Prashant P. Shinde
Research Subjects
Prashant P. Shinde has a total of 59 co-authors. Here are the top co-authors by number of shared publications:
K. Subramanya Mayya
6 shared publications- Shashishekar P. Adiga
NextGen Projects Samsung Advanced Institute of Technology Samsung Semiconductor India Research 560048 Bangalore India
6 shared publications Carlo A. Pignedoli
4 shared publicationsHyeon-Jin Shin
Inorganic Materials Lab, Samsung Advanced Institute of Technology 2 , Suwon 433-803, Republic of Korea
4 shared publicationsSeongjun Park
Inorganic Materials Lab, Samsung Advanced Institute of Technology 2 , Suwon 433-803, Republic of Korea
4 shared publicationsShanthi Pandian
Device Materials Simulations (SAIT-India), Samsung Research Institute 1 , Bangalore 560037, India
4 shared publications
Latest Publications
- Open Access
Defect detection in photolithographic patterns using deep learning models trained on synthetic data
- Open Access
Defect Detection in Photolithographic Patterns Using Deep Learning Models Trained on Synthetic Data
In the photolithographic process vital to semiconductor manufacturing, various types of defects appear during EUV pattering. Due to ever-shrinking pattern size, these defects are extremely small and cause false or missed detection during inspection. Specifically, the lack of defect-annotated quality data with good representation of smaller defects has prohibited deployment of deep learning based defect detection models in fabrication lines. To resolve the problem of data unavailability, we artificially generate scanning electron microscopy (SEM) images of line patterns with known distribution of defects and autonomously annotate them. We then employ state-of-the-art object detection models to investigate defect detection performance as a function of defect size, much smaller than the pitch width. We find that the real-time object detector YOLOv8 has the best mean average precision of 96% as compared to EfficientNet, 83%, and SSD, 77%, with the ability to detect smaller defects. We report the smallest defect size that can be detected reliably. When tested on real SEM data, the YOLOv8 model correctly detected 84.6% of Bridge defects and 78.3% of Break defects across all relevant instances. These promising results suggest that synthetic data can be used as an alternative to real-world data in order to develop robust machine-learning models.
Open AccessDefect Detection in Photolithographic Patterns Using Deep Learning Models Trained on Synthetic Data
In the photolithographic process vital to semiconductor manufacturing, various types of defects appear during EUV pattering. Due to ever-shrinking pattern size, these defects are extremely small and cause false or missed detection during inspection. Specifically, the lack of defect-annotated quality data with good representation of smaller defects has prohibited deployment of deep learning based defect detection models in fabrication lines. To resolve the problem of data unavailability, we artificially generate scanning electron microscopy (SEM) images of line patterns with known distribution of defects and autonomously annotate them. We then employ state-of-the-art object detection models to investigate defect detection performance as a function of defect size, much smaller than the pitch width. We find that the real-time object detector YOLOv8 has the best mean average precision of 96% as compared to EfficientNet, 83%, and SSD, 77%, with the ability to detect smaller defects. We report the smallest defect size that can be detected reliably. When tested on real SEM data, the YOLOv8 model correctly detected 84.6% of Bridge defects and 78.3% of Break defects across all relevant instances. These promising results suggest that synthetic data can be used as an alternative to real-world data in order to develop robust machine-learning models.
- Open Access
- 11 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 39969-39974
Wafer Defect Localization and Classification Using Deep Learning Techniques
Accurate detection and classification of wafer defects constitute an important component in semiconductor manufacturing. It provides interpretable information to find the possible root causes of defects and to take actions for quality management and yield improvement. Traditional approach to classify wafer defects, performed manually by experienced engineers using computer-aided tools, is time-consuming and can be low in accuracy. Hence, automated detection of wafer defects using deep learning approaches has attracted considerable attention to improve the performance of detection process. However, a majority of these works have focused on defect classification and have ignored defect localization which is equally important in determining how specific process steps can lead to defects in certain locations. To address this, we evaluate the state-of-the-art You Only Look Once (YOLO) architecture to accurately locate and classify wafer map defects. Experimental results obtained on 19200 wafer maps show that YOLOv3 and YOLOv4, the variants of YOLO architecture, can achieve >94% of classification accuracy in real-time. For comparison, other architectures, namely ResNet50 and DenseNet121 are also evaluated for wafer defect classification and they give accuracies 89% and 92% respectively, however, without localization abilities. We find that the object detection methods are very useful in locating and classifying defects on semiconductor wafers.
- 11 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 39969-39974
Open AccessWafer Defect Localization and Classification Using Deep Learning Techniques
Accurate detection and classification of wafer defects constitute an important component in semiconductor manufacturing. It provides interpretable information to find the possible root causes of defects and to take actions for quality management and yield improvement. Traditional approach to classify wafer defects, performed manually by experienced engineers using computer-aided tools, is time-consuming and can be low in accuracy. Hence, automated detection of wafer defects using deep learning approaches has attracted considerable attention to improve the performance of detection process. However, a majority of these works have focused on defect classification and have ignored defect localization which is equally important in determining how specific process steps can lead to defects in certain locations. To address this, we evaluate the state-of-the-art You Only Look Once (YOLO) architecture to accurately locate and classify wafer map defects. Experimental results obtained on 19200 wafer maps show that YOLOv3 and YOLOv4, the variants of YOLO architecture, can achieve >94% of classification accuracy in real-time. For comparison, other architectures, namely ResNet50 and DenseNet121 are also evaluated for wafer defect classification and they give accuracies 89% and 92% respectively, however, without localization abilities. We find that the object detection methods are very useful in locating and classifying defects on semiconductor wafers.
- 29 March 2021
- Published by AIP Publishing in Applied Physics Letters
- Vol. 118 (13) , 133504
Surface roughness mediated specularity parameter of thin Cu films
Roughness-induced resistivity variation of thin metal films is conveniently described by the Fuchs–Sondheimer model, where the phenomenological parameter p is used to quantify the extent of specular scattering at surfaces. However, p is a lumped parameter and does not include microscopic information that characterizes roughness, viz., auto-correlation function, root-mean-square height, and correlation length. In this work, we extract these roughness parameters for Cu films of thickness ranging from 31 nm to 95 nm. We find that the roughness–roughness correlation function is an exponential with a characteristic correlation length that increases monotonically with the film thickness. Using this, we predict the roughness parameter-dependent specularity coefficient, which has an implicit thickness dependency. This alters the resistivity scaling compared to the prevailing model of resistivity scaling, where p is assumed to be a constant.
- 29 March 2021
- Published by AIP Publishing in Applied Physics Letters
- Vol. 118 (13) , 133504
Surface roughness mediated specularity parameter of thin Cu films
Roughness-induced resistivity variation of thin metal films is conveniently described by the Fuchs–Sondheimer model, where the phenomenological parameter p is used to quantify the extent of specular scattering at surfaces. However, p is a lumped parameter and does not include microscopic information that characterizes roughness, viz., auto-correlation function, root-mean-square height, and correlation length. In this work, we extract these roughness parameters for Cu films of thickness ranging from 31 nm to 95 nm. We find that the roughness–roughness correlation function is an exponential with a characteristic correlation length that increases monotonically with the film thickness. Using this, we predict the roughness parameter-dependent specularity coefficient, which has an implicit thickness dependency. This alters the resistivity scaling compared to the prevailing model of resistivity scaling, where p is assumed to be a constant.
- 10 November 2020
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 128 (18) , 185103
Non-specular scattering of carriers from surface defects in thin metal interconnects
Non-specular scattering of carriers, despite being one of the key contributors to metal film resistivity, is conveniently characterized by a single parameter (referred to as specularity parameter) in the literature without providing a means to separate out relative contributions of surface defects to it. In this work, we report a theoretical formalism to predict contributions to (non-specular) reflection (scattering) of carriers from different types of chemical and structural defects on the film surface. Establishing specular reflection as a special case of the more generalized reflection from the surface, we show that elastic scatterings contribute to the non-specular reflections by randomizing the forward momentum along the transport direction. We predict density and angle-dependent specularity coefficient for various types of defects and show that large angle reflections tend to be non-specular in the presence of isotropic surface defects. Using this formalism, we predict the probability of specular reflection for technologically relevant metal–metal and metal–insulator interfaces. Calculated specularity coefficients are in agreement with values obtained from experimental measurements. Our methodology will help in designing next generation interconnects and engineering the interfaces with minimal diffuse reflection.- 10 November 2020
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 128 (18) , 185103
Non-specular scattering of carriers from surface defects in thin metal interconnects
Non-specular scattering of carriers, despite being one of the key contributors to metal film resistivity, is conveniently characterized by a single parameter (referred to as specularity parameter) in the literature without providing a means to separate out relative contributions of surface defects to it. In this work, we report a theoretical formalism to predict contributions to (non-specular) reflection (scattering) of carriers from different types of chemical and structural defects on the film surface. Establishing specular reflection as a special case of the more generalized reflection from the surface, we show that elastic scatterings contribute to the non-specular reflections by randomizing the forward momentum along the transport direction. We predict density and angle-dependent specularity coefficient for various types of defects and show that large angle reflections tend to be non-specular in the presence of isotropic surface defects. Using this formalism, we predict the probability of specular reflection for technologically relevant metal–metal and metal–insulator interfaces. Calculated specularity coefficients are in agreement with values obtained from experimental measurements. Our methodology will help in designing next generation interconnects and engineering the interfaces with minimal diffuse reflection.- 2 October 2020
- Published by American Physical Society (APS) in Physical Review B
- Vol. 102 (16) , 165102
Electrical resistivity of atomically smooth single-crystal Cu films
We investigate using the Boltzmann transport equation the electrical resistivity of atomically smooth single-crystal and films with thickness 1–45 nm. Our transport calculations show that, in the absence of grain-boundary scattering and surface perturbations, the resistivity of Cu films of thickness comparable to the electron mean free path (40 nm for bulk copper) is close to the bulk resistivity. The resistivity increases trivially for sub-40-nm-thick films. These single-crystal Cu films show an intrinsic limit of resistivity. A steep increase in resistivity observed for ultrathin films is due to enhanced finite-size effects. We explain the cause of increase in resistivity by investigating the electronic structure at the Fermi level.- 2 October 2020
- Published by American Physical Society (APS) in Physical Review B
- Vol. 102 (16) , 165102
Electrical resistivity of atomically smooth single-crystal Cu films
We investigate using the Boltzmann transport equation the electrical resistivity of atomically smooth single-crystal and films with thickness 1–45 nm. Our transport calculations show that, in the absence of grain-boundary scattering and surface perturbations, the resistivity of Cu films of thickness comparable to the electron mean free path (40 nm for bulk copper) is close to the bulk resistivity. The resistivity increases trivially for sub-40-nm-thick films. These single-crystal Cu films show an intrinsic limit of resistivity. A steep increase in resistivity observed for ultrathin films is due to enhanced finite-size effects. We explain the cause of increase in resistivity by investigating the electronic structure at the Fermi level.- Open Access
- 5 March 2019
- Published by Springer Nature in Scientific Reports
- Vol. 9 (1) , 3488
Effect of encapsulation on electronic transport properties of nanoscale Cu(111) films
The stiff compromise between reliability and conductivity of copper interconnects used in sub-nanometer nodes has brought into focus the choice of encapsulation material. While reliability was the primary driver so far, herein, we investigate how electronic conductivity of Cu(111) thin films is influenced by the encapsulation material using density functional theory and Boltzmann transport equation. Atomically thin 2D materials, namely conducting graphene and insulating graphane both retain the conductivity of Cu films whereas partially hydrogenated graphene (HGr) results in reduction of surface density of states and a reduction in Cu film conductivity. Among transition metal elements, we find that atoms in Co encapsulation layer, which essentially act as magnetic impurities, serve as electron scattering centres resulting in a decrease in conductivity by at least 15% for 11 nm thick Cu film. On the other hand, Mo, Ta, and Ru have more favorable effect on conductivity when compared to Co. The cause of decrease in conductivity for Co and HGr is discussed by investigating the electronic band structure and density of states. Our DFT calculations suggest that pristine graphene sheet is a good encapsulation material for advanced Cu interconnects both from chemical protection and conductivity point of view.
- 5 March 2019
- Published by Springer Nature in Scientific Reports
- Vol. 9 (1) , 3488
Open AccessEffect of encapsulation on electronic transport properties of nanoscale Cu(111) films
The stiff compromise between reliability and conductivity of copper interconnects used in sub-nanometer nodes has brought into focus the choice of encapsulation material. While reliability was the primary driver so far, herein, we investigate how electronic conductivity of Cu(111) thin films is influenced by the encapsulation material using density functional theory and Boltzmann transport equation. Atomically thin 2D materials, namely conducting graphene and insulating graphane both retain the conductivity of Cu films whereas partially hydrogenated graphene (HGr) results in reduction of surface density of states and a reduction in Cu film conductivity. Among transition metal elements, we find that atoms in Co encapsulation layer, which essentially act as magnetic impurities, serve as electron scattering centres resulting in a decrease in conductivity by at least 15% for 11 nm thick Cu film. On the other hand, Mo, Ta, and Ru have more favorable effect on conductivity when compared to Co. The cause of decrease in conductivity for Co and HGr is discussed by investigating the electronic band structure and density of states. Our DFT calculations suggest that pristine graphene sheet is a good encapsulation material for advanced Cu interconnects both from chemical protection and conductivity point of view.
- 7 May 2018
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry C
- Vol. 123 (46) , 28106-28113
MoS2 Quantum Dots: Effect of Hydrogenation on Surface Stability and H2S Release
- 7 May 2018
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry C
- Vol. 123 (46) , 28106-28113
MoS2 Quantum Dots: Effect of Hydrogenation on Surface Stability and H2S Release
- Top Cited Papers
- 1 February 2017
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 11 (2) , 1380-1388
On-Surface Synthesis and Characterization of 9-Atom Wide Armchair Graphene Nanoribbons
The bottom-up approach to synthesize graphene nanoribbons strives not only to introduce a band gap into the electronic structure of graphene but also to accurately tune its value by designing both the width and edge structure of the ribbons with atomic precision. We report the synthesis of an armchair graphene nanoribbon with a width of nine carbon atoms on Au(111) through surface-assisted aryl–aryl coupling and subsequent cyclodehydrogenation of a properly chosen molecular precursor. By combining high-resolution atomic force microscopy, scanning tunneling microscopy, and Raman spectroscopy, we demonstrate that the atomic structure of the fabricated ribbons is exactly as designed. Angle-resolved photoemission spectroscopy and Fourier-transformed scanning tunneling spectroscopy reveal an electronic band gap of 1.4 eV and effective masses of ≈0.1 me for both electrons and holes, constituting a substantial improvement over previous efforts toward the development of transistor applications. We use ab initio calculations to gain insight into the dependence of the Raman spectra on excitation wavelength as well as to rationalize the symmetry-dependent contribution of the ribbons’ electronic states to the tunneling current. We propose a simple rule for the visibility of frontier electronic bands of armchair graphene nanoribbons in scanning tunneling spectroscopy.
- 1 February 2017
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 11 (2) , 1380-1388
Top Cited PapersOn-Surface Synthesis and Characterization of 9-Atom Wide Armchair Graphene Nanoribbons
The bottom-up approach to synthesize graphene nanoribbons strives not only to introduce a band gap into the electronic structure of graphene but also to accurately tune its value by designing both the width and edge structure of the ribbons with atomic precision. We report the synthesis of an armchair graphene nanoribbon with a width of nine carbon atoms on Au(111) through surface-assisted aryl–aryl coupling and subsequent cyclodehydrogenation of a properly chosen molecular precursor. By combining high-resolution atomic force microscopy, scanning tunneling microscopy, and Raman spectroscopy, we demonstrate that the atomic structure of the fabricated ribbons is exactly as designed. Angle-resolved photoemission spectroscopy and Fourier-transformed scanning tunneling spectroscopy reveal an electronic band gap of 1.4 eV and effective masses of ≈0.1 me for both electrons and holes, constituting a substantial improvement over previous efforts toward the development of transistor applications. We use ab initio calculations to gain insight into the dependence of the Raman spectra on excitation wavelength as well as to rationalize the symmetry-dependent contribution of the ribbons’ electronic states to the tunneling current. We propose a simple rule for the visibility of frontier electronic bands of armchair graphene nanoribbons in scanning tunneling spectroscopy.