JP2945448B2 - Surface shape measuring instrument - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention is contained in, for example, a three-dimensional image of a surface of a semiconductor chip or an optical disk, or in ultrapure water for cleaning a semiconductor existing on a flat surface. The present invention relates to a surface shape measuring device for measuring a volume of a body such as an amount of deposited impurity residue with high accuracy in a minute dimension, and particularly to a means for obtaining a reference surface for measurement.

[Prior art] Height measurement system (optical height measurement system using astigmatism method or critical angle method, mechanical height measurement system using stylus, scanning tunnel microscope, atomic force microscope, etc.) Height measurement system, etc.) and a sample table on which a sample such as a semiconductor chip to be measured is placed facing each other, and these are placed in a plane perpendicular to the height direction (Z). 2. Description of the Related Art A device that relatively scans in two orthogonal directions (X, Y) to obtain a three-dimensional image of a sample on the sample stage is known.

For example, Japanese Patent Application Laid-Open No. 64-56408 discloses that an object to be measured placed on a sample table is moved in two horizontal directions orthogonal to each other by an XY stage configured by stacking an X table and a Y table.
The object to be scanned and scanned by XY is measured by a height measuring optical system using, for example, a defocus detection method, and height information for each position is obtained. By obtaining three-dimensional image data of the object to be measured by doing, and displaying this as a bird's-eye view or contour map on a display device such as a CRT, a numerical value indicating roughness is calculated and the value is displayed. A shape measuring device is disclosed.

According to the above-mentioned apparatus, it is possible to measure the surface shape of the object to be measured with extremely high precision at a fine dimension of the order of “nm”.

[Problem to be Solved by the Invention] The above-mentioned device has the following points to be improved. That is, since the above-mentioned device is a high-sensitivity measuring device capable of high-accuracy measurement of the surface shape with fine dimensions on the order of “nm”, if there is any slight vertical fluctuation or inclination in the XY scanning system, this becomes “ The measurement data is superimposed as a “swell component”. For example, in an apparatus configured to perform XY scanning by gradually moving in the Y direction while reciprocating the stage on which the sample is mounted in the X direction by a pulse motor, vertical movement due to pitching and rolling of the stage occurs. It is superimposed on the measurement data as an irregular undulating waveform. Further, a temperature drift due to a change in the ambient temperature during the measurement may be superimposed on the data. Such a phenomenon,
It can be said that this phenomenon is difficult to avoid in a high-precision measurement system. However, when the above phenomenon is remarkable, the obtained three-dimensional image becomes greatly distorted, and may exhibit an unsightly state. Therefore, the improvement has been strongly desired.

By the way, since the surface shape measuring device can precisely measure a fine shape on a flat surface portion, a means for obtaining deviation data which is an amount of deviation from a reference surface is provided to this device. If a means for integrating is added, it is possible to measure the volume of a convex portion, the volume of a concave portion, and the like existing in the flat surface portion. However, in this case, it is necessary to determine a volume or a reference plane for measuring the volume in advance. The reference plane is generally obtained from the surface of the sample table. However, when the upper surface of the sample itself placed on the sample table has a flat surface shape and the body volume of the uneven portion existing on the flat surface portion is to be measured, the flat surface portion of the sample upper surface is required. In some cases, a reference plane is created from As a general method of creating a reference plane, a three-reference point designation method is used. First, three points on the flat surface are designated as reference points, and the coordinates of each point are detected. Then, one reference plane is created from these coordinate values. The reference plane thus obtained is, of course, a perfect plane. Therefore, if the “swell component” and “temperature drift” do not contribute to the measurement data, the measurement of the body volume can be performed without any trouble by using the reference surface obtained as described above.

However, in reality, the “swell component” and the “temperature drift” as described above are superimposed on the measurement data. Therefore, the reference surface for measuring the body volume must naturally have a distorted form affected by the "swell component" and the like. That is, if the measurement of the body volume is performed using the reference plane composed of the planes designated by three points which are generally used, an error occurs. Thus, it was difficult to accurately measure the body volume of the uneven portion with the conventional device.

An object of the present invention is to provide a surface shape measuring apparatus capable of removing the influence of noise such as “waviness component” due to mechanical fluctuation of an XY scanning system and “temperature drift” due to a change in ambient temperature.

[Means for Solving the Problems] In order to solve the above problems and achieve the object, the present invention employs the following means. This will be described with reference to the claim correspondence diagram of FIG.

(1) a sample stage 401 on which a sample to be measured is placed;
Height measuring system 402 for measuring the height of the sample on the sample stage
And a scanning system for relatively scanning the sample stage and the height measurement system in two directions (X, Y) orthogonal to each other in a plane perpendicular to the height direction (Z). Is scanned in the XY direction, and in a surface shape measuring device 400 for measuring a three-dimensional image of the sample surface on the sample stage, measurement of the XY array obtained by the height measuring system by scanning the scanning system in the XY direction A memory 406 for storing data 405, a data string extracting means 407 for extracting an X-direction data string and a Y-direction data string of a flat area serving as a reference plane from the measurement data of the XY array, respectively. Based on the extracted X-direction data sequence and Y-direction data sequence, reference surface data 40 composed of a curved surface including a noise component
Reference plane acquisition means 408 for determining 9 is provided.

(2) Further, an averaging processing means 420 is further provided which averages a plurality of data strings adjacent to the data string in the X or Y direction extracted by the data string extracting means 407 to obtain a single data string in the X or Y direction. The reference plane data 409 may be obtained using the averaged data sequence.

(3) Alternatively, X or Y extracted by the data string extracting means
Smoothing processing means for obtaining a smoothing curve from a data string in the direction
430 may be further provided, and the data 409 of the reference plane may be obtained based on the smoothed curve.

(4) Alternatively, a deviation data creating means for obtaining the deviation data 441 of the XY array by subtracting the data 409 of the reference plane from the measurement data 405 of the XY array may be further provided.

(5) Alternatively, there may be further provided a body volume calculating means for calculating a volume of a convex portion or a volume 451 of a concave portion present on the sample from the measurement data 405 of the XY array and the data 409 of the reference plane. .

[Operation] As a result of taking the above-described means, the following operation occurs.

The XY scanning system usually has an X table and a Y table which are provided independently of each other, which are vertically overlapped. Therefore, the “undulation component” corresponding to the shift amount from the reference plane due to the mechanical fluctuation of the XY scanning system can be separated into the undulation component in the X direction and the undulation component in the Y direction. Therefore, the waviness component in the X direction can take any position in the Y direction,
Only the absolute value changes, but the pattern does not change. Similarly, the waviness component in the Y direction has the same pattern at any position in the X direction, only the absolute value changes. The “temperature drift” due to changes in the ambient temperature is
In a scanning system provided such that the X table scans N times while the scanning system performs one scanning of the Y table, it almost affects only the data string in the Y direction.

Thus, by taking the above-mentioned means (1), an X-direction data row and a Y-direction data row relating to a flat surface portion such as the sample table surface or the sample upper surface are extracted from the XY array measurement data 405, and the reference plane data 409 is obtained. Is generated, reference plane data 409 including noise such as “swell component” and “temperature drift” is obtained. Since the surface shape and the unevenness can be determined centering on the reference plane, it is possible to obtain a measurement result from which noise has been removed.

Further, if the means (2) or (3) is taken, the data 40 of the reference plane more faithfully affected by the above-mentioned noise can be obtained.
You will get 9. Therefore, the measurement result becomes more reliable.

Furthermore, if the means of (4) is taken, deviation data 441 as surface shape data with improved accuracy is obtained by removing the influence of noise such as the "undulation component" and "temperature drift".

Alternatively, if the means of the above (5) is taken, data 451 of the body volume of the uneven portion on the flat surface portion from which the influence of noise such as the “swell component” and the “temperature drift” is removed is obtained.

Embodiment FIG. 2 is a diagram showing a configuration centered on an optical system according to an embodiment of the present invention. In FIG. 2, reference numeral 10 denotes a sample table on which a sample 1 to be measured is placed. This sample stage 10
Is installed on the XY scanning mechanism 20. This XY scanning mechanism 20
Will be described in detail later.
The sample table placed on X table 21
10 That is, place the sample 1 in a plane perpendicular to the height direction (Z),
Scanning is performed in two orthogonal directions (X, Y).
A main optical system 30 is provided so as to face the sample 1 on the sample stage 10 from above.

The main optical system 30 includes an objective lens 31 and a quarter-wave plate 3 on the optical axis 3.
2. A semi-transparent mirror 33 is provided, and is configured to irradiate the surface of the sample 1 with light and take out the reflected light. The main optical system 30 is coupled to a height measuring optical system 40 and an observation optical system 50 described below, and is a system commonly used for both optical systems 40 and 50.

The height measuring optical system 40 has a beam splitter 41 disposed on the optical axis 4 orthogonal to the optical axis 3 and deflected by 90 ° by the semi-transparent mirror 33 so as to face the light incident end of the beam splitter 41. A light source 43 that emits a linearly polarized light beam via an optical element 42 such as a cylindrical lens for shaping a beam shape is disposed, and a pair of critical angle prisms 44, 45 through 1st and 2nd
The two-divided light receiving elements 46 and 47 are disposed, and a defocus detecting device using a so-called critical angle method is provided. As the light source 43, a semiconductor laser which dislikes vibration, has high sensitivity, and can be easily reduced in size is used. The outputs of the two-divided light receiving elements 46 and 47 are subjected to primary processing by a signal processing circuit 60, and thereafter, are processed by a computer 61 as a control circuit and supplied to a display 65.

The observation optical system 50 includes a filter 51 on the extension optical axis of the optical axis 3,
A semi-transparent mirror 52, an imaging lens 53, and a prism 54 are provided, and an illumination lamp 56 is arranged via a condenser lens 55 on an optical axis 5 that is orthogonal to the extended optical axis and passes through the center of the semi-transparent mirror 52. Along with
An eyepiece 57 is arranged on the refraction output axis 6 of the prism 54. The sample surface image obtained through the eyepiece 57 can be appropriately observed with the naked eye 80.
A sample surface image obtained by passing straight through the prism 54 is photographed by a video camera 90, and can be monitored by a video monitor 91.

The laser beam emitted from the light source 43 in the height measurement optical system 40, that is, the linearly polarized beam, is converted into parallel light having a circular cross section by the beam shaping optical element 42, enters the beam splitter 41, and is reflected. Light along the optical axis 4. This light is reflected by the semi-transparent mirror 33 and becomes light along the optical axis 3.

On the other hand, the light emitted from the illumination lamp 56 enters the semi-transparent mirror 52 via the lens 55, is reflected there, passes through the semi-transparent mirror 33 via the filter 51, and passes through the laser beam on the optical axis 3. And become one.

The combined laser light and illumination light pass through the quarter-wave plate 32 and enter the objective lens 31. When passing through the quarter-wave plate 32, the laser light is converted from linearly polarized light to circularly polarized light. The laser light is condensed by the objective lens 31 and projected as a minute spot on the surface of the sample 1 on the XY scanning mechanism 20. The illumination light illuminates the entire visual field through the objective lens 31.

The illumination light reflected from the sample 1 passes through an objective lens 31, a quarter-wave plate 32, a semi-transmissive mirror 33, a filter 51, and a semi-transmissive mirror 52, is imaged by an imaging lens 53, is refracted by a prism 54, and is refracted by an eyepiece. 57
To the field stop surface. The light transmitted through the prism 54 is incident on a video camera 90 having a CCD image sensor and the like,
It is imaged. The image signal is sent to the video monitor 91 and displayed. The quarter-wave plate 32 is set in a state slightly inclined from a direction perpendicular to the optical axis 3. Thus, the illumination light from the illumination lamp 56, which is a light source for observation illumination, is not directly reflected and enters the observation optical system,
A fresh field observation image without flare is obtained.

The laser light reflected from the sample 1 passes through the objective lens 31 and the quarter-wave plate 32. At this time, the laser light is
It becomes 90 degree rotated linearly polarized light. The laser light reflected by the semi-transparent mirror 33 enters the beam splitter 41 and is split into two. One of them is projected onto a two-piece element 46 via a critical angle prism 44, and the other is projected onto a two-piece light receiving element 47 via a critical angle prism 45. The output signals of each of the two divided light receiving elements 46 and 47 are converted into a signal as height information by a signal processing circuit 60, then processed by a control circuit 61, sent to a display 65, and sent as a three-dimensional image. Is displayed.

As a means for obtaining the three-dimensional image, a so-called defocus detection method is applied similarly to the means disclosed in JP-A-59-90007 and JP-A-60-38606. The outline will be described below.

When the surface measurement point of the sample 1 is at the focal position of the objective lens 31, the reflected light transmitted through the objective lens 31 becomes a parallel light beam.
When the surface measurement point of the sample 1 is located closer to the focal position of the objective lens 31, the light passing through the objective lens 31 becomes a divergent light beam. At a distant position, the light passing through the objective lens 31 becomes a convergent light flux. In other words, when deviated from the focal position, both become non-parallel light beams and the critical prisms 44, 45
Incident on. The reflection surfaces of the critical angle prisms 44 and 45 are set in advance so as to form a critical angle with respect to the parallel light beam. Therefore, when a non-parallel light beam enters the critical angle prisms 44 and 45, the central ray enters at the critical angle, but the light beam shifted to one side from the center has an incident angle smaller than the temporary angle, and a part of the light is prismatic. The light goes out and the remaining light is reflected. Further, the light beam deviated from the center to the other has an incident angle larger than the critical angle and is totally reflected. By repeating such an operation several times in the temporary angle prism, light incident at an angle smaller than the critical angle,
The difference between the light incident at an angle equal to or larger than the critical angle and the detected light amount is enlarged. Moreover, in this case, the magnitude relationship is reversed between the case where the surface measurement point of the sample 1 is closer to and farther than the focal position of the objective lens 31. When such light is received by the two-divided light receiving elements 46 and 47, respectively, and the photoelectrically converted signals are processed by the signal processing circuit 60, a substantially linear relationship with the height of the unevenness on the surface of the sample 1 is obtained. Is obtained. Therefore, the height information of each measurement point is sequentially taken in as data while scanning the sample 1 by XY scanning by the XY scanning mechanism 20, and this is processed by the computer 61 at a high speed.
A three-dimensional stereoscopic image or the like can be obtained.

FIG. 3 is an exploded perspective view showing a main part of the XY scanning mechanism 20. As shown in the figure, a pair of guide blocks 24a and 24b are installed in parallel along both side edges on the slide plate 23 and screwed. Rectangular leaf springs 25a and 25b are screwed to both ends of the pair of guide blocks 24a and 24b, respectively. Both ends of the Y table 22 are sandwiched between the rectangular leaf springs 25a and 25b. The Y table 22 is coupled to a displacement end of a laminated piezoelectric actuator 26 whose base end is fixed on a slide plate 23, and is driven to be displaced in the Y direction. Also Y
A pair of guide blocks 27 along both side edges on the table 22
a, 27b are installed in parallel and screwed. Rectangular leaf springs 28a, 28b are screwed to both ends of the pair of guide blocks 27a, 27b, respectively. The X table 21 is sandwiched between both ends of the rectangular leaf springs 28a and 28b. The X table 21 is coupled to a displacement end of a laminated piezoelectric actuator 29 having a base end fixed on a Y table 22, and is driven to be displaced in the X direction. Therefore, the XY scanning mechanism 20 is mounted on the Y table 22 while the Y table 22 on the slide plate 23 is reciprocated in the Y direction by the driving force of the piezoelectric actuator 26 mounted on the slide plate 23. The X table 21 on the Y table 22 is reciprocated in the X direction by the driving force of the piezoelectric actuator 29.

A voltage obtained by amplifying the sawtooth wave having a period T is applied to the piezoelectric actuator 26 for driving the Y table, and a voltage obtained by amplifying a triangular wave having a period T / N is applied to the piezoelectric actuator 29 for driving the X table. You. Thus, in one direct movement in the Y direction by a sawtooth wave of T period, a plurality of (eg, 115) sinusoidal movements in the X direction by a triangular wave of T / N period are performed. Repeated.

FIG. 4 is a diagram schematically showing the state of the XY scanning. Considering the XY scan as the movement of the light spot P set at the center of the field of view of the observation optical system 50 as shown in the figure, the light spot P is a solid line path from the center of the measurement area 100 as indicated by arrows a to d. To follow. During this time, data is captured. Among the acquired measurement data, data collected on a movement path a from the light spot P position to the raster scan start position, data collected on a movement path d from the raster scan end position to the P position, and both ends in the X-axis direction. The data collected in the region Δx and the data collected in the both end regions Δy in the Y-axis direction are all deleted as unnecessary data. Therefore, the measurement data actually used effectively is the measurement data in the effective measurement area 101 surrounded by a broken line in FIG. The measurement data of the XY arrangement is stored in the RAM in the computer 61, and after data processing, is displayed on the display 65 as a contour map or the like.

FIG. 5 is a block diagram showing the overall configuration of the control system in this embodiment.

In FIG. 5, a computer 61 performs all of the basic processing of the apparatus, and stores a main program, a data collection program for measurement, a data processing program for processing obtained data, and the like.
It has a built-in ROM, a RAM used for temporary storage in various calculations and storage of obtained XY array measurement data, and the like. An input interface 62 and an output interface 63 are provided. The computer 61 has a hard disk 64, a display 65, a mouse 66
Etc. are connected. A keyboard 67, a signal processing circuit 60, and the like are connected to an input terminal of the computer 61, and a printer 68, a high-voltage amplifier 70 for a piezoelectric actuator, and the like are connected to an output terminal.

The high voltage amplifier for voltage actuator 70 is a computer
Amplify the X-axis drive pulse Px, Y-axis drive pulse Py, and Z-axis drive signal Pz from 61,
9 and 26 and the Z-axis voltage actuator 71 in the main optical system 30 and the height measuring optical system 40.

The output signal from the height measuring optical system 40 is output to a signal processing circuit 60.
Is converted into a height information signal, and supplied to the inside of the computer 61 via the input interface 62.
At this time, it is synchronized with the X-axis drive pulse Px and is sampled by the subdivided sampling pulse Pt.

Incidentally, the data processing program stored in the computer 61 has a "correction process" for correcting the measurement data of the XY array obtained by the measurement as shown in FIG. Includes programs.

Next, the operation of the present embodiment configured as described above will be described in detail.

This apparatus is almost the same as the apparatus disclosed in JP-A-64-56408 except for a part of the data processing program stored in the computer 61. Therefore, a detailed description thereof will be omitted, but a high-precision measurement of the surface shape with a fine dimension of the order of “nm” can be performed optically similarly to the above-mentioned apparatus.

That is, in the data collection process, the XY scanning mechanism 20 is operated on the sample 1 set on the sample stage 10 to acquire the height information measured by the height measuring optical system 40, and the fourth information is acquired.
As explained in the figure, after deleting unnecessary data,
The obtained measurement data of the XY array is stored in the RAM of the computer 61. In addition, the data processing process is a graphic display such as a contour map, a bird's-eye chart, a cross-sectional view, etc. by performing arithmetic processing on the measurement data of the XY array in accordance with a routine in which the menu screen of the display 65 is selected by the mouse 66, Numerical displays such as variance, correlation, roughness, and volume are output to the display 65 and the printer 68.

By the way, the measurement data of the XY array obtained in the above data collection process includes noise components such as “undulation component” due to mechanical vibration in XY scanning and “temperature drift” due to temperature change of the measuring device. .

If the "correction processing" for removing such a noise component is selected as necessary, the data correction processing routine is executed according to the flow of FIG. 6 prior to the above routine in the data processing process. Is done.

That is, when “correction processing” is selected, step S1
To enter the screen 300 of the display 65 as shown in FIG.
Displays a contour diagram 310, a cross-sectional view 320 in the X direction, a cross-sectional view 330 in the Y direction, a numerical value display section 340 for body volume, etc., and a menu selection section 350 based on the measurement data of the XY arrangement stored in the RAM.
At this time, the lines 312 and 313 indicating the data string extraction positions in the X direction and the Y direction are initially set to intersect at the center of the contour map, and the cross-sectional view at the position indicated by these lines 312 and 313 is shown in the X direction. Are displayed as a cross-sectional view 320 in the Y direction.

Steps S2, S3, S4, and S5 are steps for extracting a data string. By moving the mouse 66, the screen 30 shown in FIG.
Move the cursor 360 shown in 0 to draw the contour map 31
When an arbitrary point on 0 is picked and an extraction position is input (step S2), the lines 312 and 313 indicating the previous extraction positions disappear, and new lines 312 and 313 intersecting the new pick position.
Are drawn on the contour map 310, and at the same time, the corresponding X
A cross section 321 is newly drawn as a data row in the direction and a cross section 331 is drawn as a data row in the Y direction on the respective cross-sectional views 320 and 330 (step 3). These cross sections 321 and 331 must be selected so as to avoid convex portions and concave portions on the measurement data. Therefore, when the extracted data string is determined to be inappropriate in step 4 (step S4), steps S2 to S4 are repeated until an appropriate data string is extracted. The determination of OK in step S4 is made by picking the corresponding field of the menu selection unit 350 with the mouse 66. When OK is selected, the X, Y data string is extracted (step S5) and stored in a specific area in the RAM.

Steps S6 and S7 are steps of data string averaging processing. If "Yes" is selected in step S6, the extracted X and Y data strings and the two adjacent data strings in the XY array measurement data are read, and a total of five data strings are averaged. To create one column of data (step S7). The extracted X, Y data strings stored in the RAM are rewritten with these new X, Y data strings.

Steps S8, S9, and S10 are steps of a data string smoothing process. If "Yes" is selected in step S8, the degree of the smoothing curve is input using the keyboard 67 (step S8).
9). This order input can be omitted by programming a predetermined order in advance. The smoothing process is performed by calculating a multi-degree curve having a predetermined order corresponding to the X, Y data strings stored in the area using the least squares method (step 10). The calculated constants and coefficients of the two multiple-order curves are stored in another area.

Subsequently, the process proceeds to step S11, where a reference plane is calculated. The calculation of the reference plane is performed based on the X, Y data strings stored in the area of the characteristic in the RAM, but when the data is subjected to the smoothing process, the two data stored in different areas are stored. This is performed based on the constants of the multi-degree curve and each coefficient to obtain reference plane data in an XY array (step S1
1).

Steps S12, S13, and S14 are steps for creating deviation data. If Yes is selected in step S12, deviation data is created by subtracting the XY array reference plane data from the XY array measurement data (step S13). This deviation data is displayed as a three-dimensional image on the screen 300 (step S14).

Steps S15 to S19 are steps for calculating body volume.
If Yes is selected in step S15, step S1 is performed based on the volume obtained by integrating the measurement data of the XY array in step S16.
The body volume data is created by subtracting the volume obtained by integrating the reference plane data in the XY array in step 7 (step S18). This body volume data is displayed on the numerical value display section 340 in the screen 300 (step S19).

The above is the contents of the “correction processing” routine. As you can see from this routine, the process to calculate the reference plane is
As shown in (a), (b), (c), and (d) in FIG. 8, four types are selected, and the data strings to be calculated are different.

By the way, each processing in the above-mentioned "correction processing" routine is programmed based on the following concept,
Is being processed.

That is, the XY scanning mechanism 20 of this embodiment is configured such that an independently provided X table 21 and Y table 22 are vertically overlapped as shown in FIG. Therefore, the "undulation component" due to the mechanical fluctuation of the XY scanning mechanism 20
Has features as shown in FIG.

FIG. 9 is a schematic diagram in which the degree of undulation is maximized so that the "undulation component" generated by the movement of the X table 21 and the Y table 22 can be easily understood. As shown in the figure, the “swell component” is a “swell component” x (21) in the X direction caused by the movement of the X table 21 and a “swell component” y (22) in the Y direction caused by the movement of the Y table 22. ). In addition, the curve indicating the “undulation component” x (21) in the X direction can be expressed as x at any position (y1, y2, y3) in the Y direction.
The patterns are the same except that their absolute values change as shown in (y1), x (y2), x (y3). Similarly, the curve indicated by the “swell component” y (22) in the Y direction indicates that y (1x), y (x), regardless of the position (x1, x2, x3) in the X direction.
2) As shown in the figure as y (x3), the pattern is the same, only its absolute value changes.

On the other hand, “temperature drift” due to changes in ambient temperature
Since the X table 21 is configured to scan N times while the Y table 22 performs one scan of the scanning mechanism 20 as described above, this affects almost only the data string in the Y direction.

Here, the method of calculating the reference plane will be described with reference to FIGS. 8, 10, and 11 (a) and 11 (b) as appropriate.

FIG. 10 is a perspective view schematically showing a reference surface S presenting a smooth curved surface in the X direction and a step surface having a step H in the middle in the Y direction in order to simplify the description and illustration. is there. However, it goes without saying that the Y-direction actually shows a smooth curve.

First, description will be made in accordance with the simplest case of FIG. 8 (a).

 First, the reference plane S (i, j) is defined as follows.

 S (i, j) = C + f (i) + g (j) (1) where C: offset value (constant) of reference plane S f, g: f (0) = g (0) = 0 i = 0,1 , 2,..., M−1 j = 0, 1, 2,..., N−1 M and N are the number of sampling points in the X and Y directions, respectively.

The above equation (1) is an equation of a curved surface having independent “undulation components” in the X direction and the Y direction. The basis for this definition is that the X table 21 and the Y table 22 in the XY scanning mechanism 20 are provided independently in two orthogonal directions, The temperature drift of the data is due to being effective only in the Y direction.

When the reference plane S defined as described above is obtained, the X-direction and the Y-direction are basically obtained from a region where the uneven portion or the like of the sample 1 to be measured does not exist, that is, a flat surface portion.
Select one line at a time from each direction. These are the X direction reference line LX and the Y direction reference line LY. The above two criteria LX and LY
And the coordinates of the intersection with (i ₀ , j ₀ ). Then, since the reference lines LX and LY are curves existing on the reference plane S, they can be expressed by the equations (2) and (3), respectively.

LX (i, j ₀ ) = C + f (i) + g (j ₀ ) (i = 0,1,.
··, M-1) (2 ) LY (i 0, j) = C + f (i 0) + g (j) (j = 0,1, ·
···, N-1) (3) Substituting i = 0 and i = i ₀ into the above equation (2) to obtain the following equations (2 ′) and (2 ″) LX (0, j ₀ ) = C + f (0) + g (j ₀ ) = C + g (j ₀ ) (2 ′) LX (i ₀ , j ₀ ) = C + f (i ₀ ) + g (j ₀ ) (2 ″) Similarly, in the equation (3), The following equation (3 ') is obtained by substituting j = 0.

LY (i ₀ , 0) = C + f (i ₀ ) + g (0) = C + f (i ₀ ) (3 ′) From (2 ′), (2 ″), (3 ′), C = LX (0, j ₀₎ ) + LY (i ₀ , 0) −LX (i ₀ , j ₀ ) f (i ₀ ) = LX (i ₀ , j ₀ ) −LX (0, j ₀ ) g (j ₀ ) = LY (i ₀ , j ₀ ) −LY (i ₀ , 0) (4). By substituting this into equations (2) and (3), f (i) = LX (i, j ₀ ) −LX (0, j ₀ ) g (i) = LY ( i 0, j) -LY (i 0, 0) (5) is obtained. (4), by substituting (5) into (1), the reference plane S is following Is required.

S (i, j) = LX (i, j 0) + LY (i 0, j) -LX (i 0, j 0)
(6) The above is the case where the flat surface portion of the measurement data is smooth. However, when the flat surface portion includes high frequency noise due to vibration during measurement or fine dust on the sample surface, a more accurate reference is used. In order to obtain a surface, it is necessary to smooth the extracted data sequence before obtaining a reference surface. As an example, an example in which smoothing is performed by the least square method will be described with reference to the case of FIG. 8B.

First, as in FIG. 8 (a), one data row in each of the X and Y directions is extracted, and m-th and n-th curves are respectively applied to the data rows by the least squares method. Lines LX ₂ and LY _{2 are} assumed.

Where C ₀ , d ₀ : constant C _K , d ₁ : coefficient m: order of the reference line in the X direction n: order of the reference line in the Y direction i = 0, i = i ₀ , (8 ) respectively by substituting j = _0, j = j 0 in equation (7 '), (7'), (8 '),
(8 ″) is obtained.

From the above equation, the reference plane S is calculated from (7), (8), and (7 ″) in the same manner as when equation (6) is derived. Or from (7), (8), (8 ") Is required. Thus, when the data string is smoothed, strictly speaking, as shown in equations (9) and (10), 2
A reference plane is required. The difference between the two is the constant term, that is, the difference between the offsets in the height direction. Since this difference is usually small, either equation (9) or (10) can be used. If necessary, a method of averaging two constant terms is also conceivable.

When the measurement data is not smooth, in addition to the above-described smoothing, a method of calculating a reference plane after forming an average of a plurality of adjacent data strings to form a data string, smoothing the data string and setting it as a reference line. Thereby, a more accurate reference plane can be obtained. This procedure is shown in FIG.

FIG. 11 (a) shows that the reference line LX has a common pattern at any position in the Y direction, and the difference is that there is a difference H in the absolute value. Similarly, (b) shows that the reference line LY has a common pattern at any position in the X direction, and a different point is a point having a difference G in the absolute value.

In this manner, the reference surface S including a curved surface including noise such as “undulation component” and “temperature drift” is obtained. If the surface shape or the like is determined with reference to the reference plane S thus obtained, a measurement result that does not include distortion due to noise can be obtained. That is, by obtaining the reference plane data S by the procedure as shown in FIG. 8 and correcting the measurement data with this, a three-dimensional image without distortion can be obtained.

FIG. 12 is an actual measurement diagram showing an example of a three-dimensional image when the above correction is performed, and FIG. 13 is an actual measurement diagram showing a three-dimensional image of the same object as in FIG. 12 when the above correction is not performed. It is. As is clear from a comparison between FIG. 12 and FIG. 13, it will be obvious at a glance how useful the correction by the reference plane S is.

Next, a method of obtaining the volume of the convex portion existing on the flat surface portion will be described with reference to FIGS. 14 (a) and (b) to FIG.

Now, a case where the volume of the convex portion 112 existing on the flat surface portion 111 on the upper surface of the sample table 10 as shown in FIGS. As described above, two orthogonal lines are determined on the region where the convex portion 112 does not exist, and the reference lines LX and L
The reference plane S is calculated based on Y.

The volume of the convex portion 112 is obtained by calculating a volume V2 from the base surface 113 of the sample stage 10 to the surface S2 including the convex portion 112 created by the sampling value, and a volume V1 from the base surface 113 to the reference surface S1, It can be obtained by subtracting V1. Here, the “base surface” refers to a temporary surface serving as a reference for height measurement, such as the lower end or the zero point of the measurement range of the height measurement system.

The volume V2 up to the surface S2 including the protrusion 112 is obtained as follows.

FIG. 15 is a plan view schematically showing a range sampled with respect to a measurement region including the convex portion 112. This example is M × N
3 shows an example of sampling. In FIG. 15, when one of the hatched sampling blocks is extracted, it takes a form as shown in FIG.

S2 shown in FIG. 16 indicates a part of the surface S2 including the above-described protrusion 112. The surface is distorted as shown.
Four corners S21, S22, S23, S24 forming this surface S2
Are sample values of S21 ... Z _{(i, j)} S22 ... Z _{(i + 1, j)} S23 ... Z _{(i + 1, j + 1)} S24 ... Z _{(i, j + 1)} Note that (0 ≦ i ≦ M−2) and (0 ≦ j ≦ N−2).

Therefore, the volume v _ij from the base surface 113 to the convex surface S2 in one sampling block is obtained by the following equation.

The volume v _S2 up to the surface S2 including the convex portion of the entire data sampled by M × N is obtained by the following equation.

In the above description, the sampling interval is [1] and the measurement range in the height direction is dimensionless such that 0 ≦ Z _{(i, j)} ≦ R (R is, for example, 1000).
_S2 is the correct the [μ _m ^3], V _E shown in the following equation (13) to (12)
Should be multiplied.

v _E = lx · ly · rng / (M-1) (N-1) R [μm ³ ]
(13) That is, V _S2 = v _S2 × v _E [μm ³ ] (14) where lx is the sampling length “μm” in the X direction and ly is Y
The sampling length [μm] and rng in the direction are the measurement range [μm] in the height direction.

Since the constant of the reference plane S _{i, j (i, i)} is determined by the above-described equations (6) to (9) and (10), the volume v _S from the base plane 113 to the reference plane _S is Similarly to equation (12), Is required. In addition, there is also a method in which v _S is obtained by the following integration on the reference plane of Expression (9) or Expression (10).

However, when equation (9) is used for S _{(i, j)} in (16), And when taking equation (10), It is.

V _{S in} equation (16) is a dimensionless quantity. This is the actual volume
To convert it to V _S [μm ³ ], multiply v _S by v _E in equation (13). That is, V _S = v _S × v _E (17) Thus, the volume V of the convex 112 is obtained by the following equation.

V = V _S2 −V _S = (v _S2 −v _S ) × v _E (18) By the way, according to the correction processing routine of FIG. 6, the extraction processing of the X and Y data strings in step S5 is as follows. A data string in the X direction and a data string in the Y direction extracted from the measurement data Z (i, j) are respectively converted into a one-dimensional array LX '(i), LY'.
(J).

In the averaging process in step S7, when averaging is performed on five adjacent data columns, the result is expressed as LX '(i), L
Y ′ (j).

In the smoothing process in step 10, the LX '(i) and LY' (j) obtained in this way are subjected to the least-squares method by fitting the m-th order expression and the n-th order expression to the smoothed data sequence. LX
(I) and LY (j) are obtained.

When smoothing is not performed, LX '(i) and LY' (j) are directly substituted for LX (i) and LY (j).

LX (i) and LY (j) obtained as described above are expressed by equations (2) and (3). In step 11, the reference plane S _{(i, j)} is obtained in accordance with equation (6).

When the deviation data is created in step 13, Z ′ _{(i, j)} = Z _{(i, j)} −S _{(i, j)} (i = 0,1,..., M−1) (j = 0 , 1, ..., N-1) (23) to obtain deviation data Z '(i, j).

When obtaining the body volume, the processing of the equation (12) is performed in step 16, the processing of the equation (15) is performed in step 17, and the processing of the equation (18) is performed in step 18.

“Experimental example (1)” As shown in FIG. 17, a model having a size of L / 2 × L / 2 and a full range (1000 A model 200 having a convex portion 202 having a height of 30% (300) of ()) was prepared.

In the X direction of the sampling area 201 of this model 200, a sine wave having an amplitude of 1% of the full range and a spatial frequency of 0.9 L is applied, and in the Y direction of the same area 201, an amplitude of 6% of the full range and a spatial frequency of Was a model in which a 2.3L sine wave was placed and sampled at a sampling number of 2 × 512.

While changing the orders m and n of the reference lines LX and LY in the range of 1 to 20, a reference surface is obtained by the case of FIG.
Was calculated, the error with respect to the true value was 1.0
% Or less when m, n ≧ 18. Similar results were obtained with other test data at different frequencies. Therefore, it can be said that it is desirable that both the orders m and n be about 20.

Routine processing time for fitting a 20th order curve to the 512-point data by the least squares method was about 12 seconds.

The absolute value of the error in the volume calculation of the convex portion 202 is determined by how appropriately the reference plane is determined. In addition, the influence is uniform regardless of the size of the convex portion 202. For this reason, the smaller the convex portion 202 to be measured, the larger the error with respect to the true value. Therefore, it is necessary to pay attention to the size of the measurement target.

In the case of FIG. 8 (d), the order m, n of the reference line LX, LY
Are set to 20 and the reference line LY is determined from five lines.
When the volume of the convex portion 202 of the model 200 was measured, the error with respect to the true value was only 0.6%. It can be said that such measurement accuracy is comparable to the measurement accuracy of the measuring device.

"Experimental example (2)" A silicon wafer was prepared as a sample, and the area of no protrusion was measured over a range of 800 µm x 800 µm. Ideally, the volume of the convex portion should be 0, but the measurement result obtained data of "100 μm ³ or less". Dividing this data by the measurement area gives 1000 μm ^{3 / (} 800 μm × 800 μm) = 1.56 [nm]. This is a value sufficiently smaller than the resolution of 10 nm in the height direction of the measuring device, and it has been found that the measurement accuracy is sufficient.

The present invention is not limited to the above embodiment.
For example, in the above-described embodiment, the description has been made on the assumption that a complex quadratic curve is used as a curve to be applied to the extracted data string by the least square method, but an arc or the like may be used depending on the measurement object. Further, instead of the least square method, a smoothing means such as a moving average method can be employed. Further, in the above-described embodiment, the two-dimensional trapezoidal formula is used as an integration method for obtaining the volume, but a more accurate calculation may be performed using a two-dimensional Simpson method or the like.

Further, the present invention is not limited to an apparatus employing an optical type as in the present embodiment as a height measuring system, but may be a stylus type or ST type.
Naturally, it is also effective in a device employing a sensor such as M or AFM. In addition, it goes without saying that various modifications can be made without departing from the spirit of the present invention.

[Effect of the Invention] The present invention relates to a device for measuring a three-dimensional image of a sample surface and / or a body volume such as an uneven portion using a height measurement system, from an XY array data sequence obtained by XY scanning. Since each column is configured to extract a specific row and create a reference plane based on this, if the extraction of the data row is performed on a flat surface such as the sample table surface or the sample upper surface, It is possible to obtain a reference surface that shows the influence of noise such as "waviness component" due to mechanical fluctuations and "temperature drift" due to changes in ambient temperature. It is possible to accurately measure the body volume of the uneven portion.

[Brief description of the drawings]

FIG. 1 is a view corresponding to claims, FIGS. 2 to 17 are views showing an embodiment of the present invention, FIG. 2 is a structural view mainly showing an optical system of a surface shape / body volume measuring device, FIG. FIG. 4 is an exploded perspective view showing the main part of the XY scanning mechanism, FIG. 4 is a diagram schematically showing an XY scanning condition, FIG. 5 is a block diagram showing the configuration of the entire control system, and FIG. FIG. 7 is a flowchart showing a correction processing routine for correcting data by obtaining a reference plane. FIG. 7 is a view showing a display on a display screen. FIG. 8 is a flowchart showing a procedure for obtaining a reference plane. FIG. 10 is a schematic view for explaining a “swell component” generated by the movement of the X table and the Y table, FIG. 10 is a perspective view of a reference plane schematically shown, and FIGS. Explanatory drawing of the reference line pattern according to FIG. 12, FIG. 13 and FIG. 14 (a) and (b) to 16 show a case where the volume of a convex portion existing on a flat surface portion is measured. FIG. 17 is a plan view showing an outline of a model used in “Experimental example (1)”. 1 ... Sample, 3-6 ... Optical axis, 10 ... Sample table, 20 ... XY scanning mechanism, 21 ... X table, 22 ... Y table, 30 ... Main optical system, 40 ... Height Measurement optical system, 50: Observation optical system, 60: Signal processing circuit, 61: Computer,
65 ... Display

Claims (5)

(57) [Claims] 1. A sample table on which a sample to be measured is placed, a height measuring system for measuring the height of the sample on the sample table, and a height direction of the sample table and the height measuring system. (Z) has a scanning system that relatively scans in two directions (X, Y) orthogonal to each other in a vertical plane, scans the scanning system in the XY direction, and scans the surface of the sample on the sample stage. In a surface profile measuring device for measuring a three-dimensional image, a memory for storing the XY array measurement data obtained by the height measurement system by scanning the scanning system in the XY direction, Data string extracting means for extracting an X-direction data string and a Y-direction data string of a flat area serving as a reference plane; and a noise component based on the X-direction data string and the Y-direction data string extracted by the extracting means. And a reference plane obtaining means for obtaining a reference plane composed of a curved surface. Profilometer. 2. An averaging processing means for averaging a plurality of data strings adjacent to the data string in the X or Y direction extracted by the data string extracting means to obtain a single data string in the X or Y direction. 2. The surface shape measuring apparatus according to claim 1, wherein the reference plane is obtained using the averaged data sequence. 3. The apparatus according to claim 1, further comprising: a smoothing processing means for obtaining a smoothing curve from the data string in the X or Y direction extracted by the data string extracting means, wherein the reference plane is obtained based on the smoothing curve. The surface shape measuring device according to claim 1. 4. The surface shape measurement according to claim 1, further comprising a deviation data creating means for obtaining deviation data of the XY array by subtracting data of the reference plane from the measurement data of the XY array. apparatus. 5. The apparatus according to claim 1, further comprising a body volume calculating means for calculating a volume of a convex portion or a volume of a concave portion present on the sample from the measurement data of the XY array and the reference plane. The surface shape measuring device according to the above.