A through-silicon via depth measurement method and device based on a pre-objective reference module

Abstract

This invention discloses a method and apparatus for measuring the depth of through-silicon vias (TSVs) based on a front-mounted objective reference module. The method includes: S1: Light emitted from a light source unit sequentially passes through a tube lens, objective lens, and beam splitter / combiner element before illuminating a non-hole region of a sample with a TSV. A spectral acquisition unit receives the returned light, obtains the interference spectrum signal, and processes it to obtain the system bias optical path characteristic corresponding to the round-trip optical path difference between the reference reflector and the non-hole region; S2: The sample is moved so that the light source unit is positioned on the TSV as described above, and the system bias optical path characteristic corresponding to the round-trip optical path difference between the reference reflector and the bottom of the TSV is obtained; S3: The physical depth of the TSV is calculated; S4: Anomalies occurring in steps S1 and S2 are handled. This method features high stability and a high signal-to-noise ratio, which helps promote the large-scale use of high-precision TSV measurement equipment in complex production line environments.

Description

Technical Field

[0001] This invention relates to the field of through-silicon via (TSV) depth measurement, and particularly to a method and apparatus for measuring TSV depth based on a front-mounted objective reference module. Background Technology

[0002] As Moore's Law approaches its physical limits, three-dimensional integrated circuits (3D ICs) have become the core path for advanced semiconductor products such as high-performance computing, artificial intelligence chips, and high-bandwidth storage to achieve continuous performance improvement. Through-silicon vias (TSVs), as a key process for vertical interconnection between chips, require high-precision non-destructive measurement of their depth, which is a crucial link in production line yield control.

[0003] Currently, the mainstream candidate methods for measuring the depth of through-silicon vias (TSVs) are spectral reflectance or spectral interferometry based on broadband or swept-frequency light sources. In traditional TSV interferometry, to obtain a reference light to interfere with the light reflected from the bottom of the TSV, the industry mainly uses the following three reference methods: 1. Sample Surface Reference Method: The light reflected from the upper surface of the sample under test is directly used as the reference light and interferes with the light reflected from the bottom of the through-silicon via in the main optical path. This method has a simple structure, but the amplitude, phase and reflectivity of the reference light all depend on the surface condition of the specific sample. The roughness of the sample surface, the difference in coating, the reflectivity of different materials and local contamination may all affect the quality of the reference light, resulting in low contrast of interference fringes and poor signal-to-noise ratio.

[0004] 2. Discrete long reference arm method: A beam splitter is used to build an independent, relatively long reference arm (such as the traditional Michelson interference structure) in the main optical path, and the reference mirror is placed at a position far away from the sample measurement optical path. This method can provide a stable reference light that is independent of the sample surface state. However, since the measurement arm and the reference arm are spatially separated, the differential optical path between the entire reference optical path and the measurement optical path is easily affected by environmental disturbances such as vibration, airflow, and temperature drift in the production line environment, resulting in poor phase stability.

[0005] 3. Integrated Interference Objective: Using a Mirau-type or Linnik-type interference objective, the beam splitter and reference reflection element are integrated inside the objective. This method has a certain degree of compactness and common-path disturbance resistance, but the reflectivity of the reference reflection element is usually fixed by the internal structure of the objective and is difficult to adjust independently to adapt to the weak echo at the bottom of the through-silicon via. At the same time, the reference reflection element of the Mirau-type interference objective is usually arranged on the optical axis of the objective, which may affect the effective light transmission area of ​​the main optical path.

[0006] In summary, the industry urgently needs a through-silicon via (TSV) depth interferometric measurement device and method that can simultaneously possess common optical path anti-disturbance capability, independently adjustable reference reflection element, and be well adapted to weak echo scenarios at the bottom of TSVs. Summary of the Invention

[0007] To address the aforementioned issues, this invention provides a method and apparatus for measuring the depth of through-silicon vias (TSVs) based on a front-mounted objective reference module. This method overcomes technical bottlenecks in constructing the interference optical environment by introducing a beam splitter / combiner unit and a reference reflection element between the objective and the sample. It provides stable reference light while maximizing the preservation of the optical path characteristics of coaxial interference, exhibiting high stability and a high signal-to-noise ratio. This will help promote the large-scale use of high-precision TSV measurement equipment in complex production line environments.

[0008] According to one aspect of the present invention, a method for measuring the depth of through-silicon vias based on a front objective reference module is provided, comprising the following steps: S1: The light emitted from the light source unit passes through the tube lens, objective lens, and beam splitter / combiner element in sequence and then shines on the non-pore region of the sample under test with a silicon through-hole. The spectral acquisition unit receives the returned light, obtains the interference spectral signal at this time, and processes it to obtain the system bias optical path characteristic quantity corresponding to the round-trip optical path difference between the reference reflection element and the non-pore region. S2: Move the sample to be tested so that the light emitted from the light source unit passes through the tube lens, objective lens, and beam splitter / combiner element in sequence and then shines on the silicon via of the sample to be tested. The spectral acquisition unit receives the returned light, obtains the interference spectral signal at this time, and processes it to obtain the system bias optical path characteristic quantity corresponding to the round-trip optical path difference between the reference reflection element and the silicon via. S3: Calculate the physical depth of the through-silicon via; S4: Handle any anomalies that occur in steps S1 and S2; In steps S1 and S2, the beam splitter / splitter element splits the illumination light into a measurement light and a reference light. The measurement light is transmitted or reflected to the sample to be tested, and the reference light is reflected or transmitted to the reference reflection element. The measurement light and the reference light are combined at the beam splitter / splitter element and then returned to the spectral acquisition unit.

[0009] In some implementations, in steps S1 and S2, the differential optical path between the measurement light and the reference light is located within a short optical path range within the objective lens front reference module. This is advantageous because it describes the limitation range of the differential optical path between the measurement light and the reference light.

[0010] In some implementations, in steps S1 and S2, the spectral acquisition unit processes the interference spectral signal using at least one of the following algorithms: Algorithm 1: After wavenumber domain nonlinear mapping and equal-interval resampling, apply wavenumber domain phase dispersion compensation containing at least second-order terms, then perform fast Fourier transform and peak finding; Algorithm 2: Wavenumber domain non-uniform Fourier transform and peak finding; Algorithm 3: Chirp-Z transform or scaled Fourier transform and peak finding; Algorithm 4: Cross-correlation between wavenumber domain spectral signals and candidate model functions, and peak finding; Algorithm 5: Maximum likelihood or least squares fitting based on multi-peak spectral models; Algorithm 6: Demodulation based on the phase slope of the interferometric spectrum; Algorithm 7: End-to-end deep regression based on deep learning neural networks.

[0011] Its advantage lies in describing how the spectral acquisition unit processes the interference spectral signal. In some implementations, the formula for calculating the physical depth of the through-silicon via (TSV) in step S3 is: D = (z2 z1) / (2·n_g); Where D is the physical depth of the through-silicon via (TSV), n_g is the group refractive index of the medium within the TSV near the center wavelength, z1 is the system bias optical path characteristic corresponding to the round-trip optical path difference between the reference reflective element and the non-via region, and z2 is the system bias optical path characteristic corresponding to the round-trip optical path difference between the reference reflective element and the bottom of the TSV. Its advantage lies in describing the method for calculating the physical depth of the TSV.

[0012] In some implementations, in step S4, when signal processing fails to stably extract the system bias optical path characteristic corresponding to the via location, the abnormal handling method involves controlling the processing unit to automatically use at least one of the following as a substitute: the system bias optical path characteristic of the adjacent non-via region, the autofocus height map, the wafer warpage height map, or the local surface height obtained from the via edge scan. Its advantage lies in describing an abnormal handling method under abnormal conditions.

[0013] In some implementations, in step S4, when the energy matching state between the reference light and the reflected light from the bottom of the through-silicon via (TSV) does not meet the preset energy matching conditions, the abnormal handling method is to control the processing unit to trigger at least one of the following: reselection of the reflectivity of the reference reflection element, reselection of the splitting ratio of the beam splitter / combiner element, abnormal marking, or production line alarm. Its advantage lies in describing another abnormal handling method under abnormal conditions.

[0014] In some implementations, in step S4, when the system bias calibration result of the objective lens front reference module drifts beyond a preset tolerance within the operating temperature range, the abnormal handling method is at least one of the following: the control processing unit triggers system bias recalibration, abnormal marking, or production line alarm. Its advantage lies in describing another abnormal handling method under abnormal conditions.

[0015] According to one aspect of the present invention, a silicon via depth measurement device based on a front objective reference module is provided for implementing the above-described measurement method, comprising: The optical measurement unit includes a light source unit, a tube mirror, an objective lens, and a spectral acquisition unit. The light source unit, tube mirror, objective lens, and sample to be measured are arranged in sequence, and the spectral acquisition unit is capable of receiving interference spectral signals. The objective lens front reference module includes a beam splitter / combiner and a reference reflection element disposed between the objective lens and the sample under test, and the reference reflection element is arranged outside the light transmission area or in the lateral area of ​​the light beam emitted by the light source unit. The control and processing unit is connected to the spectral acquisition unit and is capable of performing median filtering, wavenumber domain conversion, Fourier transform, and peak finding operations.

[0016] In some embodiments, beam splitters are provided between the light source unit and the tube lens, and between the tube lens and the objective lens. The advantage of this is that beam splitters can be used to divide the optical path as needed.

[0017] In some implementations, a camera and an illumination source are respectively provided on both sides of at least one beam splitter. This is advantageous because the camera and illumination source can be used to assist the beam splitter in dividing the optical path.

[0018] In some implementations, the working distance of the objective lens and the maximum structural thickness of the objective lens front reference module along the optical axis satisfy the following: WD ≥ d_module + h_clearance, Where WD is the working distance of the objective lens, d_module is the maximum structural thickness of the objective lens front reference module along the optical axis, and h_clearance is the distance between the side of the objective lens front reference module closest to the sample and the upper surface of the sample. Its advantage lies in describing the relevant requirements for the working distance of the objective lens.

[0019] In some embodiments, the reference reflective element is one or more of the following: a fixed plane mirror, a low-reflectivity coated mirror with a preset reflectivity, an adjustable reflective element, a liquid crystal variable reflector, a combination of a neutral density filter and a fixed mirror, or a multi-mirror array. The advantage is that it describes the possible types of reference reflective elements.

[0020] In some embodiments, the objective lens front reference module includes a rigid support substrate made of a material with a low coefficient of thermal expansion, on which both the beam splitter / splitter and the reference reflector are rigidly fixed. The advantage is that the rigid support substrate allows for the fixation of the beam splitter / splitter and the reference reflector.

[0021] In some embodiments, a movable sample positioning unit is also included, on which the sample to be tested is placed. The advantage of this is that the sample positioning unit can support and move the sample to be tested. Attached Figure Description

[0022] Figure 1 This is a structural diagram of the optical measurement unit and the objective lens front reference module of a silicon via depth measurement device based on an objective lens front reference module according to one embodiment of the present invention. Figure 2 for Figure 1 The diagram shows a technical concept of a through-silicon via depth measurement method based on a front-mounted objective reference module. Figure 3 This is a flowchart of the signal processing for through-silicon via (TSV) depth measurement.

[0023] In the figure: 1. Light source unit; 2. Tube lens; 3. Objective lens; 4. Sample to be tested; 5. Beam splitter / combiner unit; 6. Reference reflector element; 7. Beam splitter; 8. Camera; 9. Illumination source. Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings.

[0025] The present invention provides a silicon through-hole depth measurement device based on an objective lens front reference module, which is used to implement the corresponding measurement method. The device mainly includes an optical measurement unit, an objective lens front reference module, and a control and processing unit.

[0026] The optical measurement unit consists of a light source unit, a tube lens, an objective lens, and a spectral acquisition unit (some structures are not shown in the figure). The light source unit, tube lens, objective lens, and sample to be tested are arranged in sequence. The tube lens and objective lens form a microscopic focusing optical path, which can concentrate the illumination spot onto the surface of the sample to be tested.

[0027] The light source unit is configured to provide measurement light with a predetermined spectral bandwidth, which is either broadband light with a predetermined spectral bandwidth or swept light that scans the predetermined spectral bandwidth within the measurement period. The light source unit can be any one or a combination of several of the following: a halogen lamp, a supercontinuum laser, a broadband SLD, and a swept laser, with its operating spectral range between 400 nm and 1700 nm.

[0028] The spectral acquisition unit is selected from any one or a combination of several of the following: spectrometer, spectral camera, linear array detector, area array detector, photodetector working synchronously with a swept frequency light source, balanced detector, or equivalent spectral acquisition component, and it is capable of receiving raw interference spectral signals.

[0029] Preferably, the working distance of the objective lens is not less than 5 mm.

[0030] Preferably, beam splitters are provided between the light source unit and the tube lens, and between the tube lens and the objective lens. On both sides of at least one beam splitter, there are cameras carrying imaging lenses and illumination sources for illumination. The cameras and illumination sources can be used to assist the beam splitters in dividing the optical path and performing image detection.

[0031] The objective lens front reference module is arranged between the objective lens and the sample to be tested. It mainly includes a beam splitter / combiner and a reference reflection element, and the reference reflection element is arranged outside or to the side of the light transmission area of ​​the measurement beam emitted by the light source unit.

[0032] The working distance WD of the objective lens and the maximum structural thickness d_module of the front reference module along the optical axis satisfy the following: WD ≥ d_module + h_clearance, Where h_clearance is the distance between the side of the objective lens front reference module closest to the sample and the upper surface of the sample. The above formula ensures that after the objective lens front reference module is assembled, the equivalent working distance of the objective lens is still sufficient to allow the measurement light to form an effective focused spot on the sample.

[0033] The reference reflective element is selected from one or a combination of the following structures: 1. Fixed plane mirror; 2. A low-reflectivity coated mirror with a preset reflectivity; 3. An adjustable reflectivity reflective element configured to adjust its reflectivity according to the bottom-hole echo intensity of the TSV under test; 4. Liquid crystal variable reflector; 5. Combination of neutral density filter and fixed mirror; 6. A multi-mirror array configured to provide different reflectivities or reference optical paths by switching different mirrors.

[0034] Furthermore, the reflectivity of the reference reflector and the splitting ratio of the beam splitter / combiner are configured such that the energy of the reference light returned from the reference reflector to the beam splitter / combiner and the energy of the light reflected from the bottom of the through-silicon via (TSV) returned by the beam splitter / combiner satisfy one of the following energy matching conditions: 1. The energy matching range is set such that at least one of the following: the contrast of the interference fringe envelope acquired by the spectral acquisition unit, the signal-to-noise ratio corresponding to the processed peak, or the full width at half maximum (FWHM) of the processed peak satisfies a preset quality criterion. 2. The energy ratio between the reference light and the reflected light from the bottom of the through-silicon via falls within at least one value or a sub-range in the range of 1:1 to 100:1, preferably within at least one value or a sub-range in the range of 1:1 to 50:1.

[0035] Preferably, the objective lens front reference module includes a rigid support substrate, which is made of a material with a low coefficient of thermal expansion. The material with a low coefficient of thermal expansion is selected from Invar alloy, super Invar alloy, zero-expansion glass ceramic, fused silica, or other materials with a coefficient of thermal expansion not greater than a preset threshold. The beam splitter / splitter and the reference reflection element are both rigidly fixed on the rigid support substrate, so that the relative position between the two remains stable within the operating temperature range.

[0036] The control and processing unit can be a computer or the like, which is connected to the spectral acquisition unit and can perform median filtering, wavenumber domain conversion, Fourier transform and peak finding operations. It can be used to process the raw interference spectral signal acquired by the spectral acquisition unit to obtain the depth of the through silicon via.

[0037] In addition, it includes a sample positioning unit (not shown in the figure) for supporting the sample to be tested and for moving the sample to be tested so that the measurement position can be switched between the non-pore area of ​​the sample to be tested, the through-silicon via position or the preset calibration position.

[0038] like Figure 2 As shown, the silicon through-hole depth measurement method based on the objective lens front reference module of the present invention includes several main steps, which are described below.

[0039] S1: System bias calibration.

[0040] In the optical measurement unit, the light emitted from the light source unit passes through the beam splitter / combiner of the tube lens, objective lens, and objective lens pre-reference module, and then illuminates the non-pore region (i.e., the surface outside the silicon via) of the sample under test with a silicon via. The beam splitter / combiner splits the illumination light into measurement light and reference light. The measurement light is transmitted or reflected to the sample under test, while the reference light is reflected or transmitted to the reference reflection element. Then, the measurement light and the reference light are combined at the beam splitter / combiner and return to the spectral acquisition unit to obtain the interference spectral signal at this time.

[0041] Then, the spectral acquisition unit processes the interference spectral signal to obtain the system bias optical path characteristic z1 corresponding to the round-trip optical path difference between the reference reflective element and the non-aperture region.

[0042] S2: Move the sample to be tested so that the light emitted from the light source unit passes through the beam splitter / splitter of the tube lens, objective lens, and objective lens pre-reference module and then illuminates the silicon through-hole of the sample. The beam splitter / splitter splits the illumination light into measurement light and reference light. The measurement light is transmitted or reflected to the sample, while the reference light is reflected or transmitted to the reference reflection element. Then, the measurement light and reference light are combined at the beam splitter / splitter and return to the spectral acquisition unit along the same path to obtain the interference spectral signal at this time.

[0043] Then, the spectral acquisition unit processes the interference spectral signal to obtain the system bias optical path characteristic z2 corresponding to the round-trip optical path difference between the reference reflective element and the bottom of the silicon via.

[0044] In steps S1 and S2, the differential optical path between the measurement light and the reference light is mainly limited to the short optical path range within the objective lens front reference module.

[0045] In steps S1 and S2, the spectral acquisition unit processes the interference spectral signal using at least one of the following algorithms: Algorithm 1: After wavenumber domain nonlinear mapping and equal-interval resampling, apply wavenumber domain phase dispersion compensation containing at least second-order terms, then perform fast Fourier transform and peak finding; Algorithm 2: Wavenumber domain non-uniform Fourier transform and peak finding.

[0046] Algorithm 3: Chirp-Z transform or scaled Fourier transform and peak finding.

[0047] Algorithm 4: Cross-correlation between wavenumber domain spectral signal and candidate model function and peak finding.

[0048] Algorithm 5: Maximum likelihood or least squares fitting based on multi-peak spectral models.

[0049] Algorithm 6: Demodulation based on the phase slope of the interference spectrum.

[0050] Algorithm 7: End-to-end deep regression based on deep learning neural networks.

[0051] S3: Calculate the physical depth of the through-silicon via.

[0052] The calculation formula is D = (z2) z1) / (2·n_g).

[0053] Where D is the physical depth of the through-silicon via (TSV), and n_g is the group refractive index of the medium within the TSV near the center wavelength.

[0054] S4: Perform abnormal handling.

[0055] If the above steps fail to obtain a reasonable physical depth value for through-silicon vias, it is necessary to identify any anomalies in the above scheme and perform anomaly handling operations.

[0056] The abnormal handling measures include the following: 1. When the signal processing cannot stably extract the system bias optical path feature z1 corresponding to the location of the through-silicon via, the control processing unit automatically uses at least one of the following as the system bias replacement: the system bias optical path feature of the adjacent non-via region, the autofocus height map, the wafer warpage height map, or the local surface height obtained by scanning the aperture edge. 2. When the energy matching state between the reference light and the reflected light from the bottom of the through-silicon via does not meet the preset energy matching conditions, the control processing unit triggers at least one of the following processing methods: reselection of the reflectivity of the reference reflection element, reselection of the splitting ratio of the beam splitter / combiner element, abnormal marking, or production line alarm. 3. When the system bias calibration result of the objective lens front reference module drifts beyond the preset tolerance within the operating temperature range, the control processing unit triggers at least one of the following processing methods: system bias recalibration, abnormal marking, or production line alarm.

[0057] The following is a comparative analysis of the key TSV depth measurement indicators obtained by the objective lens front reference module method of this invention, the traditional sample surface reference method, the traditional Michelson-type discrete long reference arm method, and the Mirau-type integrated interferometric objective method under the same light source, the same spectrometer, and the same objective lens conditions. Typical comparison results are shown in Table 1: Table 1

[0058] As shown in Table 1, the core advantages of the objective lens front reference module method described in this invention are: it has the ability to resist disturbances in the common optical path, the independent adjustability of the reference reflection element, the modularity and detachability, and the ability to flexibly adapt to weak echo scenarios at the bottom of silicon vias. The specific degree of improvement depends on the selected light source, spectrometer, objective lens, TSV structure, environmental conditions and test conditions. Those skilled in the art can optimize the parameters according to the specific application scenario.

[0059] The method and apparatus for measuring the depth of through-silicon vias based on a front-mounted objective lens reference module in this invention have the following advantages: 1. By compressing the reference optical path into the short module at the front of the objective lens, the measurement light and the reference light return coaxially in the main optical path, which helps to reduce the influence of environmental vibration, airflow and temperature drift on the interference phase; 2. The reference reflective element is an independent component, and its reflectivity can be selected or adjusted independently, which helps to perform energy matching for the bottom echo intensity of different samples under test; 3. Compared with traditional sample surface reference methods, the reference light quality of this invention does not depend on the specific sample surface condition, which can improve the contrast of interference fringes and the signal-to-noise ratio; 4. Compared to the traditional Michelson-type discrete reference arm method, the differential optical path of the present invention is mainly limited to the short optical path range of the objective lens front end, which can improve the phase stability of the system in the production line environment; 5. Compared with the integrated interferometer objective lens of the Mirau or Linnik type, the reference reflection element of the present invention can be adjusted independently as an independent element and can be arranged outside or to the side of the main beam of measurement, which helps to achieve flexible matching between the reference energy and the weak echo at the bottom of the silicon via without affecting the effective light transmission of the main optical path. 6. The objective lens front reference module is constructed as a detachable add-on module, which can be added to the existing microscopic measurement platform as an add-on component, which helps to reduce the cost of production line upgrades and equipment development; 7. By conditionally extracting local system bias from the same TSV measurement spectrum, local height fluctuations of the sample at the location of the through-silicon via can be compensated. 8. The device of the present invention maintains a certain degree of versatility and can be superimposed with related technologies in the same field, further expanding the miniaturization, anti-interference and high aspect ratio adaptability of measurement equipment.

[0060] The above descriptions are merely some embodiments of the present invention. Those skilled in the art can make various modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the scope of protection of the present invention.

Claims

1. A method for measuring the depth of through-silicon vias based on a front-view objective reference module, characterized in that: Includes the following steps: S1: The light emitted from the light source unit passes through the tube lens, objective lens, and beam splitter / combiner element in sequence and then shines on the non-pore region of the sample under test with a silicon through-hole. The spectral acquisition unit receives the returned light, obtains the interference spectral signal at this time, and processes it to obtain the system bias optical path characteristic quantity corresponding to the round-trip optical path difference between the reference reflection element and the non-pore region. S2: Move the sample to be tested so that the light emitted from the light source unit passes through the tube lens, objective lens, and beam splitter / combiner element in sequence and then shines on the silicon via of the sample to be tested. The spectral acquisition unit receives the returned light, obtains the interference spectral signal at this time, and processes it to obtain the system bias optical path characteristic quantity corresponding to the round-trip optical path difference between the reference reflection element and the bottom of the silicon via. S3: Calculate the physical depth of the through-silicon via; S4: Handle any anomalies that occur in steps S1 and S2; In steps S1 and S2, the beam splitter / splitter element splits the illumination light into a measurement light and a reference light. The measurement light is transmitted or reflected to the sample to be tested, while the reference light is reflected or transmitted through the reference reflection element. The measurement light and the reference light are combined at the beam splitter / splitter element and then returned to the spectral acquisition unit.

2. The method for measuring the depth of through-silicon vias based on a front-mounted objective reference module according to claim 1, characterized in that: In steps S1 and S2, the differential optical path between the measurement light and the reference light is located within the short optical path range of the objective lens front reference module.

3. The method for measuring the depth of through-silicon vias based on a front-mounted objective lens reference module according to claim 1, characterized in that: In steps S1 and S2, the spectral acquisition unit processes the interference spectral signal using at least one of the following algorithms: Algorithm 1: After wavenumber domain nonlinear mapping and equal-interval resampling, apply wavenumber domain phase dispersion compensation containing at least second-order terms, then perform fast Fourier transform and peak finding; Algorithm 2: Wavenumber domain non-uniform Fourier transform and peak finding; Algorithm 3: Chirp-Z transform or scaled Fourier transform and peak finding; Algorithm 4: Cross-correlation between wavenumber domain spectral signals and candidate model functions, and peak finding; Algorithm 5: Maximum likelihood or least squares fitting based on multi-peak spectral models; Algorithm 6: Demodulation based on the phase slope of the interferometric spectrum; Algorithm 7: End-to-end deep regression based on deep learning neural networks.

4. The method for measuring the depth of through-silicon vias based on a front-mounted objective lens reference module according to claim 1, characterized in that: In step S3, the formula for calculating the physical depth of the through-silicon via is: D = (z2 z1) / (2·n_g): Where D is the physical depth of the through-silicon via, n_g is the group refractive index of the medium inside the through-silicon via near the center wavelength, z1 is the system bias optical path characteristic corresponding to the round-trip optical path difference between the reference reflective element and the non-via region, and z2 is the system bias optical path characteristic corresponding to the round-trip optical path difference between the reference reflective element and the bottom of the through-silicon via.

5. The method for measuring the depth of through-silicon vias based on a front-mounted objective lens reference module according to claim 1, characterized in that: In step S4, when the signal processing cannot stably extract the system bias optical path feature corresponding to the location of the through-silicon via, the abnormal handling method is to control the processing unit to automatically use at least one of the following to correct the system bias optical path feature: the system bias optical path feature of the adjacent non-via region, the autofocus height map, the wafer warpage height map, or the local surface height obtained by scanning the aperture edge, so as to obtain an alternative local system bias optical path feature.

6. The method for measuring the depth of through-silicon vias based on a front-mounted objective lens reference module according to claim 1, characterized in that: In step S4, when the energy matching state between the reference light and the reflected light from the bottom of the through-silicon via does not meet the preset energy matching conditions, the abnormal handling method is at least one of the following: the control processing unit triggers the reselection of the reflectivity of the reference reflection element, the reselection of the splitting ratio of the beam splitting / combining element, the abnormal marking, or the production line alarm.

7. A method for measuring the depth of through-silicon vias based on a front-mounted objective reference module according to claim 1, characterized in that: In step S4, when the system bias calibration result of the objective lens front reference module drifts beyond the preset tolerance within the operating temperature range, the abnormal handling method is at least one of the following: the control processing unit triggers the recalibration of the system bias, abnormal marking, or production line alarm.

8. A silicon via depth measurement device based on a front objective reference module, used to implement the measurement method according to any one of claims 1-7, characterized in that: include The optical measurement unit includes a light source unit, a tube mirror, an objective lens, and a spectral acquisition unit. The light source unit, tube mirror, objective lens, and sample to be measured are arranged in sequence, and the spectral acquisition unit is capable of receiving interference spectral signals. The objective lens front reference module includes a beam splitter / combiner and a reference reflection element disposed between the objective lens and the sample under test, and the reference reflection element is arranged outside the light transmission area or in the lateral area of ​​the light beam emitted by the light source unit. The control processing unit is connected to the spectral acquisition unit and is configured to perform optical path difference domain inversion on the interference spectral signal to obtain the system bias optical path characteristics and the aperture bottom optical path characteristics.

9. A silicon via depth measurement device based on a front objective reference module according to claim 8, characterized in that: Beam splitters are installed between the light source unit and the tube lens, and between the tube lens and the objective lens.

10. A silicon via depth measurement device based on a front objective reference module according to claim 9, characterized in that: A camera and an illumination source are respectively provided on both sides of at least one beam splitter.

11. A silicon via depth measurement device based on a front objective reference module according to claim 8, characterized in that: The working distance of the objective lens and the maximum structural thickness of the objective lens front reference module along the optical axis satisfy the following: WD ≥ d_module + h_clearance, Where WD is the working distance of the objective lens, d_module is the maximum structural thickness of the objective lens front reference module along the optical axis, and h_clearance is the distance between the side of the objective lens front reference module closest to the sample and the upper surface of the sample.

12. A silicon via depth measurement device based on a front objective reference module according to claim 8, characterized in that: The reference reflective element is one or more of the following: a fixed plane mirror, a low-reflectivity coated mirror with a preset reflectivity, an adjustable reflective element, a liquid crystal variable reflector, a combination of a neutral density filter and a fixed mirror, or a multi-mirror array.

13. A silicon via depth measurement device based on a front objective reference module according to claim 8, characterized in that: The objective lens front reference module includes a rigid support substrate made of a material with a low coefficient of thermal expansion, and the beam splitter / combiner element and the reference reflection element are rigidly fixed on the rigid support substrate.

14. A silicon via depth measurement device based on a front objective reference module according to claim 8, characterized in that: It also includes a movable sample positioning unit on which the sample to be tested is placed.

15. A silicon via depth measurement device based on a front objective reference module according to claim 8, characterized in that... The reflectivity of the reference reflective element and the splitting ratio of the beam splitter / combiner element are configured such that the reference light energy and the light energy reflected from the bottom of the through-silicon via satisfy a preset energy matching condition. The preset energy matching conditions include at least one of the following: interference fringe envelope contrast, optical path difference domain peak signal-to-noise ratio, or peak full width at half maximum (FWHM) satisfying a preset quality criterion; or the energy ratio between the reference light and the aperture bottom reflected light falling within at least one value or a sub-range in the range of 1:1 to 100:1.

Images