JPS6273103A - Film-thickness measuring method - Google Patents

Abstract

PURPOSE:To measure the thickness of a film, by measuring the spectroscopic characteristics of the reflected optical component through an optical window and the spectroscopic characteristics of the reflected optical component from a reflecting plate beforehand, subtracting a window signal from the measured signal and a Planck signal, and performing normalizing operation. CONSTITUTION:Parallel light, which is formed by a light projecting part, is made to pass an optical window 12 and projected on a plate measuring film 15. The light, which is reflected by the film 15, becomes parallel light and enters a planar diffraction grating 7. The image of the light in the specified wavelength range among the lights, divided by the grating 7, is formed on an image sensor 9. The waveform of the image having a specified spectroscopic intensity is sent to an operating part (b). The spectroscopic intensity W(lambda) is measured when nothing is placed as an object to be measured. The spectroscopic intensity B(lambda) is measured when a reflecting plate is placed. The spectroscopic intensity F(lambda) is measured when a thin film is placed. The operation of A(lambda) is carried out according to the expression in the Figure in a processing part (a), and the thickness of the thin film is measured.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a method for measuring the thickness of a thin film. The method according to the invention is particularly characterized by the ability to perform membrane piece measurements with high precision.

(Prior Art) Conventionally, it has been known to use light interference phenomena as a method of measuring the thickness of a thin film (Jibun Sho 56-115).
905, etc.). In this method, white parallel light enters a thin film at an angle of 13 The waveform of (a) is calculated, and we decide to find the wave 1 mo corresponding to the adjacent (he mermaid (bL < +A-, j, small dot) of this spectral intensity waveform n'i: I' F”
It is based on the principle of making a bell.

This principle will be explained below using mathematical formulas.

When white light is incident on a thin film at an angle θ, it is divided into light that does not enter the film but reflects on the surface, and light that enters the film and then reflects on the back of the film and is emitted from the surface. .

The reflected light is other than the light mentioned above, and there are many QJ inside the film.
Since there is light that comes out from the surface after the light is released, or the intensity is weak, we decided not to consider it here.

An optical path difference Δ occurs between the light reflected on the film surface and the light reflected on the back surface, and Δ is expressed by equation (1).

Δ=2d5■7]7[...(1) Then, as the above two lights change, the spectral intensity becomes stronger or weaker, or the optical path difference Δ becomes an integer multiple of the wavelength. The spectral intensity at the matching wavelength decreases to (integer +] / 2
) times the spectral intensity becomes (~) (This is because the phase of the light reflected from the back surface is reversed, and in the case of transmitted light, there is a relationship between phase reversal and leverage. is reversed).

In the spectral intensity waveform obtained in this way, the wavelengths of the two adjacent mermaids (or minimum points) at J3 are λ and λ.
2, the following equation (2) holds true (where λ1>λ
2).

By rearranging equations (1) and (2), the following equation (3) can be obtained.

In other words, by finding the wavelengths λ and λ2 of two adjacent (or mermaid (or) small points) in the spectral intensity waveform, the film thickness can be calculated by 61 υ based on equation (3). .

) 1 is '12' according to the method to find the wavelength corresponding to the car mermaid
The method described in Japanese Patent No. 135331/1983 is known. This method uses an image sensor as a detector of the waveform of the spectrometer's 1 attenuation, A/D converts the output of the image sensor, and imports it into the micro combicoater. 11F is determined from the output value of the cell after I'llJ of that cell.

(Invention or problem to be solved) However, is this measurement actually 1? The resulting waveform is shown in Figure 2 (it is not a round waveform, it is shown in Figure 3(a)). , the measurement system waits for the wide spectral characteristics shown in FIG. 3(b) as the total time 11, so when a thin film is measured with this measurement system, the film thickness is measured at 11 in FIG. 3(b). Interference 1h based on the 1st sample of the bass film answer
The waveform shown in Figure 3 (a) is inevitable. Therefore, if the maximum point and the minimum point of the spectral intensity waveform are found based on the waveform of FIG. 3(a), only the correct IIs!thickness can be measured.

As a solution to this problem, a method disclosed in Japanese Utility Model Application Publication No. 58-72610 is known.

This is fJ, shown in Figure 3(b) IF+! The spectrum of the optical system excluding 17T is B(λ) (hereinafter referred to as a blank signal), and the spectral characteristics during thin film measurement shown in FIG. ), A(λ)-F
(λ)/B(λ) A modulation signal A(λ) corresponding to Fig. 2 is obtained by the calculation, and the film thickness is measured from the wavelength position of the minimum point of this modulation signal A(λ). I'll go.

In principle, there is no problem with the above method, but in practice this method also has the following drawbacks.

That is, an actual film 17 measuring device is a skewer that houses an optical system in one segment and projects and receives light through an optical window 12, as shown in the schematic groove diagram of FIG. In this case, the optical window 12 is indispensable for dust-proofing, moisture-proofing, etc., but the intensity of the reflected light inside due to the optical window becomes too strong to ignore.

In the system shown in FIG. 1, when measuring without placing any object to be measured, the spectral characteristic W(λ) (hereinafter referred to as
7 wind signal thorns. ) is obtained.

Also, the blank signal B(λ) obtained with this system is B'
(λ) with the optical window removed [as a cow, B(λ)=8'(λ)+W(λ).

Similarly, the measurement signal F(λ) obtained with this system is F
' (λ) is the spectral characteristic through the thin film when the optical window is removed, then E (λ) = F' (λ) + W (λ).

Now, the modulation signal we want to obtain here is F'(λ)/B'(λ), or 1°△(λ)-F(λ)/B(λ)(F'( λ)10W(λ)) (B'(λ)+W(λ>)), and the true modulated signal cannot be calculated by 1q.

For this reason, the correct maximum and minimum points of the interference waveform were not determined, which caused errors. This phenomenon is approximately F'(λ):W(λ)=1:0.3 for a thin film with a small amplitude of the interference waveform and a large film thickness (6 to 7μ or more), W </! , > of) ii threet15p I
I'm with J Kenshu.

The purpose of the present invention is to solve the problems of the prior art by using the fourteen phenomena of light.
1) If the thickness of the film 1'7 is measured at l'n'i and the film 1'7 is measured at 4 degrees high, it is considered to be '1!1 mold'.

In other words, the present invention aims white light at a constant human QJ angle through an optical window on the pleura, collects the reflected light through the optical window, and then spectrally it to obtain a waveform of the spectral intensity. The interference phenomenon is measured from the wavelength position of the strength and weakness of the spectral intensity! )
d3 on how to measure thickness, Hiroshi 11! , anti-0 by 3
The spectral intensity of the 4 lights is F2(λ), the spectral intensity of the thin 11F, the spectral intensity of the 0'J tentatively placed in place of 3 is B(λ), and the spectral intensity of the light is B(λ), and the spectral intensity of the light is B(λ). The most spectral strength is W(λ)
An IB>thickness measurement method in which the 13 characteristics are to measure the film thickness from the strong and weak wavelength positions of the modulation signal Δ(λ) obtained based on the formula.

It provides:

(Means for Solving the Problems) A case in which the method according to the present invention is applied to an optical interference type film thickness meter will be described with reference to the drawings. In the following explanation, the membrane Q measurement target is a film, but the method according to the present invention is of course not limited to films.
It is also applicable to measuring the thickness of glass thin films and other thin films.

Figure 1 shows the optical interference type +1! +! Jri measurement system as a whole! i+
! It consists of a measuring system (J), a measuring section having a light projecting section and a light receiving section, and a calculation processing section.

It is composed of a light projection part iJ, a light source 1, a pinhole 2 for forming parallel light, J5, and a lens 3.

The light receiving section includes a condensing lens 4, a pinhole 5 for forming parallel light, a lens 6, a plane diffraction grating 7, an imaging lens 8,
An image sensor 9, an image sensor drive circuit 10, and a buffer amplifier 11 are grown. Also, an optical window 12 made of quartz glass is installed in the measurement section including the light projecting section and the light receiving section.

Further, the programmable speed embedding section (YA), 'D converter' 13, microcomputer 141.BJ: and input/output devices, which are omitted in this figure, are connected to the input/output device.

The parallel light (J) formed in the light projecting section is projected onto the film to be measured 15 through the optical window 12, and is reflected by the film to be measured 15. The pinhole 5 is focused and centered at the focal length of the condenser lens 4, and then the lens 6 is rounded by δ (the distance between the lens 6 and the pinhole 5 (J, equal to the focal length of the lens 6). ), the plane of light that is anti-CIJ on the film 15 to be measured (:'i acid component Mikahira ()) is reflected by the name j'-7. Planar diffraction Of the light separated by Rakuj''-7, a predetermined wavelength range is placed on the image sensor 9+!'ll'l A lens collection lens ε3 is placed so that the image is formed on the image sensor 9. Waveform of spectral intensity (Mi1, image sensor drive circuit 1
0 is sequentially read out for each rail and sent to the performance burying section via the buffer 1 amplifier 11.

In the raccoon section of the performance department (well, this signal is the A y' D converter) 3
(This word is read into the microcomputer 14, and processing is performed.

Next, FIG. 4 shows a flowchart of wavy line processing, which is the main part of the method according to the present invention.

ipj pa
), measurement of the wind signal and blank signal (t) performed before measuring the sample, measurement of the 4-number number (C), and the calculation of the data (done).

Here, ``Rindo (with issue 1g), the first
In the measurement system shown in the figure, a skewer of the spectroscopic light reflected by the optical window 12 is placed at the measurement position.

In addition, the blank signal is the spectral property [JU's skewer] of the entire optical system measured by placing a suitable reflector in place of the film 15 on the measuring rack in the measuring system shown in FIG.

The actual move ;;1j″i will be explained below.

It is better to measure the wind signal or to measure with nothing in the measuring position. Let this result be W(λ).

Next, to measure the blank signal, prepare a plate made of the same +A material as the film to be measured and thick enough to be unaffected by interference currents, place it at the position of the film, and measure. Let this be B(λ).

In this case, the reflectance plate may be a mirror, but the reflectance between the mirror and the f-number is very different, so the dynamic range during ripple measurement becomes narrow. Therefore, it is better to use a reflector made of the same material as the sample.

Also, this is done before starting the India measurement and blank measurement, but is it possible to check the change over time of the light source lamp? If
It is preferable to periodically re-measure for compensation. - Then perform the sample measurement.

After placing the film 15 to be measured at the measurement position δ, the image sensor 19 is activated according to the command from the microcomputer 14.
The output is sent to the four-node processing unit via the buffer 7 amplifier '11, and read out sequentially via the A and /'D converters 13 to be measured (;-E-SF(λ)).

The procedure for measuring W(λ) and B(λ) is also the same as the procedure for measuring F(λ).

Next, we will explain data processing.

As shown in FIG. 4, data integration is performed in the following steps.

(a), ゞI'-sliding (11), f-in-in 1 (i!/Shi catch I
F(c), i′I manifestation (d), Yamaya (1 Kurumajin, ) to small) Samurai head detection (C), escape)
'2 performance i; East and below, each /"l] footer provides a detailed explanation of the Noh.

First, smoothing 1H is performed. Is there an f method for smoothing l(J, +J, etc.? It's good to have a similar 0.1 interval difference like a moving average of J11. Also, the point 1 to Select teaching as 2 (n=1.2,...) and select J
'3 allows v1 to be shifted by pitch j~, which is preferable in terms of speeding up. '3. At this time, the image sensor (
It is preferable to create a dark cell in which parts other than the jUt portion are electrically sensitive to light, and then subtract the output of this dark cell from each output to compensate for the dark current of the image sensor.

Data smoothed in this way is called smoothed data7.
ball.

This smoothing necessarily removes minute noises, resulting in a smoothed signal QF(λ). Note that this smoothing operation is ``11 also for the wind signal W (λ) and the blank 1 word bow B (λ), so that J'3<.

J′3, this smoothing is essential and CJ, no-Czu)
(- is fine. From now on, F(λ) and "(λ)(!!-) will be written as F(λ) to distinguish between them.

Next, the wind signal is corrected.

However, the smoothed signal B(λ) and the blank signal B(λ)
Then subtract the wind signal W(λ) according to equation (4),
True measurement signal F'(λ>, true 1-rank signal B'
(λ) is multiplied by 1q.

Next, the modulated signal A(λ) is multiplied by 1 according to equation (5).

A(λ)=F'(λ)/B'(λ)...(5
) This is called normalization, and by this operation an interference waveform corresponding to FIG. 2 is obtained.

Next, it is necessary to identify the peak and valley (maximum, minimum) positions of the interference waveform, but by performing operations up to the previous step, an interference waveform with little noise or noise at the edge was obtained, so the slope of the signal could be determined. Due to changes,
It is very easy to find the value position of the peaks and valleys by i
'E I can know for sure.

Here, the position of the value 4 to q is the cell number of the image sensor, but since the position of the Φ value can be sold to the wavelength, the relationship between the cell number of the image sensor and the wavelength is It is necessary to know the correspondence.This correspondence can be found by placing an interference filter with a known wavelength on the reflection plate when measuring blank data, and determining the peak position.

In this way, if you can find multiple maximum wavelengths, ) to small wavelengths, the adjacent ones (from the dwarf wavelength (or #l*small wavelength) (3)
Regarding M in the formula [4! I'5' ”: * I C read. At this time, it is also possible to combine multiple results and find the average value.

In addition, for cases where the spectral waveform enters the measurement wavelength region for one period or less, such as a very thin film, equation (3) should be changed.
? The film thickness is 81 cm with the maximum wavelength and
It is preferable to edit it.

(Function) As described above, the method according to the present invention (jl, optical system, microcomputer, etc., is used to produce a spectrum of light components reflected by an optical window, and to produce a spectroscopic light component of the reflected light component by an optical window, and to generate a spectroscopic light component of the reflected light component by an optical window.
The spectral property 1i of the anti-rAJ light component due to anti-rAJ is measured beforehand, the wind signal is subtracted from the measurement signal a3 and the blank signal, and then normalization is performed.1. The film j7 is calculated from the position of the extreme value.

(Example) Hereinafter, the present invention will be explained in detail with reference to Examples.

Using a halogen lamp as a light source, the blaze wavelength is sufficiently sensitive to 550-850 nm: 500 nm, groove a: 1200 lines/ml, effective area I:
Use a 30mm x 30mm /'T (1yo/L,
Image sensor January 31, CC consisting of 2048 reels
A type D was used. A full-scale 12-bit (/1096 steps) A/D converter was used. Rough smoothing was done using a simple moving average and a moving average of 23-8 points.

FIG. 5 shows the spectral waveforms of the wind signal and blank signal. This means that the characteristics of the optical system are not uniform, and
It turns out that Wind No. 12 has a size that cannot be ignored.

Figure 6 shows various types! 14! Regarding If, how to pay (
5A-72610) and the waveform of the modulation multiplier measured by the method of the present invention, and the film thickness measured by hi from the peak and valley positions of the waveform.

In addition, in this figure, (nominal ++S>thickness) is the so-called width-hung average, where a small amount of a certain area is measured and the thickness is returned by υ1.

All of the above steps are automatically controlled by a microcomputer.

(Effects of the Invention) As explained above, the film thickness measuring method of the present invention requires the J and Q4 light components to be reflected in the optical window, which is a concave circle that avoids low F during rapid burial. After measuring and recording F, L, and blank signals, subtract the wind signal from one word of measurement, perform normalization processing, and find the positions of peaks and valleys based on the modulated signal. Since it uses a method that calculates the thickness, highly accurate measurements can be made.

Note that this effect becomes noticeable when the film thickness becomes large (6 to 7 μm or more).

[Brief explanation of drawings]

Figure '1 shows the overall strategy of the measurement system for implementing the method according to the present invention. 'l' 4 diagram, Figure 2 IJ, 1% li et al. Figure 13(b) shows the spectral waveform of the film measured in January, and Figure 13(b) shows the spectral waveform of the optical system. []-1-1・- Fig. 5(a) shows the wind signal according to the embodiment of the present invention. Fig. 1, Fig. 5(b)
) shows the blank signal according to the embodiment of the present invention.
Figure 1; J, , II) Thickness: 4.25 μm, 7.4 μm,
Measurement of 12.2 μm 3-layer n1 film by conventional method
-74 is a comparison diagram. 1...Light source 2...Pinhole 3...Lens/1...Condensing lens 5...Hinhole 6...Remator lens 7...Plane diffraction + 8-element ε3...Ya Imaging sensor' 0...Image sensor (S)○...Image sensor drive circuit 11...Buffer l-amplifier 12...Optical window 13...A/[) conversion: self]4...・Micro Combicola 15...Measurement row ♀ Film 1-4 edition Applicant Toshi Co., Ltd.'Jtyzm, 4nK+=Ko 3,000! Mugi no 4N Koteki Cedar Figure 5

Claims (1)

[Claims] A thin film is irradiated with white light at a constant angle of incidence through an optical window,
A method in which the reflected light is collected through the optical window, and then separated into spectra to measure the waveform of the spectral intensity, and the film thickness is measured from the wavelength position of the strength and weakness of the spectral intensity caused by the interference phenomenon caused by the thin film, The spectral intensity of the light reflected by the thin film is F (λ), the spectral intensity when a reflector is placed in place of the thin film is B (λ), and the spectral intensity of the light reflected only by the optical window without anything is W (λ). ), and A(λ)={F(λ)-W(λ)}/{B(λ)-W(
A film thickness measuring method characterized in that the film thickness is measured from the wavelength positions of the strengths and weaknesses of a modulation signal A(λ) obtained based on the formula of λ)}.