An official website of the United States government
Surface acoustic wave devices have many applications in signal processing, radio frequency (RF) communications, and sensing [1]. The most common utilization of these devices is for filtering electromagnetic signals in communications systems. However, since the physical dimensions of the inter-digitated transducers (IDT) determine the frequency response, it is very difficult to attain tunable devices for programmable applications. Great effort has been made to achieve an integrated solution to this in III-V semiconductors. One such work utilizes the piezoelectric the GaN buffer layer in an AlGaN/GaN epi for acoustic propagation, while a metal-insulator-semiconductor (MIS) structure is used to tune the SAW response [2]. Unfortunately, the MIS structure results in a weak interaction only achieving a phase tunability of 0.07%. Recent work, uses thin film Zinc Oxide (ZnO) as a piezoelectric on top of n-type ZnO on GaN achieving a high tunability of. 9% [3]. In this work, we demonstrate a ZnO on AIGaN/GaN heterostructure capable of achieving high tunability as well as impacting properties of the SAW filter not previously reported.
more »
« less
Quantum and classical machine learning investigation of synthesis–structure relationships in epitaxially grown wide band gap semiconductors
Several hundred plasma-assisted molecular beam epitaxy synthesis experiments of GaN and ZnO thin film crystals were organized into data sets that correlate the operating parameters selected for growth to two figures of merit: a binary determination of surface morphology, and a continuous Bragg–Williams measure of lattice ordering (S2). Quantum as well as conventional supervised machine learning algorithms were optimized and trained on the data, enabling a comparison of their generalization performance. The models displaying the best generalization performance on each data set were subsequently used to predict each figure of merit across the ZnO and GaN processing spaces.
more »
« less
- Award ID(s):
- 2003581
- PAR ID:
- 10549056
- Publisher / Repository:
- Springer Link
- Date Published:
- Journal Name:
- MRS Communications
- Volume:
- 14
- Issue:
- 4
- ISSN:
- 2159-6867
- Page Range / eLocation ID:
- 660 to 666
- Subject(s) / Keyword(s):
- machine learning quantum machine learning crystal growth molecular beam epitaxy
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The size, weight, power density, cost, and efficiency are crucial factors that should be considered when designing or employing power electronics converters for a specific application. Therefore, comparing different converters to investigate which converter provides better figures of merit at the same application and operating condition is essential. This paper uses theoretical, simulation and experimental comparisons between the two-level and multilevel converters. The DC-DC two-level buck and the flying capacitor multilevel (FCML) buck converters are chosen to carry out the theoretical, simulation and experimental prototypes when both employ Gallium Nitride (GaN) power semiconductor switches. It was found that the FCML converter inherently provides superior performance and figures of merit over both the two-level and multilevel converters. Simulation and experimental results that validate each other are provided in this paper.more » « less
-
The outstanding properties of Gallium Nitride (GaN) have enabled considerable improvements in the performance of power devices compared to traditional silicon technology, resulting in more efficient and highly compact power converters. GaN power technology has rapidly developed and is expected to gain a significant market share in an increasing number of applications in the coming years. However, despite the great progress, the performance of current GaN devices is still far from what the GaN material could potentially offer, and a significant reduction of the device on-resistance for a certain blocking voltage is needed. Conventional GaN high-electron-mobility-transistors are based on a single two-dimensional electron gas (2DEG) channel, whose trade-off between electron mobility and carrier density limits the minimum achievable sheet resistance. To overcome such limitations, GaN power devices including multiple, vertically stacked 2DEG channels have recently been proposed, showing much-reduced resistances and excellent voltage blocking capabilities for a wide range of voltage classes from 1 to 10 kV. Such devices resulted in unprecedented high-power figures of merit and exceeded the SiC material limit, unveiling the full potential of lateral GaN power devices. This Letter reviews the recent progress of GaN multi-channel power devices and explores the promising perspective of the multi-channel platform for future power devices.more » « less
-
The proliferation of Artificial Intelligence (AI) models such as Generative Adversarial Networks (GANs) has shown impressive success in image synthesis. Artificial GAN-based synthesized images have been widely spread over the Internet with the advancement in generating naturalistic and photo-realistic images. This might have the ability to improve content and media; however, it also constitutes a threat with regard to legitimacy, authenticity, and security. Moreover, implementing an automated system that is able to detect and recognize GAN-generated images is significant for image synthesis models as an evaluation tool, regardless of the input modality. To this end, we propose a framework for reliably detecting AI-generated images from real ones through Convolutional Neural Networks (CNNs). First, GAN-generated images were collected based on different tasks and different architectures to help with the generalization. Then, transfer learning was applied. Finally, several Class Activation Maps (CAM) were integrated to determine the discriminative regions that guided the classification model in its decision. Our approach achieved 100% on our dataset, i.e., Real or Synthetic Images (RSI), and a superior performance on other datasets and configurations in terms of its accuracy. Hence, it can be used as an evaluation tool in image generation. Our best detector was a pre-trained EfficientNetB4 fine-tuned on our dataset with a batch size of 64 and an initial learning rate of 0.001 for 20 epochs. Adam was used as an optimizer, and learning rate reduction along with data augmentation were incorporated.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1557/s43579-024-00590-z
