Semiconductor wafer pick-and-place device and pick-and-place method

Abstract

The present invention provides a semiconductor wafer pick-and-place device and method, belonging to the field of semiconductor pick-and-place technology. The device includes: a laser detection module for scanning the pick-and-place port and relevant areas inside the FOUP where the wafer is located using a three-dimensional laser illuminator; an adjustment and analysis module for performing unit length cuts along the central axis of the pick-and-place wafer routing path and analyzing each cutting unit to determine whether the fingertip span needs to be adjusted; a motion adjustment module for determining a safe pick-and-place starting position based on wafer deformation, controlling the fingertips to perform pick-and-place actions according to the pick-and-place wafer routing path, monitoring the fingertips' pick-and-place actions in real time, adjusting the pick-and-place actions if there are changes in the wafer's position or shape, and controlling the fingertips' rhythm of lifting the wafer based on differences in support point distances; and a verification module for controlling the laser detection module to perform another scan and verification after pick-and-place is completed. This improves the yield rate of wafer pick-and-place.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor pick-and-place technology, and in particular to a semiconductor wafer pick-and-place device and a pick-and-place method. Background Art

[0002] Wafer handling is a critical step in the semiconductor wafer production process. Current handling mechanisms suffer from significant flaws. First, the fingertip openings are poorly designed. Too small an opening can easily cause the wafer edge to sag and come into contact with the handling components (fingertips). Enlarging the opening to a more appropriate size (e.g., 200mm) limits the robot's path, making it prone to impacting the wafer transport pod (FOUP) during handling. Second, during the handling process, fingertips enter at the point of maximum wafer deformation and differing support point distances create collision risks. Existing devices and methods struggle to effectively address these issues, necessitating innovative designs to optimize the wafer handling process and ensure production safety and efficiency.

[0003] Therefore, the present invention provides a semiconductor wafer pick-and-place device and a pick-and-place method. Summary of the Invention

[0004] The present invention provides a semiconductor wafer pick-and-place device and a pick-and-place method to solve the above-mentioned technical problems.

[0005] The present invention provides a semiconductor wafer pick-and-place device, comprising:

[0006] The laser detection module is used to perform three-dimensional laser scanning of the pick-and-place port and the relevant area inside the FOUP where the wafer is located based on a three-dimensional laser illuminator, and pre-process the detection data to reconstruct the pick-and-place structure model;

[0007] A path planning module is used to plan a pick-and-place routing path based on the pick-and-place structure model, wafer, and FOUP parameters;

[0008] An adjustment and analysis module is used to perform unit length cutting along the central axis of the pick-and-place piece routing path, and analyze each cutting unit to determine whether the fingertip span needs to be adjusted. If necessary, the fingertip assembly is driven to adjust to the ideal span;

[0009] The action adjustment module is used to determine the safe pick-and-place starting position based on the wafer deformation, and control the fingertips to perform the pick-and-place action according to the pick-and-place wafer routing path. The fingertips perform the pick-and-place action in real time. If there are changes in the wafer position or shape, the pick-and-place action is adjusted, and the rhythm of the fingertips lifting the wafer is controlled based on the difference in the support point distance;

[0010] The verification module is used to control the laser detection module to scan again after the placement is completed to verify the wafer placement status and store the placement process data for subsequent process optimization and device parameter adjustment;

[0011] Among them, the wafer manufacturing process includes: wafer boat support structure and fingertip structure;

[0012] The wafer boat support structure is developed and designed based on a set platform, which is compatible with the standard wafer boat installation holes, and the lifting teeth on both sides of the wafer boat adopt a circular arc long tooth structure;

[0013] At the same time, the length of rod-1 and rod-3 supports is increased based on the standard crystal boat structure;

[0014] The fingertip structure adopts a 4-point lifting structure, combined with an arc fork finger shape, and increases the span of the fingertips to at least 200 mm.

[0015] Preferably, the laser detection module includes:

[0016] An initial construction unit, configured to construct original laser coordinates and map detection data to the original laser coordinates to obtain an initial structure;

[0017] A first interference unit is configured to perform three-dimensional substructure decomposition on the initial structure, obtain point cloud data of each three-dimensional substructure, and calculate an interference coefficient of each position point to construct an interference set;

[0018] an interference cancellation unit, configured to perform interference cancellation processing based on the interference set to obtain a first structure;

[0019] a distance difference construction unit, configured to respectively obtain a two-dimensional array of edge positions of each connecting line on the three-dimensional substructure, and perform distance difference construction on the two-dimensional array of edge positions to obtain a distance difference vector;

[0020] The second interference unit is used to calculate the fitting value of each distance difference vector to obtain a reference set of each three-dimensional substructure;

[0021] An adjustment unit is configured to modify the interference set based on the reference set to determine an adjustment position point, and adjust the first structure to obtain a second structure, wherein the second structure is a pick-and-place structure model.

[0022] Preferably, the adjustment analysis module includes:

[0023] The extraction unit is used to extract the central axis L of the pick-and-place routing path, establish a path coordinate system, and perform unit length calculation along the direction of the central axis L. Cutting to form n cutting units Ci, where i=1,2,…,n;

[0024] Constraint establishment unit, used to establish the spatial constraint equation for each cutting unit Ci :

[0025]

[0026] in, Represents the three-dimensional space area corresponding to the cutting unit Ci; , Represents the boundary constraint values ​​based on the y-axis and z-axis respectively; Represents the constraint value of the x-axis of the cutting unit Ci; Represents spatial coordinates;

[0027] Function definition unit, used to define the spatial fitness function:

[0028]

[0029] in, 、 、 is the weight coefficient; It represents the safe distance between the fingertip span d and the space boundary; represents the support stability function, which is related to the support point spacing and the center of gravity distribution of the wafer; represents the deformation influence function, considering the inhibitory effect of span on wafer deformation;

[0030] The span solving unit is used to solve the optimal span d_i* for each cutting unit Ci:

[0031] d_i*=argmaxFi,dmin <d_i*<dmax

[0032] Where dmin and dmax are the span ranges limited by the physical limitations of the device;

[0033] Function building unit, used to build span adjustment cost function:

[0034]

[0035] in, is the span change cost coefficient; is the time cost coefficient; is the actual span value of the cutting unit Ci; Time spent adjusting for span;

[0036] Sequence solving unit, used to solve the global optimal span sequence using dynamic programming algorithm: .

[0037] Preferably, the action adjustment module includes:

[0038] The starting position determination unit is used to obtain the three-dimensional topography of the wafer, build a deformation model, calculate the deformation of each area, and select the area with the smallest deformation as the safe pick-and-place starting position;

[0039] Threshold establishment unit, used to establish a safety threshold function: ,in, is the warpage; is the curvature; 、 is the weight coefficient;

[0040] Trajectory generation unit, used to plan the pick-and-place path based on the safe starting position and safety threshold function, and generate the optimal trajectory curve .

[0041] Preferably, the action adjustment module further includes:

[0042] Support point optimization unit, used to analyze support point distance differences and establish a mechanical balance model: ,in, represents the supporting force at point i; G is the weight of the wafer; is uida stress; is the penalty coefficient; n represents the total number of support point symbols;

[0043] The adaptive adjustment unit is used to build an adaptive control algorithm through real-time monitoring through multi-sensor fusion during the pick and place process: ,in, is the error vector, and They are wafer position offset, shape change, and support point force distribution; 、 、 is the PID parameter matrix;

[0044] The control unit is used to control the rhythm of fingertip lifting according to the difference in support point distance and real-time force conditions.

[0045] Preferably, the control unit includes:

[0046] Rhythm generation unit for online optimization of the stage target force of piecewise functions based on reinforcement learning and time constant , generating the fingertip lifting rhythm Ft;

[0047]

[0048] Where t represents the time point during the fingertip placement and removal process; k0 represents the k0th segment.

[0049] Preferably, it also includes:

[0050] The sequence determination unit is used to determine the force application order of the support points through the distance-force coordination priority function Pj:

[0051]

[0052] Wherein, j is the support point number; 、 is the dynamic weight; is the distance difference between the jth support point and other support points; is the maximum value of the distance difference among all support points; is the force deviation of the j-th support point; It is the maximum value of the force deviation of all support points; the dynamic optimization unit is used for real-time feedback correction and dynamic optimization of force control parameters to suppress wafer deformation.

[0053] The present invention provides a semiconductor wafer pick-and-place method, comprising:

[0054] Step 1: Use a 3D laser illuminator to perform 3D laser scanning on the pick-and-place port and the relevant area inside the FOUP where the wafer is located, and pre-process the detection data to reconstruct the pick-and-place structure model;

[0055] Step 2: Plan the pick-and-place routing path based on the pick-and-place structure model, wafer, and FOUP parameters;

[0056] Step 3: Perform unit length cutting along the central axis of the pick-and-place routing path, and analyze each cutting unit to determine whether the fingertip span needs to be adjusted. If necessary, drive the fingertip assembly to adjust to the ideal span;

[0057] Step 4: Determine the safe pick-and-place starting position based on the wafer deformation, and control the fingertips to perform the pick-and-place action according to the pick-and-place routing path. Monitor the fingertips' pick-and-place action in real time. If there are changes in the wafer position or shape, adjust the pick-and-place action, and control the rhythm of the fingertips lifting the wafer based on the difference in the support point distance.

[0058] Step 5: After the placement is completed, the laser detection module is controlled to scan again to verify the wafer placement status and store the placement process data for subsequent process optimization and device parameter adjustment;

[0059] Among them, the wafer manufacturing process includes: wafer boat support structure and fingertip structure;

[0060] The wafer boat support structure is developed and designed based on a set platform, which is compatible with the standard wafer boat installation holes, and the lifting teeth on both sides of the wafer boat adopt a circular arc long tooth structure;

[0061] At the same time, the length of rod-1 and rod-3 supports is increased based on the standard crystal boat structure;

[0062] The fingertip structure adopts a 4-point lifting structure, combined with an arc fork finger shape, and increases the span of the fingertips to at least 200 mm.

[0063] Preferably, step 1 comprises:

[0064] Constructing laser original coordinates and mapping detection data to the laser original coordinates to obtain an initial structure;

[0065] Decomposing the initial structure into three-dimensional substructures to obtain point cloud data of each three-dimensional substructure, and calculating the interference coefficient of each position point to construct an interference set;

[0066] performing interference cancellation processing based on the interference set to obtain a first structure;

[0067] Obtain a two-dimensional array of edge positions of each connecting line on the three-dimensional substructure, and construct a distance difference between the two-dimensional arrays of edge positions to obtain a distance difference vector;

[0068] Calculate the fitting value of each distance difference vector to obtain a reference set of each three-dimensional substructure;

[0069] The interference set is modified based on the reference set to determine an adjustment position point, and the first structure is adjusted to obtain a second structure, wherein the second structure is a pick-and-place structure model.

[0070] Preferably, step 2 includes:

[0071] Extract the central axis L of the pick-and-place routing path, establish a path coordinate system, and perform unit length along the direction of the central axis L. Cutting to form n cutting units Ci, where i=1,2,…,n;

[0072] For each cutting unit Ci, establish the spatial constraint equation :

[0073]

[0074] in, Represents the three-dimensional space area corresponding to the cutting unit Ci; , Represents the boundary constraint values ​​based on the y-axis and z-axis respectively; Represents the constraint value of the x-axis of the cutting unit Ci; Represents spatial coordinates;

[0075] Define the spatial fitness function:

[0076]

[0077] in, 、 、 is the weight coefficient; It represents the safe distance between the fingertip span d and the space boundary; represents the support stability function, which is related to the support point spacing and the center of gravity distribution of the wafer; represents the deformation influence function, considering the inhibitory effect of span on wafer deformation;

[0078] For each cutting unit Ci, solve the optimal span d_i*:

[0079] d_i*=argmaxFi,dmin <d_i*<dmax

[0080] Where dmin and dmax are the span ranges limited by the physical limitations of the device;

[0081] Establish the span adjustment cost function:

[0082]

[0083] in, is the span change cost coefficient; is the time cost coefficient; is the actual span value of the cutting unit Ci; Time spent adjusting for span;

[0084] Use dynamic programming algorithm to solve the global optimal span sequence: .

[0085] Compared with the prior art, the present invention has the following advantages:

[0086] Starting with the precise construction of a pick-and-place structure model using 3D lasers, planning a reasonable pick-and-place path, dynamically adjusting fingertip span, real-time monitoring and adjustment of pick-and-place movements, and finally verifying and accumulating data for optimization, the system can make the semiconductor wafer pick-and-place process more accurate and stable, reducing problems such as collisions and deformation, and improving wafer pick-and-place yield. Data accumulation can also be used to continuously optimize the overall semiconductor manufacturing process and equipment, improving the efficiency and quality of the entire semiconductor production process.

[0087] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0088] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0090] Figure 1 This is a structural diagram of a semiconductor wafer pick-and-place device according to an embodiment of the present invention;

[0091] Figure 2 This is a flow chart of a semiconductor wafer pick-and-place method according to an embodiment of the present invention;

[0092] Figure 3 A diagram of a pick-and-place structure in an embodiment of the present invention;

[0093] Figure 4 This is a laser measured structure diagram in an embodiment of the present invention;

[0094] Figure 5 This is a diagram showing fingertip lifting in an embodiment of the present invention;

[0095] Figure 6 A top view comparison diagram of the front end of a wafer sagging due to gravity in an embodiment of the present invention;

[0096] Figure 7 This is a diagram showing an embodiment of the arc long-tooth crystal boat of the present invention;

[0097] Figure 8 This is another embodiment of the arc long-tooth crystal boat of the present invention. DETAILED DESCRIPTION

[0098] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0099] The present invention provides a semiconductor wafer pick-and-place device, such as Figure 1 Shown, including:

[0100] The laser detection module is used to perform three-dimensional laser scanning of the pick-and-place port and the relevant area inside the FOUP where the wafer is located based on a three-dimensional laser illuminator, and pre-process the detection data to reconstruct the pick-and-place structure model;

[0101] A path planning module is used to plan a pick-and-place routing path based on the pick-and-place structure model, wafer, and FOUP parameters;

[0102] An adjustment and analysis module is used to perform unit length cutting along the central axis of the pick-and-place piece routing path, and analyze each cutting unit to determine whether the fingertip span needs to be adjusted. If necessary, the fingertip assembly is driven to adjust to the ideal span;

[0103] The action adjustment module is used to determine the safe pick-and-place starting position based on the wafer deformation, and control the fingertips to perform the pick-and-place action according to the pick-and-place wafer routing path. The fingertips perform the pick-and-place action in real time. If there are changes in the wafer position or shape, the pick-and-place action is adjusted, and the rhythm of the fingertips lifting the wafer is controlled based on the difference in the support point distance;

[0104] The verification module is used to control the laser detection module to scan again after the placement is completed, verify the wafer placement status and store the placement process data for subsequent process optimization and device parameter adjustment.

[0105] Preferably, the laser detection module includes:

[0106] An initial construction unit, configured to construct original laser coordinates and map detection data to the original laser coordinates to obtain an initial structure;

[0107] A first interference unit is configured to perform three-dimensional substructure decomposition on the initial structure, obtain point cloud data of each three-dimensional substructure, and calculate an interference coefficient of each position point to construct an interference set;

[0108] an interference cancellation unit, configured to perform interference cancellation processing based on the interference set to obtain a first structure;

[0109] a distance difference construction unit, configured to respectively obtain a two-dimensional array of edge positions of each connecting line on the three-dimensional substructure, and perform distance difference construction on the two-dimensional array of edge positions to obtain a distance difference vector;

[0110] The second interference unit is used to calculate the fitting value of each distance difference vector to obtain a reference set of each three-dimensional substructure;

[0111] an adjusting unit, configured to modify the interference set based on the reference set to determine an adjustment position point, and adjust the first structure to obtain a second structure, wherein the second structure is a pick-and-place structure model;

[0112] Among them, the wafer manufacturing process includes: wafer boat support structure and fingertip structure;

[0113] The wafer boat support structure is developed and designed based on the set platform, which is compatible with the standard wafer boat installation holes, and the lifting teeth on both sides of the wafer boat adopt an arc long tooth structure, such as Figure 7 and 8 shown.

[0114] At the same time, the length of rod-1 and rod-3 supports is increased based on the standard crystal boat structure;

[0115] The fingertip structure adopts a 4-point lifting structure, combined with an arc fork finger shape, and increases the span of the fingertips to at least 200 mm.

[0116] Preferably, the adjustment analysis module includes:

[0117] The extraction unit is used to extract the central axis L of the pick-and-place routing path, establish a path coordinate system, and perform unit length calculation along the direction of the central axis L. Cutting to form n cutting units Ci, where i=1,2,…,n;

[0118] Constraint establishment unit, used to establish the spatial constraint equation for each cutting unit Ci :

[0119]

[0120] in, Represents the three-dimensional space area corresponding to the cutting unit Ci; , Represents the boundary constraint values ​​based on the y-axis and z-axis respectively; Represents the constraint value of the x-axis of the cutting unit Ci; Represents spatial coordinates;

[0121] Function definition unit, used to define the spatial fitness function:

[0122]

[0123] in, 、 、 is the weight coefficient; It represents the safe distance between the fingertip span d and the space boundary; represents the support stability function, which is related to the support point spacing and the center of gravity distribution of the wafer; represents the deformation influence function, considering the inhibitory effect of span on wafer deformation;

[0124] The span solving unit is used to solve the optimal span d_i* for each cutting unit Ci:

[0125] d_i*=argmaxFi,dmin <d_i*<dmax

[0126] Where dmin and dmax are the span ranges limited by the physical limitations of the device;

[0127] Function building unit, used to build span adjustment cost function:

[0128]

[0129] in, is the span change cost coefficient; is the time cost coefficient; is the actual span value of the cutting unit Ci; Time spent adjusting for span;

[0130] Sequence solving unit, used to solve the global optimal span sequence using dynamic programming algorithm: .

[0131] Preferably, the action adjustment module includes:

[0132] The starting position determination unit is used to obtain the three-dimensional topography of the wafer, build a deformation model, calculate the deformation of each area, and select the area with the smallest deformation as the safe pick-and-place starting position;

[0133] Threshold establishment unit, used to establish a safety threshold function: ,in, is the warpage; is the curvature; 、 is the weight coefficient;

[0134] Trajectory generation unit, used to plan the pick-and-place path based on the safe starting position and safety threshold function, and generate the optimal trajectory curve .

[0135] Preferably, the action adjustment module further includes:

[0136] Support point optimization unit, used to analyze support point distance differences and establish a mechanical balance model: ,in, represents the supporting force at point i; G is the weight of the wafer; is uida stress; is the penalty coefficient; n represents the total number of support point symbols;

[0137] The adaptive adjustment unit is used to build an adaptive control algorithm through real-time monitoring through multi-sensor fusion during the pick and place process: ,in, is the error vector, and They are wafer position offset, shape change, and support point force distribution; 、 、 is the PID parameter matrix;

[0138] The control unit is used to control the rhythm of fingertip lifting according to the difference in support point distance and real-time force conditions.

[0139] Preferably, the control unit includes:

[0140] Rhythm generation unit for online optimization of the stage target force of piecewise functions based on reinforcement learning and time constant , generating the fingertip lifting rhythm Ft;

[0141]

[0142] Where t represents the time point during the fingertip placement and removal process; k0 represents the k0th segment.

[0143] Preferably, it also includes:

[0144] The sequence determination unit is used to determine the force application order of the support points through the distance-force coordination priority function Pj:

[0145]

[0146] Wherein, j is the support point number; 、 is the dynamic weight; is the distance difference between the jth support point and other support points; is the maximum value of the distance difference among all support points; is the force deviation of the j-th support point; It is the maximum value of the force deviation of all support points; the dynamic optimization unit is used for real-time feedback correction and dynamic optimization of force control parameters to suppress wafer deformation.

[0147] In this embodiment, the crystal boat can be represented by Boat, and the fingertip can be represented by Finger.

[0148] In this embodiment, the supporting teeth on both sides of the wafer boat adopt a circular arc long tooth structure, which can completely eliminate the contact friction with the back alloy of the wafer compared to the supporting teeth of an ordinary wafer boat.

[0149] Droop suppression effect: This structure can reduce the front-end droop of Taiko wafers caused by gravity by about 50%. For example, when the front-end droop of the wafer is 11.88mm due to ordinary wafer boats, the wafer boat with arc-shaped long teeth structure can reduce the droop to about 6.89mm. Figure 6 shown.

[0150] By optimizing the support structure, the wafer boat loading capacity is increased by approximately 25% compared to the standard boat design. The specific data is as follows:

[0151] When the wafer warpage is 2.5mm, the standard wafer boat can hold 100 wafers, while the special wafer boat can hold up to 125 wafers.

[0152] When the wafer warpage is 4mm, the standard wafer boat can hold 78 wafers, while the special wafer boat can hold up to 100 wafers.

[0153] When the wafer warpage is 6mm, the standard wafer boat can hold 65 wafers, while the special wafer boat can hold up to 85 wafers.

[0154] When the wafer warpage is 8mm, the standard wafer boat can hold 50 wafers, while the special wafer boat can hold up to 65 wafers.

[0155] The existing wafer boat cannot meet the support requirements of Taiko wafers, and the pitch value needs to be increased to about 15mm to adapt to the deformation and support requirements of the wafer.

[0156] In this embodiment, the newly designed fingers adopt a four-point lifting structure, combined with the arc-shaped forked fingers, so that the wafer contacts the fingers only at four points on the ring, reducing damage to the wafer surface.

[0157] Span design: To address the issue of the wafer edge sagging and contacting the fingertips due to the fingertip opening being too small when removing the wafer, the fingertip span is increased to at least 200mm to accommodate the size and deformation of Taiko wafers.

[0158] To address the risk of collision caused by fingertips entering the FOUP from the point of maximum wafer deformation when placing or taking wafers, the position and path of fingertips entering the FOUP are optimized to avoid entering from the point of maximum deformation.

[0159] Support point distance adjustment: Considering the problem of wafer lifting asynchrony caused by the difference in support point distance when taking the wafer, the fingertip motion control is optimized to ensure smooth wafer lifting and reduce the risk of collision, such as Figure 5 As shown, the pitch size > front end sag + wafer lift, where Wafer refers to the wafer.

[0160] The wafer boat support structure and wafer fingertip design used in wafer manufacturing effectively address issues such as deformation, damage, and collision of Taiko wafers during support and handling by optimizing support point distribution, adopting a circular arc-shaped long tooth structure, increasing support length, adjusting fingertip span, and optimizing the pick-and-place path. This improves wafer manufacturing yield and production efficiency. Furthermore, the wafer boat design increases wafer loading capacity and reduces production costs.

[0161] In this embodiment, for the laser measured structure such as Figure 4 As shown, the pick-and-place structure is as follows Figure 3 As shown, A01 indicates that the opening of the fingertip is too small, A02 indicates contact with the finger, and A03 indicates a 5.1mm sag here.

[0162] In this embodiment, a 3D laser illuminator is a device that emits laser light and receives reflected light to construct 3D spatial information. For example, in a semiconductor wafer handling scenario, this is like "taking a picture" of the FOUP interior and the access area, but using lasers to capture 3D data. Common industrial 3D laser scanners, like these, can accurately scan the 3D contours of an object.

[0163] A FOUP (Front Opening Unified Pod) is a wafer transfer box used to store and transport wafers in semiconductor manufacturing. For example, it acts as a protective box, safely transporting wafers from one process step to the next. Its internal structure supports the wafers and prevents damage during transport.

[0164] The pick-and-place structure model is a virtual model created by processing the scanned 3D data of the FOUP interior, the pick-and-place opening, and related areas. This model is used to simulate the pick-and-place environment and actions. For example, it can create a "digital twin" of the actual pick-and-place scenario, allowing subsequent pick-and-place path planning to be simulated within this model.

[0165] The original laser coordinates are a coordinate system established by the 3D laser illuminator itself, serving as the basis for the detection data. It's like drawing a coordinate system on paper. The data obtained by the laser scan is "placed" within this coordinate system to record its position. For example, the coordinates of a scanned point in this coordinate system are (x=10mm, y=20mm, z=30mm).

[0166] The initial structure is the preliminary three-dimensional structure obtained after the detection data is mapped to the original coordinates of the laser.

[0167] A 3D substructure is a smaller 3D component obtained by breaking down the initial structure. For example, if you split the 3D model of a FOUP's internal structure into the wafer support shelf and the box, each component is a 3D substructure, making it easier to analyze and process it separately.

[0168] Point cloud data is a collection of coordinates of numerous points generated by 3D laser scanning. These points form the outline of an object and other information. For example, when scanning a sphere, the point cloud data is the coordinates of the densely packed points on the sphere's surface, and these points can be used to reconstruct the sphere's shape.

[0169] The interference coefficient measures the degree of interference (such as noise and reflection anomalies) at each point in a 3D substructure. For example, if dust reflects the laser during scanning, causing inaccurate data at the corresponding point, the interference coefficient for that point will be high, making it unreliable to use it as a marker.

[0170] The interference set is a collection of relevant interference information such as the interference coefficients of all location points, marking points that may have problems.

[0171] The first structure is a three-dimensional structure after eliminating interference, which is closer to the actual situation than the initial structure, but may still have deficiencies and need further correction.

[0172] A connection line is a line that connects three-dimensional substructures. For example, where the support frame and the box body are connected, there is a virtual connection line to indicate their connection.

[0173] The edge position 2D array records the coordinates of the edge points of the connecting line on a 2D plane (which can be projected based on 3D coordinates). For example, project the connecting line onto a plane (such as the xy plane) and record the x and y coordinates of the edge points to form an array, which facilitates analysis of the connection.

[0174] The distance difference vector is calculated by calculating the distance difference between the data in the two-dimensional array of edge positions. It reflects the change in distance at the connection. For example, for points on both sides of a connection line, the distance difference between them in a certain direction is calculated. These differences are then combined into a vector to determine whether the connection is smooth.

[0175] Fitted values ​​are obtained by performing a mathematical fit (e.g., linear or curve fitting) on ​​the distance difference vectors. They are used to evaluate the regularity of the distance difference. For example, if the distance difference vector data roughly fits a straight line, the fitted values ​​are the relevant parameters of this line, such as the slope and intercept, which can be used to determine whether the connection is normal.

[0176] The reference set is a set consisting of information such as the fitting value of each three-dimensional substructure, which serves as the basis for correcting the interference set.

[0177] Second structure (pick and place structure model): After correction and adjustment, the model accurately reflects the pick and place scenario structure. It is the basis for subsequent steps and can more accurately simulate the actual pick and place environment.

[0178] In this embodiment, the pick-and-place routing path refers to the movement path of the tip when picking up and placing the wafer. For example, the path from the wafer's position in the FOUP to the wafer's processing position in the equipment must be planned reasonably to avoid collisions.

[0179] The central axis L is the central axis of the routing path for picking and placing pieces. It is an important reference for planning the path. The path is distributed around it. For example, if the routing path is a space similar to a pipe, the central axis is the line in the center of the pipe.

[0180] The path coordinate system is used to plan the routing path for pick-and-place applications. It facilitates describing the path's location and other information. Similar to a map's coordinate system, by placing the path in this coordinate system, you can use the coordinate values ​​to plan the movement of your fingertips.

[0181] Cutting units Ci are small units that divide the path along the central axis L. For example, a long path can be cut into small segments like sausages, with each segment being a cutting unit, making it easier to analyze them one by one.

[0182] The spatial constraint equation is an equation that describes the limitations of the three-dimensional space where the cutting unit Ci is located. For example, the cutting unit has limits in the x, y, and z directions in space, and this equation specifies these limits.

[0183] The spatial fitness function evaluates the suitability of parameters like fingertip span by comprehensively considering factors such as safety distance, support stability, and deformation suppression. Like an examiner, it assigns scores to different span options, with higher scores being considered more appropriate.

[0184] The weighting factor is used to adjust the importance of safety distance, support stability, and deformation suppression in the evaluation. For example, when removing ultra-thin wafers, deformation suppression may be more important, so a larger value may be set.

[0185] The safety distance is the distance between the fingertip span (d) and the spatial boundary (such as the inner wall of a FOUP), ensuring that the fingertip does not collide during movement. For example, requiring a safety distance of at least 2mm means that the fingertip must be at least 2mm away from the surrounding boundary.

[0186] The support stability function evaluates support stability based on the spacing between support points (the distance between the fingertips supporting the wafer) and the distribution of the wafer's center of gravity. For example, if the spacing between support points is reasonable and the wafer's center of gravity is in the right position, the function value is high, indicating stable support.

[0187] The deformation influence function is a function that takes into account the effect of the span on wafer deformation. If the span is appropriate, the wafer deformation can be suppressed, and the function value is good.

[0188] The optimal span d_i* is the fingertip span that maximizes the spatial fitness function within each cutting unit Ci. It is the optimal fingertip span within this small unit that ensures safe and stable access.

[0189] The span adjustment cost function measures the "cost" (e.g., time, mechanical wear, etc.) of adjusting the fingertip span. For example, adjusting the span takes time, and the longer the time, the higher the cost. This function quantifies these costs.

[0190] The span change cost coefficient is used to adjust the effect of span change on cost. Large span changes may result in large mechanical losses, and this coefficient reflects this effect.

[0191] The time cost coefficient is a coefficient that adjusts the impact of span time on cost. The longer the time, the higher the cost corresponding to this coefficient.

[0192] The global optimal span sequence is a set of optimal fingertip span adjustment sequences and values ​​obtained by combining the conditions of each cutting unit through a dynamic programming algorithm.

[0193] In this embodiment, unit length cutting involves cutting the path along the centerline of the pick-and-place path at fixed, small lengths (e.g., 5 mm), dividing the path into multiple small units. This is like using a ruler to mark fixed lengths to segment the path.

[0194] Cut unit analysis examines each cut unit to determine whether the fingertip span needs to be adjusted. This analysis determines whether the fingertip span within that unit meets spatial constraints, adaptability, and other requirements.

[0195] The fingertip assembly is the part of the pick-and-place mechanism that directly touches the wafer during placement and is capable of adjusting its span. Like a human finger, it is mechanical, capable of opening and closing to adjust its spacing (span) to suit different situations.

[0196] The ideal span is the fingertip span that is most suitable for picking up and placing wafers in a certain cutting unit after analysis, and it can balance factors such as safety, stability, and efficiency.

[0197] In this embodiment, the wafer deformation refers to the change in shape of the wafer during the placement and handling process, such as whether it is bent or warped. For example, the wafer may bulge in the middle or droop at the edges due to its own stress or previous processing.

[0198] The safe pick and place starting position is determined based on wafer deformation. For example, if the wafer edge is sagging, you cannot start at the most severe sagging area. You should find a location with minimal deformation that allows for stable pick and place. This area should be used as the starting position.

[0199] Real-time monitoring continuously monitors the wafer's position, shape, and other changes during the pick-and-place process. It's like installing a "real-time camera + sensor" to monitor the wafer's condition.

[0200] The difference in support point distances refers to the different distances between the multiple points where the tip supports the wafer. For example, if there are three support points, the distances between each two are different. This difference will affect the support stability of the wafer and should be taken into account when controlling the lifting rhythm.

[0201] The rhythm of wafer lifting is the pattern of force and speed of movement when lifting the wafer with the fingertips. For example, some support points are lifted slowly first, and others are lifted later. There is a rhythm of sequence and force to ensure the wafer is lifted smoothly.

[0202] In this embodiment, the wafer placement state refers to the state of the wafer at the target position after the placement is completed, such as whether it is placed correctly, whether there is any offset, whether the deformation has changed, etc. For example, the wafer must be accurately placed on the chuck of the equipment without being tilted, which means that the placement state is good.

[0203] Pick-and-place process data is a collection of data recorded during the pick-and-place process, such as fingertip position, span, force applied, wafer position changes, and deformation data. This data acts like a "diary," recording the details of the pick-and-place process.

[0204] Process optimization involves improving the overall semiconductor manufacturing process based on data from the pick-and-place process. For example, if a wafer is found to be prone to deformation at a certain step during pick-and-place, the preceding processing steps can be adjusted to avoid this.

[0205] Device parameter adjustment involves adjusting the parameters of the pick-and-place device (such as the fingertip assembly and laser detection module) based on the data from the pick-and-place process. For example, if the data indicates that the span adjustment is not accurate enough, the parameters controlling the span in the device will be adjusted to ensure more accurate subsequent pick-and-place operations.

[0206] The beneficial effects of this technical solution include: starting with the precise construction of a pick-and-place structure model using 3D lasers, planning a reasonable pick-and-place path, dynamically adjusting fingertip span, real-time monitoring and adjustment of pick-and-place movements, and finally verifying and accumulating data for optimization. This makes the semiconductor wafer pick-and-place process more precise and stable, reducing problems such as collisions and deformation, and improving wafer pick-and-place yield. Data accumulation also allows for continuous optimization of the overall semiconductor manufacturing process and equipment, enhancing the efficiency and quality of overall semiconductor production.

[0207] The present invention provides a semiconductor wafer pick-and-place method, such as Figure 2 Shown, including:

[0208] Step 1: Use a 3D laser illuminator to perform 3D laser scanning on the pick-and-place port and the relevant area inside the FOUP where the wafer is located, and pre-process the detection data to reconstruct the pick-and-place structure model;

[0209] Step 2: Plan the pick-and-place routing path based on the pick-and-place structure model, wafer, and FOUP parameters;

[0210] Step 3: Perform unit length cutting along the central axis of the pick-and-place routing path, and analyze each cutting unit to determine whether the fingertip span needs to be adjusted. If necessary, drive the fingertip assembly to adjust to the ideal span;

[0211] Step 4: Determine the safe pick-and-place starting position based on the wafer deformation, and control the fingertips to perform the pick-and-place action according to the pick-and-place routing path. Monitor the fingertips' pick-and-place action in real time. If there are changes in the wafer position or shape, adjust the pick-and-place action, and control the rhythm of the fingertips lifting the wafer based on the difference in the support point distance.

[0212] Step 5: After the placement is completed, the laser detection module is controlled to scan again to verify the wafer placement status and store the placement process data for subsequent process optimization and device parameter adjustment;

[0213] Among them, the wafer manufacturing process includes: wafer boat support structure and fingertip structure;

[0214] The wafer boat support structure is developed and designed based on a set platform, which is compatible with the standard wafer boat installation holes, and the lifting teeth on both sides of the wafer boat adopt a circular arc long tooth structure;

[0215] At the same time, the length of rod-1 and rod-3 supports is increased based on the standard crystal boat structure;

[0216] The fingertip structure adopts a 4-point lifting structure, combined with an arc fork finger shape, and increases the span of the fingertips to at least 200 mm.

[0217] Preferably, step 1 comprises:

[0218] Constructing laser original coordinates and mapping detection data to the laser original coordinates to obtain an initial structure;

[0219] Decomposing the initial structure into three-dimensional substructures to obtain point cloud data of each three-dimensional substructure, and calculating the interference coefficient of each position point to construct an interference set;

[0220] performing interference cancellation processing based on the interference set to obtain a first structure;

[0221] Obtain a two-dimensional array of edge positions of each connecting line on the three-dimensional substructure, and construct a distance difference between the two-dimensional arrays of edge positions to obtain a distance difference vector;

[0222] Calculate the fitting value of each distance difference vector to obtain a reference set of each three-dimensional substructure;

[0223] The interference set is modified based on the reference set to determine an adjustment position point, and the first structure is adjusted to obtain a second structure, wherein the second structure is a pick-and-place structure model.

[0224] Preferably, step 2 includes:

[0225] Extract the central axis L of the pick-and-place routing path, establish a path coordinate system, and perform unit length calculations along the direction of the central axis L. Cutting to form n cutting units Ci, where i=1,2,…,n;

[0226] For each cutting unit Ci, establish the spatial constraint equation :

[0227]

[0228] in, Represents the three-dimensional space area corresponding to the cutting unit Ci; , Represents the boundary constraint values ​​based on the y-axis and z-axis respectively; Represents the constraint value of the x-axis of the cutting unit Ci; Represents spatial coordinates;

[0229] Define the spatial fitness function:

[0230]

[0231] in, 、 、 is the weight coefficient; It represents the safe distance between the fingertip span d and the space boundary; represents the support stability function, which is related to the support point spacing and the center of gravity distribution of the wafer; represents the deformation influence function, considering the inhibitory effect of span on wafer deformation;

[0232] For each cutting unit Ci, solve the optimal span d_i*:

[0233] d_i*=argmaxFi,dmin <d_i*<dmax

[0234] Where dmin and dmax are the span ranges limited by the physical limitations of the device;

[0235] Establish the span adjustment cost function:

[0236]

[0237] in, is the span change cost coefficient; is the time cost coefficient; is the actual span value of the cutting unit Ci; Time spent adjusting for span;

[0238] Use dynamic programming algorithm to solve the global optimal span sequence: .

[0239] The beneficial effects of this technical solution include: starting with the precise construction of a pick-and-place structure model using 3D lasers, planning a reasonable pick-and-place path, dynamically adjusting fingertip span, real-time monitoring and adjustment of pick-and-place movements, and finally verifying and accumulating data for optimization. This makes the semiconductor wafer pick-and-place process more precise and stable, reducing problems such as collisions and deformation, and improving wafer pick-and-place yield. Data accumulation also allows for continuous optimization of the overall semiconductor manufacturing process and equipment, enhancing the efficiency and quality of overall semiconductor production.

[0240] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A semiconductor wafer pick-and-place device, characterized in that: include: The laser detection module is used to perform three-dimensional laser scanning of the pick-and-place port and the relevant area inside the FOUP where the wafer is located based on a three-dimensional laser illuminator, and pre-process the detection data to reconstruct the pick-and-place structure model; A path planning module is used to plan a pick-and-place routing path based on the pick-and-place structure model, wafer, and FOUP parameters; An adjustment and analysis module is used to perform unit length cutting along the central axis of the pick-and-place piece routing path, and analyze each cutting unit to determine whether the fingertip span needs to be adjusted. If necessary, the fingertip assembly is driven to adjust to the ideal span; The action adjustment module is used to determine the safe pick-and-place starting position based on the wafer deformation, and control the fingertips to perform the pick-and-place action according to the pick-and-place wafer routing path. The fingertips perform the pick-and-place action in real time. If there are changes in the wafer position or shape, the pick-and-place action is adjusted, and the rhythm of the fingertips lifting the wafer is controlled based on the difference in the support point distance; The verification module is used to control the laser detection module to scan again after the placement is completed to verify the wafer placement status and store the placement process data for subsequent process optimization and device parameter adjustment; Among them, the wafer manufacturing process includes: wafer boat support structure and fingertip structure; The wafer boat support structure is developed and designed based on a set platform, which is compatible with the standard wafer boat installation holes, and the lifting teeth on both sides of the wafer boat adopt a circular arc long tooth structure; At the same time, the length of rod-1 and rod-3 supports is increased based on the standard crystal boat structure; The fingertip structure adopts a 4-point lifting structure, combined with an arc fork finger shape, and increases the span of the fingertips to at least 200 mm.

2. The semiconductor wafer pick-and-place device according to claim 1, wherein: The laser detection module includes: An initial construction unit, configured to construct original laser coordinates and map detection data to the original laser coordinates to obtain an initial structure; A first interference unit is configured to perform three-dimensional substructure decomposition on the initial structure, obtain point cloud data of each three-dimensional substructure, and calculate an interference coefficient of each position point to construct an interference set; an interference cancellation unit, configured to perform interference cancellation processing based on the interference set to obtain a first structure; a distance difference construction unit, configured to obtain a two-dimensional array of edge positions of each connecting line on the three-dimensional substructure, and perform distance difference construction on the two-dimensional array of edge positions to obtain a distance difference vector; The second interference unit is used to calculate the fitting value of each distance difference vector to obtain a reference set of each three-dimensional substructure; An adjustment unit is configured to modify the interference set based on the reference set to determine an adjustment position point, and adjust the first structure to obtain a second structure, wherein the second structure is a pick-and-place structure model.

3. The semiconductor wafer pick-and-place device according to claim 1, wherein: The adjustment analysis module includes: The extraction unit is used to extract the central axis L of the pick-and-place routing path, establish a path coordinate system, and perform unit length along the direction of the central axis L. Cutting to form n cutting units Ci, where i=1,2,…,n; Constraint establishment unit, used to establish the spatial constraint equation for each cutting unit Ci : in, Represents the three-dimensional space area corresponding to the cutting unit Ci; , Represents the boundary constraint values ​​based on the y-axis and z-axis respectively; Represents the constraint value of the x-axis of the cutting unit Ci; Represents spatial coordinates; Function definition unit, used to define the spatial fitness function: in, 、 、 is the weight coefficient; It represents the safe distance between the fingertip span d and the space boundary; represents the support stability function, which is related to the support point spacing and the center of gravity distribution of the wafer; represents the deformation influence function, considering the inhibitory effect of span on wafer deformation; The span solving unit is used to solve the optimal span d_i* for each cutting unit Ci: d_i*=argmaxFi,dmin <d_i*<dmax Where dmin and dmax are the span ranges limited by the physical limitations of the device; Function building unit, used to build span adjustment cost function: in, is the span change cost coefficient; is the time cost coefficient; is the actual span value of the cutting unit Ci; Time spent adjusting for span; Sequence solving unit, used to solve the global optimal span sequence using dynamic programming algorithm: .

4. The semiconductor wafer pick-and-place device according to claim 1, wherein: The action adjustment module includes: The starting position determination unit is used to obtain the three-dimensional topography of the wafer, build a deformation model, calculate the deformation of each area, and select the area with the smallest deformation as the safe pick-and-place starting position; Threshold establishment unit, used to establish a safety threshold function: ,in, is the warpage; is the curvature; 、 is the weight coefficient; Trajectory generation unit, used to plan the pick-and-place path based on the safe starting position and safety threshold function, and generate the optimal trajectory curve .

5. The semiconductor wafer pick-and-place device according to claim 4, wherein: The action adjustment module further includes: Support point optimization unit, used to analyze support point distance differences and establish a mechanical balance model: ,in, represents the supporting force acting on point i; G is the weight of the wafer; is uida stress; is the penalty coefficient; n represents the total number of support point symbols; The adaptive adjustment unit is used to build an adaptive control algorithm through real-time monitoring through multi-sensor fusion during the pick and place process: ,in, is the error vector, and They are wafer position offset, shape change, and support point force distribution; 、 、 is the PID parameter matrix; The control unit is used to control the rhythm of fingertip lifting according to the difference in support point distance and real-time force conditions.

6. The semiconductor wafer pick-and-place device according to claim 5, wherein: The control unit comprises: Rhythm generation unit for online optimization of the stage target force of piecewise functions based on reinforcement learning and time constant , generating the fingertip lifting rhythm Ft; Where t represents the time point during the fingertip placement and removal process; k0 represents the k0th segment.

7. The semiconductor wafer pick-and-place device according to claim 5, wherein: Also includes: The sequence determination unit is used to determine the force application order of the support points through the distance-force coordination priority function Pj: Where j is the support point number; 、 is the dynamic weight; is the distance difference between the jth support point and other support points; is the maximum value of the distance difference among all support points; is the force deviation of the j-th support point; It is the maximum value of the force deviation of all support points; the dynamic optimization unit is used for real-time feedback correction and dynamic optimization of force control parameters to suppress wafer deformation.

8. A semiconductor wafer pick-and-place method, characterized in that: include: Step 1: Use a 3D laser illuminator to perform 3D laser scanning on the pick-and-place port and the relevant area inside the FOUP where the wafer is located, and pre-process the detection data to reconstruct the pick-and-place structure model; Step 2: Plan the pick-and-place routing path based on the pick-and-place structure model, wafer, and FOUP parameters; Step 3: Perform unit length cutting along the central axis of the pick-and-place routing path, and analyze each cutting unit to determine whether the fingertip span needs to be adjusted. If necessary, drive the fingertip assembly to adjust to the ideal span; Step 4: Determine the safe pick-and-place starting position based on the wafer deformation, and control the fingertips to perform the pick-and-place action according to the pick-and-place routing path. Monitor the fingertips' pick-and-place action in real time. If there are changes in the wafer position or shape, adjust the pick-and-place action, and control the rhythm of the fingertips lifting the wafer based on the difference in the support point distance. Step 5: After the placement is completed, the laser detection module is controlled to scan again to verify the wafer placement status and store the placement process data for subsequent process optimization and device parameter adjustment; Among them, the wafer manufacturing process includes: wafer boat support structure and fingertip structure; The wafer boat support structure is developed and designed based on a set platform, which is compatible with the standard wafer boat installation holes, and the lifting teeth on both sides of the wafer boat adopt a circular arc long tooth structure; At the same time, the length of rod-1 and rod-3 supports is increased based on the standard crystal boat structure; The fingertip structure adopts a 4-point lifting structure, combined with an arc fork finger shape, and increases the span of the fingertips to at least 200 mm.

9. The semiconductor wafer pick-and-place method according to claim 8, wherein: Step 1 includes: Constructing laser original coordinates and mapping detection data to the laser original coordinates to obtain an initial structure; Decomposing the initial structure into three-dimensional substructures to obtain point cloud data of each three-dimensional substructure, and calculating the interference coefficient of each position point to construct an interference set; performing interference cancellation processing based on the interference set to obtain a first structure; Obtain a two-dimensional array of edge positions of each connecting line on the three-dimensional substructure, and construct a distance difference between the two-dimensional arrays of edge positions to obtain a distance difference vector; Calculate the fitting value of each distance difference vector to obtain a reference set of each three-dimensional substructure; The interference set is modified based on the reference set to determine an adjustment position point, and the first structure is adjusted to obtain a second structure, wherein the second structure is a pick-and-place structure model.

10. The semiconductor wafer pick-and-place method according to claim 9, wherein: Step 2 includes: Extract the central axis L of the pick-and-place routing path, establish a path coordinate system, and perform unit length along the direction of the central axis L. Cutting to form n cutting units Ci, where i=1,2,…,n; For each cutting unit Ci, establish the spatial constraint equation : in, Represents the three-dimensional space area corresponding to the cutting unit Ci; , Represents the boundary constraint values ​​based on the y-axis and z-axis respectively; Represents the constraint value of the x-axis of the cutting unit Ci; Represents spatial coordinates; Define the spatial fitness function: in, 、 、 is the weight coefficient; It represents the safe distance between the fingertip span d and the space boundary; represents the support stability function, which is related to the support point spacing and the center of gravity distribution of the wafer; represents the deformation influence function, considering the inhibitory effect of span on wafer deformation; For each cutting unit Ci, solve the optimal span d_i*: d_i*=argmaxFi,dmin <d_i*<dmax Where dmin and dmax are the span ranges limited by the physical limitations of the device; Establish the span adjustment cost function: in, is the span change cost coefficient; is the time cost coefficient; is the actual span value of the cutting unit Ci; Time spent adjusting for span; Use dynamic programming algorithm to solve the global optimal span sequence: .

Images