JPS62201304A - Measuring method for film thickness - Google Patents

Abstract

PURPOSE:To obtain a device which has a small error in measurement by passing a quantized reflection factor through a digital filter and correcting a phase according to its frequency. CONSTITUTION:Detectors 11A and 11B outputs the quantity of projection light and reflected light as electrical signals, which are amplified by amplifiers 12A and 12B and then inputted to A/D converters 14A and 14B. A wave number scanner 13 scans a preset number of waves to irradiate a sample with homogeneous light and data sequences which are quantized and digitized according to the pulses are inputted to the CPU of a computer. A reflection factor calculation part 15 calculates a reflection factor and a digital filter part 16 removes high- order components. Then, a frequency and phase detection part 17 detects the frequency of data roughly and a phase correction part 18 corrects the phase delay of the data according to the detected frequency. A film thickness calculation part 19 calculates film thickness by using the data.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Application) The present invention relates to a method for measuring the thickness of thin films, particularly micrometers or less, such as oxide films on silicon in semiconductor manufacturing processes.

The basic 311 constant principle for film thickness calculation used in the present invention will be explained below with reference to FIG.

The refractive index of the medium is nl, n2 . Let n3 be the thickness to be measured now and be d. The incident angle of each wavelength used (wavelength incidence in vacuum is 0) is OL, 02.03. The amplitude reflectance γ at this time is as follows. [Reference: M. Born and E. Wolf, “Pr1n
ciples of 0ptics”3rd e
dition, PERGAMON PRESS, 62L
tl, where γ12 is the Fresnel reflection coefficient at the boundary between mediums l and 2, and γ23 is at the boundary between mediums 2 and 3), and the measurable quantity is the reflection intensity, that is, R=lγ12 (
(usually called reflectance) and is of first order.

From formula (2), Il! 2 The thickness d is as follows.

−=−=−−−(3) Here N is an integer.

The reflectance R and d when changing the wavelength input 0 used are shown in Figure 1 (
b) and (C). Therefore, the film thickness d at each wavelength
By setting the film thickness to the average value dAV, highly accurate measurement with good reproducibility can be performed.

FIG. 2 is a block diagram of a system according to an embodiment of the present invention. The fiber probe 3 illuminates the light separated by the spectrometer 1 onto the sample surface placed on the sample stage 21, and the fiber probe 3 receives the light reflected from the sample surface again. The photoelectric conversion detector 11B converts it into an electric signal.The controller 4 controls the wavelength setting of the spectrometer l and the movement of the sample stage 2, and converts the electric signal of the photoelectric conversion detector 11B of the spectrometer l into a digital signal. Then, the position of the sample stage 2 is confirmed.Furthermore, IIA is a light emitting detector that detects the amount of light emitted to the fiber probe 3 in the spectrometer 1.A computer 5 controls the controller 4 and It receives the digital signal obtained from , calculates the reflectance R, and calculates the film thickness d from the equation (3).The results are output to the CRT 8, floppy disk drive 9, and printer 10.Operation panel 6 and full keyboard 7 The computer 5 specifies conditions such as the refractive index, the wavelength of the spectrometer 1, the amount of movement of the sample stage 2, and commands to start measurement.

The operation panel 6 is used for measurements under commonly used conditions, and has the minimum number of keys required to prevent the operator from operating the device erroneously.

From equation (2), the reflectance R is proportional to the period of 2β=2π.

In other words, the reflectance R is a periodic function of =2n2d・cos θ2, and has peaks at equal wavenumber intervals.
In calculating the film thickness in this embodiment, calculation is performed using this wave number.

By the way, in order to measure the film thickness, it is necessary to know the spectral reflectance between certain wave numbers, and therefore it is necessary to change the wave number and determine the reflectance for each equal wave number interval. At that time, W+1
Some of the major factors that cause fixed errors are as follows.

(1) Electrical noise that occurs when converting, amplifying, and quantizing the amount of light into an electrical signal in order to detect the amount of light.

(2) Changes in light intensity caused by equipment that measures the amount of projected light and reflected light, or by vibration of the sample.

If film thickness calculations are performed using data that includes such errors, the calculated film thickness will also include errors, and during the calculation process, it may be possible to find an incorrect point when determining the wave number at which the reflectance takes an extreme value. There is a possibility that it will be judged as an extreme value.

One way to remove these errors is to use an analog filter that cuts frequencies above the required frequency before performing A/D conversion, but in this method, the frequency components of the data depend on the wave number scanning speed, so You need to change the cutoff frequency of the filter,
In addition, there was a problem in that the A111 constant had to be scanned at a constant speed. In other words, even if the reflectance is the same, the scanning speed is doubled.

The scan time is halved and the frequency components are doubled. Furthermore, if the scanning speed changes during the measurement, the frequency components will be higher at faster speeds and lower at slower speeds.

Another method is to measure multiple times and use the average, but this method takes time to measure and cannot eliminate errors that occur in synchronization with measurements, such as vibrations that occur during scanning. There was a problem.

(Problems to be Solved by the Invention) It is an object of the present invention to eliminate the above-mentioned problems and provide an all-reflectance setup with less measurement error.

(Means for Solving Problems) The present invention mainly includes a digital filter in order to solve the above-mentioned problems. The digital filter consists of a filter section and a phase correction section.

Furthermore, when passing through a filter, a phase lag occurs, which must be corrected, but generally the phase lag is not exactly linear with respect to the frequency, so there are cases where the frequency of the target signal is unknown or the signal varies. Accurate correction cannot be made if the signal contains frequency components. However, in the spectral reflectance of a sample with a film, it can be considered that a specific frequency is superimposed on an extremely low frequency. By determining the amount and correcting it, almost accurate correction becomes possible. In reflectance measurement, the phase shift is a wave number shift, so the wave number used to calculate the film thickness differs from the actual measured point, but the film thickness meter 'f1
There is no problem in the north.

(Description of structure and function) FIG. 3 shows an uninvented embodiment in which the amount of light projected and the reflected light hole are converted into electrical signals by detectors 11A and 11B, respectively.
After being amplified by amplifiers 12A and 12B, it is input to an A/D converter. The wave number scanner 13 scans between preset wave numbers to irradiate the sample with monochromatic light, and sends conversion timing pulses to the A/D converters 13A and 13B at equal wave number intervals. Each data string that has been spermized and digitized according to a timing pulse from the wavelength scanner 13 is taken into the CPU, and the reflectance x (i) is calculated at 15 .

16 passes through a digital filter to remove high-order components.

Next, the approximate frequency of the data is detected in the frequency detection step 17, and the phase delay of the data is corrected in accordance with the detected frequency in the phase correction step 18. Using the data, the film thickness is calculated in step 19.

is calculated, reflection: Bx data string x (1) ~x
(n) is calculated. K(+) is a reflectance correction i+Q and is a constant. This situation is shown in FIG. 4(A).

New reflectance data sequences y(1) to y(n) are obtained by applying the data sequences x(1) to x(n) to a digital filter 16 that removes frequencies similar to a predetermined frequency. This is shown in FIG. 4(B). Here, the predetermined frequency is preferably set to the highest frequency obtained from an ideal spectral reflectance. This is because frequency components north of that point can be considered noise. As shown in Figure 4 (B), the data string y (1) to y (n) passing through the filter is shown in Figure 4 (A).
It can be seen that compared to x (1) to x (n), the data is rounded, that is, high-order noise components have been removed, and the total 'J'1 error is reduced.

Furthermore, in order to correct the phase of this data string y(1) to y(n), it is applied to the frequency detection unit 17, and as described in 17i, a specific frequency component is detected, and the amount of phase d is calculated for that frequency. The phase is corrected in the correction section 18. For example, FIG. 4C shows how the calculated delay j, I=P, is corrected.

An example of a digital filter is the second-order Butterworth (B
The calculation formula when using the Utter Worth) filter is shown below.

The transmission function aH(S) of this filter is expressed as (Wc: each cutoff frequency, S: jW). Let x (n) and y (n) be the nth input data and output data, respectively.

The following equation is obtained by Z-transforming and rearranging equation (4) and performing inverse Z-transformation.

A= 1 +, /'7waC+W, 1 (2B=2-W,
c2-2 C= 1-, /TWac+WaC2(5)[)=Wa(
2 The output data string y(n) is obtained by equation (5). (5
) Since the data sequence v (n) obtained by the equation includes a phase delay, this must be corrected.

(4) Equation (7) H(S) ヲZ transformation, z=ejw
It can be found by substituting wo and taking the inverse nE tangent of the ratio of the imaginary part to the real part, and the formula is as follows. W here is the angular frequency of the signal, and the value obtained by frequency detection in step 17 is entered. Since θ is expressed as an angle, it is necessary to convert it into a wave number. The wave number 8 DW is found by θ(W) DW=-(7).

The following processes are performed by the frequency 1 phase detection section 17.

(Effects) According to the present invention, in the spectral reflection specification 'A11 constant of the film thickness measuring device, by passing the quantized reflectance through a digital filter and performing phase correction according to its frequency,
High-order noise on reflectance data can be removed without depending on the wavenumber scan cannon speed, and at the same time, accurate film thickness calculations are possible because the phase is corrected according to the frequency of the data.

[Brief explanation of drawings]

Figures 1 (a), (b), and (c) are diagrams explaining the principle of film thickness calculation used in the present invention, Figure 2 is an overall block diagram of an example of the present invention, and Figure 3 is a partial detail thereof. The block diagram and FIG. 4 are data diagrams for explaining its operation. 11A and IIB are light emitting parts and light receiving detectors, respectively, 14A and 14B are A/D converters, 15 is a reflectance calculation part, 16 is a digital filter part, 17 is a frequency component detection part, and 18 is a phase correction part. , 19 indicates a film thickness calculation section. Town λ Mixed ■λ

Claims (1)

[Claims] 1. In an apparatus for measuring the film thickness on a sample from the spectral reflectance of the sample, after removing high-order noise components by removing frequencies above a predetermined frequency from the spectral reflectance measurement data. A film thickness measurement method characterized by calculating film thickness. 2. The film thickness measuring method according to claim 1, wherein an approximate frequency is detected from the data from which the high-order noise components have been removed, and phase correction is performed accordingly.