Explore the vast range of titles available.

Models of artificial intelligence (AI) that have billions of parameters can achieve high accuracy across a range of tasks 1 , 2 , but they exacerbate the poor energy efficiency of conventional general-purpose processors, such as graphics processing units or central processing units. Analog in-memory computing (analog-AI) 3 – 7 can provide better energy efficiency by performing matrix–vector multiplications in parallel on ‘memory tiles’. However, analog-AI has yet to demonstrate software-equivalent (SW eq ) accuracy on models that require many such tiles and efficient communication of neural-network activations between the tiles. Here we present an analog-AI chip that combines 35 million phase-change memory devices across 34 tiles, massively parallel inter-tile communication and analog, low-power peripheral circuitry that can achieve up to 12.4 tera-operations per second per watt (TOPS/W) chip-sustained performance. We demonstrate fully end-to-end SW eq accuracy for a small keyword-spotting network and near-SW eq accuracy on the much larger MLPerf 8 recurrent neural-network transducer (RNNT), with more than 45 million weights mapped onto more than 140 million phase-change memory devices across five chips. A low-power chip that runs AI models using analog rather than digital computation shows comparable accuracy on speech-recognition tasks but is more than 14 times as energy efficient.

NAD⁺ (oxidized form of NAD:nicotinamide adenine dinucleotide)–reducing soluble [NiFe]-hydrogenase (SH) is phylogenetically related to NADH (reduced form of NAD⁺):quinone oxidoreductase (complex I), but the geometrical arrangements of the subunits and Fe–S clusters are unclear. Here, we describe the crystal structures of SH in the oxidized and reduced states. The cluster arrangement is similar to that of complex I, but the subunits orientation is not, which supports the hypothesis that subunits evolved as prebuilt modules. The oxidized active site includes a six-coordinate Ni, which is unprecedented for hydrogenases, whose coordination geometry would prevent O₂ from approaching. In the reduced state showing the normal active site structure without a physiological electron acceptor, the flavin mononucleotide cofactor is dissociated, which may be caused by the oxidation state change of nearby Fe–S clusters and may suppress production of reactive oxygen species.

The Southern Ocean (44-75° S) plays a critical role in the global carbon cycle, yet remains one of the most poorly sampled ocean regions. Different approaches have been used to estimate sea-air CO2 fluxes in this region: synthesis of surface ocean observations, ocean biogeochemical models, and atmospheric and ocean inversions. As part of the RECCAP (REgional Carbon Cycle Assessment and Processes) project, we combine these different approaches to quantify and assess the magnitude and variability in Southern Ocean sea-air CO2 fluxes between 1990-2009. Using all models and inversions (26), the integrated median annual sea-air CO2 flux of -0.42 ± 0.07 Pg C yr-1 for the 44-75° S region, is consistent with the -0.27 ± 0.13 Pg C yr-1 calculated using surface observations. The circumpolar region south of 58° S has a small net annual flux (model and inversion median: -0.04 ± 0.07 Pg C yr-1 and observations: +0.04 ± 0.02 Pg C yr-1 ), with most of the net annual flux located in the 44 to 58° S circumpolar band (model and inversion median: -0.36 ± 0.09 Pg C yr-1 and observations: -0.35 ± 0.09 Pg C yr-1 ). Seasonally, in the 44-58° S region, the median of 5 ocean biogeochemical models captures the observed sea-air CO2 flux seasonal cycle, while the median of 11 atmospheric inversions shows little seasonal change in the net flux. South of 58° S, neither atmospheric inversions nor ocean biogeochemical models reproduce the phase and amplitude of the observed seasonal sea-air CO2 flux, particularly in the Austral Winter. Importantly, no individual atmospheric inversion or ocean biogeochemical model is capable of reproducing both the observed annual mean uptake and the observed seasonal cycle. This raises concerns about projecting future changes in Southern Ocean CO2 fluxes. The median interannual variability from atmospheric inversions and ocean biogeochemical models is substantial in the Southern Ocean; up to 25% of the annual mean flux, with 25% of this interannual variability attributed to the region south of 58° S. Resolving long-term trends is difficult due to the large interannual variability and short time frame (1990-2009) of this study; this is particularly evident from the large spread in trends from inversions and ocean biogeochemical models. Nevertheless, in the period 1990-2009 ocean biogeochemical models do show increasing oceanic uptake consistent with the expected increase of -0.05 Pg C yr-1 decade-1 . In contrast, atmospheric inversions suggest little change in the strength of the CO2 sink broadly consistent with the results of Le Quéré et al. (2007).

The oceanic uptake and resulting storage of the anthropogenic CO2 (Cant) that humans have emitted into the atmosphere moderates climate change. Yet our knowledge about how this uptake and storage has progressed in time remained limited. Here, we determine decadal trends in the storage of Cant by applying the eMLR(C*) regression method to ocean interior observations collected repeatedly since the 1990s. We find that the global ocean storage of Cant grew from 1994 to 2004 by 29 ± 3 Pg C dec−1 and from 2004 to 2014 by 27 ± 3 Pg C dec−1 (±1σ). The storage change in the second decade is about 15 ± 11% lower than one would expect from the first decade and assuming proportional increase with atmospheric CO2. We attribute this reduction in sensitivity to a decrease of the ocean buffer capacity and changes in ocean circulation. In the Atlantic Ocean, the maximum storage rate shifted from the Northern to the Southern Hemisphere, plausibly caused by a weaker formation rate of North Atlantic Deep Waters and an intensified ventilation of mode and intermediate waters in the Southern Hemisphere. Our estimates of the Cant accumulation differ from cumulative net air‐sea flux estimates by several Pg C dec−1, suggesting a substantial and variable, but uncertain net loss of natural carbon from the ocean. Our findings indicate a considerable vulnerability of the ocean carbon sink to climate variability and change. Plain Language Summary The ocean takes up about 30% of the anthropogenic CO2 that is emitted to the atmosphere by human activities. The removal of this anthropogenic CO2 from the atmosphere counteracts climate change. The rate at which the ocean takes up anthropogenic CO2 is controlled by its transport from the surface to the depth of the ocean, where most of it accumulates. Thus, we can quantify and understand the oceanic uptake by keeping track of the accumulation of anthropogenic CO2 in the ocean interior. In this study, we use a global collection of measurements of CO2 in seawater to infer the temporal evolution of this accumulation between 1994 and 2014. We find that the ocean continued to act as a strong sink for CO2 over this period, removing, on average, nearly 30 billion tons of carbon per decade. However, we also detect a possible weakening of this uptake, since the accumulation of anthropogenic CO2 during the second decade was not as large as expected from the increase in atmospheric CO2. Our findings suggest that the ocean sink for CO2 might further shrink as climate change progresses. Key Points The global ocean storage of anthropogenic carbon grew by 29 ± 3 and 27 ± 3 Pg C dec−1 from 1994 to 2004 and 2004 to 2014, respectively The change in oceanic storage of anthropogenic carbon relative to the atmospheric CO2 growth decreased by 15 ± 11% from the first to the second decade This reduction is attributed to a decrease of the ocean buffer capacity and changes in ocean circulation

Tuning the nerves Different neurons initiate their action potentials (or spikes) at different points down their axons, but the functional implications of this have been unclear. Kuba et al . focus on nerve cells in a bird's auditory system: the nucleus laminaris is a binaural coincidence detector, and is a good model for examining this phenomenon. They find that neurons that initiate spikes closer to the cell body (or soma) are tuned to sounds with lower frequencies. Computer modelling suggests that spike initiation sites may be key to coincidence detection by other neurons as well. Study of nerve cells in a bird's auditory system shows that those where spikes initiate closer to the cell body (or soma) are tuned to sounds with lower frequencies. Computer modelling suggests that spike initiation sites may also be key to coincidence detection by other neurons. Neurons initiate spikes in the axon initial segment or at the first node in the axon 1 , 2 , 3 , 4 . However, it is not yet understood how the site of spike initiation affects neuronal activity and function. In nucleus laminaris of birds, neurons behave as coincidence detectors for sound source localization and encode interaural time differences (ITDs) separately at each characteristic frequency (CF) 5 , 6 , 7 . Here we show, in nucleus laminaris of the chick, that the site of spike initiation in the axon is arranged at a distance from the soma, so as to achieve the highest ITD sensitivity at each CF. Na + channels were not found in the soma of high-CF (2.5–3.3 kHz) and middle-CF (1.0–2.5 kHz) neurons but were clustered within a short segment of the axon separated by 20–50 μm from the soma; in low-CF (0.4–1.0 kHz) neurons they were clustered in a longer stretch of the axon closer to the soma. Thus, neurons initiate spikes at a more remote site as the CF of neurons increases. Consequently, the somatic amplitudes of both orthodromic and antidromic spikes were small in high-CF and middle-CF neurons and were large in low-CF neurons. Computer simulation showed that the geometry of the initiation site was optimized to reduce the threshold of spike generation and to increase the ITD sensitivity at each CF. Especially in high-CF neurons, a distant localization of the spike initiation site improved the ITD sensitivity because of electrical isolation of the initiation site from the soma and dendrites, and because of reduction of Na + -channel inactivation by attenuating the temporal summation of synaptic potentials through the low-pass filtering along the axon.

Chondroitin sulfate and heparan sulfate proteoglycans are major components of the cell surface and extracellular matrix in the brain. Both chondroitin sulfate and heparan sulfate are unbranched highly sulfated polysaccharides composed of repeating disaccharide units of glucuronic acid and N -acetylgalactosamine, and glucuronic acid and N -acetylglucosamine, respectively. During their biosynthesis in the Golgi apparatus, these glycosaminoglycans are highly modified by sulfation and C5 epimerization of glucuronic acid, leading to diverse heterogeneity in structure. Their structures are strictly regulated in a cell type-specific manner during development partly by the expression control of various glycosaminoglycan-modifying enzymes. It has been considered that specific combinations of glycosaminoglycan-modifying enzymes generate specific functional microdomains in the glycosaminoglycan chains, which bind selectively with various growth factors, morphogens, axon guidance molecules and extracellular matrix proteins. Recent studies have begun to reveal that the molecular interactions mediated by such glycosaminoglycan microdomains play critical roles in the various signaling pathways essential for the development of the brain.

This paper introduces a new Earth's atmosphere‐ionosphere coupled model that treats seamlessly the neutral atmospheric region from the troposphere to the thermosphere as well as the thermosphere‐ionosphere interaction including the electrodynamics self‐consistently. The model is especially useful for the study of vertical connection between the meteorological phenomena and the upper atmospheric behaviors. As an initial simulation using the coupled model, we have carried out a 30 day consecutive run in September. The result reveals that the longitudinal structure of the F‐region ionosphere varies on a day‐to‐day basis in a highly complex way and that a four‐peak structure of the daytime equatorial ionization anomaly (EIA) similar to the recent observations appears as an averaged feature. The simulation reproduces and thus confirms the vertical coupling processes proposed so far with respect to the formation of the averaged EIA longitudinal structure; the excitation of solar nonmigrating tides in the troposphere, their propagation through the middle atmosphere, and the modulation of ionospheric dynamo, which in turn affects EIA generation. The simulation result indicates that not only the ionospheric averaged longitudinal structure but also the day‐to‐day variation can be modulated significantly by the lower atmospheric effect.

The reaction characteristics, including hydration reaction rate and performance over repeated reaction cycles, of CaO extracted from different local limestone samples in Japan were examined to identify the optimal material to be used in chemical heat storage and pumps. The factors influencing the chemical reaction, including the chemical composition, crystal grain structure, collapse due to hydration/dehydration reactions of CaO/Ca(OH) 2 , specific surface area, and mean pore diameter, were compared and discussed. Based on the results, it was observed that CaO derived from limestone from Hiroshima, previously utilized by our research group, and Kawara CaO with a high purity exhibit similar reaction characteristics. Garou CaO, which has more impurities than that in the others tested, exhibits a lower initial hydration reaction rate and conversion change than those of Hiroshima/Okayama and Kawara CaO due to the presence of magnesium in Garou CaO as dolomite, which initially fills the pores. However, it was shown that Garou CaO can be used as a practical material for chemical heat storage and pumps with relatively high strength and reactivity compared with the other limestone samples because of its columnar joint texture and larger pore size following activation by four water vapor reaction cycles.

A linkage between climate change in the Atlantic and the Pacific oceans during the 1990s is investigated using three versions of the coupled climate model MIROC and CMIP5 multi‐model ensemble. From the early 1990s to the early 2000s, the observed sea surface temperature (SST) shows warming in the North Atlantic and a La Niña‐like pattern in the Pacific. Associated with the SST pattern, the observations indicate a strengthened Walker circulation in the tropical Pacific and enhanced precipitation in the tropical Atlantic. These SST and precipitation patterns are simulated well by hindcast experiments with external forcing and an initialized ocean anomaly state but are poorly simulated by uninitialized simulation with external forcing only. In particular, the observed La Niña‐like SST pattern becomes prominent in ensemble members with large amplitudes of Atlantic Multidecadal Oscillation (AMO) index during 1996–1998. Our results suggest that ocean initialization in both the Pacific and the Atlantic plays an important role in predicting the Pacific stepwise climate change during the 1990s, which contributes to the accurate estimation of global temperature change in the coming decade. Forecasting typhoon frequency or marine fisheries production in the coming decade may be possible by improving the predictive skill of stepwise climate change. Key Points Pacific decadal climate change is predictable with a help of North Atlantic Decadal ENSO variability is predictable beyond 1 year Initialization contributes to the accurate temperature change

It is well established that the ocean plays an important role in absorbing anthropogenic carbon C ant from the atmosphere. Under global warming, Earth system model simulations and theoretical arguments indicate that the capacity of the ocean to absorb C ant will be reduced, with this constituting a positive carbon–climate feedback. Here we apply a suite of sensitivity simulations with a comprehensive Earth system model to demonstrate that the surface waters of the shallow overturning structures (spanning 45°S–45°N) sustain nearly half of the global ocean carbon–climate feedback. The main results reveal a feedback that is initially triggered by warming but that amplifies over time as C ant invasion enhances the sensitivity of surface pCO₂ to further warming, particularly in the warmer season. Importantly, this “heat–carbon feedback” mechanism is distinct from (and significantly weaker than) what one would expect from temperature-controlled solubility perturbations to pCO₂ alone. It finds independent confirmation in an additional perturbation experiment with the same Earth system model. There mechanism denial is applied by disallowing the secular trend in the physical state of the ocean under climate change, while simultaneously allowing the effects of heating to impact sea surface pCO₂ and thereby CO₂ uptake. Reemergence of C ant along the equator within the shallow overturning circulation plays an important role in the heat–carbon feedback, with the decadal renewal time scale for thermocline waters modulating the feedback response. The results here for 45°S–45°N stand in contrast to what is found in the high latitudes, where a clear signature of a broader range of driving mechanisms is present.