Toward Quantitative Surface-Enhanced Raman Scattering with Plasmonic Nanoparticles: Multiscale View on Heterogeneities in Particle Morphology, Surface Modification, Interface, and Analytical Protocols.

Surface-enhanced Raman scattering (SERS) provides significantly enhanced Raman scattering signals from molecules adsorbed on plasmonic nanostructures, as well as the molecules' vibrational fingerprints. Plasmonic nanoparticle systems are particularly powerful for SERS substrates as they provide a wide range of structural features and plasmonic couplings to boost the enhancement, often up to >108-1010. Nevertheless, nanoparticle-based SERS is not widely utilized as a means for reliable quantitative measurement of molecules largely due to limited controllability, uniformity, and scalability of plasmonic nanoparticles, poor molecular modification chemistry, and a lack of widely used analytical protocols for SERS. Furthermore, multiscale issues with plasmonic nanoparticle systems that range from atomic and molecular scales to assembled nanostructure scale are difficult to simultaneously control, analyze, and address. In this perspective, we introduce and discuss the design principles and key issues in preparing SERS nanoparticle substrates and the recent studies on the uniform and controllable synthesis and newly emerging machine learning-based analysis of plasmonic nanoparticle systems for quantitative SERS. Specifically, the multiscale point of view with plasmonic nanoparticle systems toward quantitative SERS is provided throughout this perspective. Furthermore, issues with correctly estimating and comparing SERS enhancement factors are discussed, and newly emerging statistical and artificial intelligence approaches for analyzing complex SERS systems are introduced and scrutinized to address challenges that cannot be fully resolved through synthetic improvements.

Synthesis, Assembly, Optical Properties, and Sensing Applications of Plasmonic Gap Nanostructures.

Plasmonic gap nanostructures (PGNs) have been extensively investigated mainly because of their strongly enhanced optical responses, which stem from the high intensity of the localized field in the nanogap. The recently developed methods for the preparation of versatile nanogap structures open new avenues for the exploration of unprecedented optical properties and development of sensing applications relying on the amplification of various optical signals. However, the reproducible and controlled preparation of highly uniform plasmonic nanogaps and the prediction, understanding, and control of their optical properties, especially for nanogaps in the nanometer or sub-nanometer range, remain challenging. This is because subtle changes in the nanogap significantly affect the plasmonic response and are of paramount importance to the desired optical performance and further applications. Here, recent advances in the synthesis, assembly, and fabrication strategies, prediction and control of optical properties, and sensing applications of PGNs are discussed, and perspectives toward addressing these challenging issues and the future research directions are presented.

Hierarchic Interfacial Nanocube Assembly for Sensitive, Selective, and Quantitative DNA Detection with Surface-Enhanced Raman Scattering.

Surface-enhanced Raman scattering (SERS)-based sensing is promising in that it has potential to allow for highly sensitive, selective, and multiplexed detection and imaging. However, the controlled assembly and gap formation between plasmonic particles for generating strong SERS signals in a quantitative manner is highly challenging, especially on biodetection platforms, and particle-to-particle variation in the signal enhancement can vary by several orders of magnitude in a single batch, largely limiting the reliable use of SERS for practical sensing applications. Here, a hierarchic-nanocube-assembly based SERS (H-Cube-SERS) bioassay to controllably amplify the electromagnetic field between gold nanocubes (AuNCs) is developed. Based on this strategy, H-Cube-SERS assay allows for detecting target DNA with a wide dynamic range from 100 aM to 10 pM concentrations in a stable and reproducible manner. It is also found that the uniformly formed AuNCs with flat surfaces are much more suitable for highly sensitive, reliable, and quantitative biodetection assays due to faster DNA binding kinetics, sharper DNA melting transition, wider hot spot regions, and less dependence on light polarization direction than spherical Au nanoparticles with curved interfaces. This work paves the pathways to the quantitative and sensitive biodetection on a SERS platform and can be extended to other particle assembly systems.

Dealloyed Intra-Nanogap Particles with Highly Robust, Quantifiable Surface-Enhanced Raman Scattering Signals for Biosensing and Bioimaging Applications.

Uniformly controlling a large number of metal nanostructures with a plasmonically enhanced signal to generate quantitative optical signals and the widespread use of these structures for surface-enhanced Raman scattering (SERS)-based biosensing and bioimaging applications are of paramount importance but are extremely challenging. Here, we report a highly controllable, facile selective-interdiffusive dealloying chemistry for synthesizing the dealloyed intra-nanogap particles (DIPs) with a ∼2 nm intragap in a high yield (∼95%) without the need for an interlayer. The SERS signals from DIPs are highly quantitative and polarization-independent with polarized laser sources. Remarkably, all the analyzed particles displayed the SERS enhancement factors (EFs) of ≥1.1 × 108 with a very narrow distribution of EFs. Finally, we show that DIPs can be used as ultrasensitive SERS-based DNA detection probes for detecting 10 aM to 1 pM target concentrations and highly robust, quantitative real-time cell imaging probes for long-term imaging with low laser power and short exposure time.

Dual-memory neural networks for modeling cognitive activities of humans via wearable sensors.

Wearable devices, such as smart glasses and watches, allow for continuous recording of everyday life in a real world over an extended period of time or lifelong. This possibility helps better understand the cognitive behavior of humans in real life as well as build human-aware intelligent agents for practical purposes. However, modeling the human cognitive activity from wearable-sensor data stream is challenging because learning new information often results in loss of previously acquired information, causing a problem known as catastrophic forgetting. Here we propose a deep-learning neural network architecture that resolves the catastrophic forgetting problem. Based on the neurocognitive theory of the complementary learning systems of the neocortex and hippocampus, we introduce a dual memory architecture (DMA) that, on one hand, slowly acquires the structured knowledge representations and, on the other hand, rapidly learns the specifics of individual experiences. The DMA system learns continuously through incremental feature adaptation and weight transfer. We evaluate the performance on two real-life datasets, the CIFAR-10 image-stream dataset and the 46-day Lifelog dataset collected from Google Glass, showing that the proposed model outperforms other online learning methods.