Quantum on silicon; energy-efficient AI chip; growing thin film materials.
Researchers at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS) developed a platform to probe and control qubits in silicon for quantum networks, after an earlier discovery that defects in silicon could be used to send and store quantum information over widely used telecommunications wavelengths.
The device uses an electric diode to manipulate qubits inside a commercial silicon wafer. Researchers can then explore how the defect responds to changes in the electric field, tune its wavelength within the telecommunications band, and turn it on and off. The team also developed a diagnostic tool to image how the defects embedded in the device change in space as the electric field is applied.
“We found that the way we’re modifying the electric environment for the defects has a spatial profile, and we can image it directly by seeing the changes in the intensity of light being emitted by the defects,” said Aaron Day, a Ph.D. candidate at SEAS who co-led the work with Madison Sutula, a research fellow at Harvard, in a release. “By using so many emitters and getting statistics on their performance, we now have a good understanding of how defects respond to changes in their environment. We can use that information to inform how to build the best environments for these defects in future devices.” [1]
Researchers from Oregon State University, the University of Michigan, University of Oklahoma, Cornell University, and Pennsylvania State University developed an energy efficient AI chip, which is a tunable and stable memristor based on entropy-stabilized oxides (ESOs). Their work was funded by the National Science Foundation (NSF).
Memristors are similar to biological neural networks because there is no external memory source, so no energy is lost moving data back and forth. ESOs comprise over six elements – single-crystalline (Mg,Co,Ni,Cu,Zn)O films grown on an epitaxial bottom electrode – which allows their memory capabilities to be finely tuned, and means they can handle both computation and data storage tasks. By optimizing the ESO composition for AI jobs, ESO-based chips can perform tasks with far less energy than a CPU.
Another benefit is that these artificial neural networks can process time-dependent information, such as data for audio and video. The device can work on a varied time scale because of the ability to tune the ESOs’ composition. [2]
Researchers from USC Viterbi introduced a technique called hybrid pulsed laser deposition to grow thin film chalcogenides, which have the electronic and photonic properties needed for high-performance devices.
Growing the thin films without defects is challenging, especially if the chalcogenides contain sulfur because its precursor materials are toxic, reactive, and difficult to control. The new technique harnesses a non-corrosive organosulfur compound as the sulfur source for thin film growth.
In a release, Mythili Surendran, a Ph.D. researcher in the Mork Family Department of Chemical Engineering and Materials Science, explained why they are switching to an organic precursor which has sulfur in it. “It’s efficient, as it has optimal vapor pressure and also, people don’t have to worry about using hydrogen sulfide in their chambers. You can also control it precisely and decompose it at much lower temperatures. This is an important need for the semiconductor industry as we have to grow these materials at low temperatures to integrate these devices directly on our current silicon-based electronic chips without degrading those chips.”
Surendran said it is a much more efficient process than using hydrogen sulfide. The team is now investigating how hybrid pulsed laser deposition can be improved, scaled up, and applied to grow thin films of different material classes to a similarly high standard. [3]
[1] Day, A.M., Sutula, M., Dietz, J.R. et al. Electrical manipulation of telecom color centers in silicon. Nat Commun 15, 4722 (2024). https://doi.org/10.1038/s41467-024-48968-w
[2] Yoo, S., Chae, S., Chiang, T. et al. Efficient data processing using tunable entropy-stabilized oxide memristors. Nat Electron (2024). https://doi.org/10.1038/s41928-024-01169-1
[3] Surendran, Mythili, Shantanu Singh, Huandong Chen, Claire Wu, Amir Avishai, Yu‐Tsun Shao, and Jayakanth Ravichandran. “A hybrid pulsed laser deposition approach to grow thin films of chalcogenides.” Advanced Materials (2024): 2312620. https://doi.org/10.1002/adma.202312620
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Leave a Reply