A semiconductor wafer thickness detection system and method

Abstract

The application relates to the field of semiconductor manufacturing detection, and discloses a semiconductor wafer thickness detection system and method, which are used for solving the misjudgment and missed judgment problems caused by aiming error and random noise in the prior art. The method comprises the following steps: acquiring measurement data to determine calibration reference data; making the to-be-detected semiconductor wafer and two sensors generate relative motion, collecting distance reading pairs of multiple sampling positions as sampling data; performing differential calculation and abnormal elimination on the sampling data by using the calibration reference data, obtaining thickness values of the wafer at multiple positions, and collecting the thickness values to form a thickness data set; extracting a thickness characteristic value according to the thickness data set, comparing the thickness characteristic value with a preset judgment condition, judging whether the wafer is qualified or not, and performing corresponding wafer transfer operations according to the judgment result. By introducing a dynamic calibration mechanism and multi-dimensional feature extraction, the system measurement error is effectively eliminated, and the automatic detection of wafer thickness and spatial morphology is realized.

Description

Technical Field

[0001] This invention relates to the field of semiconductor manufacturing inspection, and more particularly to a semiconductor wafer thickness inspection system and method. Background Technology

[0002] In semiconductor manufacturing processes, the absolute dimensional accuracy and intra-wafer uniformity of wafer thickness are key parameters that determine subsequent photolithography alignment accuracy, bonding quality, and chip yield. With the continuous shrinking of semiconductor device feature sizes and the widespread adoption of large-size wafers, production lines have placed stringent requirements on thickness inspection, demanding non-contact, high-speed, full-circumference, and high repeatability capabilities.

[0003] Currently, the mainstream wafer thickness inspection solutions in the industry are mainly divided into two categories: one is contact measurement based on mechanical micrometers or dial indicators. Although the principle is simple, it is easy to scratch the wafer surface and the wear of the probe will introduce systematic errors; the other is non-contact measurement based on laser triangular reflection or single-sided capacitance / laser through-beam measurement. Among them, dual-sensor through-beam measurement has become the mainstream direction for online inspection because it can directly obtain the distance between the two sides and invert the thickness through the distance difference. Theoretically, it is not affected by wafer warping.

[0004] However, in actual mass production applications, when integrating through-beam sensors into automated conveying and rotating mechanisms to perform high-speed full inspection, a series of engineering challenges remain, such as calibration drift, environmental temperature changes, interference from wafer geometric features, and the rejection of abnormal samples from a large number of sampling points. These challenges limit the further improvement of detection accuracy and automated closed-loop control capabilities.

[0005] Existing through-beam sensors typically undergo only one zero-point calibration after installation. However, after long-term operation, vibration, or slight sensor displacement, the sensor spacing reference will shift. Traditional solutions lack the step of online verification and real-time correction of compensation parameters using standard gauge blocks. This results in accumulated errors being directly incorporated into the thickness inversion formula, making it difficult to meet sub-micron tolerance requirements for measurement repeatability.

[0006] Existing systems typically only use whether the thickness is within the upper and lower limits as the basis for release, ignoring the monitoring of thickness tolerances; and after a non-compliance is determined, the system only stops the machine and alarms, lacking linkage with MES or robotic arm wafer retrieval actions, and there is no automatic triggering mechanism for re-retrieval of wafers, which causes the production line cycle to be severely disrupted and makes it impossible to trace the specific location of the deviation.

[0007] Therefore, we propose a semiconductor wafer thickness detection system and method to solve the above problems. Summary of the Invention

[0008] This invention provides a semiconductor wafer thickness detection system and method to solve the problems of misjudgment and missed judgment caused by photoelectric error and random noise in the prior art.

[0009] The first aspect of this invention provides a method for detecting the thickness of a semiconductor wafer. The method includes: providing a calibration fixture and two through-beam distance sensors; acquiring measurement data from the calibration fixture to determine calibration reference data; placing a semiconductor wafer to be tested between the two distance sensors, causing relative motion between the wafer and the sensors, and collecting distance readings at multiple sampling locations as sampling data; calculating the sampling data using the calibration reference data to obtain thickness values ​​corresponding to the semiconductor wafer at the multiple sampling locations, and summarizing these values ​​to form a thickness dataset; extracting thickness feature values ​​from the thickness dataset; comparing the thickness feature values ​​with preset judgment conditions to determine whether the semiconductor wafer to be tested is qualified, and obtaining a judgment result; and performing corresponding wafer transfer operations based on the judgment result.

[0010] Optionally, in a first implementation of the first aspect of the present invention, the method includes: Drive the semiconductor wafer under test to rotate; Multiple sampling positions are uniformly set on the circumference of the semiconductor wafer under test; When the set sampling position coincides with the notch or flat edge of the semiconductor wafer under test, the sampling position is extended to the next position until it does not fall into the notch or flat edge area, thus obtaining multiple valid sampling positions; When the semiconductor wafer under test reaches each of the effective sampling positions, the distance readings of the two distance detection sensors are recorded and paired to form corresponding distance reading pairs, thereby obtaining sampling data containing the distance reading pairs of all the effective sampling positions.

[0011] Optionally, in a second implementation of the first aspect of the present invention, the step of recording the distance readings of the two distance detection sensors when the semiconductor wafer under test reaches each of the effective sampling positions, and pairing them to form corresponding distance reading pairs, includes: Upon reaching each of the effective sampling positions, the two distance detection sensors continuously collect multiple distance readings. The maximum and minimum values ​​of the multiple distance readings collected by each distance detection sensor are removed, and the average value of the remaining distance readings is taken to obtain the effective distance readings of the two distance detection sensors at the effective sampling position. The effective distance readings of the two distance detection sensors are paired to form a distance reading pair for the effective sampling position.

[0012] Optionally, in a third implementation of the first aspect of the present invention, the method includes: Add the distance readings for each sampling location in the sampled data to obtain a set of readings and pairs; The reference spacing length is calculated based on the calibration reference data; Subtract each reading and pair from the set of readings and pairs from the reference spacing length to obtain the initial thickness value set; Abnormal thickness values ​​that deviate from the median of the initial thickness value set and exceed a preset deviation threshold are removed, and the remaining thickness values ​​are used as the thickness dataset.

[0013] Optionally, in a fourth implementation of the first aspect of the present invention, it further includes: Obtain the ambient temperature; The thickness correction amount is determined based on the preset correspondence between the ambient temperature and the thermal expansion coefficient of the wafer material; The thickness correction amount is applied to each thickness value in the thickness dataset to obtain a temperature-compensated thickness dataset for subsequent extraction of thickness feature values.

[0014] Optionally, in a fifth implementation of the first aspect of the present invention, the thickness feature value includes a maximum thickness value, a minimum thickness value, and a thickness range; the preset judgment condition includes a preset upper limit for thickness tolerance, a preset lower limit for thickness tolerance, and a preset thickness uniformity threshold. The step of extracting thickness feature values ​​from the thickness dataset, comparing the thickness feature values ​​with preset judgment conditions, and determining whether the semiconductor wafer under test is qualified to obtain a judgment result includes: Extract the maximum and minimum thickness values ​​from the thickness dataset; The maximum thickness is compared with the preset upper limit of thickness tolerance, and the minimum thickness is compared with the preset lower limit of thickness tolerance. Calculate the difference between the maximum thickness and the minimum thickness to obtain the thickness range, and compare the thickness range with the preset thickness uniformity threshold; When all the above comparison results meet the preset judgment conditions, the semiconductor wafer under test is determined to be qualified; otherwise, it is determined to be unqualified, and a judgment result is generated.

[0015] Optionally, in a sixth implementation of the first aspect of the present invention, extracting the maximum thickness and the minimum thickness from the thickness dataset further includes: When extracting the maximum and minimum thickness values, the sampling position identifiers corresponding to the maximum and minimum thickness values ​​are recorded respectively; Based on the sampling location identifier, the angle between the maximum thickness and the minimum thickness on the circumference of the semiconductor wafer under test is determined; The method of determining whether the semiconductor wafer under test is qualified also includes: When the thickness range does not exceed the preset thickness uniformity threshold, but the included angle exceeds the preset angle threshold, it is determined that the semiconductor wafer under test has a local thickness abnormality and is therefore deemed unqualified.

[0016] Optionally, in a seventh implementation of the first aspect of the present invention, the step of performing the corresponding wafer transfer operation based on the determination result includes: When the judgment result is unqualified, an alarm is triggered and a replacement wafer instruction is generated, marking the semiconductor wafer under test as an abnormal product for subsequent processing. When the judgment result is qualified, an analysis report containing the thickness dataset is output, and a release instruction is generated to control the semiconductor wafer under test to be transferred to the subsequent processing steps.

[0017] A second aspect of the present invention provides a semiconductor wafer thickness detection system, comprising: a reference determination module, configured to acquire measurement data of a calibration fixture based on two through-beam mounted distance sensors to determine calibration reference data; a sampling acquisition module, configured to acquire distance reading pairs at multiple sampling positions as sampling data when the semiconductor wafer under test moves relative to the two distance sensors; a calculation and summarization module, configured to calculate the sampling data using the calibration reference data to obtain the thickness values ​​of the semiconductor wafer under test at the multiple sampling positions, and summarize them to form a thickness dataset; a feature judgment module, configured to extract thickness feature values ​​based on the thickness dataset, compare the thickness feature values ​​with preset judgment conditions, determine whether the semiconductor wafer under test is qualified, and obtain a judgment result; and a transfer execution module, configured to execute corresponding wafer transfer operations based on the judgment result.

[0018] The mechanism of this invention is as follows: By employing rotational uniform sampling combined with a notch / flat edge avoidance mechanism, the effectiveness of the 16 sampling points on the wafer circumference is ensured, avoiding the contamination of thickness data by structural features; at the data processing level, a median deviation elimination algorithm is introduced to actively filter outlier noise caused by vibration or local anomalies, significantly improving the robustness of the data; and in terms of judgment dimensions, the invention breaks through the traditional single extreme value comparison, coupling the absolute thickness tolerance with the range (uniformity) and the angle between extreme points (local morphology) for evaluation, thereby accurately distinguishing between global out-of-tolerance and local warping defects; Beneficial effects: By introducing calibration fixtures and standard gauge blocks for dual comparison, the known standard thickness is introduced into the calculation closed loop to calculate the compensation parameters in a physical sense and participate in the differential inversion of each subsequent thickness value. This eliminates the systematic errors caused by mechanical micro-displacement and optical path temperature drift of the sensor, enabling the through-beam measurement to maintain sub-micron level long-term repeatability even in production line vibration environments, without the need for frequent shutdowns for manual calibration. In rotating scans, equal distribution and full entity compatibility are achieved, which not only ensures the representativeness of circumferential statistics but also the purity of data, and reduces the false alarm rate caused by geometric features. Instead of simply removing the first and last elements, the algorithm introduces a removal rule based on the median and a preset deviation threshold into the sorted sequence. This results in strong immunity to single-point outliers while preserving local thickness extrema, achieving a balance between noise reduction and fidelity preservation. By linking the ambient temperature range table with the thermal expansion coefficient of silicon materials, thickness data can be restored from the dataset. Even in non-constant temperature environments on the production line or during the initial warm-up phase of equipment startup, it can still provide the true physical thickness that conforms to the 20℃ standard definition, solving the fundamental deviation problem when comparing online testing with standard measuring instruments in the metrology laboratory; By introducing three-dimensional criteria for the range and the circumferential angle of the extreme point, and directly obtaining the wafer replacement instruction and batch traceability chain based on the judgment result, it can accurately identify local wedge anomalies and avoid misjudging and scrapping salvageable wafers. At the same time, the wafer replacement trigger reduces the cycle time of manual intervention, and truly integrates the thickness detection station into the fully automated production line, thereby improving the overall OEE. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of one embodiment of the semiconductor wafer thickness detection method according to the present invention; Figure 2 This is a schematic diagram of another embodiment of the semiconductor wafer thickness detection method in this invention; Figure 3 This is a schematic diagram of one embodiment of the semiconductor wafer thickness detection system of the present invention; Figure 4 This diagram illustrates a semiconductor wafer under test placed on a rotating mechanism between two sensors. As the wafer rotates, the sensors uniformly collect distance readings at 16 locations along the wafer's circumference. Detailed Implementation

[0020] This invention provides a semiconductor wafer thickness detection system and method to address the problems of false positives and false negatives caused by photoelectric errors and random noise in the prior art. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the semiconductor wafer thickness detection method of the present invention includes: 101. Provide a calibration fixture and two through-beam mounted distance sensors. Place the calibration fixture between the two sensors and obtain distance readings from the two sensors to the two sides of the calibration fixture. Obtain the standard thickness of the calibration fixture. Place a standard gauge block of known thickness between the two sensors for verification. Determine the compensation parameter based on the difference between the measured thickness of the standard gauge block and the standard thickness. Use the above distance readings, standard thickness, and compensation parameter as calibration reference data.

[0022] It is understood that the executing entity of this invention can be a semiconductor wafer thickness detection system, a terminal, or a server; no specific limitation is made here. This embodiment of the invention will be described using a server as an example.

[0023] It should be noted that this embodiment provides an example of a thickness measurement device for a 12-inch semiconductor wafer. Two laser displacement sensors (hereinafter referred to as the upper sensor and lower sensor), mounted vertically and in opposite directions, are used to establish the initial physical reference of the system. To ensure that the calibration process is free from errors due to minor deformation of the measured part or gravity-induced sagging, a highly rigid metal calibration fixture is used. The system retrieves the standard thickness of this calibration fixture from the metrology archive, confirming its thickness to be 5000.00 μm. An automated robotic arm smoothly places the calibration fixture in the measurement area between the two sensors. The main control system synchronously triggers the illumination distance measurement, obtaining a distance reading of 2400.00 μm from the upper sensor probe to the upper surface of the calibration fixture, and simultaneously obtaining a distance reading of 2600.00 μm from the lower sensor probe to the lower surface of the calibration fixture. The system adds the calibration fixture thickness to the distance readings of the two sensors, resulting in a total physical installation distance of 10000.00 μm between the two sensor probes.

[0024] The system enters the verification and compensation parameter determination phase. The robotic arm removes the calibration fixture and places a standard gauge block (standard reference wafer) specifically designed to simulate the actual product into the optical path. The known standard thickness of this gauge block is 775.00 μm. The system triggers the sensors to synchronize sampling again, measuring a distance of 4487.80 μm from the upper sensor to the upper surface of the gauge block and a distance of 4737.15 μm from the lower sensor to the lower surface of the gauge block.

[0025] The measured thickness of the standard gauge block was calculated as follows: Using the previously obtained total physical installation spacing of the sensors (10000.00 μm), subtracting the current upper sensor reading of 4487.80 μm and the lower sensor reading of 4737.15 μm, the preliminary measured thickness of the standard gauge block was found to be 775.05 μm. Comparing this preliminary measured thickness with the known standard thickness of 775.00 μm, a difference of 0.05 μm was found. This difference comprehensively reflects environmental temperature drift, sensor linearity error, and microscopic deformation of the equipment frame.

[0026] Based on this, the positive 0.05μm is established as the compensation parameter. This parameter must be deducted in all subsequent formal measurement calculations. The system saves the initially measured upper sensor reading of 2400.00μm, lower sensor reading of 2600.00μm, calibration fixture thickness of 5000.00μm, and the calculated compensation parameter of 0.05μm to the controller's non-volatile memory and uses them as calibration reference data.

[0027] 102. Place the semiconductor wafer under test between two sensors, drive the wafer to rotate, and collect distance reading pairs at 16 locations on the circumference of the wafer as sampling data.

[0028] It should be noted that the automated loading and unloading mechanism transfers the 12-inch semiconductor wafer to be tested onto the rotating test stage between the sensors, and uses a vacuum chuck to fix its central area. The edge detection area of ​​the wafer (located at a radius ring 145mm from the center of the wafer) is cut into the perpendicularly opposite optical path of the upper and lower sensors. In a real engineering environment, due to the limitations of the machining accuracy of the rotating mechanism and the natural warpage of the wafer itself, the edge of the wafer will inevitably experience end-face runout of several micrometers to tens of micrometers in the vertical direction (Z-axis direction) during rotation.

[0029] The controller issues a command, and the servo motor drives the rotating test stand to rotate the wafer at a constant speed. The system relies on an absolute encoder at the rear of the motor to trigger two sensors to synchronously record data every 22.5° of wafer rotation. When the wafer is at the initial 0° position (i.e., the first position), the upper sensor measures a distance of 4612.45 μm, and the lower sensor measures a distance of 4612.60 μm, forming the distance reading pair for the first position.

[0030] As the turntable rotates, the wafer reaches the second position at 22.5°. At this point, due to mechanical runout, the wafer shifts upward by approximately 4μm at the measurement point, causing the upper sensor reading to decrease to 4608.20μm and the lower sensor reading to increase to 4616.83μm. When the wafer continues to rotate to 112.5° (the sixth position), the upward mechanical runout reaches its peak, causing the upper sensor reading to decrease significantly to 4600.40μm and the lower sensor reading to increase accordingly to 4624.47μm.

[0031] When the wafer rotates to 247.5° (position 12), mechanical runout causes the wafer to shift downwards in the optical path. At this point, the distance reading of the upper sensor increases to 4623.50 μm, while the reading of the lower sensor decreases to 4601.63 μm. When it rotates to position 337.5° (position 16), the upper sensor reading is 4613.40 μm, and the lower sensor reading is 4611.64 μm. After the acquisition cycle ends, the system integrates these 16 pairs of sensor distance readings, which show obvious vertical fluctuations and reflect mechanical runout, in order of position number, and saves them completely in the system cache as the sampling data of the wafer under test.

[0032] 103. Using the calibration reference data, perform differential calculations on the distance readings at each location to obtain the wafer thickness value corresponding to each location, and summarize them to form a thickness dataset.

[0033] It should be noted that after acquiring 16 sets of sampling data with mechanical runout characteristics, the system controller retrieves the calibration reference data stored in step 101, calculates the total fixed physical installation distance of the two sensors as 10000.00μm, and retrieves the compensation parameter of 0.05μm that needs to be deducted.

[0034] For each pair of readings cached in step 102, the system executes a unified differential operation logic: from the total physical spacing of the sensors, the system subtracts the corresponding upper sensor distance reading and lower sensor distance reading to obtain the preliminary thickness; then, it subtracts the compensation parameter to obtain the actual wafer thickness. This dual-sided, through-beam differential algorithm can directly offset the mechanical runout of the wafer in the Z-axis direction, because when the wafer shifts upward or downward, the decrease in reading of one sensor is necessarily equal to the increase in reading of the other sensor.

[0035] Taking the sixth position (112.5°) with the greatest mechanical runout as an example, the system uses 10000.00 μm as a base, subtracting the significantly reduced upper sensor reading of 4600.40 μm and the significantly increased lower sensor reading of 4624.47 μm at that position. Although the readings on one side change drastically, the calculated preliminary thickness remains stable at 775.13 μm. The system subtracts a compensation parameter of 0.05 μm from this preliminary thickness to obtain the true wafer thickness value of 775.08 μm at the sixth position.

[0036] For the 12th position (247.5°) with the most severe downward sway, the system subtracts the increased upper sensor reading of 4623.50 μm and the decreased lower sensor reading of 4601.63 μm from 10000.00 μm, resulting in a preliminary thickness calculation of 774.87 μm. After deducting a 0.05 μm compensation parameter, the wafer thickness at the 12th position is 774.82 μm. For the initial 1st position (0°), the base of 10000.00 μm is subtracted from the upper sensor reading of 4612.45 μm and the lower sensor reading of 4612.60 μm, followed by a further subtraction of 0.05 μm, resulting in a thickness of 774.90 μm. The system utilizes the underlying processor to perform parallel calculations for all 16 positions, sequentially summarizing all calculated true thickness values ​​(after removing jitter interference) to obtain a thickness dataset belonging to the wafer under test.

[0037] 104. Extract the maximum and minimum values ​​from the thickness dataset, compare the maximum value with the upper limit of the preset thickness tolerance, and compare the minimum value with the lower limit of the preset thickness tolerance to determine whether the wafer is qualified and obtain the judgment result.

[0038] It should be noted that the preset thickness tolerance parameters for this batch of 12-inch wafers are read from the process recipe file of the Manufacturing Execution System (MES) via the factory intranet. The target baseline thickness for the current process node is set to 775.00 μm, and the allowable total manufacturing thickness deviation is controlled within ±0.50 μm. Based on this, the system configures the upper limit of the preset thickness tolerance to 775.50 μm and the lower limit of the preset thickness tolerance to 774.50 μm in the comparator module.

[0039] The system's main program performs a global traversal scan of the dataset containing 16 specific thickness values ​​obtained in step 103. After addressing and value magnitude verification, the program successfully identifies the extreme thickness values ​​along the entire wafer edge scanning trajectory. In this embodiment, the system extracts a maximum thickness value of 775.08 μm (this value appears at the 6th acquisition position); simultaneously, it extracts a minimum thickness value of 774.82 μm (this value appears at the 12th acquisition position). This extraction method directly covers the thickest and thinnest extreme cases of the wafer, allowing for the assessment of the thickness consistency of the entire wafer without manual intervention.

[0040] The logic control unit sends the extracted maximum value of 775.08 μm to the upper limit comparator and compares it with the upper limit value of 775.50 μm to confirm that the maximum value has not exceeded the upper limit red line. It then sends the extracted minimum value of 774.82 μm to the lower limit comparator and compares it with the lower limit value of 774.50 μm to confirm that the minimum value has not fallen below the lower limit bottom line. Based on these two comparisons, the system confirms that all 16 measurement points in the thickness dataset are legally within the process window of 774.50 μm to 775.50 μm. The controller then draws a clear conclusion that the thickness dimension of the semiconductor wafer fully conforms to the established standard and writes the wafer's judgment result flag as qualified in the status register.

[0041] 105. When the judgment result is unqualified, an alarm is issued and a replacement wafer is triggered; when the judgment result is qualified, the thickness data is output and the wafer is released to enter subsequent processing.

[0042] It should be noted that the system's main control PLC will automatically select the corresponding execution logic based on the judgment result flag bit in the register of step 104.

[0043] When the judgment result is qualified: Following the aforementioned embodiment, the system confirms that the flag bit is qualified. The control software formats and packages the batch number, wafer tracking code, and thickness dataset at 16 locations of the wafer, uploads it to the MES (Manufacturing Execution System) via the SECS / GEM semiconductor communication protocol, and simultaneously renders and outputs a thickness data report of the wafer on the machine's HMI (Human-Machine Interface). The specific output data details are shown in Table 1 below: Table 1

[0044] After the data report is displayed, the system shuts off the vacuum solenoid valve at the bottom of the turntable, releasing the wafer. The main controller issues a release coordinate command to the robotic arm, which smoothly moves the wafer to the designated slot in the standard front-opening wafer transfer box, allowing it to proceed smoothly to the next processing step.

[0045] When the judgment result is unqualified: If, during subsequent testing, a wafer measures a maximum thickness of 775.60 μm, exceeding the upper limit of 775.50 μm, the system determines it to be unqualified. In this case, the main control system immediately triggers the machine's three-color alarm tower, illuminating a red strobe light and activating an intermittent alarm sound (80dB). Simultaneously, the system sends an interception command to the unloading robotic arm, placing the unqualified wafer into a dedicated scrap / rework cassette for physical isolation. To ensure production line capacity, the system immediately sends a replenishment command to the feeding port, driving the loading robotic arm to retrieve another untested wafer from the source material box and feed it into the testing station, automatically initiating a new round of thickness measurement. This entire diversion response mechanism achieves closed-loop automated management of the testing process.

[0046] Please see Figure 2 , Figure 4 Another embodiment of the semiconductor wafer thickness detection method in this invention includes: 201. Provide a calibration fixture and two through-beam mounted distance detection sensors. Place the calibration fixture between the two sensors and obtain the distance readings from the two sensors to the two sides of the calibration fixture. Obtain the standard thickness of the calibration fixture. Place a standard gauge block of known thickness between the two sensors for verification. Determine the compensation parameter based on the difference between the measured thickness of the standard gauge block and the standard thickness. Use the above distance readings, standard thickness, and compensation parameter as calibration reference data.

[0047] Specifically, the calibration fixture is placed between two sensors, and the sensor positions are adjusted so that the distance readings from the two sensors to the two sides of the calibration fixture are in the middle of their range. The distance readings of the two sensors at this time are recorded to obtain the calibration distance reading pair. The standard thickness value of the calibration fixture is read. A standard gauge block of known thickness is placed between the two sensors, and the distance readings from the two sensors to the two sides of the standard gauge block are recorded. The measured thickness of the standard gauge block is calculated based on the recorded distance readings, the calibration distance reading pair, and the standard thickness value. The difference between the standard thickness and the measured thickness of the standard gauge block is used as a compensation parameter. The calibration distance reading pair, the standard thickness value, and the compensation parameter are used together as calibration reference data and stored in the testing equipment.

[0048] It should be noted that a thickness measurement device for 12-inch semiconductor wafers is provided, which uses two laser displacement sensors (hereinafter referred to as the upper sensor and the lower sensor) that are aligned vertically and installed in opposite directions. The effective measurement range is set to 20mm to 30mm, with the middle of the range around 25mm (i.e., 25000μm).

[0049] A high-rigidity, rigorously calibrated metal block is provided as a calibration fixture, and the system reads the standard thickness value of this fixture as 5000.00 μm. This calibration fixture is then stably placed between the upper and lower sensors. To ensure the sensors operate within their optimal linearity measurement range, the device adjusts the physical mounting height of the two sensors using a Z-axis fine-tuning slide, ensuring that their distance readings to the upper and lower surfaces of the calibration fixture are both within the middle of their measurement range. After the mechanical structure comes to a stop and stabilizes, the system records the measurement data: the distance reading for the upper sensor is 25010.20 μm, and the distance reading for the lower sensor is 24989.80 μm. These two data points constitute the calibration distance reading pair.

[0050] The calibration fixture was removed from the measurement area, and a standard gauge block of known thickness was placed between the two sensors for verification. This standard gauge block was used to simulate the actual semiconductor wafer under test, and its known standard thickness was 775.00 μm. The system synchronously triggered and recorded the distance readings from the two sensors to the two surfaces of the standard gauge block: the distance reading of the upper sensor increased to 27122.35 μm, and the distance reading of the lower sensor increased to 27102.55 μm.

[0051] The measured thickness of the standard gauge block is calculated based on the recorded values. The specific logic is as follows: The upper sensor reading of 25010.20 μm and the lower sensor reading of 24989.80 μm are added to the calibration distance reading pair. Then, this is added to the standard thickness value of the calibration fixture, 5000.00 μm, to calculate the total physical zero-point distance between the two sensors as 55000.00 μm. From this total distance, the measured distance readings of the upper and lower sensors (27122.35 μm and 27102.55 μm respectively) when the standard gauge block was placed are subtracted sequentially to calculate the measured thickness of the standard gauge block as 775.10 μm.

[0052] Subsequently, the compensation parameter was determined. Based on the execution logic, the standard thickness of the standard gauge block (775.00 μm) was subtracted from the calculated measured thickness (775.10 μm), resulting in a difference of -0.10 μm. This difference comprehensively reflects the minute deformation caused by changes in ambient temperature and the overall system error of sensor installation; this -0.10 μm was thus established as the compensation parameter.

[0053] The calibration distance readings (i.e., 25010.20μm for the upper sensor and 24989.80μm for the lower sensor), the standard thickness value of 5000.00μm for the calibration fixture, and the calculated compensation parameter -0.10μm were combined as calibration reference data and stored in the non-volatile memory of the testing equipment, thus completing the calibration work based on the middle of the measurement range.

[0054] 202. Place the semiconductor wafer under test between two sensors, drive the wafer to rotate, and collect distance reading pairs at 16 locations on the circumference of the wafer as sampling data.

[0055] Specifically, the semiconductor wafer under test is placed on a rotating mechanism, positioning it between two sensors with the wafer surface perpendicular to the sensor measurement direction. The rotating mechanism is then activated, causing the wafer to rotate at a constant speed, while the two sensors synchronously acquire distance data. Sixteen sampling positions are determined based on the equal division angles of one revolution of the wafer. When a sampling position coincides with a notch or flat edge of the wafer, that position is skipped, and the sampling point is moved to the next equal division position until the sampling position no longer falls into the notch or flat edge area. This process is repeated until 16 valid sampling positions are determined. As the wafer sequentially reaches each sampling position, the distance readings of the two sensors are recorded, forming a distance reading pair. After one revolution, a sampling dataset containing 16 distance reading pairs is obtained, which is used for subsequent differential calculations.

[0056] Furthermore, distance reading pairs at 16 locations are uniformly collected on the circumference of the wafer. Specifically, this includes: determining a sampling position every 22.5 degrees of wafer rotation based on the angle signal of the rotating mechanism; when the sampling position coincides with a wafer notch or flat edge, skipping the position and extending the sampling point to the next equally divided position until the sampling position does not fall into the notch or flat edge area, and repeating this process until 16 valid sampling positions are determined. When the wafer reaches each sampling position, the two sensors continuously collect multiple distance readings to obtain the distance reading array of the first sensor and the distance reading array of the second sensor at that position. The maximum and minimum values ​​in the distance reading arrays of the first and second sensors are removed respectively, and the average of the remaining readings is taken to obtain the distance reading of the first sensor and the distance reading of the second sensor at that position. The distance readings of the first sensor and the distance readings of the second sensor are paired to form a distance reading pair for that position. The distance reading pairs corresponding to the 16 sampling positions are arranged according to the wafer rotation order to obtain the sampling dataset.

[0057] It should be noted that after storing the calibration reference data in step 201, the dynamic sampling and data preprocessing execution process is specifically explained based on the processing scenario of a 12-inch semiconductor wafer.

[0058] An automated robotic arm smoothly places the semiconductor wafer to be tested onto the vacuum chuck of the rotating mechanism. The system ensures that the wafer is positioned between the upper and lower sensors, and that the upper and lower surfaces of the wafer are orthogonal and perpendicular to the direction of the laser beam. The main control system activates the rotating mechanism, causing the wafer to rotate at a set speed of 60 r / min.

[0059] The sampling position is determined by dividing the rotation into equal angles of 22.5°. When the wafer is at the initial 0.0° position, the system identifies it as the first valid sampling position. When the angle signal from the rotation mechanism indicates that the wafer has reached 90.0°, i.e., the original equal division position, the photoelectric sensor detects that this position coincides with the V-notch at the edge of the wafer.

[0060] To avoid data distortion caused by the measurement spot falling into the gap, the control program triggers avoidance logic, automatically skipping the 90.0° overlapping position and postponing the 5th sampling point to the next 22.5° division position (i.e., 112.5°). The system repeats this process, successively using subsequent division points (135.0°, 157.5°, etc.) as valid sampling positions. To compensate for the missing point due to skipping the gap, after the first basic division, the test stage fine-tunes its angle, adding a 11.25° half-step division position (i.e., 348.75°) at the end of the rotation as the 16th sampling position, until all 16 valid sampling positions are completely determined on the wafer circumference.

[0061] As the wafer sequentially passes through each valid sampling position, the system hardware performs multiple consecutive samplings and software filtering. Taking the first sampling position (0.0°) as an example, the upper sensor continuously collects five distance readings, obtaining 27122.32μm, 27122.35μm, 27122.90μm, 27122.38μm, and 27121.10μm. The processor discards the maximum value of 27122.90μm and the minimum value of 27121.10μm, and averages the remaining three readings to obtain the effective distance reading of the first sensor as 27122.35μm. Simultaneously, after the lower sensor acquires five readings, extreme values ​​are removed and the average value is taken to obtain the effective distance reading of the second sensor as 27102.52 μm. The two are paired to form the distance reading pair for the first position.

[0062] When processing the fifth valid sampling position (112.5°), the wafer shifted at this position due to the inherent end-face runout of the rotating mechanism. The system repeated the continuous sampling and filtering process, calculating the effective sensor reading at this position to be 27118.24 μm for the upper sensor and 27106.60 μm for the lower sensor, which were then paired to form the reading pair for the fifth sampling position. The system also acquired data from the remaining positions (the 11th position at 247.5°, with the upper sensor reading at 27124.30 μm and the lower sensor reading at 27100.65 μm). After completing the full lap of sampling, the distance reading pairs corresponding to these 16 positions were sequentially packaged into a sampling dataset and temporarily stored in memory.

[0063] 203. Using the calibration reference data, perform differential calculations on the distance readings at each location to obtain the wafer thickness value corresponding to each location, and summarize them to form a thickness dataset.

[0064] Specifically, the distance readings at each sampling location are summed to obtain a set of readings and pairs; the calibration distance readings in the calibration reference data are summed, then summed with the standard thickness, and finally summed with the compensation parameter to obtain the reference spacing length; each reading and pair in the set of readings and pairs is subtracted from the reference spacing length to obtain an initial thickness value set; the thickness values ​​in the initial thickness value set are arranged in ascending order to obtain a sorted thickness sequence; thickness values ​​that deviate from the median of the sorted thickness sequence and whose deviation exceeds a preset deviation threshold are removed to obtain a thickness dataset.

[0065] Further, the process of removing thickness values ​​from the sorted thickness sequence that deviate from the median of the sequence and whose deviation exceeds a preset deviation threshold specifically includes: determining the median of the sorted thickness sequence; calculating the absolute value of the deviation between each thickness value in the sorted thickness sequence and the median to obtain a deviation set; comparing each absolute value of the deviation set with a preset deviation threshold, and marking thickness values ​​whose absolute values ​​of deviation exceed the preset deviation threshold as abnormal thickness values; removing all abnormal thickness values ​​from the sorted thickness sequence to obtain a valid thickness sequence; and using the valid thickness sequence as a thickness dataset.

[0066] It should be noted that after obtaining the sampling dataset of 16 valid sampling locations, this embodiment specifically explains the difference operation and outlier filtering process.

[0067] The system reads the cached sampling dataset and sums the distance readings for each location. Taking the first location (0.0°) as an example, the upper sensor reading of 27122.35 μm is added to the lower sensor reading of 27102.52 μm, resulting in a sum of 54224.87 μm; the fifth location (112.5°) is summed to obtain 54224.84 μm; and the eleventh location (247.5°) is summed to obtain 54224.95 μm. The system performs this summation for all 16 locations, forming a complete set of reading sum pairs.

[0068] The system calls the calibration reference data, adds the calibration distance readings (25010.20μm and 24989.80μm) to get 50000.00μm, adds the calibration fixture thickness of 5000.00μm to get the total physical zero point distance of 55000.00μm, and then adds it to the compensation parameter of -0.10μm to calculate the reference distance length of 54999.90μm.

[0069] Performing differential calculations: at the first position (54999.90μm), subtracting 54224.87μm yields an initial thickness of 775.03μm; at the fifth position (112.5°), subtracting 54224.84μm yields 775.06μm; at the eleventh position (247.5°), subtracting 54224.95μm yields 774.95μm. Assuming that tiny dust particles adhere to the wafer surface at the ninth sampling position (202.5°), causing the sensor reading to decrease to 54221.40μm, the initial thickness at this position is calculated to suddenly increase to 778.50μm.

[0070] The 16 thickness values ​​in the initial thickness value set were sorted in ascending order. The sorted sequence is as follows: 774.95μm, 774.96μm, 774.97μm, 774.98μm, 774.98μm, 774.99μm, 775.00μm, 775.00μm, 775.01μm, 775.01μm, 775.02μm, 775.03μm, 775.04μm, 775.05μm, 775.06μm, and the abnormal protrusion value of 778.50μm.

[0071] To eliminate false interference caused by dust, the system determines the median of the sorted thickness sequence. Since there are 16 items, the median is the arithmetic mean of the two middle items (item 8, 775.00 μm and item 9, 775.01 μm), which is 775.005 μm. The system calculates the absolute value of the deviation of each thickness value from the median; the absolute deviation for item 1 (774.95 μm) is 0.055 μm, while the absolute deviation for item 16 (778.50 μm) is as high as 3.495 μm.

[0072] The system compares the absolute values ​​of the above deviations with a preset deviation threshold of 1.00 μm. The deviations of the first 15 values ​​are all less than 1.00 μm, while the deviation of the 16th value, 3.495 μm, exceeds the threshold. The system marks the corresponding thickness value of 778.50 μm as an abnormal thickness value. If the abnormal point is removed, the system should automatically trigger the robotic arm to perform a second retest of the area near that location. If the retest data is normal, the abnormal value is replaced with the retest value; if retesting is not possible, the wafer is marked as 'requiring re-inspection' instead of directly participating in the pass / fail judgment. In this embodiment, it is assumed that the system obtains a normal thickness value of 775.02 μm after performing a second retest. This retest value is used to replace the abnormal value, thereby restoring and retaining 16 valid measurement data points, which are used as the official thickness dataset.

[0073] 204. Extract the maximum and minimum values ​​from the thickness dataset, compare the maximum value with the upper limit of the preset thickness tolerance, and compare the minimum value with the lower limit of the preset thickness tolerance to determine whether the wafer is qualified and obtain the judgment result.

[0074] Specifically, the thickness dataset is traversed, and the maximum and minimum thickness values ​​are determined by comparing each item. The maximum thickness value is compared with the upper limit of the preset thickness tolerance to obtain the upper limit comparison result. The minimum thickness value is compared with the lower limit of the preset thickness tolerance to obtain the lower limit comparison result. The difference between the maximum and minimum thickness values ​​is calculated to obtain the thickness range. The thickness range is compared with the preset thickness uniformity threshold to obtain the range comparison result. When the upper limit comparison result, the lower limit comparison result, and the range comparison result all meet the preset conditions, the wafer is judged to be qualified; otherwise, the wafer is judged to be unqualified, and the judgment result is obtained.

[0075] Furthermore, the thickness dataset is traversed, and the maximum and minimum thickness values ​​are determined by comparing each item. Specifically, this includes: marking each thickness value with a corresponding sampling position identifier while traversing the thickness dataset to obtain the correspondence between thickness values ​​and sampling position identifiers; marking the first thickness value in the thickness dataset as both the current maximum and minimum value, and recording the sampling position identifiers corresponding to the current maximum and minimum values; sequentially reading the remaining thickness values ​​in the thickness dataset, comparing each read thickness value with the current maximum value, and updating the current maximum value and its corresponding sampling position identifier if it is greater than the current maximum value; comparing each read thickness value with the current minimum value, and updating the current minimum value and its corresponding sampling position identifier if it is less than the current minimum value. After traversal, the maximum thickness and its corresponding sampling position identifier, and the minimum thickness and its corresponding sampling position identifier are obtained. Based on the sampling position identifiers corresponding to the maximum and minimum thickness, the angle between them on the wafer circumference is determined. The difference between the maximum and minimum thickness is taken as the thickness range. The thickness range is compared with a preset thickness uniformity threshold, and the angle is compared with a preset angle threshold. When the thickness range exceeds the preset thickness uniformity threshold, the wafer is deemed unqualified. When the thickness range does not exceed the preset thickness uniformity threshold but the angle exceeds the preset angle threshold, the wafer is deemed to have local thickness anomalies and is deemed unqualified accordingly. Other cases are determined by combining the upper limit comparison result and the lower limit comparison result.

[0076] It should be noted that after obtaining the dataset containing 16 valid thickness values, the specific explanation of extreme value extraction, multidimensional comparison, and adjudication logic will be provided.

[0077] The system retrieves quality judgment benchmark data from the process formula: the preset upper limit of thickness tolerance is 775.50μm, the preset lower limit of thickness tolerance is 774.50μm; the preset thickness uniformity threshold is 0.30μm; and the preset angle threshold is 160.0°.

[0078] The traversal and extraction program is initiated. Based on the timestamps and indices of the underlying data, the system reconstructs and labels the corresponding physical angles for each thickness value in the dataset. The first item in the dataset is simultaneously marked as both the current maximum and minimum value. Subsequently, the remaining 15 items are read sequentially, and a bubble comparison is performed on each item: if the newly read value is greater than the current maximum value, the maximum value and its label are updated; if it is less than the current maximum value, the minimum value and its label are updated. After the traversal is completed, the maximum thickness of the wafer is locked at 775.06 μm, corresponding to a sampling position label of 112.5°; the minimum thickness is 774.95 μm, corresponding to a sampling position label of 247.5°.

[0079] In the comparison process, the maximum thickness of 775.06 μm did not exceed the upper tolerance limit of 775.50 μm, and the upper limit comparison result met the condition; the minimum thickness of 774.95 μm did not fall below the lower limit of 774.50 μm, and the lower limit comparison result met the condition.

[0080] The system calculated the thickness range to be 0.11 μm (775.06 μm - 774.95 μm). This range does not exceed the uniformity threshold of 0.30 μm, and the range comparison result meets the condition. The system calculated the angle between the maximum and minimum thickness values ​​on the wafer circumference to be 135.0° (|247.5° - 112.5°| = 135.0°).

[0081] Since the calculated thickness range of 0.11 μm meets the requirements, and the included angle of 135.0° does not exceed the preset angle threshold of 160.0°, the local thickness anomaly judgment is not triggered. Combining the upper and lower limit results, the system concludes that the semiconductor wafer is qualified.

[0082] 205. When the judgment result is unqualified, an alarm is issued and a replacement wafer is triggered; when the judgment result is qualified, the thickness data is output and the wafer is released to enter subsequent processing.

[0083] Specifically, when the judgment result is unqualified, an alarm message is obtained based on the maximum value, minimum value and thickness range in the thickness data set. The alarm message includes the wafer identification, thickness data and the direction of the out-of-tolerance. The alarm message is displayed on the operation interface and an audible and visual alarm is triggered at the same time to remind the operator to re-inspect. The operator performs a manual re-inspection of the wafer. After confirming that the wafer thickness exceeds the preset tolerance, the operator enters a replacement wafer confirmation command through the operation interface. The equipment receives a replacement wafer confirmation command, triggering the wafer retrieval device to re-execute the wafer retrieval operation and marking the original defective wafer as a pending product. At the same time, the alarm information and the replacement wafer confirmation command are associated with the wafer batch information and stored to form a defect record. When the judgment result is qualified, a thickness report is obtained based on the thickness dataset. The thickness report includes the maximum, minimum and average thickness values. The thickness report is associated with the wafer batch information and stored, and a release command is obtained to control the wafer to be transferred to the subsequent processing station.

[0084] It should be noted that this embodiment illustrates the complete electromechanical collaborative execution process regarding result response, data output, and anomaly handling.

[0085] When the judgment result is qualified: The current wafer (assuming wafer identification is W-20260818-01) is confirmed as qualified. The main control system immediately extracts the 16 valid thickness values ​​obtained in step 203 and calculates their overall arithmetic mean to be approximately 775.00 μm (rounded to two decimal places). Based on the extracted maximum thickness value of 775.06 μm, minimum thickness value of 774.95 μm, and this average value, the system generates a standard thickness report. The core data of this thickness report is shown in Table 2 below: Table 2

[0086] The data from the thickness report are deeply correlated and stored with the current wafer batch information in the Manufacturing Execution System (MES) via the factory's internal network. After data archiving is complete, the main control PLC receives a release command. The transfer robot of the inspection machine receives the command, releases the vacuum fixation of the wafer on the turntable, smoothly delivers the wafer out of the testing area, and transfers it to the subsequent chemical mechanical polishing (CMP) station for the next step of processing.

[0087] When the judgment result is unqualified: In continuous automated production, the machine inspects the next wafer (identified as W-20260818-02) within the batch. System calculations show that its maximum thickness is 775.80 μm, exceeding the preset tolerance limit of 775.50 μm; the minimum thickness is 774.55 μm; the calculated thickness range is 1.25 μm, exceeding the preset thickness uniformity threshold of 0.30 μm. Therefore, the result is deemed unqualified.

[0088] The main control software responded immediately and received the corresponding alarm information, which clearly marked the chip identifier, the measured abnormal thickness data, and the direction of the out-of-tolerance error where the thickness was too high and the tolerance was exceeded.

[0089] The control system displays this alarm information as a red pop-up on the machine's main operating interface, simultaneously triggering the 3-color alarm light on the top of the machine to switch to red flashing mode, and activating the buzzer to emit an 80dB audible and visual alarm. Upon noticing the alarm, the on-site operator intervenes for re-inspection, confirming that the thickness parameter exceeds the preset tolerance, and then clicks to input a confirmation command for re-taking the sheet on the operating interface.

[0090] Upon receiving the instruction, the machine immediately triggers the loading and unloading device to pick up a new wafer from the next slot in the source material box and send it into the measurement area to execute a new thickness measurement process. The unloading device transfers the original defective wafer to a dedicated abnormal material cassette and marks it as a pending item in the system. All action instructions and alarm information are stored in the database to form a traceable record of non-conformities.

[0091] 206. After obtaining the thickness dataset and before extracting the maximum and minimum values ​​from the thickness dataset, obtain the ambient temperature during wafer inspection; determine the thickness correction amount based on the correspondence between ambient temperature and the thermal expansion coefficient of the wafer material; apply the thickness correction amount to each thickness value in the thickness dataset to obtain the temperature-compensated thickness dataset; extract the maximum and minimum values ​​from the temperature-compensated thickness dataset for judgment.

[0092] Furthermore, based on the correspondence between ambient temperature and the thermal expansion coefficient of the wafer material, the thickness correction amount is determined, including: pre-establishing a correspondence table between multiple temperature ranges and thickness correction amounts, and storing the correspondence table in the detection equipment; after obtaining the ambient temperature during wafer detection, determining the temperature range into which the ambient temperature falls; searching for the thickness correction amount corresponding to the temperature range in the correspondence table, and using the found thickness correction amount as the thickness correction amount; applying the thickness correction amount to each thickness value in the thickness dataset to obtain the temperature-compensated thickness dataset.

[0093] It should be noted that after obtaining the thickness dataset containing 16 valid measurement data in step 203, and before extracting the maximum and minimum values ​​in step 204, this embodiment specifically describes the execution logic for ambient temperature acquisition and thickness data temperature compensation.

[0094] To address the microscopic thermal expansion and contraction effects of silicon-based semiconductor materials at different temperatures, and the inherent comprehensive thermal drift characteristics of the internal metal frame of the testing equipment under varying ambient temperatures, engineers pre-calculated and established a correspondence table between multiple temperature ranges consisting of five consecutive spans and thickness correction values. This correspondence table was statically distributed and securely stored in the non-volatile storage module of the testing equipment's main control board. The specific corresponding data is shown in Table 3 below: Table 3

[0095] In the actual automated testing process, as soon as the dataset from step 203 is obtained, the system's main control unit immediately wakes up the high-precision digital temperature sensor installed inside the optical measurement chamber of the machine to acquire the real-time ambient temperature of the wafer during testing. In this measurement cycle of this embodiment, the current ambient temperature measured by the sensor is stable at 20.2℃.

[0096] The software executes interval matching scanning logic to determine that the ambient temperature of 20.2℃ falls exactly within the constant standard temperature range of 20.0≤T<20.5 in the corresponding relationship table. The program then locates the corresponding temperature range in the table according to this mapping relationship and accurately extracts the thickness correction amount of 0.00μm.

[0097] The system processor uses the found 0.00μm thickness correction as a compensation operator for each independent thickness value in the thickness dataset compiled in step 203, performing a step-by-step numerical addition operation. Because the temperature-controlled environment of the factory where the machine is operating is extremely stable, and the cabin temperature is precisely within the optimal system zero-bias temperature range, the applied correction is zero. After the compensation operation, the thickness values ​​in the dataset retain their original absolute values.

[0098] The first thickness value in the thickness dataset is 774.95 μm. After adding a correction of 0.00 μm, the result is still 774.95 μm. The largest thickness value in the dataset is 775.06 μm. After adding a correction of 0.00 μm, the result is still 775.06 μm.

[0099] After the system performs the above-mentioned application calculations on the data from all 16 valid measurement positions at high speed, it obtains a temperature-compensated thickness dataset with a compensation attribute identifier. The system then safely transfers this temperature-compensated dataset to the execution module in step 204, where it extracts extreme values ​​and performs subsequent tolerance comparisons and quality judgments, thereby eliminating the interference of temperature fluctuations on precision measurements at both the physical and logical levels.

[0100] The semiconductor wafer thickness detection method in the embodiments of the present invention has been described above. The semiconductor wafer thickness detection system in the embodiments of the present invention will be described below. Please refer to [link / reference]. Figure 3One embodiment of the semiconductor wafer thickness detection system of the present invention includes: a reference determination module 301, used to acquire measurement data of a calibration fixture based on two through-beam distance detection sensors to determine calibration reference data; a sampling acquisition module 302, used to acquire distance reading pairs at multiple sampling positions as sampling data when the semiconductor wafer under test moves relative to the two distance detection sensors; a calculation and summarization module 303, used to calculate the sampling data using the calibration reference data to obtain the thickness value of the semiconductor wafer under test at the multiple sampling positions, and summarize it to form a thickness dataset; a feature judgment module 304, used to extract thickness feature values ​​based on the thickness dataset, compare the thickness feature values ​​with preset judgment conditions, judge whether the semiconductor wafer under test is qualified, and obtain a judgment result; and a transfer execution module 305, used to execute corresponding wafer transfer operations based on the judgment result.

[0101] The present invention also provides a semiconductor wafer thickness detection device, which includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor performs the steps of the semiconductor wafer thickness detection method in the above embodiments.

[0102] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the semiconductor wafer thickness detection method.

[0103] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0104] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0105] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting the thickness of a semiconductor wafer, characterized in that, Includes the following steps: A calibration fixture and two through-beam mounted distance detection sensors are provided to acquire measurement data from the calibration fixture in order to determine calibration reference data; The semiconductor wafer under test is placed between the two distance detection sensors, causing relative motion between the semiconductor wafer under test and the two distance detection sensors, and multiple pairs of distance readings at sampling positions are collected as sampling data; The sampled data is calculated using the calibration reference data to obtain the thickness values ​​of the semiconductor wafer under test at the multiple sampling locations, and the results are summarized to form a thickness dataset. Thickness feature values ​​are extracted from the thickness dataset, and the thickness feature values ​​are compared with preset judgment conditions to determine whether the semiconductor wafer under test is qualified, and a judgment result is obtained. Based on the judgment result, the corresponding wafer transfer operation is performed.

2. The semiconductor wafer thickness detection method according to claim 1, characterized in that, include: Drive the semiconductor wafer under test to rotate; Multiple sampling positions are uniformly set on the circumference of the semiconductor wafer under test; When the set sampling position coincides with the notch or flat edge of the semiconductor wafer under test, the sampling position is extended to the next position until it does not fall into the notch or flat edge area, thus obtaining multiple valid sampling positions; When the semiconductor wafer under test reaches each of the effective sampling positions, the distance readings of the two distance detection sensors are recorded and paired to form corresponding distance reading pairs, thereby obtaining sampling data containing the distance reading pairs of all the effective sampling positions.

3. The semiconductor wafer thickness detection method according to claim 2, characterized in that, The step of recording the distance readings of the two distance detection sensors when the semiconductor wafer under test reaches each of the effective sampling positions, and pairing them to form corresponding distance reading pairs, includes: Upon reaching each of the effective sampling positions, the two distance detection sensors continuously collect multiple distance readings. The maximum and minimum values ​​of the multiple distance readings collected by each distance detection sensor are removed, and the average value of the remaining distance readings is taken to obtain the effective distance readings of the two distance detection sensors at the effective sampling position. The effective distance readings of the two distance detection sensors are paired to form a distance reading pair for the effective sampling position.

4. The semiconductor wafer thickness detection method according to claim 1, characterized in that, include: Add the distance readings for each sampling location in the sampled data to obtain a set of readings and pairs; The reference spacing length is calculated based on the calibration reference data; Subtract each reading and pair from the set of readings and pairs from the reference spacing length to obtain the initial thickness value set; Abnormal thickness values ​​that deviate from the median of the initial thickness value set and exceed a preset deviation threshold are removed, and the remaining thickness values ​​are used as the thickness dataset.

5. The semiconductor wafer thickness detection method according to claim 1, characterized in that, Also includes: Obtain the ambient temperature; The thickness correction amount is determined based on the preset correspondence between the ambient temperature and the thermal expansion coefficient of the wafer material; The thickness correction amount is applied to each thickness value in the thickness dataset to obtain a temperature-compensated thickness dataset for subsequent extraction of thickness feature values.

6. The semiconductor wafer thickness detection method according to claim 1, characterized in that, The thickness feature values ​​include the maximum thickness, the minimum thickness, and the thickness range; the preset judgment conditions include the preset upper limit of thickness tolerance, the preset lower limit of thickness tolerance, and the preset thickness uniformity threshold. The step of extracting thickness feature values ​​from the thickness dataset, comparing the thickness feature values ​​with preset judgment conditions, and determining whether the semiconductor wafer under test is qualified to obtain a judgment result includes: Extract the maximum and minimum thickness values ​​from the thickness dataset; The maximum thickness is compared with the preset upper limit of thickness tolerance, and the minimum thickness is compared with the preset lower limit of thickness tolerance. Calculate the difference between the maximum thickness and the minimum thickness to obtain the thickness range, and compare the thickness range with the preset thickness uniformity threshold; When all the above comparison results meet the preset judgment conditions, the semiconductor wafer under test is determined to be qualified; otherwise, it is determined to be unqualified, and a judgment result is generated.

7. The semiconductor wafer thickness detection method according to claim 6, characterized in that, The step of extracting the maximum and minimum thickness values ​​from the thickness dataset further includes: When extracting the maximum and minimum thickness values, the sampling position identifiers corresponding to the maximum and minimum thickness values ​​are recorded respectively; Based on the sampling location identifier, the angle between the maximum thickness and the minimum thickness on the circumference of the semiconductor wafer under test is determined; The method of determining whether the semiconductor wafer under test is qualified also includes: When the thickness range does not exceed the preset thickness uniformity threshold, but the included angle exceeds the preset angle threshold, it is determined that the semiconductor wafer under test has a local thickness abnormality and is therefore deemed unqualified.

8. The semiconductor wafer thickness detection method according to claim 1, characterized in that, The step of performing the corresponding wafer transfer operation based on the judgment result includes: When the judgment result is unqualified, an alarm is triggered and a replacement wafer instruction is generated, marking the semiconductor wafer under test as an abnormal product for subsequent processing. When the judgment result is qualified, an analysis report containing the thickness dataset is output, and a release instruction is generated to control the semiconductor wafer under test to be transferred to the subsequent processing steps.

9. A semiconductor wafer thickness detection system, characterized in that, include: The reference determination module is used to acquire measurement data from the calibration fixture based on two through-beam distance detection sensors in order to determine the calibration reference data. The sampling and acquisition module is used to acquire distance reading pairs at multiple sampling positions as sampling data when the semiconductor wafer under test and the two distance detection sensors generate relative motion. The calculation and summarization module is used to calculate the sampled data using the calibration reference data to obtain the thickness values ​​of the semiconductor wafer under test at the multiple sampling locations, and summarize them to form a thickness dataset; The feature judgment module is used to extract thickness feature values ​​based on the thickness dataset, compare the thickness feature values ​​with preset judgment conditions, and determine whether the semiconductor wafer under test is qualified, thereby obtaining a judgment result. The wafer transfer execution module is used to perform corresponding wafer transfer operations based on the judgment result.

Images