Search

Highly sensitive, low-power, and chip-scale H2 gas sensors are of great interest to both academia and industry. Field-effect transistors (FETs) functionalized with Pd nanoparticles (PdNPs) have recently emerged as promising candidates for such H2 sensors. However, their sensitivity is limited by weak capacitive coupling between PdNPs and the FET channel. Herein we report a nanoscale FET gas sensor, where electrons can tunnel between the channel and PdNPs and thus equilibrate them. Gas reaction with PdNPs perturbs the equilibrium, and therefore triggers electron transfer between the channel and PdNPs via trapping or de-trapping with the PdNPs to form a new balance. This direct communication between the gas reaction and the channel enables the most efficient signal transduction. Record-high responses to 1–1000 ppm H2 at room temperature with detection limit in the low ppb regime and ultra-low power consumption of ~ 300 nW are demonstrated. The same mechanism could potentially be used for ultrasensitive detection of other gases. Our results present a supersensitive FET gas sensor based on electron trapping effect in nanoparticles. Highly sensitive, low-power, and chip-scale gas sensors are of great interest to both academia and industry. The authors developed a nanotransistor-based gas sensor with high sensitivity enabled by electron trapping effect in nanoparticles. [ABSTRACT FROM AUTHOR]

Air pollution monitoring using mobile ground-based measurement platforms can provide high-quality spatiotemporal air pollution information. As mobile air quality monitoring campaigns extend to entire fleets of vehicles and integrate smaller-scale air quality sensors, it is important to address the need to assess these measurements in a scalable manner. We explore the collocation-based evaluation of air quality measurements in a mobile platform using fixed regulatory sites as a reference. We compare two approaches: a standard collocation assessment technique, in which the mobile platform is parked near the fixed regulatory site for a period of time, and an expanded approach, which uses measurements while the mobile platform is in motion in the general vicinity of the fixed regulatory site. Based on the availability of fixed-reference-site data, we focus on three pollutants (ozone, nitrogen dioxide, and nitric oxide) with distinct atmospheric lifetimes and behaviors. We compare measurements from a mobile laboratory with regulatory site measurements in Denver, CO, USA, and in the San Francisco Bay Area, CA, USA. Our 1-month Denver dataset includes both parked collocation periods near the fixed regulatory sites and general driving patterns around the sites, allowing a direct comparison of the parked and mobile collocation techniques on the same dataset. We show that the mobile collocation approach produces similar performance statistics, including coefficients of determination and mean bias errors, to the standard parked collocation technique. This is particularly true when the comparisons are restricted to specific road types, with residential streets showing the closest agreement and highways showing the largest differences. We extend our analysis to a larger (yearlong) dataset in California, where we explore the relationships between the mobile measurements and the fixed reference sites on a larger scale. We show that using a 40 h running median converges to within ± 4 ppbv of the fixed reference site for nitrogen dioxide and ozone and up to about 8 ppbv for nitric oxide. We propose that this agreement can be leveraged to assess instrument performance over time during large-scale mobile monitoring campaigns. We demonstrate an example of how such relationships can be employed during large-scale monitoring campaigns using small sensors to identify potential measurement biases. [ABSTRACT FROM AUTHOR]

Analog memory presents a promising solution in the face of the growing demand for energy-efficient artificial intelligence (AI) at the edge. In this study, we demonstrate efficient deep neural network transfer learning utilizing hardware and algorithm co-optimization in an analog resistive random-access memory (ReRAM) array. For the first time, we illustrate that in open-loop deep neural network (DNN) transfer learning for image classification tasks, convergence rates can be accelerated by approximately 3.5 times through the utilization of cooptimized analog ReRAM hardware and the hardware-aware Tiki-Taka v2 (TTv2) algorithm. A simulation based on statistical 14 nm CMOS ReRAM array data provides insights into the performance of transfer learning on larger network workloads, exhibiting notable improvement over conventional training with random initialization. This study shows that analog DNN transfer learning using an optimized ReRAM array can achieve faster convergence with a smaller dataset compared to training from scratch, thus augmenting AI capability at the edge. [ABSTRACT FROM AUTHOR]