Large Footprint Research Articles

Ultra-compact plasmonic electron accelerator on chip

Electron accelerators are essential equipment in science and industry. However, their large footprint, ranging in length up to kilometers, limits their availability. Hence, compact solution to this problem is one of the major trends in advance accelerators. In this paper, a design of micro-scale electron accelerator based on surface plasmons is proposed. Due to the near-field enhancement effect of plasmons, this structure establishes extremely strong field for acceleration within less than 1μm. Simulation results shows that driven by infrared laser, the designed accelerator with 5 GV/m peak field deliver more than 0.39 keV energy to electrons. Cascaded in chain, this structure can realize fully electron acceleration, and has the potential of becoming the critical element of future integrated free electron light source lab on chip.

Miniaturization of microwave planar circuits using composite microstrip/coplanar-waveguide transmission lines

The rapid development of modern communication systems propels the demand of high performance and miniaturization of RF/microwave components. Power dividers and filters are indispensable circuit components in amplifiers and antenna feed-networks which occupy large footprint of circuit broads, particularly at low frequency. The purpose of this study is to initiate a new type of planar transmission line with unique dual signal-path characteristics and ultimately achieve circuit miniaturization. In this paper, a novel technique of consolidating microstrip lines (MLs) and coplanar waveguide (CPW) lines on a single dielectric substrate, designated as the composite microstrip/CPW line (CM/CPW line), is proposed. With reference to the average line length of ML and CPW, it warrants a physical length reduction of 48.3% and further achieves a reduction of 80.8% after zigzagging CM/CPW line. The proof-of-concept design example of a two-way Wilkinson power divider (WPD) has demonstrated the effectiveness of the miniaturization technique. The simulation, in combination with the experimental results, validates that a size reduction factor of 86.9% when compared to conventional one has been achieved while maintaining a fractional bandwidth of 57.7% for the WPD.

High-Production-Rate Fabrication of Low-Loss Lithium Niobate Electro-Optic Modulators Using Photolithography Assisted Chemo-Mechanical Etching (PLACE)

Integrated thin-film lithium niobate (LN) electro-optic (EO) modulators of broad bandwidth, low insertion loss, low cost and high production rate are essential elements in contemporary interconnection industries and disruptive applications. Here, we demonstrated the design and fabrication of a high performance thin-film LN EO modulator using photolithography assisted chemo-mechanical etching (PLACE) technology. Our device shows a 3-dB bandwidth over 50 GHz, along with a comparable low half wave voltage-length product of 2.16 Vcm and a fiber-to-fiber insertion loss of 2.6 dB. The PLACE technology supports large footprint, high fabrication uniformity, competitive production rate and extreme low device optical loss simultaneously, our result shows promising potential for developing high-performance large-scale low-loss photonic integrated devices.

A Reconfigurable Two‐WSe2‐Transistor Synaptic Cell for Reinforcement Learning

Reward-modulated spike-timing-dependent plasticity (R-STDP) is a brain-inspired reinforcement learning (RL) rule, exhibiting potential for decision-making tasks and artificial general intelligence. However, the hardware implementation of the reward-modulation process in R-STDP usually requires complicated Si complementary metal-oxide-semiconductor (CMOS) circuit design that causes high power consumption and large footprint. Here, a design with two synaptic transistors (2T) connected in a parallel structure is experimentally demonstrated. The 2T unit based on WSe2 ferroelectric transistors exhibits reconfigurable polarity behavior, where one channel can be tuned as n-type and the other as p-type due to nonvolatile ferroelectric polarization. In this way, opposite synaptic weight update behaviors with multilevel (>6 bit) conductance states, ultralow nonlinearity (0.56/-1.23), and large Gmax /Gmin ratio of 30 are realized. By applying positive/negative reward to (anti-)STDP component of 2T cell, R-STDP learning rules are realized for training the spiking neural network and demonstrated to solve the classical cart-pole problem, exhibiting a way for realizing low-power (32 pJ per forward process) and highly area-efficient (100 µm2 ) hardware chip for reinforcement learning.

Recycling forward and backward frequency-multiplexed modes in a waveguide coupled to phased time-perturbed microrings for low-footprint neuromorphic computing

Optical structures can serve as low-power high-capacity alternatives of electronic processors for more efficient neuromorphic computing, but can suffer from large footprints and weak scalability. In this work, properly phased time-perturbed microrings side-coupled to a waveguide are utilized to realize a compact processor for linear transformations. We build up a synthetic frequency dimension to provide sufficient degrees of freedom, where the linear time-varying structures enable the linear intermixing and transformation of frequency-multiplexed data. Moreover, non-reciprocal and asymmetric flow of data in the forward and backward modes, due to phasing of the perturbations, helped to build up another synthetic dimension and to avoid physically repeating the processing elements, thus enabling a much more compact and scalable linear processor.

Using groundwater in the hyper-arid Atacama Desert to discover buried hydrothermal mineral deposits

ABSTRACT Epithermal deposits occur in many metallogenic belts globally and are economically important sources of precious and base metals. However, post-mineral cover (both transported and younger deposits) also occurs in many prospective districts and hinders exploration efforts. Hydrogeochemistry is a potentially useful exploration tool in these environments, with water–rock interaction shown from other studies to produce large hydrogeochemical footprints. Groundwaters from in and around the Vaquillas (Au-Ag) epithermal deposit in the pre-cordillera of northern Chile are typically Ca-SO4 2- type with meteoric upflow groundwaters interacting with the ore body and producing a significant downflow (down hydraulic head) mobility (up to 3 km) of (oxy)anions of arsenic and molybdenum, which contrasts with proximal trace cation anomalies. Multiple sulphur endmembers contribute to the dissolved sulphate concentrations, including evaporites, primary hypogene sulphide minerals, and a single δ34S(SO4) value of 30.1 ‰ which is potentially associated with disproportionated sulphur in sulphate. This study reports anomalism from one of the most dilute endmember groundwater compositions from Chile reported to date. The sub-alkaline and weakly oxygenated groundwaters are common in many aquifers globally, making the results applicable to many other metallogenic belts. It is also the first hydrogeochemical case study of an intermediate-sulfidation epithermal deposit. Despite the highly dilute groundwaters recharging aquifers at Vaquillas, we conclude hydrogeochemistry to be a highly effective exploration tool for epithermal and porphyry deposits in areas of significant post-mineral cover.

Boosting climate change mitigation potential of perennial lignocellulosic crops grown on marginal lands

Abstract Nitrogen fertilizer (NF) is a major uncertainty surrounding the greenhouse gas (GHG) emissions of lignocellulosic biofuels. NF enhances agronomic yields and soil C inputs via plant litters, but results in soil organic carbon (SOC) decomposition, soil N2O fluxes, and a large fossil energy footprint. Thus, whether NF is beneficial or detrimental to the GHG mitigation of biofuels is unknown. Here we show the potential GHG mitigation of fertilizing switchgrass (Panicum virgatum) at the NF rate that minimizes net GHG emissions across 7.1 million ha of marginal lands in the Midwest US, with long-term production advantages surpassing emitted GHG by 0.66 Mg CO2e ha-1 yr-1 on the aggregate. Marginal lands limited by poor N fertility could see a much greater benefit, but not SOC-rich lands, limited by low precipitation, or short growing seasons. The objectives of maximizing yield and minimizing GHG overlap only in a few environments, suggesting that maximum yield will reduce the climate benefit of cellulosic biofuels.

Compact RSFQ microwave pulse generator based on an integrated RF module for controlling superconducting qubits

An on-chip superconducting microwave pulse generator (MPG), incorporating rapid single-flux-quantum (RSFQ) logic gates and passive radio frequency (RF) components, is an energy-efficient solution for controlling superconducting quantum bits (qubits). Traditional RSFQ MPG designs, however, suffer from a large footprint as multiple RF components (including impedance matching networks, filters, etc.) are separately implemented one after another. In this Letter, we propose a compact on-chip MPG by introducing an integrated RF module. Using a generalized filter synthesis technique, the integrated RF module combines the functionalities of impedance matching and bandpass filtering as a single inductor-capacitor (LC) lumped network. Plenty of circuit elements can, thus, be saved, and the resultant LC network is applicable to both real and complex terminal impedance cases. For validation, a superconducting MPG adopting a fourth-order, 5 GHz centered, 100 MHz bandwidth integrated RF module is designed and fabricated, whose correct operations are confirmed through both simulation and measurement. These results raise the availability of controlling large-scale quantum systems by low-power superconductor technology in the future.

Sub-field of view surface thermal modeling of Cassini CIRS observations of Rhea during south polar winter

Rhea's exosphere is thought to originate from sources of carbon, water ice and other volatiles that arrived at Rhea by bombardment. Its seasonal variability is directly driven by polar surface temperatures allowing surface adsorption, and the persistence of source volatiles require that seasonal temperatures remain sufficiently cold to retain them. Cassini CIRS detected temperatures in Rhea's winter polar region of 23 K, amongst the coldest measured in the Solar System, but with a relatively large footprint we seek to add value to these observations by modeling sub field-of-view (FOV) temperature distribution and examining how the coldest scenes evolve on a seasonal basis.A simple, rough surface 1-dimensional thermal model is developed using a digital elevation map as input to a thermal model. We compared averaged rough and flat modeled FOVs to CIRS temperatures for a set of case study observations in the south polar region in winter darkness and found they both agree within expected CIRS error in all cases. We develop an asymmetric estimate of CIRS FOV temperature uncertainty, which is particularly important for very cold spectra to accurately represent the sensitivity of the instrument. This approach provides a conservative upper temperature limit to spectral fits whilst highlighting there is often little information to constrain the lower bound of temperature uncertainty in these cases.The distribution of modeled sub-FOV facet temperatures is explored for the full range of azimuth angles and slope gradients. More extensive and cooler temperatures were found in the rough model, complemented by fewer but warmer areas than the flat model. We find temperature contrasts of individual model facets of up to 15 K warmer and 11 K colder within some CIRS FOV when the rough model was compared flat, in a case study of scenes in winter darkness. This is due to the persistence of the seasonal thermal wave beneath the surface. We tested model sensitivity to thermal input parameters (thermal inertia and bolometric Bond albedo). These values are challenging to constrain due to limited observations and measurement noise and are expected to vary with subtle changes in surface characteristics. We found that within 20° of the pole, temperatures do not rise significantly above 80 K all year, implying that a variety of simple organic molecules, linear amides and other carbon-containing compounds would remain stable on a quasi-permanent basis, potentially until photolytic or radiolytic mechanisms liberate by-products which provide a source for exospheric constituents. The simple rough model indicates that some facets experience very different summer/winter season lengths than a flat model can represent, which is important in terms of the exospheric sequestration process.We use threshold temperatures of 55 K for CO2 and 30 K for O2 ice as proxies for their relative stability at the surface. The surface coverage of these temperatures with time was compared to modeled exospheric abundance by Teolis and Waite (2016). They are anti-correlated which is expected, but with slight differences in timings of minima and the equinoctial maxima. We find that the rough model suggests larger areas able to adsorb these species than the flat model, and that the cumulative influence of topography would have a direct relationship on the timing and abundance of the exosphere.We place Rhea's polar environment in context with other icy moons in the Saturnian system using the flat model. The satellites Tethys, Dione and Enceladus experience similar polar thermal regimes to Rhea (with exceptionally cold winters) and would potentially benefit from the consideration of topography in relation to exospheric modeling.

Determinants of Food Consumption Water Footprint in the MENA Region: The Case of Tunisia

Tunisia, like most countries in the Middle East and North Africa (MENA) region, has limited renewable water resources and is classified as a water stress country. The effects of climate change are exacerbating the situation. The agricultural sector is the main consumer (80%) of blue water reserves. In this study, to better understand the factors that influence the food water footprint of Tunisian consumers, we used a multiple linear regression model (MLR) to analyze data from 4853 households. The innovation in this paper consists of integrating effects of socio-economic, demographic, and geographic trends on the food consumption water footprint into the assessment of water and food security. The model results showed that regional variations in food choices meant large differences in water footprints, as hypothesized. Residents of big cities are more likely to have a large water footprint. Significant variability in water footprints, due to different food consumption patterns and socio-demographic characteristics, was also noted. Food waste is also one of the determining factors of households with a high water footprint. This study provides a new perspective on the water footprint of food consumption using “household” level data. These dietary water footprint estimates can be used to assess potential water demand scenarios as food consumption patterns change. Analysis at the geographic and socio-demographic levels helps to inform policy makers by identifying realistic dietary changes.

SRDiff: Single image super-resolution with diffusion probabilistic models

Single image super-resolution (SISR) aims to reconstruct high-resolution (HR) images from given low-resolution (LR) images. It is an ill-posed problem because one LR image corresponds to multiple HR images. Recently, learning-based SISR methods have greatly outperformed traditional methods. However, PSNR-oriented, GAN-driven and flow-based methods suffer from over-smoothing, mode collapse and large model footprint issues, respectively. To solve these problems, we propose a novel SISR diffusion probabilistic model (SRDiff), which is the first diffusion-based model for SISR. SRDiff is optimized with a variant of the variational bound on the data likelihood. Through a Markov chain, it can provide diverse and realistic super-resolution (SR) predictions by gradually transforming Gaussian noise into a super-resolution image conditioned on an LR input. In addition, we introduce residual prediction to the whole framework to speed up model convergence. Our extensive experiments on facial and general benchmarks (CelebA and DIV2K datasets) show that (1) SRDiff can generate diverse SR results with rich details and achieve competitive performance against other state-of-the-art methods, when given only one LR input; (2) SRDiff is easy to train with a small footprint(The word “footprint” in this paper represents “model size” (number of model parameters).); (3) SRDiff can perform flexible image manipulation operations, including latent space interpolation and content fusion.

Vertical Gentrification: A 3D Analysis of Luxury Housing Development in New York City

New York City has experienced a boom in elite “luxury” housing development. With large apartments and expansive amenity spaces, luxury buildings offer uncrowded living in an otherwise densely populated landscape. However, making space for these luxuries requires novel engineering, especially high-rise development. This article maps the expanding footprint of luxury real estate in three dimensions, analyzing 943 housing projects built between 2000 and 2020. It assesses how construction of taller buildings with larger footprints increased height and volume of the built environment, and how these landscape changes interact with social changes related to gentrification. On average, new build luxury development increased height by 6.8 stories and more than doubled building volumes. Building heights and volumes are also significantly larger than neighboring structures. The resulting intensification of land investment leads to new kinds of displacements, especially middle class displacement. Vertical development is closely associated with super-gentrification, the further intensification of gentrification processes in already gentrified or otherwise middle class neighborhoods.

Contingent stimulus delivery assay for zebrafish reveals a role for CCSER1 in alcohol preference.

Alcohol use disorders are complex, multifactorial phenomena with a large footprint within the global burden of diseases. Here, we report the development of an accessible, two-choice self-administration zebrafish assay (SAZA) to study the neurobiology of addiction. Using this assay, we first demonstrated that, although zebrafish avoid higher concentrations of alcohol, they are attracted to low concentrations. Pre-exposure to alcohol did not change this relative preference, but acute exposure to an alcohol deterrent approved for human use decreased alcohol self-administration. A pigment mutant used in whole-brain imaging studies displayed a similar relative alcohol preference profile; however, mutants in CCSER1, a gene associated with alcohol dependence in human genetic studies, showed a reversal in relative preference. The presence of a biphasic response (hormesis) in zebrafish validated a key aspect of vertebrate responses to alcohol. SAZA adds a new dimension for discovering novel alcohol deterrents and studying the neurogenetics of addiction using the zebrafish.

Tuberculosis care quality in urban Nigeria: A cross-sectional study of adherence to screening and treatment initiation guidelines in multi-cadre networks of private health service providers.

Nigeria has a high burden of tuberculosis (TB) and low case detection rates. Nigeria's large private health sector footprint represents an untapped resource for combating the disease. To examine the quality of private sector contributions to TB, the USAID-funded Sustaining Health Outcomes through the Private Sector (SHOPS) Plus program evaluated adherence to national standards for management of presumptive and confirmed TB among the clinical facilities, laboratories, pharmacies, and drug shops it trained to deliver TB services. The study used a standardized patient (SP) survey methodology to measure case management protocol adherence among 837 private and 206 public providers in urban Lagos and Kano. It examined two different scenarios: a "textbook" case of presumptive TB and a treatment initiation case where SPs presented as referred patients with confirmed TB diagnoses. Private sector results were benchmarked against public sector results. A bottleneck analysis examined protocol adherence departures at key points along the case management sequence that providers were trained to follow. Except for laboratories, few providers met the criteria for fully correct management of presumptive TB, though more than 70% of providers correctly engaged in TB screening. In the treatment initiation case 18% of clinical providers demonstrated fully correct case management. Private and public providers' adherence was not significantly different. Bottleneck analysis revealed that the most common deviations from correct management were failure to initiate sputum collection for presumptive patients and failure to conduct sufficiently thorough treatment initiation counseling for confirmed patients. This study found the quality of private providers' TB case management to be comparable to public providers in Nigeria, as well as to providers in other high burden countries. Findings support continued efforts to include private providers in Nigeria's national TB program. Though most providers fell short of desired quality, the bottleneck analysis points to specific issues that TB stakeholders can feasibly address with system- and provider-level interventions.

Sex-dependent influence of postweaning environmental enrichment in Angelman syndrome model mice.

IntroductionAngelman syndrome (AS) is a rare neurodevelopmental disorder caused by mutation or loss of UBE3A and marked by intellectual disability, ataxia, autism‐like symptoms, and other atypical behaviors. One route to treatment may lie in the role that environment plays early in postnatal life. Environmental enrichment (EE) is one manipulation that has shown therapeutic potential in preclinical models of many brain disorders, including neurodevelopmental disorders. Here, we examined whether postweaning EE can rescue behavioral phenotypes in Ube3a maternal deletion mice (AS mice), and whether any improvements are sex‐dependent.MethodsMale and female mice (C57BL/6J Ube3atm1Alb mice and wild‐type (WT) littermates; ≥10 mice/group) were randomly assigned to standard housing (SH) or EE at weaning. EE had a larger footprint, a running wheel, and a variety of toys that promoted foraging, burrowing, and climbing. Following 6 weeks of EE, animals were submitted to a battery of tests that reliably elicit behavioral deficits in AS mice, including rotarod, open field, marble burying, and forced swim; weights were also monitored.ResultsIn male AS‐EE mice, we found complete restoration of motor coordination, marble burying, and forced swim behavior to the level of WT‐SH mice. We also observed a complete normalization of exploratory distance traveled in the open field, but we found no rescue of vertical behavior or center time. AS‐EE mice also had weights comparable to WT‐SH mice. Intriguingly, in the female AS‐EE mice, we found a failure of EE to rescue the same behavioral deficits relative to female WT‐SH mice.ConclusionsEnvironmental enrichment is an effective route to correcting the most penetrant phenotypes in male AS mice but not female AS mice. This finding has important implications for the translatability of early behavioral intervention for AS patients, most importantly the potential dependency of treatment response on sex.

Software Modeling of Direct Wind to Energy Generator

The Wind Energy is one of the cleanest sources of electrical energy. Commercial wind turbines deliver the promise of being practically realizable with the cost of having an extremely large footprint and being haunted by the Betz Limit. This paper gives an idea of Direct Wind to Energy Conversion which aims to produce electrical energy by harnessing wind’s kinetic energy, devoid of any rotational mechanism. The method in question utilizes the technique of Electro-Hydro-Dynamic Atomization (EHDA) to atomize charged working fluid to deposit its particles on collector electrodes using wind as a carrier. A comparative analysis of several working fluids was conducted to optimize energy production. Simulations on the charged particle trajectory were conducted based on historical average wind speed values to enhance electrode placement which in turn led to design iterations and increased efficiency.

Generative Feature Extraction From Sentinel 1 and 2 Data for Prediction of Forest Aboveground Biomass in the Italian Alps

Aboveground biomass (AGB) is an important forest attribute directly linked to the forest carbon pool. The use of satellite remote sensing (RS) data has increased for AGB prediction due to their large footprint and low-cost availability. However, they have been limited due to saturation effect that leads to low prediction precision. In this study, we propose an innovative and dynamic architecture based on generative neural network that extracts target oriented generative features for predicting forest AGB using satellite RS data. These features are more resilient to mixed forest types and geographical conditions as compared to the traditional features and models. The effectiveness of the proposed features was assessed by experiments performed using multispectral (MS), Synthetic aperture Radar (SAR) and combined dual-source (DS) datasets. The proposed model achieved best performance in terms of precision, model agreement and overfitting as compared to the other conventional models for all analyzed datasets. The t-distributed stochastic neighbor embedding (t-SNE) scatterplots of the generative features clearly show one dimension of the feature space associated with the target AGB. Feature importance analysis indicated that the produced generative features were more significant than the conventional analytical features. Also, the model provided a robust framework for homogeneous fusion of multi-sensor features from satellite RS data for predicting AGB.

Construction of dPCR and qPCR integrated system based on commercially available low-cost hardware.

Fluorescent quantitative PCR (qPCR) and digital PCR (dPCR) are two mainstream nucleic acid quantification technologies. However, commercial dPCR and qPCR instruments have a low integration, a high price, and a large footprint. To solve these shortcomings, we introduce a compound PCR system with both qPCR and dPCR functions. All the hardware used in this compound PCR system is commercially available and low-cost, and free software was used to realize the absolute quantification of nucleic acids. The compound PCR provides two working modes. In the qPCR mode, thermal cycling is realized by controlling the reciprocating motion of the x axis. The heating rate is 1.25 °C s-1 and the cooling rate is 1.75 °C s-1. We performed amplification experiments of the PGEM-3zf (+)1 gene. The performance level was similar to commercial qPCR instruments. In the dPCR mode, the heating rate is 0.5 °C s-1 and the cooling rate is 0.6 °C s-1. We performed the UPE-Q gene amplification and used the sequential actions of the two-dimensional mechanical sliders to scan the reaction products and used the method of regional statistics and back-inference threshold to get test results. The result we got was 1208 copies per μL-1, which was similar to expectations.

Spatial Heterogeneity of Albedo at Subpixel Satellite Scales and its Effect in Validation: Airborne Remote Sensing Results From HiWATER

Characterizing the subpixel heterogeneity within satellite pixels is a key issue in validation. Nevertheless, it is challenging due to multi-scale problems in the geological description based on remote sensing. Based on an airborne platform, the multi-scale variation laws of several key indicators in validation including spatial heterogeneity (SH), representativeness errors, and representative area with subpixel size were analyzed and discussed. Furthermore, this article discussed the optimal subpixel size to assess SH within a coarse pixel and the optimal footprint of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in situ</i> measurements for building dense and sparse validation networks. SH decreases with the increase of subpixel size. And a reduction of about 10% can be obtained from 5 m <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times 5$ </tex-math></inline-formula> m to 150 m <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times150$ </tex-math></inline-formula> m subpixel size, depending on the degree of SH within the typical satellite pixels. And the sensitiveness of SH to subpixel size decreases gradually with the increasing of subpixel size. Ideally, SH should be assessed using maps with pixel sizes corresponding to the footprint of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in situ</i> measurements. Regarding the deployment of future validation networks, the footprint of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in situ</i> sites should be designed at least larger than 25 m for dense networks. And much larger footprints (e.g., 100 m) are preferred in designing sparse networks. The representativeness error is not fully related to subpixel sizes because it is affected by many factors. The findings are also transferable to model evaluation when comparing model grid values to local observations.

Searching for CNN Architectures for Remote Sensing Scene Classification

Convolutional neural network (CNN) models for remote sensing (RS) scene classification are largely built on pretrained networks that are trained on the general-purpose ImageNet dataset in computer vision. The pretrained networks can easily be adapted for transfer learning in RS scene classification. However, the accuracy of transfer learning may decline as RS images are considerably different from other images. Thus, the pretrained CNN model learned on ImageNet may not be sufficient for the accurate classification of RS image scenes. Furthermore, most of the pretrained models have large memory footprints, which place a further burden on computational requirements. In this work, we explore SLGE-based random search with early stopping in the search for CNN architectures for both single-label and multilabel RS scene classification tasks. In SLGE, the architecture search space is capable of representing multipath Inception-like modular cells with skip-connections similar to human-expert designs. The experimental results on four RS scene classification benchmarks show that the automatically discovered networks demonstrate the promising capability in classifying multispectral satellite image scenes compared with fine-tuned pretrained CNN models. Using fewer parameters with 0.56B FLOPS, our best network achieves a classification accuracy rate of 96.56% and 96.10% on NWPU-RESISC45 single-label and AID single-label RGB aerial image datasets, respectively, and the classification accuracy rate of 99.76% and 93.89% on EuroSAT single-label and BigEarthNet multilabel multispectral satellite image datasets, respectively. The results position our approach among the best of the state of the art.