Element analyzing method

Description: TECHNICAL FIELD [Detailed description of the invention]

【0001】

[Industrial application field]

The present invention relates to an elemental analysis method using fluorescent X-rays, and more particularly to an elemental analysis method capable of accurately identifying a light element and calculating a concentration.

【0002】

[Conventional technology]

A fluorescence X-ray analysis method has been used as a non-destructive elemental analysis of an object (sample) to be measured. In addition, a total reflection fluorescence X-ray analysis method has been developed with the aim of improving its sensitivity, and its application to pollution control in semiconductor processes is being studied. Among the total internal reflection fluorescence X-ray analysis methods, the energy-dispersed fluorescence X-ray analysis method can measure the spectrum of a wide range of energy regions, so a single semiconductor detector (SSD) placed directly above the sample can be used. Simultaneous analysis of multiple elements is possible. In addition, the energy-dispersed fluorescence X-ray analysis method does not require crystal spectroscopy, so the sample and SSD can be brought close to each other, and higher sensitivity can be expected compared to, for example, the wavelength-dispersive fluorescence X-ray analysis method. Has.

【0003】

In the total reflection fluorescence X-ray analysis method, the spectral resolution is inferior to that of the wavelength dispersive type, and the total power count obtained is also low. In addition, escape peaks, peaks due to the influence of diffraction lines, sum peaks (2Kα; Kα + Kβ), peaks due to the effect of Compton scattering, etc. are likely to appear as interference peaks.

【0004】

[Problems to be Solved by the Invention]

In the conventional energy-dispersed fluorescence X-ray analysis method, it is easy to distinguish between element peaks and interfering peaks when the concentration of pollutant elements is high, but light elements are particularly required in semiconductor processes10.⁹atoms / cm²It is difficult to identify the following concentrations. The reasons for this are as follows. (1) The probability of K-ray generation (fluorescence yield) is 0.02 for Na, which is 1/20 of 0.4 for Zn. Therefore, even if an element having the same concentration as a heavy element exists in the light element (Na to Cl), the total count number (Kα ray) obtained is lower than that of the heavy element. (2) A Be window is attached to the tip of the semiconductor detector to maintain a high vacuum inside the detector and prevent surface contamination of the Si crystal of the material of the semiconductor detector, and Kα by this Be window. Line absorption occurs. That is, when there is a Be window with a thickness of ~ 10 μm, light elements below the F element cannot be detected due to absorption of Kα rays, and those above the Cl element are not affected by Kα ray absorption. Then, in the element between the F element and the Cl element, absorption occurs according to the mass absorption coefficient. (3) In the semiconductor process, the sample is mainly Si, and there is a peak of Na to Cl elements at the base of the Kα spectrum due to the Si element. Therefore, it is necessary to take measures to suppress the half width of the Si peak as much as possible, for example, to set the X-ray incident angle sufficiently smaller than the critical angle. Being susceptible to.

【0005】

The present invention has been made in consideration of such points, and an object of the present invention is to provide an elemental analysis method capable of accurately identifying analytical elements, particularly light elements, and calculating their concentrations.

【0006】

[Means for solving problems]

In the present invention, each analysis is based on the step of smoothing the measurement waveform of the fluorescence X-ray obtained from the object to be measured and the energy resolution of the detection system with respect to the fluorescence X-ray energy value of each analysis element. A step of obtaining the device function of the detection system for each element, a step of deconvolving the smoothed measurement waveform with the device function of the detection system, and specifying the analysis element from the measured waveform subjected to the deconvolution process. In addition, it is an element analysis method consisting of a step of determining the concentration and a step of determining the concentration.

【0007】

[Action]

According to the present invention, it is possible to correct the waveform distortion caused by the energy resolution of the detection system by performing the deconvolution process using the device function of the detection system.

【0008】

[Example]

Hereinafter, examples of the present invention will be described with reference to the drawings. (First Example) FIGS. 1 to 3 are diagrams showing a first embodiment of the elemental analysis method according to the present invention.

【0009】

In FIG. 3, a sample 1 (object to be measured) such as a semiconductor wafer made of Si is irradiated with X-rays, and fluorescent X-rays generated from the sample 1 are detected by the semiconductor detector 2 (SSD). A signal processing device 3 for processing an electric signal output from the semiconductor detector 2 is connected to the semiconductor detector 2. A Be window (~ 10 μm thickness) is attached to the tip of the semiconductor detector 2. Further, a FET, a preamplifier, an A / D converter, and the like are built in the signal processing device 3.

【0010】

Next, a pollutant element analysis method using this energy-dispersed fluorescence X-ray analyzer will be described with reference to FIGS. 1 to 3.

【0011】

First, the sample 1 containing the contaminating element is irradiated with X-rays, and the fluorescent X-rays generated from the sample 1 are detected by the semiconductor detector 2. Next, the measurement waveform of the fluorescence X-ray is input to the signal processing device 3.

【0012】

The signal processing device 3 performs the following signal processing on the measured waveform of the fluorescent X-ray. First, in the input measurement waveform data, absorption by the Be window occurs in the fluorescence X-ray energy region of the light element. For example, in a Be window with a thickness of ~ 10 μm, 100% of the fluorescence X-ray energy is absorbed at an energy value of 0.55 KeV, and 0% of the fluorescence X-ray energy is absorbed at an energy value of 2.5 KeV. That is, I / I₀= exp {-(μ / ρ) ρ Z} ...... (1) Relationship is established.

【0013】

I / I here₀Is the absorption coefficient, μ / ρ is the mass absorption coefficient of Be, and μ is the window degree of Be (1.85 g / cm).³), Z is the thickness of the Be window (here 10 μm)^tCalculated as).

【0014】

Here, as shown in FIG. 1, the absorption coefficient I / I described above is given to each count (Ci) of the channel corresponding to the energy region of 0.55 KeV to 2.5 KeV from the initially input measurement waveform data.₀By multiplying by the reciprocal of (Ai), the count value (Ci') of the fluorescent X-ray that should originally pass through the Be window is corrected.

【0015】

Ci = Ai^-1× Ci ...... (2) (Where i is the channel number).

【0016】

After performing this correction calculation, the measured waveform is weighted and smoothed at five points to mainly reduce background noise.

【0017】

Next, the waveform distortion improvement (deconvolution processing) in the detection system consisting of SSD ~ FET ~ preamplifier ~ A / D converter is executed. That is, the energy resolution of the SSD for the 2KeV energy value is ~ 100eV, and the resolution for the 6KeV energy value is ~ 150eV.

【0018】

Therefore, based on the energy resolution of the detection system with respect to the luminous X-ray energy value of each analytical element, a device function of the detection system is prepared for each analytical element, and the measurement waveform smoothed using this device function is obtained. On the other hand, deconvolution processing (deconvolution processing) is performed.

【0019】

The device function of the detection system is represented by a Gaussian function G (X) having an energy resolution in half width.

【0020】

[Number 1]

<img file="JP2848751B2_D0001.tif" />Where X is the channel number; Wh is the half width of energy corresponding to the Kα spectral value of each element; C_NIs the noise component; C is the constant; E is the energy value of the fluorescence X-ray spectrum of each element.

【0021】

Deconvolution processing is performed on the smoothed measurement waveform that has been smoothed as described above, using the above Gaussian function as a device function. Specifically, in the measurement waveform, the value of E is set to 6.398KeV, 7.471KeV, 8.630KeV for each analytical element to be measured such as Fe, Ni, Zn, and the device function is obtained by the formulas (3) and (4). .. Then, deconvolution processing is performed on the measured waveform in the range of ± 500 eV centering on each spectrum value. The Gauss-Seidel method is used for the actual operation of deconvolution processing.

【0022】

That is, when the actual fluorescence X-ray waveform data is I and the measured waveform data measured via the device function G (X) is 0, 0 = I × G (X) ...... (5).

【0023】

Therefore, the actual waveform data I Obtained by I = 0 / G (X) ...... (6).

【0024】

Next, after the deconvolution process, the measured waveform is subjected to a smoothing differential process. The number of processing points for smoothing is 7 in consideration of the energy width and energy resolution of the channel (each point on the horizontal axis of the measurement waveform). Next, the zero point is detected from the differential waveform, and this is used as a candidate for the peak of the measurement waveform.

【0025】

Next, a model function, for example, a Gaussian function is prepared for each peak of the measurement waveform, and a model waveform is constructed by the linear sum of the Gaussian functions. In this case, each Gaussian function has a peak energy position u, a peak height h, and a half-value half width w as initial parameters, and these initial parameters are variables of each Gaussian function.

【0026】

z (i, p) = Σ_jh exp {-ln2 (ν-u)²/ w²}.

【0027】

(j corresponds to each detected peak) The peak energy position u of each Gaussian function is roughly determined from the channel number i corresponding to each peak of the waveform data. The peak height h is roughly determined from the peak of the waveform data, and the half-value half-width w is roughly determined based on the point where the third derivative is O.

【0028】

Next, the nonlinear optimization process is performed so that the sum of the residual squares of the model waveform composed of the linear sum of the Gaussian functions prepared for the number of peaks of the measurement waveform and the measurement waveform is minimized, and the above-mentioned initial stage is performed. Determine the parameters and obtain the separated waveform of each Gaussian function. That is, the sum of the residual squares of the model waveform z (i, p) and the measured waveform y (i) is defined as the objective function e (p) as follows: e (p) = Σ {z (i, p) -y (i)}²here z (i, p) = Σ_jh exp {-ln2 (ν-u)²/ w²}.

【0029】

(j corresponds to each detected peak) The vector variable p that minimizes this objective function e (p) is also found by using the nonlinear optimization method using the simplex method. In addition, another method (DEP method or the like) may be used instead of the simplex method.

【0030】

Next, based on the separation waveform of each Gaussian function obtained by the nonlinear optimization process, the energy position of the separation waveform and the kαX-ray peak position of each element are sequentially compared. At this time, a certain margin is provided in consideration of the chemical shift, and when a Gaussian function corresponding to the kαX-ray peak position of the element is found, this Gaussian function is specified as the element (contamination element). Peaks without corresponding elements are specified as escape peaks, thumb peaks, and background noise. Next, the integrated intensity (area) of the Gaussian function corresponding to each element is obtained. In this case, the Gaussian function is integrated in the range of ± 4w. Where w is the half width of the Gaussian function. (Specific Example) Next, a specific example of the first embodiment will be described with reference to FIG. Figure 2 (a) shows the initially input measurement waveform data in the energy region 0.41 to 1.52 KeV. Figure 2 (b) shows the measured waveform data after correcting the waveform distortion of light elements caused by Be window absorption using the absorption coefficient. Next, FIG. 2 (c) shows the measured waveform data after the 5-point weighted smoothing process. Next, Fig. 2 (d) shows the measured waveform data after deconvolution processing of the waveform distortion caused by the energy resolution of the detection system using the device function. Furthermore, the measured waveform data after peak separation by the nonlinear optimization process is shown in Fig. 2 (e).

【0031】

As shown in FIG. 2, it is difficult to identify the peaks of Na and Mg in the initial measured waveform data (Fig. 2 (a)), whereas in the separated waveform data after waveform distortion correction shown in the present invention, it is difficult. It can be seen that the peaks of the light elements can be easily identified.

【0032】

As described above, according to the present embodiment, the light element is identified and its light element is identified by correcting the waveform distortion caused by the absorption of the Be window of the SSD and the waveform distortion caused by the energy resolution of the detection system, which are particularly likely to occur in the light element. The concentration can be calculated accurately.

【0033】

In addition, if hydrocarbon-based contamination in the Be window or an ice film is formed on the Si crystal surface of the sample due to cooling, the sensitivity of the Na-Kα spectrum deteriorates or the light element region The detection sensitivity is reduced. In such a case, the waveform distortion correction in the light element region according to the present invention has an important effect. (Second Example) Hereinafter, a second example of the elemental analysis method according to the present invention will be described with reference to FIGS. 4 to 6.

【0034】

First, a clean blank sample is irradiated with X-rays. Then, the input measurement waveform of the blank sample is weighted and smoothed at 5 points to reduce background noise. Next, based on the energy resolution of the detection system with respect to the luminous X-ray energy value of each analytical element, a device function of the detection system is prepared for each analytical element, and the measurement waveform smoothed using this device function is obtained. On the other hand, deconvolution processing is performed.

【0035】

Since the smoothing process and the deconvolution process are performed by the method shown in the first embodiment, detailed description thereof will be omitted.

【0036】

In this case, the same point of the blank sample is continuously measured up to 10 times, and the fluctuation range of the channel number of the spectrum corresponding to each element is obtained. This fluctuation range is obtained as the detection error due to the detection system.

【0037】

The actual sample is processed according to the flowchart shown in FIG. That is, after inputting the measurement waveform, smoothing processing (weighting averaging processing of 5 points) is performed over the entire range. Next, the channel fluctuation range corresponding to the analytical element obtained in advance in the blank sample is input.

【0038】

Next, the device function used in the deconvolution process is determined for each target analytical element. After that, deconvolution processing is performed on the measured waveform in the ± 500 eV range centered on the energy value of the Kα spectrum of the target analytical element using the device function.

【0039】

When identifying the target analytical element for the measured waveform after deconvolution processing, if the peak of the waveform data exists within the fluctuation range of the channel number previously obtained for the blank sample, it is automatically said that the target analytical element could be specified. judge.

【0040】

Further, when calculating the element concentration from this peak, the model waveform is constructed by the linear sum of the Gaussian functions prepared for the number of peaks of the measurement waveform. Then, nonlinear optimization processing is performed so that the sum of the residual squares of the measured waveform and the model waveform after the deconvolution processing is minimized, the initial parameters of each Gaussian function are obtained, and the separation waveform of each Gaussian function is obtained. Next, by obtaining the integrated intensity of each Gaussian function, element identification and concentration calculation can be automatically processed. (Specific Example) Next, a specific example of the second embodiment will be described with reference to FIGS. 5 and 6.

【0041】

Fig. 5 (a) shows the initially input measurement waveform, Fig. 5 (b) shows the measurement waveform after the 5-point weighted smoothing process, and Fig. 5 (c) shows the measurement waveform data after the deconvolution process. .. In FIGS. 5 (a) to 5 (c), the energy range is 6.01 to 7.13 KeV, with Fe-Kα peaks at the energy position of 6.398 KeV and Fe-Kβ peaks at the energy position of 7.057 KeV. That is, while the total count number is 0 to 15 (cts) in all channels in the input initial measurement waveform (Fig. 5 (a)), it corresponds to Fe-Kα of the measurement waveform after deconvolution processing. The channel count is ~ 200 (cts) (Fig. 5 (c)).

【0042】

Therefore, as shown in FIGS. 5 (a) to 5 (c), 10⁹atoms / cm²The presence of peaks can be clearly identified even in the case of low levels of contamination.

【0043】

As a result of measuring the 5-point weighted smoothing process and the deconvolution process 10 times continuously at the same point of the blank sample, it was found that there are ± 3 channels as the channel fluctuation range. This fluctuation corresponds to the fluctuation error caused by the device.

【0044】

Figure 6 shows the results of performing nonlinear optimization processing on the measured waveform data after deconvolution processing to obtain the integrated intensity of a specific element (Fe element). As shown in FIG. 6, as compared with the conventional method in which the deconvolution treatment is not performed, the concentration can be detected with high accuracy by performing the deconvolution treatment as in the present invention.

【0045】

[Effect of the invention]

As described above, according to the present invention, it is possible to correct the waveform distortion caused by the energy resolution of the detection system by performing the deconvolution process using the device function of the detection system. Therefore, the element can be specified and its concentration can be calculated with high accuracy.

[Simple explanation of drawings]

[Figure 1]

The flowchart which shows 1st Example of the elemental analysis method by this invention.

[Figure 2]

The figure which shows the specific example in the 1st Example.

[Fig. 3]

The schematic which shows the energy distribution type fluorescence X-ray analyzer.

[Fig. 4]

The flowchart which shows the 2nd Example of the elemental analysis method by this invention.

[Fig. 5]

The figure which shows the specific example in the 2nd Example.

[Fig. 6]

A calibration curve of Fe element showing a specific example in the second embodiment.

[Explanation of symbols]

1 sample 2 Semiconductor detector 3 Signal processing device