X-Ray Intensity Correlation Defect Detection Using a Single Speckle Pattern

Objective X- ray is a powerful tool to analyze the internal structure of macroscopic objects and has been widely used in many fields, such as biomolecular imaging and micro/ nanostructure detection. Traditional X-ray analysis techniques often have high requirements for light flux and coherence, which are difficult to be applied in a table-top source and thus limit their application. X-ray Fourier-transform ghost imaging (XFGI) has a low requirement for spatial coherence and enables table- top X-ray microscopic detection and imaging. In recent years, researchers have focused on spatial multiplexing, nonlocal modulation, preset speckle field, and other aspects in the field of XFGI, and it is shown that high-quality imaging can be achieved by using preset speckle patterns at low flux. XFGI via preset speckle patterns needs to measure a mass of speckle fields, and imaging consumes time. A single speckle field has a low signal-to-noise ratio and cannot be used to retrieve the image independently. However, certain sample structure information can be extracted from the spatial distribution of the single speckle field, which can be employed to realize rapid sample defect detection. We aim to propose a method for sample defect detection by using a single speckle pattern, which will be helpful for micro/nanostructure detection and analysis. Methods In this paper, a defect detection method based on speckle field distribution by single detection is proposed, and the correlation coefficient between the detected speckle field distribution of test samples and standard samples is used as the evaluation function for sample defect detection. The second-order autocorrelation detection of intensity fluctuation is simulated with the energy of 1095 eV, and defect detection samples have two types: samples with ten holes and circuit samples. Since experimental noise, such as shot noise, can affect the image contrast of the speckle field, the effect of speckle contrast is analyzed under different signal- to-noise ratios. To improve the reliability of this defect detection method, this paper compares the influence of different detail enhancement methods, such as Butterworth high-pass filtering, Wiener filtering, and guided image filtering, on the correlation coefficient detection, so as to select the appropriate detail enhancement method. Results and Discussions Samples with ten holes and circuit samples that have rotation angles of 60 degrees and 103 degrees, respectively, are inserted into the optical path to get the detected speckle field distributions and then process the defect detection ( Fig. 2). The matching angle and the correlation coefficient of defect detection results are consistent with the presupposition (Fig. 4). For image contrast, this paper corrects simulated speckle field distributions by gamma transform, and the correlation coefficient between detected speckle field distribution and simulated speckle field distribution will change with the image contrast correction coefficient. When the speckle field with a low signal-to-noise ratio is corrected by appropriate gamma transform, the correlation coefficient has a maximum value, which is the upper limit that this sample defect detection method can raise the correlation coefficientto (Fig. 6). For different detail enhancement methods, the results of guided image filtering are similar to Wiener filtering results in most details. However, for details inconsistent with the simulated speckle field distribution of the standard sample, the filtering results will be different from the detected speckle field distribution. The detection details corresponding to the sample defects will be weakened or even disappeared under the guidance of the simulated speckle field distribution of the standard samples. Therefore, the guided image filtering method has a better defect detection effect ( Fig. 7 and Table 1). At the same time, guided image filtering can smooth signals with a lower signal- to-noise ratio to avoid the misjudgment of detection (Fig. 7). Conclusions In this paper, based on the autocorrelation principle of intensity fluctuation, a fast defect detection method is developed by using a single preset speckle pattern. At the same time, based on this method, the second-order autocorrelation detection optical path of X-ray intensity fluctuation is simulated, and the defect detection of samples with ten holes and circuit samples is simulated. The accurate rotation matching angle and defect detection results can be obtained. In view of the actual situation, the images will have different contrasts under different signal- to-noise ratios, and the change in the contrast will affect the reduction of the measured value of the correlation coefficient. The gamma transform can correct the contrast and improve the upper limit of the correlation coefficient when the details are enhanced, so as to improve the detection accuracy. By comparing different detail enhancement methods, this paper also finds that the details consistent with the standard image will be enhanced, and those inconsistent with the standard image will be smoothed when the guided image filtering method is used to process the detected speckle field distribution. The measured correlation coefficient of the standard samples can be increased to more than 0. 95. However, the measured value of the correlation coefficient of samples with defects cannot be improved to such a level, and this property greatly improves the reliability of this sample defect detection method. In principle, this method does not require a coherent source, which makes it possible for applying table-top X-ray detection in semiconductor devices, integrated circuits, high- performance materials, and other aspects.