Statistical Topological Gradient and Shape Optimization for Robust Metal–Semiconductor Contact Reconstruction
L. Afraites, A. Hadri, M. Hrizi, J. F. T. Rabago · 2026
Read Statistical Topological Gradient and Shape Optimization for Robust Metal–Semiconductor Contact Reconstruction by L. Afraites, A. Hadri, M. Hrizi, J. F. T. Rabago (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-04216.txt.
Introduction
Afraites and others develop statistical topological gradient and shape tuning for robust metal-semiconductor contact rebuild. They build a robust method that reconstructs metal-semiconductor contact regions via topological gradients. The analysis shows deterministic balance of the topological gradient under measurement noise, and extends to a data setting with several independent notes.
Catalog record
| CREATOR | L. Afraites, A. Hadri, M. Hrizi, J. F. T. Rabago |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 88039 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.04216
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Statistical Topological Gradient and Shape Optimization for Robust Metal–Semiconductor Contact Reconstruction L. Afraites, A. Hadri, M. Hrizi, J. F. T. Rabago Abstract We develop a statistically robust framework for reconstructing metal–semiconductor contact regions using topological gradients. The inverse problem is formulated as the identification of an unknown contact region from boundary measurements governed by an elliptic model with piecewise coefficients. Deterministic stability of the topological gradient with respect to measurement noise is established, and the analysis is extended to a statistical setting with multiple independent observations. A central limit theorem in a separable Hilbert space is proved for the empirical topological gradient, yielding optimal n − 1 / 2 n^{-1/2} convergence and enabling the construction of confidence intervals and hypothesis tests for contact detection. To further refine the reconstruction, a shape optimization procedure is employed, where the free parameter β \beta in the CCBM formulation plays a crucial role in controlling interface sensitivity. While β \beta affects both topological and shape reconstructions, its influence is particularly pronounced in the shape optimization stage, allowing more accurate estimation of the size and geometry of the contact subregion. The proposed approach provides a rigorous criterion for distinguishing true structural features from noise-induced artifacts, and numerical experiments demonstrate the robustness, precision, and enhanced performance of the combined statistical, topological, and β \beta -informed shape-based reconstruction. Keywords: Geometric inverse problem, topological optimization, shape optimization, coupled complex boundary method, contact resistivity MSC codes: 49Q12, 49N45, 49Q10, 49K40, 35A15, 35B20, 65R32 1 Introduction 1.1 The contact resistivity problem In very-large-scale integration (VLSI) circuits (Figure 1(a) ), electrical contacts form the critical interfaces between semiconductor devices and metallic interconnects. As device dimensions shrink into the deep-submicron regime, parasitic contact resistance increasingly limits overall electrical performance (Figure 1(b) ). In its simplest approximation, the contact resistance is given by R_{\mu}=\frac{\rho_{\mu}}{A}, where A is the contact area and \rho_{\mu} is the specific contact resistivity. Since A typically scales with the square of the minimum feature size \lambda , this model predicts the ideal scaling law R_{\mu}\propto\lambda^{-2} (see, e.g., [ 45 , 46 ] ). In practice, this ideal scaling fails due to non-uniform current injection near the contact edges, a phenomenon known as current crowding . The spatial extent of current spreading is characterized by the transfer length l_{t} . When the lateral dimensions of the contact become comparable to or smaller than l_{t} , current crowding intensifies, causing the effective contact resistance to increase more rapidly than predicted by geometric scaling. This effect ultimately sets a fundamental physical limit on contact miniaturization in advanced integrated technologies. Figure 1(b) schematically illustrates a simplified metal–semiconductor contact structure. (a) Interface Aluminum (metal) Silicon (semiconductor) VLSI contact \rho_{\mu} (b) (c) Figure 1 : (a) A VLSI circuit; (b) simplified metal/semiconductor contact; (c) a 2D slab contact model with interface at z=0 and diffusion-layer thickness z_{j} . Adapted from [ 46 ] . In deeply scaled integrated circuits, contact resistivity governs device performance, making its accurate characterization essential. Yet direct measurement is fundamentally constrained by the buried nature of the contact interface is buried within the device structure, precluding physical access for local probing. Consequently, researchers employ indirect methods: prescribed boundary currents and measured potentials provide the data for inferring internal contact properties. This inverse problem, recovering internal parameters from boundary observations, has been extensively studied, with seminal works establishing the mathematical foundation for contact resistivity determination [ 45 , 46 ] . Related inverse problems for semiconductor devices can be found in [ 13 , 14 , 23 ] and the references therein. 1.2 Mathematical formulation Although the physical metal–semiconductor contact system is inherently three-dimensional (3D), the extremely shallow junction depths encountered in modern VLSI processes justify an effective two-dimensional (2D) reduction. As illustrated in Figure 1(c) , the contact interface is idealized as a planar surface orthogonal to the z -axis, while the semiconductor diffusion region of thickness z_{j} exhibits a depth-dependent conductivity profile \sigma=\sigma(z) that is laterally uniform in the (x,y) -plane. This assumption is well supported by contemporary fabrication techniques, where implantation and diffusion produce nearly planar doping distributions across contact windows. Following the thin-film reduction of Loh et al. [ 46 ] , we consider the asymptotic regime in which the vertical-to-lateral aspect ratio tends to zero. Vertical integration of the 3D current continuity equation yields the effective sheet resistance R_{s}=\left(\int_{0}^{z_{j}}\sigma(z)\,\mathrm{d}z\right)^{-1}\ >0, which characterizes lateral current transport. This averaging maps the 3D potential v(x,y,z) to an effective 2D potential u(x,y)=R_{s}\int_{0}^{z_{j}}\sigma(z)\,v(x,y,z)\,\mathrm{d}z. The reduced model governs the potential u=u(x,y) in an open, bounded semiconductor domain \varOmega\subset\mathbb{R}^{2} , with \omega\Subset\varOmega an open, bounded subdomain representing the metal–semiconductor contact. Outside the contact region, \varOmega\setminus\overline{\omega} , lateral current spreads without vertical transfer and u satisfies \Delta u=0 in \varOmega\setminus\overline{\omega} , with purely two-dimensional current density \mathbf{J}=-R_{s}^{-1}\nabla u . Inside the contact region \omega , vertical current injection introduces a reaction term, and u solves \Delta u-l_{t}^{-2}u=0 in \omega , where the transfer length l_{t}=\sqrt{\rho_{\mu}/R_{s}} characterizes the depth of current penetration. At the contact boundary \partial\omega , continuity of the potential and conservation of the normal current flux impose u_{|_{\partial\omega_{-}}}=u_{|_{\partial\omega_{+}}} and R_{sk}^{-1}\,n_{z}\!\cdot\nabla u_{|_{\partial\omega_{-}}}=R_{s}^{-1}\,n_{z}\!\cdot\nabla u_{|_{\partial\omega_{+}}} , where n_{z} denotes the unit vector orthogonal to the contact boundary and tangential to the (x,y) -plane (cf. Figure 1(c) ), and R_{sk} is the sheet resistance directly beneath the contact. Collecting the governing relations, the fully reduced boundary value problem can be written compactly as -\Delta u+{\mu_{0}}\,\chi_{\omega}\,u=0\quad\text{in }\varOmega,\qquad\partial_{\nu}u=g\geq 0\quad\text{on }\partial\varOmega,\qquad(g\not\equiv 0), where {\mu_{0}}\coloneqq 1/l_{t}^{\,2}=R_{s}/\rho_{\mu} , \chi_{\omega} is the characteristic function of the contact region, g denotes the prescribed boundary current density, and \partial_{\nu}u=\nabla u\cdot\nu represents the outward normal flux, with \nu the unit outward normal to \partial\varOmega . The term (1/l_{t}^{2})\chi_{\omega}u accounts for vertical current exchange at the metal-semiconductor interface. The transfer length l_{t} controls current crowding: during the transition from vertical injection beneath the contact to lateral transport in the semiconductor, current concentrates near the contact edges, leading to strongly non-uniform current densities that dominate the effective contact resistance in modern device geometries. 1.3 The geometric inverse problem The general inverse problem aims to recover both the contact region \omega\Subset\varOmega and the parameter {\mu_{0}}>0 from Cauchy boundary data (g,f) . This problem is severely ill-posed, as different pairs (\chi_{\omega},{\mu_{0}}) can produce identical measurements [ 37 ] . In this work, we adopt the standard assumption that {\mu_{0}} is known a priori, reducing the problem to the geometric identification of the contact region \omega (or equivalently \chi_{\omega} ) from the boundary data. We implicitly assume the existence of an exact contact region \omega^{\star} generating the measurements. Throughout the paper, unless stated otherwise, \varOmega\subset\mathbb{R}^{2} is a bounded domain with Lipschitz boundary, and \omega\Subset\varOmega is a bounded open subdomain with Lipschitz boundary. Problem 1.1 . Let {\mu_{0}}>0 be a given constant. Given nontrivial Cauchy boundary data (g,f)\in H^{-\frac{1}{2}}(\partial\varOmega)\times H^{\frac{1}{2}}(\partial\varOmega) , find the subdomain \omega and a function u\in H^{1}{(\varOmega)} such that -\Delta u+{\mu_{0}}\chi_{\omega}u=0\ \text{in }\varOmega,\qquad\partial_{\nu}u=g,\quad u=f\ \text{on }\partial\varOmega. (1) The analysis of this reduced inverse problem naturally involves three closely related aspects: (i) Identifiability. A fundamental question is whether the boundary data (g,f)\in H^{-\frac{1}{2}}(\partial\varOmega)\times H^{\frac{1}{2}}(\partial\varOmega) uniquely determine the contact region \omega . Uniqueness has been established for certain classes of admissible subdomains [ 22 , 41 , 42 ] . Notably,