JPH07104956B2 - High-accuracy peak certification method - Google Patents

Description

Detailed Description of the Invention [Industrial applications]

The present invention relates to a method for recognizing a spectral peak with high accuracy (high reliability) from spectral data obtained by EPMA or the like.

[Prior art]

In addition to the detection signal obtained from the characteristic X-ray from the measurement sample, the spectrum obtained by EPMA contains noise and detection signals from other X-rays incident from the surrounding area.
Since the signal from the characteristic X-ray is randomly incident,
If there is statistical fluctuation and a sufficient integration time is not applied, the spectrum data of a smooth curve cannot be obtained. Usually, since the measurement is performed with an integration time such that the peak can be determined, the obtained spectrum is jagged. Therefore, the spectral data was first subjected to smoothing processing to form an ideal curved line without jaggedness, and then the portion protruding from the background was identified as a spectral peak.

[Problems to be Solved by the Invention]

In the above-described conventional example, since the protruding portion in the smoothed spectrum data is simply identified as a peak, a noise peak that is accidentally present may be mistaken for a true peak, or conversely, it may be identified as a peak. The reliability was not sufficient because we sometimes missed the true peaks below the selection level. It is an object of the present invention to make the detection peak more reliable by checking again.

[Means for Solving the Problems]

Differentiate the smoothed spectrum data obtained by X-ray spectroscopic analysis, search for the point where the differential value changes from + to − (differential inflection point), and recognize that differential inflection point as the peak apex. , The detection signal strength value at the same peak apex is (PK), the half-value width is estimated from the same (PK), and the detection signal strength value at a point separated by n times the half-value width on both sides from the peak apex is set to the background. Value [(BG-), (BG
+)] And the peak signal strength value (PK) is When satisfying, the (PK) was approved as the peak value.

[Action]

The spectrum data is smoothed, the smoothed spectrum data is differentiated, the differential inflection point is searched, and it is determined whether or not the peak existing at the searched differential inflection point is the detected peak. In this determination, the peak is detected by inverting the positive / negative of the differential signal, rather than detecting the peak by the method of recognizing it as a peak if it protrudes by a certain value (selection level) or more from the baseline of the smoothed spectrum. , Don't miss a minute peak. On the other hand, this may lead to over detection of peaks. Therefore, in the present invention, the following peak test is further performed to confirm the peak by double detection. In radiation measurement in which a certain substance is irradiated with an electron beam and the generated radiation is detected by a detector, there is a statistical fluctuation error in the detection signal. The variation error has a Gaussian distribution with respect to the true value. Therefore, the error standard deviation σ is Becomes From this, it is possible to obtain about 90% (87%) of the accuracy if the measured value is within the range of 1.5σ, so when the measured value within the 1.5σ range does not intersect with the background 1.5σ range. Since it is determined that the measured value has a significant difference from the background, the determination can be performed using the following determination formula. Also, in the case of a small peak, peak intensity ≈ background intensity, so And the above formula is Therefore, empirically, this formula has the highest determination accuracy. As mentioned above, the detection signal of the point that was once identified as the peak
Whether the PK is higher than the value obtained by adding k times the square root of the background to the background at the peak position interpolated from the detection signal at the position that is considered to be the background part that is appropriately separated on both sides. By deciding whether or not it is, it is checked again whether or not the initially recognized peak is a true peak.

Embodiment FIG. 1 to FIG. 4 show an embodiment of the data processing method of the present invention. FIG. 1 shows unprocessed spectral data obtained by an electron beam microanalyzer (EPMA) or the like. The unprocessed spectrum data is smoothed and converted into spectrum data as shown in FIG. Differentiating this spectrum data results in a differential curve (increase / decrease function curve) shown in FIG. Since the peak top is a point where the peak changes from increase to decrease, it is considered to be at a point (differential inflection point) where + changes to − in the differential curve. Therefore, in the differential curve, the differential inflection point is searched, and the detection signal strength value (PK) at the differential inflection point is checked and stored, and
As shown in FIG. 4, the half-value width corresponding to the detection signal strength (PK) at the peak point is estimated, the detection signal strength value at the point three times the half-value width away from the peak apex is examined, and the same strength value is calculated. Use the background value [(BG-), (BG +)].
The peak signal strength value (PK) When the inequality is satisfied by substituting into, the (PK) is recognized as the peak value. The measurement position of the background will be described. A normal spectrum peak waveform also has a Gaussian distribution. However, in terms of intensity, even at a point 1.5 σ away from the true value, the intensity is about 32% of the peak intensity, and at a point 3 σ away, the intensity is about 1% of the peak intensity. In X-ray intensity measurement by EPMA, the measurement error is usually set to 1% or less in consideration of problems such as equipment stability, sample smoothness, and cleanliness. Therefore, as for the background position, when there is no other peak in between, the influence of the peak can be almost eliminated by setting a point 3σ apart. The full width at half maximum is about 1.2σ.

【effect】

According to the present invention, the qualification of the detection peak becomes more accurate.

[Brief description of drawings]

1 is an unprocessed spectrum data of one embodiment of the present invention, FIG. 2 is a spectrum data of the above embodiment subjected to smoothing processing, FIG. 3 is a differential curve (increase / decrease function curve) of the above embodiment, and FIG. FIG. 4 is an explanatory diagram of data processing of the above embodiment.

Claims (1)

[Claims] 1. A method for differentiating smoothed spectral data obtained by X-ray spectroscopic analysis or the like to search for a point (differential inflection point) at which the differential value changes from + to −, and the differential inflection point is the peak apex. However, the detection signal strength value at the same peak apex is defined as (PK), the half width is estimated from the same (PK), and the detection signal strength at the point n times the half width on both sides from the peak apex. The background value [(BG
-), (BG +)], and the above peak signal strength value (PK)
But, A high-accuracy peak qualification method characterized by certifying (PK) as a peak value when satisfying