JPH0854223A - Apparatus for measuring width of semiconductor pattern - Google Patents

Abstract

PURPOSE:To measure the line width correctly by subjecting an image signal obtained through an electronic microscope to A/D conversion, storing the converted signal in am image memory and referring to a semiconductor pattern and a waveform data related to a semiconductor pattern being measured. CONSTITUTION:A previously written reference pattern is scanned by means of an electronic microscope and an analog image signal therefrom is digitized before being stored in an image memory 3. The image signal is then processed by a CPU 7 according to a calculation procedure stored in a procedure data memory 6 thus storing a reference pattern edge position and a required data in a reference pattern data memory 4. At the time of measurement, a required data of pattern to be measured is stored in a pattern data memory 5 to be measured according to the similar procedure. Edge position of a pattern to be inspected, corresponding to that of the reference pattern, is then detected by collating the patterns through the use of these data and the collation results are presented on a display 8.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a semiconductor pattern width measuring apparatus, and more particularly to a semiconductor pattern width measuring apparatus capable of easily measuring the length of a circuit pattern having a complicated edge waveform.

[0002]

2. Description of the Related Art In a semiconductor manufacturing process, in general, in order to improve product reliability and yield, the pattern line width of a photoresist or a formed film after etching is measured and quality is checked. Recently, semiconductor patterns have been miniaturized, and the resolution of optical microscopes conventionally used for the purpose of this measurement has been insufficient.
High resolution electron microscopes have been used. Further, development of an automatic device for measuring line width using this electron microscope has been actively conducted. As a typical dimension measurement method of these devices, a peak detection method using a peak of a mountain of a one-dimensional waveform obtained from an observation pattern as a pattern edge (“high-precision line width measuring device”, Semiconductor)
World 1984.12, pp.87-93), and a linear approximation method in which a slope line and a base line of a one-dimensional waveform are linearly approximated and an intersection thereof is defined as an edge (“Super LSI micro-size measurement system MEA-3000”, Semiconductor World 1985.1) , pp.
120-128).

[0003]

In the above-mentioned conventional method,
At the time of length measurement, it is assumed that a pattern dedicated to length measurement is used. That is, in the peak detection method, one peak exists for the one-dimensional waveform corresponding to one edge of the pattern, and in the linear approximation method, one peak exists for the one-dimensional waveform corresponding to one edge of the pattern. It is assumed that one slope line and one baseline can be approximated. However, such an assumption can be applied only to a measurement pattern having a simple shape. On the other hand, recently, there is a strong demand for measuring the length of an actual circuit pattern having a complicated shape for the purpose of failure analysis. In this case, since a plurality of circuit patterns are duplicated, the one-dimensional waveform near the edge is often a rather complicated waveform having a plurality of peaks and valleys. Therefore, for example, in the peak detection method, it is not known which peak corresponds to the pattern edge to be measured, and it is often difficult to find the baseline by the linear approximation method. SUMMARY OF THE INVENTION An object of the present invention is to provide a semiconductor pattern width measuring apparatus which solves such a conventional problem and makes it possible to measure the length of a circuit pattern having a complicated edge waveform.

[0004]

In order to achieve the above object, a semiconductor pattern width measuring apparatus according to the present invention scans a prepared semiconductor pattern and another semiconductor pattern to be measured by scanning them with an electron microscope. Signal converting means for converting the obtained image signal from analog to digital, an image memory for storing the digital signal converted by the signal converting means, and an address for generating an address of the image memory using a scanning synchronization signal of an electron beam Generating means, a computer for processing an image stored in the image memory, and extracting necessary data at an edge position of the semiconductor pattern and the semiconductor pattern to be measured according to a calculation procedure, and a waveform relating to the semiconductor pattern obtained by the computer A first memory for storing data, and the semiconductor pattern to be measured obtained by the computer; A second memory for storing input waveform data concerning the input signal, and, as the calculation procedure, in order to compare the respective signals of the semiconductor pattern and the measured semiconductor pattern, after removing noise of each signal with a Gaussian filter, Obtaining each approximate waveform by changing the standard deviation value of the Gaussian filter, obtaining the coordinates of the inflection point of each approximate waveform, detecting the pair of inflection points corresponding to each approximate waveform of each pattern by matching, Obtain corresponding points of both patterns, detect the start point and end point of the measured semiconductor pattern corresponding to the start point and end point set in the semiconductor pattern among the corresponding points, and measure the width from the difference between the end point and the start point, a series of Memorize the procedure of the third
It is characterized by including the memory of.

[0005]

In the present invention, in order to obtain a starting point for measuring the line width of a semiconductor pattern, global matching and detailed matching of a complicated one-dimensional waveform are dynamically performed. Edge detection corresponding to waveforms of various shapes can be performed as compared with the conventional method of finding an edge position based on a relative magnitude. Also, even for a one-dimensional waveform in which noise exists, by appropriately setting the lower limit values of the filter size, Σ _s and σ _s , peaks and valleys smaller than そ れ_s and σ _s are not detected.
Stable position detection becomes possible. Further, by interpolating the correspondence between the feature positions by polynomial approximation, when the one-dimensional waveform is deformed in the variable direction, for example, when there is a change in the magnification of the electron microscope or a change in the edge cross-sectional shape of the pattern to be measured. However, the position can be detected.

[0006]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 3 is an overall configuration diagram of a semiconductor pattern width measuring device showing an embodiment of the present invention. In FIG. 3, 1
Is an A / D converter that samples an analog image signal 1s of a pattern to be measured or a reference pattern from an electron microscope (SEM) 9 at an appropriate timing and converts it into a digital image signal 1d. An address generating circuit for generating an address 2a of the image memory 3 using the 2s, a first memory 4 for storing data relating to a reference pattern, and a fifth memory 5 for storing data relating to a pattern to be measured. 2 is a memory, 6 is a third memory for storing a measurement procedure according to the present invention, 7 is a computer (CPU) for processing data, and 8 is a display device for displaying measurement results as numerical data and the like. , 7a is an address bus, and 7d is a data bus. Next, an outline of the measurement procedure of this embodiment will be described. The reference pattern written in advance is scanned by an electron microscope, the analog image signal is digitized, and stored in the image memory 3. Then, the image is processed by the computer 7 according to the calculation procedure of the memory 6, and the edge position of the reference pattern and necessary data are stored in the reference pattern data memory 4. At the time of measurement, the required data of the measured pattern is stored in the measured pattern data memory 5 in substantially the same processing procedure. Next, the patterns are compared using these data, the edge position of the pattern to be inspected corresponding to the edge position of the reference pattern is detected, and the result is displayed on the display device 8.

1 and 2 are flow charts showing the procedure stored in the procedure data memory of FIG. 3, and FIG. 4 is a diagram showing the waveforms of the reference pattern and the measured pattern and their approximate waveforms. . FIG. 1A is a flow of a processing procedure of a reference pattern obtained by scanning a prewritten semiconductor pattern with an electron microscope.
FIG. 2B is a flow of the processing procedure of the input pattern obtained by scanning the measured semiconductor pattern with an electron microscope, and FIG.
Is a flow of a processing procedure for comparing the reference pattern and the pattern to be measured. In the first step of FIG. 1 (a), 1-dimensional waveform of predetermined and are the position of the reference pattern from the image memory 3, line _{profile: F (X) (X s} ≦ X
≤X _e ) is taken out. A (thick solid line) in FIG. 4A shows an example of this waveform. Next, in the second step, an approximate waveform: G (X; Σ) obtained by applying a Gaussian filter with a standard deviation Σ shown in the following equation is obtained.

[Equation 1]

In the third step,

[Equation 2] The position coordinates satisfying are obtained. These coordinates represent the changing points of the peaks and valleys of the original waveform, that is, the positions of the inflection points, and are herein referred to as reference characteristic positions. The second and third steps are repeated for the required standard deviation (Σ _s ≦ Σ ≦ Σ _e ), and the reference feature position (XΣ ₁ , ... X) for each standard deviation is repeated.
Σ _i , ..., XΣ _n ) is obtained. Naturally, the number n of reference feature positions is different for each standard deviation. The waveform of the thin line drawn on A of FIG. 4 (a) is obtained when Σ is changed, and · represents the characteristic position. In the figure, the mountain and valley positions are also shown in ° for reference. From the figure, it can be seen that the general / detail of the waveform can be described by the size of the standard deviation. Although the signal obtained from the low-acceleration electron microscope used for this type of measurement has a considerably low S / N ratio, it is possible to obtain a noise-free waveform by applying a Gaussian filter with an appropriate standard deviation Σ. Is possible. Therefore, the minimum value 考慮 す る_s of the standard deviation to be considered is set to this value. In the fourth step, a person inputs the edge position (Xd) on the original waveform. Fifth
In the step, data on the reference pattern collected so far, that is, the reference feature position ((X
Σ ₁ , ... XΣ _i , ... XΣ _n ), the number n thereof, and the edge position (Xd) are stored in the memory 4.

Next, the procedure of data processing when the pattern to be measured is input will be described. In this case as well, the processing is executed in substantially the same procedure as in the case of the reference pattern. That is, the analog image signal from the electron microscope of the measured pattern is digitized and stored in the image memory 3. Then, the image data is processed by the computer 7 in accordance with the calculation procedure stored in the memory 6. The calculation procedure will be described with reference to FIG. In the first step, the one-dimensional waveform (line profile) f (x) (x _s ≤x≤x _e ) of the measured pattern is taken out from the image memory 3. FIG.
A 'in (b) shows an example of this input waveform.
Due to slight differences in the charge state and magnification of the pattern, even if the waveform sampling position is the same position as the reference pattern, deformation occurs. In the second step, an approximate waveform g (x; σ) obtained by applying a Gaussian filter with a standard deviation σ is obtained.

In the third step,

(Equation 3) The position coordinates satisfying are obtained. Here, this coordinate will be called a characteristic position. Then, the second and third steps are repeated for the required standard deviation (σ _s ≦ σ ≦ σ _e ), and the characteristic position (xσ ₁ , ... Xσ _i , ...) For each standard deviation is repeated.
... xσ _m ) is calculated. Naturally, the number m of characteristic positions is different for each standard deviation. The waveform of the thin line drawn on A 'in FIG. 4B is obtained when σ is changed, and indicates a characteristic position. In the figure, the mountain and valley positions are also indicated by ° for reference. Minimum standard deviation to consider σ
_s is set to remove the noise of this waveform.
In the fourth step, the data related to the measured pattern obtained so far, that is, the characteristic position (x
σ ₁ , ・ ・ ・ x σ _i , ... ・ ・ ・ x σ _m ) and its number m
Are stored in the memory 5.

Next, the procedure for collating the data of the measured pattern with the reference pattern obtained by the above-mentioned procedure will be described with reference to FIG.
This will be described with reference to (a) and (b). As described above, when the standard deviation value of the Gaussian filter is large (upper part of the figure), it represents the general shape of the waveform, and when it is smaller (lower part of the figure), it represents the detailed shape close to the sampled waveform. . Therefore, the reference characteristic position of the reference waveform and the characteristic position of the input waveform are collated even if the standard deviation value is large. A specific method of detecting the matching feature position will be described later. A mark with a black circle in a white circle in FIGS. 4A and 4B is the first matching feature position. Next, the matching feature positions are collated again in detail. That is, from the waveform start point
a and a waveform start point-a ', a-b and a'-b', b-c and b'-c ', c-waveform end point and c'-waveform end point, standard deviation values for each part between coincident feature positions Match while making the smaller. The start point and the end point of each partial waveform when the standard deviation value is changed are determined by the first matching feature position (a, b,
c / a ', b', and c '), the coordinates slightly change as indicated by arrows in FIGS. 4A and 4B. The results of the second verification are indicated by black circles. By repeating such collation recursively up to the lower limit values Σ _s and σ _s of the standard deviation values, it is possible to perform comprehensive collation to detailed collation in a unified manner.

Next, a program in which the collation procedure is performed will be described. The language used is the C language, and the declaration of variables and data transfer descriptions irrelevant to the essence of the present invention are omitted or simplified, and do not always adhere to the grammar. MATCH (Σ ₀ , Σ ₁ , X ₀ , X ₁ , σ ₀ , σ ₁ , x ₀ , x ₁ ) / * Σ ₀ , Σ ₁ / σ ₀ , σ ₁ : Minimum and maximum standard deviation of reference and input waveforms * // * X ₀ , X ₁ / x ₀ , x ₁ : Minimum and maximum inflection point search coordinates of reference and input waveforms * / ｛/ * Stop condition-step 1-* / if (Σ ₁ = lower limit standard deviation value) ) ｛Goto end; （if (σ ₁ = lower limit standard deviation) ｔｏ goto end; Σ if (no feature position between Σ ₀ , Σ ₁ , X ₀ , X ₁ ) ｛goto end;｝ if (σ ₀ , Σ ₁ , x ₀ , x ₁ ) {goto end;} / * collation-step 2-* / for (σ = σ ₀ , err0 = 9999; σ ≦ σ ₁ ; σ ⁺⁺ ) {/ * in the case of the standard deviation values σΣ, X ₀ of the reference and input waveform, X between _1, x _0, matching feature position X between x ₁ [i], and x [i], the error e at that time Seek r. * / Lesq (σ, Σ, X ₀ , X ₁ , x ₀ , x ₁ , X [i], x [i], err); / * Stores each data of the minimum error (1 ≦ i ≦ N) * / If (err <err0) {/ * large error * / goto end;} / * recursive matching-step 3-* / for (i = 1; i ≦ N-1; i ⁺⁺ ) {MATCH (Σ ₀ , Σ ′, X ′ [i], X ′ [i + 1], σ ₀ , σ ′, x ′ [i], x ′ [i + 1]);} end:}

Step 1 of this program is to determine a collation stop condition. When this program is called recursively and reaches the lower limit of the standard deviation value, the execution of this program ends. Further, in step 2, for each combination within the range of the standard deviation value, in the subroutine lessq, collation is performed while extracting the data of the previously obtained reference feature position and the feature position from the memory 4 or the memory 5,
The matching reference characteristic position X [i], the characteristic position x [i], and the error err at that time are output. Next, of the data thus obtained, only the data having the smallest error are matched with the standard deviation values Σ ′ and σ ′ of the coincident characteristic position X ′ [i] x ′.
[I] is stored. If the minimum error itself at this time is larger than a certain value (const), since the deformation between the two waveforms is too large, the obtained data is invalidated, and the comparison at this stage ends. In step 3, the matching feature positions X '[i] and x' [i] obtained as a result of the previous step are used to re-match the partial regions of both waveforms. The upper limit values of the standard deviation values at this time are Σ ′ and
σ ′, and more detailed waveform matching is performed. The initial values of this program are Σ ₀ = Σ _s , Σ ₁ = Σ _e , X ₀ = X _s ,
Let X ₁ = X _e , σ ₀ = σ _e , σ ₁ = σ _e , x ₀ = x _s , x ₁ = x _e .

Next, a subroutine l in the above program
n reference feature positions (XΣ ₁ ,..., X ， _i ,..., X の_n ) of the reference pattern with the standard deviation Σ corresponding to esq
(However, X ₀ ≦ XΣ ₁ <・ <XΣ _i <・ <XΣ _n ≦ X
₁ ) and m characteristic positions (xσ ₁ ,..., Xσ _i ,..., Xσ _m ) of the pattern to be measured having a standard deviation σ (where x ₀
<Xσ ₁ <·· <xσ _i <·· <xσ _m ≦ x ₁ ) The matching method will be described. Here, n reference feature positions (XΣ ₁ ,..., XΣ _i ,..., XΣ _n ) of the reference pattern (FIG.
(A)) are subjected to coordinate transformations for n ′ points, and these are converted to m characteristic positions (xσ ₁ ,... X) of the pattern to be measured.
σ _i ,... xσ _m ) (FIG. 4 (b)) corresponds to n ′ (however, 3 ≦ n ′ ≦ min [n, m]). Assuming that the coordinate transformation equation is a linear equation of the input variable X, the polynomial T (X) when minimizing the squared error δ as in the following equation is the optimal coordinate transformation equation.

[Equation 4] However, T (X) = A ₀ + A ₁ X Note that A ₀ and A ₁ are constants, A ₀ represents a change due to a position shift, and A ₁ represents a change due to magnification, strain, and the like. By the above coordinate conversion formula, gain fluctuation, deflection distortion,
It can also correct for fluctuations such as beam drift. Then, for each combination taking n ′ from the reference characteristic position of the reference pattern and the characteristic position of the pattern to be measured, the minimum 2
The square error δ at that time is obtained by applying the multiplication method,

(Equation 5) Finds the smallest combination (α: constant). However, n 'itself is also changed from 3 to min [n, m]. simply,
Instead of finding the minimum value of δ, the weighting factor

(Equation 6) The reason that the error is evaluated is that the error δ inevitably becomes smaller as n ′ becomes smaller, so that the error is properly evaluated. Therefore, the above coefficients may be different as long as this purpose is realized.

FIG. 2 is a flow chart showing the collation procedure described above. In the first step, a feature position whose standard deviation value is larger than X _{0 and} smaller than X ₁ in Σ is extracted. In the second step, the initial combination element variable n '
In the 3, it leaves substituting error over large value possible in the variable err, n pieces of reference feature positions of the reference pattern in the third step _{(XΣ 1, ··· XΣ i,} ···· XΣ n) from n ′ and m feature positions (xσ ₁ ,.
·· xσ _i, choose n 'number from ···· xσ _m). In the fourth step, the least-squares error δ and δ ′ = α · δ / (n ′) ² in this combination are calculated, and only when δ ′ <err in the fifth to sixth steps, err =
After setting δ ′, the combination at that time is output argument X [i],
x [i]. In the seventh step,
Check whether all the combinations that take n'have finished,
If not completed, return to the fourth step and repeat the same process. When the processing is completed, n 'is increased by 1 in the eighth step, and the same operation as described above is repeated until n' = min [n, m]. FIG. 5 is a diagram showing the correspondence between the final feature positions obtained by this method. If this correspondence is known, the edge position (X
d) can be easily calculated by using an appropriate interpolation or the coordinate conversion formula T (X) in the partial waveform including the interpolation. Then, the result is displayed on the display device 8.

In the present embodiment, an example of collation for a one-dimensional waveform is shown, but the present invention is not limited to this, and can be easily applied to collation of a two-dimensional waveform. This will be described below. 6 (a) (b)
Is a binary reference line image P (X, Y) and an input line image p
FIG. 6 is a diagram illustrating an example of collation of (x, y). Here, the line image refers to an image drawn with a line having a width of one pixel. Such a line image can be expressed by functions X = S (L) and Y = T (L) using on-line coordinates L from a starting point on a certain line as parameters. Therefore, the functions after applying a Gaussian filter of each standard deviation value (Σ _sフ ィ ル タ Σ Σ _e ) to the functions S (L) and T (L), S ′ (L; Σ) and T ′ (L; If Σ) is obtained, the shape of the line image represented by the functions S ′ (L; Σ) and T ′ (L; Σ) at that time will be global or detailed depending on the magnitude of the standard deviation value. Therefore, the curvature Θ = 0, or the L coordinate LΣ ₁ , satisfying dΣ / dL = 0, which can be calculated using the functions S ′ (L; 座標) and T ′ (L; Σ) at each standard deviation value. LΣ _i, lσ ₁ reference feature coordinates ···· LΣ _n, similarly obtained input line image, ··· lσ _i, ·
... If lσ _m (1: on-line coordinate) is used as the characteristic coordinate, it can be reduced to the above one-dimensional waveform collation method.

Naturally, the line image is decomposed into partial line images by using the result of the first collation, and the collation in the same process is performed again for this partial line image. If the line image is a closed loop, it is necessary to allow the correspondence beyond the starting point when associating the reference feature coordinates with the feature coordinates. In that case, the perimeter of the line drawing is added to the exceeded coordinates. In this manner, the global shape and the detailed shape of the line image can be compared on each line, and the line image having the deformed shape can be compared. FIG. 7 is a diagram showing a correspondence relationship between final characteristic positions obtained by the above method. In each of the above embodiments, a Gaussian filter is used as a filter. However, another simple averaging filter may be used as long as the filter can describe a general or detailed waveform.

As described above, in the present embodiment, the criteria 1
An apparatus for measuring the width of a semiconductor pattern by comparing a dimensional waveform F (X) with an input one-dimensional signal waveform f (x) and detecting a specific position of the input one-dimensional signal waveform f (x). , At each position of the reference one-dimensional waveform F (X),
The first step of performing filtering with a certain filter size Σ to obtain an approximate waveform G (X), and the approximate waveform G
In (X)

(Equation 7) The second step of obtaining X coordinates (X ₁ , ... X _i , ... _Xn ) satisfying the above as standard feature coordinates and each position of the input one-dimensional signal waveform f (x). In the third step of performing the filtering of the filter size σ to obtain the approximate waveform g (x), and in the approximate waveform g (x)

[Equation 8] The fourth step of obtaining x coordinates (x ₁ , ... X _i , ..., X _m ) satisfying the _above as characteristic coordinates and the reference characteristic coordinates or the characteristic coordinates when there is no characteristic coordinate correspond to the collation result. The fifth step of setting no position, and T (X) being a coordinate conversion expression expressed by a polynomial of X, and the number of the reference feature coordinates (X ₁ , ... X _i , ... X _n ) , And a combination (X ₁ , ... X _i , ... X _h ) of some of the feature coordinates (x ₁ , ... X _i , ... X _m ) and (x ₁ , ...・
· X _i, ···· x _h) of the square error

[Equation 9] If the error δ at this time is less than or equal to a certain value, the corresponding position is taken as the matching result, and the error δ
When the value exceeds the certain value, it is possible to automatically detect the start point and the end point by the procedure having the sixth step of making the matching result no corresponding position and measure the semiconductor pattern width by the difference. it can. Therefore, the semiconductor pattern width can be accurately measured even when the compound circuit patterns overlap and the one-dimensional waveform near the edge has a plurality of peaks and troughs. Further, in the present embodiment, the inflection point of the approximate waveform is detected as the position for detecting the start point and the end point, but the inflection point is hardly affected by the noise because the staying time of the noise is short. There is an advantage.

[0019]

As described above, according to the present invention,
Since the start point and the end point can be automatically detected even in a semiconductor pattern having a complicated edge waveform without relying on the conventional peak detection method or linear approximation method, the line width of the semiconductor pattern can be accurately measured.

[Brief description of drawings]

FIG. 1 is an operation flowchart of reference pattern processing and measured pattern processing according to an embodiment of the present invention.

FIG. 2 is an operation flowchart for performing a matching process between the two in FIG. 1;

FIG. 3 is an overall block diagram of a semiconductor pattern width measuring apparatus showing an embodiment of the present invention.

FIG. 4 is a diagram showing waveforms of a reference pattern, a measured pattern, and their approximate waveforms.

FIG. 5 is a diagram showing a correspondence between final characteristic positions.

FIG. 6 is a diagram showing an example of collation of a binary reference line image and an input line image.

FIG. 7 is a diagram showing correspondence between characteristic positions in FIG.

[Explanation of symbols]

1 ... A / D converter, 2 ... Address generation circuit, 3 ... Image memory, 4 ... Reference pattern data memory, 5 ... Measured pattern data memory, 6 ... Calculation procedure data memory, 7 ... C
PU, 8 ... Display device, 9 ... Electron microscope.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G06T 7/60 9061-5H G06F 15/70 350 J (72) Inventor Hidenori Inouchi Higashi Koikeku, Kokubunji, Tokyo 1-280, Central Research Laboratory, Hitachi, Ltd.

Claims (4)

[Claims] 1. A signal conversion means for scanning a prepared semiconductor pattern and a semiconductor pattern to be measured separately from the prepared semiconductor pattern with an electron microscope, and converting an obtained image signal from analog to digital, and conversion by the signal conversion means. An image memory for storing the obtained digital signal, an address generating means for generating an address of the image memory using a scanning synchronization signal of an electron beam, and processing the image stored in the image memory. A computer for extracting data necessary for an edge position of the pattern and the semiconductor pattern to be measured; a first memory for storing waveform data relating to the semiconductor pattern obtained by the computer; and a semiconductor pattern to be measured obtained by the computer A second memory for storing input waveform data relating to In order to check each signal of the semiconductor pattern and the semiconductor pattern to be measured, after removing noise of each signal with a Gaussian filter, a standard deviation value of the Gaussian filter is changed to obtain each approximate waveform, and each approximate waveform is obtained. The coordinates of the inflection point of each pattern are obtained, the inflection point pair corresponding to each approximate waveform of each pattern is detected by collation, the corresponding points of both patterns are obtained, and the corresponding points are set in the semiconductor pattern. And a third memory for storing a series of procedures for detecting a start point and an end point of the semiconductor pattern to be measured corresponding to the start point and the end point, and measuring a width from a difference between the end point and the start point. Pattern width measuring device. 2. A first step of obtaining an approximate waveform by performing filtering of a certain filter size at each position of a reference one-dimensional waveform in the third memory, and an X coordinate of an inflection point in the approximate waveform. A second step of obtaining an approximate waveform by performing filtering of a certain filter size at each position of the input one-dimensional signal waveform, and a step x of an inflection point in the approximate waveform. A fourth step of obtaining coordinates as feature coordinates;
In order to eliminate the influence of the difference in magnification and the distortion caused between the two-dimensional waveform and each approximated waveform of the input one-dimensional signal waveform, a predetermined coefficient of a coordinate conversion expression expressed by a polynomial of X is assumed, and each coefficient is assumed. A fifth step of specifying a coefficient that minimizes the error of the coordinate conversion equation including the above, and a sixth step of obtaining a corresponding set of reference characteristic coordinates and characteristic coordinates in each approximate waveform including the coefficient.
And a seventh step of obtaining a corresponding set of both signal waveforms from the corresponding set in the approximate waveform and detecting a specific position of the input one-dimensional signal waveform based on the corresponding set. The semiconductor pattern width measuring device according to claim 1, wherein 3. In the procedure for obtaining a corresponding set of characteristic coordinates in the sixth step among the procedures stored in the third memory, the filter size is reduced for each corresponding set of each partial waveform. 3. The semiconductor pattern width measuring apparatus according to claim 2, wherein the collation is performed again on the partial region of the waveform to perform more detailed waveform collation. 4. In the procedure for obtaining the coordinates of the inflection point among the procedures stored in the third memory, the X coordinate at which the differential value of the approximate waveform becomes zero, or the second differential or higher differential 3. The semiconductor pattern width measuring device according to claim 1, wherein the X coordinate that becomes zero is used.