from inefficient beam coupling of scattered light from the defects, but also significant power loss owing to beam leakage while it propagates in the waveguide. These fundamental problems limit the precision of this measurement system as 10-µm which was the minimum measurable size of defects. In this investigation, we developed an alternative approach using a laser scatterometer comprising a polygonal-shaped waveguide, so that it enables substantially decreasing the beam leakage in the waveguide, as verified via a ray-tracing simulation. Furthermore, the double-slit aperture installed in the waveguide effectively discriminates between the normal and defective surfaces. The performance of this laser scatterometer was evaluated using standard specimens with digs and scratches. Then we validated the developed measurement system by installing it in a pilot production line for inspection of touch-screen panels.

2. Basic principle and system implementation

Figure 1(a) shows the optical configuration of the polygonal-shaped waveguide based scatterometer in this investigation. The system consists of a laser scanning part and a detection part with the waveguide. A collimated beam having wavelength of 630 nm from a diode laser is continuously scanned via reflection on the facets of a rotating mirror. A photodetector (PD) is installed so that trigger signals are generated at the end timing of every scanning line by each facet of the rotating mirror. The beam after reflection from the rotating mirror is incident on an f-theta lens for focusing on a flat target while maintaining a constant the spot size on the entire scanning line. After the focused beam is reflected from the target, the beam is incident on a double-slit attached to the waveguide. Figure 1(b) shows the basic principle of discrimination between a normal surface without defects and an imperfect surface with defects by the waveguide with the double-slit aperture. The double-slit aperture is designed so that only the scattered beam from the defect can be coupled with the waveguide.

 

Fig. 1. Basic operating principle of the laser scatterometer using a waveguide with a double-slit aperture. (a) Optical configuration of the system. (b) Cross-section of the waveguide with or without beam coupling attributable to normal or defective targets. (c) Beam propagation in the waveguide with forward or backward coupling. PMT: photomultiplier tube, PD: photodetector, f-θ lens: f-theta lens

Download Full Size | PDF

The specular reflection from the normal surface without defects is blocked by the double-slit aperture. The coupled beam then propagates in the waveguide and is detected by the photomultiplier tube (PMT), as shown in Fig. 1(c-1). On the other side of the waveguide, an end mirror is attached in order to reflect back the beam coupled in the reverse direction, as shown in Fig. 1(c-2).

The coupled beam propagating in the waveguide suffers from inevitable power loss owing to the leakage of rays through the double-slit aperture. As shown in Fig. 2(a), the path of the coupled beam in the waveguide is analyzed by simulation using the ray-tracing tool (Zemax) with non-sequential mode [17]. It was found that severe beam loss occurred at the beginning of the beam coupling, and the loss reduced as the beam propagated in the waveguide (Fig. 2(b)). The loss is originated from the retro-reflective effect owing to the symmetric geometry of the cross-section of the circular waveguide [18]. At the beginning of the beam coupling, most of the beam returns back to the entrance aperture of the double-slit and decouples from the waveguide. In order to reduce the beam loss in the waveguide, the geometry with breaking symmetry should be exploited to prohibit the retro-reflective effect. We adopted representative geometries in our ray-tracing simulation, including circular, triangular, square, and asymmetric rectangular (polygon), as shown in Fig. 3. We used 100,000 numbers of rays as scattered light emanating from the defect. The cone angle of the scattered light was set to 0.7° considering f/# of the f-theta lens and bidirectional scatter distribution function (BRDF) induced by micrometer sized defect [19]. The coating option of the interior faces of the waveguides was set to Au, provided by the material library in the ray-tracing tool after the geometries of waveguides imported from the CAD program. Owing to the high reflectivity of Au coating at 630 nm, we discarded the transmitted light through the coating, which induces multiple reflections inside the waveguide structure. We placed the detector viewer at the PMT position with setting 500 × 500 resolution without the smoothing option. The detection efficiency was evaluated by the ratio between the number of rays incident on the detector viewer and number of rays coupled to the waveguide. The simulation results showed that only 3.7% of rays from the scattered light were incident on the PMT in the case of the circular-shaped waveguide, while more than 50% of the rays were incident on the PMT in the case of the polygonal-shaped waveguide.

 

Fig. 2. (a) Simulation of ray leakage in propagating beam in circular-shaped waveguide. (b) Visualization of light leakage through the double-slit aperture as the coupled beam propagates in the waveguide. The white dotted line depicts the waveguide.

Download Full Size | PDF

 

Fig. 3. Simulation of detection efficiency of scattered light from the defect depending on waveguide shape: (a) circular, 3.7% (b) triangular, 9.8% (c) square 11.6% (d) asymmetric rectangular (polygon), 51.8%. The white dotted lines denote the cross-sectional shape of the waveguide. The orange solid lines denote effective detection area of the PMT.

Download Full Size | PDF

Figure 4 illustrates the detection part design, including the polygonal-shaped waveguide with a double-slit aperture and the PMT. The polygonal waveguide was fabricated by assembling four Au-coated acrylic plates. One of the Au-coated acrylic plates facing the target has a double-slit, as shown in the sub-diagram in Fig. 4. The entrance apertures were separated by 2.5 mm, which was identical to the beam size at the waveguide in order to block the specularly reflected beam from the target. We determined the size of the entrance apertures as 1 mm by a trade-off between the effective coupling of the scattered light and the leakage as the beam propagates in the waveguide.

 

Fig. 4. Hardware configuration of the detection part including polygonal-shaped waveguide and PMT.

Download Full Size | PDF

The waveguide holder was designed for a mechanical connection between the waveguide and PMT, as well as for mounting an optical diffuser and optical filter. The optical filter enables detection of only light having 630 nm wavelength by the PMT from the diode laser while discarding unnecessary stray light from the measurement environment. The optical diffuser was used for spreading the distribution of beam incident on the PMT detection area. The PMT (Hamamatsu, R11101-02) converted the intensity of the beam into an electrical signal with a gain of 10,000 through 500 V supply voltage. The signal from the PMT and the trigger signal from PD in the laser scanning part (Fig. 2(a)) were sampled using a two-channel high-speed digitizer (NI, PCI5122) with a sampling rate of 100 MHz of sampling resolution of 8-bits.

3. Experiments and results

In this investigation, we exploited a rotating mirror with four facets in the laser scanner part, and the rotation speed was set to 400 revolutions per second (∼24,000 RPM). The distance between the laser scanning part and target was adjusted to focus the laser beam on the target with a 10 µm beam size in the full-width-half maximum (FWHM). The duty cycle in our scanning was 50%, which is the ratio of the effective scanning time on the target and total scanning time [20]. The resulting laser scanning range limited by the effective scanning time was 250 mm, while preserving the size of the focused spot on the target. The focused laser spot scans the scanning ranges every 0.625 ms (=1/400 Hz/4 facets), producing 31,250 data (0.625 ms·100 MHz·0.5) on a single scanning line. The resulting longitudinal sampling resolution on the target was 8 µm (250 mm/ 31,250). The target is translated perpendicular to the laser scanning direction, as shown in Fig. 1(a) using a motorized stage (Sigma Koki, SGSP 20-85). The velocity of the target translation was set to 15 mm/s, and the resulting transverse resolution was 9.4 µm. The spatial resolutions in both the longitudinal and transverse directions were smaller than focused laser beam size, so that the scanning beam covered the entire target area without any missing between sampled data.

Figure 5 shows the calibration results of the developed scatterometer by measuring dig defects of tens of micrometers to hundreds of micrometers size on a standard specimen (Edmund Optics, #53-197). The standard specimen contains 7 rectangular sections and 7 circular sections with 5∼100 dig numbers and 10 to 160 scratch numbers, respectively, on an acrylic plate. At the center of the circular polished surfaces, extruded spot defects are located, and the intensity peaks are clearly shown at the defects position (Fig. 5(a)). The height of a digs are quantified using the value of peak intensity as well as the dig diameter by counting the pixels of the cross-sectional FWHM diameter, as shown in Fig. 5(c). For comparison purposes, we measured an identical specimen using a commercial surface profiler (Panasonic, UA3P); this instrument is based on point-by-point contact measurement providing sufficiently high sub-nm resolution, to calibrate the developed scatterometer (Fig. 5(d)) [21]. Both measurement results show good agreement with the variation in diameter and height depending on the dig number from 10 to 50. The values of the diameter and height of the digs measured using the two methods are plotted (Fig. 5(e,f)). The peak intensities measured by the scatterometer and heights measured by the commercial surface profiler show good linearity with a slope of 0.96 µm/count(Fig. 5(e)). The FWHM diameters of the digs (blue, right y-axis), it shows linearity with a slope of 7.3 µm/pixel(Fig. 5(f)).

 

Fig. 5. (a) Measurement results of digs (10, 20, 40, 70, and 100) of the standard specimen using the developed scatterometer (b) Magnified 3D view of Dig10 and Dig100, respectively. (c) Cross-sectional profiles measured by the scatterometer. (d) Cross-sectional profiles measured by commercial surface profiler (Panasonic, UA3P). (e) Calibration results of peak intensity vs. height (f) Calibration results of number of pixels of the diameter vs. dig diameter from the results presented in (c) and (d).

Download Full Size | PDF

Figure 6(a) shows the results of the scratch defects measured by the developed scatterometer. The standard specimen (Edmund Optics, #53-197) provides the extruded scratch defects, and the cross-sectional profile was calibrated using AFM (atomic force microscopy, PSIA, XE-100), as shown in Fig. 6(b) [22]. The FWHM width and height of the scratch defects span from 20 µm to 3 µm and from 2.3 µm to 0.4 µm, depending on the scratch number from 160 to 40. Figure 6(c) and (d) show the areal intensity map of scratch160 and scratch40 measured by the scatterometer, respectively. Micrometer-sized linewidths of the scratches are clearly shown at the location where the defects exist.

 

Fig. 6. (a) Measurement results of the scratch defects from 40 to 160 of the standard specimen using the developed scatterometer (b) Cross-sectional profile of the scratch defects calibrated using commercial AFM (PSIA, XE-100). (c,d) Magnified aerial intensity profiles of the scratch number 160 and 40, respectively, measured by the developed scatterometer.

Download Full Size | PDF

After the calibration and performance evaluation of the scatterometer using the standard specimen in the laboratory environment, we applied the scatterometer system to the inspection line of touch screen panels (TSPs), as shown in Fig. 7(a). The laser scanner part was mounted on a manual angle stage in order to adjust the incident angle on the targets. Rotation of the waveguide part was enabled by a rotational stage so that the double-slit on the waveguide was positioned along the direction of the reflected beam from the targets. Depending on the coupled direction of scattered light from the target, there was inevitable efficiency variation, because the scattered light coupled in the reverse direction propagates longer distance than the light couple in forward direction, as shown in Fig. 1(c). In order to compensate the efficiency variation, we measured an intensity profile of the reflected light from glass plates without defects by rotating the waveguide. The size of the TSPs in this inspection system was 250×140 mm² with an attached flexible printed circuit board (FPCB), as shown in Fig. 7(b). After the alignment of our measurement system, the TSP targets were translated under the scatterometer system at a speed at 15 mm/s. The inspection system was designed to inspect two TSPs at a time, and measurement time was less than 10 s per single target. The measurement time was limited by scanning speed of the beam on the target, which could be enhanced by increasing rotation speed or number of facets of the rotating mirror in the laser scanner part along with devising higher sampling rate of the digitizer in order to preserve the spatial resolution of the current system. Figure 7(c) shows the measurement result of the TSPs in the inspection line. The various defects on the TSPs, including particles, fingerprints, bubbles, and edges at the end of the glass surface, were successfully detected (Fig. 7(c)). Because the TSP target is transparent, the defect features of the bottom and top surface are superposed. The inspection system exploiting the scatterometer should make the pass/fail determination of a TSP target using the measurement result. The superposed result would allow the determination by a single decision-making process.

 

Fig. 7. (a) System for inspecting TSPs (touch screen panels) comprising the scatterometer (b) TSP samples (c) Measurement results of detecting various types of defects on TSPs in the inspection line. P: particles, F: Finger prints, B: Bubbles, E: Edges of the TSPs.

Download Full Size | PDF

4. Conclusion

We demonstrated a laser scatterometer using a polygonal-shaped waveguide for flat surface inspection. Breaking the symmetry of the rectangular cross-section of the waveguide enabled substantial improvement in the detection efficiency, thus resulting in sub-µm detection precision. By identifying defects in a standard target and through comparisons of the results obtained using the developed scatterometer and a commercial surface profiler, we observed that the scatterometer results show good linearity in terms of the size and height of the defect structures. The scatterometer was successfully applied to a production system for inspection of TSPs with a size of 250×140 mm²; it detected various defects on the TSPs with measurement time less than 10 s per single target. The improved scatterometer with an optimized shaped waveguide would be beneficial in the various applications requiring rapid and precise measurements of surface quality with a measurement range exceeding hundreds of millimeters.