To address the demand for sub-100-nm overlay accuracy in wafer bonding for 3D integration, this study proposes an extended overlay assessment model integrating physical mechanisms and data-driven approaches, along with a correlation analysis methodology with process parameters. Rigid-body models inadequately characterize systematic deformations from crystalline anisotropy and process stresses. To overcome this, we construct an extended overlay model based on Zernike polynomials, incorporating physically meaningful terms for precise description of non-uniform wafer deformation. An innovative Zernike term selection strategy combining physics-guided pre-screening and AIC-optimized stepwise regression resolves overfitting/underfitting, enhancing generalizability and interpretability. Validation using patterned wafer geometry (PWG) data shows the model achieves R & sup2; > 0.70 for both net bonding deformation and lithography-compensable components, demonstrating excellent deformation decomposition. Correlation analysis of multiple process experiments reveals strong correlations (|r| > 0.85) between key process parameters (e.g., peak bonding head force) and specific Zernike modes, providing evidence for suppressing detrimental deformations via process optimization. This research establishes a complete framework from theory to experimental verification and process traceability, laying a foundation for mechanism diagnosis, predictive compensation, and closed-loop control in high-precision wafer bonding.

Objective With advancements in three-dimensional integration technology, wafer stacking has become a critical process for enhancing semiconductor device performance in the post-Moore era. The reliability of interfacial electrical interconnections depends on the bonding overlay accuracy, which is now primarily limited by residuals at the 50 nm level. This study addresses the challenge of bonding residuals in high-precision wafer bonding, which arise from the coupled effects of wafer elastic deformation, clamping constraints, and bond wave propagation. Existing models often lack comprehensive multiphysics coupling or fail to establish a link between specific process parameters and residual formation, limiting their use in process optimization. Therefore, developing a high-fidelity coupled model is essential for understanding the residual generation mechanism and devising effective suppression strategies. Methods A multiphysics coupling analysis model was developed that comprehensively considers wafer anisotropy, clamping boundary effects, and bond wave propagation behavior. The framework integrates anisotropic thin-plate elasticity (incorporating crystal orientation transformation tensors), gas film dynamics (governed by a modified Reynolds equation with bonding stress), and contact mechanics (solved via the augmented Lagrangian method). Bond wave propagation is governed by an energy criterion at the wavefront, balancing effective bonding energy against strain energy and the work performed by gas film and mechanical contact pressures. A finite element model for the 300 mm wafer bonding process was developed, achieving submicron accuracy. Key numerical strategies included a staggered iterative scheme for updating the wavefront, bonding force, gas pressure, and structural deformation; adaptive time stepping based on residual variations; and stabilization damping to suppress rigid body motion. Model validity was confirmed through comparison between simulation predictions and experimental pattern wafer geometry measurements. Results The simulation accurately captured the nonuniform bond wave propagation induced by wafer anisotropy. The stress and residual distributions exhibited a distinct fourfold symmetry consistent with the crystallographic orientation. Residuals were primarily concentrated near the wafer edge, with additional significant residuals observed at the center-consistent with previous reports. Experimental validation showed strong agreement between simulated and measured residual distribution patterns. Systematic parameter studies revealed that using a flat bond head reduced the 2-norm of the residual vector by 41% compared with a spherical head (0.63 u03BCm vs. 1.07 u03BCm). Employing a lower-stiffness material (polyethylene) for the bond head reduced the residual 2-norm by 23% compared with PEEK plastic. Moreover, an increase in the initial wafer gap correlated with a higher residual 3u03C3 value. Conclusions This study establishes a robust multiphysics coupling and process co-optimization framework for high-precision wafer bonding. The proposed model effectively captures the combined effects of wafer anisotropy, gas film dynamics, and contact mechanics on residual formation, enabling high-fidelity simulation of the bonding process and quantitative analysis of key process parameters. The findings demonstrate that optimizing the bond head design-with a flat surface and low-stiffness material-and minimizing the initial wafer gap can significantly suppress bonding residuals. This work provides a theoretical basis and practical design guidelines for optimizing wafer bonding processes to achieve superior overlay accuracy.

Direct wafer bonding is essential for high-performance 3D integrated circuits (3D ICs), but process-induced wafer distortion, which is driven by mechanical deformation during bonding, remains a key challenge that degrades alignment accuracy and interconnect quality. This distortion is influenced by parameters such as bonding gap, applied force, and chuck constraints. While prior models capture basic dynamics, they overlook practical conditions such as the restriction imposed by the top chuck on the top wafer. We address this by introducing an improved, computationally efficient axisymmetric model that incorporates process factors via tailored boundary conditions, including a penalty term for the top chuck’s height constraint. The model accounts for bonding gap, central loading force, and partial chuck release, enabling rapid prediction of wafer deformation throughout bonding. Experimental validation with a custom setup shows strong agreement between simulated and measured wafer shapes across various conditions. Notably, the model allows calibration of hard-to-measure parameters (e.g., adhesion energy) from a single experiment and generalizes well within the same wafer batch. Though limited to axisymmetric cases, it provides a fast, accurate, and practical alternative to conventional finite-element methods, significantly reducing experimental costs and offering deeper insight into bonding mechanics.

This paper proposes a Rapid and Robust Iterative Feedback Tuning ($\mathbf{R}^{2}$ IFT) method to address the slow convergence and limited robustness of classical IFT. Robustness is enforced by explicitly constraining the magnitude of the closedloop sensitivity function, and the resulting constrained problem is reformulated as an unconstrained optimization via a smooth SquarePlus penalty. To improve convergence efficiency, a dualloop iterative scheme is developed: the outer loop minimizes the performance criterion through experiments, while the inner loop iteratively approaches the gradient zero point without requiring additional experiments. The proposed $\mathbf{R}^{2}$ IFT is fully data-driven and does not rely on a parameterized plant model. Simulation results demonstrate significantly improved convergence speed together with guaranteed robustness of the tuned controller.

Human locomotion recognition (HLR) is essential for the self-paced treadmill and other human-robot interactive systems. The generalization capability of recognition algorithms should be carefully considered due to the diversity of locomotion patterns for different subjects. Conventional CNN- or LSTM-based pipelines suffer a marked loss of accuracy once a new treadmill user is encountered, and domain-adaptation approaches such as DANN still rely on collecting unlabeled data from that user. In response, we previously introduced a Hybrid Spatial–Temporal Graph Convolutional Network (HSTGCN) that preserves the natural topology of plantar-pressure sensors through an adaptive spatial graph, extracts modality-intrinsic features by processing pressure data with a spatial–temporal GCN and inertial data with an LSTM, and then fuses these heterogeneous streams through a temporal LSTM to produce a compact, user-invariant representation of locomotion. The present study is the first to rigorously assess whether these architectural choices truly deliver ready-to-use generalization. The proposed HSTGCN is validated on a dataset consisting of eight subjects with five locomotion modes. Under this test, the HSTGCN retained 97.9

In the field of precision motion control, particularly for systems with stringent motion accuracy requirements such as wafer stages, position-related model perturbations have become a major factor limiting further performance improvements. To enhance servo performance, this paper proposes a data-driven position-dependent feedforward control strategy. The proposed method utilizes spatio-temporal basis functions to linearly parameterize the feedforward controller, thereby effectively accommodating model perturbations while maintaining robustness to task variations. A data-driven iterative algorithm based on the Gauss-Newton method is derived to optimize the controller parameters without relying on a parameterized model. Convergence analysis reveals that the proposed algorithm does not require highly accurate system model information, thereby significantly reducing experimental costs. The effectiveness and practicality of the proposed method are validated through multi-die exposure trajectory tracking experiments with high transient characteristics on an ultra-precision wafer stage. The experimental results demonstrate that the proposed method significantly improves servo performance under varying trajectory parameters, exposure die positions and die sizes.

The manufacturing of advanced chips, particularly 3D-IC, presents increasingly stringent requirements regarding overlay errors. In-plane distortion (IPD) emerges as a critical factor contributing to these overlay errors, making its compensation a focal point for both lithography and wafer bonding processes. An effective strategy for addressing this issue involves predicting IPD through the deformation information of the wafer, thereby facilitating feedforward compensation. However, existing gradient model is limited to predicting IPD only in axisymmetric cases, performing ineffective for non-axisymmetric cases. This paper introduces a novel analytical model for predicting IPD based on wafer shape, derived from an in-depth analysis of stress distribution within the film. This model not only maintains equivalence with the conventional gradient model under axisymmetric conditions but also extends its applicability by accurately predicting IPD in non-axisymmetric cases. Utilizing the finite element method, we conducted simulations of wafer deformation and the resultant IPD, demonstrating the robustness and accuracy of our proposed model. This work significantly advances the understanding of IPD mechanisms and offers a reliable approach for enhancing overlay accuracy in advanced chip manufacturing processes.

Objective In the cutting-edge semiconductor manufacturing domain targeting nodes below 40 nm, laser spike annealing (LSA) has emerged as a key technology. As the industry relentlessly continues to pursue miniaturization and performance enhancement, LSA plays a critical role in fabricating high-performance logic devices by enabling precise dopant activation in wafers. However, existing LSA processes-commonly employing high-energy density laser beams at large incident angles-frequently result in highly non-uniform temperature distributions, particularly in the edge regions of wafers. This leads to serious challenges such as edge burn and wafer cracking. Edge burn damages the wafer edges and introduces defects that degrade electrical properties, whereas wafer cracking renders wafers unusable, leading to material and time losses. These issues significantly undermine wafer integrity, quality, and production yield, increasing costs and delaying deliveries. Thus, there is an urgent need to develop innovative and effective approaches for optimizing the scanning trajectory design and temperature control parameters. The primary objective of this study aims to enhance temperature uniformity across the wafer, with a particular emphasis on improving thermal stability in the edge regions. Improvement is critical for ensuring reliable, high-performance semiconductor manufacturing processes and meeting the industry's increasing demands for advanced devices. Methods This research introduces a comprehensive and systematic methodology. First, a novel scanning trajectory was designed. In contrast to the conventional straight-line scanning approach, a three-segment tangent line-arc-line trajectory was adopted. This design mitigates direct laser irradiation on the wafer edges-a common issue with straight-line scanning-while also reducing vibration problems associated with purely arc-based trajectories. By precisely defining parameters such as the light spot length, outer reference circle radius, and the azimuth angle of the initial trajectory, a detailed trajectory planning method was established. Subsequently, the relationships between the azimuth angle and key optical factors, such as incident light intensity and absorptance, were thoroughly investigated through geometric and polarization optical analyses. Using these relationships and considering the maximum energy absorbed at the wafer edge, an optimal azimuth angle range was calculated. In addition, a power-switching mechanism was integrated into the system. This mechanism dynamically and precisely adjusts the laser power in real time as the laser spot moves across the wafer boundary. Finally, to address the persistent issue of temperature overshoot during open/closed-loop switching processes, a second-orderScurve temperature trajectory was designed. This temperature trajectory allows for smooth, gradual changes in temperature response, optimizing temperature control performance. Results and Discussions The experimental findings clearly highlight the effectiveness of the proposed method. Regarding azimuth angle optimization, a gradual increase in the azimuth angle from a relatively small value to the optimal range of 57.3 degrees -90 degrees results in a significant and continuous reduction in the peak temperature at the wafer edge during the first scanning. This demonstrates that proper adjustment of the azimuth angle can effectively regulate the interaction between the laser beam and the wafer edge. By increasing the azimuth angle, the incident light intensity and absorptance at the wafer edge are optimized, effectively suppressing excessive heat generation in the edge region. In terms of temperature trajectory optimization, the introduction of the second-order Scurve temperature trajectory yields highly favorable results. It effectively mitigates the temperature overshoot problem commonly encountered in traditional control methods. This approach stabilizes the temperature within a reasonable range and enhances the overall temperature control stability during the scanning process. Conclusions This research represents a significant advancement in LSA for semiconductor manufacturing. By addressing the persistent issue of edge burn, the study offers a set of highly effective solutions that significantly enhance the quality and reliability of semiconductor production processes. The meticulous optimization of scanning trajectory parameters, combined with the innovative design of the temperature trajectory, leads to a marked improvement in wafer temperature uniformity. This, in turn, enhances the processing quality and performance of high-performance logic devices. Although the experiments were conducted under specific ion implantation conditions, the methodologies and insights derived from this study provide valuable references for a broader range of semiconductor manufacturing processes. Furthermore, the proposed approach has significant potential for application in other laser scanning technologies. The ability to improve processing efficiency and quality opens up new avenues for technological advancement in the manufacturing sector.

Due to its high hardness, brittleness, and high melting point, high-precision surface processing of silicon carbide (SiC) presents significant challenges. To improve processing efficiency and surface quality, laser-assisted machining techniques have been widely applied in SiC treatment. In this study, ultraviolet nanosecond laser was used to modify the surface of 4H-SiC wafers, and the effects of laser energy density, scanning speed, and step overlap rate on the surface roughness of SiC were investigated. As the laser energy density increases, the silicon content on the SiC surface remains largely unchanged, while the oxygen content rises (25.27 %) and the carbon content decreases. The thickness of the oxide layer also grows, reaching a maximum of approximately 830 nm. Reducing scanning speed and increasing step overlap rate also enhanced the oxygen content on the SiC surface. Comparing the two wafers, the SiC wafer with a larger surface roughness has a thicker oxide layer and higher oxygen content. The nanoindentation test results show that plastic deformation occurred on the SiC surface during laser treatment, leading to a reduction in hardness, which is beneficial for subsequent processes. For the finished wafers, the optimal process parameters are an energy density of 10 J/cm2, scanning speed of 350 mm/s, and step overlap rate of 8 %. For the process wafer, the energy density is 8 J/cm2, with the other parameters remaining the same. The research results provide experimental support for the improvement of SiC surface quality and subsequent precision manufacturing processes.

Wafer bonding is a critical process in 3D integration, and overlay (OVL) metrology is essential for its success. Accurately positioning the centre of OVL targets is fundamental for effective metrology. However, the identification and localization of target centres become challenging due to complex shapes and unexpected features, such as rounded corners, that can arise during manufacturing. An algorithm is proposed to tackle this challenge by employing customizable shape fitting. This method begins with the extraction of sub-pixel edge points, followed by applying a Hough transform to group and smooth these points, thereby enhancing contour quality. By parameterizing the target shape based on specific points, the algorithm integrates sub-pixel traversal techniques with an optimization objective, achieving sub-pixel accuracy in centre positioning. Simulation results indicate that the algorithm can achieve a positioning accuracy of +/- 0.03 pixels and demonstrates robustness against noise and blur. Finally, the proposed algorithm was used to test the OVL target pair arrays fabricated by electron beam etching, confirming an accuracy of +/- 0.04 pixels (+/- 6.9 nm). These results validate the algorithm's capability to meet high precision requirements for OVL target centre positioning in wafer applications.

Reaching mode, which guides system states toward the sliding mode surface, is a crucial aspect of sliding mode control. This paper proposes an optimal reaching filter designed to achieve an optimal reaching mode while accounting for control input constraints. The optimal reaching law is formulated as an optimization problem to ensure efficient and constrained control execution. Using a double integrator system with a linear sliding mode surface as a representative example, the effectiveness and superiority of the proposed approach are validated through rigorous theoretical analysis and simulation. Additionally, an experimental setup is developed for further verification, with comparative results demonstrating that the proposed method significantly enhances the speed at which system states converge to the sliding mode surface.

Ball screw drive (BSD) is a precision transmission mechanism widely used in high-precision positioning or tracking systems. The dynamic behavior of BSD varies with position and load, which causes tracking errors and poor robustness. Therefore, this paper proposes a polytopic linear parameter varying (LPV) model to express the varying dynamic behavior of BSD. The parameters of the LPV model are identified by the closed-loop frequency domain method and Levenberg–Marquardt iterative algorithm. Based on the polytopic LPV model, an LPV gain scheduling (GS) H∞ controller is proposed for the BSD with varying dynamics. Specifically, the controller is designed through polytope-based GS representation and mixed sensitivity synthesis. The most significant part is the proposal of a GS H∞ control algorithm to implement controller parameters that change with changing dynamics. Moreover, the stability of the closed-loop system is achieved by quadratic stabilization with state feedback. Finally, identification experiments and trajectory-tracking comparative experiments are carried out. The experimental results demonstrate that the proposed polytopic LPV modeling and GS H∞ control synthesis are effective in achieving accurate trajectory tracking and excellent robustness.

In this article, a data-driven multiple-input–multiple-output (MIMO) feedforward control approach is synthesized to enhance the tracking performance of precision MIMO motion systems. Specifically, a MIMO feedforward controller parameterized with polynomial basis functions is employed to address the coupling of MIMO systems. A new data-driven feedforward tuning algorithm for the MIMO feedforward controller is then developed based on the measured step response of the process sensitivity function. The proposed approach requires only one tracking experiment in each iteration, resulting in an experimentally efficient feedforward parameter optimization w.r.t. a user-defined and tracking-performance-related criterion through iterative learning from the measured data. Finally, application to an industrial three degrees-of-freedom motion stage illustrates that the proposed approach outperforms a data-driven single-input–single-output (SISO) feedforward control scheme in terms of tracking performance and achieves good performance robustness against the reference variation.

Due to its exceptional physical properties, silicon carbide (SiC) is pivotal in the semiconductor industry. Laser annealing technology is utilized to enhance ohmic contacts on SiC surfaces. This study focuses on the influence of laser parameters, such as energy density and scanning speed, on the temperature of the Ni/SiC contact system. Simulations and experiments explore the impacts of varying energy densities and scanning speeds under constant conditions. The results from both methods delineate the optimal process conditions necessary for establishing robust and efficient ohmic contacts, providing a defined range of suitable parameters. This research underscores the critical role of precise laser annealing in improving semiconductor device interfaces.

Accurate wafer alignment is the key to achieving wafer bonding accuracy. High-precision wafer alignment systems typically use vision to locate the aligned Mark on two wafers, and use complex mechanisms to perform multiple composite movements to achieve functionality, making the coupling effect of multiple types of errors more complex and challenging the error budget of wafer alignment systems. This paper proposes an error budgeting method for such vision based multibody precision systems. This method takes the homogeneous transformation matrix (HTM) method as the core to model the system error and establish two types of error transfer chains. For the error chain involving visual measurement, an analysis method based on geometrical optics is proposed to consider the influence of the position and orientation errors of the optical path components. Then organize the possible error sources in the system and model the parameters of each error based on actual test results. Combined with the process flow, customize the error model for each link. Finally, perform Monte Carlo simulation. Using the aforementioned method to budget errors for a certain configuration of wafer alignment system, main error sources were identified, and accuracy indicators were proposed based on the alignment accuracy requirements of +/- 200 nm. The rationality of the error budget conclusion in this study has been verified through experiments on the construction machine.

The rigid-flexible coupled variable mechanical system, as a typical multi-input multi-output (MIMO) system, poses significant challenges in system modeling, control design, theoretical analysis, and practical engineering implementation due to its complex dynamic characteristics and the interactions among multiple variables. To address these challenges, this paper proposes a subspace closed-loop identification method for MIMO linear parameter-varying (LPV) systems based on the minimum Euclidean distance of the state sequence. First, through the subspace closed-loop identification algorithm, a non-unique form of the linear time-invariant state-space (L TI-SS) model with the minimum Euclidean distance state sequence is obtained. Then, using the least squares method, a similarity transformation matrix is computed to convert the non-unique form of the LTI-SS model into the standard model. Finally, the global linear parameter-varying state-space (LPV -SS) model is obtained through cubic spline interpolation. The proposed method is experimentally validated on a single-degree-of-freedom ball screw test rig and simulated on a two-degree-of-freedom motion system. Through simulation and experimentation, the calibration determination coefficient of the LPV MIMO model is above 0.98, demonstrating the superior accuracy of the model. Compared with traditional modal standard models and controllable standard models, the proposed method requires fewer data to achieve the specified model accuracy, improves the computational efficiency of the identification process, and reduces experimental and data collection costs. This provides an effective solution for modeling rigid-flexible coupled variable mechanical systems.

This paper proposes a new online calibration method for the differential-drive mobile robot equipped with a laser scanner. Our algorithm can jointly estimate the 3-DoF extrinsic parameters of the laser scanner and the intrinsic parameters (radii and wheel spacing) of the differential-drive kinematic model. Applying the pre-integration theory initially developed the for IMU sensor to the differential-drive kinematic model, we adopt iterative nonlinear optimization to minimize the cost derived from laser and encoder measurements. Experiments results confirm that the proposed method can do online calibration precisely.

The aim of this article is to propose a novel rational feedforward tuning method, by directly mapping the feedforward signal learned by dual-loop iterative learning control (DILC) onto the corresponding reference, that achieves high performance for varying trajectory tracking tasks. The DILC algorithm is first developed by paralleling the standard iterative learning control (ILC) with an additional iterative loop. Different from the standard ILC, DILC can learn an ideal feedforward signal eliminating the reference-induced error even though a robustness filter presents for the robust convergence against model uncertainties. Then, based on the reference and the feedforward signal learned by DILC, an instrumental variable-based algorithm is developed for the parameter tuning of the rational feedforward controller, which leads to unbiased estimates and optimal accuracy in terms of variance. The proposed method combines the performance of DILC with the flexibility of rational feedforward controllers. Comparative simulation and application to an ultraprecision wafer stage illustrate the enhanced performance of the proposed approach compared to the preexisting results.

In intelligent vehicle cooperative systems, the mismatch in driving characteristics between a human and a machine and the driver misoperation caused by this mismatch result in human–machine conflicts, which significantly affect driving safety. Therefore, an intelligent vehicle human–machine cooperative steering torque control method is proposed herein. To adapt the intelligent system to the varying previewing characteristics of a human, a time-varying previewing driver model is constructed, and a penalty factor for human–machine intervention is designed based on fuzzy rules to assign driving control rights by assessing the driver’s state. Consequently, a human–vehicle–road model with driver preview time and penalty factor as varying parameters is established. Based on gain-scheduling control, a human–machine cooperative steering torque controller is designed to adapt to the varying previewing characteristics of a human and the change in human–machine intervention. The stability and robustness of the entire parameter space are guaranteed by constraining the poles in a certain region. Finally, the proposed human–machine cooperative control scheme demonstrates the effective alleviation of conflicts between the driver and the intelligent driving system.

The wafer stage is a crucial component in lithography machines, requiring nanometer-level motion control accuracy. To overcome the performance limitation between low-frequency trajectory tracking performance and high-frequency noise amplification of fixed-value hybrid integrator-gain system (HIGS), a novel hybrid time-varying integrator-gain control (HTVIGC) strategy is proposed for an ultra-precision wafer stage. Specially, a time-varying parameter design method is newly introduced into HIGS, and the integrator frequency and gain value in HTVIGC vary with the trajectory phase. Furthermore, the time-domain properties and approximate frequency-domain characteristics of HTVIGC are analyzed in detail, and the stability condition of the closed-loop system is obtained through an analysis similar to the circle criterion. Finally, experimental results on wafer stage demonstrate that the proposed method can achieve excellent control performance in both the acceleration and constant velocity phases of the trajectory.