JPS6385305A - Apparatus for measuring depth of fine groove - Google Patents

Abstract

PURPOSE:To accurately measure the depth of a fine groove on the way of a process, by forming optical route data from the spectrum signal due to the output of a spectroscope and calculating the depth of the fine groove on the basis of said data. CONSTITUTION:In a fine groove depth measuring apparatus used in inspection after the groove of a silicon substrate is formed by etching, the light emitted from a light source 1 passes through the lens system of a microscope to be condensed to the substrate 2 having a fine groove formed thereto. The reflected light from the substrate 2 is guided to a diffraction lattice 4 through the lens system and a slit 3 to be spectrally diffracted. The measurement of spectral intensity is performed at once using a multichannel type photodetector 5 without performing mechanical scanning. The signal obtained by the 5 is digitalized in a data processing part 6 to be operationally processed into a predetermined wave form. By this method, the depth of the fine groove can be accurately monitored on the way of a process.

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial Application Field> The present invention relates to an inspection technique for a VLSI manufacturing process, and more particularly to a technique for measuring the depth of microgrooves.

<Prior art> Currently, in order to realize megabit-class dynamic RAMt-%, three-dimensional
Dimensional device formation techniques are being developed. For example, instead of the conventional method of forming elements on the silicon surface, by forming fine grooves on the silicon surface and forming elements on the sides of the groove, the effective surface area is increased and the degree of integration is improved. There is a way to do it. Typical dimensions of the silicon trench formed by this method are as narrow as 1100 nm in width and several μm in depth in order to achieve higher integration, and the aspect ratio of the trench is (groove depth to groove width ratio) reaches 5 to 10. In order to manufacture a VLSI using this method, it is necessary to accurately measure the depth of this groove and perform process control based on that data.

A stylus type surface roughness measuring device has conventionally been used as a device for measuring the depth of a groove. In the case of the stylus type measuring device,
Since the radius of curvature of the tip of the stylus is on the order of several μm, there is a problem in that the stylus cannot enter the submicron groove as described above, making it impossible to measure the depth. Another disadvantage is that even if the stylus enters the groove, the substrate is likely to be contaminated because it is a contact measurement. In order to solve the above problems, the depth of the fine grooves is generally measured by splitting the substrate, scanning the cross section, and taking an electron micrograph (hereinafter referred to as an SEM cross-sectional photograph).

<Problems to be Solved by the Invention> The method of measuring the groove depth by taking SEM cross-sectional photographs described above requires destroying the sample, so
cannot be used constantly for inspection of manufacturing processes. Therefore, the groove depth has to be controlled loosely in the manufacturing process, which makes it necessary to estimate a large margin for the process, which also has a negative impact on the yield.

<Means for Solving the Problems> The present invention has been made to solve the above-mentioned problems, and it splits the reflected light obtained by irradiating light onto a substrate in which grooves are formed, thereby creating a multi-channel light-receiving element. The output waveform is calculated using a waveform processing technique such as fast Fourier transform or maximum entropy method to obtain an optical path difference. The present invention provides a micro groove depth measuring device that detects groove depth based on the optical path difference.

<Function> As described above, by measuring the groove depth by irradiating light and using information obtained from the reflected light, there is no need to contact the substrate to be measured, so there is no need to contaminate the substrate. Moreover, since there is no need to destroy the substrate, the yield can be improved.

<Embodiment> FIG. 1 is an embodiment of the present invention, and is a block diagram of a micro-groove depth measuring device used for inspection after trenching and etching of a silicon substrate. Light emitted from a light source 1 is focused onto a substrate 2 through a lens system of a microscope. The substrate 2 is machined with fine grooves.

The reflected light from the substrate 2 is guided to the diffraction grating 4 through the lens system and the slit 3, and is separated into spectra. The measurement of spectral intensity is carried out at once without mechanical scanning using a multi-channel type light receiving element 5. Multichannel photodetectors have the following advantages: (1) Spectral time is short (within 3 seconds including calculation), and (2) information is used efficiently (bright). In the wood example, the area from 400 nm to 800 nm is approximately 400 nm.
A linear multi-channel photodetector is used.

The signal obtained by the light-receiving element 5 is digitized in the data processing section 6, and then subjected to arithmetic processing using a waveform processing technique described later.

FIG. 2 shows the reflection spectrum intensity obtained from the multichannel light receiving element 5 for a pattern (sample 5-1) in which grooves with a width of 0.8 μm are repeated at a period of 1.6 μm, and FIG. 1.2μm wide groove is 2.4μm
Pattern repeated with a period of m (samp/L/5-2
), the reflection spectrum intensity obtained from the multi-channel/l' type light receiving element 5 is shown. Figure 2,
In Figure 3, the horizontal axis is the wave number, and the spectral area is approximately 20μ.
Hire Shimo.

A 10x objective lens is used for the microscope. By optimizing the numerical aperture of the optical system, even in a groove pattern with a large aspect ratio, the reflected light from the deep groove bottom and the surface will interfere, and the second
It was discovered that interference spectra such as those shown in Fig. 3 and Fig. 3 can be obtained.

Here, sample S-1 was obtained by removing 5i02 used as a mask material after etching, and sample S-1 was obtained by removing 5i02 used as a mask material after etching.
-2 is immediately after etching.

In other words, in Sump/L/S-1, one type of interference appears repeatedly, whereas in Figure 3, which shows the reflected spectral nuvector intensity of Sump/L'S-2, two or more types of interference appear repeatedly. A signal with . This is because the sample S-2 has the cross-sectional view shown in FIG.
Assuming the reflected light from the surface ■, the reflected light from the 5i02 and Si interface ■, and the reflected light from the groove bottom ■, there are three optical path differences corresponding to the SiO□ film thickness and the Si etching groove depth. It happens because it happens.

By the way, it is desirable to inspect the groove depth immediately after etching. This is because if the groove depth is shallower than the specifications, additional etching can be performed immediately. It is difficult to immediately obtain information about the groove depth from the interference spectrum shown in FIG. Therefore, we attempted a method to separate the signals and obtain the etching depth using a spectrum analysis method. In this embodiment, fast Fourier transform (hereinafter referred to as FFT) is employed as a signal analysis method, and is executed in the data processing section 6.

FFT numerical operations are described using the Panucal language.80
This was done using an object program compiled for 86 series CPU. This progera takes about 10 seconds. This is fast enough to perform inspections during the manufacturing process.

FIG. 5 shows the results of performing FFT on the reflection spectrum μ signals of FIGS. 2 and 3. FIG. The vertical axis indicates the reflection spectrum intensity, and the horizontal axis corresponds to optical path difference information. In addition, in the peak observed in Figure 5, sample/l'
If the peak of S-1 is named A1 sample IV and the two peaks of S-2 are named B and C, then peak A is S
This is the output that appears based on the interference between the reflected light from the i-substrate surface and the reflected light from the groove bottom, and the value on the horizontal axis indicates the optical path difference between the two reflected lights, and the peak B is as shown in Figure 4. It shows the optical path difference between the lights ■ and ■, and peak C shows the optical path difference between the lights ■ and ■ in FIG. Based on this optical path difference, the Si groove depth and the 5i02 film thickness can be obtained.

Table 1 shows the optical path difference obtained by FFT processing, the groove depth or 5i02 membrane thickness calculated from the optical path difference, and each measurement obtained from the SEM cross-sectional photograph for samples S-1, s, and -2. The results are shown along with the resolution. Here, the calculation was performed assuming that the refractive index of 5i02 was 1.45. - Table 1 [Unit: μm] □□□= Here, the resolution of Fourier transform is a quantity given by the theory of discrete Fourier transform, and corresponds to the reciprocal of the bandwidth of the signal before transform. Therefore, as long as spectroscopy is performed in the visible region, the resolution of Fourier transform is 800 in optical path difference.
It is large, around nm.

Therefore, as a second example of arithmetic processing of spectral signals,
The maximum entropy method (hereinafter M EM ) was adopted.

Similar to the above-mentioned FFT, MEM was also performed using an object program written in the Pascal language t- and compiled for an 8086 series CPU.

-, it is about 70 seconds. This is also fast enough to perform inspections during the manufacturing process.

FIG. 6 shows the results of applying MEM to the reflection spectrum signals of FIGS. 2 and 3. FIG. The vertical axis shows the reflection spectrum spectrum)/the intensity, and the horizontal axis corresponds to the optical path difference information.
In addition, among the peaks observed in Fig. 6, if the peak of sample S-1 is named AI, and the two peaks of sample S-2 are named B+ and C+, respectively, the peak AI
appears based on the interference between the reflected light from the Si substrate surface and the reflected light from the bottom of the groove, and the value on the horizontal axis indicates the optical path difference between the two reflected lights, and the peak Bl is between the light ■ and ■ in Fig. 4. The peak C1 shows the optical path difference between the lights ``■'' and ``■'' in FIG. Based on this optical path difference, the Si groove depth and 5i
02 film thickness is obtained.

Table 2 shows the optical path difference obtained by performing the MEM process, the groove depth and 5i02 film thickness calculated from the optical path difference, and each measurement result obtained from the SEM cross-sectional photograph, together with the resolution. Here, the calculation was performed assuming that the refractive index of 5i02 was 1.45.

By doing so, the computational resolution can be made as small as possible.
The groove depth and SiO2 film thickness calculated based on the optical path difference obtained from MEM and the groove depth and 5i02 film thickness obtained from the SEM cross-sectional photograph are each within the range of measurement accuracy. The results are in good agreement with each other, indicating that the micro groove depth can be measured with high accuracy by this example using MEM.

The selection of the order of the filter is important. We tried the following three methods to select the filter order.

(1) The filter order is increased sequentially from 1 and the filter order that gives the minimum value of the final prediction error is selected.

(2) Calculate all final prediction errors within the set filter order range and select the filter order that gives the minimum value.

(3) Calculate the filter order by fixing it to a preset value, regardless of the dependence of the final prediction error on the prediction error filter order.

Method (1) is fast in calculation time, but when observing a noisy waveform, the behavior of the final prediction error due to noise may be mistaken for a local minimum value, and as a result, correct nuvector analysis may not be possible. Method (2) yielded good results in most cases in this signal analysis. However, if the area of the fine groove is small and the signal from the groove bottom is smaller than the signal from the surface of the groove, the optical path difference corresponding to the groove depth may be determined to be noise and erased. (3) is effective when analysis cannot be performed using (2), and Nuvector analysis is possible even from a small signal riding on a large amplitude. However, in rare cases, noise may be mistaken for a signal and a false optical path difference may be given. In order to avoid this problem, it is preferable to set the order of the expected error filter in the range of 30 to 60 and select the filter order from this range. From the results of measurements of microgrooves of various structures, it has been found that it is appropriate to use (2) and (3) depending on the situation, keeping in mind the above characteristics. Practically speaking, if an inspection mark that can be easily analyzed by spectroscopy is made on a silicon substrate (such as a pattern in which grooves with a width of around 1 μm are lined up at a pitch of about 2 μrrL), the groove depth can be measured without failure using method (2). It was confirmed that it is possible. In order to identify a false signal in (J), 7Ht = 5fi
It is useful to take the product of the signal with the result of .

In this embodiment, the case of measuring the groove depth after trenching and etching of a silicon substrate has been described, but it is obvious that the present invention can be used for measuring the depth of general fine grooves.

It is also believed that the present invention is effective in measuring the optical thickness of each volatile film formed on the bottom surface of a microgroove formed on a general multilayer film or substrate.

Effects> By using the present invention, it has become possible to accurately monitor the depth of microgrooves during the process.

As a result, the reliability of the LSI manufacturing process using microgrooves is improved, and the yield of VLSIs typified by highly integrated memories is improved.

[Brief explanation of the drawing]

Fig. 1 is a block diagram according to an embodiment of the present invention, Fig. 2 is a diagram plotting the reflection spectral intensity of sample s-1 against the wave number, and Fig. 3 is a diagram plotting the reflection spectral intensity of sample S-2 against the wave number. Figure 4 is a cross-sectional view of the sample /l/S-2, Figure 5 is a diagram showing the results of fast Fourier transform of the reflection spectroscopy signal of Figures 2 and 3. FIG. 6 is a diagram showing the results of processing the reflection spectroscopic spectrum signals of FIGS. 2 and 3 using the maximum entropy method. 1: Light source 2: Substrate 3 Nislit 4: Diffraction grating 5: Multi-channel light receiving element 6: Data processing department agent Patent attorney Takeshi Sugiyama (and 1 other person)
-9%ml Figure 2 /3 15 /7 /9 2/
27 25' defeat CX10''& 3 Figure 3 Figure 4 0 5000 10000
15 CODt roaring flood [nm, 1 to 5 figure 0 〃〃 /ω〃
Figure 6

Claims (1)

[Claims] 1) A light source for irradiating light onto a surface with microgrooves, a microscope that guides the irradiated light and reflected light from the light source to the surface with microgrooves, and receives the reflected light derived by the microscope. a reflection spectrometer that performs spectroscopy, a calculation means that forms optical path difference information from a spectral signal output from the spectrometer, and a calculation means that calculates a microgroove depth based on the optical path difference information. A micro-groove depth measuring device that is characterized by 2) The micro-groove depth measuring device according to claim 1, wherein the light-receiving element of the reflection spectrometer is a multi-channel type light-receiving element. 3) The fine groove depth measuring device according to claim 1, wherein the calculation means for forming the optical path difference information uses fast Fourier transform. 4) The micro groove depth measuring device according to claim 1, wherein the calculation means for forming the optical path difference information uses a maximum entropy method.