Search for: All records

Abstract Two key challenges in the development of data‐driven gravity‐wave parameterizations are generalization, how to ensure that a data‐driven scheme trained on the present‐day climate will continue to work in a new climate regime, and calibration, how to account for biases in the “host” climate model. Both problems depend fundamentally on the response to out‐of‐sample inputs compared with the training dataset, and are often conflicting. The ability to generalize to new climate regimes often goes hand in hand with sensitivity to model biases. To probe these challenges, we employ a one‐dimensional (1D) quasibiennial oscillation (QBO) model with a stochastic source term that represents convectively generated gravity waves in the Tropics with randomly varying strengths and spectra. We employ an array of machine‐learning models consisting of a fully connected feed‐forward neural network, a dilated convolutional neural network, an encoder–decoder, a boosted forest, and a support‐vector regression model. Our results demonstrate that data‐driven schemes trained on “observations” can be critically sensitive to model biases in the wave sources. While able to emulate accurately the stochastic source term on which they were trained, all of our schemes fail to simulate fully the expected QBO period or amplitude, even with the slightest perturbation to the wave sources. The main takeaway is that some measures will always be required to ensure the proper response to climate change and to account for model biases. We examine one approach based on the ideas of optimal transport, where the wave sources in the model are first remapped to the observed one before applying the data‐driven scheme. This approach is agnostic to the data‐driven method and guarantees that the model adheres to the observational constraints, making sure the model yields the right results for the right reasons. 

Abstract This paper presents a system-level efficiency analysis, a rapid design methodology, and a numerical demonstration of efficient sub-micron, spin-wave transducers in a microwave system. Applications such as Boolean spintronics, analog spin-wave-computing, and magnetic microwave circuits are expected to benefit from this analysis and design approach. These applications have the potential to provide a low-power, magnetic paradigm alternative to modern electronic systems, but they have been stymied by a limited understanding of the microwave, system-level design for spin-wave circuits. This paper proposes an end-to-end microwave/spin-wave system model that permits the use of classical microwave network analysis and matching theory towards analyzing and designing efficient transduction systems. This paper further compares magnetostatic-wave transducer theory to electromagnetic simulations and finds close agreement, indicating that the theory, despite simplifying assumptions, is useful for rapid yet accurate transducer design. It further suggests that the theory, when modified to include the exchange interaction, will also be useful to rapidly and accurately design transducers launching magnons at exchange wavelengths. Comparisons are made between microstrip and co-planar waveguide lines, which are expedient, narrowband, and low-efficiency transducers, and grating and meander lines that are capable of high-efficiency and wideband performance. The paper concludes that efficient microwave-to-spin-wave transducers are possible and presents a meander transducer design on YIG capable of launching $$\varvec{\lambda = 500}\,$$ λ = 500 nm spin waves with an efficiency of − 4.45 dB and a 3 dB-bandwidth of 134 MHz.