Direct bonding is a widely used technique in the semiconductor industry. Many materials can be assembled and stacked together with this technique, enabling the fabrication of many useful structures. The direct bonding technique is, for instance, a major step of the Smart Cut ™ process flow enabling the industrial-scale production of Silicon-On-Insulator (SOI) substrates [1][2]. Material diversification in the microelectronics industry (i.e. “More than Moore” approach) imposes new technological constraints. The substrate bonding process control has to be improved calling for more and more accurate bonding energy measurement protocols.In this work, a new technique is assessed in order to measure, at the wafer scale, direct bonding energies. It is derived from the standard Double Cantilever Beam (DCB) method [3] and uses interferometry in confocal IR laser source microscopy to measure crack openings (Fig. 1A) [4][5][6]. Such a bonding energy measurement protocol shows a better accuracy compared to other techniques (Fig. 1B & 1D). This is due to a better confocal microscopy resolution and the high intensity of the laser source. The elastic energy stored in bent wafers is obtained by measuring the beam curvature [7][8]. DCB deformation models are discussed from short-range crack opening to long distance beam-bending theories. Comparison is done between results from analytic models, Finite Element Model (FEM) and experiments (Fig. 1C).Near-bonding edge zone modeled with FEM simulation accurately matches experimental data. A new mechanical and universal bending model, valid at both small and large scales, is proposed. With this model, bonding energy extraction is then reduced to a simple scaling parameter extraction directly linked to the stored elastic strain energy (Fig 1C & 1D) [9]. This work opens prospects for deformation and bonding energy measurements in various configurations including bonded heterostructures, polymer, and metal bonding. [1] Bruel, Electron. Lett. 31, 1201 (1995).[2] K. Celler, A.-J. Auberton-Hervé, B. Aspar, C. Lagahe-Blanchard, and C. Maleville, in Wafer Bonding Applications and Technology, Springer Series in Materials Science, edited by M. Alexe and U. Gosele, Springer, Berlin (2004).[3] Maszara, G. Goetz, A. Caviglia, and J. B. Mckitterick, J. Appl. Phys. 64, 4943 (1988).[4] Wu, S. Gowrishankar, R. Huang et al., Int. J. Fract. 202, 1–19 (2016).[5] Gowrishankar, H. Mei, K. M. Liechti et al., Int. J. Fract. 177, 109–128 (2012).[6] Pallares, L. Ponson, A. Grimaldi, M. George, G. Prevot, and M. Ciccotti, Int. J. Fract. 156, 11–20 (2009).[7] Bertholet, “Measurement, optimization and multiscale modeling of silicon wafer bonding interface fracture resistance,” Ph.D. thesis (Catholique de Louvain, 2006).[8] Olbrechts, B. Lejeune, Y. Bertholet, T. Pardoen, and J.-P. Raskin, ECS Trans. 3(6), 279–289 (2006).[9] Colonel, A. Calvez, F. Fournel et al., J. Appl. Phys. 132, 215106 (2022). Fig. 1: A – (Top) Four pasted IR confocal microscope images acquired during the DCB testing (the vertical red dashed line shows the closing point’s fitted position). (Bottom) Air gap height measured as a function of pixel position.B – Simulated reflected intensity (left y-axis) and experimental reflected intensity (right y-axis), both as functions of the estimated distance from the closure point x. The vertical dashed grey line shows the real experimental closing point.C – Air gap height as a function of the position/thickness ratio (in log/log scales) for different analytical and numerical models and experimental data.D – Bonding energies including errors obtained with the standard length measurement method, the curvature interpolation method, the scaling method, or J-integral calculation from FEM as functions of the annealing temperature. Figure 1
Read full abstract