JP2017078641A - Shape measurement method and shape measurement device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a shape measurement method capable of precisely determining and correcting a jump area even when measuring a fine shape in a lateral direction.SOLUTION: The shape measurement method includes: step S102 to acquire a piece of coordinate data by measuring probe coordinates while scanning probe with a probe in contact with the surface to be measured; step S103 to convert the coordinate data into a piece of frequency data; step S105 to set an abnormal segment which is determined on the basis of a preset frequency threshold level with respect to frequency data; step S106 to correct data included in the abnormal segment in the coordinate data; and a step to calculate measurement data which represents a shape of the surface of the object on the basis of the corrected coordinate data.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a shape measuring method for measuring the shape of an optical element or a mold with high accuracy.

  As optical products become more sophisticated, optical elements such as lenses and mirrors mounted on optical products are becoming more accurate. Such optical elements have been required to have high shape accuracy of about 100 nm, which is a fraction of the wavelength of light (human visible light is about 400 to 800 nm). In order to manufacture a highly accurate optical element, it is necessary to measure the surface shape with high accuracy. Molding is also widely performed as a method for producing a large amount at a low cost. In this case, it is required to measure a molding die with high accuracy. In order to measure the surface shape of an optical element or a molding die with high accuracy, a probe-type shape measuring apparatus is widely used. The probe-type shape measuring apparatus uses an optical element as an object to be measured, scans the surface to be measured while contacting the probe, and acquires the behavior of the probe as coordinate data. The shape data to be evaluated is extracted from the coordinate data to obtain measurement data. This method achieves nanometer-level measurement accuracy. Note that the force applied to the probe during scanning in the direction in which the probe is pressed against the surface to be measured is referred to as measurement force.

  However, it is known that the measurement accuracy deteriorates by scanning a scratch on the surface to be measured, attached dust, or the like.

  Usually, when a flaw or dust is scanned, the probe easily jumps. If it is a small jump, the measurement force applied to the probe is instantaneously pushed back to the surface to be measured to make contact with the ground, so the influence on the measurement accuracy is negligible. However, if the jump is large, the influence on the measurement accuracy cannot be ignored. In addition, when the jumping splash is large, it may jump again after touching the surface to be measured, and the jumping and touching may be repeated for a certain period. Since the section in which jumping and grounding are repeated does not represent the shape of the surface to be measured, the section in which the measurement error is large continues.

  In order to perform highly accurate measurement, a method of determining a section with a large measurement error as an abnormal section and correcting coordinate data is performed. For example, the data in the abnormal section is removed, or a label indicating that the data is in the abnormal section is attached and is not used for evaluation. Alternatively, the data in the vicinity of the abnormal section may be used to interpolate the data in the abnormal section for substitution.

  As a method for determining an abnormal section, Patent Document 1 discloses a method in which an approximate curve for probe coordinate data is created, and a section in which the height difference between the approximate curve and the coordinate data exceeds a specified threshold is determined as an abnormal section. Is described.

  The approximate curve can be created by removing high frequency components from the coordinate data. As specific methods, various methods such as digital filter processing, function approximation, and Fourier transform are conceivable. The boundary frequency to be removed as a high frequency component is generally called a cut-off frequency. By adjusting the cut-off frequency, it is possible to change how finely the coordinate data is approximated. When a high-frequency component is gradually attenuated over a certain frequency band, one representative point in the frequency band is defined as a cutoff frequency.

Assume that coordinate data having a large measurement error in a certain section is acquired. Then, it is assumed that the cut-off frequency is set to a frequency Q (= 1 / W) [Hz (= mm ⁻¹ )] corresponding to the width W [mm] of the section or higher than that. In this case, the approximate curve is greatly attracted to the measurement error. Even if an abnormal section is determined based on an approximate curve that is largely drawn by measurement errors, it is difficult to determine accurately. In other words, the cut-off frequency needs to be set lower than the frequency corresponding to the maximum width to be determined as the abnormal section.

  On the other hand, if the cut-off frequency is too low, the shape component having a fine width originally present on the surface to be measured cannot be approximated. That is, a shape component that cannot be evaluated is generated. In other words, the cutoff frequency needs to be set higher than the frequency corresponding to the minimum width to be evaluated as the shape component.

  As described above, in the prior art, an appropriate cut-off frequency is adjusted according to the maximum width to be determined as an abnormal section and the minimum width to be evaluated as a shape component. In this way, it is possible to create an approximate curve including the shape component to be evaluated while reducing the shape component of the measurement error. As described in Patent Document 1, it is possible to enhance the effect by repeating creation of an approximate curve, determination of an abnormal section, and removal.

Japanese Patent No. 3719634

As the accuracy of optical elements has increased, there has been an increasing need to evaluate finer shape components. For example, there is a need to evaluate a wavy shape with a period of 1 mm or less. Therefore, if the minimum width W1 to be evaluated as a shape component is 0.1 mm, the cutoff frequency Q needs to be higher than 10 Hz (= mm ⁻¹ ).

On the other hand, the section W2 having a large measurement error caused by the jumping of the probe may continue for about 0.2 mm. In this case, the cut-off frequency Q needs to be lower than 5 Hz (= mm ⁻¹ ).

  In the above example, the cutoff frequency is set so that the two conditions of the frequency corresponding to the minimum width to be evaluated as the shape component (10 Hz or more) and the frequency corresponding to the maximum width to be determined as the abnormal section (5 Hz or less) are satisfied simultaneously. It cannot be adjusted.

  As described above, in the conventional technique, when trying to evaluate a shape component having a finer width, it may not be possible to adjust to an appropriate cutoff frequency. If it is not possible to adjust to an appropriate cutoff frequency, the abnormal section cannot be determined accurately. Therefore, highly accurate measurement cannot be performed.

  The present invention provides a shape measurement method that can determine an abnormal section more accurately and perform highly accurate measurement even when evaluating a shape component having a fine width.

In order to achieve the above object, the invention according to the present application is a shape measuring method for measuring the shape of a surface to be measured,
Scanning while contacting a probe to the surface to be measured and measuring coordinate of the probe to obtain coordinate data;
Converting the coordinate data into frequency data;
Setting an abnormal interval determined based on a frequency threshold set for the frequency data;
Correcting the data included in the abnormal section in the coordinate data;
Calculating measurement data representing the shape of the surface to be measured based on the corrected coordinate data;
A shape measuring method comprising:

  When the probe is scanning the surface to be measured without jumping (or including a minute jump), the probe vibration frequency is the tilt angle of the surface to be measured. It was found to correlate with. In general, the inclination angle of the optical element changes gently as a whole. Therefore, for example, when attention is paid to a local section of 0.1 mm, the change in the inclination angle is minute. That is, the change in frequency is minute.

  Further, even if the surface to be measured has a wavy shape, and the shape changes in height of 100 nm to a width of 0.1 mm, the change in the tilt angle is less than 0.1 degree. If the tilt angle is less than 0.1 degree, the frequency change is very small.

  That is, even when a shape component having a fine width is evaluated, the change in the inclination angle of the local section is very small.

  On the other hand, the frequency changes even in a section with a large measurement error. For example, when the probe bounces 100 nm, the vibration frequency of the probe decreases by about 100 Hz. Such a change in frequency occurs over a section with a large measurement error.

  However, even if a section with a large measurement error continues for about 0.2 mm, the frequency change in the vicinity is gentle, so that the frequency change can be easily found.

  As described above, the present invention utilizes the characteristic that the frequency of vibration of the probe changes gently over the entire optical element when the probe is normally scanned. And since the local frequency change of the section with a large measurement error is found, the abnormal section can be determined more accurately.

  Therefore, according to the present invention, even when a shape component having a fine width is evaluated, an abnormal section can be determined more accurately and highly accurate measurement can be performed.

It is a flowchart of the shape measuring method in connection with the first embodiment of the present invention. It is a figure explaining the data processing in connection with 1st Example of this invention. It is a figure explaining the shape measuring apparatus in connection with this invention. It is a figure explaining the behavior of the probe under measurement. It is a figure explaining the result measured using a first embodiment of the present invention.

  Hereinafter, a first embodiment according to the present invention will be described with reference to the drawings.

  Note that the approximate curve described below is similarly effective even if it is replaced with an approximate line. When the data to be approximated changes linearly, it is desirable to approximate with a straight line.

  First, paying attention to the coordinate data acquired on the surface of an optical element such as a lens, it was confirmed that vibration with an amplitude of about several tens of nanometers occurred when the probe scanned along the surface to be measured. . It was also found that the main frequency of this vibration continuously changes according to the inclination angle of the surface to be measured.

  As described above, the probe jumps temporarily. It has been found that this jump is a kind of vibration, and the frequency at which it vibrates decreases relative to the transition frequency.

  First, the procedure of the shape measuring method according to the present invention will be described.

  The present invention is a shape measuring method for measuring the shape of a surface to be measured, and acquires coordinate data by scanning while contacting a probe to the surface to be measured and measuring the position of the probe. Then, measurement data representing the shape of the surface to be measured is calculated based on the coordinate data.

  The shape measuring apparatus used for executing the shape measuring method is as follows.

  The movable stage includes a movable probe stage 62, a probe 61 held on the probe stage via a housing 65 and a leaf spring 64 (spring member), a work stage 68, and a calculation unit 66 as a control unit. . In acquiring coordinate data, position information is acquired at a specific sampling rate, and the position information is recorded as point sequence data. The configuration of the shape measuring apparatus will be described in more detail later.

  As will be described in more detail in the following embodiments, the present invention is characterized by a series of processing of acquired coordinate data.

(First Example)
FIG. 1 is a flowchart of a shape measuring method according to the first embodiment of the present invention. FIG. 2 is a diagram for explaining data processing according to the first embodiment of the present invention. The shape measuring method will be described with reference to FIGS.

  In S101, measurement is started. This step includes preparation for measuring the shape of the surface to be measured, such as placing the surface to be measured on the work stage of the shape measuring device and performing alignment of the shape measuring device.

  In S102, coordinate data is acquired. In this step, the coordinate data is acquired by scanning while making the probe contact the surface to be measured and measuring the coordinates of the probe. Hereinafter, coordinate data is represented as M and design shape D. The coordinate data M minus the design shape D, that is, the residual MD may be treated as coordinate data again. This residual MD indicates how far the measured surface shape is deviated from the design shape D, which is the target value of the shape of the measured surface. Therefore, the shape of the surface to be measured is reflected in the same manner as the coordinate data M itself.

  In S103, the acquired coordinate data is converted into frequency data. FIG. 2A shows the coordinate data and the waviness shape. X is a lateral position of the rear end of the probe, and is about 0.1 mm from the origin of the X axis to the right end. Z represents the height of the surface to be measured, and the distance from the origin of the Z axis to the upper end is about 100 nm. The wavy shape 20 of the surface to be measured is a shape in which the height varies by 100 nm with a lateral width of 0.1 mm. On the other hand, the coordinate data 21 vibrates with a width that is finer than a width of 0.1 mm and changes in height by several hundred nm. This vibration is derived from the vibration of the probe while scanning the surface to be measured as described above.

  FIG. 2B shows an approximate curve of coordinate data. An approximate curve 23 approximates the coordinate data 21. The coordinate data 21 includes a region where the probe jumps and the amplitude increases (hereinafter, jump region 22). For this reason, in the vicinity of the jumping region 22, the approximate curve 23 is attracted to the large amplitude of the coordinate data 21, and is biased upward from the waviness shape 20.

  As a specific process in S103, first, an extreme value in the coordinate data 21 is extracted as extreme value data 24. As a specific method of extracting extreme values, the inclination at each point of the coordinate data 21 is calculated, and the point where the sign of the inclination has changed is extracted as extreme value data 24.

For example, when the point i constituting the coordinate data 21 is Z _i at the position of X _i , paying attention to the point i + 1 adjacent to the point i, (Z _{i + 1} −Z _i ) / (X _{i + 1} − X _i) [Equation 1]
Is calculated for all points (i = 1, 2,..., N), and the points where a change in sign is seen may be extracted. Assuming that the slope value belongs to the younger one (point i) of the points i and i + 1, the slope is calculated for the entire measurement data and the slope of each point is calculated. Of course, the inclination value may be unified so as to belong to either the point i or the point i + 1, or may be newly organized and recorded as a data point sequence of inclination instead of as data belonging to each point. Further, instead of the point i + 1, a point i + g (an integer where g ≠ 0) may be used.

  Next, at each point of the extracted extreme value data 24, the time between the extreme values is obtained from the extreme value interval 25, the sampling frequency, and the probe scanning speed. Here, the interval between the maximum value and the adjacent maximum value was used.

  In FIG. 2B, the interval 25 between the extreme values is shown as the interval between the maximum value and the maximum value in the X direction, but the interval between the minimum value and the minimum value or the interval between the maximum value and the minimum value is doubled. May be used.

Next, the reciprocal of the time between extreme values is calculated, and this is used as the frequency of vibration. Since the sampling rate of measurement is known, the time t between the extreme value E _k and the extreme value E _{k + 1} can be calculated. Therefore, the reciprocal 1 / t is calculated as the vibration frequency [Hz]. It is assumed that the vibration frequency belongs to the younger one (Ek) of the extreme value E _k and the extreme value E _{k + 1} . Similar to the above-described inclination data, the data attribution and organization method may be a desired method.

  Hereinafter, data representing the frequency at each point of the extreme value data 24 will be described as frequency data.

  Further, data closest to the z value between the maximum value and the minimum value may be extracted. When the intermediate z value is extracted, data can be extracted at a location where the gradient is large, so that it can be accurately extracted. In the step of extracting the extreme value data 24, if a low pass filter is applied in advance to the coordinate data 21, the extreme value can be extracted by removing white noise having a frequency higher than that of the probe vibration. Further, by using the extreme value of the spline curve obtained from the coordinate data, the position of the extreme value data in the X direction may be determined with a resolution higher than the interval of the coordinate data.

  In this way, the coordinate data is converted into frequency data.

  Subsequently, as described below, an abnormal section determined for the frequency threshold set for the frequency data is set.

  In S104, an approximate curve 27 of frequency data is obtained (FIG. 2 (c)).

  In S105, an area where the frequency data is away from the approximate curve 27 by a predetermined threshold or more is set as an abnormal section.

  As an example of a threshold setting method, the threshold value may be the reciprocal of the probe dwell time when the probe's measuring force F, the probe mass m, and the probe bounces in the Z direction by an amplitude h.

More specifically, since the probe is controlled so as to come into contact with the surface to be measured with the same measurement force as much as possible during measurement, it can be considered as a simple physical model similar to throwing up an object. That is, the equation of motion m (d / dt) ² z = −F can be expressed, and the initial values of z and d / dtz can be solved as ₀ and v ₀ , respectively, and the relationship between z and t can be uniquely obtained ( z = − (F / 2m) t ² + v ₀ t).

From this equation, the time from the moment of jumping at t = 0 to the time when the probe reaches the highest point is t ₀ = mv ₀ / F, and the height at the highest point is h = mv ₀ ² / (2F). Accordingly, the time t ₁ = 2t ₀ = (8 hm / F) ^1/2 that has elapsed until the probe that has jumped back to the original position is obtained. Since the measurement force F of the probe, the probe mass m, and the probe have an amplitude h in the Z direction, the amplitude h can be considered as a measurement error that can be accepted by the user.

For this reason, the value 1 / t ₁ = (F / 8 hm) ^1/2 should be used as a threshold for identifying an abnormal interval from the above-described frequency data, or as a reference for determining the threshold. Can do. Furthermore, the dwell time may be calculated from a model that takes into account the air resistance and the inclination angle of the surface to be measured.

  Further, the abnormal section is defined from the position in the X direction where the vibration frequency deviates from the approximate curve to a threshold value or more to the position in the X direction of the adjacent frequency data.

  FIG. 2C illustrates the relationship between the frequency data obtained by conversion from the coordinate data and the abnormal section. An approximate curve 27 is obtained for the frequency data 26. As shown in FIG. 2C, when the coordinate data is converted into frequency data, the change in the frequency data with respect to X becomes more gradual than the change in the coordinate data with respect to X. Although the details of this reason will be described later, it is caused by the fact that the vibration frequency of the probe changes according to the inclination angle of the surface to be measured. Since the change is gradual, the frequency data can be accurately approximated by a low-order function.

  Although the threshold value 29 may have various setting methods, as an example, a threshold value obtained by offsetting the approximate curve 27 may be set. If the appropriate threshold value is known in advance, it may be set directly without being associated with the approximate curve 27.

  The abnormal section 28 is a section including all data deviating from the threshold value 29.

  As described above, when the threshold value 29 is set for the approximate curve 27 for the frequency data, the data can be easily and accurately selected. Of course, the setting method of the threshold value 29 is not limited to this, and any frequency threshold value set for the frequency data may be used.

  In S106, the coordinate data of the abnormal section is corrected, and measurement data representing the shape of the surface to be measured is calculated.

  FIG. 2D is a diagram for explaining the corrected coordinate data. In the corrected coordinate data 21 ', the data of the abnormal section is removed. As a result, the degree of the approximate curve 23 ′ of the coordinate data after correction is attracted to the frequency data of the jumping region is reduced, and the corrected undulation shape 20 ′ can be expressed well. That is, the shape of the measured surface could be calculated.

  In addition, although the example which removed the data of the abnormal area was shown, you may interpolate by the nearby coordinate data.

  In S107, the measurement ends.

  In the method described above, the approximate curve is generated after the coordinate data is converted into frequency data. Although the coordinate data is converted to a temporal frequency, the same effect can be obtained even if correction processing is performed after converting the coordinate data to a spatial frequency.

  As a result of intensive studies on the relationship between the vibration of the probe and the tilt angle of the surface to be measured, the following was found for this coordinate data.

  That is, it has been found that when the probe scans normally following the surface to be measured, the vibration frequency of the probe depends on the inclination angle of the surface to be measured. If the tilt angle of the surface to be measured changes, the vibration frequency also changes according to the tilt angle.

  Even if the inclination angle of the surface to be measured includes a fine waviness shape, the change is minute. For example, a wavy shape of 0.1 mm width and 100 nm has a great influence on the optical characteristics of an optical element such as a lens. However, this swell is approximately 0.06 degrees (= 0.1 / 100 = 0.001 [rad]) when converted to an inclination angle. If it is about 0.06 degree, the change of the frequency of vibration is small enough.

  Therefore, the vibration frequency of the probe changes according to the approximate shape of the surface to be measured. In the conventional method of approximating coordinate data, the coordinate data changes according to the shape. For this reason, it is necessary to express coordinate data with a fine resolution in the X direction. On the other hand, in this embodiment, frequency data changes according to a rough shape. For this reason, the transition of the frequency data can be accurately represented even with a coarse resolution in the X direction.

  Since the approximation is performed with a low resolution, an approximation curve can be drawn with only a slight influence of the frequency change caused by jumping. For this reason, it is possible to accurately obtain an approximate curve representing a state where scanning is possible without jumping.

  Therefore, the abnormal section can be effectively identified by converting the coordinate data into frequency data and processing the data in the frequency space instead of processing the coordinate data itself. In other words, since the approximation accuracy of the frequency data is high, it is possible to accurately determine the relative decrease in the frequency in the jumping region. Since the jumping and splashing area can be determined more accurately, highly accurate measurement can be performed.

  In the above-described method, the horizontal axis is described as the position X. However, if the correspondence with the frequency data such as the inclination angle (deg) and the measurement time (t) can be expressed, the frequency data can be arbitrarily converted. May be. The order in which data is acquired may be plotted on the horizontal axis.

  If the transition of the frequency data has a linear tendency, it is preferable to approximate it with a straight line.

  Although described using frequency, a period may be used.

  FIG. 3 is a diagram for explaining an example of a shape measuring apparatus according to the present invention. The probe 61 is held by a leaf spring 64 so as to be movable in the Z direction. The leaf spring 64 is attached to the housing 65. The housing 65 is fixed to the probe stage 62. The probe stage 62 is attached to the scanning stage 63 and is driven in two axes in the XY direction or one of them. The object to be measured 67 is mounted on the work stage 68. More preferably, the work stage is placed on a slow shaking table or the like to reduce vibration from the floor. In addition, an example in which a laser length measuring device L is mounted is illustrated as an example of a length measuring device.

  The position information in the Z direction of the rear end of the probe and the position information in the Z direction of the probe stage are measured by a laser length measuring device L or the like, and are taken into a calculation unit 66 having a memory and a CPU. The calculation unit 66 calculates measurement data representing the shape of the surface to be measured based on the behavior of the probe, and stores it in the memory. The memory may be provided in the apparatus or may be externally connected via a computer network. In executing the above-described shape measuring method of the present invention, a program for causing a computer to execute each process is stored in a memory. In processing the coordinate data acquired by the shape measurement and calculating the shape of the surface to be measured, the program is executed, and the shape of the surface to be measured is calculated based on the corrected coordinate data.

  That is, the shape measuring apparatus of the present invention is a shape measuring apparatus including a movable probe stage, a probe supported by the probe stage, a work stage, and a control unit. And it is a shape measuring apparatus which equips the control part with the program which performs each process of the above-mentioned shape measuring method so that execution is possible. Having the program executable in the control unit is not limited to providing all the programs offline in the control unit. That is, it includes a case where an external computer connected online has a part of the program, and the shape measuring apparatus can execute the program when performing the measurement.

  In addition, the measurement data is displayed on a monitor or printed on paper as necessary.

  Further, the computing unit calculates a deviation between the position information of the probe 61 in the Z direction and the position information of the probe stage 62 in the Z direction, and sends a drive signal to the probe stage so that the deviation becomes constant. The probe stage is driven in the Z direction based on the drive signal. When the deviation becomes constant, the amount of deflection of the plate spring 64 of the shape measuring probe becomes constant, and the probe and the surface to be measured come into contact with each other at a constant pressure.

  In the threshold value setting method described in S105 described in the first embodiment, a setting method based on a physical model of the probe motion is shown. Instead, a method based on statistical information of coordinate data or preliminary measured data is also possible. Hereinafter, these methods will be described.

(Second embodiment)
A method for setting the threshold based on the statistical information of the coordinate data will be described.

  As will be described in detail below, the method of this embodiment is effective when the residual data in the frequency space follows a normal distribution.

  The processes from S101 to S104 are performed in the same manner as in the first embodiment, and it is assumed that the frequency data 26 and the approximate curve 27 in FIG.

First, based on the approximate curve 27 the frequency data 26, the value _{A i} in the approximation curve 27 at the coordinates _{X i,} the value _{E i} of the frequency data 26 in same coordinates _{X i,} and calculates the difference _A i -E _i. Since it is only necessary to obtain the difference from one to the other, either may be subtracted.

  Hereinafter, this difference is referred to as residual data in the frequency space. The residual data in the frequency space is obtained by removing a gradual change with respect to the position X of the frequency data. Next, the standard deviation s of the residual data in the frequency space is calculated, and the value of the standard deviation s is set as a threshold for determining the abnormal section. In calculating the standard deviation, the residual data may be divided for each section in the X direction and calculated for each section.

  Assuming that the variation in the Z direction of the residual data in the frequency space follows a normal distribution by setting the threshold values as described above, 15.87% of the entire data is determined to be noise as can be seen by referring to the normal distribution table. .

  In the above description, the standard deviation is set as a threshold value, but the ratio of data determined as noise can be changed by setting the standard deviation multiplied by a predetermined coefficient as the threshold value. That is, the threshold value can be set according to the ratio of the number of data points to the entire coordinate data of the data to be determined as noise.

  As described above, the approximate curve for the frequency data is calculated, the standard deviation of the residual data, which is the difference between the approximate curve and the frequency data, is calculated, and the frequency threshold based on the standard deviation is set.

  This method can be executed immediately in the case where the residual data in the frequency space follows a normal distribution, and it is easy to identify an abnormal interval derived from the jumping that is noise. That is, it is possible to easily specify the coordinate data to be corrected.

(Third embodiment)
A method for setting a threshold based on preliminary measured data will be described. In this method, a sample having a surface different from the surface to be measured whose shape is to be measured (hereinafter referred to as a standard sample) is prepared. By performing the steps S101 to S104 a plurality of times on the standard sample, an approximate curve for each measurement is obtained. Next, the standard deviation of the approximate curve for each measurement is calculated, and the standard deviation is set as a threshold value. When the approximate curve is a straight line, the standard deviation of the intercept for each measurement is calculated. As described above, an appropriate threshold value can be set in consideration of a plurality of actually measured data.

  That is, in calculating an approximate curve for frequency data, coordinate data obtained by measuring a standard sample is used.

  Since the frequency of vibration changes with respect to the inclination angle of the surface to be measured, an approximate curve used when measuring the surface to be measured can be obtained using the approximate curve obtained with the standard sample. For example, an approximate curve with the horizontal axis as the tilt angle is calculated from the relationship between the approximate curve obtained with the standard sample and the tilt angle of the standard sample. This can be combined with information on the tilt angle of the surface to be measured to obtain the frequency change of the surface to be measured.

  By using actual measurement data of a standard sample with a small measurement error, an approximate curve with reduced influence of the measurement error can be acquired. For this reason, even when the measurement error is large due to the influence of the material of the surface to be measured, a highly accurate approximate curve can be used for discrimination. Therefore, it is possible to more accurately determine the relative decrease in frequency.

(Probe behavior and vibration mechanism)
The behavior of the probe during measurement will be described. FIG. 4 is a diagram for explaining the behavior of the probe during measurement. In FIGS. 4A and 4B, the probe at each time is drawn, but the one in the + X direction is the one when the time has advanced.

  In a place where jumping does not occur, there are three main spring elements related to the probe under measurement: the elasticity of the leaf spring, the bending rigidity of the probe itself, and the contact rigidity between the probe and the surface to be measured. It was found that when an external force such as a collision with the surface to be measured due to movement in the XY direction or external vibration is applied to this system, the probe resonates using the force of the three spring elements as a restoring force. Hereinafter, this vibration is referred to as flexural vibration. In flexural vibration, the amount of deflection of the probe in the XY direction changes, so that the probe rear end is displaced in the Z direction along the surface to be measured. For this reason, as shown in FIG. 4A, the trajectory 33 at the rear end of the probe is obtained by adding displacement in the moving direction 32 and Z displacement due to flexural vibration. The displacement 34 shown in FIG. 4C is displayed by removing the tilt component from the trajectory of the probe rear end. One period 35 of the flexural vibration indicates one period of the vibration motion indicated by the displacement 34.

  The frequency of the flexural vibration is higher than the frequency band of the shape of the surface to be measured. According to the examination, the frequency of the flexural vibration is 200 to 600 Hz, which is higher than the frequency of the shape (about several tens of Hz).

  A characteristic of the flexural vibration is that the amplitude is equal to or less than the pushing amount because the probe is always in contact with the surface to be measured. Of the three spring elements described above, the direction of the contact stiffness changes depending on the inclination angle, and thus it has been found that the vibration frequency of the probe during shape measurement corresponds to the inclination angle of the surface to be measured. The inclination angle is only about 0.06 degrees in a fine waviness shape such as the waviness of about 100 nm with respect to the width of about 0.1 mm as described above.

  From such a slight change in the tilt angle, the frequency of vibration is affected only minutely.

  Third, the vibration does not depend on the amplitude because it vibrates at a specific frequency composed of the three spring elements.

  As described above, the probe performs a flexural vibration at a frequency corresponding to the inclination angle of the surface to be measured at a location where no jump splash occurs.

  On the other hand, the behavior of the probe at the location where the jumping occurs will be described. When the probe collides with a fine scratch on the surface to be measured, it jumps. Then, among the three spring elements that generate the restoring force of the flexural vibration, the contact rigidity is released. Therefore, while the probe is away from the surface to be measured, vibration is performed using the force of the two spring elements of the leaf spring and the bending rigidity as a restoring force. After that, contact with the surface to be measured and repeat the flexural vibration and jumping until it is sufficiently damped. The above movement is hereinafter referred to as “jumping vibration”. In the jumping and bending vibration, one cycle is a period from when the probe jumps to the ground and then jumps again. As shown in FIG. 4B, the trajectory 303 at the rear end of the probe is obtained by adding displacement due to jumping to Z displacement due to displacement in the moving direction 32 and flexural vibration. A displacement 304 in FIG. 4D is displayed by removing the tilt component from the probe rear end locus. As shown in FIG. 4D, one period 305 of the jumping and bending vibration is the sum of the probe dwell time 307 and the time 308 during which the bending vibration is performed. Among these, during the time 308 when the flexural vibration is performed, the vibration is caused under the influence of the contact stiffness, whereas during the dwell time 307, the contact stiffness is released. For this reason, during the dwell time 307, the restoring force in the Z direction is reduced.

  The characteristic of the jumping and bending vibration is that the amplitude is larger than the amplitude of the bending vibration because the probe and the surface to be measured are not in contact with each other. Further, since the restoring force is reduced by releasing from the contact rigidity, the frequency becomes lower than the flexural vibration. Therefore, it becomes lower than the frequency according to the inclination angle.

  Therefore, as described above with reference to the first to third embodiments, the vibration frequency is significantly lower than the frequency corresponding to the inclination angle, particularly in the abnormal section in FIG. Therefore, by adopting the method exemplified in the present embodiment, it is possible to specify the abnormal section to be corrected with high accuracy. As a result, highly reliable measurement can be realized by executing the present invention.

(Experimental example)
Hereinafter, the results measured using the first embodiment of the present invention will be described with reference to the drawings. FIG. 5 is a diagram for explaining coordinate data measured by executing the shape measuring method according to the first embodiment of the present invention. In FIGS. 5A to 5C, scales in the horizontal (X) direction are aligned. In FIGS. 5A and 5C, the scales in the vertical (Z) direction are aligned.

  FIG. 5A shows the coordinate data 70. The coordinate data 70 includes data 72 having a large measurement error. For this reason, the filter curve 71 obtained by smoothing the coordinate data 70 has a deviation in the + Z direction in the vicinity of the data 72 having a large measurement error.

  FIG. 5B shows frequency data 76 obtained by converting the coordinate data 70 and an approximate curve 75 thereof. In the vicinity of the data 72 having a large measurement error, the frequency is greatly reduced. For this reason, in the vicinity of the data 72 having a large measurement error, the frequency data 76 is set as an abnormal section separated from the approximate curve 75 by a predetermined threshold or more.

  FIG. 5C shows coordinate data 73 from which data in the abnormal section is removed. Data in the vicinity of the data 72 having a large measurement error is removed. Further, it can be confirmed that the bias in the + Z direction in the vicinity of data having a large measurement error is reduced in the filter curve 74 as compared with the filter curve 71. Measurement errors can be reduced, and highly accurate measurements can be performed.

20 Waveform 21 Coordinate Data 22 Jump Area 23 Shape Approximate Curve 24 Extremum Data 25 Extremum Interval 26 Frequency Data 27 Frequency Data Approximate Curve 28 Abnormal Section 29 Threshold 30 Probe 31 Measurement Surface 32 Probe Movement Direction 33 Probe Displacement Trajectory 34 Probe Displacement Trajectory (Inclined Component Elimination) 35 Flexural Vibration Period 305 Jump Splash Bending Vibration Period 307 Dwell Time 308 Time for Performing Flexural Vibration 61 Probe 62 Probe Stage 63 Scanning Stage 64 Plate Spring 65 Housing 66 Calculator 67 Measured object 68 Work stage 70 Coordinate data 71 Shape approximate curve 72 Data with large measurement error 75 Frequency data approximate curve 76 Frequency data

Claims (6)

A shape measuring method for measuring the shape of a surface to be measured,
Scanning while contacting a probe to the surface to be measured and measuring coordinate of the probe to obtain coordinate data;
Converting the coordinate data into frequency data;
Setting an abnormal interval determined based on a frequency threshold set for the frequency data;
Correcting the data included in the abnormal section in the coordinate data;
Calculating measurement data representing the shape of the surface to be measured based on the corrected coordinate data;
A shape measuring method comprising:   The shape measuring method according to claim 1, wherein an approximate curve for the frequency data is calculated, and the frequency threshold is set for the approximate curve.   An approximate curve for the frequency data is calculated, a standard deviation of residual data that is a difference between the approximate curve and the frequency data is calculated, and a frequency threshold based on the standard deviation is set. 2. The shape measuring method according to 1.   4. The shape measuring method according to claim 3, wherein coordinate data obtained by measuring a surface different from the surface to be measured is used in calculating the approximate curve for the frequency data.   The shape measurement method according to claim 1, wherein the frequency data is temporal frequency data. A shape measuring device comprising a movable probe stage, a spring member fixed to the probe stage, a probe held on the probe stage via the spring member, a work stage, and a control unit,
A shape measuring apparatus comprising a program for executing each step of the shape measuring method according to claim 1 in the control unit.