 |
|
||||
BiographyLado Filipovic is an Associate Professor and the Director of the Christian Doppler Laboratory for Multi-Scale Process Modeling of Semiconductor Devices and Sensors at the Institute for Microelectronics, TU Wien. Lado’s research is centered around Integrated Semiconductor Sensors and Process Technology Computer Aided Design (TCAD). He obtained his venia docendi (habilitation) in Semiconductor Based Integrated Sensors and his doctoral degree (Dr.techn.) in Microelectronics from TU Wien in 2020 and 2012, respectively. He holds a Master’s degree in Applied Sciences (MASc.) from Carleton University in Ottawa, Canada, which he obtained in 2009. Lado is currently heading several research projects from a wide range of technology readiness levels (TRLs) including basic research – funded by the Austrian Science Fund (FWF); applied research – funded by the Christian Doppler Forschungsgesellschaft (CDG), the Austrian Research Promotion Agency (FFG), and the European Union (EU); and industry research – funded by direct industry collaborations. He is a Senior Member of the IEEE and is an active member of the Technical Program Committee for many outstanding IEEE sponsored conferences. He has served as a reviewer for several European funding agencies, has edited two books on Miniaturized Transistors, and is an active reviewer for many leading journals. His research team has released several open-source scientific software tools under the ViennaTools moniker, such as the process simulator ViennaPS and the device simulator ViennaEMC, which have been applied for studying the fabrication and operation of advanced nanoelectronic devices. The research group currently collaborates with many industry partners (e.g., Silvaco, Infineon Technologies, Global TCAD Solutions, Fuji Electric) as well as academia from around the world (e.g., MIT, Arizona State University, University of Glasgow, University of Groningen, University of Vienna, Chinese Academy of Sciences). Lado's primary research interest is studying the operation, stability, and reliability of novel semiconductor-based sensors using advanced process and device simulations. An additional pillar of his research is the multi-scale modeling of processes involved in the fabrication of semiconductor devices and sensors. This involves combining atomistic modeling with Monte Carlo and continuum approaches, as well as merging physical and empirical modeling in a single framework, specifically in process TCAD. He is also actively investigating metal oxide semiconductors and novel two-dimensional (2D) materials, e.g., graphene, MoS2, and phosphorene for the detection of biomarkers and environmental pollutants. In particular, his group is investigating the impact of the adsorption of ambient gas molecules on the surfaces of 2D semiconducting films and on the performance of devices and sensors based on these. Software Development:Process Simulator ViennaPS Research Topics:Semiconductor Sensors Running Public Projects:FWF - Adsorbate-Dependent Conductivity of MoS2 CDG - CDL for Multi-Scale Process Modeling of Semiconductor Devices and Sensors |
|||||
Process Simulation and Virtual Fab
Process simulation and virtual fabrication are essential pillars of modern semiconductor development, enabling predictive modeling of fabrication steps and helping bridge the gap between design and manufacturing. Our efforts in this domain focus on high-fidelity simulation frameworks that span from atomistic to feature and equipment scales. These frameworks allow us to model plasma etching, deposition, ion implantation, and annealing processes with physical accuracy and computational efficiency. We develop and maintain tools such as ViennaPS and ViennaLS, integrating advanced models for transport, surface reactions, and topography evolution, supporting the exploration of novel device architectures and process conditions.
This year, our modeling capabilities were significantly enhanced. A new multi-scale approach couples chamber-scale plasma simulations with feature-scale etching using spline-interpolated fluxes, enabling fast and accurate emulation of Cl2/Ar processes, shown in the figure below. In ion beam etching, newly introduced angular yield models and ion reflection mechanisms revealed how process parameters shape nanostructures like blazed gratings. On the atomic scale, molecular dynamics simulations unraveled the defect evolution during SiC annealing, while a cluster-based semi-empirical model improved the description of dopant activation in 4H-SiC. Additional advancements include energy-dependent yield functions and refined modeling of surface evolution under oblique incidence.
The ViennaPS engine was also extended with major usability and performance features. The integrated ray tracing module, ViennaRay, now supports GPU acceleration via NVIDIA OptiX, delivering orders-of-magnitude speedups in high-resolution simulations. Furthermore, ViennaPS is now available via PyPI, allowing researchers to install it easily with pip install viennaps and integrate it directly into Python-based design flows and digital twin platforms.
Fig 1: Integrated multi-scale modeling framework linking reactor-scale plasma simulations with feature-scale process simulations using ViennaPS. A multi-variable spline interpolation model captures the dependence of ion and neutral fluxes on reactor parameters, enabling fast and accurate prediction of etch behavior in complex structures without re-running full plasma simulations.
Semiconductor- and 2D-Based Sensors
The integration of advanced sensing capabilities into semiconductor platforms is a key enabler for applications ranging from environmental monitoring to wearable electronics. Our research targets the development of highly sensitive and selective sensors based on both conventional semiconductors and emerging two-dimensional (2D) materials. These atomically thin materials, such as graphene, MoS2, and phosphorene, exhibit unique surface reactivity and electronic tunability that make them ideal candidates for gas sensing and other detection mechanisms. Through atomic-level engineering, including substitutional doping and defect control, we tailor their electrical response and interface properties. This work is supported by a full experimental–theoretical pipeline, including ultra-high vacuum synthesis, electron microscopy, electrical characterization, and multi-scale simulations linking quantum-mechanical calculations with device-level models.
Over the past year, we successfully demonstrated atomic-level substitutional doping of 2D materials using a two-step method combining ion irradiation and metal evaporation, enabling precise incorporation of single platinum (Pt) and gold (Au) atoms into the MoS2 lattice, shown in the figure below. This breakthrough was confirmed through atomic-resolution STEM imaging and spectroscopy, demonstrating the stability and structural configuration of the dopants. Building on this, we theoretially explored how dopant–gas interactions influence charge trapping dynamics, supporting a new sensing mechanism based on hysteresis modulation in 2D transistors. Complementary theoretical work refined our understanding of trap states in doped phosphorene and MoS2 through ab initio calculations, while device-level simulations linked these effects to measurable electrical signatures. Additionally, we extended our materials modeling to investigate phonon scattering and hot carrier cooling mechanisms in lead halide perovskites, deepening our insights into their potential for high-efficiency optoelectronic applications.
Fig 2: Controlled substitutional doping of MoS2 using a two-step ion irradiation and metal evaporation process. (a–c) Schematic of the doping process: (a) experimental setup combining a plasma source and e-beam evaporator; (b) low-energy He+ irradiation creates chalcogen vacancies; (c) Pt atoms are deposited and diffuse into vacancy sites. (d–f) Atomic-resolution HAADF-STEM images: (d) pristine MoS2; (e) after defect creation; (f) after Pt substitution, showing successful incorporation of dopants at specific lattice positions. Insets and arrows highlight individual vacancies and dopant atoms. This process enables precise engineering of 2D material properties at the atomic level for next-generation sensing and electronic applications.
AI for Microelectronics
Artificial Intelligence is playing an increasingly central role across all levels of semiconductor research, from atomistic materials modeling to equipment-scale process simulation. Our group focuses on physics-informed machine learning (ML) approaches that accelerate simulation workflows and enhance predictive accuracy. A cornerstone of our efforts is the development of machine-learned interatomic potentials (ML-IAPs), which bring first-principles accuracy to large-scale atomistic simulations that would be computationally intractable with traditional methods. In parallel, we develop digital twins of fabrication equipment—surrogate models that emulate full reactor simulations with high fidelity and speed, enabling inverse design and real-time optimization when integrated with feature-scale TCAD tools such as ViennaPS.
Over the past year, we made substantial advances in both atomistic and equipment-level modeling. At the equipment scale, we introduced a digital twin framework that couples physics-based plasma simulations with neural network surrogates to model ion fluxes and energy/angular distributions. These surrogates allow rapid process parameter exploration and inverse modeling. We also applied Sobol indeces and SHAP-based explainability to interpret surrogate predictions and identify key control parameters.
On the materials side, we used unsupervised learning and ML-based structural descriptors to study defect formation and clustering in amorphous SiO2, shedding light on the atomic precursors to trap generation and long-term device degradation. We further expanded our ML-IAP capabilities to wide-bandgap semiconductors, including 4H-SiC and GaN, enabling predictive atomistic modeling of point defects, diffusion mechanisms, and implantation damage. In 4H-SiC, we benchmarked multiple ML-IAP frameworks to evaluate accuracy-efficiency tradeoffs in defect and amorphous structure prediction. For Mg-doped GaN, we developed a high-fidelity Gaussian Approximation Potential that reproduces defect formation energies and migration barriers, laying the foundation for predictive simulations of dopant activation. These new ML-based tools now make it possible to simulate large-scale, defect-rich systems with near-DFT accuracy, supporting improved process control and reliability in advanced power and optoelectronic devices.
Fig 3: Molecular dynamics workflow for simulating thermal oxidation of silicon using ML-based interatomic potentials. (a) Planar Si(100) surfaces and (b) silicon nanowires are exposed to molecular oxygen in large-scale MD simulations. (c) O2 molecules spontaneously react and dissociate at the surface, forming a conformal oxide layer. The simulation includes periodic energy minimization and gas refilling steps to maintain thermal and pressure conditions.