Defect inspection device and sensor data processing method

Abstract

A defect inspection device has a signal processing unit 105 that comprises a signal integration unit 302 that integrates sensor data from photoelectric conversion units 103 of a plurality of detection units 102 at which an image of scattered light from the same position on a sample W is formed. The signal integration unit performs filter processing on the sensor data from each of the plurality of detection units and adds all the sensor data that has undergone the filter processing. The filter processing takes a cross-correlation with a filter that has been set to the respective detection unit and has a filter coefficient that has been set in a pixel arrangement direction, the filter being selected in accordance with the height displacement of the sample from a height measurement unit and a pixel position for a light reception surface at which the sensor data is detected.

Description

Defect inspection device and sensor data processing method

[0001] The present invention relates to a defect inspection device and a sensor data processing method. For example, it relates to a defect inspection device and its sensor data processing method for inspecting minute defects present on the surface of a sample, determining the position, type, and dimensions of the defects, and outputting them.

[0002] In a manufacturing line of semiconductor substrates, thin film substrates, etc., in order to maintain and improve the product yield, inspection of defects present on the surface of semiconductor substrates, thin film substrates, etc. is carried out.

[0003] In Patent Document 1, in order to detect scattered light generated from minute defects, a plurality of detection systems are arranged in a direction inclined with respect to the sample surface, and an image of linear illumination irradiated on the sample surface is imaged at the sensor position by each detection system to perform defect determination. When the optical system is arranged so that the image of the linear illumination can be detected obliquely, the working distance between the detection unit and the linear illumination unit of the sample surface changes within the field of view. At this time, defocus occurs, so the resolution of the image formed on the sensor surface decreases. In order to suppress this, it is described that by tilting the sensor so as to be conjugate with the inspection target substrate W according to the inclination of the detection direction with respect to the inspection target substrate W, the image on the inspection target substrate W can be imaged obliquely from above on the detector.

[0004] Japanese Patent Application Laid-Open No. 2007-033433

[0005] Defect inspection used in the manufacturing processes of semiconductors and other materials requires the detection of minute defects, high-precision measurement of the dimensions of detected defects, non-destructive inspection of samples, obtaining substantially consistent inspection results regarding, for example, the number, location, dimensions, and type of detected defects when inspecting the same sample, and inspecting a large number of samples within a certain time. However, in the method disclosed in Patent Document 1, fluctuations in the height of the sample surface that occur during inspection greatly affect the inspection results. During inspection, the sample rotates at a high speed of several thousand RPM (Rotations Per Minute), and the position of the sample fluctuates in the height direction, that is, in the direction intersecting the sample surface, due to sample warping, airflow, and runout of the chuck and rotation axis of the sample stage. Such height fluctuations occur at a frequency of up to the same frequency as the rotation period of the sample. When the height of the sample fluctuates, the sample surface moves out of the focus of the detection optical system, reducing the image resolution, causing the image to extend beyond the pixels and reducing the signal output by the sensor. Furthermore, if there are magnification errors or eccentricity of optical elements in the detection optical system, the image shape may change depending on the position on the sensor.

[0006] The objective of the present invention is to suppress the decrease in resolution caused by fluctuations in the height of the sample surface in a defect inspection device.

[0007] To solve the above problems, for example, the configuration described in the claims may be adopted. The present invention includes several means for solving the above problems, but to give one example, it includes a stage on which a sample is mounted and moved, an illumination unit that illuminates the sample with an illumination spot, a plurality of detection units that detect scattered light from the sample illuminated by the illumination spot, a height measuring unit that measures the height displacement of the sample illuminated by the illumination spot, and a signal processing unit. The detection unit comprises a photoelectric conversion unit and a detection optical system that images the scattered light onto the light-receiving surface of the photoelectric conversion unit, the light-receiving surface of the photoelectric conversion unit is divided into a plurality of pixels, and the imaged scattered light is converted into an electrical signal for each pixel and output as sensor data. The signal processing unit comprises a signal integration unit that integrates the sensor data from the photoelectric conversion units of the plurality of detection units onto which scattered light from the same position on the sample is imaged. The signal integration unit performs a filtering process on each of the sensor data from the plurality of detection units, taking cross-correlation with a filter set for each detection unit and having a filter coefficient defined in the direction of the pixel arrangement, and adds the filtered sensor data from the plurality of detection units together. The filter is selected according to the height displacement of the sample from the height measuring unit and the pixel position on the light-receiving surface where the sensor data was detected.

[0008] This invention provides a defect inspection apparatus that suppresses the reduction in resolution caused by variations in the height of the sample surface. Other challenges and novel features will become apparent from the description and accompanying drawings herein.

[0009] This is a schematic diagram of the defect inspection device. This is a diagram showing an example of the illumination intensity distribution shape realized by the illumination unit. This is a diagram showing an example of the illumination intensity distribution shape realized by the illumination unit. This is a diagram showing an example of the illumination intensity distribution shape realized by the illumination unit. This is a diagram showing an example of the illumination intensity distribution shape realized by the illumination unit. This is a diagram showing an example of the illumination intensity distribution shape realized by the illumination unit. This is a diagram showing the illumination distribution shape on the sample surface and the scanning direction. This is a diagram showing the trajectory of the illumination spot due to scanning. This is a side view of the arrangement and detection direction of the detection unit. This is a top view of the arrangement and detection direction of the detection unit. This is a side view of the arrangement and detection direction of the detection unit. This is a diagram showing an example of the configuration of the detection unit in Example 1. This is a functional block diagram of the signal processing unit. This is a diagram showing a first configuration example of the illumination spot and the photoelectric conversion unit. This is a diagram showing a first configuration example of the illumination spot and the photoelectric conversion unit. This is a diagram showing a second configuration example of the illumination spot and the photoelectric conversion unit. This is a diagram explaining the arrangement of the detection unit. This is a diagram showing an example of the image shape and filter with respect to the sensor pixel array direction. This is a diagram showing an example of the image shape and filter with respect to the sensor pixel array direction. This is a diagram showing an example of the image shape and filter with respect to the sensor pixel array direction. This diagram illustrates the height displacement of the sample surface in the detection unit. This diagram shows an example of image shape change when sample surface height displacement occurs. This diagram shows an example of image shape change when sample surface height displacement occurs. This is a detailed functional block diagram of the signal integration unit. This is an example of the data structure of the filter database. This is a functional block diagram of the control unit related to filter creation. This diagram shows simulation data (image shape relative to the sensor pixel array direction). This diagram shows simulation data (image shape relative to the sensor pixel array direction). This diagram shows an example of the detection unit configuration in Example 3. This diagram shows the image shape relative to the sensor pixel array direction. This diagram shows the image shape relative to the sensor pixel array direction.

[0010] Embodiments of the present invention will be described below with reference to the figures. Note that the present invention is not limited to the embodiments described below, and includes various modifications. The embodiments described below are described in detail to make the present invention easier to understand, and are not necessarily limited to those having all the described configurations. Furthermore, it is possible to replace parts of the configuration of one embodiment with those of another embodiment, and to add other embodiments to the configuration of one embodiment. Also, it is possible to add, delete, or replace parts of the configuration of each embodiment with other configurations.

[0011] Figure 1 shows a schematic example of a defect inspection apparatus. The defect inspection apparatus mainly consists of an illumination unit 101, a detection unit 102 (including a detection optical system and a photoelectric conversion unit 103), a stage 104 on which the placed sample W can be moved perpendicular to the surface by an actuator, a signal processing unit 105, a control unit 53, a display unit 54, and an input unit 55. The illumination unit 101 includes, for example, a laser light source 2, an attenuator 3, an output light adjustment unit 4, a beam expander 5, a polarization control unit 6, and an illumination intensity distribution control unit 7. The laser light beam emitted from the laser light source 2 is adjusted to a desired beam intensity by the attenuator 3, adjusted to a desired beam position and beam propagation direction by the output light adjustment unit 4, adjusted to a desired beam diameter by the beam expander 5, adjusted to a desired polarization state by the polarization control unit 6, and adjusted to a desired intensity distribution by the illumination intensity distribution control unit 7, and is irradiated onto the inspection target area of ​​the sample W.

[0012] The angle of incidence of illumination light on the sample surface is determined by the position and angle of the reflective mirror of the output light adjustment unit 4, which is located in the optical path of the illumination unit 101. The angle of incidence of illumination light is set to an angle suitable for detecting minute defects. The larger the angle of incidence of illumination, that is, the smaller the angle of elevation of illumination (the angle between the sample surface and the illumination light axis), the weaker the scattered light from minute irregularities on the sample surface (called haze), which becomes noise in response to scattered light from minute foreign matter on the sample surface, making it suitable for detecting minute defects. For this reason, if scattered light from minute irregularities on the sample surface hinders the detection of minute defects, it is preferable to set the angle of incidence of illumination light to 75 degrees or more (elevation angle of 15 degrees or less). On the other hand, in oblique incidence illumination, the smaller the angle of incidence of illumination, the larger the absolute amount of scattered light from minute foreign matter. Therefore, if insufficient scattered light from defects hinders the detection of minute defects, it is preferable to set the angle of incidence of illumination light to 60 degrees or more and 75 degrees or less (elevation angle of 15 degrees or more and 30 degrees or less). Furthermore, when performing oblique incidence illumination, the polarization control in the polarization control unit 6 of the illumination unit 101 controls the polarization of the illumination to P-polarization, which increases scattered light from defects on the sample surface compared to other polarizations. Also, if scattered light from minute irregularities on the sample surface interferes with the detection of minute defects, the polarization of the illumination to S-polarization reduces scattered light from minute irregularities on the sample surface compared to other polarizations.

[0013] Furthermore, if necessary, as shown in Figure 1, the illumination path can be changed by inserting a mirror 107 into the optical path of the illumination unit 101 and arranging other mirrors as appropriate, so that illumination light is irradiated from a direction substantially perpendicular to the sample surface (vertical illumination). At this time, the illumination intensity distribution on the sample surface is controlled by the illumination intensity distribution control unit 7v in the same way as oblique incidence illumination. To obtain scattered light from concave defects on the sample surface (polishing scratches or crystal defects in crystalline materials), vertical illumination, which is incident substantially perpendicular to the sample surface, is suitable.

[0014] For laser light source 2, to detect minute defects near the sample surface, a high-power laser with an output of 2W or more is used that emits a short-wavelength (wavelength 355 nm or less) ultraviolet or vacuum ultraviolet laser beam, which does not easily penetrate into the sample. The output beam diameter is about 1 mm. To detect defects inside the sample, a laser light source that emits a visible or infrared laser beam, which penetrates into the sample easily, is used.

[0015] The attenuator 3 comprises, for example, a first polarizer, a half-wave plate rotatable around the optical axis of the illumination light, and a second polarizer. Light incident on the attenuator 3 is converted to linear polarization by the first polarizer, its polarization direction is rotated in any direction according to the lagging axis azimuth angle of the half-wave plate, and it passes through the second polarizer. By controlling the azimuth angle of the half-wave plate, the light intensity is attenuated by an arbitrary ratio. The first polarizer is not necessarily required if the degree of linear polarization of the light incident on the attenuator 3 is sufficiently high. The attenuator 3 used is one in which the relationship between the input signal and the attenuation rate has been calibrated in advance. As the attenuator 3, it is possible to use an ND filter with a gradient density distribution, or to switch between and use multiple ND filters with different densities.

[0016] The light emission adjustment unit 4 is equipped with multiple reflective mirrors. Here, an embodiment consisting of two reflective mirrors is described, but it is not limited to this, and three or more reflective mirrors may be used. Here, we will tentatively define a three-dimensional Cartesian coordinate system (XYZ coordinates) and assume that the incident light to the reflective mirror is traveling in the +X direction. The first reflective mirror is set up to deflect the incident light in the +Y direction (incident and reflection in the XY plane), and the second reflective mirror is set up to deflect the light reflected by the first reflective mirror in the +Z direction (incident and reflection in the YZ plane). The position and direction of travel (angle) of the light emitted from the light emission adjustment unit 4 are adjusted by translation and tilt angle adjustment of each reflective mirror. As described above, by arranging the incident and reflection surfaces (XY plane) of the first reflective mirror and the incident and reflection surfaces (YZ plane) of the second reflective mirror to be orthogonal, the position and angle of the light emitted from the light emission adjustment unit 4 (traveling in the +Z direction) in the XZ plane and the position and angle of the light in the YZ plane can be adjusted independently.

[0017] The beam expander 5 has two or more lens groups and has the function of expanding the diameter of the incident parallel light beam. For example, a Galilean-type beam expander equipped with a combination of concave and convex lenses is used. The beam expander 5 is installed on a translation stage with two or more axes and its position can be adjusted so that the center coincides with a predetermined beam position. In addition, a tilt angle adjustment function for the entire beam expander 5 is provided so that the optical axis of the beam expander 5 coincides with the predetermined beam optical axis. The degree of expansion of the light beam diameter can be controlled by adjusting the spacing between the lenses (zoom mechanism). If the light incident on the beam expander 5 is not parallel, the expansion of the light beam diameter and collimation (conversion of the light beam to quasi-parallel light) are performed simultaneously by adjusting the spacing between the lenses. Collimation of the light beam may also be performed by installing a collimating lens upstream of the beam expander 5, independently of the beam expander 5. The beam diameter expansion ratio by the beam expander 5 is about 5 to 10 times, and a beam with a diameter of 1 mm emitted from the light source is expanded to about 5 mm to 10 mm.

[0018] The polarization control unit 6 is composed of a half-wave plate and a quarter-wave plate, and controls the polarization state of the illumination light to any desired polarization state. Along the optical path of the illumination unit 101, the state of the light incident on the beam expander 5 and the light incident on the illumination intensity distribution control unit 7 is measured by the beam monitor 22.

[0019] Figures 2 to 6 show schematic diagrams of the positional relationship between the illumination optical axis 120 guided from the illumination unit 101 to the sample surface and the shape of the illumination intensity distribution. Note that the configuration of the illumination unit 101 in Figures 2 to 6 shows only a part of the configuration of the illumination unit 101, and the output light adjustment unit 4, mirror 21, beam monitor 22, etc. are omitted. Figure 2 shows a schematic diagram of the cross section of the incident surface (the surface including the illumination optical axis and the normal to the sample surface) of oblique incident illumination. Oblique incident illumination is inclined with respect to the sample surface within the incident surface. The illumination unit 101 creates a substantially uniform illumination intensity distribution within the incident surface. The length of the portion with uniform illumination intensity is approximately 100 μm to 4 mm in order to inspect a large area per unit time.

[0020] Figure 3 shows a schematic cross-section of a plane that includes the normal to the sample surface and is perpendicular to the incident plane of oblique incidence illumination. Within this plane, the illumination intensity distribution on the sample surface forms an illumination intensity distribution where the intensity is weaker at the periphery than at the center. More specifically, it is an intensity distribution similar to a Gaussian distribution that reflects the intensity distribution of light incident on the illumination intensity distribution control unit 7, or an intensity distribution similar to a first-order Bessel function of the first kind or a sin function that reflects the aperture shape of the illumination intensity distribution control unit 7. The length of the illumination intensity distribution within this plane (the length of the region with an illumination intensity of 13.5% or more of the maximum illumination intensity) is shorter than the length of the portion where the illumination intensity is uniform within the incident plane, and is approximately 2.5 μm to 20 μm, in order to reduce haze generated from the sample surface. The illumination intensity distribution control unit 7 includes optical elements such as aspherical lenses, diffractive optical elements, cylindrical lens arrays, and light pipes, which will be described later. The optical elements constituting the illumination intensity distribution control unit 7 are installed perpendicular to the illumination optical axis, as shown in Figures 2 and 3.

[0021] A variation of the illumination intensity distribution created on the sample surface by the illumination unit 101 will be described. As an alternative to the illumination intensity distribution described above, which is long in one direction (linear) and has substantially uniform intensity in the longitudinal direction, it is also possible to use an illumination intensity distribution that has a Gaussian distribution in the longitudinal direction. A Gaussian distribution illumination that is long in one direction is formed by having a spherical lens in the illumination intensity distribution control unit 7 and forming a long elliptical beam in one direction with the beam expander 5, or by configuring the illumination intensity distribution control unit 7 with multiple lenses including a cylindrical lens. By setting some or all of the spherical lens or cylindrical lens of the illumination intensity distribution control unit 7 parallel to the sample surface, an illumination intensity distribution that is long in one direction and narrow in the direction perpendicular to it is formed on the sample surface. Compared to the case of creating a uniform illumination intensity distribution, the fluctuation of the illumination intensity distribution on the sample surface due to fluctuations in the state of light incident on the illumination intensity distribution control unit 7 is small, the stability of the illumination intensity distribution is high, and the light transmittance is high and the efficiency is good compared to the case in which diffractive optical elements or microlens arrays are used in the illumination intensity distribution control unit 7.

[0022] The state of the illumination light in the illumination unit 101 is measured by beam monitors 22 and 23. Beam monitors 22 and 23 measure and output the position and angle (direction of propagation) of the illumination light that has passed through the output light adjustment unit 4, and the position and wavefront of the illumination light incident on the illumination intensity distribution control unit 7, respectively. The position of the illumination light is measured by measuring the centroid of the light intensity of the illumination light. Specific position measurement means include a position-sensitive detector (PSD), or an image sensor such as a CCD sensor or CMOS sensor. The angle of the illumination light is measured by a position-sensitive detector or image sensor installed at a position farther from the light source than the aforementioned position measurement means, or at the focusing position of the collimating lens. The illumination light position and illumination light angle detected by these sensors are input to the control unit 53 and displayed on the display unit 54. If the illumination light position or angle deviates from a predetermined position or angle, the output light adjustment unit 4 adjusts it back to the predetermined position.

[0023] Wavefront measurement of the illumination light is performed to measure the parallelism of the light incident on the illumination intensity distribution control unit 7. If the wavefront measurement reveals that the light incident on the illumination intensity distribution control unit 7 is not quasi-parallel light but diverges or converges, the lens group of the preceding beam expander 5 can be displaced in the optical axis direction to bring it closer to quasi-parallel light. Furthermore, if the wavefront measurement reveals that the wavefront of the light incident on the illumination intensity distribution control unit 7 is partially tilted, a spatial light phase modulation element, a type of spatial light modulation element (SLM), can be inserted before the illumination intensity distribution control unit 7. By applying an appropriate phase difference at each position in the cross-section of the light beam so that the wavefront becomes flatter, the wavefront can be brought closer to flat, i.e., the illumination light can be brought closer to quasi-parallel light. Through the above wavefront accuracy measurement and adjustment means, the wavefront accuracy of the light incident on the illumination intensity distribution control unit 7 (deviation from a predetermined wavefront (design value or initial state)) can be kept below λ / 10rms.

[0024] The illumination intensity distribution on the sample surface, adjusted by the illumination intensity distribution control unit 7, is measured by the illumination intensity distribution monitor 24. Similarly, even when using vertical illumination, as shown in Figure 1, the illumination intensity distribution on the sample surface, adjusted by the illumination intensity distribution control unit 7v, is measured by the illumination intensity distribution monitor 24. The illumination intensity distribution monitor 24 detects the sample surface as an image by forming an image on an image sensor such as a CCD sensor or CMOS sensor via a lens. The image of the illumination intensity distribution detected by the illumination intensity distribution monitor 24 is processed by the control unit 53, and the centroid position of the intensity, the maximum intensity, the position of the maximum intensity, the width and length of the illumination intensity distribution (width and length of the illumination intensity distribution region that is above a predetermined intensity or above a predetermined ratio to the maximum intensity value) are calculated and displayed on the display unit 54 along with the contour shape and cross-sectional waveform of the illumination intensity distribution.

[0025] When performing oblique incidence illumination, height displacement of the sample surface causes displacement of the illumination intensity distribution and disturbance of the illumination intensity distribution due to defocusing. To suppress this, the height of the sample surface is measured by the height measurement unit 106, and if the height is off, the displacement is corrected by the illumination intensity distribution control unit 7 or by adjusting the height along the Z axis of the stage 104.

[0026] The illuminance distribution shape (illumination spot 20) formed on the sample surface by the illumination unit 101 and the method of scanning the sample with the illumination spot 20 will be explained using Figures 7 and 8. A circular semiconductor silicon wafer is assumed as the sample W. The stage 104 includes a translation stage, a rotation stage, and a Z stage for adjusting the sample surface height (none of which are shown). As described above, the illumination spot 20 has an illumination intensity distribution that is long in one direction, which is denoted as S2, and the direction substantially perpendicular to S2 is denoted as S1. The rotational motion of the rotation stage scans in the circumferential direction S1 of a circle centered on the rotation axis of the rotation stage, and the translational motion of the translation stage scans in the translational direction S2 of the translation stage. While the sample is rotated once by scanning in the scanning direction S1, the illumination spot 20 is scanned in the scanning direction S2 by a distance less than or equal to the longitudinal length of the illumination spot 20, so that the illumination spot traces a helical trajectory T on the sample W, and the entire surface of the sample W is scanned.

[0027] Next, the detection unit 102 will be described. Multiple detection units 102 are arranged to detect scattered light from multiple directions emitted from the illumination spot 20. An example of the arrangement of the detection unit 102 relative to the sample W and the illumination spot 20 will be explained using Figures 9 and 10. Figure 9 shows a side view of the arrangement of the detection unit 102. The angle between the detection direction (direction of the center of the detection aperture) made by the detection unit 102 and the normal to the sample W is defined as the detection zenith angle. The detection unit 102 has multiple detection units to cover scattered light in multiple directions. Figure 10 shows a plan view of the arrangement of the detection unit 102. In a plane parallel to the surface of the sample W, the angle between the direction of propagation of obliquely incident illumination and the detection direction is defined as the detection azimuth angle. The detection unit 102 appropriately includes a front detection unit 102f, a rear detection unit 102b, and front detection units 102f' and rear detection units 102b' positioned symmetrically with respect to the illumination incident surface. For example, the forward detection unit 102f is installed with a detection azimuth angle of 0 degrees or more and 90 degrees or less, and the rear detection unit 102b is installed with a detection azimuth angle of 90 degrees or more and 180 degrees or less. However, it is not limited to this, and the number and position of the detection units may be changed as appropriate. As shown in Figure 11, multiple detectors may be placed relative to the detection zenith angle. For example, a high-angle detection unit 102h with a detection zenith angle of 45 degrees or less and a low-angle detection unit 102l with a detection zenith angle of 45 degrees or more may be used as appropriate.

[0028] Using Figure 12A, a specific example of the detection optical system of the detection unit 102 will be explained. Scattered light generated from the illumination spot 20 is focused by the objective lens 1021, and the polarization direction is controlled by the polarization control filter 1022. As the polarization control filter 1022, for example, a half-wave plate whose rotation angle can be controlled by a drive mechanism such as a motor is used. To efficiently detect scattered light, it is preferable that the detection NA of the objective lens 1021 be 0.3 or higher. The lower end of the objective lens 1021 is notched as necessary so that it does not interfere with the sample surface W. The imaging lens 1023 forms an image of the illumination spot 20 at the position of the aperture 1024. The aperture 1024 is set to allow only the light of the region to be detected by the photoelectric conversion unit 103 to pass through the image formed by the illumination spot 20. If the illumination spot 20 has a Gaussian distribution profile in its longitudinal direction S2, the aperture 1024 allows only the central part of the Gaussian distribution, where the light intensity is strong in the longitudinal direction S2, to pass through, while blocking the weaker light intensity region at the beam edge. In addition, in the direction S1 perpendicular to the longitudinal direction S2, the aperture is made to be approximately the same size as the image formed by the illumination spot 20, suppressing disturbances such as air scattering that occur when the illumination passes through the air. The focusing lens 1025 focuses the image formed by the aperture 1024 again. The polarizing beam splitter 1026 separates the light whose polarization direction has been converted by the polarization control filter 1022 according to its polarization direction. The imaging lens 1027 forms an image of the illumination spot 20 on the photoelectric conversion unit 103. It is also possible to use a cylindrical lens as the imaging lens 1027 to perform imaging in only one direction. In this embodiment, a combination of a half-wave plate as the polarization control filter 1022 and a polarizing beam splitter 1026 was used to detect only light with a specific polarization direction from the light focused by the objective lens 1021 in the photoelectric conversion unit 103. However, as an alternative, for example, the polarization control filter 1022 could be a wire grid polarizer with a transmittance of 80% or more, and only light with a desired polarization direction could be extracted without using the polarizing beam splitter 1026.

[0029] The signal processing unit 105 shown in Figure 1 is a computer that processes detection signals input from the photoelectric conversion unit 103, and is composed of ROM, RAM, other memory, as well as a CPU, FPGA, timer, etc. While it is conceivable that the signal processing unit 105 is a single computer forming a unit with the main body of the defect inspection device (stage, lighting unit, detection unit, sensors, etc.), it may also be composed of multiple computers. In this case, a server can be used as one of the multiple computers. This is an example of including a server as a component of the defect inspection device. For example, a computer attached to the main body of the device can acquire defect detection signals from the main body, process the detection data as needed, and send it to the server, where the server can perform processing such as defect detection and classification.

[0030] Figure 12B shows a functional block diagram of the signal processing unit 105. The signal processing unit 105 includes a lighting spot position analysis unit 301, a signal integration unit 302, a defect detection unit 303, a data memory 311, and a filter memory 312. The lighting spot position analysis unit 301, the signal integration unit 302, and the defect detection unit 303 can be configured as functional units realized, for example, by a processor executing processing according to a program. Alternatively, all or part of each functional unit may be configured using an FPGA. Furthermore, for example, the processing of the defect detection unit 303 may be executed on an external server.

[0031] The illumination spot position analysis unit 301 performs AD conversion on the electrical signal from the photoelectric conversion unit 103 and analyzes the position of the illumination spot 20 on the sample based on the sensor data input to the signal processing unit 105. The illumination spot position analysis unit 301 stores the sensor data input from the photoelectric conversion unit 103 and the position data calculated by the illumination spot position analysis unit 301 as scattered light data in the data memory 311. The signal integration unit 302 integrates and calculates multiple scattered light data sets with different illumination spot positions output from the same sensor based on the scattered light data stored in the data memory 311, and integrates and calculates scattered light data sets that have been similarly integrated for different sensors. In this embodiment, prior to the integration calculation of scattered light data, filtering is performed to correct aberrations contained in the sensor data. The filters necessary for the filtering process are registered in the filter memory 312. Details of the signal integration unit 302 will be described later. The defect detection unit 303 extracts areas on the sample surface that show high brightness at high frequencies as defects based on the scattered light data after the integration calculation.

[0032] Let me elaborate on the processing of the signal integration unit 302. Integrating and calculating multiple scattered light data with different illumination spot positions output from the same sensor means adding the scattered light intensities of data scanned in the longitudinal direction of a linear illumination spot 20 at the same coordinate on the sample surface. Specifically, since the defect inspection device scans the sample W in a helical manner, if the coordinates on the sample surface are represented in the rθ coordinate system, the scattered light intensities of the same coordinate are added for scattered light data with the same θ coordinate but different r coordinates. Note that in the rθ coordinate system, the r coordinate is the radial coordinate on the sample surface, and the θ coordinate is the azimuth coordinate on the sample surface, which is a different concept from the detected zenith angle θ in Figure 16. Even with scattered light data output from the same sensor, for data with different r coordinates, the imaging position of the scattered light from the same coordinate (i.e., the pixel that receives the scattered light from the same coordinate) will shift by the amount of movement of the stage 104 during one rotation of the sample W. Therefore, based on the amount of movement of this stage 104, the correspondence between the pixels that are the source of the signal to be added as scattered light intensity at the same coordinates is calculated in advance.

[0033] The control unit 53 is a computer that comprehensively controls the defect inspection device, and like the signal processing unit 105, it is composed of ROM, RAM, other memory, as well as a CPU, FPGA, timer, etc. The control unit 53 is connected to the input unit 55, display unit 54, and signal processing unit 105 by wire or wireless. The functions of the signal processing unit 105 may be incorporated into the control unit 53, so that the control unit 53 also functions as the signal processing unit 105. The input unit 55 is a device for the user to input various operations, and various input devices such as a keyboard, mouse, and touch panel can be used as appropriate.

[0034] The control unit 53 receives input from the input unit 55, including encoders for the rotary and translational stages, and inspection conditions input from the input unit 55 in response to operator operations. Inspection conditions include, for example, the type, size, shape, and material of the sample W, as well as lighting conditions and detection conditions. The control unit 53 also outputs signals to command the operation of the detection unit 102, the stage 104, and the lighting unit 101 according to the inspection conditions of the sample W, and outputs coordinate data of the lighting spot 20 synchronized with the defect detection signal to the signal processing unit 105. The control unit 53 also displays the defect inspection results from the signal processing unit 105 on the display unit 54.

[0035] The following describes a configuration in which the photoelectric conversion unit 103 is positioned obliquely to the optical axis. Figures 13 and 14 show examples of the arrangement of the photoelectric conversion unit 103. The optical axis 121 is inclined with respect to the normal direction of the light receiving unit 1031. The light receiving unit 1031 of the photoelectric conversion unit 103 is positioned parallel to the longitudinal direction of the optical image 27 that is imaged on the light receiving unit 1031 by the detection unit 102 from a linear illumination spot 20 irradiated onto the surface of the sample W. The pair of cylindrical lenses 10210 and 10211 constitute a cylindrical beam expander, making the spread in the short direction γ of the optical image 27 imaged by the illumination spot 20 smaller than the spread in the longitudinal direction of the optical image 27.

[0036] The light-receiving section of the photoelectric conversion unit 103 is divided into multiple pixels, and the optical image 27 is converted into an electrical signal for each pixel and output as sensor data. The photoelectric conversion unit 103 is equipped with an array of light-receiving sections and an anti-reflective coating at a position conjugate to the illumination spot 20 irradiated onto the surface of the sample W. The light-receiving surface of the photoelectric conversion unit 103 is not conjugate to the sample surface in the short-side direction γ. However, since the short-side direction γ is also the short-side direction of the illumination spot 20, the image height of the optical image 27 is low, and almost no focus shift occurs. Therefore, by increasing the imaging magnification of the optical image 27 in the short-side direction γ, the variation in the incident angle to the light-receiving section 1031 can be reduced.

[0037] Figure 15 shows an example in which an adjustment area camera 1032 is placed on the detection unit 102. The area camera 1032 is positioned at an angle so as to be conjugate to the surface of the sample W, the optical path is branched by a half mirror 10212, and the illumination spot 20 is imaged onto the detection surface of the area camera 1032 by an imaging lens 1028. By using the area camera 1032, the short-side direction of the illumination spot 20 can also be observed.

[0038] Figure 16 shows a schematic diagram of the three-dimensional arrangement of the sample W and the detection unit 102. The optical axis 121 of the detection unit 102 is inclined by an angle θ with respect to the normal direction Z of the sample W. The projection of the optical axis 121 onto the sample surface is inclined by an angle φ with respect to the longitudinal direction S2 of the illumination spot 20.

[0039] If the optical axis 121 that detects light from the detection unit 102 is offset by an angle θ with respect to the normal direction Z of the sample W and by an angle φ with respect to the longitudinal direction S2 of the illumination spot 20, then in three-dimensional space, this optical axis 121 can be represented by the vector v0 in (Equation 1).

[0040] The angle α between this vector v0 and the longitudinal direction S2 of the lighting spot 20 can be found by (Equation 2).

[0041] At this time, the defect inspection device detects a section of 2L in the longitudinal direction S2 of the illumination spot 20 using the photoelectric conversion unit 103. Depending on the position x from the center of the field of view, the working distance (distance between the sample W and the detection unit 102) changes as shown in (Equation 3) Δz.

[0042] Note that in (Equation 3), |x| < L indicates that the distance from the center of the field of view is L or less, that is, it is a position within the illumination spot 20. The same applies to (Equation 4).

[0043] The imaging magnification (lateral magnification) M is determined by the objective lens 1021 and the imaging lens 1027. The position of the image formed here is represented by (Equation 4).

[0044] Generally, a line sensor is arranged to be orthogonal to the optical axis, which is the center of the light beam emitted by the imaging lens. However, in this embodiment, by tilting the photoelectric conversion unit 103, imaging detection without focus deviation is realized regardless of the change in the field of view of the working distance. At this time, the optical axis 121 incident on the photoelectric conversion unit 103 and the pixel array vector v1 of the light receiving surface are in the plane spanned by the longitudinal direction S2 of the illumination spot 20 and the vector v0, and the angle β formed by the vector v1 and the vector v0 is set to satisfy (Equation 5).

[0045] Here, the angle α is the angle formed by the vector v2 and the vector v0, and satisfies the relationship of (Equation 6).

[0046] From (Equation 5), when the imaging magnification M increases, the angle β formed by the vector v0 and the vector v1 decreases and the incident angle increases. When M = 2, the light ray with the largest incident angle enters the photoelectric conversion unit 103 at an angle close to 90 degrees. The absorption rate of the antireflection film of the photoelectric conversion unit 103 has an incident angle dependence, and the absorption rate becomes very small at an incident angle close to 90 degrees. To prevent this, the imaging magnification M is set to 2 times or less.

[0047] Under these conditions, the angle between the optical axis 121 incident on the photoelectric conversion unit 103 from the imaging lens 1027 and the vector v1 can be maximized. If the numerical aperture of the incident light beam of the objective lens 1021 is N, the spread of the light beam emitted to the photoelectric conversion unit 103 is obtained by multiplying the spread of the light beam emitted to the imaging lens 1027 by the reciprocal of the imaging magnification M. As mentioned above, in order to keep the imaging magnification M at 2x or less, if a lens with a large numerical aperture is used as the objective lens 1021, light will be incident on the photoelectric conversion unit 103 from a wide range of directions. Due to the characteristics of the anti-reflective coating on the photoelectric conversion unit 103, if the range of incident light angles to the photoelectric conversion unit 103 is wide, the light absorption rate of the photoelectric conversion unit 103 decreases, making high sensitivity difficult. For this reason, the imaging magnification M is set to 1x or higher.

[0048] In this embodiment, the aberration W(x) when scattered light from a defect located at position x in the S2 direction from the center of the field of view is imaged onto the light-receiving surface of the photoelectric conversion unit 103 is expressed as (Equation 7).

[0049] Here, M is the imaging magnification, α is the angle between the optical axis vector v0 and the longitudinal direction S2 of the illumination spot 20, and r is the radius of the pupil plane (Fourier transform plane) of the objective lens 1021. From (Equation 7), it can be seen that when the imaging magnification M is not 1x, aberrations that depend on the location of the defect occur.

[0050] The defect inspection device is equipped with multiple detection units 102. However, the image magnification (lateral magnification) M of each detection unit 102 may deviate from the specified value due to errors in the focal length of the lens or distortion aberration. In addition, eccentricity of the optical elements may occur due to environmental fluctuations such as atmospheric pressure and temperature, or errors during optical system adjustment. As a result, blurring and positional variations occur in the images formed on the light-receiving surface of the photoelectric conversion unit 103 by the multiple detection units 102. In other words, the pixel positions on the light-receiving surface corresponding to the same coordinate on the sample surface W shift. Consequently, the signal processing unit 105 becomes unable to add the scattered light intensities of the same coordinates between the multiple detection units 102, resulting in a decrease in sensitivity.

[0051] Figures 17 and 18 illustrate the change in image formation when there is a magnification error. Figures 17 and 18 show the image formed when scattered light from a point on the sample (corresponding to the defect location) is projected onto the light-receiving surface of the photoelectric conversion unit 103, obtained by simulation assuming that the inspection optical system is composed of an ideal lens satisfying the sinusoidal condition. Here, the specified value of the imaging magnification (lateral magnification) M is set to 1x, the angle α = 56.8 deg between the optical axis 121 and the longitudinal direction S2 of the illumination spot 20 on the sample surface, and the angle β = 56.8 deg between the optical axis 121 and the pixel arrangement direction of the photoelectric conversion unit 103. The pixel size of the photoelectric conversion unit 103 is set to 5 μm. In all cases, the height displacement of the sample surface is assumed to be zero (z = 0 μm).

[0052] Figure 17 shows the case with no magnification error. The upper part of Figure 17 shows the image 401 on the light-receiving surface of the photoelectric conversion unit 103 when the defect position in the S2 direction x = 0 μm (center of the field of view), and the image 402 on the light-receiving surface of the photoelectric conversion unit 103 when the defect position in the S2 direction x = 600 μm. In each case, the horizontal axis is the position in the direction conjugate to the sensor pixel array direction (vector v1), that is, the scanning direction of the sample (longitudinal direction S2 of the illumination spot 20). Figure 18 shows the case with a magnification error of 1%. Similarly, the upper part of Figure 18 shows the image 411 on the light-receiving surface of the photoelectric conversion unit 103 when the defect position in the S2 direction x = 0 μm, and the image 412 on the light-receiving surface of the photoelectric conversion unit 103 when the defect position in the S2 direction x = 600 μm.

[0053] When comparing the case with a magnification error of 1% (Figure 18) and the case with no magnification error (Figure 17), nearly identical images are obtained at the defect position x = 0 μm. However, at the defect position x = 600 μm, blurring and positional shift occur in image 412 due to the aberration shown in (Equation 7). Specifically, because the image is blurred and no longer fits into a single pixel, the signal intensity per pixel decreases, and the peak position of image 412 is shifted by one pixel (5 μm) compared to image 402 without magnification error. Therefore, if image 412 from the detection unit 102 with magnification error and image 402 from the detection unit 102 without magnification error are combined as is, the sensitivity will decrease.

[0054] To reduce the impact of such magnification errors on sensitivity, the sensor data output from the photoelectric conversion unit 103 in the sensor pixel array direction v1 (a direction conjugate to the scanning direction of the sample) is filtered using filters corresponding to image blurring and positional shift due to aberrations, thereby suppressing the effects of magnification errors. The filtered sensor data from each detection unit 102 are then added together. Specifically, the filtering process involves taking the cross-correlation between the sensor data a(n) and the filter b(n), as shown in (Equation 8).

[0055] Here, n represents the pixel position aligned in the direction of vector v1 on the light-receiving surface of the photoelectric conversion unit 103, and m represents the amount of pixel position shift. Furthermore, a(n) represents the sensor data at pixel position n, and b(n) represents the filter coefficient at pixel position n.

[0056] The lower panels of Figures 17 and 18 show filters 403 and 413 when the defect position in the S2 direction x = 0 μm, and filters 404 and 414 when the defect position in the S2 direction x = 600 μm. The filter shapes are examples only. Since the image shape changes and positional shifts caused by the optical system state differ for each pixel position on the light-receiving surface, a different filter is applied to each pixel position on the light-receiving surface.

[0057] If there is a magnification error, as shown in Figure 18, the image at x = 600 μm is blurrier than the image at x = 0 μm, even though the scattered light images are similar. The method for creating the filter will be described later, but the filter coefficient is determined based on the shape obtained by actually measuring the image when scattered light from a point on the sample is imaged on the light-receiving surface of the photoelectric conversion unit 103, or by simulation. For example, in the optical system state of Figure 18, a filter applied to the image at x = 600 μm is determined to reflect the shape and position of the image 412 at x = 600 μm. As a result, the filter 414 applied to the image 412 at x = 600 μm has a wider width (width in the v1 direction) than the filter 413 applied to the image 411 at x = 0 μm, and the peak of the filter 414 is at 605 μm, in line with the peak position of the image 412 at x = 600 μm being at 605 μm. When the cross-correlation between the image 412 and the filter 414 is taken according to (Equation 8), a waveform with a strong peak at x = 600 μm is obtained. By applying such a filter, blurred images can be restored, image positional shifts can be corrected, and as a result, a decrease in sensitivity when the signals between the detection units 102 are integrated in the signal integration unit 302 can be suppressed. This applies not only to signal integration in multiple detection units 102, but also to cases where signals detected at different pixel positions of the photoelectric conversion unit 103 are integrated in a single detection unit 102.

[0058] The filters in this embodiment have an asymmetrical shape to reflect the changes in image shape and positional displacement caused by the optical system state. Furthermore, even for images 401 and 411 at a field center x = 0 μm and image 402 at a field center x = 600 μm, where no changes in image shape or positional displacement occur due to magnification errors, filters 403, 413, and 404, which have an asymmetrical shape, are applied to suppress peak intensity. This is because, as will be described later, changes in image shape occur when there is a height fluctuation of the sample surface. Therefore, even when there is no height displacement, filtering is performed to blur the signal and suppress the highest sensitivity, thereby suppressing sensitivity variation.

[0059] Optical system conditions that cause shape changes or positional shifts in the image are not limited to magnification errors. For example, eccentricity (shift or tilt) of sensors and optical elements may occur due to environmental fluctuations or adjustment errors. Figure 19 shows images 421, 422 and filters 423, 424 on the light-receiving surface of the photoelectric conversion unit 103 when the magnification error is 1% and the angle β between the optical axis 121 and the pixel arrangement direction of the photoelectric conversion unit 103 changes to 56.3 degrees. Image 421 is the image when the defect position x in the S2 direction is 0 μm, and image 422 is the image when the defect position x in the S2 direction is 600 μm. When comparing image 422 with image 412 (see Figure 18) due to the change in the tilt angle of the photoelectric conversion unit 103, the blur width of the image has decreased, while the peak position of the image has shifted from 605 μm to 610 μm. The filter 424 applied to the image 422 is defined to reflect the shape and position of the image 422. Therefore, to reflect changes in the shape and positional shift of the image caused by environmental fluctuations and adjustment errors, the width of the filter coefficients of filter 424 is smaller than that of filter 414, and the peak of the filter coefficients also shifts in accordance with changes in the peak position of the image.

[0060] In addition to the magnification errors and sensor and optical element eccentricity illustrated in Figures 17 and 18, another optical system condition that causes shape changes and positional shifts in the image is the height displacement of the sample surface (Z-axis displacement) of several micrometers to several hundred micrometers, occurring at frequencies of several tens to several hundred Hz. Causes of height displacement include misalignment of the rotation axis and vibration of the stage. This height displacement of the sample surface causes the position of the image formed on the light-receiving surface of the photoelectric conversion unit 103 to shift. Hereinafter, the height displacement of the sample surface is characterized as an error factor that fluctuates during the inspection period of the sample.

[0061] Figures 20-22 will be used to explain the height displacement of the sample surface and the positional shift of the image formation. Figures 21-22 show the image formed when scattered light from a point on the sample (corresponding to the defect location) is projected onto the light-receiving surface of the photoelectric conversion unit 103, obtained by simulation assuming that the inspection optical system is composed of an ideal lens satisfying the sinusoidal condition. The optical system state is the same as the optical system state in Figure 18, but with the added height displacement of the sample surface. Figure 20 shows a state where the height of the sample surface has been displaced, and the position of the defect to be observed has been displaced from a height of 1041d (z = 0 μm) to a height of 1041e (z = -4 μm).

[0062] The upper panel of Figure 21 shows images 431 at x = 0 μm, 432 at x = 150 μm, and 433 at x = -150 μm, when the height of the sample surface is displaced and the position of the observed defect is displaced to z = 4 μm. Due to the height displacement of the sample surface, image blurring occurs, the image no longer fits into a single pixel, and the signal intensity per pixel decreases. In addition, a shift in the peak position of the image also occurs.

[0063] On the other hand, when the height of the sample surface is displaced and the position of the observed defect is displaced to z = -4 μm, the images 441 at x = 0 μm, 442 at x = 150 μm, and 443 at x = -150 μm are shown in the upper panel of Figure 22. Comparing the upper panel of Figure 21 with the upper panel of Figure 22, it can be seen that the scattered light images from the defect at the same position are significantly different. In other words, if these images are simply added together, a decrease in sensitivity will occur.

[0064] To reduce the effects of height displacement of the sample surface, filtering is performed to correlate the sensor data output from the photoelectric conversion unit 103 with filters corresponding to image blurring and positional shift due to aberrations. The lower part of Figures 21 and 22 shows examples of filters 434 and 444 when x = 0 μm, filters 435 and 445 when x = 150 μm, and filters 436 and 446 when x = -150 μm.

[0065] The filter shape is asymmetrical, reflecting the shape of the image. If there is a magnification error, the image shape changes and the image position shifts depending on the location of the defect, so different filters are used depending on the pixel position in the light receiving section of the photoelectric conversion unit 103. Also, in the cases of z = 4 μm (Figure 21) and z = -4 μm (Figure 22), the image shape is shifted in opposite directions relative to the sensor pixel array direction, and the image width is also different. Therefore, when a height displacement of the sample surface occurs, a filter that reflects each image shape is applied. Typically, the larger the height displacement of the sample surface, the larger the width of the filter coefficient and the greater the shift in the centroid position. Also, the further the defect location (position on the sensor surface) is from the center of the field of view, the larger the width of the filter coefficient and the greater the shift in the centroid position.

[0066] The image shape changes according to the height displacement of the sample surface, and the height displacement fluctuates during the inspection period of the defect inspection device. Therefore, the filter that suppresses the effect of the height displacement of the sample surface monitors the height of the sample surface during the inspection period and sets the appropriate filter in real time according to the height displacement. The height displacement of the sample surface is the value measured by the height measurement unit 106.

[0067] Figure 23A is a detailed functional block diagram of the signal integration unit 302 (see Figure 12B) of the signal processing unit 105. The control unit 53 includes a filter setting unit 501 and a filter database 502. The filter setting unit 501 can be configured as a functional unit realized, for example, by a processor executing processing according to a program. The filter setting unit 501 searches the filter database 502 and stores the filters used by the defect inspection device for signal integration in the filter memory 312.

[0068] Figure 23B shows an example of the data structure of the filter database 502. For example, the filter database 502 has a filter set set for each detection unit 102 and for each optical system state. A filter set is a set of filters that are set according to the height displacement of the sample surface and the pixel position of the light-receiving surface of the photoelectric conversion unit 103. Here, each is simplified to three levels, but it is not limited to this. The filter setting unit 501 selects one of the filter sets stored in the filter database 502 according to the optical system state of each detection unit 102 that performs inspection of the sample W, and stores it in the filter memory 312 of the signal processing unit 105 as a filter set to be used for signal integration from each detection unit 102. The optical system state of the detection unit 102 may be specified by the input unit 55, or the control unit 53 may estimate the optical system state of each detection unit 102 by the optical system state estimation unit described later. For example, if the optical system state of detection unit 102-1 is optical system state B and the optical system state of detection unit 102-2 is optical system state A, then filter sets 602 and 603 are stored in the filter memory 312. Furthermore, the optical system state described here is not particularly limited to any optical system state that causes shape changes or positional shifts in the image, as explained in Figures 17-19, but which can be considered not to change during the inspection period.

[0069] The signal integration unit 302 includes a filter selection unit 511, a filter processing unit 512, and a signal addition unit 513. The height measurement unit 106 (e.g., an optical sensor) detects the height displacement of the sample W in real time and inputs the detected height displacement to the filter selection unit 511. The filter selection unit 511 selects a filter to apply to each detection unit 102 from the filter memory 312 according to the height displacement of the sample W.

[0070] Here, the first sensor data 520-1 stored in the data memory 311 is the sensor data acquired by the detection unit 102-1, and the second sensor data 520-2 is the sensor data acquired by the detection unit 102-2. Furthermore, the first sensor data 520-1 is filtered by filter processing units 512-1-1 to 512-3, and the second sensor data 520-2 is filtered by filter processing units 512-2-1 to 512-2-3. The three filter processing units that perform filtering on the first sensor data 520-1 or the second sensor data 520-2 are each separated according to the pixel position on the light-receiving surface of the photoelectric conversion unit 103. For example, filter processing unit 512-1(2)-1 receives sensor data from pixels whose pixel position is included in the p1 region, filter processing unit 512-1(2)-2 receives sensor data from pixels whose pixel position is included in the p2 region, and filter processing unit 512-1(2)-3 receives sensor data from pixels whose pixel position is included in the p3 region.

[0071] On the other hand, if the height displacement input to the filter selection unit 511 is +a1, then filters 1B1 to 1B3 are input from the filter set 602 to the filter processing units 512-1-1 to 512-3, and filtering is performed. Similarly, filters 2A1 to 2A3 are input from the filter set 603 to the filter processing units 512-2-1 to 512-2-3, and filtering is performed.

[0072] The signal summing unit 513-1 adds the filtered sensor data output from the filter processing units 512-1-1 and 512-2-1, the signal summing unit 513-2 adds the filtered data output from the filter processing units 512-1-2 and 512-2-2, and the signal summing unit 513-3 adds the filtered data output from the filter processing units 512-1-3 and 512-2-3. This prevents image blurring due to height displacement of the sample surface and sensitivity reduction due to image position shift. The signal integration unit 302 can operate in real time at frequencies of several tens to several hundreds of Hz during inspection, preventing sensitivity reduction due to Z-axis displacement that occurs during inspection.

[0073] Here, we have described an example of integrating sensor data from multiple detection units 102. However, even when adding sensor data from the same detection unit 102 where the θ coordinate is the same but the r coordinate is different in the rθ coordinate system, a decrease in sensitivity due to Z-axis displacement occurring during inspection can be prevented by integrating the sensor data after filtering in the same way.

[0074] Environmental fluctuations such as temperature, or vibrations and shocks during equipment transport, can cause changes in magnification errors and eccentricity of optical elements. In such cases, the image shape on the light-receiving surface of the photoelectric conversion unit 103 changes, and therefore the optimal filter coefficient also changes. To prepare for such cases, the system is designed to create appropriate filters according to the state of the optical system. Figure 24 is a functional block diagram of the control unit 53 that creates filters. The control unit 53 includes an image measurement unit 701, an optical system state estimation unit 702, a filter creation unit 703, and a simulation database 704. The image measurement unit 701, the optical system state estimation unit 702, and the filter creation unit 703 can be configured as functional units realized, for example, by a processor executing processing according to a program.

[0075] The image measurement unit 701 measures the image of scattered light via the input unit 55. While an example using the photoelectric conversion unit 103 is shown, an area camera 1032 (see Figure 15) positioned conjugate to the sample surface may also be used. The optical system state estimation unit 702 compares the measured image with image simulation data stored in the simulation database 704. The simulation database 704 stores simulation data for various cases where the optical system state of the detection unit 102 is changed. By extracting the simulation data that most closely approximates the measured image, the optical system state can be estimated.

[0076] Figures 25A and 25B show examples of simulation data stored in the simulation database 704. Similar to Figure 17, Figures 25A and 25B show images obtained by simulation of scattered light from a point on the sample (corresponding to the defect location) focused onto the light-receiving surface of the photoelectric conversion unit 103, assuming that the inspection optical system is composed of an ideal lens satisfying the sinusoidal condition. Figure 25A shows images 711 when scattered light is incident on the sensor pixel position at 0 μm, images 712 when scattered light is incident on the sensor pixel position at 150 μm, and images 713 when scattered light is incident on the sensor pixel position at -150 μm, under optical system conditions with a magnification error of 1%, no displacement in the height direction of the sample surface, and no optical axis direction shift of the light-receiving surface of the photoelectric conversion unit 103. Figure 25B shows images 721 when scattered light is incident on a sensor pixel position with 0 μm, images 722 when scattered light is incident on a sensor pixel position with 150 μm, and images 723 when scattered light is incident on a sensor pixel position with -150 μm, under optical system conditions of a magnification error of 1%, no displacement in the height direction of the sample surface, and an optical axis direction shift of 5 μm of the light receiving surface of the photoelectric conversion unit 103.

[0077] In this way, by using simulation data of different S2 direction defect locations under different optical system conditions for comparison, the optical system condition can be estimated with greater accuracy. Furthermore, the optical system may be readjusted as needed based on the estimation results of the optical system condition. The filter creation unit 703 creates a filter set based on the shape of the simulation data image that best approximates the measured shape, using the height displacement of the sample surface and the shape of the simulation data image for each pixel position as filters, and stores it in the filter database 502. By using the newly created filters as described above, an appropriate filter can be applied even when the optical system condition changes, thereby suppressing a decrease in inspection sensitivity.

[0078] Figure 26 shows an example of a detection unit 102 (vertical detection system) positioned relative to the normal direction of the sample surface. Scattered light generated from the illumination spot 20 is focused by the objective lens 1021, and the polarization direction is controlled by the polarization control filter 1022. The imaging lens 1023 forms an image of the illumination spot 20 at the position of the aperture 1024. The aperture 1024 is set to allow only the light in the region to be detected by the photoelectric conversion unit 103 to pass through the image formed by the illumination spot 20. The focusing lens 1025 focuses the image of the aperture 1024 again. The polarization beam splitter 1026 separates the light whose polarization direction has been converted by the polarization control filter 1022 according to its polarization direction. The imaging lens 1027 forms an image of the illumination spot 20 on the photoelectric conversion unit 103. The light-receiving surface of the photoelectric conversion unit 103 is positioned perpendicular to the optical axis. If it is desired to separate and detect a portion of the scattered light as reflected light and a portion as transmitted light, an aperture 1070 is placed. The aperture 1070 has a shape that transmits a portion of the light and reflects a portion. The aperture 1070 is placed on the Fourier transform plane of the objective lens 1021. The scattered light transmitted by the aperture 1070 is focused to the photoelectric conversion unit 103 by the imaging lens 1028. The scattered light reflected by the aperture 1070 is focused to the photoelectric conversion unit 103' by the imaging lens 1028'. By utilizing the fact that the directional distribution of defect scattered light and wafer surface scattered light are different, and by using an aperture 1070 of an appropriate shape, defect scattered light and wafer surface scattered light can be separated.

[0079] Figures 27A and 27B illustrate the changes in image shape and positional shift that occur when a height displacement (z = 4 μm) occurs on the sample surface. The images were calculated using an ideal lens. Figure 27A shows the case with no magnification error, with image 801 when the S2 defect position x = 0 μm and image 802 when the S2 defect position x = 600 μm. Figure 27B shows the case with a magnification error of 1%, with image 811 when the S2 defect position x = 0 μm and image 812 when the S2 defect position x = 600 μm. The peak position of image 812 with a magnification error of 1% is 605 μm, and compared to image 802 with no magnification error, it can be seen that the image has shifted.

[0080] Therefore, in the rθ coordinate system, when adding sensor data with the same θ coordinate but different r coordinates, or when integrating images from the vertical detection system and the oblique detection system, a decrease in sensitivity occurs. For this reason, the detection unit 102 of the vertical detection system also stores a filter set in the filter database 502 for each optical system state. The filter set consists of multiple filters configured according to the height displacement of the sample surface and the light-receiving surface position of the photoelectric conversion unit 103. The filter selection unit 511 selects a filter, and the filter processing unit 512 applies it to the sensor data. The details of the processing are the same as those described in Figure 23A. This allows a filter with a shifted peak position of the filter coefficient to be applied to the sensor data, suppressing the effect of image misalignment. The amount of shift in the peak position of the filter coefficient varies depending on the light-receiving surface position. Furthermore, if the degree of image blurring changes depending on the defect location due to the effects of aberrations such as distortion, the width of the shift changes depending on the image position on the sensor surface. Since it becomes more difficult to correct aberrations towards the edges of the field of view, it is desirable to increase the width of the filter coefficient as you move away from the center of the field of view.

[0081] Depending on the distribution of the defect scattered light and wafer surface scattered light to be separated, it is effective to make the aperture 1070 asymmetric with respect to the optical axis. When the aperture 1070 has an asymmetric shape, the image shape becomes asymmetrically blurred and the imaging position shifts due to the height displacement of the sample surface. Therefore, by using a filter that estimates the signal from the shape of the aperture 1070, it becomes possible to correct the image blur and image position shift caused by the height displacement of the sample surface and integrate the signals. Also, when the aperture 1070 is positioned diagonally with respect to the optical axis, even if the center of the aperture 1070 is positioned on the Fourier transform plane of the objective lens 1021, the edges of the aperture 1070 will be offset from the Fourier transform plane. Therefore, the image shape changes depending on whether the defect is on the optical axis or off-axis. In addition, the image shape and image position shift also change depending on the image position due to the height displacement of the sample surface. By applying a filter that reflects the image shape according to the height displacement of the sample surface and the position on the sensor, it becomes possible to correct the image blur and image position shift and integrate the signals, thereby suppressing a decrease in sensitivity.

[0082] The present invention is not limited to the embodiments described above, and various modifications are included. For example, the embodiments described above are explained in detail to make the present invention easier to understand, and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to add, delete, or replace some of the configurations in the embodiments with other configurations.

[0083] 2: Laser light source, 3: Attenuator, 4: Emitted light adjustment unit, 5: Beam expander, 6: Polarization control unit, 7: Illumination intensity distribution control unit, 20: Illumination spot, 21: Mirror, 22: Beam monitor, 23: Beam monitor, 24: Illumination intensity distribution monitor, 27: Optical image, 53: Control unit, 54: Display unit, 55: Input unit, 101: Illumination unit, 102: Detection unit, 103: Photoelectric conversion unit, 104: Stage, 105: Signal processing unit, 106: Height measurement unit, 107: Mirror, 120: Illumination optical axis, 121: Optical axis, 301: Illumination spot position analysis unit, 302: Signal integration unit, 303: Defect detection unit, 311: Data memory, 312: Filter memory, 501: Filter setting unit, 502 : Filter database, 511: Filter selection unit, 512: Filter processing unit, 513: Signal addition unit, 520: Sensor data, 701: Image measurement unit, 702: Optical system state estimation unit, 703: Filter creation unit, 704: Simulation database, 1021: Objective lens, 1022: Polarization control filter, 1023: Imaging lens, 1024: Aperture, 1025: Focusing lens, 1026: Polarization beam splitter, 1027: Imaging lens, 1028: Imaging lens, 1031: Light receiving unit, 1032: Area camera, 1041: Height, 1070: Aperture, 10210: Cylindrical lens, 10211: Cylindrical lens, 10212: Half mirror.

Claims

A stage for loading and moving the sample, A lighting unit that illuminates the sample with a lighting spot, Multiple detection units for detecting scattered light from the sample irradiated by the aforementioned illumination spot, A height measuring unit for measuring the height displacement of the sample irradiated by the aforementioned lighting spot, It has a signal processing unit, The detection unit comprises a photoelectric conversion unit and a detection optical system that images the scattered light onto the light-receiving surface of the photoelectric conversion unit. The photoelectric conversion unit has a light-receiving surface divided into multiple pixels, and converts the imaged scattered light into an electrical signal for each pixel and outputs it as sensor data. The signal processing unit includes a signal integration unit that integrates the sensor data from the photoelectric conversion units of multiple detection units, each of which scattered light from the same position on the sample is imaged. The signal integration unit performs a filtering process on each of the sensor data from the plurality of detection units, taking into account the cross-correlation with a filter set for each detection unit and having filter coefficients defined in the direction of the pixel arrangement, and adds the filtered sensor data from the plurality of detection units together. The filter is a defect inspection device selected according to the amount of height displacement of the sample from the height measuring unit and the pixel position of the light-receiving surface where the sensor data is detected. 　 In claim 1, The signal processing unit maintains a filter set that performs the filtering process on each of the sensor data from the plurality of detection units. The filter set is a defect inspection device that includes a plurality of filters set according to the height displacement of the sample and the pixel position of the light-receiving surface where the sensor data is detected. 　 In claim 2, The signal processing unit is a defect inspection device that holds the filter set according to the optical system state of each of the multiple detection units. 　 In claim 1, The detection unit is positioned obliquely to the sample, A defect inspection device wherein the light-receiving surface of the photoelectric conversion unit is conjugate to the illumination spot that irradiates the sample, and the normal direction of the light-receiving surface is inclined with respect to the optical axis of the detection optical system. 　 In claim 1, The aforementioned filter has an asymmetrical shape in this defect inspection device. 　 In claim 2, A defect inspection device in which the filter included in the filter set is a filter whose filter coefficient width and centroid position differ depending on the pixel position of the light-receiving surface. 　 In claim 6, A defect inspection device in which the filter included in the filter set is a filter whose filter coefficient width increases as the pixel position on the light-receiving surface moves away from the center. 　 In claim 6, A defect inspection device in which the filter included in the filter set is a filter in which the shift of the centroid position of the filter coefficient increases as the pixel position of the light-receiving surface moves away from the center. 　 In claim 1, A defect inspection device in which the filter coefficient of the filter is set based on the shape obtained by actually measuring the image when scattered light from a point on the sample is imaged onto the light-receiving surface of the photoelectric conversion unit, or by simulation. 　 In claim 1, By changing the optical system state of the detection unit through simulation, the image obtained when scattered light from a point on the sample is imaged onto the light-receiving surface of the photoelectric conversion unit is pre-determined as simulation data. The optical system state of the detection unit is estimated by comparing the measured shape obtained by actually measuring the image formed when scattered light from a point on the sample is imaged onto the light-receiving surface of the photoelectric conversion unit with the simulation data. A defect inspection device in which the filter coefficients of the filter, according to the estimated optical system state, are set based on the shape of the image of the simulation data that best approximates the measured shape. 　 In claim 1, The detection unit is arranged perpendicular to the sample, A defect inspection device in which the light-receiving surface of the photoelectric conversion unit is conjugate to the illumination spot that irradiates the sample, and the normal direction of the light-receiving surface is positioned perpendicular to the optical axis of the detection optical system. 　 A method for processing sensor data in a defect inspection device, The defect inspection apparatus comprises a stage on which a sample is mounted and moved, an illumination unit that irradiates the sample with an illumination spot, a plurality of detection units that detect scattered light from the sample irradiated by the illumination spot, a height measuring unit that measures the amount of height displacement of the sample irradiated by the illumination spot, and a signal processing unit. The detection unit comprises a photoelectric conversion unit and a detection optical system that images the scattered light onto the light-receiving surface of the photoelectric conversion unit. The photoelectric conversion unit has a light-receiving surface divided into multiple pixels, and converts the imaged scattered light into an electrical signal for each pixel and outputs it as sensor data. The signal processing unit includes a signal integration unit that integrates the sensor data from the photoelectric conversion units of multiple detection units, each of which scattered light from the same position on the sample is imaged. The signal integration unit performs a filtering process on each of the sensor data from the plurality of detection units, taking into account the cross-correlation with a filter set for each detection unit and having filter coefficients defined in the direction of the pixel arrangement, and adds the filtered sensor data from the plurality of detection units together. The filter is a sensor data processing method selected according to the amount of height displacement of the sample from the height measuring unit and the pixel position of the light-receiving surface where the sensor data is detected. 　 In claim 12, The signal processing unit maintains a filter set that performs the filtering process on each of the sensor data from the plurality of detection units. The filter set includes a plurality of filters set according to the height displacement of the sample and the pixel position of the light-receiving surface where the sensor data is detected. 　 In claim 13, The signal processing unit is a sensor data processing method that holds the filter set according to the optical system state of each of the multiple detection units. 　 In claim 12, The detection unit is positioned obliquely to the sample, A sensor data processing method wherein the light-receiving surface of the photoelectric conversion unit is arranged conjugate to the illumination spot that irradiates the sample, and the normal direction of the light-receiving surface is inclined with respect to the optical axis of the detection optical system.

Images