matOptimize: A parallel tree optimization method enables online phylogenetics for SARS-CoV-2

Phylogenetic tree optimization is necessary for precise analysis of evolutionary and transmission dynamics, but existing tools are inadequate for handling the scale and pace of data produced during the COVID-19 pandemic. One transformative approach, online phylogenetics, aims to incrementally add samples to an ever-growing phylogeny, but there are no previously-existing approaches that can efficiently optimize this vast phylogeny under the time constraints of the pandemic. Here, we present matOptimize, a fast and memory-efficient phylogenetic tree optimization tool based on parsimony that can be parallelized across multiple CPU threads and nodes, and provides orders of magnitude improvement in runtime and peak memory usage compared to existing state-of-the-art methods. We have developed this method particularly to address the pressing need during the COVID-19 pandemic for daily maintenance and optimization of a comprehensive SARS-CoV-2 phylogeny. Thus, our approach addresses an important need for daily maintenance and refinement of a comprehensive SARS-CoV-2 phylogeny.

Multiterminal Pathfinding in Practical VLSI Systems with Deep Neural Networks

A multiterminal obstacle-avoiding pathfinding approach is proposed. The approach is inspired by deep image learning. The key idea is based on training a conditional generative adversarial network (cGAN) to interpret a pathfinding task as a graphical bitmap and consequently map a pathfinding task onto a pathfinding solution represented by another bitmap. To enable the proposed cGAN pathfinding, a methodology for generating synthetic dataset is also proposed. The cGAN model is implemented in Python/Keras, trained on synthetically generated data, evaluated on practical VLSI benchmarks, and compared with state-of-the-art. Due to effective parallelization on GPU hardware, the proposed approach yields a state-of-the-art like wirelength and a better runtime and throughput for moderately complex pathfinding tasks. However, the runtime and throughput with the proposed approach remain constant with an increasing task complexity, promising orders of magnitude improvement over state-of-the-art in complex pathfinding tasks. The cGAN pathfinder can be exploited in numerous high throughput applications, such as, navigation, tracking, and routing in complex VLSI systems. The last is of particular interest to this work.

Study of the impact of interface traps associated with SiNX passivation on AlGaN/GaN MIS-HEMTs

In this work, we show that a bilayer SiNx passivation scheme which includes a high-temperature annealed SiNx as gate dielectric, significantly improves both ON and OFF state performance of AlGaN/GaN MISHEMTs. From devices with different SiNx passivation schemes, surface and bulk leakage paths were determined. Temperature-dependent MESA leakage studies showed that the surface conduction could be explained using a 2-D variable range hopping mechanism along with the mid-gap interface states at the GaN(cap)/ SiNx interface generated due to the Ga-Ga metal like bonding states. It was found that the high temperature annealed SiNx gate dielectric exhibited the lowest interface state density and a two-step C-V indicative of a superior quality SiNx/GaN interface as confirmed from conductance and capacitance measurements. High-temperature annealing helps in the formation of Ga-N bonding states, thus reducing the shallow metal-like interface states. MISHEMT measurements showed a significant reduction in gate leakage and a 4-orders of magnitude improvement in the ON/OFF ratio while increasing the saturation drain current (IDS) by a factor of 2. Besides, MISHEMTs with 2-step SiNx passivation exhibited a relatively flat transconductance profile, indicative of lower interface states density. The dynamic Ron with gate and drain stressing measurements also showed about 3x improvements in devices with bilayer SiNx passivation.

Comparing different nonlinearities in readout systems for optical neuromorphic computing networks

Nonlinear activation is a crucial building block of most machine-learning systems. However, unlike in the digital electrical domain, applying a saturating nonlinear function in a neural network in the analog optical domain is not as easy, especially in integrated systems. In this paper, we first investigate in detail the photodetector nonlinearity in two main readout schemes: electrical readout and optical readout. On a 3-bit-delayed XOR task, we show that optical readout trained with backpropagation gives the best performance. Furthermore, we propose an additional saturating nonlinearity coming from a deliberately non-ideal voltage amplifier after the detector. Compared to an all-optical nonlinearity, these two kinds of nonlinearities are extremely easy to obtain at no additional cost, since photodiodes and voltage amplifiers are present in any system. Moreover, not having to design ideal linear amplifiers could relax their design requirements. We show through simulation that for long-distance nonlinear fiber distortion compensation, using only the photodiode nonlinearity in an optical readout delivers BER improvements over three orders of magnitude. Combined with the amplifier saturation nonlinearity, we obtain another three orders of magnitude improvement of the BER.

Inverse Dirichlet Weighting Enables Reliable Training of Physics Informed Neural Networks

We characterize and remedy a failure mode that may arise from multi-scale dynamics with scale imbalances during training of deep neural networks, such as Physics Informed Neural Networks (PINNs). PINNs are popular machine-learning templates that allow for seamless integration of physical equation models with data. Their training amounts to solving an optimization problem over a weighted sum of data-fidelity and equation-fidelity objectives. Conflicts between objectives can arise from scale imbalances, heteroscedasticity in the data, stiffness of the physical equation, or from catastrophic interference during sequential training. We explain the training pathology arising from this and propose a simple yet effective inverse Dirichlet weighting strategy to alleviate the issue. We compare with Sobolev training of neural networks, providing the baseline of analytically ε-optimal training. We demonstrate the effectiveness of inverse Dirichlet weighting in various applications, including a multi-scale model of active turbulence, where we show orders of magnitude improvement in accuracy and convergence over conventional PINN training. For inverse modeling using sequential training, we find that inverse Dirichlet weighting protects a PINN against catastrophic forgetting.

The Role of Cloud Computing in a Modern Reverse Engineering Workflow at the 5nm Node and Beyond

A modern reverse engineering (RE) workflow contains many challenges, especially as process nodes drop below the 5nm node. With increased complexity, more circuitry is packed into a smaller area, requiring large quantities of raw data collected and subsequently processed to help reconstruct the original schematics. By leveraging inexpensive cloud computing, orders of magnitude improvement in throughput were achieved for 2D image registration and high quality image segmentation was achieved using machine learning.

Rotational Seismology with a Quartz Rotation Sensor

We describe the construction and performance of a new high-precision ground- or platform-rotation sensor called the Quartz Rotation Sensor (QRS). The QRS is a mechanical angular accelerometer that senses rotational torque with an inherently digital, load-sensitive resonant quartz crystal. The noise floor is measured to be ∼45 pico-radians/root (Hz) near 1 Hz, and the resonant period of the sensor is about 10 s, making it a broadband sensor. Among similarly sized broadband rotation sensors, this represents more than two orders of magnitude improvement in noise floor near 0.1 Hz. We present measurements of rotational components of teleseismic waves recorded with the sensor at a vault. The QRS is useful for rotational seismology and for improving low-frequency seismic isolation in demanding applications such as the Laser Interferometer Gravitational-Wave Observatories.

Iodide Functionalized Paper-Based SERS Sensors for Improved Detection of Narcotics

An inkjet-printed paper-based Surface-enhanced Raman scattering (SERS) sensor is a robust and versatile device that provides trace sensing capabilities for the detection and analysis of narcotics and drugs. Such sensors generally work well for analytes with good binding affinity towards the Au or Ag plasmonic nanoparticles (NPs) resident in the sensors. In this report, we show that iodide functionalization of the printed sensors helps to remove adsorbed contaminants from AuNP surfaces enabling superior performance with improved detection of narcotics such as fentanyl, heroin and cocaine by SERS. SERS signals are easily doubled with the iodide-functionalized sensors which also showed orders of magnitude improvement in detection limit. In this report, we show that a short (90 s) iodide treatment of the sensors significantly improved the detection of heroin. We propose that iodide functionalization be integrated into field detection kits through the solvent that wets paper-based sensor prior to swabbing for narcotics. Alternatively, we have also demonstrated that iodide functionalized sensors can be stored in ambient for up to 1 week and retain the improved performance towards heroin detection. This report will help to significantly improve the performance of paper-based sensors for field detection of narcotic drugs.

Energy-efficient ultrafast nucleation of single and multiple antiferromagnetic skyrmions using in-plane spin polarized current

We numerically investigate the ultrafast nucleation of antiferromagnetic (AFM) skyrmion using in-plane spin-polarized current and present its key advantages over out-of-plane spin-polarized current. We show that the threshold current density required for the creation of AFM skyrmion is almost an order of magnitude lower for the in-plane spin-polarized current. The nucleation time for the AFM skyrmion is found to be $$12-7$$ 12 - 7 ps for the corresponding current density of 1–$$3\times 10^{13}~\text{A/m}^{2}$$ 3 × 10 13 A/m 2 . We also demonstrate ultrafast nucleation of multiple AFM skyrmions that is possible only with in-plane spin polarized current and discuss how the current pulse width can be used to control the number of AFM skyrmions. The results show more than one order of magnitude improvement in energy consumption for ultrafast nucleation of AFM skyrmions using in-plane spin-polarized current, which is promising for applications such as logic gates, racetrack memory, and neuromorphic computing.

High AC and DC Electroconductivity of Scalable and Economic Graphite–Diamond Polylactide Nanocomposites

Two types of graphite/diamond (GD) particles with different ash content was applied to prepare new electroconductive polylactide (PLA)-based nanocomposites. Four samples of nanocomposites for each type of GD particles with mass fraction 0.01, 0.05, 0.10, and 0.15 were prepared via an easily scalable method—melt blending. The samples were subjected to the studies of electrical properties via broadband dielectric spectroscopy. The results indicated up to eight orders of magnitude improvement in the electrical conductivity and electrical permittivity of the most loaded nanocomposites, in reference to the neat PLA. Additionally, the influence of ash content on the electrical conductivity of the nanocomposites revealed that technologically less-demanding fillers, i.e., of higher ash content, were the most beneficial in the light of nanofiller dispersibility and the final properties.