Method and system for nanoscale imaging based on second harmonic signal generation, and through-focus scanning optical microscope

Abstract

This specification describes methods and systems that improve the detection of target defects and the measurement of structures embedded within complex three-dimensional semiconductor structures. Through-focus scanning optical microscopes (TSOMs) that use nonlinear second-harmonic generation (SHG) optical signals emitted from a sample provide interface-selective sensitivity for the measurement and inspection of advanced semiconductor device structures. The TSOM / SHG system includes a spectral filter that allows collected light at wavelengths corresponding to SHG emission to pass through. In some embodiments, the TSOM / SHG system includes an ultrafast pulsed laser source emitting ultraviolet to near-infrared wavelengths to efficiently induce SHG at surface interfaces. By halving the wavelength specific to SHG, the illumination wavelength can be doubled without compromising resolution. In further embodiments, the TSOM / SHG measurement system includes an external illumination source, an external field source, or both to induce a DC electric field at one or more interfaces of the illuminated structure, thereby improving SHG emission.

Description

Technical Field

[0001] Cross - reference to Related Applications This patent application claims priority under 35 U.S.C. § 119 to U.S. Provisional Patent Application No. 63 / 460,032, filed on April 18, 2023, with the title "Nanoscale Imaging Apparatus Combining Through - Focus Scanning Optical Microscopy (TSOM) and Second Harmonic Generation (SHG)", the subject matter of which is incorporated herein by reference in its entirety.

[0002] The described embodiments relate to methods and systems for the measurement of semiconductor specimens, and more particularly, to semiconductor wafer inspection and metrology modes.

Background Art

[0003] Semiconductor devices, such as logic and memory devices, are typically manufactured by a series of processing steps applied to a substrate or wafer. These processing steps form various features and multiple structural levels of the semiconductor device. Among them, lithography is one of the semiconductor manufacturing processes, which includes, for example, generating patterns on a semiconductor wafer. Further examples of semiconductor manufacturing processes include, without limitation, chemical mechanical polishing, etching, deposition, and ion implantation. Multiple semiconductor devices can be manufactured on a single semiconductor wafer and then separated into individual semiconductor devices.

[0004] To improve yield, inspection and measurement processes for detecting and measuring defects on wafers are used at various steps in the semiconductor manufacturing process. As design rules and process window sizes decrease, inspection and measurement systems need to capture a wider range of physical defects while maintaining high throughput. Inspection systems, such as unpatterned wafer inspection systems and patterned wafer inspection systems, illuminate wafers to inspect for undesirable defects. As semiconductor design rules evolve, the size of the minimum defect requiring detection continues to decrease. Semiconductor architectures are shifting from two-dimensional floating-gate architectures to fully three-dimensional shapes. In some examples, film stacks and etched structures become very deep (e.g., up to 6 micrometers or more). Such high aspect ratio structures present problems for patterned wafer inspection. To achieve the desired performance levels and device yield, it is essential to be able to measure defects embedded within these structures.

[0005] Furthermore, semiconductor manufacturing of semiconductor metal-oxide-semiconductor field-effect transistor (MOSFET) devices is evolving from vertical fin-based fin field-effect transistor (FinFET) architectures to gate-all-around (GAA) architectures to achieve better performance, lower power consumption, smaller area, and lower costs. This is driven by increasingly stringent dimensional measurement and defect inspection requirements for process control. FET devices manufactured according to the GAA architecture, such as nanosheet FET devices, forksheet FET devices, and complementary nanosheet FET devices, are characterized by multi-channel level structures, reduced feature size, and small spaces between channels in GAA FET devices. In the GAA architecture, channels are vertically aligned and separated by very small spaces. These spaces are currently, for example, about 10 nanometers and are trending towards being even smaller. Buried defects can be generated during the channel-release etching process step. The defect feature size is typically less than 10 nanometers. In some examples, the defect feature size is about 1 nanometer. The small size and embedded nature of the defects, along with the channel spacing in the GAA structure, pose significant challenges for current defect inspection and dimensional measurement systems.

[0006] Figure 1A is a simplified diagram of a defect-free GAA nanosheet device structure 10 after the GAA channel release process step. The GAA nanosheet device structure 10 includes a silicon substrate 13, source and drain structures 11 and 12, an interlayer dielectric layer 14, gate spacer layers 15 and 16, internal spacers 17A to 17C, and silicon channels 18A to C.

[0007] Figure 1B is a simplified diagram of a GAA nanosheet device structure 20 having etching residual defects 21 after the GAA channel release process step. The similarly numbered elements shown in Figure 1B are the same as the elements shown in Figure 1A. The etching residual defects are the residues of sacrificial silicon germanium (SiGe) that were not completely etched during the GAA channel release process step.

[0008] Figure 1C is a simplified diagram of a GAA nanosheet device structure 30 having channel erosion defects 31 after the GAA channel release process step. The similarly numbered elements shown in Figure 1C are the same as the elements shown in Figure 1A. Channel erosion defects occur when germanium (Ge) diffuses from sacrificial silicon germanium (SiGe) into silicon lattice channels, which are then removed by etching during the GAA channel release process step, resulting in erosion of a portion of the silicon lattice channel structure.

[0009] Figure 1D is a simplified diagram of a GAA nanosheet device structure 40 having high channel roughness defects 41 after the GAA channel release process step. Similar numbered elements shown in Figure 1D are similar to those shown in Figure 1A. Channel roughness defects occur when sacrificial silicon maggelgermanium (SiGe) diffuses into the silicon lattice channel structure, which is then removed by etching during the GAA channel release process step, eroding a portion of the silicon lattice channel structure. If the surface roughness of the silicon channel after the channel release etching process step is about 1 nanometer or more, this surface roughness reduces carrier mobility. Surface roughness of silicon channels is a defect that should be controlled to less than 0.5 nanometers in modern semiconductor manufacturing processes.

[0010] Figure 1E is a simplified diagram of a GAA nanosheet device structure 50 having void defects 51 adjacent to the channel 18B after deposition of high dielectric layers 51A to D and metal gate layers 52A to D. Elements of similar number shown in Figure 1E are similar to those shown in Figure 1A. As shown in Figure 1E, the high dielectric layers 51A to D and metal gate layers 52A to D are deposited after the GAA channel release process step. In the example shown in Figure 1E, voids 51 remain after metal gate filling.

[0011] The defects shown in Figures 1B to 1E are embedded defects in the high aspect ratio (HAR) stacked structure. The defect feature size is typically less than 10 nanometers, and can be close to 1 nanometer. These exemplary defects are adjacent to channels and negatively impact device yield and performance. From a process control perspective, these defects are unacceptable. In addition to defect detection, process challenges related to GAA, as shown in Figures 1A to 1E, include measuring the critical dimensions of individual channels (e.g., CD shown in Figure 1A) and the spacing between channels (e.g., S shown in Figure 1A).

[0012] In some cases, electronic testing is used to detect defects embedded within three-dimensional structures. However, multiple device layers must be manufactured before electronic testing can be performed. Therefore, defects cannot be detected early in the manufacturing cycle. As a result, performing electronic testing, especially during the R&D and manufacturing process startup phases where rapid defect assessment is crucial, incurs enormous costs.

[0013] In several other cases, embedded defects within three-dimensional structures can be detected using X-ray-based measurement techniques. For example, X-ray diffraction systems or coherent X-ray imaging systems can be used to detect embedded defects. While X-ray-based measurement techniques have the advantage of being non-destructive, their throughput remains extremely low.

[0014] In some other cases, atomic force microscopy (AFM) can be used as a tapping and non-contact technique for defect inspection. Unfortunately, positioning the probe tip of an AFM system within trenches less than 10 nanometers wide is extremely difficult. Attempts to inspect within such narrow trench features result in a short probe tip lifespan and frequent tool maintenance. Furthermore, the scanning operation mode of AFM systems is very slow. As a result, the buried defect inspection capability of AFM at the scale of current GAA architectures is extremely limited.

[0015] Optical bright-field (BF) inspection is a wafer inspection technique widely adopted in modern semiconductor manufacturing facilities. Optical BF inspection is a wide-field and low-dose exposure far-field inspection method commonly performed in modern semiconductor manufacturing facilities.

[0016] In some cases, optical BF inspection systems utilize broadband plasma (BBP) light sources that emit light at wavelengths below 193 nanometers. However, the transmission of light at these wavelengths for detecting buried defects on patterned wafers is limited. This results in high noise and low image contrast, thus limiting the ability to detect buried defects.

[0017] Other recent bright-field techniques, such as scanning near-field optical microscopy (SNOM), offer improved image resolution, for example, down to λ / 20 (where λ is the wavelength of the illumination light). Unfortunately, similar to optical BF inspection methods, SNOM illumination light has limited ability to penetrate embedded defects, requiring more precise depth of field (DOF) positioning compared to conventional far-field microscopy.

[0018] In several other examples, by utilizing the device-specific hyperbolic metamaterial structure, buried defect detection may be possible on three-dimensional NAND flash memory devices by leveraging the epsilon near-zero effect and hyperbolic Bloch mode formation. Spectral analysis in the longer wavelength infrared region reveals clear hyperbolic guided mode resonance features, which may enable the identification of defects at the multi-micrometer level across the entire depth of the device. The hierarchical structure of three-dimensional NAND flash memory presents periodically coupled surface plasmon modes, i.e., vertical Bloch modes, which are not available in logic GAA FET structures. Therefore, this method cannot be applied to the detection of buried defects within GAA targets.

[0019] In summary, optical BF inspection systems face issues of transparency and resolution when attempting to detect embedded defects adjacent to the channel structure of GAA devices.

[0020] One of the further wafer inspection technologies being implemented in modern semiconductor manufacturing facilities is electron beam inspection (EBI). Also known as scanning electron microscopy (SEM) technology, EBI can detect and characterize defects on the surface of a sample with a feature size of up to 1 nanometer.

[0021] In some other examples, electron beam inspection (EBI) is used directly to detect buried defects within three-dimensional structures. Backscattered electrons can be used to detect buried defects. Unfortunately, high electron landing energy is required for the backscattered electrons to penetrate to the location of the buried defect and provide enough energy to escape the defect region and return to the detector. Furthermore, there are limitations to the resolution of backscattered electrons generated by high landing voltages.

[0022] In practice, EBI has very limited ability to detect defects deeper than approximately 1 micrometer. In many cases, EBI is limited to depths well below 1 micrometer (e.g., less than 50 nanometers). This limitation is due to the practical limitations of electron dose before sample strain or fracture occurs. Therefore, the effectiveness of EBI as a defect detection tool for thick three-dimensional structures is limited.

[0023] Furthermore, electron beam inspection has a very narrow field of view, which limits throughput and hinders its implementation as an in-line production measurement system. Multiple electron beam columns are used as an attempt to increase wafer throughput. However, even with multiple columns, the throughput limitations of EBI are not resolved.

[0024] Through-focus scanning optical microscopy (TSOM) is an imaging technique based on conventional microscopes. This allows conventional optical microscopes to collect dimensional information by combining two-dimensional optical images acquired at a series of through-focus positions. The through-focus images are superimposed with respect to the focal point, resulting in a three-dimensional space containing optical information. From this three-dimensional space, a two-dimensional cross-sectional TSOM image is extracted through the defect location of the object of interest in a given direction. The defect signal is obtained by subtracting the selected images containing the defect from the corresponding baseline TSOM image.

[0025] The lateral and longitudinal measurement sensitivities of TSOM are generally considered to be sub-nanometer. In some examples, the measurement accuracy of nanodot and nanoparticle size structures using TSOM has been demonstrated to be comparable to that of SEM, provided that a reasonably quality TSOM image database is available.

[0026] TSOM involves the analysis of numerous images. In some cases, this allows for the reduction of optical cross-correlations, such as the correlation between linewidth and lineheight, resulting in reduced measurement uncertainty. In these examples, the reduction of cross-correlations overcomes the limitations of conventional optical measurement tools.

[0027] Using TSOM, the diffraction limit of conventional optical imaging systems can be avoided. Therefore, dimensional measurement resolution on the nanometer or sub-nanometer scale can be achieved with a standard wide-field optical microscope. TSOM has been demonstrated to have nanometer-level sensitivity to a wide variety of target materials and shapes, from simple nanoparticles to complex semiconductor structures including diffraction grating critical dimensions, overlays, pattern defect detection and analysis, FinFETs, nanoparticles, photomask linewidths, thin film thicknesses, through-silicon vias (TSVs), and high aspect ratio (HAR) targets.

[0028] However, due to the complexity of the buried GAA structure and the narrow spacing between channels, the sensitivity achievable by TSOM measurement for buried defects is limited. Due to intense diffraction within the GAA structure, it is difficult to separate the measured defects from the influence of neighboring structures, making defect detection complex and inaccurate. Furthermore, optical interference between measurement channels and variations between complex structures surrounding the target defect within the GAA structure also contribute to detection inaccuracies. Differences in the structures surrounding the target defect within the wafer, between wafers, between lots, and between products result in inconsistencies with the baseline criteria. This leads to unacceptable uncertainties in the detection and measurement of defects buried within the GAA structure.

[0029] Furthermore, it is difficult to specify the illumination wavelength and intensity for generating an image with a high signal-to-noise ratio at a depth far beyond the location of the defect. Typically, only light with a near-infrared or longer wavelength can achieve a sufficient penetration depth into the GAA transistor channel. Unfortunately, long-wavelength illumination light is not desirable because it causes limitations in image resolution.

Prior Art Documents

Patent Documents

[0030]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0031] In short, the shift to gate-all-around (GAA) architectures in semiconductor manufacturing increases the number and types of defects that can occur during manufacturing, making it more difficult to detect these defects. Buried feature dimensional measurement and buried defect inspection are challenging, requiring methods to achieve process control goals in advanced semiconductor manufacturing facilities. Generally, defects within GAA structures, such as those shown in Figures 1B to 1E, are difficult to detect and put a strain on the capabilities of current detection systems in terms of sensitivity, specificity, processing speed, and capture rate. Therefore, wafer defect inspection and measurement systems for buried defects adjacent to device channels in GAA structures are desired to achieve consistent device performance and yield. [Means for solving the problem]

[0032] This specification describes improved methods and systems for detecting target defects (DOIs) and measuring structures embedded within complex three-dimensional semiconductor structures. Through-focus scanning optical microscopes (TSOMs) that utilize nonlinear second-harmonic generation (SHG) optical signals emitted from a sample achieve interface-selective sensitivity for the measurement and inspection of structures manufactured according to advanced semiconductor device manufacturing techniques. Therefore, the ability to measure and inspect defects is improved in specific regions that emit SHG signals.

[0033] In one embodiment, a TSOM / SHG measurement system excites and detects SHG signals emitted from a sample to characterize structural defects and dimensions embedded within complex semiconductor structures. A TSOM microscope, based on an image of the detected SHG light, is an interface-based defect inspection and measurement imaging device. TSOM provides non-destructive, high-throughput focus scanning capabilities. The SHG light signal achieves high-resolution capabilities while reducing sensitivity to structural disturbances and changes around the target interface region. The SHG microscope acquires contrast images from changes in SHG signal generation of the sample activated by incident light.

[0034] In some cases, SHG optical signals originate from the material interface. In some of these cases, SHG optical signals are sensitive to the properties of structures and defects adjacent to inversion-symmetric materials such as silicon, while being largely unaffected by matrix and background interference from surrounding structures and materials. Such measurements detect and measure defects and structures based on their proximity to the inversion-symmetric material. Therefore, the significant limitations of TSOM in measuring structures embedded in complex semiconductor structures and detecting defects are minimized by using SHG optical signal imaging.

[0035] In some embodiments, the TSOM / SHG system includes a high-intensity laser light source for penetrating a semiconductor sample and exciting an SHG optical signal. In some embodiments, the laser light source is a powerful wavelength-tunable pulsed laser, such as a Ti-sapphire laser or a Yb-KGW laser. In some embodiments, the wavelength range of the laser illumination source extends from ultraviolet to near-infrared wavelengths. In some embodiments, the laser light source generates illumination light as a time-series of pulses repeated at a certain repetition rate. In some examples, the illumination light has a pulse repetition rate in the kilohertz to megahertz range. Each pulse is characterized by a pulse duration, such as the full width at half maximum (FWHM) of optical power versus time. In some examples, the illumination light has an ultrafast pulse duration, such as a pulse duration measured in picoseconds, femtoseconds, or attoseconds. In one example, the illumination light has a pulse repetition rate of less than 100 femtoseconds at an average optical power of at least 1 watt. Ultrafast pulses provide high peak power, making them suitable for efficiently inducing SHGs at surface interfaces embedded deep within semiconductor samples during measurement.

[0036] SHG emission occurs at half the illumination wavelength, i.e., twice the frequency, at the energy combination position. Therefore, the wavelength of the probe illumination beam is selected to be relatively long in order to allow deeper penetration into the sample without compromising resolution. In fact, by halving the wavelength inherent to SHG, the wavelength of the illumination probe beam can be doubled without compromising resolution. In some embodiments, the wavelength of the probe illumination beam is in the wavelength band beyond the visible light to avoid the transmission limitations inherent in ultraviolet-visible light to typical semiconductor materials such as SiGe and polysilicon.

[0037] In a further embodiment, the TSOM / SHG measurement system includes an external illumination source that generates external illumination directed at the sample being measured to excite the SHG signal. The external illumination can cause charge separation at one or more interfaces of the structure under illumination, thereby inducing a DC electric field. The DC electric field can enhance the SHG emission. In this way, the intensity of SHG generation and the measurement sensitivity at interfaces of inversion-symmetric materials such as silicon are enhanced by illumination from the external illumination source.

[0038] In a further embodiment, the TSOM / SHG measurement system includes an external field source that generates an external electric field across the sample under measurement to excite the SHG signal. In this way, the intensity of SHG generation and the measurement sensitivity at the interface of inversion-symmetric materials such as silicon are enhanced by the fluctuating electric field induced by the external field source.

[0039] In a further embodiment, the TSOM / SHG system includes a spectral filter in the collection optical path. The spectral filter blocks light of wavelengths corresponding to the illumination probe beam and the external pump illumination beam, while allowing collected light of wavelengths corresponding to the SHG emission in the sample being measured to pass through. In this way, the spectral filter separates the SHG collected light, which is half a wavelength of the probe illumination light, from all other light collected from the sample, such as reflected and scattered light, allowing only the SHG light to pass to the detector.

[0040] The TSOM / SHG system generates a three-dimensional image of a thick semiconductor structure from volumes measured in two lateral dimensions (e.g., parallel to the wafer surface) and a depth dimension (e.g., perpendicular to the wafer surface). The output from the detector is placed in a volume dataset corresponding to the measured volume.

[0041] In some embodiments, defects within the target structure are detected by comparing a target SHG image with a reference SHG image. In one example, defects are detected based on the difference between the reference image and the target image. In other examples, defects are identified by comparing a target SHG image with three-dimensional images of one or more simulated defects. In some other examples, a measurement library is generated that matches measurement data with defects measured by a reliable reference measurement system. In one example, the reliable reference measurement system is defect verification performed after focused ion beam etching of the specimen under consideration. Once the library is generated, defects relevant to subsequent measurements are estimated based on library matching.

[0042] In a further embodiment, defects are classified based on difference images obtained from SHG images generated under different measurement conditions, i.e., different optical modes. In some examples, SHG images are generated using different characteristics of incident illumination, such as polarization, wavelength, and angle of incidence; different characteristics of external illumination, such as different intensity levels; and different characteristics of external electric fields, such as different intensities.

[0043] The above is a summary and therefore inevitably includes simplifications, generalizations, and omissions of details. Those skilled in the art will understand that the summary is illustrative and not limiting. Other aspects of the devices and / or processes described herein, features of the invention, and advantages will become apparent in the non-limiting detailed description provided herein. [Brief explanation of the drawing]

[0044] [Figure 1A]This is a simplified diagram of the GAA nanosheet device structure in various defect states after the GAA channel release process step. [Figure 1B] This is a simplified diagram of the GAA nanosheet device structure in various defect states after the GAA channel release process step. [Figure 1C] This is a simplified diagram of the GAA nanosheet device structure in various defect states after the GAA channel release process step. [Figure 1D] This is a simplified diagram of the GAA nanosheet device structure in various defect states after the GAA channel release process step. [Figure 1E] This is a simplified diagram of the GAA nanosheet device structure in various defect states after the GAA channel release process step. [Figure 2] This is a simplified schematic diagram of an embodiment of a TSOM / SHG system configured to excite and detect SHG emission from an embedded semiconductor structure, and to classify, measure, or both target defects (DOIs) based on the detected SHG image. [Figure 3] This is a simplified schematic diagram of another embodiment of a TSOM / SHG system configured to excite and detect SHG emission from an embedded semiconductor structure and to classify, measure, or both target defects (DOIs) based on the detected SHG image. [Figure 4] This is a simplified schematic diagram of one embodiment of a system for defect detection and measurement recipe optimization for inspection of three-dimensional semiconductor structures. [Figure 5] This flowchart illustrates an exemplary method for detecting buried defects and measuring buried structures based on a three-dimensional image of second-harmonic generated light. [Modes for carrying out the invention]

[0045] Herein, we will refer in detail to some background examples and embodiments of the present invention, as illustrated in the attached drawings.

[0046] This specification describes methods and systems for improving the detection of target defects (DOIs) and the measurement of embedded structures within complex three-dimensional semiconductor structures. In particular, through-focus scanning optical microscopes (TSOMs) using nonlinear second-harmonic generation (SHG) optical signals emitted from a sample achieve interface-selective sensitivity for the measurement and inspection of structures manufactured according to advanced semiconductor device manufacturing techniques. This improves the measurement and defect inspection capabilities in specific regions that emit SHG signals. TSOM microscopes using SHG optical signals are small-spot-size, high-throughput technologies with in-line wafer inspection and measurement capabilities to meet the demands of modern semiconductor device mass production. Both design rule targets and actual device targets can be inspected and measured according to the methods and systems described herein.

[0047] TSOM microscopy using SHG optical signals is used to measure the structural dimensions and inspect for periphery defects of metal oxide semiconductor field-effect transistor (MOSFET) devices manufactured according to many architectures, including but not limited to planar FETs, FinFETs, gate-all-around (GAA) nanosheets, forksheets, and complementary nanosheet FETs. In addition to MOSFET devices, TSOM microscopy using SHG optical signals can also be used to perform structural dimensional measurements and periphery defect inspections of through-silicon vias (TSVs), 3D flash and 3D dynamic random-access memory (DRAM) memories, ferroelectric FETs, two-dimensional (2D) material-based FET devices where very thin monolayers are used for their channel, source, drain, and gate dielectrics, HfO2 film layers doped with Zr, Al, Gd, La, Si, Sr, and Y, and HfO2 multilayers with oxides of Zr, Al, Gd, La, Si, Sr, and Y, such as SiO2 / HfO2 / AlO3 and HfO2 / ZrO2 / HfO2. As a non-limiting example, types of defects on two-dimensional material-based FET devices include film lattice misalignment, etching residue, distortion and bending of the two-dimensional channel, and etching roughness.

[0048] As a non-limiting example, TSOM microscopy using SHG optical signals can be used to measure structures and detect defects at various process steps in the GAA nanosheet manufacturing process. This includes, but is not limited to, channel release etching residual defects, channel roughness, and channel shape remaining after channel release etching; high dielectric constant metal layer defects and metal gate layer defects after multiple steps of high dielectric constant metal gate formation, after SiGe recess etching and internal spacer formation; and source / drain epitaxial defects after source / drain epitaxial growth steps.

[0049] Inspecting and measuring embedded defects in patterned wafers is more complex and challenging than inspecting and measuring defects on the wafer surface. The complex structural geometry within patterned wafers and the various materials used at different depths make characterizing defects and features difficult. TSOM identifies target defects at multiple depths within the structure under high-throughput inspection. Because three-dimensional optical inspection captures signal propagation within the wafer, it can distinguish actual DOIs from false or noise, even in relatively thick samples (e.g., three-dimensional NAND wafers with layered structures thicker than 3 micrometers).

[0050] Second harmonic generation (SHG) is a nonlinear optical process in which two photons of the same wavelength interact with an optically nonlinear material, energy-coupled, and generate a new photon with twice the energy and half the wavelength of the original photon. SHG is nondestructive and is used for characterizing semiconductor surfaces and interfaces.

[0051] Generally, the crystal structure of silicon is a rhombic-cubic lattice with inversion symmetry. As a result, the electric field and polarization vector within bulk silicon do not fluctuate in the inversion system, and therefore SHG cannot be displayed. On the other hand, SHG can be generated from higher-order nonlinear responses from inversion-symmetric materials. In one example, SHG can be generated from electric dipoles on the silicon surface. In another example, SHG can be generated from an electric quadrupole response from bulk silicon excited by an externally applied electric field. In these examples, the inversion symmetry breaks down along the normal direction at the silicon surface so that the secondary surface susceptibility at the interface does not become zero. Interfacial breakdown in the surface region enables highly sensitive SHG, ranging from the breakdown of inversion symmetry at the surface of the inversion-symmetric material to the breakdown of inversion symmetry at interfaces with other materials.

[0052] In one embodiment, a TSOM / SHG measurement system excites and detects SHG signals emitted from a sample to characterize defects embedded within complex semiconductor structures and the dimensions of the structure. A TSOM microscope, based on an image of the detected SHG light, is an interface-based defect inspection and measurement imaging device. TSOM provides non-destructive, high-throughput focus scanning capabilities. The SHG light signal provides high-resolution capabilities with reduced sensitivity to structural disturbances and variations around the boundary region of interest. While an SHG microscope acquires a contrast image from changes in the SHG signal generation of a sample activated by incident light, a conventional optical microscope acquires a contrast image from which changes in the sample's optical density, path length, or refractive index are detected.

[0053] TSOM / SHG provides a rich image dataset of every location within deep and complex semiconductor structures, whereas conventional light scattering-based TSOM images are contaminated by matrix and background interference from surrounding structures and materials. This problem worsens as the complexity of the underlying structural geometry increases. In contrast, SHG signals occur under very specific structural conditions compared to conventional scattering. Therefore, SHG signals are more selective.

[0054] In some cases, SHG optical signals originate from the material interface. In some of these cases, SHG optical signals are highly sensitive to the properties of structures and defects adjacent to reflection-symmetric materials such as silicon, and considerably less sensitive to matrix and background interference from surrounding structures and materials. Such measurements enable defect detection and structural measurement based on proximity to inversion-symmetric materials. Therefore, the significant limitations of TSOM with respect to structural measurement and detection of defects embedded in complex semiconductor structures can be minimized by using SHG optical signal imaging.

[0055] TSOM microscopy using SHG optical signals improves the measurement and defect inspection of nanosheet GAA transistor channels and their surrounding structures. In some cases, charge and trapping properties at the channel surface or interface generate SHG optical signals that are sensitive to adjacent defects, surface roughness characteristics, surface erosion, channel stress / strain changes, and channel structural characteristics such as bending and surface roughness.

[0056] In some cases, TSOM measurements using SHG signals enable defect detection and structural measurement based on their approximation to silicon channels embedded in semiconductor structures. SHG signals emitted at interfaces of inversion-symmetric materials such as silicon are highly sensitive to the presence of additional dipoles near the interface. In some cases, surface bonding between different materials results in different SHG behaviors. In one example, disilane or hydrogen bonds to the surface of bulk silicon during an epitaxial process step. The behavior of SHG at the interface is highly dependent on the quality of the surface bonding. As an example, it has been observed that a detectable variation in SHG intensity occurs when the hydrogen process pressure is changed from 0.8 to 1.5 Torr.

[0057] Figure 2 is a simplified schematic diagram of an embodiment of a TSOM / SHG system 100 configured to excite and detect SHG emission from an embedded semiconductor structure and to classify, measure, or both, target defects (DOIs) based on the detected SHG image. The TSOM / SHG measurement system 100 includes a computer system, a wafer placement system, and an optical measurement subsystem including an illumination subsystem, an acquisition subsystem, and one or more detectors. The illumination subsystem includes an illumination source 101 and all optical elements in the illumination optical path from the illumination source to the wafer. The acquisition subsystem includes all optical elements in the acquisition optical path from the detector to each detector. For simplification, some optical components of the system have been omitted. Examples may include folding mirrors, polarizers, beamforming optics, additional light sources, additional concentrators, and detectors. All such variations are within the scope of the invention described herein. The measurement system described herein can be used for inspection, measurement, or both of patterned wafers and reticles.

[0058] As shown in Figure 2, the wafer 103 is illuminated by a perpendicularly incident beam 104 generated by one or more illumination sources 101. Alternatively, as shown in Figure 3, the illumination subsystem may be configured to direct the light rays to the detector at an oblique angle of incidence. In some embodiments, the system 100 may be configured to direct multiple light rays, such as obliquely incident and perpendicularly incident light rays, to the specimen. The multiple light rays may be directed to the specimen substantially simultaneously or sequentially.

[0059] Figure 3 is a simplified schematic diagram of an embodiment of a TSOM / SHG system 200 configured to excite and detect SHG emission from an embedded semiconductor structure and to classify, measure, or both target defects (DOIs) based on the detected SHG image. Similar numbered elements shown in Figure 3 are similar to those shown in Figure 2. As shown in Figure 3, the illumination beam 111 is incident on the surface of the wafer 103 at an oblique angle α, i.e., at an angle α with respect to an axis N perpendicular to the surface of the wafer 103. As shown in Figure 3, the illumination beam 111 is focused onto the wafer 103 by an illumination focusing optical system 147. As described with respect to Figure 2, the objective 109 collects SHG light 148 from the wafer 103.

[0060] Examples of illumination sources 101 include broadband laser sustained plasma light sources, lasers, supercontinuum lasers, diode lasers, helium-neon lasers, argon lasers, solid-state lasers, diode-pumped solid-state (DPSS) lasers, xenon arc lamps, gas discharge lamps, LED arrays, and incandescent lamps. The light source may be configured to emit nearly monochromatic or broadband light. In some embodiments, the illumination subsystem may also include one or more spectral filters that can limit the wavelength of light directed to the specimen. One or more spectral filters may be bandpass filters and / or edge filters and / or notch filters. Illumination may be provided to the specimen over any suitable wavelength range. In some examples, the illumination light includes wavelengths in the range of 260 nanometers to 950 nanometers. In some examples, the illumination light includes wavelengths beyond 950 nanometers (e.g., extending to 2,500 nanometers) to capture defects in high aspect ratio structures.

[0061] In some embodiments, the TSOM / SHG system includes a high-intensity laser light source for penetrating the semiconductor sample and exciting the SHG optical signal. In some embodiments, the illumination source 101 is a powerful, wavelength-tunable pulsed laser, such as a Ti-sapphire laser or a Yb-KGW laser. In some embodiments, the wavelength range provided by the laser illumination source extends from ultraviolet to near-infrared wavelengths.

[0062] In general, the illumination source 101 can be configured to generate illumination light 104 having any suitable temporal profile. In some embodiments, the illumination source 101 generates illumination light 104 as time-series pulses repeated at a certain repetition rate. In non-limiting examples, illumination light 104 has a pulse repetition rate in the kilohertz to megahertz range. Each pulse is characterized by its pulse duration, e.g., full width at half maximum (FWHM) of optical power versus time. In non-limiting examples, illumination light 104 has an ultrafast pulse duration measured in picoseconds, femtoseconds, or attoseconds. In one example, the pulse repetition rate of illumination light 104 is less than 100 femtoseconds, and the average optical power is at least 1 watt. Ultrafast pulses achieve high peak power suitable for efficiently inducing SHGs at deeply embedded surface interfaces of the semiconductor sample under measurement.

[0063] As described above, SHG emission occurs at a position of energy combination that is half the wavelength of the illumination wavelength, i.e., twice the frequency. Therefore, the wavelength of the probe illumination beam, for example illumination beam 104, is selected to be relatively long in order to allow deeper penetration into the sample without compromising resolution. In practice, by halving the wavelength inherent to SHG, it becomes possible to double the wavelength of the illumination probe beam without compromising resolution.

[0064] In some embodiments, the wavelength of the probe illumination beam is in a wavelength range beyond the visible light to avoid the inherent transmission limitations of ultraviolet-visible light to typical semiconductor materials such as SiGe and polysilicon.

[0065] In some embodiments, the wavelength of the probe illumination beam 104 exceeds 400 nanometers. In some embodiments, the peak wavelength of the probe illumination beam is in the range of 770 to 810 nanometers. Within this wavelength range, it is possible to penetrate deep into the semiconductor structure. Furthermore, the peak of SHG emission at the Si / SiGe interface is approximately 400 nanometers, and SHG emission occurs at half the wavelength of the probe illumination beam. Therefore, with probe illumination close to 800 nanometers, peak or near-peak SHG emission occurs at the Si / SiGe interface.

[0066] As shown in Figure 2, the computer system 130 communicates an instruction signal 122A to the illumination source 101. In response, the illumination source 101 adjusts the spectral range(s) of the illumination light 104. Generally, the beam 111 incident on the wafer 103 may differ from the illumination light 104 emitted by the illumination source 101 in one or more respects, such as polarization, intensity, size, and shape.

[0067] The beam 104 generated by the illumination source 101 is directed to the beam splitter 105. The beam splitter 105 directs the beam to the objective lens 109. The objective lens 109 focuses the beam 111 onto the incident point 119 on the wafer 103. The incident point 119 is defined (i.e., its shape and size) by projecting the light emitted by the illumination source 101 onto the surface of the wafer 103.

[0068] Generally, the illumination subsystem includes one or more optical elements for controlling the parameters of the illumination beam 111. These parameters include intensity, peak wavelength, or more generally, spectral characteristics, polarization, spot size on wafer 103, and incidence angle on wafer 103. Illumination optical elements include, but are not limited to, one or more polarizers, one or more spectral filters, one or more spatial filters, or one or more apodizers. The illumination optical elements may be placed at any suitable position in the illumination path, including but not limited to the pupil plane and field of view.

[0069] As shown in Figure 2, the TSOM / SHG system 100 includes an arbitrary illumination aperture 124. As shown in Figure 2, the computer system 130 communicates instruction signals 122C to the illumination aperture 124. Accordingly, the illumination aperture 124 adjusts the illumination direction and beam shape provided to the surface of the wafer 103. In one embodiment, the illumination aperture 124 is an assembly that provides various aperture shapes controlled by instruction signals 122C communicated from the computer system 130.

[0070] In some embodiments, the illumination subsystem may further include one or more polarization optics that control the polarization of the illumination light directed onto the specimen. In the embodiment shown in Figure 2, the TSOM / SHG system 100 includes a selectable illumination deflection element 180. In one example, a computer system 130 communicates an instruction signal 122E to the illumination deflection element 180 or higher. Accordingly, the deflection element 180 adjusts the polarization of the illumination light directed onto the surface of the wafer 103.

[0071] In some embodiments, a probe beam 104 generated by a pulsed laser-based illumination source is converted to linearly polarized, circularly polarized, or elliptically polarized light by a deflection element 180. The polarized illumination light is focused onto the wafer 103. In some embodiments, the deflection element 180 may be configured to generate radially and azimuthally polarized light directed onto the wafer 103.

[0072] As shown in Figure 2, the TSOM / SHG system 100 includes an optional illumination power attenuator 102 that controls the illumination power supplied to the wafer 103. In some other embodiments, the illumination power density attenuator 102 is a beam shaping element that resizes the illumination spot 119 to reduce the density of illumination power supplied to the wafer 103. In some other embodiments, a combination of illumination power reduction and beam sizing is used to reduce the illumination power density supplied to the wafer 103. As shown in Figure 2, the computer system 130 communicates a control signal 122B to the illumination power attenuator 102 to control the illumination power based on a three-dimensional image detected by the detector 125.

[0073] In some embodiments, the TSOM / SHG system 100 may include a deflector (not shown) in the illumination path. In one embodiment, the deflector may be an acousto-optic deflector (AOD). In other embodiments, the deflector may include a mechanical scanning assembly, an electronic scanner, a rotating mirror, a polygon-based scanner, a resonant scanner, a piezoelectric scanner, a garbomamirror, or a galvanometer. The deflector scans the ray over the specimen. In some embodiments, the deflector may scan the ray over the specimen at a substantially constant scanning speed.

[0074] In a further embodiment, the TSOM / SHG measurement system includes an external illumination source that generates external illumination on the sample being measured to excite the SHG signal from the sample. The external illumination can cause charge separation at one or more interfaces of the structure under the illumination spot 119, thereby inducing a DC electric field. The DC electric field can enhance SHG emission. Thus, illumination from the external illumination source improves the generation and measurement sensitivity of SHG, for example, at interfaces of inversion-symmetric materials such as silicon.

[0075] Figure 2 shows an external illumination source 120 that generates external illumination light 117 directed onto the wafer 103 via an incident point 119. Examples of external illumination sources 120 include broadband laser sustained plasma sources, lasers, supercontinuum lasers, diode lasers, helium-neon lasers, argon lasers, solid-state lasers, diode-pumped solid-state (DPSS) lasers, xenon arc lamps, gas discharge lamps, LED arrays, and incandescent lamps. The light source may be configured to emit substantially monochromatic or broadband light. In some embodiments, the illumination source may also include one or more spectral filters that can limit the wavelength of light directed onto the specimen. One or more spectral filters may be bandpass filters and / or edge filters and / or notch filters. Illumination may be provided to the specimen over any suitable wavelength range. In some embodiments, the external illumination source 120 is a tunable pulsed laser source configured to generate illumination light 117 characterized by the ultraviolet region, the visible region, or both.

[0076] As shown in Figure 2, the computer system 130 communicates an instruction signal 129 to the external illumination source 120 to control the intensity of the illumination light 117 directed at the wafer 103 to any desired level. In response, the external illumination source 120 generates illumination light 117 according to the instruction signal 129. In some examples, the external illumination source 120 generates illumination light 117 directed at the wafer 103 during measurement. In some examples, the external illumination source 120 generates illumination light 117 in a time-dependent manner, for example, in an on-off sequence at a specified frequency, which modulates the SHG emission and improves the SHG intensity. In some examples, the external illumination source 120 generates illumination light 117 in a time-dependent manner, which modulates the SHG emission and improves measurement sensitivity, reduces the signal-to-noise ratio, etc.

[0077] In a further embodiment, the TSOM / SHG measurement system includes an external field source that generates an external electric field over the entire sample during measurement to excite the SHG signal from the sample. Thus, by changing the electric field induced by the external field source, SHG generation and measurement sensitivity at interfaces of inversion-symmetric materials such as silicon can be improved.

[0078] Figure 2 shows an external electric field source 142 that generates an external electric field 144 over the entire wafer 103 near the incident point 119. In the embodiment shown in Figure 2, the external electric field source 142 is a parallel plate capacitor structure and has an upper plate placed above the wafer 103 and a lower plate (not shown) placed below the wafer 103. As shown in Figure 2, the upper plate includes an aperture 143 that allows illumination light 111 to reach the wafer 103 without interference and SHG emission to reach the objective 109. The voltage source 140 includes voltage nodes 141 and 146. Voltage node 146 is electrically coupled to the upper plate of the parallel plate capacitor, and voltage node 141 is electrically coupled to the lower plate of the parallel plate capacitor.

[0079] As shown in Figure 2, the computer system 130 communicates an instruction signal 145 to the voltage source 140. In response, the voltage source 140 generates a voltage difference between voltage nodes 146 and 141. This induces a corresponding electric field 144 perpendicular to the wafer 103 according to the instruction signal 145. In this way, the electric field strength of the induced electric field 144 is precisely controlled by the computer system 130.

[0080] In some examples, the external field source 142 generates an external field 144 during the period in which the measurement is performed. In some examples, the external field source 142 generates the external field 144 in a time-dependent manner, for example, in an on-off sequence at a specified frequency to modulate the SHG emission to increase the SHG intensity. In some examples, the external field source 142 generates the external field 144 in a time-dependent manner to modulate the SHG emission and achieve improved measurement sensitivity, reduced signal-to-noise ratio, etc.

[0081] An external electric field 144 induces SHG emission at structural interfaces, such as the Si / SiO2 interface. External electric field-induced SHG emission is a third-order nonlinear process that depends on the interaction between the electric field and the incident photon. SHG emission is enhanced by a DC electric field applied to a symmetry-broken surface, due to the contribution of the third-order electric susceptibility induced by the externally applied electric field.

[0082] The TSOM / SHG system 100 includes a focusing optical system 118 that collects SHG light emitted from the wafer 103, as well as light scattered and / or reflected from the wafer 103, and focuses the light onto a detector array 125. The output signal 127 of the detector 125 is communicated to a computer system 130 for processing and determining the presence and location of defects, measurement results, or both.

[0083] The focusing optical system 118 may be a lens, a composite lens, or any suitable lens known in the art. Alternatively, the focusing optical system 118 may be a reflective or partially reflective optical component such as a mirror. Furthermore, although specific collection angles are shown in Figure 2, it should be understood that the focusing optical system can be positioned at any suitable collection angle. The collection angle may vary, for example, based on the angle of incidence and / or the shape characteristics of the specimen.

[0084] Generally, the acquisition subsystem includes one or more acquisition optical elements for controlling the parameters of the acquired light. These parameters include, but are not limited to, intensity, peak wavelength, more generally, spectral characteristics, polarization, acquisition position on wafer 103, or acquisition angle on wafer 103. For example, the acquisition optical elements include, but are not limited to, one or more polarizers, one or more spectral filters, one or more spatial filters, or one or more apodizers. The acquisition optical elements may be placed at any suitable position in the acquisition ray path, including, but not limited to, the pupil plane, the field of view plane, etc.

[0085] As shown in Figure 2, the TSOM / SHG system 100 includes an optional Fourier filter 106. The computer system 130 communicates an instruction signal 122D to the Fourier filter 106. In response, the Fourier filter 106 adjusts its Fourier filtering characteristics (for example, by changing specific Fourier filter elements placed in the acquisition beam path).

[0086] As shown in Figure 2, the TSOM / SHG system 100 includes an arbitrary collection aperture 182. As shown in Figure 2, the computer system 130 communicates an instruction signal 122G to the collection aperture 182. Accordingly, the collection aperture 182 adjusts the amount of light collected from the surface of the wafer 103 that is transmitted to the corresponding detector. In one embodiment, the collection aperture 182 is an assembly that provides various aperture shapes controlled by the instruction signal 122G communicated from the computer system 130.

[0087] In a further embodiment, the TSOM / SHG system 100 includes a spectral filter 183 in the collection optical path. The spectral filter 183 blocks light of wavelengths corresponding to the illumination probe beam 111 and the external pump illumination beam 117, and allows collected light of wavelengths corresponding to the SHG emission in the sample being measured to pass through. In this way, the spectral filter 183 separates the SHG collected light, which is half the wavelength of the probe illumination light, from all other light collected from the sample, such as reflected and scattered light, and allows only the SHG light to pass to the detector 125.

[0088] In some embodiments, the spectral filter 183 includes one or more spectral filters, such as dielectric filters. These include, but are not limited to, bandpass filters that selectively allow SHG light to pass through, bandreject filters that selectively block the spectrum of the incident illumination beam 111, or lowpass filters (e.g., lowpass wavelength filters) that block the spectrum of the incident illumination beam 111 and allow the spectrum of SHG light to pass through. In another example, the spectral filter 183 includes a dispersion element that spectrally disperses the light emitted from the wafer 103, followed by a spatial filter that selectively allows SHG light to pass through.

[0089] In some embodiments, the acquisition subsystem may further include one or more polarizing optics that control the polarization of the acquired light directed to the detector. In the embodiment shown in Figure 2, the TSOM system 100 includes a selectable acquisition deflection element 181 in the acquisition optical path. In one example, the computer system 130 communicates an instruction signal 122F to the acquisition deflection element 181. In response, the acquisition deflection element 181 adjusts the polarization of the acquired light directed to the active surface of the detector 125. The SHG light intensity at the detector depends on the polarization of the illumination probe beam 111 and non-centrosymmetric structural interface information, including material composition and defect properties. In some embodiments, the deflection element 180 and the acquisition deflection element 181 each include two polarizers and one waveplate. Thus, the two polarizers and one waveplate are located in both the illumination optical path and the acquisition optical path. This configuration provides measurement flexibility to optimize the polarizer-waveplate combination in both the illumination optical path and the acquisition optical path, resulting in optimal data acquisition.

[0090] The detector 125 generally functions to convert the collected SHG emission into an electrical signal representing an intensity image of the SHG emission on the active surface of the detector. Generally, the detector 125 can include substantially any photodetector known in the art. However, a specific detector used in one or more embodiments of the invention may be selected based on the desired performance characteristics of the detector, the type of object being inspected, and the lighting configuration. For example, when the amount of light available for inspection is relatively small, using an efficiency-enhancing detector such as a time-delay integral (TDI) camera may improve the system's signal-to-noise ratio and throughput. However, depending on the amount of light available for inspection and the type of inspection performed, other detectors such as charge-coupled device (CCD) cameras, photodiodes, photocells, and photomultiplier tubes (PMTs) may be used. In at least one embodiment of the invention, a photomultiplier tube is used to detect light scattered from the object. Each detector may include only one sensing region, but may in some cases include multiple sensing regions (e.g., a detector array or a multi-node PMT).

[0091] In some embodiments, the detector 125 includes, but is not limited to, a single-pixel device such as a photodetector, avalanche photodiode, or photomultiplier tube. In some embodiments, the detector 125 includes, but is not limited to, a multi-pixel device such as a charge-coupled device (CCD) or a complementary metal-oxide-semiconductor (CMOS) device. In some embodiments, the detector 125 includes a spectrometer suitable for measuring the spectrum of light emitted from the wafer 103 in response to an incident illumination beam 111. Generally, the TSOM / SHG system 100 may include any number or type of detector 125. Thus, the TSOM / SHG system 100 may be suitable for additional measurements beyond SHG measurements, more generally, but not limited to, Raman spectroscopy or photoluminescence.

[0092] In the embodiment shown in Figure 2, the TSOM / SHG system 100 is configured to operate in bright-field imaging mode. For example, in the detected embodiment, the detector 125 generates a bright-field image. As shown in Figure 2, a certain amount of light scattered at a narrow angle from the surface of the wafer 103 is focused by the objective lens 109. This light passes through the objective lens 109 and returns to the beam splitter 105. The beam splitter 105 sends a portion of the light to the focusing optical system 118, which focuses the light onto the detector 125. In this way, the detector array 125 generates a bright-field image. The focusing optical system 118 includes an imaging lens 107 that images the light collected by the objective lens 109 onto the detector array 125. An aperture 182, a Fourier filter 106, or both are located at the back focal plane of the objective lens 109.

[0093] However, generally speaking, the TSOM / SHG system 100 can be configured to operate in various imaging modes, such as brightfield, darkfield, and phase contrast. These various imaging modes can be achieved by using different illumination apertures 124, acquired textures, Fourier filters 106, or combinations thereof. The configuration of the imaging mode, such as illumination direction and imaging acquisition solid angle, can be determined based on the acquired three-dimensional SHG image. These imaging modes are described in more detail in Patent Documents 1 and 2, incorporated herein by reference. In another embodiment, an additional detector is configured to generate a darkfield image by imaging the imaging light acquired at a larger field of view. These imaging modes are described in more detail in Patent Document 3, incorporated herein by reference.

[0094] The TSOM / SHG system 100 further includes various electronic components (not shown) necessary for processing the SHG signals detected by the detector 125. For example, system 100 may include an amplification circuit for receiving output signals from the detector 125 and amplifying those output signals by a predetermined amount, and an analog-to-digital converter (ADC) for converting the amplified signals into a digital format suitable for use within the processor 131. In one embodiment, the processor may be directly coupled to the ADC by a transmission medium. Alternatively, the processor may receive signals from other electronic components coupled to the ADC. Thus, the processor may be indirectly coupled to the ADC by a transmission medium and intervening electronic components.

[0095] In the embodiment shown in Figure 2, the wafer placement system 114 moves the wafer 103 under the beam 111 based on instructions 126 received from the computer system 130. The wafer placement system 114 includes a wafer chuck 108, a motion controller 113, a rotary stage 110, a linear movement stage 112, and a z-linear movement stage 121. The z-linear movement stage 121 is configured to move the wafer 103 in a direction perpendicular to the surface of the wafer 103 (e.g., the z-direction of the reference system 123). The linear movement stage 112 and the rotary stage 110 are configured to move the wafer 103 in a direction parallel to the surface of the wafer 103 (e.g., the x-direction and y-direction of the reference system 123). In some other embodiments, the wafer 103 moves in a plane direction (e.g., the x-direction and y-direction) by the coordinated motion of a plurality of linear movement stages.

[0096] The wafer 103 is supported on the wafer chuck 108. In some embodiments, the wafer 103 is positioned such that its geometric center substantially coincides with the axis of rotation of the rotating stage 110. In this way, the rotating stage 110 rotates the wafer 103 around its geometric center at a specified angular velocity ω within the tolerance. The linear moving stage 112 moves the wafer 103 at a predetermined speed V TThe wafer 103 is moved linearly in a direction substantially perpendicular to the rotation axis of the rotary stage 110. The motion controller 113 coordinates the rotation of the wafer 103 by the rotary stage 110 and the linear movement of the wafer 103 by the linear movement stage 112 to achieve a desired in-plane scanning operation of the wafer 103 within the TSOM / SHG system 100. The motion controller 113 also adjusts the movement of the wafer 103 by the linear movement stage 121 to achieve a desired out-of-plane scanning operation of the wafer 103 within the TSOM system 100. In some embodiments, the linear movement stage 121 is driven by a piezoelectric actuator.

[0097] The wafer 103 can be positioned in various modes relative to the optical subsystem of the TSOM / SHG system 100. In inspection mode, the wafer 103 is repeatedly scanned laterally (e.g., in the X and Y directions) at different z positions. In some of these embodiments, the entire wafer 103 is scanned by the incident illumination beam 111 by controlling the position, orientation, or both of one or more illumination optical elements, such as a Garbo mirror coupled to an fθ lens.

[0098] In some examples, wafer 103 is scanned through a layered structure at two or more different z-positions corresponding to two or more depths (e.g., distance below the wafer surface). In measurement mode, wafer 103 is positioned at fixed positions in the x and y directions while being scanned in the z direction. In this way, a three-dimensional image is generated based on measurement data across the depth range within the structure being measured at the fixed lateral position of wafer 103. Measurement mode is typically used to perform a more detailed investigation of defects (e.g., higher image resolution, higher depth of field resolution, or both).

[0099] In some embodiments, the wafer is moved to a number of different z-positions relative to the focal plane of the TSOM / SHG system in order to image different depths of the wafer stack. In some other embodiments, the position of the focal plane of the TSOM / SHG system is optically adjusted relative to the wafer in order to image different depths of the wafer stack. In some embodiments, an objective, e.g., objective 109, and any focal lenses positioned in the illumination and acquisition paths, e.g., illumination focal lens 147, are moved to adjust the position of the focal plane of the TSOM / SHG system relative to the wafer 103. In some embodiments, a deformable mirror may be used to adjust the position of the focal plane of the TSOM / SHG system relative to the wafer 103.

[0100] Generally, the optical subsystem 140, which includes both an illumination subsystem and an acquisition subsystem, generates focused optical images at each of several focal planes located at several different depths of the structure being measured (e.g., a vertically stacked structure). Alignment of the focal planes of the optical subsystems at each different depth is achieved by optical adjustments that move the focal planes in the z direction, position the detectors in the z direction, or both. One or more detectors detect the light collected at each of the several different depths and generate several output signals indicating the amount of light collected at each of the several different depths. The images collected at each focal position form a three-dimensional volumetric image of the thick semiconductor structure measured at two lateral dimensions (e.g., parallel to the wafer surface) and several different depths (i.e., different Z positions).

[0101] The TSOM / SHG system 100 generates a three-dimensional image of a thick semiconductor structure from volumes measured in two lateral dimensions (e.g., parallel to the wafer surface) and a depth dimension (e.g., perpendicular to the wafer surface). In the embodiment shown in Figure 2, the computer system 130 places the outputs from one or more measurement channels (e.g., from the detector 125) into a dataset corresponding to the measured volume.

[0102] In one measurement example, a series of images are acquired at the same (x,y) position at different wafer locations within the focal plane of the TSOM / SHG system. In this example, the computer system 130 generates a three-dimensional image of the measured volume by assembling a stack of a series of two-dimensional images acquired for each different focus offset. The focus offset is the relative distance between the most reflective surface of the specimen and the focal plane of the TSOM / SHG system. Generally, the parameters scanned are not limited to the focus offset. In other examples, the sensor's axial position, spectral band, illumination direction, etc., can be scanned to form a three-dimensional defect image. In some embodiments, the computer system 130 generates defect images of four dimensions or more. In one example, both the focus offset and illumination direction are scanned for a specified (x,y) position. In one example, the computer system 130 generates a four-dimensional image of the measured volume by assembling a series of two-dimensional images acquired for each different focus offset and illumination angle into a fourth-order tensor. In some examples, a series of images are collected for a predefined set of focus offsets without changing the illumination intensity and other system parameters.

[0103] In one inspection example, a series of images are acquired at various (x,y) positions across multiple wafer locations within the focal plane of the TSOM / SHG system. To generate a three-dimensional image, it is necessary to minimize image misalignment between different focus offsets. In some examples, this is achieved by acquiring data on a wafer stage precisely positioned for different depth measurements. However, this method can significantly reduce throughput. In some other examples, images corresponding to the same lateral position at different focus offsets are aligned using alignment targets after data acquisition.

[0104] Generally, three-dimensional images are processed by algorithms to identify and classify defects or to estimate the values ​​of one or more target parameters that characterize the structure being measured. In some examples, the processor 131 is configured to detect and classify defects from three-dimensional images. The processor may include any suitable processor known in the art. Furthermore, the processor may be configured to use any suitable defect detection and classification algorithm or method known in the art. For example, the processor may utilize comparison of the target image to a reference image. This is, for example, a comparison of the target image to a wafer target image, or to another wafer reference image of the target image, or to a simulated image of the target image. Examples include library matching with a simulated image library, three-dimensional filtering, clustering algorithms such as principal component analysis or spectral clustering, thresholding algorithms, deep learning algorithms, or other suitable algorithms. This allows for the detection and classification of defects based on features extracted from the target image, deep learning algorithms, or other suitable algorithms for directly detecting and classifying defects from the target image. As a non-limiting illustration, any of the defects shown in Figures 1B to 1E can be identified and classified based on appropriate techniques using images of the SHG optical signal captured by the TSOM system.

[0105] In some embodiments, one or more reference structures are illuminated by an incident illumination beam 111, and SHG light is collected accordingly. The collected images of one or more SHG lights (TSOM / SHG images) are detected by a detector 125. The reference structure comprises one or more inversion-symmetric materials, and the collected SHG light is associated with one or more interfaces between the one or more inversion-symmetric materials and additional features of the one or more reference structures. In some embodiments, the reference structure is known to be free of symmetric defects. In some embodiments, the reference image is generated based on a combination of multiple images of one or more reference structures on the same or different wafers, for example, an average.

[0106] Furthermore, one or more target structures are illuminated by the incident illumination beam 111, and SHG light is collected accordingly. Images of the collected SHG light are detected by the detector 125. The target structures include one or more inversion-symmetric materials, and the collected SHG light is associated with one or more interfaces between the one or more inversion-symmetric materials and one or more additional features of the target structures. Generally, one or more target structures are substantially identical in structure to one or more reference structures. That is, the differences between the target and reference structures are due to process differences, not nominal design differences.

[0107] In some embodiments, defects within the target structure are detected by comparing a target TSOM / SHG image with a reference TSOM / SHG image. In one example, defects are detected based on the difference between the reference image and the target image. The reference image and the target TSOM / SHG image may differ based on the acceptable variation between the reference image and the target structure within the manufacturing tolerance. Alternatively, defects within the target structure may be identified by measurable differences between the reference image and the target image.

[0108] In other examples, defects are identified by comparing the target TSOM / SHG image with simulated three-dimensional images of one or more defects. For example, a computer system 130 performs a rigorous coupled-wave analysis (RCWA) to simulate the measured defect response. This analysis may be performed recursively to minimize the error between the measured and simulated responses and to identify the defects.

[0109] In several other examples, a measurement library is generated that matches measurement data with defects measured by a trusted reference measurement system. In one example, the trusted reference measurement system is defect verification performed after focal ion beam etching of the specimen under consideration. Once the library is generated, defects relevant to subsequent measurements are estimated based on the library matching.

[0110] In a further embodiment, the three-dimensional image is filtered before defect analysis to improve the signal-to-noise ratio (SNR). In some examples, a computer system analyzes the assembled three-dimensional image using a three-dimensional digital filter or other appropriate numerical technique to detect the intrinsic three-dimensional structure resulting from the defect. This effectively improves the SNR of the defect, allowing for more effective isolation of the target defect from noise and false influences.

[0111] In some embodiments, defects are classified based on difference images obtained from TSOM / SHG images generated under different measurement conditions, i.e., different optical modes. In some examples, TSOM / SHG images are generated using different characteristics 111 of the incident illumination beam, such as polarization, wavelength, and angle of incidence; different characteristics of the external illumination 117, such as different intensity levels; and different characteristics of the external electric field 144, such as different intensities.

[0112] Different defect types affect the generation of SHG light under different measurement conditions. In these embodiments, these differences are used to detect and classify defects. Furthermore, difference images can be used with any classification technique, including the use of design data, defect depth information, pattern recognition techniques, machine learning techniques, or a combination thereof.

[0113] In some examples, the difference image may be generated by subtracting a first TSOM / SHG image generated with incident illumination light 111 of a first wavelength from a second TSOM / SHG image generated with incident illumination light 111 of a second wavelength. Furthermore, the reference difference image and the target difference image may be generated by subtracting the first and second reference TSOM / SHG images, and subtracting the first and second target TSOM / SHG images, respectively. In this way, as described above, defects can be identified based on a comparison of the reference image and the target difference image.

[0114] In some embodiments, defects are classified based on design data of the structure under inspection. For example, the location of a defect in a target TSOM / SHG image, or in a comparison image between a reference image and the target TSOM / SHG image, may be associated with specific features based on design data that may include the intended layout in the relevant process step in which the image was generated. Such information may then be used in the relevant process step to classify the defect based on a known defect mechanism within the relevant region. Furthermore, depth information provided based on the wavelength or combination of wavelengths used to generate the image as described herein may be used to further isolate the location of the defect in three dimensions to facilitate classification (e.g., based on a known defect mechanism within the relevant region in the relevant process step).

[0115] In some embodiments, defects are classified based on patterns in the target TSOM / SHG image (e.g., pattern recognition) or on comparison images of a reference and the target TSOM / SHG image. For example, multiple defects of different types may appear in a common region of the target TSOM / SHG image. Such defects may result in different variations in the SHG light emitted from adjacent inversion-symmetric material interfaces. More specifically, different types of defects may have different effects on the electric dipoles involved in the generation of SHG light at the interface, based on differences in the surface structure of the interface, the composition of the associated materials, the thickness of the materials, etc. As a result, different types of defects may generate identifiable patterns suitable for classification. Furthermore, when multiple wavelengths are used as described herein, different types of defects may generate patterns in one or more difference images, which can be used for defect classification.

[0116] In some embodiments, defects are classified using supervised or unsupervised machine learning techniques. For example, a supervised machine learning algorithm can be trained on any combination of reference TSOM / SHG images, target TSOM / SHG images of samples with known defects, comparison images (e.g., images generated by subtracting the reference and test SHG images), and difference images associated with different measurement parameter settings. The trained supervised machine learning algorithm can then classify defects identified in the new test structure using corresponding input images of the same type used for training but generated from the new test structure. As another example, defects may be directly identified by an unsupervised machine learning technique based on an analysis of such inputs generated from any test structure.

[0117] TSOM images (TSOM / SHG images) of SHG signals emitted from deep and complex semiconductor structures minimize interference from the matrix and background because the measured signal captures structural details only at interfaces, such as around the nanosheet GAA transistor channel structure. Since many defects in GAA nanosheet transistor manufacturing are close to the FET channel region, TSOM / SHG images can be used to monitor devices during manufacturing at various process steps and to implement yield control of the manufacturing process based on TSOM / SHG measurement results to control the performance and yield of FET transistors.

[0118] Figure 1A shows the SHG emission regions 210A to C associated with the three channels of the GAA nanosheet transistor. Differences in the interface or surface due to the presence or absence of etching residual defects 21 after channel release etching are reflected in different TSOM / SHG images collected from the GAA nanosheet device structure 10 shown in Figure 1A and the GAA nanosheet device structure 20 shown in Figure 1B. Similarly, differences in the interface or surface due to the presence or absence of channel erosion defects 31 after channel release etching are reflected in different TSOM / SHG images collected from the GAA nanosheet device structure 10 shown in Figure 1A and the GAA nanosheet device structure 30 shown in Figure 1C. Similarly, differences in the interface or surface due to the presence or absence of surface roughness defects 41 after channel release etching are reflected in different TSOM / SHG images collected from the GAA nanosheet device structure 10 shown in Figure 1A and the GAA nanosheet device structure 40 shown in Figure 1D. Similarly, the differences in interfaces or surfaces due to the presence or absence of void defects 51 after the deposition of the high dielectric constant dielectric layers 51A to D and the metal gate layers 52A to D are reflected in the different TSOM / SHG images collected from the GAA nanosheet device structure 10 shown in Figure 1A and the GAA nanosheet device structure 50 shown in Figure 1E.

[0119] In general, defects in other nanosheet GAA device process steps, including SiGe recess etching and source / drain EPI growth, can be detected based on TSOM / SHG imaging. For EPI growth processes, detectable defects include over-inner spacer defects, source / drain underfill, void and non-merged defects, and HKMG void defects. Defects in the ferroelectric layer of ferroelectric FETs and the two-dimensional layer of two-dimensional FET devices can also be detected based on TSOM / SHG imaging.

[0120] Generally, defects are detected from three-dimensional images by applying a defect detection algorithm. In some embodiments, defect detection is performed directly from image data generated by the TSOM / SHG system 100. In some embodiments, one or more feature vectors are extracted from the collected image data, and defect detection is performed based on the measured feature vectors. Generally, a feature vector is an n-th order vector of numerical features representing an object (e.g., a defect in the object, a nominal structure, etc.). In some examples, the defect detection algorithm includes one or more selectable thresholds to adjust the sensitivity of the defect detection algorithm. If a very restrictive threshold is selected, the defect detection algorithm will detect fewer target defects from the set of three-dimensional images. If a more tolerant threshold is selected, the defect detection algorithm will detect more target defects from the same set of three-dimensional images. If too few defects are detected, there is a high possibility that actual defects will be missed, and if too many defects are detected, many false (e.g., ill-advised) defects will be captured. Therefore, optimized measurement recipes tailored to specific measurement applications also include the selection of detection algorithm thresholds that maximize the detection rate of actual defects while simultaneously minimizing the detection rate of false (i.e., sham) defects.

[0121] As explained with respect to Figure 2, the computer system 130 generates and communicates G from the instruction signal 122A so that the illumination power, illumination palette, collection wavelength, spectral band, Fourier filter, illumination polarization, collection polarization, or any combination thereof are selected according to the specified optical mode. Furthermore, measurement systems such as the TSOM / SHG system 100 include other selectable optical system settings such as the angle of incidence and azimuth angle. Each different combination of optical system settings is referred to as a separate optical mode of the TSOM / SHG system 100. In particular, the images produced by the TSOM / SHG system are sensitive to polarization, e.g., single polarization, multiple polarizations, or difference images between different polarizations, waveplate selection, probe beam wavelength, use of external illumination light, use of external electric field, etc.

[0122] In practice, TSOM / SHG measurement systems, such as the TSOM / SHG system 100 described herein, offer over 10,000 different optical modes, and it is necessary to select one mode from thousands of mode candidates to achieve one or more performance objectives. Examples of performance objectives include, but are not limited to, minimizing the response of nominal structures in a 3D image, enhancing the response of defect signals in a 3D image, minimizing the response of wafer noise or pseudo-signals in a 3D image, distinguishing defect responses from wafer noise or pseudo-signals in a 3D image, improving the accuracy of the physical location of defects estimated from a 3D image, or any combination thereof. Therefore, a measurement recipe tailored to a specific measurement application includes the selection of the optimal optical mode.

[0123] As mentioned above, optimizing the measurement recipe for three-dimensional optical systems such as the TSOM / SHG system 100 involves selecting the optimal optical mode from thousands of system configuration candidates and selecting the detection threshold. Because DOIs can exist throughout the entire depth of the three-dimensional semiconductor structure, the amount of available image data is enormous (i.e., not just two-dimensional images, but three-dimensional images), and defect verification is extremely time-consuming (i.e., wafer de-processing is required to verify defects), thus presenting significant practical challenges regarding defect detection and recipe optimization.

[0124] In one embodiment, the three-dimensional volume of the semiconductor wafer subject to defect detection and verification is reduced by saving images associated with a subset of the total depth of the semiconductor structure being measured. Consequently, the amount of image data that needs to be collected and analyzed as part of the measurement recipe optimization process is reduced.

[0125] In some cases, the TSOM / SHG measurement system receives values ​​for one or more parameters corresponding to the initial TSOM / SHG measurement recipe.

[0126] In the embodiment shown in Figure 2, the computer system 130 receives information regarding the measurement application 136 under consideration from a user input source 135. Typically, the user input source 135 is an entity such as a user or operator that has knowledge of the structure under inspection and the expected defects. In non-limiting examples, the structural information 136 may include the expected stack depth of the defect, the wafer-level characteristics of the defect, and the refractive index of the three-dimensional stack. In one embodiment, the TSOM system 100 includes peripheral devices useful for receiving input from the operator (e.g., a keyboard, mouse, touchscreen, communication port, etc.) for the structural information 136 to be communicated from the user to the TSOM / SHG system 100.

[0127] The user further communicates an initial set of optical modes for the TSOM / SHG system 100. The user of the TSOM / SHG system 100 typically performs pre-modeling or leverages past experience to arrive at the initial set of optical modes for the TSOM / SHG system 100 that is most likely to yield the best measurement results. Typically, the initial set of optical modes includes parameter values ​​corresponding to dozens of different optical modes, which is far fewer than thousands of optical mode candidates. In some examples, the user communicates one or more initial focus levels to the TSOM / SHG system 100. These one or more initial focus levels include the focus level where the target defect should be located. In some examples, the user communicates parameter values ​​characterizing an initial measurement recipe, including characteristics of the illumination beam 104, such as intensity, wavelength, polarization, spot size on wafer 103, and incidence angle on wafer 103. In some examples, the user communicates parameter values ​​characterizing an initial measurement recipe, including characteristics of the collected SHG light 148, such as polarization, not limited to these. In some examples, the user communicates parameter values ​​characterizing the initial measurement recipe, including the characteristics of the detector 125, non-limitingly such as gain settings. In some examples, the user communicates parameter values ​​characterizing the initial measurement recipe, including the characteristics of the spectral filter 183, non-limitingly such as cutoff wavelength, bandpass wavelength, etc. In some examples, the user communicates parameter values ​​characterizing the initial measurement recipe, including the characteristics of the external illumination source 120, non-limitingly such as intensity, wavelength, polarization, spot size of wafer 103, incident angle of wafer 103, etc. In some examples, the user communicates parameter values ​​characterizing the initial measurement recipe, including the characteristics of the external electric field source 142, non-limitingly such as electric field strength, angle relative to wafer 103, etc.

[0128] In some cases, a TSOM / SHG measurement system performs measurements of one or more semiconductor structures based on the values ​​of one or more parameters corresponding to an initial TSOM / SHG measurement recipe.

[0129] In response to user input 136, the TSOM / SHG system 100 performs measurements of the wafer 103 at each of the initial sets of optical modes and one or more initial focus levels. Typically, the inspection is performed in scanning mode, and a wide area of ​​the wafer (e.g., the entire wafer) is inspected at each of the one or more initial focus levels. The threshold of the defect detection algorithm employed during the initial inspection is set to a very tolerant value that identifies many defects (i.e., both actual and false defects).

[0130] After the initial inspection is performed, the computer system 130 selects some of the most likely defects identified in the initial inspection. The most likely defects are those of interest that best match the expected defects provided by the user of the TSOM / SHG system 100. The TSOM / SHG system 100 performs a through-focus review of the selected target defects by positioning the wafer 103 against the optical inspection subsystem 140 so that the selected target defects are within the field of view of the TSOM / SHG system 100. A series of measurements are performed at multiple focus levels across the entire structure being measured. Based on the results of the through-focus review, the computer system 130 determines one or more focal planes or focal ranges that best capture the target defects. In some examples, one or more focal planes or focal ranges are determined based on the best match between the measured defect markings (e.g., images or feature vectors) and the expected defect markings.

[0131] After determining one or more focal planes or focal ranges, the TSOM / SHG system 100 records image patches associated with the defect locations identified in each of the initial inspections at one or more focal planes or focal ranges, rather than the entire depth of the structure. In some examples, 100 million or more defect locations are imaged and recorded at multiple focal levels. In this way, the amount of data recorded in relation to defect detection is limited to a subset of the depth. The recorded data is used in the subsequent defect verification and recipe optimization process. By limiting the amount of data recorded, the subsequent defect verification and recipe optimization process is significantly simplified.

[0132] In a further embodiment, the number of optical modes considered is reduced based on a comparison of one or more measured wafer-level defect markings with one or more expected wafer-level defect markings. For example, after performing an initial inspection, the computer system 130 selects several of the most likely optical modes for further consideration (e.g., five or fewer optical modes). For example, the computer system 130 varies the threshold of the defect detection algorithm for each optical inspection mode. The computer system 130 varies the threshold value to best match the measured wafer-level defect markings with the expected wafer-level defect markings for each optical mode. In an example, wafer-level defect markings are wafer defect maps that show areas of the wafer area where defects are more or less concentrated. Generally, wafer-level defect markings include any markings of wafer defects expressed across the entire wafer area under inspection.

[0133] The optical mode that best matches the expected defect indication is selected for further consideration, and the other optical modes are discarded. Thus, as mentioned above, the number of modes selected for recording is reduced. Consequently, the amount of inspection data considered during subsequent defect verification and recipe optimization processes is further reduced. In some examples, based on wafer-level indication analysis, five or fewer optical modes are selected for further consideration.

[0134] In yet another embodiment, the number of optical modes considered is reduced based on the measured defect signal-to-noise ratio. For example, after performing initial inspection and wafer-level marking matching, the computer system 130 selects several of the most likely optical modes for further consideration (e.g., three or fewer optical modes). For example, the computer system 130 analyzes the signal-to-noise ratio associated with each selected optical mode at one or more focal planes or focal levels. The computer system 130 selects the optical mode with the best signal-to-noise ratio for further consideration, and the other optical modes are discarded. Thus, as mentioned above, the number of modes selected for recording is reduced. Consequently, the amount of inspection data considered in subsequent defect verification and recipe optimization processes is further reduced. In some examples, three or fewer optical modes are selected for further consideration based on wafer-level marking analysis.

[0135] In some cases, the TSOM / SHG measurement system performs measurements of one or more semiconductor structures based on the values ​​of one or more parameters corresponding to the selected TSOM / SHG measurement recipe.

[0136] In some cases, defects are identified based on referencing and testing TSOM / SHG images generated using selected values ​​for TSOM / SHG system parameters associated with each selected TSOM / SHG measurement recipe. In other cases, defects are identified, as described above, based on differences in images associated with different optical modes, comparison with simulated images, library matching, etc.

[0137] In yet another embodiment, the number of optical modes to be considered is reduced based on the SEM review. The computer system 130 receives the defect markings verified by the SEM review tool and selects the optical mode with the highest detection rate of verified defects and the lowest detection rate of false defects for further consideration. Other optical modes are discarded. In this way, as described above, the number of modes selected for recording is reduced. Therefore, the amount of inspection data considered during the subsequent defect verification and recipe optimization process is further reduced. In some examples, three or fewer optical modes are selected for further consideration based on wafer-level marking analysis.

[0138] In some cases, the reference system performs measurements of one or more semiconductor structures measured by the TSOM / SHG measurement system to verify whether defects have been correctly identified. In some cases, tunneling electron microscopy (TEM), scanning electron microscope (SEM), TSOM microscope, and optical bright-field imaging microscope are used to verify the identified defects.

[0139] In some cases, the difference between defects identified by the TSOM / SHG measurement system, associated with each different measurement recipe, is compared to the defects verified by a reference measurement system. If the difference is below a predetermined threshold, for example, a threshold for the percentage of correctly identified defects, the measurement recipe is adopted for inline production measurement. If the difference is greater than the predetermined threshold, the measurement recipe is updated, and the TSOM / SHG measurement is performed according to the updated measurement recipe, and the results are evaluated.

[0140] In another embodiment, the verified defect images / features are mapped to the corresponding defects identified by the TSOM / SHG system 100. The verified defects and recorded 3D images are used to train a mock-up filter and further optimize the measurement recipe.

[0141] Figure 4 is a simplified schematic diagram of one embodiment of a system 150 for defect detection and measurement recipe optimization for inspection of three-dimensional semiconductor structures. System 150 includes the TSOM / SHG system 100, defect verification tool 151, and computer system 160, as described with reference to Figure 2. In some embodiments, tasks described herein as being performed by computer system 160 are performed by computer system 130 or another computer system.

[0142] In some embodiments, the defect verification tool 151 is an electron beam-based analysis tool. In some other embodiments, the defect verification tool 151 is an X-ray-based analysis tool. In these embodiments, a material removal tool may not be necessary to visualize buried defects for the X-ray-based analysis tool. Therefore, the associated material removal tool is optional.

[0143] In some examples, defect verification is performed by delaminating the wafer 103 and inspecting the exposed defects with the TSOM / SHG system 100. In these examples, a separate defect verification tool 151 may not be necessary. In some embodiments, defect verification tools, such as SEM review tools, may be integrated with the TSOM / SHG system 100 as a single wafer processing tool, separated into different wafer processing systems, or in any combination.

[0144] Computer system 130 coordinates the inspection process and performs analysis, data processing, and communication tasks. Similarly, computer system 160 coordinates the material removal and verification process, performs analysis, and performs data processing and communication tasks.

[0145] Defect verification can be performed in various ways. In some embodiments, voltage contrast testing is performed for defect verification. In these embodiments, wafers are decorated according to a small sample plan, and voltage contrast measurements are performed on the decorated wafers using a voltage contrast testing tool.

[0146] In some other embodiments, after wafer manufacturing is complete, a bitmap test is performed on the finished wafer to verify for defects.

[0147] In some other embodiments, the wafer is delaminated to remove layers of the multilayer structure under consideration. Delamination can be achieved by chemical, mechanical, or both processes. In one example, a focused ion beam (FIB) tool is used to remove material from the surface of the wafer. The wafer is delaminated until embedded defects are located on or near the surface of the wafer and can be effectively imaged by a defect verification tool 151, such as an SEM review tool or a TSOM / SHG system 100. The location of the defects and associated defect images 152 related to the defect verification measurement are stored in memory (e.g., memory 162 on a board computer system 160). In some embodiments, the defect information is stored in the format of a KLA result file (KLARF). A KLARF file is a flat ASCII file generated by the defect verification tool 151. The same KLARF file format is also used to store defect information from the TSOM / SHG system 100.

[0148] In a further embodiment, defect information 141 related to defects identified by the TSOM / SHG system 100 as part of defect detection is communicated to a computer system 160. The computer system 160 samples the identified defects to generate a diverse set of DOIs 153 and communicates them to a defect verification tool 151. In some embodiments, the computer system 160 groups the defects identified by the TSOM / SHG system 100 during defect detection (e.g., more than 100 million DOIs), selects a small number of defects from each group to generate a diverse set of DOIs 153. The diverse set of DOIs 153 is stored in memory (e.g., memory 162 on the computer system 160).

[0149] Defect verification measurements are performed on a diverse set of DOIs. The location of defects and associated defect images from the defect verification measurements are stored in memory (e.g., memory 162 on computer system 160). In some embodiments, the defect information associated with the diverse set of DOIs is also stored in KLARF file format.

[0150] Defect verification data from diverse sets of DOIs, other verified defect sets, or combinations thereof is mapped to stored through-focus defect image patches and their corresponding feature vectors. The defect verification data, along with the corresponding through-focus defect image patches and their corresponding feature vectors, is used to train a pseudo-removal filter.

[0151] In one example, computer system 160 trains a through-focus image-based machine learning network to remove pseudo-defects. In some of these examples, the machine learning network is trained based on defective images. In non-limiting examples, the appropriate machine learning network may be implemented as a neural network, support vector machine model, decision tree model, etc.

[0152] In another example, computer system 160 trains a focus feature-based autoclassifier to remove false defects. In some of these examples, the autoclassifier is feature-based rather than image-based. As a non-restrictive example, a suitable through-focus feature-based autoclassifier can be implemented as a pre-trained random forest algorithm, for example.

[0153] In another example, computer system 160 implements a rule-based tree classifier to eliminate false defects. In some of these examples, the rule-based tree classifier is feature-based rather than image-based. As an unrestricted example, a suitable rule-based tree classifier is implemented based on manually generated rules.

[0154] A trained pseudo-rejection filter 142 is communicated to the TSOM / SHG system 100 and applied to the stored defect images associated with each system optical mode under consideration. In this way, defect detection is emulated using through-focus defect events recorded during defect detection. The detection threshold associated with each optical mode is adjusted to achieve a desired pseudo-rejection rate. In one example, the detection threshold associated with each optical mode is adjusted to achieve a pseudo-rejection rate of approximately 30%, and the optical mode that achieves the best defect marking matching and actual defect capture rate is selected for implementation as a production measurement recipe for the measurement application under consideration. In the embodiment shown in Figure 2, the TSOM / SHG system 100 implements the pseudo-rejection filter 142 and the selected manufacturing measurement recipe to identify and classify defects based on analysis of a three-dimensional image of a thick semiconductor structure in a manufacturing setting.

[0155] Generally, computer systems 130 and 160 are configured to detect and classify defects based on feature vectors derived from through-focus images of SHG light or TSOM / SHG images. Computer systems 130 and 160 may include any suitable processor(s) known in the art. Furthermore, computer systems 130 and 160 may be configured to use suitable defect detection algorithms or methods known in the art. For example, computer systems 130 and 160 may use comparison of the die with a database or a thresholding algorithm to detect defects in a specimen.

[0156] Furthermore, the TSOM / SHG system 100 may include peripheral devices useful for receiving input from the operator (e.g., keyboard, mouse, touchscreen, etc.) and displaying output to the operator (e.g., display monitor). Input instructions from the operator may be used by the computer system 130 to adjust thresholds used to control the lighting power. The resulting power levels may be graphically displayed to the operator on the display monitor.

[0157] Figure 5 shows a flowchart of an exemplary method 300 useful for detecting buried defects and measuring buried structures based on three-dimensional images of SHG light. In some non-limiting examples, TSOM / SHG systems 100 and 200, described with reference to Figures 2 and 3, are configured to perform method 300. However, in general, the performance of method 300 is not limited by the specific embodiments described herein.

[0158] In block 301, according to a first optical mode, a first quantity of probe illumination light is supplied to the semiconductor wafer at each of the multiple focal planes within the complex three-dimensional structure arranged on the semiconductor wafer. The first quantity of probe illumination light excites second harmonic generation (SHG) emission around material interfaces embedded deep within the complex three-dimensional structure.

[0159] In block 302, the first quantity of SHG emission from the complex three-dimensional structure to the detector at each of the multiple focal planes is imaged.

[0160] In block 303, the first imaged SHG light is detected at each of several focal planes.

[0161] In block 304, a three-dimensional volumetric image of a complex three-dimensional structure is generated based on the first SHG light imaged at each of the multiple focal planes.

[0162] In block 305, buried defects are identified or the values ​​of parameters of interest characterizing the buried structure are estimated, at least partially based on the three-dimensional volumetric image.

[0163] It should be recognized that the various steps described herein may be performed by a single computer system 130 or by multiple computer systems 130. Furthermore, different subsystems of the TSOM / SHG systems 100 and 200, such as the sample placement system 114, may include computer systems suitable for performing at least some of the steps described herein. Therefore, the foregoing description should not be construed as limiting the invention, but is merely illustrative. Furthermore, one or more computer systems 130 may be configured to perform any other step(s) of any of the method embodiments described herein.

[0164] Furthermore, the computer system 130 can be communicatively coupled to the detector 125, illumination source, illumination optics, collection optics, and sample placement system 114 in any manner known in the art. For example, one or more computer systems 130 may be coupled to computer systems associated with the detector 125, illumination source, illumination optics, collection optics, and sample placement system 114, respectively. In another example, any of the detector 125, illumination source, illumination optics, collection optics, and sample placement system 114 may be directly controlled by a single computer system coupled to the computer system 130.

[0165] The computer system 130 of the TSOM / SHG systems 100 and 200 may be configured to receive and / or acquire data or information from subsystems of the system (e.g., detector 125, illumination source, illumination optics, collection optics, sample placement system 114, etc.) via a transmission medium which may include wired and / or wireless components. In this way, the transmission medium may function as a data link between the computer system 130 and other subsystems of systems 100 and 200.

[0166] The computer system 130 of the TSOM / SHG systems 100 and 200 may be configured to receive and / or acquire data or information (e.g., measurement results, modeling inputs, modeling results, etc.) from other systems via a transmission medium which may include wired and / or wireless components. Thus, the transmission medium may function as a data link between the computer system 130 and other systems (e.g., memory onboard measurement system 100, external memory, or external system). For example, the computer system 130 may be configured to receive measurement data (e.g., output signals 127) from a storage medium (i.e., memory 132) via the data link. For example, image results obtained using the detector 125 may be stored in a persistent or semi-persistent memory device (e.g., memory 132). In this regard, measurement results may be imported from onboard memory or an external memory system. Furthermore, the computer system 130 may transmit data to other systems via the transmission medium. For example, defect locations and sample parameter values ​​determined by the computer system 130 may be stored in a persistent or semi-persistent memory device. In this regard, measurement results may be exported to another system.

[0167] Computer systems 130 and 160 include, but are not limited to, personal computer systems, mainframe computer systems, workstations, image computers, parallel processors, network computer systems, or other devices known in the art. Generally, the term “computer system” can be broadly defined to encompass any device having one or more processors that execute instructions from a memory medium.

[0168] Program instructions 134 implementing the methods described herein may be transmitted via a transmission medium such as a wired, cable, or wireless transmission link. For example, as shown in Figure 2, program instructions stored in memory 132 are transmitted to the processor 131 via bus 133. Program instructions 134 are stored in a computer-readable medium (e.g., memory 132). Examples of computer-readable mediums include read-only memory, random-access memory, magnetic disks, optical disks, solid-state memory, and magnetic tape.

[0169] The TSOM / SHG system 100 includes a processor 131 and a computer-readable amount of memory 132. The processor 131 and the memory 132 can communicate via a bus 133. The memory 132 includes a certain amount of memory 134 that, when executed by the processor 131, stores an amount of program code that causes the processor 131 to perform the detection, classification, and depth estimation functions described herein.

[0170] System 150 includes a processor 161 and a certain amount of computer-readable memory 162. The processor 161 and memory 162 can communicate via a bus 163. Memory 162 includes a certain amount of memory 164 that, when executed by the processor 161, stores a certain amount of program code that causes the processor 161 to perform the defect detection, classification, and depth estimation functions described herein.

[0171] In general, the three-dimensional imaging techniques described herein can be applied during the research and development, production, and mass production phases of semiconductor device manufacturing, and are applicable to any optical image-based measurement techniques. In some embodiments, the TSOM / SH inspection and measurement described herein are implemented as part of manufacturing process tools. Examples of manufacturing process tools include, but are not limited to, lithography exposure tools, deposition tools, implantation tools, and etching tools. Thus, the results of TSOM / SHG measurements are used to control the manufacturing process. For example, TSOM / SHG measurement data collected from one or more targets is transmitted to the manufacturing process equipment. X-ray data is analyzed, and the results are used to adjust the operation of the manufacturing process equipment.

[0172] Regardless of the specific type of manufacturing process, defects must be detected as early as possible in the process at all levels of the multilayer stack. Certain inspection embodiments preferably include the detection of defects throughout the entire stack, including the stack surface, and across various depths of the stack. The thickness of a vertical ONON or OPOP stack during inspection is limited only by the depth of transmission of the illumination light. Transmission through oxide-nitride-oxide-nitrite (ONON) or oxide-polysilicon-oxide-polysilicon (OPOP) stacks is less limited by absorption at longer wavelengths. Therefore, longer illumination wavelengths can be used to effectively inspect extremely deep structures.

[0173] The three-dimensional imaging techniques described herein are applicable to complex vertical stack structures, including but not limited to 3D negative AND (NAND) gate memory devices. While TSOM / SHG systems and techniques are described herein as applicable to specific types of vertical NAND (VNAND) memory structures, it should be understood that embodiments of the present invention are applicable to any suitable 3D or vertical semiconductor structure, such as NAND or NOR memory devices formed using terabit cell array transistors (TCAT), vertical stack array transistors (VSAT), bit cost scalable (BiCST) technology, tubular BiCS (P-BiCS) technology, etc. Vertical direction generally refers to the direction perpendicular to the substrate surface. While specific manufacturing steps, processes, and materials for forming such 3D structures are described, the inspection embodiments can be applied at any point in the manufacturing flow where multiple layers are formed on the substrate, and such layers may include any number and types of materials.

[0174] This specification describes various embodiments of inspection systems or tools that can be used to inspect specimens, or measuring systems or tools that can be used to measure specimens. The term “specimen” is used herein to mean a wafer, reticle, or any other sample that can be inspected or measured for defects, features, or other information known in the art (e.g., the amount of haze or film properties).

[0175] As used herein, the term “wafer” generally refers to a substrate formed of a semiconductor or non-semiconductor material. Examples include, but are not limited to, single-crystal silicon, gallium arsenide, and indium phosphide. Such substrates may be commonly found in and / or processed in semiconductor manufacturing facilities. In some cases, a wafer may consist of only a substrate (i.e., a bare wafer). Alternatively, a wafer may consist of one or more layers of different materials formed on a substrate. The one or more layers formed on a wafer may be “patterned” or “unpatterned.” For example, a wafer may consist of multiple layers having repeatable pattern features.

[0176] A “reticle” can be a reticle at any stage of the reticle manufacturing process, or a finished reticle that is commercially available or not commercially available for use in a semiconductor manufacturing facility. A reticle, or “mask,” is generally defined as a substantially transparent substrate on which a substantially opaque area composed of a pattern is formed. The substrate may include, for example, a glass material such as quartz. A reticle may be placed on a resist-covered wafer so that the pattern on the reticle is transferred to the resist during the exposure step of the lithography process.

[0177] In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored or transmitted as one or more instructions or codes by a machine-readable medium, i.e., a computer-readable medium. A computer-readable medium includes both computer storage media and communication media, including any media that facilitate the transfer of computer programs from one location to another. The storage media may be any available medium accessible by a computer. As a non-limiting example, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM, or other optical disk storage devices, magnetic disk storage devices, or other magnetic storage devices, or any other medium used to carry or store desired program code in the form of instructions or data structures, and accessible by a computer. Furthermore, any connection is appropriately referred to as a computer-readable medium. For example, if software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, or microwave, then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, or microwave are included in the definition of media. As used herein, disks and discs include compact discs (CDs) and discs, laserdiscs (registered trademark) and optical discs, digital multipurpose discs (DVDs) and discs, where a disc optically reproduces data using a laser, while a disk typically reproduces data magnetically. Combinations of the above should also be included within the scope of computer-readable media.

[0178] While specific embodiments are described above for illustrative purposes, the teachings in this patent document have general applicability and are not limited to the specific embodiments described above. In one example, the detector may include a fiber array. In one example, the TSOM / SHG system 100 may include a plurality of light sources (not shown). The light sources may be configured differently or similarly. For example, the light sources may be configured to produce light having different characteristics that can illuminate at the same or different times and at the same or different angles of incidence, and directed onto the wafer in the same or different areas. The light sources may be configured according to any of the embodiments described herein. Furthermore, one of the light sources may be configured according to any of the embodiments described herein, and another light source may be any other light source known in the art. In some embodiments, the TSOM / SHG system may illuminate a wafer simultaneously across a plurality of illumination areas. The plurality of illumination areas may overlap spatially. The plurality of illumination areas may be spatially separated. In some embodiments, the TSOM / SHG system may illuminate a wafer across a plurality of illumination areas at different times. The different illumination areas may overlap in time (i.e., illuminated simultaneously over a period of time). Different illumination areas can be separated in time. Generally, the number of illumination areas is arbitrary, and the size, orientation, and angle of incidence of each illumination area may be equal or different. In yet another example, the TSOM / SHG system 100 may be a scanning spot system with one or more illumination areas that scan independently of the movement of the wafer 103. In some embodiments, the illumination area is scanned in a repeating pattern along a scan line. The scan line may or may not coincide with the scanning motion of the wafer 103. As presented herein, the wafer positioning system 114 realizes the motion of the wafer 103 by coordinating rotational and translational motions, but in yet another example, the wafer positioning system 114 may realize the motion of the wafer 103 by coordinating two translational motions. For example, the wafer positioning system 114 may realize motion along two orthogonal linear axes (e.g., XY motion).

[0179] Therefore, various modifications, adaptations, and combinations of features of the described embodiments can be implemented without departing from the scope of the invention as defined in the claims.

Claims

1. An illumination subsystem that provides a first amount of probe illumination light to a semiconductor wafer at each of a plurality of focal planes in a complex three-dimensional structure arranged on the semiconductor wafer according to a first optical mode, wherein the first amount of probe illumination light excites second harmonic generation (SHG) emission around material interfaces embedded deep within the complex three-dimensional structure, A collection subsystem for imaging a first amount of SHG emission from the complex three-dimensional structure to a detector at each of the plurality of focal planes according to the first optical mode, wherein the detector detects the imaged first SHG light at each of the plurality of focal planes, A computer system configured to generate a three-dimensional volumetric image of the complex three-dimensional structure based on the first SHG light imaged at each of the plurality of focal planes, and to identify buried defects or estimate the values ​​of parameters of interest that characterize the buried structure based at least partially on the three-dimensional volumetric image, A system equipped with these features.

2. The system according to claim 1, further comprising an external illumination source that provides a certain amount of external illumination light to the complex three-dimensional structure arranged on the semiconductor wafer for at least a portion of the time that the illumination subsystem provides the first amount of probe illumination light to the semiconductor wafer at each of the plurality of focal planes.

3. The system according to claim 1, further comprising an external electric field source that generates an external electric field over the complex three-dimensional structure disposed on the semiconductor wafer for at least a portion of the time the illumination subsystem provides the first amount of probe illumination light to the semiconductor wafer at each of the plurality of focal planes.

4. The system according to claim 1, wherein identifying the buried defect involves either comparing the three-dimensional volumetric image with a reference image or comparing the three-dimensional volumetric image with a simulated reference image or a library of simulated reference images.

5. The system according to claim 1, wherein the illumination subsystem provides a second amount of probe illumination light to the semiconductor wafer at each of the plurality of focal planes in the complex three-dimensional structure according to a second optical mode, the acquisition subsystem images a second amount of SHG emission from the complex three-dimensional structure to the detector at each of the plurality of focal planes according to the second optical mode, the detector detects the imaged first SHG light and the imaged second SHG light at each of the plurality of focal planes, and the computer system generates the three-dimensional volume image of the complex three-dimensional structure based on the difference between the imaged first SHG light and the imaged second SHG light at each of the plurality of focal planes.

6. The system according to claim 5, wherein the first optical mode and the second optical mode differ in any of the wavelength of the probe illumination light, the polarization of the probe illumination light, or the angle of incidence of the probe illumination light.

7. The system according to claim 1, wherein the collection subsystem comprises a spectral filter disposed in the collection optical path of the collection subsystem, and the spectral filter separates the SHG emission from the collection light having a wavelength corresponding to the wavelength of the first amount of probe illumination light.

8. The system according to claim 1, wherein the first amount of probe illumination light is provided to the semiconductor wafer at an angle of incidence perpendicular to the top surface of the semiconductor wafer.

9. The system according to claim 1, wherein the first amount of probe illumination light is provided to the semiconductor wafer at an oblique angle of incidence with respect to the top surface of the semiconductor wafer.

10. The system according to claim 1, wherein the complex three-dimensional structure is a gate-all-around (GAA) device structure.

11. The present invention provides a first amount of probe illumination light to a semiconductor wafer at each of a plurality of focal planes in a complex three-dimensional structure arranged on the semiconductor wafer, wherein the first amount of probe illumination light excites second harmonic generation (SHG) emission around material interfaces embedded deep within the complex three-dimensional structure. To image the first amount of SHG emission from the complex three-dimensional structure to the detector at each of the multiple focal planes, To detect the imaged first SHG light at each of the plurality of focal planes, Based on the first SHG light imaged at each of the plurality of focal planes, a three-dimensional volume image of the complex three-dimensional structure is generated. Based at least partially on the aforementioned three-dimensional volumetric image, the buried defect is identified, or the values ​​of parameters of interest characterizing the buried structure are estimated. A method that includes this.

12. The illumination subsystem further includes providing a certain amount of external illumination light to the complex three-dimensional structure arranged on the semiconductor wafer for at least a portion of the time that the illumination subsystem provides the first amount of probe illumination light to the semiconductor wafer at each of the plurality of focal planes, The method according to claim 11, wherein the aforementioned amount of external illumination light is different from the first amount of probe illumination light.

13. The method of claim 11, further comprising generating an external electric field over the complex three-dimensional structure disposed on the semiconductor wafer for at least a portion of the time that the illumination subsystem provides the first amount of probe illumination light to the semiconductor wafer at each of the plurality of focal planes.

14. The method according to claim 11, wherein identifying the buried defect involves either comparing the three-dimensional volumetric image with a reference image or comparing the three-dimensional volumetric image with a simulated reference image or a library of simulated reference images.

15. In accordance with the second optical mode, a second amount of probe illumination light is provided to the semiconductor wafer at each of the plurality of focal planes in the complex three-dimensional structure, In accordance with the second optical mode, the second amount of SHG emission from the complex three-dimensional structure to the detector at each of the multiple focal planes is imaged, The method further includes detecting the imaged first SHG light and the imaged second SHG light at each of the plurality of focal planes, The method according to claim 11, wherein generating the three-dimensional volume image of the complex three-dimensional structure is based on the difference between the imaged first SHG light and the imaged second SHG light at each of the plurality of focal planes.

16. The method according to claim 15, wherein the first optical mode and the second optical mode differ in any of the wavelength of the probe illumination light, the polarization of the probe illumination light, and the angle of incidence of the probe illumination light.

17. The method according to claim 11, further comprising filtering the SHG emission from a collected light having a wavelength corresponding to the wavelength of the probe illumination light.

18. The method according to claim 11, wherein the first amount of probe illumination light is provided to the semiconductor wafer at a perpendicular or oblique angle of incidence with respect to the top surface of the semiconductor wafer.

19. The method according to claim 11, wherein the complex three-dimensional structure is a gate-all-around (GAA) device structure.

20. An illumination subsystem that provides a first amount of probe illumination light to a semiconductor wafer at each of a plurality of focal planes in a complex three-dimensional structure arranged on the semiconductor wafer according to a first optical mode, wherein the first amount of probe illumination light excites second harmonic generation (SHG) emission around material interfaces embedded deep within the complex three-dimensional structure, A collection subsystem for imaging a first amount of SHG emission from the complex three-dimensional structure to a detector at each of the plurality of focal planes according to the first optical mode, wherein the detector detects the imaged first SHG light at each of the plurality of focal planes, When executed by one or more processors, the one or more processors will Based on the first SHG light imaged at each of the plurality of focal planes, a three-dimensional volume image of the complex three-dimensional structure is generated. Based at least partially on the aforementioned three-dimensional volumetric image, buried defects are identified, or the values ​​of parameters of interest that characterize the buried structure are estimated. A non-temporary computer-readable medium for storing instructions, A system equipped with these features.

Images