JP2000221008A - Method and device for inspecting transparent substrate - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform rapidly 100% inspection of appropriateness of a cross section in convex working by use of a polarization observation method. SOLUTION: First, with the use of a reference sample of transparent substrate, an analyzer 22 or a polarizer 21 is rotated so as to maximize a luminance in the center of the same cross section of the reference sample, then the luminance is computed. By inserting a 1/2-wavelength plate 24 into an optical path at an angle of 22.5 deg., an optical phase difference is shifted by 45 deg., then the luminance in the center is computed. A sinusoidal waveform passing through both points of the luminance maximum value in the center and the luminance value for the phase difference 45 deg. is computed. This sinusoidal waveform is regarded as a reference luminance distribution waveform in the same cross section of the reference sample provided by visualization of the phase difference. Next, with the use of a crystal blank 1 to be inspected, the luminance in the same cross section of the reference sample is computed, then a luminance distribution waveform is detected. By comparing this detected luminance distribution waveform with the reference luminance distribution waveform, a thickness in the same cross section of the transparent substrate is relatively inspected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for inspecting a transparent substrate, and more particularly to an apparatus and method for inspecting the thickness of a convex type transparent substrate in real time by a polarization observation method.

[0002]

2. Description of the Related Art As a technique for improving the characteristics of a crystal resonator, beveling (chamfering) the periphery of a flat-plate type crystal blank has been performed. Further, a flat plate type crystal blank is formed into a lens shape by convex processing. In order to perform beveling processing and convex processing, a cylindrical barrel method capable of mass processing is commonly adopted. As shown in FIG. 18, the abrasive g is put in the cylindrical barrel 41 together with a large number of crystal blanks 1 and the cylindrical barrel 41 is polished while rotating the cylindrical barrel 41. Polishing in a relatively short time tends to beveling processing, and polishing in a long time tends to convex processing, but cannot be concluded because various conditions work. In order to improve the characteristics of the crystal unit with good reproducibility, it is necessary to increase the dimensional accuracy of the beveling and the convex by the above-mentioned method so that the crystal unit is symmetrical left and right and up and down.

In the cylindrical barrel method, beveling and convex processing conditions are different depending on the number of quartz substrates to be charged, the type of abrasive, the number of rotations, the temperature, and the like. Even if these conditions are the same, the same beveling state or convex state is always used. Shape is not always obtained. Quartz substrate manufacturers often rely on their own know-how, but the know-how also has uncertain factors, and it does not work well, so it is not always possible to maintain the same beveling and convex dimensional accuracy. At present, elements are also mixed. In order to improve beveling or convex dimensional accuracy, it is necessary to establish an inspection method capable of evaluating the beveling state and convex shape of the quartz crystal blank surface and feeding back the results to beveling processing and convex processing.

[0004] Beveling is performed by making very small scratches and shaving off a part of the surface, so that there is no significant difference from a non-beveled part in ordinary bright-field observation, and the crystal blank surface The beveling state cannot be inspected by visual inspection of an operator or image processing technology. Convex processing is also achieved by shaving off a part of the surface in the same way, but ordinary bright-field observation does not make it possible to grasp the degree of surface polishing. And cannot be inspected with conventional image processing techniques.

For this reason, the conventional beveling inspection or convex finishing inspection is carried out by (1) pen recorder method,
Physical methods such as (2) level measurement and (3) projection have to be relied on.

(1) Pen Recorder Method This is a method in which a quartz blank surface is scanned by a probe and its locus is recorded using a pen recorder, and the surface height of the probe along the scanning direction can be measured. In the case of the convex type, since the stability is poor, fixing to the mounting table with an adhesive is performed.

(2) Level measuring method This method uses a laser height measuring machine. A silver film for reflection is vapor-deposited on the back of a quartz blank, and a laser beam is irradiated on the front of the quartz blank in the radial direction (linear direction). ), And the level on a straight line is measured from the reflected light. According to this, as shown in FIG. 19, the surface height of the crystal blank along the linear direction can be continuously measured, and the state of beveling or convex processing can be known from the surface height. In the figure, the solid line is beveling,
The dashed line indicates the convex. In the case of the convex type, since the stability is poor, fixing to the mounting table with an adhesive is performed.

(3) Projection method Carbon powder (black) is applied to the surface of a quartz blank,
This is a method in which a semi-transparent film or thin paper is pressed thereon, and the carbon powder is transferred to the film or thin paper. The transferred film is enlarged and projected on a screen. Since carbon adheres to the non-beveled part, the beveling state can be visually recognized from the transfer state of the carbon powder. This method is not applicable to convex type.

[0009]

However, the above-mentioned conventional method has the following problems.

(1) Pen Recorder Method Since the accuracy of measuring the height of the polished surface is poor and only data on a straight line obtained by scanning a quartz blank is obtained, beveling or convex inspection cannot be performed on the entire quartz blank. In addition, since the sampling inspection is performed, it takes time for the inspection, and the inspected sample cannot be used. Further, in the case of the convex type, a troublesome operation of fixing with an adhesive is required.

(2) Level Measuring Method The height of the polished surface can be measured with high accuracy, but since only data on a straight line obtained by scanning a quartz blank can be obtained, the entire surface of the quartz blank cannot be subjected to beveling or convex inspection. . If the scanning in the horizontal direction is shifted in the vertical direction, surface data will not be obtained, but missing data corresponding to the vertical pitch width cannot be avoided. Therefore, it becomes difficult to inspect the entire surface of the crystal blank. In addition, since it is necessary to deposit silver on the back surface, the inspection becomes a sampling inspection, which takes a long time, and the inspected sample cannot be used. In the case of the convex type, a troublesome operation of further fixing with an adhesive is required.

(3) Projection method Although inspection can be performed over the entire surface of a quartz blank, it is necessary to extract, transfer, and project, so that a sampling inspection is required, and the inspection takes a very long time. In addition, inspection data cannot be obtained due to visual observation. Inspection of convex processing is not possible.

As described above, any of the methods (1), (2) and (3) cannot be inspected in real time, and it takes time. In addition, since a unified inspection method cannot be established, beveling inspection or standardization of convex processing cannot be performed. In addition, since appropriate data cannot be obtained, feedback cannot be effectively provided to the beveling processing technology or the convex processing technology.

In addition, since the beveling or convex processing is a batch method in which a large amount is processed, the state of beveling or convex processing differs for each crystal blank, and variations occur even in the subsequent etching by an acid treatment. Nevertheless, since the inspection standard has not been determined, it is difficult to determine the allowable range, and there has been a problem that a non-defective product is treated as a defective product or a defective product is made a non-defective product.

By the way, in addition to the beveling inspection, a flaw inspection and an outline (dimension) inspection are also required for the quartz substrate. Recently, it has been proposed that a dark field observation method is effective for these inspections (Japanese Patent No. 282460). Then, when this method was applied to the beveling inspection, it was found that the method was effective even after the beveling process but before the etching, since the minute flaws formed during the beveling process remained on the surface. However, it was found that the dark-field observation method could not be applied because the surface was smoothed after etching and fine scratches disappeared from the surface. In addition, it was found that the method cannot be applied to a convex type quartz substrate having different thicknesses at the center and the periphery. This is because the convex type quartz substrate has a more uniform surface as compared with the beveling process, and is difficult to detect by the dark field observation method.

An object of the present invention is to provide a method and an apparatus for inspecting a transparent substrate capable of objectively inspecting the cross-sectional shape of the transparent substrate by image processing while solving the problems of the prior art. is there.

[0017]

SUMMARY OF THE INVENTION The gist of the present invention is to introduce a polarization observation method so that the finished state of a convex type transparent substrate having the same cross section and different thickness can be imaged. The cross-sectional shape of the convex type transparent substrate, which cannot be inspected by the dark-field observation method, can be inspected by image processing.

According to a first aspect of the present invention, a light from a light source is polarized by a polarizer and applied to a transparent substrate made of a convex-processed birefringent medium by a polarizer. In the method of inspecting a transparent substrate, the light having a polarization state to be formed is transmitted, a phase difference due to a difference in the thickness of the transparent substrate is visualized by polarization interference, and an inspection of a finished finish of the transparent substrate is performed. Compare the luminance distribution in the same cross section of the transparent substrate obtained by visualizing the phase difference with a reference luminance distribution obtained in advance, and inspect the thickness of the transparent substrate in the same cross section of the transparent substrate. It is an inspection method. Convex processing generally includes grinding or polishing and etching. The transparent substrate includes a crystal unit and a crystal blank for a filter. The thickness at the same cross section of the transparent substrate includes the thickness and the cross sectional shape at the center.

When a transparent substrate is subjected to convex processing to have different thicknesses in the same cross section, it is necessary to inspect the thickness as viewed from the cross section. Polarized light from a light source is given to the processed transparent substrate, and light in a polarization state forming a predetermined angle with respect to the polarized light from the light passing through the transparent substrate is passed,
When the transparent substrate is imaged with this light, a phase difference occurs due to the difference in thickness, and interference fringes consisting of light and dark appear on the image of the transparent substrate. From this image, a brightness distribution can be obtained by plotting the light and dark on the same cross section. This luminance distribution is correlated with the thickness of the transparent substrate at the same cross section.
Therefore, when the detected luminance distribution waveform of the transparent substrate is compared with a previously determined reference luminance distribution waveform, the amount of deviation of the detected luminance distribution waveform with respect to the reference luminance distribution waveform can be determined. Can be inspected.

It is preferable that a luminance distribution in the same cross section of the transparent substrate obtained by visualizing a phase difference due to a difference in thickness of the transparent substrate has a correlation with the following equation.

[0021] _{P = (n e -n 0)} t · 2π / λ However lambda: wavelength of the light source n _0: the ordinary refractive index n _e: refractive index of extraordinary ray t: thickness of the transparent substrate P: fringe Phase difference between

The inspection is preferably performed in the following procedure.

(1) First, the analyzer or the analyzer is so set that the center of the luminance distribution in the same cross section of the standard sample obtained by visualizing the phase difference due to the difference in thickness of the standard sample on the transparent substrate is maximum. By rotating the polarizer, the maximum value of the luminance at that time is obtained. (2) A half-wave plate is inserted at an angle of 22.5 ° with respect to the optical axis to shift the phase difference of light by 45 °. The luminance value of the central part at that time is obtained, and (3) the maximum value of the luminance of the central part and the phase difference 4
A sin waveform passing through two points with a luminance value of 5 ° is obtained.
The n waveform is regarded as a reference luminance distribution waveform in the same cross section of the standard sample obtained by visualizing the phase difference,
(4) Next, a luminance distribution waveform is detected from a luminance distribution in the same cross section of the transparent substrate obtained by visualizing a phase difference due to a difference in thickness of the transparent substrate to be inspected, and (5) a detected luminance distribution The thickness of the transparent substrate at the same cross section is inspected by comparing the waveform with the reference luminance distribution waveform.

A cos waveform may be used instead of the sin waveform.

In the case of inspecting the finished finish of a convex type transparent substrate, the above-described first invention is particularly useful because a dark field observation method useful for inspecting a flaw cannot be applied. The transparent substrate is particularly preferable when it is a quartz blank for a quartz oscillator.

According to a second aspect of the present invention, there is provided a polarization observation light source for irradiating a non-polarized light toward a transparent substrate made of a convex-processed birefringent medium; A polarizer that irradiates the substrate, an analyzer that detects light having a predetermined angle of polarization transmitted through the transparent substrate, and an image of the transparent substrate of polarization observation light that has passed through the analyzer, and the transparent substrate is subjected to polarization interference. Imaging means for visualizing a phase difference due to a difference in thickness of the image, image processing means for processing an image visualized by the imaging means to detect a luminance distribution in the same cross section, and a luminance distribution of a detected transparent substrate and a standard sample And a comparing means for comparing the reference luminance distribution obtained from the above with the reference luminance distribution and inspecting the thickness of the transparent substrate at the same cross section.

This allows the first method of the present invention to be appropriately carried out with a simple structure, and the equipment cost and maintenance operation cost can be reduced.

[0028]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic configuration diagram showing an embodiment in which an inspection apparatus for performing an inspection method of a transparent substrate according to the present invention is applied to a convex type crystal blank for a crystal unit.

The inspection apparatus comprises a monochromatic light source 26 having no polarization.
And a diffuser plate 2 for diffusing the light of the light source 26 and irradiating the light uniformly.
3 and a polarizer 21 for polarizing light from the diffusion plate 23
A mounting table 2 on which a quartz blank 1 as a transparent substrate made of a birefringent medium is placed; an analyzer 22 for passing light having a predetermined angle with respect to the polarized light from light passing through the quartz blank 1; It comprises a lens 12, a CCD camera 13 as an imaging means, and a monochrome image processing device 14. Here, as the monochromatic light source 26, a 660 nm red LED having a high luminance and a long life is used. Note that the polarizer 21, the mounting table 2, and the analyzer 22 are all provided rotatably so that they can be fixed after adjusting the rotation angle. The shorter the distance between the polarizer 21 and the mounting table 2 and the shorter the distance between the analyzer 22 and the mounting table 2, the better. Further, if necessary, a condenser lens for condensing light on the crystal blank 1 is provided between the diffusion plate 23 and the polarizer 21, or a compensator for making the crystal blank image stand out clearly from the background is provided on the analyzer 22 and the mounting table. 2 may be inserted. Further, between the analyzer 22 and the crystal blank 1,
By inserting the half-wave plate 24 at an angle of 22.5 ° with respect to the optical axis, the phase difference of light can be shifted by 45 °. The half-wave plate 24 may be inserted between the mounting table 2 and the polarizer 21.

The crystal blank mainly to be inspected by the inspection apparatus is a biconvex convex type crystal blank 1 having a thinner peripheral portion and a thicker central portion as shown in FIG. Then, as shown in FIG. 2B, the substantially flat type is wrapped (step 101), the peripheral portion is polished by beveling (step 102), and the peripheral portion is further etched (step 103). Except for this, there is a crystal blank 17 of a beveling process whose entire surface is smooth. As a substantially flat-shaped crystal blank subjected to beveling, for example, in the case of a strip shape, the size is usually 1 ×.
3mm-3x10mm, thickness 30μm-500μm
It is about. The shape of the crystal blank 17 is not limited to a rectangular shape or a rectangular shape, but may be a circular shape having a diameter of about 4 to 8 mm or other shapes. The shape of the convex type is almost the same.

The image processing device 14 shown in FIG.
An image of the crystal blank 1 captured by the D camera 13 is taken in, features are extracted as in, for example, a binarization process, and processing data is stored or processing finish determination processing is performed based on the extracted features.

The illumination light from the light source 26 is diffused by the diffusion plate 23, is converted into linearly polarized light by passing through the polarizer 21, and is incident on the crystal blank 1 on the mounting table 2. The linearly polarized light incident on the quartz blank 1 is divided into an ordinary ray and an extraordinary ray, and these rays cause a phase lag with respect to one another, so that interference occurs. The interference differs depending on the phase difference and the intensity (amplitude). The phase difference is determined by the thickness t and the wavelength λ of the quartz blank 1 (see FIG. 3B), and the intensity is determined by the linear polarization from the polarizer 21 and the quartz blank 1.
Is determined by the angle θ formed with the main surface of

For example, the peripheral area becomes a dark area, the central area becomes a light area, and a phase difference due to a difference in thickness is visualized.
Therefore, the light emitted from the analyzer 22 becomes strong and weak interference light due to the strengthening light and the canceling light in the white light. This interference light passes through the objective lens 12 and enters the CCD camera 13. The image of the crystal blank 1 is converted into an electric signal by the CCD camera 13 and transferred to the image processing device 14. The image processing device 14 performs statistical processing so that a dark portion generated in a peripheral portion and a bright portion generated in a central portion can be clearly distinguished. For example, a contrast enhancement process between a dark portion and a bright portion is performed by a binarization process (FIG. 3A). By measuring the dark area or the light area after the processing, the processing finish such as excessive shaving, insufficient shaving or symmetry is inspected.

As the observation method applicable to the finished processing inspection, other observation methods such as a phase difference observation method, a differential interference observation method, and a modulation contrast observation method can be considered in addition to the polarization observation method described above. The reason for adopting the polarization observation method in the above mode is that the light source may be an LED, and the polarizing filter such as a polarizer or an analyzer may be of such a size that a polarizing film is attached to a glass plate, and is the cheapest and simplest.

FIG. 4A schematically shows an image of the crystal blank 1 obtained by the polarization observation method shown in FIG. The crystal blank 1 is a convex type as described above. Using monochromatic light as the light source 26, the polarizer 21, the analyzer 22, and the mounting table 2 are rotated with respect to each other, and
Is an image when interference fringes appear most clearly at the periphery of the image. The peripheral portion having a small thickness becomes “dark” because the phase difference Δφ is an odd multiple of π, and the central portion having a large thickness becomes “bright” because the phase difference Δφ is an even multiple of π. In addition, crystal blank 1
It is considered that the four corners become "bright" because the birefringence increases due to stress due to processing, lens effect, and the like, and is a phenomenon different from the above-described interference fringes due to the thickness.

FIG. 4B shows an image of the quartz blank 1 obtained by the dark field method. Here, the dark field method is a method according to the aforementioned Japanese Patent No. 282460. In this case, only the outline is raised white, the inside of the outline is uniformly gray, and the difference in thickness is visualized inside. I can't find an image that does it. As described above, since the surface of the convex type is uniformly scratched, the dark field method effective for the scratch is not suitable for inspecting the finish of processing the curved surface of the quartz blank.

Polarized light from a light source is given to a convex type quartz blank by a polarizer, and light having a predetermined angle with respect to the polarized light from the light passing through the quartz blank is extracted from the analyzer. A fringe pattern (contour pattern) appears around the blank. Many interference fringes occur when the blank size is large and thick. However, the interference fringes decrease when the blank size is small and thin. The long side is 4 mm, the short side is 2 mm, and the thickness is 100 to 200 μm.
With a small rectangular blank, only one dark interference fringe is formed on the outer periphery. In addition, the main force of the crystal blank has shifted to a small one by the downsizing of the mounting equipment.

In the embodiment, it is assumed that there is a correlation between the shape of the interference fringes and the thickness of the quartz blank. This makes it possible to visualize the phase difference due to the difference in thickness due to the interference fringes, and to inspect the shape of the interference fringes in a planar view by inspecting the shape of the interference fringes. Feedback for improvement of the barrel shape.

In order to verify the above premise, the relationship between the interference fringes and the thickness will be considered below. For this consideration, a method is first employed in which a theoretical value is obtained, and then the theoretical value is confirmed by experiments. However, the sample used in this experiment should originally be able to use a small crystal blank according to the actual situation. However, in practice, a large convex type crystal blank was used instead of the small crystal blank. This is because the larger one is clearer and produces a plurality of interference fringes, and the larger one is easier to measure the shape (thickness) of the pen recorder than the smaller one. 5 and 6 show the large sample (long side 8.25) described above.
(mm to 7.0 mm, short side 1.5 mm to 2.0 mm, thickness 400 μm to 450 μm), showing images in which two and three black stripes are generated, respectively. FIG. 7 shows the aforementioned small sample (long side: 5.0 mm to 4.0 mm, short side: 1.5 mm to
2.5 mm, 100 μm to 200 μm in thickness), each showing an image with one black stripe. [1] Theoretical value Quartz is a birefringent device, and the light incident on it is split into an ordinary ray and an extraordinary ray whose polarization planes are orthogonal to each other and propagates.

The crystal blank has a function as a phase plate, and gives a predetermined phase difference by a difference in phase speed between the ordinary ray and the extraordinary ray. The quarter-wave plate converts linearly polarized light to circularly polarized light,
Alternatively, the half-wave plate has the function of rotating the circularly polarized light into linearly polarized light and rotating the linearly polarized light by (2θ) with respect to the angle θ between the incident polarization plane and the optical axis of the wave plate.

The wavelength of the light source is λ, the refractive index of the ordinary ray is n ₀ ,
The refractive index of the extraordinary ray When n _e, the relationship between the thickness t and the phase difference P of the crystal blank, P = a _{(n e -n 0) t ·} 2π / λ (1).

From equation (1), it can be seen that the thickness of the quartz blank can be determined if the phase difference, the difference between the refractive indices of the ordinary ray and the extraordinary ray, and the wavelength of the light source are known. The following data is obtained by determining the refractive index of the quartz crystal in the Y-axis cut for a specific wavelength.

The lambda (Å) inspection of the crystal blank refractive index _{n e 5080 1.54822 1.55746 5893 1.54424 1.55335} 7680 1.53903 1.54794 embodiment of a refractive index n ₀ extraordinary ray ordinary ray Is 660 nm for the light source
Use red LED. The above data is shown in FIG.
The wavelength is plotted on the horizontal axis, and the refractive index is plotted on the vertical axis.
When determining the difference between the refractive index of the ordinary ray and the extraordinary ray in the 0 nm, the n _e -n ₀ = 0.009. The quartz device is not an Y-axis cut, but an AT cut having a good frequency-temperature characteristic. Therefore, the difference between the refractive index of the ordinary ray and the extraordinary ray on the Y-axis indicates that AT
It is necessary to find the difference between the cut (about 35 °) quartz pieces.

The refractive index of the ordinary ray and the extraordinary ray of the AT-cut quartz plate can be obtained from FIG. 9 in which the refractive index distribution of the ordinary ray and the extraordinary ray is drawn on the YZ axis plane. X perpendicular to the paper
A circle 51 around which the refractive index distribution of the ordinary ray is drawn.
And the ellipse 52 on which the refractive index distribution of the extraordinary ray is drawn intersects the y-axis at points n _e and n ₀ , respectively, which represent the extraordinary ray refractive index and the ordinary ray refractive index on the Y axis, respectively. , The aforementioned Y
The difference between the axis of the extraordinary ray refractive index and the ordinary refractive index (n _e -n
₀ ).

Further, the AT cut surface comes at an angular position shifted about 35 ° counterclockwise from the Z axis. In the polarization observation method for a quartz blank, light passes in a direction perpendicular to this plane. Straight line 5 indicating the direction of light passing through the origin (X axis)
Each point n _e ′, n at which 3 intersects the circle 51 and the ellipse 52
₀ 'is the extraordinary ray refractive index and the ordinary ray refractive index of the AT cut quartz plate, respectively, and the length between both points is the difference between the extraordinary ray refractive index and the ordinary ray refractive index of the AT cut quartz plate (AT plate). a _{(n e '-n 0')} . First, the intersection between the ellipse and the straight line is determined, and then the intersection between the perfect circle and the straight line is determined.

The equation of the ellipse is y ² / a ² + z ² / b ² = 1 (2). Where a = n _e, a b = n _0.

The equation for the straight line of light is: z = cy (3) Here, c = tan 35 °.

[0048] When determining the intersection of the ellipse formula (2) and the linear equation (3), y = ± n e · n 0 / (n 0 2 + n e 2 · tan 2 35 °) (4) z = ± n e _{· n 0 · tan35 ° / (} n 0 2 + n e 2 · tan 2 35 °) (5) is obtained.

Similarly, in the elliptic equation (2), a = b = n
_When the intersection of the perfect circle equation to which ₀ is substituted and the linear equation (3) is obtained, y = ± n ₀ / (1 + tan ² 35 °) ^1/2 (6) z = ± n ₀ · tan 35 ° / (1 + tan ²⁾ 35 °) ^1/2 (7) is obtained.

From [0050] Equation (4) to (7), the difference between both the refractive index in a direction perpendicular to the AT plate _{(n e '-n 0')} = 0.006 (8) is obtained.

Substituting equation (8), the phase difference between interference fringes P = 2π, and the light source wavelength λ = 660 nm into equation (1), t = 660 nm / 0.006 = 110 μm (9) Is obtained. Since the sample is a biconvex convex type (biconvex type), t = 110/2 = 55 μm (10) is obtained by calculating the thickness of one side. Therefore, the change amount of the phase difference of 2π, that is, the thickness between the interference fringes is 55 μm. [2] Experimental value As shown in FIG. 10, the luminance (brightness or light amount) of the striped pattern of the sample image 61 of the rectangular crystal blank reflected on the display was measured. The measured stripe pattern is the outermost dark area 6
3 (the point at which the luminance dropped the most) and the dark portion 62 (the point at which the luminance dropped the most) one inside from the outermost. From the distance L _1, one inside of the dark part 62 from the sample center to one inside of the dark part 62 and the distance L ₂ to the outermost of the dark part 63, further the distance to the edge and the L _0. L ₀ × 2 is the long side of the sample. The same samples were bonded to the base was measured L ₁ and L between the _second height (thickness) H a pen recorder. Note that the dark portion was the portion where the luminance distribution along the center line in the longitudinal direction was the lowest. The measurement sample is the following 3
One.

[0052] (1) Sample _{A L 1 = 2.85mm L 2 =} 0.95mm L = 8.25mm H = 55μm (2) Sample _{B L 1 = 2.8mm L 2 =} 1.05mm L = 8.25mm H = 55 μm (3) Sample C L ₁ = 3.15 mm L ₂ = 0.7 mm L = 8.00 mm H = 55 μm

Each of the samples (1) to (3) described above exhibited a height H between the dark portions of 55 μm, which is the same as the theoretical value. Here we used an adhesive to fix the sample to the base,
In some cases, the horizontality of the sample could not be ensured due to the variation in the applied thickness of the adhesive. However, since the relative height between the dark portions was determined, the horizontality of the sample did not become an error. The surface roughness of the sample surface was measured in order to examine how much error was included in the aforementioned 55 μm. As shown in FIG. 11, since the half of the peak-to-peak was about 1 μm, it was found that there was no problem if the accuracy of the thickness was about 10%. Therefore, it was confirmed that the stripe pattern had a correlation with the thickness of the sample.

[Inference] The above theoretical values and experimental values were applied to a large sample having a long side of the order of 8 mm and a thickness of the order of 400 μm. Should be applicable. 4mm long side
As shown in FIG. 7, a small sample having a table, a short side of about 1 mm, and a thickness of about 100 μm to 200 μm has a stripe pattern (dark area) of 1 μm.
Only one occurs. Therefore, the thickness cannot be measured based on the above results. However, since it was found that there was a correlation between the sample thickness and the luminance appearing in the sample image, the luminance was observed when only one stripe pattern was generated by the polarization observation method, or even when no clear stripe pattern was generated. If there is a change, set an optimal threshold value and draw an isoluminance line (corresponding to a contour line of mountainous terrain) connecting points with the same intensity of interference fringes, so that the isoluminance line is the same You can show the thickness. FIG. 12 is a sample diagram depicting such an equal luminance line. These samples are about 153 μm thick,
The blank frequency is 10 MHz.

(A) Sample (FIG. 12 (a)) Long side = 04.324 mm Short side = 0.807 mm Equal luminance line shape: beautiful ellipse (b) Sample (FIG. 12 (b)) Long side = 0. 328 mm short side = 01.807 mm isoluminance line shape: slightly distorted ellipse (c) sample (Fig. 12 (c)) long side = 04.332 mm short side = 01.806 mm isoluminance line shape: greatly distorted ellipse

When the above-described strip-shaped convex oscillators (a) to (c) were assembled and subjected to an oscillation test, (a)
The samples (b) and (b) had good and stable frequency characteristics, obtained a high Q value, had good temperature characteristics, and were good. On the other hand, (c) was inferior due to unstable frequency characteristics and poor temperature characteristics. From these results, it was found that the polarization observation method is effective for thickness measurement and shape measurement even for a small sample.

In the above-described embodiment, the quality of the convex processing is determined based on the shape of the isoluminance line when the crystal blank image is viewed from above. There are also requests to determine the quality. Rather, since the frequency characteristics have a strong correlation with the cross-sectional shape, this requirement is large.

Next, the relationship between the polarization observation method and the cross-sectional shape of the quartz crystal blank will be examined. As described above, if there is a change in luminance in the polarization observation method, draw an isoluminance line connecting points of equal luminance by setting an optimal threshold value.
The same thickness is shown on the isoluminance line. If this is further advanced, if the brightness value is continuously plotted in the length direction or the width direction of the crystal blank, the change in the brightness indicates the change in the thickness itself in the same cross section. This luminance change is shown in FIGS. 5B to 7B as a luminance distribution diagram along the center line. This luminance distribution can be expressed by the sin function (or cos function) of the phase difference P according to Marieu's law, and it can be seen from the equation (1) that the phase difference P is proportional to the thickness t. 5 (b) to FIG.
The luminance distribution in (b) can be represented by a sin function.
Similarly, FIG. 7 shows an example of a small and thin sample in which only one stripe pattern is generated or no clear stripe pattern is generated.
The luminance distribution of (b) can also be represented by a sin function.

From this luminance distribution, the thickness of the quartz blank in the same section can be inspected. This will be described. 13 and 14 show the relationship between the cross section of the crystal blank and the phase of the luminance distribution waveform. Here, the attenuation of transmission is ignored for simplicity. FIG. 13 shows a case of beveling processing polished at a fixed angle, and the phases of the waveforms are at equal intervals. When the thickness is constant, the brightness is flat without any change. On the other hand, FIG. 14 shows a case of the convex processing in which the polishing amount decreases as approaching the center, and the phase of the luminance waveform gradually increases as approaching the center. As described above, the relationship between the phase and the luminance has a correlation with the equation (1).

As shown by the solid line in FIG. 14, if the luminance distribution in a plurality of periods can be observed, the thickness between the peaks (black points) of the distribution waveform becomes 55 μm.
If a method of plotting so as to increase the thickness by m and extrapolating the plot points is used, the cross-sectional shape of the crystal blank can be easily drawn. At this time, if the valleys (white points) of the distribution waveform are also plotted, the gaps between the black points can be complemented, so that the accuracy of extrapolation increases and a more correct cross-sectional shape can be obtained.

As described above, if the luminance distribution waveform can be observed for a plurality of cycles, the sectional shape can be easily drawn.
The above method cannot be used when the brightness distribution waveform cannot be observed even in one cycle as shown by the dotted line because the crystal blank is thin. Therefore, a new method has been developed to obtain the cross-sectional shape even from the luminance distribution waveform indicated by the dotted line. This will be described below.

FIG. 15 shows three types of small and thin quartz blanks (long side of 5.0 mm or less) having different convex finishes.
4.0mm, short side 1.5mm ~ 2.5mm, thickness 100
FIG. 3 shows a qualitative polishing amount distribution diagram (μm to 200 μm). The horizontal axis indicates the length of the quartz blank, and the vertical axis indicates the polishing amount. The upper half of the left side of the center of the double convex quartz crystal blank having a thicker center is shown. In the figure, SS is a standard sample, US is an undercut sample, and OS is an overcut sample. It is defective if the polishing amount is shifted from the standard sample SS and shifts largely toward the undercut sample US, or shifts too much to the sample OS side. The non-defective product includes the standard sample SS and has an allowable range in the vicinity thereof.

FIG. 16 shows a luminance distribution diagram obtained by the polarization observation method corresponding to the polishing amount distribution of these three types of quartz blanks. Since the standard sample of the quartz blank is small and thin, the luminance distribution does not have one wavelength. When adjusted so that the maximum peak of luminance comes to the center during observation, the standard sample SS drops from the center to the left end by the sin function, passes through the minimum peak, rises and cuts before reaching the maximum peak again. Is done. This is because there is no crystal blank. At this cutting point, the values of other samples similarly increase continuously or discontinuously with the sin function, but this is probably because the difference is emphasized by the lens effect as described above, rather than by the phase difference. The undercut sample US has a longer cycle than the standard sample SS, and the luminance gradually decreases from the center toward the left end. Over-sharpened sample O
Conversely, S has a longer period than the standard sample SS, and the maximum peak at the center is smaller.

As described above, since the difference in luminance distribution between the standard sample SS and the undercut sample US or the overcut sample OS clearly appears, the deviation of the detected luminance distribution of the crystal blank to be inspected from the luminance distribution of the standard sample SS is known. If this is the case, the quality of the cross-sectional shape of the convex processing can be determined.

Next, a procedure for inspecting a deviation of the detected luminance distribution from the luminance distribution of the standard sample SS will be described.

Here, the operations common to the above-mentioned processing finish inspection from the shape of the equal luminance line are as follows. FIG.
In the above, the light from the light source 26 is polarized by a polarizer, applied to a quartz blank 1 made of a convexly-reflected birefringent medium, and passed through the quartz blank 1 so as to pass light in a polarization state forming a predetermined angle with respect to the polarized light. . As a result, as shown in FIG. 3, the phase difference due to the difference in thickness can be visualized with light-dark contrast due to the polarization interference, and the contrast image of the crystal blank 1 of the polarized light that has passed therethrough is taken as a video image, and the image processing device 14 is used. , And statistically processes the binarized data in order to perform contrast enhancement processing and further obtain data for processing finish inspection.

When inspecting the finish of processing,
The light amount of the light source 26 and the rotation angles of the polarizer 21, the analyzer 22, and the mounting table 2 are adjusted with each other in advance so that interference fringes appearing on the image are optimal for the processing finish inspection.
These light amounts and angles are adjusted.

In FIG. 17, first, in order to calibrate the inspection apparatus, a standard sample of a quartz blank is mounted on the mounting table 2 (step 201). Here, the standard sample is measured with a micrometer and has a thickness that oscillates at a desired frequency. The analyzer 22 or the polarizer 21 is rotated about the optical axis so that the central portion of the standard sample in the same cross section has the maximum luminance, and the luminance peak value of the central portion at that time is obtained (steps 202 to 202). 204). After the acquisition, the half-wave plate 24 is inserted at an angle of 22.5 ° with respect to the optical axis to shift the phase difference of the light by 45 °, and the luminance value at the center at that time is acquired (Step 206). After the acquisition, the half-wave plate 24 is extracted (step 207).

In the luminance distribution diagram of the standard sample shown in FIG. 16, the luminance peak value at the center corresponds to point A. Also 45
Due to the shift, the luminance at the center decreases, but this point corresponds to point B, which is exactly 45 ° out of phase in the luminance distribution diagram of the standard sample. The reason why the angle is set to 45 ° is to enable inspection even when the luminance distribution does not correspond to one wavelength of sin. A sine waveform passing through the two points of the maximum value of the luminance at the center and the value of the phase difference of 45 ° is obtained (step 208) and displayed on the display (step 209). This is nothing but finding the sine function of the standard sample in FIG. This sin waveform is regarded as a reference luminance distribution waveform in the same cross section of the standard sample obtained by visualizing the phase difference. In this way, it is more accurate and easier to calculate the ideal reference luminance distribution waveform using a calculation formula than the raw data obtained from the actual luminance distribution, rather than using a standard sample that is expected to vary. This is because the reference value can have versatility. Note that the reference luminance distribution waveform may be used as raw data.

Next, the crystal blank to be inspected is placed on the mounting table 2 to inspect the crystal blank to be inspected (step 210). The inspection target crystal blank luminance distribution is determined along the center line, and the resulting inspection target crystal blank luminance distribution waveform is displayed (step 211). By comparing the displayed detected luminance distribution waveform with the reference luminance distribution waveform, the thickness of the quartz crystal blank in the same cross section is relatively inspected to determine pass / fail (step 212). As the inspection content, when the deviation of the detected luminance distribution waveform from the reference luminance distribution waveform exceeds the allowable value, it is determined to be defective, and when the deviation is within the allowable value, it is determined to be good. Or, if you collect convex processing data,
It is also possible to obtain a normal distribution characteristic by obtaining the above-mentioned shift amount. This makes it possible to know the relative shape of the thickness of the crystal blank to be inspected at the same cross section. Since the brightness distribution waveforms themselves are compared, the overall shape of the brightness distribution waveform, the brightness on the vertical axis, and the phase on the horizontal axis can be known.

As described above, since the crystal blank image can be determined not from above but from the cross section, it is possible to judge how much of the crystal blank has been cut, so that the quality of the convex processing can be accurately inspected. In addition, since a sin function serving as a reference luminance distribution is obtained from two luminance values having a phase difference of 45 °, even a small and thin quartz blank which does not show a luminance distribution of one wavelength or a half wavelength is suitable. Inspection of the cross-sectional shape becomes possible.
This is particularly significant for a crystal resonator that is further miniaturized and thinned.

In addition, the finished state of the convexity processing can be inspected completely by using image processing, it can be grasped instantaneously in real time, and a large amount of data can be statistically processed, so that the quality control of the crystal blank can be performed. In particular, by making use of convex type processing finish data at the time of polarization observation, it is possible to standardize convex processing, and thus to select GO / NO GO with high reliability. In addition, since statistical methods can be used, even if the flaw detection, convex processing, beveling processing and subsequent acid treatment vary,
Since the optimum threshold value can be determined, it is possible to flexibly take the allowable value, and by setting such an allowable value as the inspection standard, it is possible to make a good product a defective product or a defective product a good product. Absent.

Although the image is monochrome in the embodiment, the image may be color. Although monochromatic light of a specific wavelength has been used as the light source for polarization observation, a white light source can be used in principle. In addition, a 660 nm light source for polarized light observation is used.
However, the present invention is not limited to this. It is preferable to use light having a shorter wavelength than this in order to improve measurement accuracy because more interference fringes appear.

In the embodiment, it is assumed that interference fringes are generated. However, since there is a correlation between luminance (light quantity, brightness) and thickness, it is sufficient that a luminance distribution is generated, and the luminance distribution is displayed. Then, by comparing with the standard distribution, it is possible to inspect the processing finish of the cross-sectional shape of the convex type crystal blank.

If the wavelength of the light source is short, such as a laser, or if the wavelength is multiplied to obtain a short wavelength, a large number of interference fringes can be generated, so that the cross-sectional shape can be easily inspected. However, since both the laser light source and the multiplier are expensive, it is not practical to inspect an inexpensive crystal blank. In this respect, the device of the present invention is very practical because the cross-sectional shape can be measured only by producing light and dark using an inexpensive LED.

[0076]

According to the method of the present invention, it is possible to perform a convex type cross-sectional shape inspection by a polarization observation method.
Inspection data obtained by the polarization observation method can be used as a control index for convex processing.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of an inspection apparatus for carrying out a convex inspection method for a transparent substrate according to an embodiment.

FIGS. 2A and 2B are explanatory diagrams of a case optimal for the inspection method of the embodiment, in which FIG. 2A is a cross-sectional view of a convex type crystal blank, and FIG.

FIGS. 3A and 3B are explanatory views by polarization observation, in which FIG. 3A is a binary image, and FIG. 3B is a cross-sectional view illustrating the thickness of a quartz blank.

4A and 4B show images of a convex type crystal blank before binarization, wherein FIG. 4A shows an image of the crystal blank observed by the polarization observation method of the embodiment, and FIG. This is the blank image.

5A and 5B are explanatory diagrams showing an image of a stripe pattern and luminance of a large sample, wherein FIG. 5A is a plan view and FIG. 5B is a luminance distribution diagram along a center line.

6A and 6B are explanatory diagrams showing an image of a stripe pattern and luminance of a large sample, wherein FIG. 6A is a plan view and FIG. 6B is a luminance distribution diagram along a center line.

7A and 7B are explanatory diagrams showing an image of a striped pattern and luminance of a small sample, wherein FIG. 7A is a plan view and FIG. 7B is a luminance distribution diagram along a center line.

FIG. 8 is a plot diagram showing a Y-axis refractive index of a quartz crystal blank with respect to a specific wavelength.

FIG. 9 is a refractive index distribution diagram of an ordinary ray and an extraordinary ray on the YZ axis plane of the AT cut quartz plate.

FIG. 10 is an explanatory diagram showing a position where a stripe pattern of a sample image of a rectangular crystal blank reflected on a display is generated;
(A) is a plan view, and (b) is an enlarged sectional view of a main part.

FIG. 11 is an explanatory diagram showing the surface roughness of the sample surface.

FIG. 12 is a plan view showing images of three small samples depicting isoluminance lines.

FIG. 13 is a characteristic diagram showing a phase and a luminance distribution waveform when a crystal blank is beveled.

FIG. 14 is a characteristic diagram showing a phase and a luminance distribution waveform when a quartz blank is subjected to convex processing.

FIG. 15 is a view showing polishing characteristics of a sample according to the embodiment.

FIG. 16 is a luminance distribution diagram corresponding to the sample of FIG. 13;

FIG. 17 is a flowchart for inspection of a cross-sectional shape of a crystal blank to be inspected according to the embodiment.

FIG. 18 is an explanatory diagram of a cylindrical barrel method for performing convex (beveling) processing.

FIG. 19 is a characteristic diagram in which height data obtained by a conventional level measurement method is plotted in the length direction of a quartz blank.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 crystal blank (transparent substrate) 2 mounting table 12 objective lens 13 CCD camera 14 image processing device 21 polarizer 22 analyzer 23 diffuser plate 24 half-wave plate 26 light source

Continued on the front page F-term (reference) 2F065 AA30 AA52 BB22 CC00 DD06 FF01 FF04 FF51 FF61 GG07 GG22 HH09 JJ03 JJ26 LL33 LL34 LL35 LL49 PP13 QQ04 QQ13 QQ25 QQ29 QQ31 QQ41 SS04 TT03 2G0H GG05A02 JJ19 JJ21 JJ26 KK04 MM02 MM05 MM09 PP04 2H052 AA04 AC05 AF03 AF14 AF25

Claims (5)

[Claims] 1. A light from a light source is polarized by a polarizer and applied to a transparent substrate made of a convex-processed birefringent medium by a polarizer. In a method of inspecting a transparent substrate, which passes light and visualizes a phase difference due to a difference in thickness of the transparent substrate due to polarization interference, and inspects a finish of processing the transparent substrate, the phase difference is visualized. A method for inspecting the thickness of the transparent substrate at the same cross section by comparing the luminance distribution at the same cross section of the transparent substrate obtained at the same time with a reference luminance distribution obtained in advance. 2. The brightness distribution in the same cross section of the transparent substrate obtained by visualizing a phase difference caused by a difference in thickness of the transparent substrate is represented by a sin function correlated with the following equation. Inspection method for transparent substrate. _{P = (n e -n 0)} t · 2π / λ However lambda: wavelength of the light source n _0: the ordinary refractive index n _e: refractive index of extraordinary ray t: thickness of the transparent substrate P: position of interference fringes Difference 3. The method for inspecting a transparent substrate according to claim 2, wherein (1) first, a visual observation of a phase difference caused by a difference in thickness of the standard sample of the transparent substrate is performed on the same cross section of the standard sample. The analyzer or the polarizer is rotated so that the central part of the luminance distribution becomes maximum, and the maximum value of the luminance at that time is obtained. (2) The half-wave plate is set at 22.5 ° with respect to the optical axis. And the phase difference of light is shifted by 45 °, and the luminance value of the central part at that time is obtained. (3) Two points of the maximum value of the luminance of the central part and the luminance value of the phase difference of 45 ° Is obtained, and this sin waveform is regarded as a reference luminance distribution waveform in the same cross section of the standard sample obtained by visualizing the phase difference. (4) Next, the thickness of the transparent substrate to be inspected is determined. Visualize the phase difference due to the difference And (5) comparing the detected luminance distribution waveform with a reference luminance distribution waveform to inspect the thickness of the transparent substrate at the same cross section. Inspection method for transparent substrates. 4. The method for inspecting a transparent substrate according to claim 3, wherein the analyzer or the polarizer is rotated by 45 ° instead of inserting the half-wave plate of (2). Method. 5. A polarization observation light source for irradiating unpolarized light toward a transparent substrate made of a convex-processed birefringent medium, and irradiating the transparent substrate with polarized light from the polarization observation light source. A polarizer, an analyzer that detects light in a polarization state at a predetermined angle transmitted through the transparent substrate, and an image of the transparent substrate of the polarization observation light that has passed through the analyzer, and the thickness of the transparent substrate is determined by polarization interference. Imaging means for visualizing a phase difference due to the difference; image processing means for image-processing the image visualized by the imaging means to detect a luminance distribution in the same cross section; reference luminance obtained from the detected transparent substrate luminance distribution and a standard sample A transparent substrate inspection apparatus comprising: comparison means for comparing the distribution and the thickness of the transparent substrate at the same cross section.