Wafer defect detection method based on phase shift interference and movement difference

Abstract

The application discloses a wafer defect detection method based on phase shift interference and movement difference, collects phase shift interference pictures of a measured wafer under single visual field, and completes three-dimensional topography data recovery of the wafer surface; the sample is microscopically scanned by a horizontal displacement table according to a certain step size, the sample is repeatedly measured for multiple times, second-order difference is made by using adjacent two measurement results, then the multiple second-order difference results are sequentially cropped and superimposed, so that the wafer surface defects form a specific mode of three-pole pattern; the defect three-pole pattern is detected and marked by using frequency domain high-pass filtering and image morphological operation and other image processing methods. Compared with other wafer defect detection methods, the application has high detection sensitivity, can accurately identify defects smaller than the optical resolution scale, and can classify and judge the concave-convex state of the defects, the method is suitable for surface defect detection of wafers and transparent or opaque optical elements, and is especially suitable for defect detection of patterned wafers.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of optical precision measurement, and particularly relates to a wafer defect detection method based on phase-shifting interference and moving differential. BACKGROUND

[0002] The semiconductor manufacturing process is divided into three stages of silicon wafer manufacturing, wafer manufacturing, and packaging testing, among which the wafer manufacturing is the most challenging production stage. The current chip processing has broken through 3 nm, and the corresponding wafer manufacturing surface defect detection requirement has reached sub-micron or even nanometer level. The mainstream detection equipment is based on the bright-dark field scattering method, and the light path structure of this kind of equipment is complex. In order to realize the detection of sub-micron or even nanometer level defects, some devices use ultraviolet or extreme ultraviolet light sources, which makes the manufacturing and maintenance cost of this kind of equipment high. Other methods such as detection equipment based on optical interference can achieve sub-nanometer resolution in the depth direction of the sample, but the horizontal resolution is limited by the optical diffraction phenomenon, and it is difficult to detect surface defects smaller than the optical resolution. Although atomic force microscopy (AFM) and scanning electron microscopy (SEM) can accurately measure nanometer-level defects, they are low in efficiency and not suitable for full inspection of wafer surface, but only suitable for quantitative re-inspection analysis of local defects. Therefore, in the large-scale production process of semiconductors, there is an urgent need for a low-cost technology that can quickly locate and detect sub-micron or nanometer scale defects on the wafer surface.

[0003] Among the retrieved related prior arts, patent CN202311165674.2 proposes a pattern wafer defect detection and three-dimensional topography measurement device and method, which uses a wide-spectrum white light interferometer to perform three-dimensional topography measurement in different regions; the three-dimensional topography in different regions is spliced by an image splicing algorithm to obtain the three-dimensional topography test result of the entire pattern wafer to be detected, and the test result is compared with the standard sample test result, and a threshold algorithm is used to obtain the abnormal area; the three-dimensional topography test result of the pattern wafer to be detected is compared with the actual wafer to obtain the accurate defect area range. Finally, an atomic force microscope measurement head is used to scan the abnormal area and the defect area to accurately characterize the three-dimensional topography, and the defect three-dimensional topography data result is obtained. This method needs to perform difference between the measurement result and the standard sample, and when measuring in multiple fields of view, the mechanical movement error of the device will cause the position of the measured pattern wafer to be skewed or rotated, which will make it difficult to accurately align when performing difference calculation with the standard sample in different fields of view, thereby causing the possibility of false detection.

[0004] Patent CN202180024107.8 proposes a method for measuring the shape of DIC defects. This method uses a particle counter to detect DIC defects on the main surface of a silicon wafer and determines the position coordinates of the detected DIC defects. Subsequently, it measures the shape characteristics of the height or depth of the detected DIC defects using phase-shift interferometry. This method directly evaluates and measures defects based on phase-shift interferometry imaging results. However, due to the influence of optical diffraction limits, the system has low lateral resolution, making it difficult to accurately detect submicron-level defects. Summary of the Invention

[0005] To overcome the limitations of existing technologies, this invention proposes a wafer defect detection method based on phase-shifting interferometry and moving differential. This method leverages the unique advantage of sub-nanometer sensitivity in the Z-axis of optical phase-shifting interferometry microscopy, while combining it with the excellent noise suppression capabilities of the moving second-order differential method. This invention enhances the horizontal detection sensitivity of the phase-shifting interferometry system and uses micro-movement second-order differential to create specific tripolar patterns from micro- and nano-defects on the wafer surface. Subsequently, image processing techniques are used to enhance and identify the signals, thereby achieving the detection and localization of wafer surface defects.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A wafer defect detection method based on phase-shifting interference and moving difference includes the following steps:

[0008] Step 1: Place the wafer under test under an optical phase-shifting interferometer microscope and adjust its position so that interference fringes of 1 to 2 complete cycles appear in the field of view. Acquire multiple phase-shifting interferometer images along the height direction of the wafer under test to obtain a set of phase-shifting interferometer images.

[0009] Step 2: Process the phase-shifting interferometric image set to complete the three-dimensional morphology calculation of the surface of the wafer under test under a single field of view;

[0010] Step 3, adjust the horizontal micro-step distance. The wafer under test is slightly moved along the horizontal X direction, and then steps one and two are repeated to acquire data. A set of phase-shifting interferometric images under different fields of view, and completed Calculation of the three-dimensional morphology of the wafer surface under test in a single field of view;

[0011] Step four: Divide the three-dimensional morphology calculation results of the measured wafer surface under three adjacent fields of view into one group of data, and perform second-order difference calculation on each group of data in sequence to obtain a total of The second-order difference results are denoted as . ;according to The order of calculation and Regarding the The second-order difference results are trimmed and superimposed to obtain an intermediate result containing the tripolar pattern of wafer defects. ;

[0012] Step 5: Eliminate intermediate results A The residual background fluctuations present in the signal-to-noise ratio are reduced, resulting in an enhanced effect. ;

[0013] Step Six: Based on the enhancement results Background standard deviation Set the central detection threshold and the two side detection thresholds for the tripolar pattern formed by defects on the surface of the wafer under test, and combine the number of moving pixels. Defect detection is performed.

[0014] In one embodiment, during the phase-shifting interferometric image acquisition in steps one and three, the light source is a quasi-monochromatic light source with a bandwidth of less than 20 nm.

[0015] In one embodiment, step one involves controlling a piezoelectric ceramic displacement device to acquire multiple phase-shifted interference images along the height direction of the wafer sample under test; step two involves processing the phase-shifted interference image set using a phase-shifted (PSI) encapsulation phase extraction algorithm and a phase unencapsulation algorithm to complete the three-dimensional morphology calculation of the wafer surface under test in a single field of view.

[0016] In one embodiment, the horizontal micro-step distance , represented as:

[0017]

[0018]

[0019]

[0020] The number of moving pixels , represented as:

[0021]

[0022] in The integer is greater than 1. When the wafer being tested is a patternless wafer, The minimum reliable step distance in the X direction of the horizontal displacement stage, when the wafer being measured is a patterned wafer after processing. The minimum spatial repetition period of the fabricated pattern on the patterned wafer in the horizontal X direction; For the optical resolution of the optical phase-shifting interferometer, The magnification of the optical phase-shifting interferometer is... This represents the pixel size of the image detector in an optical phase-shifting interferometer microscope. Indicates to Rounding to the nearest whole number.

[0023] In one embodiment, step four, the second-order difference result The calculation method is as follows

[0024]

[0025] in, , , They represent the first time. The, the The, the 3D topography calculation results of the wafer surface under test under a single field of view; pixel coordinates The value on The calculation is as follows:

[0026]

[0027] All That is, to constitute an intermediate result .

[0028] In one embodiment, step five first involves... Perform a Fast Fourier Transform (FFT), then create a frequency domain high-pass filter to filter it in the frequency domain, and finally perform an Inverse Fast Fourier Transform (IFFT) to obtain the intermediate result. Finally, based on the intermediate results Background standard deviation A bilateral filter is created for intermediate results using the optical resolution of the optical phase-shifting interferometer. Perform bilateral filtering to enhance intermediate results. The signal-to-noise ratio is improved to obtain the enhanced result. .

[0029] In one embodiment, the cutoff frequency of the frequency domain high-pass filter satisfy ,in , These represent the horizontal and vertical resolutions of the phase-shifting interferometer image detector, respectively; and the spatial Gaussian standard deviation of the bilateral filter. satisfy Pixel value similarity Gaussian standard deviation satisfy .

[0030] In one embodiment, the surface defects of the wafer being tested include pit defects and bump defects, in step six:

[0031] Set the central detection threshold for the tripolar pattern formed by the pits on the surface of the wafer under test. and detection thresholds on both sides Set the central detection threshold for the tripolar pattern formed by the bumps on the surface of the wafer under test. and detection thresholds on both sides And will enhance the results. The conditions are met and The pixels identified as pit-type defects will meet the conditions. and The pixels were identified as pit-type defects.

[0032] Specifically:

[0033] Enhance the results All less than Pixels marked as suspected pit defects are indicated by the horizontal X-axis direction of these pixels. The distance is the value of the pixels on both sides of the search, and compared with... The comparison showed that only the pixel values ​​on both sides of the suspected pit defect were greater than [value missing]. Only then is the pixel location identified as a pit defect.

[0034] Enhance the results All of the above Pixels marked as suspected bump defects are indicated by the horizontal X-axis direction of these pixels. The distance is the value of the pixels on both sides of the search, and compared with... The comparison showed that only the pixel values ​​on both sides of the suspected protrusion defect were less than [value missing]. Only then is the pixel location identified as a protrusion defect.

[0035] In one embodiment, the central detection threshold satisfy detection thresholds on both sides satisfy , , .

[0036] In one embodiment, the wafer under test is a transparent or opaque planar optical element.

[0037] Compared to the aforementioned patent CN202311165674.2, this invention is based on narrowband monochromatic light phase-shifting interferometry, and the second-order shifting difference is performed on the three-dimensional height and phase information of the same sample under different fields of view, rather than on the standard sample, thus avoiding the problem of inaccurate difference with the standard sample caused by mechanical motion errors of the system. Furthermore, the optical phase-shifting interferometer system structure used in this method is simpler than the white light quantitative phase imaging device used in patent CN202311165674.2.

[0038] Compared with the aforementioned patent CN202180024107.8, this invention uses the phase-shifting interferometric imaging result of a single field of view as the raw data, and then obtains the final detection result through steps such as moving differential and superposition of repeated measurement results and frequency-domain spatial filtering. This can improve the detection sensitivity of the optical phase-shifting device, thereby detecting submicron-level surface defects.

[0039] Applying the method of this invention to defect detection on the surface of unpatterned wafers, the results show that this invention can reliably detect wafer surface defects smaller than one-third the size of the system's optical resolution, and can also distinguish between depressions and protrusions. Furthermore, compared to other methods and devices, this method has the significant advantage of low hardware costs. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of a hardware device used in an embodiment of the present invention. In the figure:

[0041] 1-Optical probe, 2-Z-Precision piezoelectric displacement device, 3-Horizontal X-axis displacement stage, 4-Horizontal X-axis and Y-axis tilt angle orthogonal adjustment mechanism, 5-Wafer under test, 6-Interference objective lens, 7-Illumination system, 8-Z-axis height adjustment mechanism.

[0042] Figure 2 The results are obtained by scanning electron microscopy observation of submicron-level, patternless wafer surface defects simulating rectangular pits in a specific embodiment of the present invention.

[0043] Figure 3 This is a phase-shifting interference image of a patternless wafer surface under a single field of view in a specific embodiment of the present invention.

[0044] Figure 4 The phase of the surface height of the patternless wafer is calculated using the five-step phase shifting method in a specific embodiment of the present invention.

[0045] Figure 5 The results are the phase unwrapping calculations based on the least squares algorithm of shearing interference in a specific embodiment of the present invention.

[0046] Figure 6This is an intermediate result obtained by multi-field second-order difference superposition of a patternless wafer defect sample in a specific embodiment of the present invention. , where (a) represents an intermediate result (Global), (b) indicates intermediate results (Rectangular area).

[0047] Figure 7 This is an intermediate result obtained by high-pass filtering in the frequency domain from a patternless wafer defect sample in a specific embodiment of the present invention. , where (a) represents an intermediate result (Global), (b) indicates intermediate results (Rectangular area).

[0048] Figure 8 This is the result obtained by spatial bilateral filtering of a patternless wafer defect sample in a specific embodiment of the present invention. , where (a) represents an intermediate result (Global), (b) indicates intermediate results (Rectangular area).

[0049] Figure 9 This is the marking result of a local submicron defect on a non-image wafer in a specific embodiment of the present invention. Detailed Implementation

[0050] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings and examples.

[0051] The method of this invention can be implemented using an optical phase-shifting interferometry microscope. Figure 1 This diagram illustrates the structure of a quasi-monochromatic phase-shifting interferometry (PSI) microscope. To ensure high contrast of the interference fringes in the phase-shifting region, the bandwidth (full width at half maximum) of the quasi-monochromatic light source should be less than 20 nm. The illumination system 7 of the phase-shifting interferometry microscope in this diagram uses broadband white light as the light source. A narrow-band interference filter is installed behind the light source to obtain quasi-monochromatic light. To ensure a uniform and stable illumination field, the illumination system employs a Kohler illumination structure. The interferometer objective 6 is mounted on the phase-shifting interferometer probe via a Z-axis precision piezoelectric transducer (PZT) 2. The PZT allows the interferometer objective 6 to move along the Z-axis, thereby changing the optical path difference between the optical reference plane and the object under test. Optical phase-shifting interferometry (PSI) data acquisition is achieved through the vertical movement of the PZT. The optical probe 1 is mounted on a Z-axis height adjustment mechanism 8 to achieve focusing of the optical system. A horizontal X-axis displacement stage 3 is mounted below the optical probe 1. The horizontal X-axis displacement stage 3 has high motion accuracy and is used to move the wafer under test 5 in small steps along the horizontal X-axis. The horizontal X and Y tilt angle orthogonal adjustment mechanism 4 below the sample is used to adjust the position of the wafer sample under test to be horizontal.

[0052] In the embodiments of the present invention, after adjusting the pose of the wafer 5 under test using the above-described device, the three-dimensional surface morphology of the wafer 5 under test in a single field of view is first calculated using optical interference phase extraction technology. Then, the wafer 5 under test is moved by second-order difference so that its surface defects form a specific tripolar pattern. Finally, the identification and location of wafer surface defects are realized by combining filtering and threshold segmentation detection techniques in digital images.

[0053] The wafer defect detection method based on phase-shifting interference and moving difference of this invention is applicable to planar optical components, whether the wafer under test is transparent or opaque. Its main steps and possible implementation methods can be described as follows:

[0054] Step 1: Place the wafer 5 to be tested on the stage of the optical phase-shifting interferometry microscope. Focus the optical system by adjusting the Z-axis height adjustment mechanism 8. Further adjust the orientation of the horizontal X-axis and Y-axis tilt orthogonal adjustment mechanisms 4 to produce interference fringes of 1-2 complete cycles within the field of view. Control the Z-axis piezoelectric ceramic displacement device 2 to slightly move the interferometer objective lens 6 in the Z-axis, and acquire the wafer 5 under the first field of view. Zhang phase-shift interferometric images are generated to form a phase-shift interferometric image set. This set includes the micro-motion distance of the Z-axis piezoelectric ceramic displacement device 2 and the number of acquisitions. It can be determined based on the specific phase extraction method used in step two.

[0055] Step 2: The phase-shifting interferometry (PSI) image set is processed using the phase extraction algorithm and the phase unwrapping algorithm to complete the three-dimensional morphology calculation of the wafer surface under single field of view.

[0056] For the wrap-around phase extraction algorithm used in this step, there are currently several mature phase information extraction techniques, such as the classic multi-step phase shift method. This method requires a single phase shift amount of [missing information]. The corresponding single-step movement distance of the piezoelectric ceramic displacement device for Taking the four-step phase shift method as an example, the method for calculating the wrapping phase of the surface height of the measured object is as follows:

[0057]

[0058] in, Indicates the first nInterference images acquired through phase shifting. While phase extraction algorithms, such as the Advanced Iterative Algorithm (AIA), which are less demanding on phase shift, can be used, these algorithms are computationally expensive and only suitable for applications where measurement efficiency is not critical. Furthermore, since wafer surface defects range from submicron to nanometer scale, their weak signals are easily masked by background noise in the detection system. Therefore, it is recommended to use a phase extraction algorithm with four or more steps to extract the wrapped phase, given its high robustness and accuracy. The phase calculated by the wrapped phase extraction algorithm is folded within the interval defined by brackets in the formula. Therefore, it is necessary to unwrap the folded phase information to obtain the true height phase information of the measured wafer surface.

[0059] For the phase unwrapping algorithm used in this step, since the wafer surface is flat and smooth, mainstream phase unwrapping algorithms can all perform reliable unwrapping calculations. These methods include the least squares method based on Fast Fourier Transform (FFT-LS), the least squares method based on Discrete Cosine Transform (DCT-LS), and the least squares method based on transverse shearing interference (LS-LS). The phase unwrapping calculation process will be explained using the LS-LS algorithm as an example. First, the wrapped phase... Construct an equivalent two-dimensional complex light field A new sheared light field is obtained by shifting the equivalent light field one pixel along the x-direction. And calculate the gradient of the wrapping phase in the x-direction. Similarly, the y-axis gradient of the wrapping phase is calculated. .

[0060]

[0061]

[0062]

[0063]

[0064] And the discrete Poisson equations are obtained. ,in For pixels The true phase of each pixel can be calculated by solving the system of equations using the discrete cosine transform. .

[0065] Step 3: Determine the horizontal micro-motion step distance of the sample based on the optical resolution of the optical phase-shifting interferometer, the minimum step size of the displacement stage in the horizontal X direction, and the pixel size of the image detector in the optical phase-shifting interferometer microscope. Number of moving pixels .according to The wafer under test 5 is slightly moved along the horizontal X direction, and then steps one and two are repeated to collect data. A set of phase-shifting interferometry images under different fields of view, and completed Calculation of the three-dimensional morphology of the surface of the wafer under test under a single field of view.

[0066] For example, in this step, the horizontal micro-motion step distance and number of moving pixels , represented as:

[0067]

[0068]

[0069]

[0070]

[0071] in The integer is greater than 1. When the wafer being tested is a patternless wafer, The minimum reliable step distance in the X direction of the horizontal displacement stage, when the wafer being measured is a patterned wafer after processing. It is the minimum spatial repetition period of the fabricated pattern on the patterned wafer in the horizontal X direction. For the optical resolution of the optical phase-shifting interferometer, The magnification of the optical phase-shifting interferometer is... This represents the pixel size of the image detector in an optical phase-shifting interferometer microscope. Indicates to Rounding to the nearest integer. This requirement determines the movement step size. The diameter of the Airy disk should be greater than twice the size of the disk diameter, and the number of pixels moved each time should be an integer to avoid inaccuracies in the overlapping of the middle and sides of the third-order pattern and the superposition of multiple fields of view during subsequent second-order difference calculations.

[0072] Step four: Divide the three-dimensional morphology calculation results of the measured wafer surface under three adjacent fields of view into one group of data, and perform second-order difference calculation on each group of data in sequence to obtain ( The second-order difference results are denoted as [a_n] . ;according to The calculation order and horizontal micro-motion step distance right The second-order difference results are trimmed and superimposed to obtain an intermediate result containing the tripolar pattern of wafer defects. .

[0073] For example, the second-order difference result of this step Calculate according to the following formula

[0074]

[0075] in, , , They represent the first time. The, the The, the Calculation results of the three-dimensional topography of the wafer surface under test under a single field of view. Pixel coordinates. The value on The calculation is as follows:

[0076]

[0077] All That is, to constitute an intermediate result .

[0078] Step 5, in order to eliminate intermediate results The residual background fluctuations present in the first place Perform a Fast Fourier Transform (FFT), then create a frequency domain high-pass filter to filter it in the frequency domain, and finally perform an Inverse Fast Fourier Transform (IFFT) to obtain the intermediate result. .

[0079] Environmental vibrations and piezoelectric displacement errors during the sampling process can cause calculation errors in the phase extraction algorithm described in step two, ultimately resulting in weak periodic background fluctuations in the intermediate results. These fluctuations can be eliminated in the frequency domain using two-dimensional Butterworth filters or Gaussian filters of order four or lower, which effectively suppress ringing effects. For example, the intermediate results... The calculation can be expressed as follows:

[0080]

[0081] in, This indicates that a two-dimensional fast Fourier transform is being performed. This indicates the inverse fast Fourier transform (FFT) in two dimensions. This means taking the real part of a complex number. This is a frequency domain high-pass filter. The following explanation uses a frequency domain Butterworth high-pass filter as an example. Creation method:

[0082]

[0083]

[0084] in, Coordinates of the points in the spectrum ( i , jDistance from the center of the spectrum graph The order of the Butterworth filter is used to avoid ringing effects. ,recommend , It is the cutoff frequency of the frequency domain high-pass filter, and satisfies... , , These represent the horizontal and vertical resolutions of the phase-shifting interferometer image detector, respectively. This filter must not only eliminate background fluctuations in the intermediate results but also, to a certain extent, enhance the model of the third-order defect pattern. The optimal choice should be made based on actual results within the constraints.

[0085] Step 6, to enhance intermediate results The signal-to-noise ratio, based on intermediate results Background standard deviation The optical resolution of the optical phase-shifting interferometer creates a bilateral filter for intermediate results. Bilateral filtering is performed to obtain the enhanced result. .

[0086] The specific method for creating a bilateral filter is as follows:

[0087] Bilateral filters can smooth background noise signals while preserving rapidly changing signals containing useful information, thereby enhancing the signal-to-noise ratio. (For intermediate results...) The result is obtained by performing bilateral filtering. , The formulas for calculating the values ​​of each pixel are as follows.

[0088]

[0089]

[0090] in, This is the sum of weights for each pixel within the current filtering window, used to normalize the weights. In the calculation formula and These are sliding windows Within the X and Y directions. and These represent the spatial Gaussian standard deviation of the bilateral filter and the Gaussian standard deviation of pixel value similarity, respectively. Determines the filtering window The size of the image is adjusted to avoid data aliasing between the sides and the middle of the three-pole graph during filtering. Should meet According to the Gaussian distribution law, the value of the pixel in the central pattern of the tripolar pattern should be... Outside the range, to highlight the tripolar pattern, a bilateral filter should be used. In calculating the standard deviation of intermediate result B First calculate separately Central and surrounding areas The standard deviation of the neutron region is calculated, and then the median of the standard deviations of these sub-regions is taken as the standard deviation. .

[0091] Step 7, based on the enhancement results Background standard deviation Set the central detection threshold for the tripolar pattern formed by the pits on the surface of the wafer under test. and detection thresholds on both sides ; will enhance the results All values ​​less than the threshold Pixels marked as suspected pit defects are indicated by the horizontal X-axis direction of these pixels. The distance is the value of the pixels on both sides of the search, and is compared with the threshold. The comparison showed that only the pixel values ​​on both sides of the suspected pit defect were greater than the threshold. Only then is the pixel location identified as a pit defect.

[0092] The specific method for setting the three-pole pattern detection threshold in this step is as follows:

[0093] The triangular pattern is clearly different from the background. According to statistical rules, the value at the center pixel of the triangular pattern should be within ± In addition, To enhance the background standard deviation of result C, the threshold can be set to... and In reality, the signal-to-noise ratio of the final three-pole image is significantly enhanced after multiple filtering steps. To avoid false detections, the aforementioned threshold can be further increased and set to [value missing]. and ,in The calculation method and steps in step six The calculation method is the same.

[0094] Step 8: Set the central detection threshold for the tripolar pattern formed by the bumps on the surface of the wafer under test. Both sides detection threshold ,satisfy: , This will enhance the results. All values ​​greater than the threshold Pixels marked as suspected bump defects are indicated by the horizontal X-axis direction of these pixels. The distance is the value of the pixels on both sides of the search, and is compared with the threshold. The comparison showed that only the pixel values ​​on both sides of the suspected protrusion defect were less than the threshold. Only then is the pixel location identified as a protrusion defect.

[0095] In one specific embodiment of the present invention, submicron-sized rectangular pit defects were fabricated on the surface of a patternless wafer using ion beam lithography equipment, and the defects were then detected using the method of the present invention. Figure 2 As shown, the actual size of the defective sample, measured by scanning electron microscopy (SEM), is 374.4 nm × 361.6 nm, and the defect depth is approximately 50 nm. The basic hardware parameters of the system used in this example are shown in Table 1.

[0096] Table 1 Example Hardware Parameter Table

[0097]

[0098] The sample to be tested is placed on a sliding stage, and the defect location is moved to the center of the field of view using the markings on the sample. The optical system is then focused, and the sample is further leveled. The final image captured by the camera is the original phase-shifted interference image of the sample, as shown below. Figure 3 As shown. A five-step fixed-step phase-shift method is proposed to extract the phase information of the sample surface height. This algorithm requires a single interference phase shift of [value missing]. Single phase shift distance Five phase-shifting interferometric images are acquired under a single field of view to obtain a phase-shifting interferometric image set, which encapsulates the phase. The calculation formula is:

[0099]

[0100] in, For the recorded number The phase-shifting interferometric image, with the wrapped phase calculation results under a single field of view, is as follows: Figure 4 As shown. Phase unwrapping is performed using the least squares (LS-LS) unwrapping algorithm based on transverse shearing interferometry proposed in step two. The phase result of the object surface height under this field of view is as follows. Figure 5 As shown.

[0101] Based on step three and the system hardware parameters, the micro-motion step size is determined. The value should be greater than 6.71 μm, so the movement step size is set to 6.9 μm, and the number of moved pixels is calculated. The sample was moved, and phase-shifting interferometry data acquisition and phase calculation were performed, obtaining sample surface height information under seven fields of view. The phase information of three adjacent fields of view were grouped together, and second-order difference calculations and the superposition of five second-order difference results were performed as described in step four, resulting in the following... Figure 6 The intermediate process results shown ,inFigure 6 The tripolar pattern within the central rectangular area is formed by submicron defects during processing, while the anomaly in the circular area in the upper right corner is caused by surface impacts. Significant background fluctuations can be observed in intermediate result A, which interferes with the identification of the tripolar pattern.

[0102] As described in step five, a third-order Butterworth high-pass frequency domain filter is constructed. The filter cutoff frequency should satisfy... Here, the cutoff frequency is set to (This frequency was determined through experimental testing and optimization.) Regarding the intermediate results... Perform frequency domain filtering to obtain, as follows Figure 7 Intermediate results shown Clearly, background fluctuations are significantly suppressed through frequency domain high-pass filtering.

[0103] As described in step six, intermediate results are taken sequentially. The standard deviations were calculated for a 150 × 150 pixel sub-region at the center, top left, top right, bottom left, and bottom right of the field of view. The results were 0.0364, 0.0350, 0.0343, 0.0375, and 0.0318, respectively. The final value was... And set a bilateral filter 0.035, For intermediate process results Bilateral filtering is performed to obtain the following: Figure 8 The results shown Compared to intermediate results , The signal strength of the tripolar pattern was preserved, while background noise in non-defect areas was significantly suppressed. Table 2 shows the intermediate results. Intermediate results Enhance results The background standard deviation and peak signal-to-noise ratio of the three-pole pattern were calculated. The results show that the background noise of the signal was suppressed and the signal-to-noise ratio was significantly improved through the above two filtering processes.

[0104] Table 2 Intermediate Results Intermediate results Enhance results Parameter statistics table

[0105]

[0106] As described in step seven, set the detection threshold for the center of the pit defect. detection thresholds on both sides Perform pit-type defect identification. Set the central detection threshold for protrusion-type defects as described in step eight and identify protrusion-type defects. Finally, mark pixels identified as pit-type defects in white. Figure 9 Demonstrates the enhancement results The marking results of the three-pole pattern formed by the processing design defect in the central 20 μm × 15 μm region show that the method successfully detected and located a patternless wafer surface defect smaller than one-third of the system's optical resolution, verifying the effectiveness of the method and demonstrating extremely high sensitivity to defect signals.

[0107] For patterned wafers, the micro-step distance can be set according to the minimum arrangement period of the pattern structure on the wafer. After performing second-order difference calculations, the influence of the wafer's own pattern structure on defect identification can be removed, thereby enabling rapid identification and location of various defects on the patterned wafer surface.

Claims

1. A wafer defect detection method based on phase shift interferometry and moving differential, characterized in that, The method comprises the following steps: Step one, place the measured wafer under the optical phase shift interference microscope and adjust the position to make 1-2 complete period interference fringes appear in the field of view, collect multiple phase shift interference images in the height direction of the measured wafer to obtain a phase shift interference image set; Step two, process the phase shift interference image set to complete the calculation of the three-dimensional topography of the surface of the measured wafer under single field of view; Step three, horizontal micro-motion step distance The measured wafer is micro-moved along the horizontal X direction, and then steps one and two are repeated to collect a phase shift interference image set under different fields of view, and complete the three-dimensional topography calculation of the measured wafer surface under the field of view; Step four, the three-dimensional topography of the wafer surface measured under three adjacent fields of view is divided into one group of data, and second-order difference calculation is sequentially performed on each group of data, a total of second-order difference results are obtained and recorded as ; according to the calculation sequence of and , the second-order difference results are cropped and superimposed to obtain an intermediate result containing a three-pole pattern of wafer defects ; Step five, eliminating intermediate results A Residual background fluctuations present in the middle, and enhance the signal-to-noise ratio, resulting in enhanced results ; Step six, setting the central detection threshold and the two side detection thresholds of the three-pole pattern formed by the surface defects of the wafer under test, in combination with the number of moving pixels Step six, setting the central detection threshold and the two side detection thresholds of the three-pole pattern formed by the surface defects of the wafer under test, in combination with the number of moving pixels Step six, setting the central detection threshold and the two side detection thresholds of the three-pole pattern formed by the surface defects of the wafer under test, in combination with the number of moving pixels Step six, setting the central detection threshold and the two side 2. The phase shift interferometry and movement differential based wafer defect detection method of claim 1, wherein, When the phase shift interference image is collected, the light source is a quasi-monochromatic light source, and the bandwidth of the quasi-monochromatic light source is less than 20 nm.

3. The phase shift interferometry and movement differential based wafer defect detection method of claim 1, wherein, In the step one, the piezoelectric ceramic displacement device is controlled to collect multiple phase shift interference images in the height direction of the measured wafer sample; in the step two, the phase shift interference image set is processed by using a wrapped phase extraction algorithm and a phase unwrapping algorithm to complete the calculation of the three-dimensional topography of the surface of the measured wafer under single field of view.

4. The phase shift interferometry and movement differential based wafer defect detection method of claim 1, wherein, The horizontal micro-motion step distance is expressed as: The number of pixels moved is expressed as: wherein is a positive integer greater than 1, when the measured wafer is a non-patterned wafer, is the minimum reliable step distance of the horizontal displacement stage X direction, when the measured wafer is a processed patterned wafer, is the minimum spatial repeat period of the processed pattern on the patterned wafer in the horizontal X direction; is the optical resolution of the optical phase shift interferometry device, is the magnification of the optical phase shift interferometry device, is the pixel size of the image detector in the optical phase shift interferometry microscope device, denotes the rounding to the nearest integer. is the rounding to the nearest integer.

5. The phase shift interferometry and movement differential based wafer defect detection method of claim 1, wherein, The step four, second order difference result The calculation method is wherein, , , respectively represent the calculation results of the three-dimensional topography of the wafer surface under the first , the second , and the third field of view; the value at the pixel coordinate is calculated as follows: ​ All of i.e. constitutes an intermediate result .

6. The phase shift interferometry and movement differential based wafer defect detection method of claim 1, wherein, In step five, firstly... Perform a Fast Fourier Transform (FFT), then create a frequency domain high-pass filter to filter it in the frequency domain, and finally perform an inverse FFT to obtain the intermediate result. Finally, based on the intermediate results Background standard deviation A bilateral filter is created for intermediate results using the optical resolution of the optical phase-shifting interferometer. Perform bilateral filtering to enhance intermediate results. The signal-to-noise ratio is improved to obtain the enhanced result. .

7. The phase shift interferometry and movement differential based wafer defect detection method of claim 6, wherein, The cut-off frequency of the frequency domain high-pass filter satisfies wherein , are the horizontal and vertical resolution of the phase-contrast interference microscope image detector, respectively; the spatial Gaussian standard deviation of the bilateral filter satisfies , the pixel value similarity Gaussian standard deviation satisfies .

8. The phase shift interferometry and movement differential based wafer defect detection method according to any one of claims 1 to 7, characterized in that, The surface defects of the measured wafer include pit defects and protrusion defects; and the step six: Set the central detection threshold for the tripolar pattern formed by the pits on the surface of the wafer under test. and detection thresholds on both sides ; will enhance the results All less than Pixels marked as suspected pit defects are indicated by the horizontal X-axis direction of these pixels. The distance is the value of the pixels on both sides of the search, and compared with... The comparison showed that only the pixel values ​​on both sides of the suspected pit defect were greater than [value missing]. Only then is the pixel location identified as a pit defect; Setting central detection threshold of three-pole pattern formed by protrusion of wafer surface under test and side detection threshold ; marking all pixel points greater than in the enhanced result as suspected protrusion defects, searching the values of pixel points at a distance of in horizontal X direction of these pixel points, and comparing with , only when the values of pixel points at both sides of suspected protrusion defects are less than , the pixel position is determined as protrusion defect.

9. The phase shift interferometry and movement differential based wafer defect detection method of claim 1, wherein, The central detection threshold satisfies , the two-side detection threshold satisfies , , .

10. The phase shift interferometry and movement differential based wafer defect detection method of claim 1, wherein, The measured wafer is a transparent or non-transparent planar optical element.

Images