Semiconductor wafer inspection using dark-field epi-illumination microscopy

Abstract

The application relates to the technical field of semiconductor detection, and discloses a drop illumination dark field illumination microscopic imaging system and method for semiconductor wafer detection, which comprises a stage, a bright field light source module, a dark field light source module, 12 collimating LED sub-sources symmetrically arranged along a hollow annular layout, 12 independent illumination paths which do not interfere with each other and have clear directions are constructed, a hollow cone lens which adopts a center hollow structure, the upper surface of the cone lens is a plane, and the lower surface is a cone surface, an illumination control module which realizes at least three dark field illumination modes which can be flexibly switched, a pulse driving module, and an automatic focusing module and an XY displacement table which are used for completing automatic focusing and field switching. The application can realize high-contrast, high-signal-to-noise-ratio and high-efficiency dark field microscopic imaging detection of different types, sizes and orientation defects on the surface of a semiconductor wafer.

Description

Technical Field

[0001] This invention relates to the field of semiconductor inspection technology, and specifically to an epi-type dark-field illumination micro-imaging system and method for semiconductor wafer inspection. Background Technology

[0002] Dark-field detection is a high-precision microscopic inspection technique based on the principle of scattered light imaging. Through a special illumination and imaging optics system, it selectively acquires only the scattered light generated by defects on the sample surface, while maximally shielding reflected light from the sample surface, achieving an imaging effect with a dark background and bright sample details. This high-contrast imaging characteristic gives it a natural advantage in scenarios requiring precise identification of minute anomalies.

[0003] In the semiconductor manufacturing field, defects such as tiny particles, minor scratches, and thin film damage on the wafer surface directly affect the yield and reliability of chips. Dark field detection technology, with its high sensitivity to capture scattered light signals, has become one of the core technologies for rapid screening and high-precision identification of such defects, providing key support for quality control in the semiconductor manufacturing process.

[0004] However, wafer inspection methods based on dark-field illumination still have significant shortcomings in practical applications: First, in commercial microscopic imaging systems, dark-field imaging is usually achieved using a transmission-type dark-field illumination method. The illumination module and the probe optical path are distributed on both sides of the sample. The illumination beam penetrates the sample from behind and enters the probe optical path to complete the imaging. This method can be used for dark-field imaging of transparent samples, but it is difficult to achieve dark-field imaging of reflective samples such as semiconductor wafers.

[0005] Secondly, the ring-shaped LED array is a common solution for reflective dark-field illumination of samples, projecting small-angle oblique incident light onto the sample surface through the ring-shaped LEDs. However, this small-angle incident illumination method does not achieve beam shaping, has extremely low energy utilization, and makes it difficult to obtain a sufficiently strong scattered light signal, directly restricting the improvement of detection efficiency and accuracy.

[0006] Third, another common approach to reflective dark-field imaging is unidirectional illumination, which uses a tilted beam projected in a single direction. The core characteristic of unidirectional dark-field illumination is that the incident angle and direction of the illumination source are fixed. However, wafer defects, due to their physical morphology, size, orientation, and the wavelength and polarization state of the illumination light, exhibit significant anisotropy in the spatial distribution of scattered light signal intensity. For example, micro-scratches extending along a specific direction have their scattered light intensity peaks concentrated perpendicular to the scratch's extension direction, while scratches distributed along the illumination direction may only produce weak scattering. For sub-micron diameter particles, the peak direction of scattered light intensity is strongly correlated with particle diameter, refractive index, and incident light wavelength. This contradiction between the fixed illumination direction and the anisotropy of the scattered signal leads to extremely low scattered signal capture efficiency and strong directionality in unidirectional dark-field illumination for wafer defect detection. Furthermore, the intensity of scattered signals captured by unidirectional illumination varies drastically for different types and orientations of defects, causing severe imaging distortion and significantly increasing measurement errors for key parameters such as defect size and morphology.

[0007] Fourth, in dark-field imaging, the intensity of the scattered signal from the observed features on the sample is directly proportional to the intensity of the illumination light. Unidirectional dark-field illumination intensity is limited by the inherent properties of the light source, making it difficult to provide high-intensity illumination. This results in weak scattered light signals from defects, leading to a low signal-to-noise ratio in the acquired raw image. This fails to meet the clear imaging requirements for nanoscale particles and low-scattering defect features, reducing the accuracy and reliability of wafer defect detection. Furthermore, while unidirectional dark-field illumination can increase instantaneous light intensity by increasing the driving current of the light source, it also causes irreversible thermal damage to the light source, reducing its stability and lifespan. Summary of the Invention

[0008] The present invention aims to provide an epi-type dark-field illumination micro-imaging system and method for semiconductor wafer inspection, in order to solve the technical problems that existing transmissive dark-field illumination cannot be applied to reflective samples, the unidirectional illumination has low efficiency and strong directionality in capturing scattered signals, and the traditional illumination has insufficient light intensity, resulting in a low signal-to-noise ratio.

[0009] To solve the above problems, the present invention adopts the following technical solution: Option 1: An epi-irradiation dark-field illumination micro-imaging system for semiconductor wafer inspection, comprising: A stage is used to hold semiconductor wafer samples to be tested. Bright field light source module, used to provide bright field illumination beam; The dark field light source module includes 12 collimated LED sub-light sources symmetrically arranged along a hollow ring layout. Each sub-light source includes a fixed wavelength high-stability LED lamp bead or a laser light source that has undergone decoherence processing, a beam expander lens, and a collimating lens arranged sequentially along the optical axis, forming an integrated beam shaping module. The 12 collimated sub-light sources construct 12 independent lighting paths that do not interfere with each other and have clear directions. The hollow conical lens has a central hollow structure. The upper surface of the conical lens is a plane and the lower surface is a cone. The 12 collimating sub-light sources are incident perpendicular to the upper surface of the conical lens. When the light beam passes through the second surface of the conical lens, it is refracted. The angle between the refracted light beam and the vertical direction is determined by the physical base angle α of the conical lens, the refractive index n of the lens, and the incident wavelength λ. The spot convergence point of the annular light beam coincides with the focal plane of the microscope objective. The central hollow aperture of the hollow conical lens is customized according to the outer diameter of the microscope objective. The microscope objective is coaxially mounted in the hollow region. The microscope objective is used to collect the scattered light signal from the sample surface. After being collected by the microscope objective, the scattered light signal passes sequentially through the first beam splitter, the second beam splitter, and the infinity imaging tube before being imaged onto the camera target surface. The lighting control module is used to independently switch and control the power of the 12 ring-shaped distributed sub-light sources, and realize at least three flexibly switchable dark field lighting modes, including ring-shaped focused light spot mode, unidirectional asymmetric lighting mode, and at least one multidirectional symmetric or asymmetric lighting mode. The pulse drive module generates a high-precision pulsed current signal using a rectangular wave pulse drive scheme, independently adjusting the brightness of each sub-light source. When the rectangular wave pulsed current reaches its peak value, it triggers the camera to start exposure, ensuring that the exposure time completely coincides with the effective emission period of the pulsed flash; and The autofocus module and XY shift stage are used to complete autofocus and field of view switching.

[0010] Beneficial effects: The hollow conical lens-based incident focusing illumination design achieves coaxial integration of the illumination and imaging optical paths, solving the problems of traditional transmission-type dark fields being unsuitable for reflective samples and the structural complexity caused by the separate layout of side illumination and objective lenses; the conical lens-based refraction focusing forms an isotropic high-brightness annular spot, significantly improving the intensity of the scattered signal and the uniformity of detection; at least three switchable illumination modes adapt to the anisotropic scattering characteristics of different types of defects; and pulse driving synchronized with camera exposure achieves instantaneous high-brightness illumination and precise timing control, greatly improving the image signal-to-noise ratio.

[0011] This invention effectively solves the problems in the prior art and achieves high-contrast, high-signal-to-noise ratio, and high-efficiency dark-field microscopic imaging detection of different types, sizes, and orientation defects on the surface of semiconductor wafers.

[0012] Preferably, the beam expander is an aspherical lens, the collimating lens is a plano-convex lens, and the sub-light source module uses a lens with a diameter d ≤ 10 mm, ensuring that the diameter of the assembled sub-light source is ≤ 12 mm.

[0013] Beneficial effects: By using a two-stage synergistic shaping process involving an aspherical beam expander and a plano-convex collimating lens, the defects of excessive beam divergence angle and uneven energy distribution of traditional light sources are effectively avoided, thereby improving beam utilization and reducing beam size and divergence angle.

[0014] Preferably, the center wavelength of the fixed-wavelength high-stability LED bead is 550nm and the half-width at half-maximum (WHM) is <20nm, or a laser source that has undergone decoherence processing is selected, wherein the WHM of the laser source is ≤1nm.

[0015] Beneficial effects: By using a dual-selection scheme of narrowband LED and decoherent laser, the incident light spectrum can be customized according to the defect type and the optical absorption and scattering characteristics of the wafer material, thereby maximizing the optical response intensity of the defect signal.

[0016] Preferably, the conical lens is made of PMMA material, the central hollow aperture of the conical lens is φ20mm~φ40mm, and the angle range of the emitted light beam from the conical lens incident on the wafer surface is 30°~60°.

[0017] Beneficial effects: By using a conical lens design made of PMMA material and reserving space for objective lens assembly in a hollow structure, an integrated coaxial design of lens and objective lens is achieved, avoiding the use of additional reflectors, directional mirrors and other components; at the same time, by precisely controlling the incident angle within the range required for dark field detection, the pure dark field illumination effect is ensured.

[0018] Preferably, the dark field lighting mode implemented by the lighting control module includes: Ring-shaped focused spot mode: All 12 sub-light sources are activated, all sub-light sources have the same power, and the illumination angle covers 360°; Unidirectional asymmetric lighting mode: Only the sub-light source in a single direction is turned on, while the rest of the sub-light sources are turned off; Two-way non-central symmetrical lighting mode: Two sets of sub-light sources with an angle of non-180° are turned on in the ring distribution, and the power of the two sets of light sources can be adjusted independently; Three-way non-centrally symmetrical lighting mode: Activate three sub-light sources with non-uniformly distributed angles, and adjust their power independently; Four-way non-central symmetrical lighting mode: Activate 4 sub-light sources with non-uniform angles in the ring distribution, and adjust the power of the four sub-light sources independently; Two-way centrally symmetrical lighting mode: Turn on two sub-light sources that are symmetrical along the annular diameter and have the same power; Three-way centrally symmetrical lighting mode: Turn on 3 sets of sub-light sources that are evenly distributed along the ring, with the same power; Four-way centrally symmetrical lighting mode: Four sets of sub-light sources, uniformly symmetrical along a ring, are activated with identical power; and Six-way centrally symmetrical lighting mode: Turn on 6 sets of sub-light sources that are uniformly symmetrical along the ring, with the same power.

[0019] Beneficial effects: Through the refined design of nine lighting modes, the dark field lighting solution system is made more complete. The appropriate lighting solution can be flexibly selected according to the different needs of defect type, detection accuracy and detection efficiency, so as to achieve the core goal of full coverage, high precision and high adaptability of semiconductor wafer defect detection.

[0020] Preferably, in the pulse driving module, when using an LED light source, within the rated peak current range, the luminous brightness of the LED is proportional to the peak current passing through it. The rapid response characteristics of the LED are utilized to inject a large instantaneous current to generate instantaneous high-brightness illumination. When using a laser light source, when the driving current of the laser is greater than the threshold current, the output optical power has a linear relationship with the driving current. By precisely controlling the peak value and pulse width of the pulse current, a short pulse width and high peak power pulse output can be obtained.

[0021] Beneficial effects: By replacing the traditional constant-on driving mode with a pulse driving scheme, a momentary ultra-large peak current is injected within a short pulse period, enabling the light source to generate instantaneous high-energy-density illumination, significantly improving the intensity of scattered light from defective surfaces; at the same time, short-pulse driving effectively avoids the thermal effect of the light source, balancing high brightness and illumination stability.

[0022] Preferably, the 12 collimated sub-light sources are uniformly distributed with adjacent sub-light sources having an angle of 30° between them.

[0023] Beneficial effects: By limiting the number and distribution angle of sub-light sources, the uniformity and isotropy of the ring illumination light field are ensured, so that defects in different directions on the wafer surface are illuminated by the same intensity of light, thus ensuring the consistency and accuracy of the test results.

[0024] Option 2: An epi-irradiation dark-field illumination micro-imaging method for semiconductor wafer inspection, using the system described above, includes the following steps: Step S1: Based on the spectral requirements of the specific detection scenario, select a fixed-wavelength high-stability LED lamp bead or a laser light source that has undergone decoherence processing as the core light-emitting device. At the output end of the light source, a beam expanding lens and a collimating lens are sequentially arranged along the optical axis to form a beam shaping module, thereby forming a sub-light source module with high parallelism and high uniformity. Step S2: Arrange the 12 collimated sub-light sources symmetrically in a hollow ring layout to construct 12 independent lighting paths that do not interfere with each other and have clear directions, forming a ring-shaped lighting light field distribution; S3: A hollow conical lens is used to focus the ring beam. The center of the conical lens has a hollow structure to reserve an assembly position for the microscope objective. The collimated sub-light source is incident perpendicular to the upper surface of the conical lens. When it passes through the second surface of the conical lens, it is refracted. The direction of the refracted beam is determined by the physical base angle α of the conical lens, the refractive index n of the lens, and the incident wavelength λ. The assembly position of the ring-shaped sub-light source module and the conical lens is adjusted so that the spot convergence point of the ring beam coincides with the focal plane of the test objective. The microscope objective is coaxially assembled in the hollow area. S4: Based on the type, symmetry characteristics, and number of directions of the defects to be detected, select the appropriate lighting scheme from multiple dark field lighting modes, and achieve lighting mode switching by independently switching and power controlling the 12 ring-shaped distributed sub-light sources; S5: A rectangular wave pulse driving scheme is used to generate a high-precision pulse current signal. When the rectangular wave pulse current reaches its peak value, the camera is triggered to start exposure, ensuring that the exposure time completely coincides with the effective emission period of the pulse flash. The camera simultaneously acquires the original dark field image and stores it in the local cache. S6: The stage moves to the next field of view according to the preset path, and after completing the autofocus, it repeats the above illumination and image acquisition process to build a closed-loop imaging system for the entire process until all fields of view are scanned.

[0025] Beneficial effects: This invention achieves a complete detection process from light source shaping, ring layout, conical lens focusing, mode selection, pulse driving to closed-loop scanning through the organic integration of method steps. The timing of each step is precisely coordinated, providing high-quality raw data for subsequent defect detection algorithms.

[0026] Preferably, the lighting scheme selection process in step S4 includes: The first step is to define the detection objectives: for rapid batch screening scenarios, a ring lighting mode is selected; for high-precision quantitative analysis scenarios, a symmetrical six-directional lighting mode is selected; and for scenarios requiring precise determination and classification of defect direction and morphology, proceed to the second step. The second step is to determine the symmetry characteristics of the defect: For regular symmetrical defects, the corresponding symmetrical lighting mode is selected according to the number of symmetrical directions; for irregular and asymmetrical defects, proceed to the third step. The third step is to select the appropriate lighting mode based on the number of defects in each direction: for defects in a single direction, use a unidirectional lighting mode; for defects in two directions that are not centrally symmetrical, use a two-way non-centrally symmetrical lighting mode; for defects in three directions that are irregular, use an asymmetrical three-way lighting mode; and for defects in four directions that are irregular, use a four-way non-centrally symmetrical lighting mode.

[0027] Beneficial effects: The three-step lighting scheme selection process establishes a systematic decision-making mechanism from detection target to defect characteristics to lighting mode selection, giving the selection of lighting modes a clear logical basis and technical orientation.

[0028] Preferably, the method for calculating the mounting height h of the upper surface of the conical lens in step S3 is as follows:

[0029] Where r1 is the distance from the center of the sub-source to the optical axis of the microscope objective; r2 is the radius of the conical lens; and α is the physical base angle of the conical lens. The angle by which the refracted beam deviates from the vertical direction.

[0030] Beneficial effects: By establishing clear formulas for calculating the optical parameters of conical lenses, a quantitative basis is provided for the design, processing, assembly, and debugging of conical lenses, ensuring the precise controllability of the angle and convergence position of the refracted beam.

[0031] (a) Advantages of the present invention This invention has the following significant advantages: First, the beam quality is precisely controllable, ensuring reliable detection. A two-stage collaborative shaping structure using a beam expander and a collimating lens enables precise control and optimization of the beam. This structure effectively avoids the common defects of traditional light sources, such as excessive beam divergence angle and uneven energy distribution, thereby solving core problems caused by these defects, such as image blurring, defect signal intensity fluctuations, and low signal-to-noise ratio. This lays a solid foundation for high-contrast differentiation between defects and background in dark-field imaging scenarios, significantly improving the accuracy and stability of defect identification.

[0032] Secondly, multi-directional collaborative illumination eliminates shadow occlusion. A symmetrical ring layout of 12 collimated sub-light sources achieves 360° omnidirectional beam coverage without blind spots. Compared to traditional unidirectional illumination schemes that easily produce shadow areas along the sample outline, this ring collaborative illumination design precisely fills shadow gaps through multi-directional complementary light fields, ensuring that all key areas of the sample surface receive uniform and sufficient illumination. This design fundamentally solves the problem of defect signal distortion caused by unidirectional illumination, effectively avoiding the risk of missed or false detections of minute defects, and providing high-precision, high-reliability data support for subsequent quantitative analysis of defect size, morphology, etc.

[0033] Third, spectral characteristics are matched to adapt to diverse detection scenarios. For the spectral response requirements of different detection scenarios such as silicon wafer scratch detection, silicon carbide particle detection, and lattice defect identification, a dual-selection solution of narrowband LED and decoherent laser is provided. The incident light with a center wavelength and narrow half-width can be customized according to the defect type and the optical absorption and scattering characteristics of the wafer material, maximizing the optical response intensity of the defect signal and significantly improving the detection rate of minute defects such as particles and scratches.

[0034] Fourth, achieving an isotropic dark-field spot. The conical lens uses a special conical structure to focus and control the annular collimated beam, converging the multi-directional annular beam onto the wafer inspection area using the principle of optical refraction. This forms a high-energy-density, uniformly focused spot, providing a high signal-to-noise ratio illumination basis for dark-field inspection and effectively enhancing the scattering signal of tiny defects. The circumferentially symmetrical structure of the conical lens ensures that the focused annular spot has perfect rotational symmetry, with the size, intensity, and divergence angle of the spot being completely consistent in all directions. This isotropic characteristic ensures that defects in different directions on the wafer surface receive the same intensity of illumination, avoiding differences in defect signals caused by deviations in the directionality of the spot, and guaranteeing the consistency and accuracy of the inspection results.

[0035] Fifth, the hollow structure design provides ample space for objective lens assembly, ensuring high optical path compatibility. The conical lens features a central hollow structure, allowing direct coaxial mounting of microscope objectives within the hollow area without altering the core layout of the original imaging optical path. It supports seamless adaptation to both low-magnification and high-magnification microscope objectives, eliminating the need to replace the main conical lens structure; simply adjusting the conical lens's mounting position achieves different resolution imaging requirements, reducing equipment upgrade and maintenance costs. This integrated coaxial design of the lens and objective lens avoids the use of additional reflectors, steering mirrors, and other components, reducing optical path loss and assembly errors, resulting in a more compact optical system structure and significantly improving the ease of optical path integration.

[0036] Sixth, the illumination angle is precisely controllable, meeting the core requirements of dark-field inspection. The direction of the emitted beam from the conical lens is determined by the cone angle α, the lens refractive index n, and the incident wavelength λ. By precisely designing these parameters, the angle of the beam that is finally incident on the wafer surface can be strictly controlled within the angle range required for dark-field inspection. This ensures that the beam is only scattered by defects before entering the objective lens for imaging, achieving a pure dark-field illumination effect, effectively suppressing background noise, and improving the contrast between defects and the background.

[0037] Seventh, the structure is simple and compact, with low processing difficulty and excellent consistency. The core component, the conical lens, adopts an integrated molding design, containing only two optical surfaces. Compared with traditional complex multi-lens focusing systems, the processing technology of this design is simpler and more efficient, and the processing accuracy of core parameters such as cone angle and hollow aperture is easier to ensure. This results in stronger product consistency during mass production and can effectively reduce production and application costs.

[0038] Eighth, multiple lighting modes are available, improving the dark field lighting system. Nine lighting modes are integrated, making the dark field lighting solution system more complete. Through comparative analysis of the characteristics of different lighting modes, suitable lighting solutions can be flexibly selected according to the differentiated needs of defect type, detection accuracy, and detection efficiency. This achieves the core goals of full coverage, high precision, and high adaptability in semiconductor wafer defect detection, significantly improving the comprehensive detection capabilities and scene adaptability of a single device.

[0039] Ninth, pulse-driven instantaneous high-brightness illumination is achieved. Injecting a short-pulse period with an instantaneous ultra-large peak current enables the light source to generate instantaneous high-energy-density illumination, significantly improving the intensity of scattered light from the defect surface. This effectively compensates for the small scattering cross-section, greatly enhances the grayscale contrast between the defect and the background, and provides sufficient signals for subsequent image algorithms to identify defects.

[0040] Tenth, precise timing control, balancing noise reduction and efficiency improvement. A control strategy that precisely synchronizes the pulsed light source and camera exposure time ensures that original images are acquired only during the illumination period, effectively suppressing ambient stray light interference and easily obtaining high-quality inspection images with clean backgrounds and high contrast. Simultaneously, data storage is completed during the intervals when the light source is off, and in conjunction with the XY stage, rapid switching of the imaging field of view is achieved, significantly improving the overall operating efficiency of the inspection system.

[0041] (II) Unexpected Aspects of the Technical Problem Solved by the Invention The unexpected aspect of this invention's solution to the technical problem lies in the fact that traditional methods typically address insufficient scattered signals by increasing the number of light sources or raising the power of a single light source. However, this invention, through a combination of a hollow ring-shaped multi-sub-light source, conical lens focusing, and nine illumination mode switching, achieves a multiplier effect on scattered signal intensity and a qualitative leap in directional adaptability without increasing the total power of the light sources. Specifically, traditional unidirectional illumination can only obtain effective scattered signals in a specific direction, while the ring layout combined with conical lens focusing in this invention allows multi-directional beams to converge into the same detection area. This not only achieves 360° omnidirectional signal coverage but also concentrates and enhances the energy density of the ring beam through the refraction focusing of the conical lens. This distributed input and concentrated output optical path design produces an unexpected improvement in light energy utilization efficiency. Furthermore, the independent control capability of the nine illumination modes allows the same hardware system to be adapted to completely different detection scenarios through software control. This flexibility of multi-purpose use is unprecedented in existing technologies.

[0042] (III) Unexpected Aspects of Technical Means The unexpected aspects of the technical means of this invention are mainly reflected in the following aspects: First, the hollow structure design of the conical lens. Traditional conical lenses used for beam shaping typically employ a solid structure. This invention, however, hollows out the center of the conical lens to accommodate the microscope objective. This seemingly simple structural change actually solves the core problem of ensuring that the illumination and imaging optical paths are coaxial yet do not interfere with each other in incident dark-field illumination. The hollow structure allows the illumination beam to be refracted and focused from the periphery of the conical lens before reaching the sample, while the scattered signal enters the objective directly from the central region. This achieves physical separation and coaxial unification of illumination and detection, something that traditional transmission-based dark-field and side-illumination methods cannot achieve.

[0043] Second, precise synchronization between pulse drive and camera exposure. Traditional pulsed illumination typically focuses only on increasing the brightness of the light source itself. This invention, however, strictly synchronizes the timing control of the pulse drive with the camera's exposure window, ensuring the camera only acquires images during the peak illumination period of the light source. This time-gating strategy produces two unexpected effects: firstly, because the light source is at its peak brightness during the exposure period, the image signal-to-noise ratio is significantly improved; secondly, the camera does not operate during the light source's off period, effectively suppressing ambient stray light and dark current noise. Simultaneously, this period is used for data storage and stage movement, achieving multi-task parallel processing in the time dimension.

[0044] Third, the nine lighting modes are systematically classified along two dimensions: symmetry or asymmetry and the number of directions. Existing lighting mode selection methods are typically based on experience or trial and error, while this invention establishes a systematic selection logic from the detection target to the defect's symmetry characteristics and then to the number of directions, transforming the selection of lighting modes from an art to a science. This systematic classification method is groundbreaking in the field.

[0045] Fourth, the incident focusing illumination of the hollow conical lens: Prior art document 1TW201132960A uses a cylindrical lens, and prior art document 2CN112098421A uses a uniform cavity and critical illumination, both of which form linear light spots; the present invention uses a hollow conical lens to refract and focus the annular collimated beam to form an isotropic annular light spot, and achieves incident (coaxial) illumination, which solves the problem that the transmission dark field cannot be used for reflecting samples.

[0046] Fifth, the lens and objective are integrated and coaxially assembled: In the prior art, the objective and illumination modules are separate; the conical lens of the present invention has a hollow center to reserve an assembly position for the objective, realizing the physical separation and coaxial unification of illumination and detection, which is not revealed in the prior art. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the system of the present invention.

[0048] Figure 2 This is a schematic diagram illustrating the working principle of the sub-light source module in the system of this invention.

[0049] Figure 3 This is a schematic diagram of the layout structure of the 12 collimated LED sub-light sources in the system of the present invention.

[0050] Figure 4 This is a diagram showing the optical path structure of the conical lens in the system of this invention.

[0051] Figure 5 This is a schematic diagram of the illumination modes formed by 12 ring-shaped sub-light sources in the system of the present invention.

[0052] Figure 6This is a schematic diagram of the illumination modes formed by 12 ring-shaped sub-light sources in the system of the present invention.

[0053] Figure 7 This is a schematic diagram of the illumination modes formed by 12 ring-shaped sub-light sources in the system of the present invention.

[0054] Figure 8 This is a schematic diagram of the illumination modes formed by 12 ring-shaped sub-light sources in the system of the present invention.

[0055] Figure 9 This is a schematic diagram of the illumination modes formed by 12 ring-shaped sub-light sources in the system of the present invention.

[0056] Figure 10 This is a schematic diagram of the illumination modes formed by 12 ring-shaped sub-light sources in the system of the present invention.

[0057] Figure 11 This is a schematic diagram of the illumination modes formed by 12 ring-shaped sub-light sources in the system of the present invention.

[0058] Figure 12 This is a schematic diagram of the illumination modes formed by 12 ring-shaped sub-light sources in the system of the present invention.

[0059] Figure 13 This is a pulse timing diagram of the system of the present invention.

[0060] Figure 14 This is a schematic diagram illustrating the working principle of the system of the present invention.

[0061] The reference numerals in the accompanying drawings include: Z-axis displacement stage 1, camera 2, extension sleeve 3, infinity imaging tube 4, autofocus module 5, second semi-transparent mirror 6, bright field light source module 7, first semi-transparent mirror 8, reflecting mirror 9, microscope objective 10, dark field light source module 11, conical lens 12, and sample stage 13. Detailed Implementation

[0062] The following detailed description illustrates the specific implementation method: As attached Figure 1 As shown: The semiconductor wafer inspection epi-irradiation dark-field illumination microscopic imaging system of the present invention includes, from top to bottom, a Z-axis displacement stage 1 (including a stepper and a lead screw), a camera 2, an extension sleeve 3, an infinity imaging tube 4, a second semi-transparent mirror 6 (a 50 / 50 beam splitter), a first semi-transparent mirror 8 (a 50 / 50 beam splitter), a microscope objective 10, a dark-field light source module 11, a conical lens 12, and a sample stage 13 (also called an XY stage). The second semi-transparent mirror 6 (i.e., the...) Figure 14 The semi-transparent mirror 2) is connected to the autofocus module 5, and the first semi-transparent mirror 8 (i.e. Figure 14The semi-transparent and semi-reflective mirror 1) is connected to the reflector 9, and the reflector 9 is connected to the bright field light source module 7.

[0063] The 13 sample stages are XY stages used to carry semiconductor wafer samples to be tested, and can move in the horizontal plane along a preset path to achieve switching of imaging field of view.

[0064] The bright-field light source module 7 provides a collimated illumination beam, which passes sequentially through the reflector 9, the first semi-transparent mirror 8, and then through the microscope objective 10 to reach the sample surface, thus achieving bright-field illumination. The reflected light signal from the sample surface is collected by the microscope objective 10 and then sequentially passes through the first semi-transparent mirror 8, the second semi-transparent mirror 6, and the infinity imaging tube 4 before being imaged onto the target surface of the camera 2.

[0065] The dark field light source module 11 includes 12 collimated LED sub-light sources arranged symmetrically along a hollow ring layout. Each sub-light source includes a fixed wavelength high-stability LED bead, a beam expander lens, and a collimating lens arranged sequentially along the optical axis.

[0066] The conical lens 12 is integrally molded from PMMA material, with a hollow structure at the center. The diameter of the hollow aperture is customized according to the outer diameter of the microscope objective 10. The upper surface of the conical lens 12 is flat, and the lower surface is conical. The collimated sub-light source is incident perpendicularly to the upper surface of the conical lens 12. After passing through the first surface of the conical lens 12, the beam transmission direction remains unchanged. When the beam reaches the second surface of the conical lens 12, it is refracted. The direction of the refracted beam is determined by the physical base angle α of the conical lens 12, the refractive index n of the lens, and the incident wavelength λ.

[0067] Adjust the assembly positions of the sub-light source module and the conical lens 12 in the ring layout so that the convergence point of the ring beam coincides with the focal plane of the test objective.

[0068] The lighting control module is used to independently switch and control the power of 12 ring-shaped distributed sub-light sources, enabling nine flexibly switchable dark field lighting modes.

[0069] The pulse drive module uses a rectangular wave pulse drive scheme to generate high-precision pulse current signals. When using an LED light source, within the rated peak current range, the LED's luminous brightness is proportional to the peak current passing through it. Utilizing the LED's fast response characteristics, a massive instantaneous current is injected to produce instantaneous high-brightness illumination. When using a laser light source, when the laser's drive current exceeds the threshold current, the output optical power exhibits a good linear growth relationship with the drive current. When the rectangular wave pulse current reaches its peak, camera 2 is triggered to start exposure, ensuring that the exposure time completely coincides with the effective emission period of the pulse flash.

[0070] The autofocus module 5 drives the Z-axis to complete focusing, and the XY displacement stage realizes the field of view switching, working in conjunction with the lighting and acquisition process to build a closed-loop imaging system for the entire process.

[0071] like Figure 2 As shown, based on the spectral requirements of the specific detection scenario, a fixed-wavelength, high-stability LED (e.g., center wavelength 550nm, half-width < 20nm) or a decoherent laser source is selected as the core light-emitting device. To optimize beam propagation characteristics, a beam expander lens and a collimating lens are sequentially integrated along the optical axis at the light source output end to form an integrated beam shaping module. The beam expander lens uses an aspherical lens to uniformly expand the divergent beam emitted from the light source, avoiding spot distortion caused by energy concentration. The collimating lens uses a plano-convex lens to collimate and calibrate the expanded beam, ultimately forming a sub-light source module with high parallelism and high uniformity. This module effectively improves beam utilization, reduces beam size and divergence angle, and provides stable and controllable incident light for subsequent optical detection. In this embodiment, to balance the sub-light source module volume and the emitted spot diameter, a lens with a diameter d ≤ 10mm is selected, ensuring that the assembled sub-light source diameter is ≤ 12mm.

[0072] like Figure 3 As shown, the 12 collimated LED sub-light sources are symmetrically arranged in a hollow ring layout, constructing 12 independent illumination paths that do not interfere with each other and have clearly defined directions. The hollow ring structure facilitates the precise coaxial assembly of the light source module and the microscope objective 10, resulting in a ring-shaped illumination light field distribution. When light acts on the observed sample, the multi-directional ring illumination beams work together to effectively fill the shadow areas of the sample contour and suppress the edge blurring problem that is prone to occur in traditional unidirectional illumination. Ultimately, a uniform illumination light field distribution with ring characteristics is formed in the imaging space, providing a stable illumination foundation for subsequent high-resolution microscopic imaging and accurate identification of sample details. Its symmetrical ring arrangement structure is shown in the figure.

[0073] This system uses a hollow conical lens 12 to focus the annular beam. The lens material, size, cone angle, and other parameters are designed and manufactured according to the incident angle required for dark-field illumination, achieving an isotropic, focused, high-brightness dark-field illumination spot. The center of the conical lens 12 has a hollow structure, reserving an assembly position for the microscope objective 10. The optical path structure is as follows... Figure 4 As shown.

[0074] The collimated sub-light source is incident perpendicularly to the upper surface (horizontal plane) of the conical lens 12. After passing through the first surface of the conical lens 12, the beam propagation direction remains unchanged. When the beam reaches the second surface of the conical lens 12, it is refracted, and the propagation direction of the refracted beam is the final output beam direction. The direction of the refracted beam is determined by the physical base angle of the conical lens. Lens refractive index n, incident wavelength The decision is made jointly. The specific calculation method is as follows: .

[0075] in is the refractive index of the cone lens at the illumination wavelength; Let be the angle of incidence of the light beam as it passes through the second surface of the conical lens. This angle is defined when the sub-source light source is incident perpendicularly to the upper surface of the conical lens. ; The refractive index of air is 1. The angle of refraction of the light beam after passing through the second surface of the conical lens is denoted as .

[0076] The angle between the refracted beam and the vertical direction The calculation method is as follows:

[0077] PMMA was chosen as the material for the design and fabrication of the cone lens. The dispersion formula for PMMA is:

[0078] Where n is the optical refractive index of the material. The wavelength is the incident light wavelength. When illuminated by a light source with a wavelength of 400 nm, the refractive index of PMMA material is 1.5010. When the illumination wavelength is 550 nm, the refractive index of PMMA material is 1.4910.

[0079] Adjust the assembly positions of the sub-light source modules and conical lenses in the ring layout so that the convergence point of the ring beam coincides with the focal plane of the test objective lens. This ensures that all directly reflected light cannot enter the objective lens, and only the scattered light from defects on the surface of the sample under test can be captured by the objective lens, thus achieving efficient dark-field illumination. The calculation method for the mounting height of the upper surface of the conical lens is as follows:

[0080] Where r1 is the distance from the center of the sub-source to the optical axis of the microscope objective; r2 is the radius of the conical lens; and α is the physical base angle of the conical lens. The angle by which the refracted beam deviates from the vertical direction.

[0081] To adapt to the optical response characteristics of different types of defects in semiconductor wafers (such as particles, scratches, cracks, and lattice defects) and overcome the limitations of traditional single illumination modes, this solution employs independent sub-light source control technology. Through precise switching and power control of 12 ring-distributed sub-light sources, nine flexibly switchable dark field illumination modes are achieved to meet the needs of defect detection in various scenarios. The illumination modes are mainly divided into: like Figure 3 As shown, the ring-shaped focused light spot activates all 12 sub-light sources (numbered 1 to 12 clockwise), with all sub-light sources having the same power (default 100%, adjustable). The illumination angle covers 360°, resulting in optimal light field uniformity and minimal intensity deviation in all directions. This design is suitable for rapid scanning across the entire field of view and general-purpose defect detection (such as surface particles and randomly distributed scratches). The 360° illumination without blind spots maximizes defect scattering signals and improves detection efficiency.

[0082] like Figure 5 As shown, unidirectional asymmetric illumination is used; only a single sub-light source in a single direction is turned on (e.g., number 1, any direction can be selected), while the rest of the sub-light sources are turned off. In this case, the incident direction is singular, suitable for determining the direction of defects (e.g., the extension direction of linear scratches) and enhancing the directional scattering signal of small defects; by switching between different directions of unidirectional illumination, the three-dimensional morphology of defects can be verified.

[0083] like Figure 6 As shown, this is a bidirectional non-centrosymmetric illumination system. Two sets of sub-light sources with an angle other than 180° are activated in a ring-shaped distribution (e.g., numbers 1 and 4, 90° angle; numbers 2 and 6, 120° angle, etc.), and the power of each set of light sources can be adjusted independently. The illumination angle covers an asymmetric range (e.g., 90°, 120°, or 150°), and the light field is non-centrosymmetric; the intensity deviation in both directions can be adjusted as needed. It is suitable for detecting non-centrosymmetric defects (e.g., tilted scratches, eccentrically distributed particle groups, asymmetric edge notches) and localized asymmetric structural regions on wafers. Symmetrical illumination light fields do not match the morphology / distribution characteristics of asymmetric defects, causing defect scattering signals to cancel each other out and details to be masked. Non-centrosymmetric light fields can break the symmetry redundancy of defect signals, highlighting irregular morphological features and improving the distinction between defects and background, especially suitable for "hidden asymmetric defects" that are difficult to identify with traditional symmetrical illumination. like Figure 7 As shown, this is a three-dimensional, non-centrosymmetric illumination system that activates three sub-light sources with non-uniformly distributed angles (e.g., numbered 1, 4, and 9, with angles of 0°, 90°, and 120° respectively). The power is independently adjustable, the incident direction is asymmetrical, and the power adjustment range is independently controllable from 0% to 100%. It is suitable for asymmetric defect detection (such as irregular cracks or locally clustered particles), breaking signal symmetry through an asymmetric light field to enhance the distinction between defects and the background.

[0084] like Figure 8 As shown, this is a four-way non-centrosymmetric illumination system. Four sub-light sources with non-uniform angles within a ring distribution (e.g., numbered 1, 3, 7, and 10, with adjacent angles of 60°, 120°, 90°, and 90° respectively) are activated, and the power of each sub-light source is independently adjustable. This system is suitable for detecting four-way irregular defects (such as tilted cross scratches, eccentric four-quadrant particle clusters, and asymmetric quadrilateral corner defects) and locally complex asymmetric structures on wafers (such as non-center-aligned chip arrays). The non-uniform four-way light field can accurately match the asymmetric distribution characteristics of defects, avoiding signal cancellation problems caused by symmetrical illumination, ensuring that scattered signals from all regions of the defect are effectively excited, and improving the detection rate of latent four-way defects.

[0085] like Figure 9As shown, bidirectional centrally symmetrical illumination activates two sub-light sources symmetrically arranged along the annular diameter (e.g., numbered 1 and 7, 2 and 8, with an included angle of 180°), maintaining consistent power for both sources. The light is incident symmetrically, resulting in a symmetrical light field. This method is suitable for detecting defects in symmetrical structures (such as wafer edge notches and symmetrically distributed lattice defects). Bidirectional complementary illumination can suppress shadows cast by unidirectional illumination, improving the integrity of defect contours.

[0086] like Figure 10 As shown, the three-dimensional centrally symmetrical illumination turns on three sets of sub-light sources (such as numbered 1, 5, and 9, with adjacent angles of 120°) that are evenly distributed along the ring. The three sub-light sources have the same power and the intensity in each direction is relatively small. This is suitable for the detection of defects in triangular lattice defects, particle groups distributed in three directions on the wafer surface, and equilateral triangular structural regions. The three-dimensional uniform illumination makes the intensity of the scattered signal of symmetrical defects equal, improving the accuracy of defect positioning and size measurement.

[0087] like Figure 11 As shown, the four-directional centrally symmetrical illumination system activates four sets of sub-light sources (numbered 1, 4, 7, and 10, with an included angle of 90°) that are uniformly symmetrical along a ring. The four sets of sub-light sources have the same power, providing uniform coverage in all four directions, and the light field is rotationally symmetrical. This system is suitable for medium-to-high resolution defect detection (such as submicron scratches and micro-cracks). Compared to ring illumination, it reduces beam crosstalk while maintaining the advantages of multi-directional illumination, thus improving the signal-to-noise ratio.

[0088] like Figure 12 As shown, the six-directional centrally symmetrical illumination turns on six sets of sub-light sources (numbered 1, 3, 5, 7, 9, and 11, with an included angle of 60°) that are uniformly symmetrical along a ring. The six sets of sub-light sources have the same power and densely cover the six directions, resulting in a uniform light field. This is suitable for high-precision quantitative analysis of defects (such as defect size measurement and depth assessment). The dense symmetrical illumination makes the defect scattering signal more stable, providing accurate data support for quantitative calculations.

[0089] LED or laser light sources employ pulsed driving schemes instead of the traditional constant-on driving mode. By precisely controlling the amplitude, repetition frequency, and duty cycle parameters of the pulse current, independent and continuous adjustment of the luminous brightness of each sub-light source can be achieved. When using LED light sources, within the rated peak current range, the luminous brightness of the LED is directly proportional to the peak current passing through it. Utilizing the fast response characteristics of LEDs, a sudden surge of current can be injected into the LED to produce instantaneous high-brightness illumination. Short-pulse driving can effectively avoid the thermal effects of the light source, balancing high brightness and illumination stability. When using laser light sources, when the laser's driving current exceeds the threshold current, the output optical power exhibits a good linear growth relationship with the driving current. Precise control of the peak value and pulse width of the pulse current can also obtain short-pulse-width, high-peak-power pulse output.

[0090] like Figure 13As shown, this system employs a rectangular wave pulse driving scheme to generate a high-precision pulsed current signal, enabling precise timing control of the illumination source and achieving accurate control over the luminous intensity and activation timing. When the rectangular wave pulsed current reaches its peak, camera 2 is triggered to start exposure, ensuring that the exposure time completely coincides with the effective emission period of the pulsed flash, maximizing the image signal-to-noise ratio and avoiding insufficient imaging brightness or signal attenuation due to timing misalignment. After camera 2 synchronously acquires the original dark-field image and stores it in its local cache, the stage moves to the next field of view along a preset path, completes autofocus, and repeats the above illumination and acquisition process, constructing a closed-loop imaging system until all fields of view are scanned, providing high-quality raw data for subsequent defect detection algorithms.

[0091] Compared with the prior art, the present invention has the following significant advantages: 1. Precise and controllable beam quality ensures reliable detection: A two-stage collaborative shaping structure using a beam expander and collimator lens enables precise control and optimization of the beam. This structure effectively avoids the common defects of traditional light sources, such as excessive beam divergence angle and uneven energy distribution, thereby solving core problems caused by these defects, such as image blurring, defect signal intensity fluctuations, and low signal-to-noise ratio. This lays a solid foundation for high-contrast differentiation between defects and background in dark-field imaging scenarios, significantly improving the accuracy and stability of defect identification.

[0092] 2. Multi-directional collaborative illumination eliminates shadow occlusion. A symmetrical ring layout of 12 collimated sub-light sources (with adjacent sub-light sources at a 30° angle) achieves 360° omnidirectional beam coverage. Compared to traditional unidirectional illumination schemes that easily produce shadow areas along the sample outline, this ring collaborative illumination design precisely fills shadow gaps through multi-directional complementary light fields, ensuring that all key areas on the sample surface—including defect depressions, edge steps, and small grooves—receive uniform and sufficient illumination. This design fundamentally solves the problem of defect signal distortion caused by unidirectional illumination, effectively avoiding the risk of missed or false detections of minor defects, and providing high-precision, high-reliability data support for subsequent quantitative analysis of defect size and morphology.

[0093] 3. Spectral Characteristic Matching for Diverse Detection Scenarios. Addressing the spectral response requirements of various detection scenarios, such as silicon wafer scratch detection, silicon carbide particle detection, and lattice defect identification, a dual-selection solution of narrowband LED and decoherent laser is provided. Based on the defect type and the optical absorption and scattering characteristics of the wafer material, the incident light with a customized center wavelength (tunable from 400nm to 850nm) and narrow half-width (FWHM) (LED ≤ 20nm, laser ≤ 1nm) can maximize the optical response intensity of the defect signal, significantly improving the detection rate of minute defects such as particles and scratches.

[0094] 4. Achieving an isotropic illumination spot. The conical lens uses a special conical structure to focus and control the annular collimated beam, converging the multi-directional annular beam onto the wafer inspection area using the principle of optical refraction. This forms a high-energy-density, uniformly focused spot, providing a high signal-to-noise ratio illumination basis for dark-field inspection and effectively enhancing the scattering signal of minute defects. The circumferentially symmetrical structure of the conical lens ensures that the focused annular spot has perfect rotational symmetry, with the size, intensity, and divergence angle of the spot being completely consistent in all directions. This isotropic characteristic ensures that defects on the wafer surface in different directions (such as scratches at any angle or annular particles) receive the same intensity of illumination, avoiding differences in defect signals caused by deviations in the directionality of the spot, and guaranteeing the consistency and accuracy of the inspection results.

[0095] 5. Hollow structure design, reserving space for objective lens assembly, achieving high optical path compatibility. The conical lens adopts a central hollow structure, with the hollow aperture customized according to the outer diameter of the microscope objective (typical value φ20mm~φ40mm). The microscope objective can be directly coaxially assembled in the hollow area without changing the core layout of the original imaging optical path. It is fully compatible with microscope objectives with NA≤0.8, and simultaneously supports the use of both high-magnification and low-magnification objectives, flexibly matching imaging scenarios of different resolutions. No replacement of the main structure of the conical lens is required; only the installation position of the conical lens needs to be adjusted to achieve different resolution imaging requirements, reducing equipment upgrade and maintenance costs. This integrated coaxial "lens-objective" design avoids the use of additional reflectors, steering mirrors, and other components, reducing optical path loss and assembly errors, making the entire optical system structure more compact, and significantly improving the convenience of optical path integration. The hollow structure ensures that the entrance pupil of the microscope objective is fully exposed and will not be blocked by the conical lens material. The incident beam is focused by the conical lens and acts on the wafer surface. The light signal scattered by the defects can directly enter the microscope objective for imaging, avoiding signal attenuation or imaging blind spots caused by structural obstruction.

[0096] 6. Precise and controllable illumination angle, adapting to the core requirements of dark-field inspection. The direction of the emitted beam from the conical lens is determined by the cone angle α, the lens refractive index n, and the incident wavelength λ (following the law of refraction). By precisely designing these parameters, the angle of the beam ultimately incident on the wafer surface can be strictly controlled within the angle range required for dark-field inspection (typical incident angle range 30°~60°, determined according to the numerical aperture of the microscope objective). This ensures that the beam is only scattered by defects before entering the objective for imaging, achieving a pure dark-field illumination effect, effectively suppressing background noise, and improving the contrast between defects and the background.

[0097] 7. Simple and compact structure, low processing difficulty and excellent consistency. The core component, the conical lens, adopts an integrated molding design, containing only two optical surfaces (the upper surface is flat and the lower surface is conical). Compared with traditional complex multi-lens focusing systems, the processing technology of this design is simpler and more efficient, and the processing accuracy of core parameters such as cone angle and hollow aperture is easier to ensure. This results in stronger product consistency during mass production and can effectively reduce production and application costs.

[0098] 8. Multiple lighting modes available, perfecting the dark field lighting system. Integrating nine lighting modes makes the dark field lighting solution system more complete. Through comparative analysis of the characteristics of different lighting modes, the appropriate lighting solution can be flexibly selected according to the differentiated requirements of defect type, detection accuracy, and detection efficiency, achieving the core goal of "full coverage, high precision, and high adaptability" in semiconductor wafer defect detection, and significantly improving the comprehensive detection capabilities and scene adaptability of a single device.

[0099] 9. Pulse-driven instantaneous high-brightness illumination. Injecting a massive peak current within a short pulse period enables the light source to generate instantaneous high-energy-density illumination, significantly increasing the intensity of scattered light from the defect surface. This effectively compensates for the small scattering cross-section, greatly improving the grayscale contrast between the defect and the background, and providing sufficient signal for subsequent image algorithms to identify defects. The pulse-driven solution, with its core advantages of precise brightness control and instantaneous high-brightness output, provides a stable, efficient, and low-loss illumination solution for dark-field defect detection on semiconductor wafers, making it particularly suitable for high-precision, high-throughput wafer inspection scenarios.

[0100] 10. Precise timing control, balancing noise reduction and efficiency improvement. Employing a control strategy that precisely synchronizes the pulsed light source and camera exposure time, the system ensures that original images are acquired only during periods when the light source is on. This effectively suppresses ambient stray light interference, easily obtaining high-quality inspection images with clean backgrounds and high contrast. Simultaneously, data storage is completed during the intervals when the light source is off, and in conjunction with the XY stage, rapid switching of the imaging field of view is achieved, significantly improving the overall operating efficiency of the inspection system.

[0101] The specific implementation process is as follows: like Figure 14As shown, based on the spectral requirements of the specific detection scenario, high-stability LED beads with a fixed wavelength (e.g., center wavelength 550nm, half-width at half-maximum < 20nm) or a decoherent laser source are selected as the core light-emitting device. To optimize beam propagation characteristics, in each sub-source structure, a beam-expanding lens 1 and a collimating lens 2 are sequentially arranged along the optical axis to form a beam-shaping module. The beam-expanding lens 1 uses an aspherical lens to uniformly expand the divergent beam emitted from the source, avoiding spot distortion caused by energy concentration. The collimating lens 2 uses a plano-convex lens to calibrate the expanded beam, ultimately forming a sub-source module with high parallelism and high uniformity, effectively improving beam utilization, reducing beam size and beam divergence angle, and providing stable and controllable incident light for subsequent optical detection. To balance the sub-source module volume and the diameter of the emitted beam spot, lenses with a diameter d ≤ 10mm are selected, ensuring that the assembled sub-source diameter is ≤ 12mm.

[0102] Twelve collimated LED sub-light sources are symmetrically arranged in a hollow ring layout, constructing twelve independent illumination paths that do not interfere with each other and have clearly defined directions. The hollow ring structure facilitates the precise coaxial assembly of the light source module and the microscope objective, resulting in a ring-shaped illumination light field distribution. When light acts on the observed sample, the multi-directional ring illumination beams work synergistically to effectively fill the shadow areas of the sample contour and suppress the edge blurring problem that is prone to occur in traditional unidirectional illumination. Ultimately, a uniform illumination light field distribution with ring characteristics is formed in the imaging space, providing a stable illumination foundation for subsequent high-resolution microscopic imaging and accurate identification of sample details. Its symmetrical ring arrangement structure is shown in the figure.

[0103] Select the bright-field / dark-field imaging mode based on the structural and size characteristics of the target to be detected.

[0104] In bright-field imaging mode: the collimated LED light source illuminates the sample surface after passing through the reflector 9, the first semi-transparent and semi-reflective mirror 8, and the microscope objective. The reflected light signal from the sample surface is collected by the microscope objective and then imaged onto the camera target surface after passing through the first semi-transparent and semi-reflective mirror 8, the second semi-transparent and semi-reflective mirror 6, and the infinity imaging tube 4.

[0105] In dark-field illumination mode: First, an illumination scheme is selected based on the sample's structural characteristics and the type of defect to be detected. The selected sub-light source provides a collimated illumination beam, which, after refraction by the conical lens 12, illuminates the focal plane of the microscope objective 10, achieving dark-field illumination. The scattered light signal from the sample surface is collected by the microscope objective 10 and then sequentially passes through the first semi-transparent mirror 8, the second semi-transparent mirror 6, and the infinity imaging tube 4 before being imaged onto the target surface of the camera 2. The illumination scheme selection process is as follows: Step 1: Define the testing objective: For batch rapid screening scenarios, the ring illumination mode should be selected directly. This mode can turn on all 12 independent power-controllable sub-light sources to construct a high-brightness, highly uniform isotropic circular light field, achieving full coverage of the detection field without blind spots; its illumination intensity can reach the system peak, and the detection time of a single sample is significantly reduced, which can maximize the compression of the batch detection cycle and significantly improve the screening efficiency and throughput of symmetrical defects, especially suitable for the full inspection process of semiconductor wafers and precision structural components.

[0106] If the target of the inspection is a high-precision quantitative analysis scenario, specifically covering the precise quantitative characterization of defect size and depth, then a symmetrical six-directional illumination mode is directly selected. This solution is based on a 60° equiangularly uniformly distributed symmetrical optical architecture, integrating six independent power closed-loop control sub-light sources. Through a balanced six-directional light field distribution, it effectively suppresses interference factors such as uneven surface reflection and edge shadows on the test piece, significantly reducing the impact of light field disturbances on the quantitative results. This ensures the accuracy and repeatability of the quantitative analysis results, meeting the core technical requirements of high-precision inspection.

[0107] If the detection target focuses on the accurate determination and classification of defect direction and shape, then proceed to the second step of defect feature subdivision and discrimination process.

[0108] Step 2: Determine the symmetry characteristics of the defect 1. Defects of regular symmetry: For bidirectional centrosymmetric defects (such as spherical particulate contaminants, symmetrical lattice structure anomalies, and centrosymmetric surface micro-defects), a symmetrical bidirectional illumination mode is selected. This mode, based on a 180° symmetrical illumination source layout, can effectively suppress unidirectional shadow interference, achieve high signal output balance and stability, accurately adapt to centrosymmetric defect feature extraction, and significantly improve the consistency of defect identification.

[0109] For bidirectional centrosymmetric defects (such as spherical submicron-sized particulate contaminants on semiconductor wafer surfaces, symmetrical lattice structure distortions, and centrosymmetric surface micro-defects), a symmetrical bidirectional illumination mode is selected. This mode is based on a 180° symmetrical optical architecture, integrating two independent power closed-loop control sub-light sources. It has low crosstalk rate and can effectively suppress unidirectional shadow interference through a balanced bidirectional light field distribution, achieving high signal output balance and stability. It accurately adapts to the feature extraction requirements of centrosymmetric defects, significantly improving the consistency of defect identification, and is particularly suitable for precision optical inspection scenarios such as quantitative detection of semiconductor wafer particle contamination and screening of lattice symmetry defects.

[0110] For triaxial symmetric defects (such as submicron-level micro-defects on the surface of triangular precision structures, equilateral uniformly distributed particle groups, and triaxial symmetric lattice structure distortions), a triaxial symmetric illumination mode is selected. This mode is based on a 120° equiangular uniformly distributed symmetric optical architecture and integrates three independent power closed-loop control sub-light sources. It can eliminate signal blind spots in the detection field of view through a balanced triaxial light field distribution, accurately match the feature extraction requirements of triaxial symmetric defects, and significantly improve the positioning accuracy and recognition consistency of symmetric structural defects. It is especially suitable for precision optical inspection scenarios such as triangular lattice defect detection and equilaterally distributed particle group recognition.

[0111] Four-way symmetric defects (such as micro-defects in square chip arrays, four-way uniformly discrete particle groups, and four-way symmetric lattice structure distortions): A symmetrical four-way illumination mode is selected. This mode is based on a 90° orthogonal uniformly distributed symmetrical optical architecture, configured with four independent controllable sub-light sources. While ensuring high-resolution feature recognition, it achieves a dynamic optimization balance between detection accuracy and detection efficiency. It can accurately match the feature extraction requirements of four-way symmetric defects, significantly improving the detection consistency and accuracy of symmetrical structural defects, and is especially suitable for batch inspection in scenarios such as semiconductor chip arrays and precision structural component surfaces.

[0112] For six-way symmetric defects (such as hexagonal semiconductor wafer lattice distortion, six-way uniformly distributed submicron-level particle clusters, and micro-defects on the surface of hexagonal precision packages), a symmetrical six-way illumination mode is selected. This mode is based on a 60° equiangularly uniformly distributed six-way symmetric optical architecture, integrating six independent power closed-loop control sub-light sources. It can suppress the signal blind zone of symmetric defects through a balanced six-way light field distribution, taking into account both the accurate extraction of six-way symmetric features and the high-precision quantitative analysis requirements. This significantly improves the accuracy and consistency of defect size / depth measurement, and is especially suitable for high-end precision inspection scenarios such as quantitative detection of semiconductor wafer lattice defects and surface quality control of hexagonal precision structural components.

[0113] 2. Irregular and asymmetric defects: Proceed to the third step.

[0114] Step 3: Select based on the number of defect directions 1. Unidirectional defects (such as linear continuous scratches, directional surface micro-defects, unidirectional lattice dislocations, etc.): Unidirectional illumination mode is selected. This mode relies on a single independent power controllable sub-light source to achieve directional light field illumination. Differential light intensity control can achieve directional enhancement of defect signals and suppression of background noise, significantly improving the accuracy of defect direction determination and the recognition rate of small unidirectional defects. It is suitable for scenarios such as unidirectional lattice defects and surface scratch screening in semiconductor wafers.

[0115] 2. Bidirectional non-centrosymmetric defects (such as tilted-angle linear scratches, bidirectional eccentrically distributed submicron-level micro-defects, and non-centrosymmetric lattice dislocations): A bidirectional non-centrosymmetric illumination mode is selected. This mode is based on an asymmetric optical architecture design with two independent power closed-loop control sub-light sources. Through differentiated light intensity control and precise matching of illumination angle, it can specifically highlight the optical characteristics of bidirectional non-centrosymmetric defects, effectively suppress background noise and stray light interference, and significantly improve the defect signal recognition and feature extraction accuracy. It is suitable for optical inspection scenarios such as quantitative detection of tilted scratches and screening of non-centrosymmetric lattice defects.

[0116] 3. Triaxial Irregular Defects (such as irregularly extended submicron-level microcracks, triaxial non-uniformly distributed surface defects, triaxial random lattice distortion defects, etc.): An asymmetric triaxial illumination mode is selected. This mode is based on an asymmetric optical architecture constructed from three independent power closed-loop controlled sub-light sources. Through differentiated light intensity control and dynamic matching of illumination angles, it can effectively break the signal redundancy and crosstalk interference of symmetric light fields, directionally enhance the optical characteristic signals of complex triaxial defects, and significantly improve the feature extraction accuracy, recognition accuracy, and effective detection rate of irregular morphological defects. It is suitable for optical inspection scenarios such as quantitative detection of microcracks and screening of triaxial random lattice defects.

[0117] 4. Four-directional irregular defects (such as multi-directional intersecting submicron linear scratches, four-quadrant eccentric discrete particle groups, and four-directional non-uniformly distributed surface micro-defects): A four-directional non-centrosymmetric illumination mode is selected. This mode is based on an asymmetric optical architecture constructed from four independent power closed-loop controlled sub-light sources. Through differentiated light intensity control and precise adaptation of illumination angle, it can significantly enhance the contrast of complex defect edge signals, optimize the detailed texture presentation of four-directional irregular defects, and greatly improve the feature extraction accuracy, edge recognition, and identification accuracy of multi-directional defects. It is suitable for optical inspection scenarios such as quantitative detection of multi-directional scratches on semiconductor wafers and screening of four-quadrant particle contamination.

[0118] After selecting the illumination scheme, the sub-light sources to be illuminated are simultaneously turned on via pulse drive, and camera 2 starts exposure synchronously. The exposure time of camera 2 is strictly synchronized with the length of the illumination pulse. When the light source is turned off, the exposure of camera 2 ends, and the raw image acquired by camera 2 is stored in the storage device. At the same time, the sample stage 13 moves horizontally to switch the field of view. The above process is repeated until the imaging of the entire wafer sample is completed.

[0119] The above descriptions are merely embodiments of the present invention, and common knowledge such as specific technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solutions of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. An epi-irradiation dark-field illumination micro-imaging system for semiconductor wafer inspection, characterized in that, include: A stage is used to hold semiconductor wafer samples to be tested. Bright field light source module, used to provide bright field illumination beam; The dark field light source module includes 12 collimated LED sub-light sources symmetrically arranged along a hollow ring layout. Each sub-light source includes a fixed wavelength high-stability LED lamp bead or a laser light source that has undergone decoherence processing, a beam expander lens, and a collimating lens arranged sequentially along the optical axis, forming an integrated beam shaping module. The 12 collimated sub-light sources construct 12 independent lighting paths that do not interfere with each other and have clear directions. The hollow conical lens has a central hollow structure. The upper surface of the conical lens is a plane and the lower surface is a cone. The 12 collimating sub-light sources are incident perpendicular to the upper surface of the conical lens. When the light beam passes through the second surface of the conical lens, it is refracted. The angle between the refracted light beam and the vertical direction is determined by the physical base angle α of the conical lens, the refractive index n of the lens, and the incident wavelength λ. The spot convergence point of the annular light beam coincides with the focal plane of the microscope objective. The central hollow aperture of the hollow conical lens is customized according to the outer diameter of the microscope objective. The microscope objective is coaxially mounted in the hollow region. The microscope objective is used to collect the scattered light signal from the sample surface. After being collected by the microscope objective, the scattered light signal passes sequentially through the first beam splitter, the second beam splitter, and the infinity imaging tube before being imaged onto the camera target surface. The lighting control module is used to independently switch and control the power of the 12 ring-shaped distributed sub-light sources, and realize at least three flexibly switchable dark field lighting modes, including ring-shaped focused light spot mode, unidirectional asymmetric lighting mode, and at least one multidirectional symmetric or asymmetric lighting mode. The pulse drive module is used to generate a high-precision pulse current signal using a rectangular wave pulse drive scheme, and to independently adjust the brightness of each sub-light source. When the rectangular wave pulse current reaches its peak value, it triggers the camera to start exposure, ensuring that the exposure time completely coincides with the effective emission period of the pulse flash. as well as The autofocus module and XY shift stage are used to complete autofocus and field of view switching.

2. The system according to claim 1, characterized in that, The beam expander is an aspherical lens, the collimating lens is a plano-convex lens, and the sub-light source module uses a lens with a diameter d ≤ 10 mm, ensuring that the diameter of the assembled sub-light source is ≤ 12 mm.

3. The system according to claim 1, characterized in that, The center wavelength of the fixed-wavelength high-stability LED bead is 550nm and the half-width at half-maximum (WHM) is <20nm, or a laser source that has undergone decoherence processing is selected, and the WHM of the laser source is ≤1nm.

4. The system according to claim 1, characterized in that, The conical lens is made of PMMA material, the central hollow aperture of the conical lens is φ20mm~φ40mm, and the angle range of the emitted beam from the conical lens incident on the wafer surface is 30°~60°.

5. The system according to claim 1, characterized in that, The dark field lighting modes implemented by the lighting control module include: Ring-shaped focused spot mode: All 12 sub-light sources are activated, all sub-light sources have the same power, and the illumination angle covers 360°; Unidirectional asymmetric lighting mode: Only the sub-light source in a single direction is turned on, while the rest of the sub-light sources are turned off; Two-way non-central symmetrical lighting mode: Two sets of sub-light sources with an angle of non-180° are turned on in the ring distribution, and the power of the two sets of light sources can be adjusted independently; Three-way non-centrally symmetrical lighting mode: Activate three sub-light sources with non-uniformly distributed angles, and adjust their power independently; Four-way non-central symmetrical lighting mode: Activate 4 sub-light sources with non-uniform angles in the ring distribution, and adjust the power of the four sub-light sources independently; Two-way centrally symmetrical lighting mode: Turn on two sub-light sources that are symmetrical along the annular diameter and have the same power; Three-way centrally symmetrical lighting mode: Turn on 3 sets of sub-light sources that are evenly distributed along the ring, with the same power; Four-way centrally symmetrical lighting mode: Four sets of sub-light sources, uniformly symmetrical along a ring, are activated with identical power; and Six-way centrally symmetrical lighting mode: Turn on 6 sets of sub-light sources that are uniformly symmetrical along the ring, with the same power.

6. The system according to claim 1, characterized in that, In the pulse driving module, when using an LED light source, within the rated peak current range, the luminous brightness of the LED is proportional to the peak current passing through it. The fast response characteristics of the LED are utilized to inject a large instantaneous current to generate instantaneous high-brightness illumination. When using a laser light source, when the driving current of the laser is greater than the threshold current, the output optical power is linearly related to the driving current. By precisely controlling the peak value and pulse width of the pulse current, a short pulse width and high peak power pulse output can be obtained.

7. The system according to claim 1, characterized in that, The 12 collimated sub-light sources are evenly distributed with adjacent sub-light sources having an angle of 30°.

8. A method for epi-irradiation dark-field illumination micro-imaging for semiconductor wafer inspection, characterized in that, The system as described in claim 1 includes the following steps: Step S1: Based on the spectral requirements of the specific detection scenario, select a fixed-wavelength high-stability LED lamp bead or a laser light source that has undergone decoherence processing as the core light-emitting device. At the output end of the light source, a beam expanding lens and a collimating lens are sequentially arranged along the optical axis to form a beam shaping module, thereby forming a sub-light source module with high parallelism and high uniformity. Step S2: Arrange the 12 collimated sub-light sources symmetrically in a hollow ring layout to construct 12 independent lighting paths that do not interfere with each other and have clear directions, forming a ring-shaped lighting light field distribution; S3: A hollow conical lens is used to focus the ring beam. The center of the conical lens has a hollow structure to reserve an assembly position for the microscope objective. The collimated sub-light source is incident perpendicular to the upper surface of the conical lens. When it passes through the second surface of the conical lens, it is refracted. The direction of the refracted beam is determined by the physical base angle α of the conical lens, the refractive index n of the lens, and the incident wavelength λ. The assembly position of the ring-shaped sub-light source module and the conical lens is adjusted so that the spot convergence point of the ring beam coincides with the focal plane of the test objective. The microscope objective is coaxially assembled in the hollow area. S4: Based on the type, symmetry characteristics, and number of directions of the defects to be detected, select the appropriate lighting scheme from multiple dark field lighting modes, and achieve lighting mode switching by independently switching and power controlling the 12 ring-shaped distributed sub-light sources; S5: A rectangular wave pulse driving scheme is used to generate a high-precision pulse current signal. When the rectangular wave pulse current reaches its peak value, the camera is triggered to start exposure, ensuring that the exposure time completely coincides with the effective emission period of the pulse flash. The camera simultaneously acquires the original dark field image and stores it in the local cache. S6: The stage moves to the next field of view according to the preset path, and after completing the autofocus, it repeats the above illumination and image acquisition process to build a closed-loop imaging system for the entire process until all fields of view are scanned.

9. The method according to claim 8, characterized in that, The lighting scheme selection process in step S4 includes: The first step is to define the detection objectives: for rapid batch screening scenarios, a ring lighting mode is selected; for high-precision quantitative analysis scenarios, a symmetrical six-directional lighting mode is selected; and for scenarios requiring precise determination and classification of defect direction and morphology, proceed to the second step. The second step is to determine the symmetry characteristics of the defect: For regular symmetrical defects, the corresponding symmetrical lighting mode is selected according to the number of symmetrical directions; for irregular and asymmetrical defects, proceed to the third step. The third step is to select the appropriate lighting mode based on the number of defects in each direction: for defects in a single direction, use a unidirectional lighting mode; for defects in two directions that are not centrally symmetrical, use a two-way non-centrally symmetrical lighting mode; for defects in three directions that are irregular, use an asymmetrical three-way lighting mode; and for defects in four directions that are irregular, use a four-way non-centrally symmetrical lighting mode.

10. The method according to claim 8, characterized in that, The method for calculating the mounting height h of the upper surface of the cone lens in step S3 is as follows: Where r1 is the distance from the center of the sub-source to the optical axis of the microscope objective; r2 is the radius of the conical lens; and α is the physical base angle of the conical lens. The angle by which the refracted beam deviates from the vertical direction.

Images