Lightnews — Scholar-powered news

Patsy DeLacey @patsyd.bsky.social · 2d

🧪 Like atomic-scale spray painting, plasma-enhanced atomic layer deposition can now be used for ScAlN.

The technique expands use of the ultra-wide band gap material to complex, 3D structures within next-gen electronic devices.

news.engin.umich.edu/2025/10/new-...

New process layers uniform ScAlN on 3D surfaces

Scandium aluminum nitride can now be integrated into high-voltage, high-frequency or piezoelectric devices with plasma-enhanced atomic layer deposition.

news.engin.umich.edu

1 2

Patsy DeLacey @patsyd.bsky.social · 4d

🧪 Using nearly a decade of high-speed footage, an unsupervised machine learning approach distinguished 9 distinct precipitation types, including varieties of rain, snow, and mixed-phases.

Results could improve weather forecasting and winter driving safety.

news.engin.umich.edu/2025/10/betw...

Between rain and snow, machine learning finds 9 precipitation types

Leveraging 1.5M minutes of precipitation data and a nonlinear method to handle complex relationships between variables, the team created a new classification system

news.engin.umich.edu

2 2 6

Patsy DeLacey @patsyd.bsky.social · May 7

Halting slow faults 🧪

When computing systems slow rather than fail outright, these slow faults can be tricky to detect.

A new solution adjusts in real time, more efficiently catching and overcoming slow faults.

cse.engin.umich.edu/stories/a-ne...

A new tool to manage slow faults

A new library dynamically adjusts to runtime operating slowdowns, reducing performance degradation under varying slow-fault conditions by 65%.

cse.engin.umich.edu

3

Patsy DeLacey @patsyd.bsky.social · Apr 29

Boundary setting is healthy for AI models too 🧪

Check out how a new communication-collective system speeds up AI and machine learning training across multiple cloud servers 👇

Presented today at NSDI ‘25

cse.engin.umich.edu/stories/perf...

Perfect is the enemy of good for distributed deep learning in the cloud

Leveraging deep learning’s resilience, approximating data lost by allowing some servers to time out speeds up model training while preserving performance.

cse.engin.umich.edu

2 2 6

Patsy DeLacey @patsyd.bsky.social · Apr 28

A new, straightforward way to make a microgel able to release multiple drugs at different times 🧪

che.engin.umich.edu/2025/04/25/a...

A simpler way to make microgels for programmable drug release - Chemical Engineering

Dunking a calcium alginate microgel in ion baths helps finely tune drug release profiles without complex equipment, opening up microgel synthesis to more researchers.

che.engin.umich.edu

1 10

Patsy DeLacey @patsyd.bsky.social · Apr 14

AI is WEIRD, but it doesn’t have to be 🧪

cse.engin.umich.edu/stories/ai-w...

AI with, for and by everyone can help maximize its benefits

Involving a wider swatch of people in AI development can help maximize its benefits.

cse.engin.umich.edu

1 4

Patsy DeLacey @patsyd.bsky.social · Apr 8

Directing heavy traffic in cloud data centers🚦 🧪

Gigaflow takes a new approach to caching incoming data flows, helping speed up heavy traffic caused by AI and machine learning workloads

Full story below 👇
cse.engin.umich.edu/stories/stre...

Streamlining cloud traffic with a Gigaflow Cache

Gigaflow improves virtual switches for programmable SmartNICs, delivering a 51% higher hit rate and 90% lower misses.

cse.engin.umich.edu

2 3

Patsy DeLacey @patsyd.bsky.social · Mar 21

Capturing carbon and value from biodiesel wastewater 🧪

Researchers test catalyst stability in an electrochemical method that leverages glycerol’s low redox potential to utilize CO2 while producing valuable chemicals like formate.

cee.engin.umich.edu/2025/03/21/b...

Biodiesel wastewater treatment: Capturing carbon and valuable chemicals - Civil and Environmental Engineering

A new experimental design tests catalyst stability in an electrochemical method for treating environmental waste streams while capturing valuable co-products

cee.engin.umich.edu

3 6

Patsy DeLacey @patsyd.bsky.social · Mar 19

Cheaper wind turbine reliability testing 🧪

A new simulation makes wind turbine reliability testing more affordable and accessible without sacrificing accuracy, helping prevent system failures like dropping a blade.

ioe.engin.umich.edu/2025/03/18/m...

More efficient wind turbine reliability simulation - Industrial & Operations Engineering

A University of Michigan-led team developed a method to make virtual testing of wind turbine components more affordable and reliable.

ioe.engin.umich.edu

1 12

Patsy DeLacey @patsyd.bsky.social · Mar 13

AI for social scientists 🧪

Natural language processing (NLP) can help social scientists keep up with the fast pace of digital communication, extracting the human behaviors that underlie social media posts, text messages or zoom transcripts.

cse.engin.umich.edu/stories/how-...

How AI tools can help us understand human behavior

Natural language processing tools can help social scientists process larger and longer-term text datasets to keep up with fast-paced digital communication.

cse.engin.umich.edu

2 5

Patsy DeLacey @patsyd.bsky.social · Mar 12

A time-based fingerprint exposes an internet censorship evasion tactic 🧪

cse.engin.umich.edu/stories/coun...

Countering a flaw in anti-censorship tools to improve global internet freedom

A time-based fingerprint can expose obfuscation proxies, prompting the need for new techniques to protect information access

cse.engin.umich.edu

2

Patsy DeLacey @patsyd.bsky.social · Mar 12

AC units balance the grid 🧪

Residential AC units in Austin, Texas balanced electricity supply and demand without sacrificing comfort.

Using small distributed energy sources rather than power plant output to grid balance can improve stability and help integrate wind and solar.

myumi.ch/E8xNJ

Home AC units can balance the electrical grid without impacting comfort

A real-world frequency regulation experiment demonstrates that managing AC units can help actively balance electricity supply and demand.

myumi.ch

7

Patsy DeLacey @patsyd.bsky.social · Feb 19

Sponge-like implant helps develop a treatment for a MS-like condition in mice 🧪

The implant captures immune cells, providing a close look at cell communication outside the CNS.

Results helped develop a targeted nanoparticle treatment that slowed or stopped disease progression.

myumi.ch/nyMn5

Multiple sclerosis: Cell-catching implant helps identify successful treatment in mice

The implant attracts diseased immune cells, enabling researchers to study and target them with a nanoparticle-based treatment.

myumi.ch

1 10

Patsy DeLacey @patsyd.bsky.social · Feb 3

Sensors with a sense of time 🧪

A system of time-aware sensors can map out chemical concentrations within long, narrow tubes according to simulations.

Memristors act like a ticking clock until they collide with the target chemical, switching off the timer.

che.engin.umich.edu/2025/02/03/t...

Thousands of tiny, time-aware sensors collectively map chemical concentrations within narrow tubes - Chemical Engineering

A new framework optimizes a system of sensors that remember when they collided with a target chemical while flowing through hard-to-reach areas

che.engin.umich.edu

1 5

Reposted by Patsy DeLacey

Science X / Phys.org @sciencex.bsky.social · Jan 29

An interactive tool evaluates the most efficient carbon-neutral aircraft for various flight scenarios, comparing battery-electric, hydrogen, and synthetic fuel options based on energy consumption. doi.org/g83j3x

Interactive tool helps find most efficient carbon-neutral aircraft for your flight

OK, we admit, we're a long way from a carbon-free grid—but when we have one, what's the most efficient way to use that energy to fly planes?

techxplore.com

1 3

Patsy DeLacey @patsyd.bsky.social · Jan 29

Given a carbon-free grid, what’s the most efficient air travel? 🧪

An interactive tool finds the best option given the distance, speed and payload.

Battery powered aircrafts are best for short flights, while hydrogen options win out for longer flights.

news.engin.umich.edu/2025/01/find...

Finding the most efficient carbon-neutral aircraft for your flight

Sustainable aviation doesn't have a one-size-fits-all solution, but an interactive tool can identify the right fit given the flight distance, speed and payload.

news.engin.umich.edu

1 2 3

Reposted by Patsy DeLacey

Derek Smith @djsmitty156.bsky.social · Jan 21

PRESS RELEASE: New water purification technology helps turn seawater into drinking water without tons of chemicals and extra reverse osmosis steps. 💧🧪news.engin.umich.edu/2025/01/new-...

New water purification technology helps turn seawater into drinking water without tons of chemicals

Cutting acid and base treatments from conventional desalination plants could save billions of dollars globally, making seawater a more affordable option for drinking water.

news.engin.umich.edu

5 9

Patsy DeLacey @patsyd.bsky.social · Jan 22

Transparent AI for nuclear engineers 🧪 ☢️

Nuclear regulators need a way to validate AI models. Starting to crack open the AI ‘black box’, a new open source python package includes explainability features and helps engineers create AI tools from their data.

ners.engin.umich.edu/2025/01/22/s...

Streamlining AI development for transparent nuclear engineering models - Nuclear Engineering & Radiological Sciences

A new tool simplifies machine learning development for nuclear engineers while incorporating model explainability features to help improve transparency.

ners.engin.umich.edu

4 22

Patsy DeLacey @patsyd.bsky.social · Jan 10

Greener cement: An electrochemical method produces calcium carbonate, the primary cement feedstock, while capturing CO2 from the air. 🧪

If used at full capacity, the process could reduce cement industry global CO2 emissions from 8% to 3%.

cee.engin.umich.edu/2025/01/10/s...

Sustainable cement: An electrochemical process to help neutralize cement industry CO2 emissions - Civil and Environmental Engineering

The reaction produces carbon-negative calcium carbonate to replace limestone as the primary feedstock and green hydrogen as the fuel for cement kilns

cee.engin.umich.edu

2 9 36

Patsy DeLacey @patsyd.bsky.social · Jan 10

Machine learning advances the search for methane storage solutions, promoting use as a cleaner vehicle fuel 🧪

A new tool evaluated 84,000 covalent organic frameworks (COFs), a porous material that condenses methane to its surface.

me.engin.umich.edu/news-events/...

A tool to optimize methane storage solutions – Mechanical Engineering

me.engin.umich.edu

1 2 6

Patsy DeLacey @patsyd.bsky.social · Dec 19

Two aligned solar probes helped researchers track changes in the Sun’s magnetic field. 🧪

As the magnetic field loses steam, it accelerates the surrounding solar wind. Results can help improve space weather predictions and potential impacts on Earth.

clasp.engin.umich.edu/2024/12/19/w...

When Solar Probes Align: Data Confirms the Sun’s Magnetic Field Accelerates Solar Wind - Climate and Space Sciences and Engineering

A new study led by University of Michigan researchers published in The Astrophysical Journal confirms the sun's magnetic field accelerates solar wind as it relaxes.

clasp.engin.umich.edu

1 10

Patsy DeLacey @patsyd.bsky.social · Dec 17

A new self-supervised learning model learns both object identity and pose using unlabeled image triplets, mimicking robotic vision.
🧪

This paves the way for more intuitive interactions between tech like robots and self-driving cars and the physical world.

cse.engin.umich.edu/stories/help...

Helping machine learning models identify objects in any pose

A new self-supervised learning benchmark and dataset improve pose estimation, paving the way for more intuitive interactions between technology and the physical world.

cse.engin.umich.edu

2 17

Patsy DeLacey @patsyd.bsky.social · Dec 11

A fair concern! Visual artifacts are the hallucinations that need to be tamped down.

To suppress visual artifacts, the researchers used conjugate gradient methods and measured how well generated images matched measurements. Keep in mind, it’s still early days with this tech.

1 1

Patsy DeLacey @patsyd.bsky.social · Dec 11

KLA contributed funding in addition to funding from U-M.

Patsy DeLacey @patsyd.bsky.social · Dec 11

A new diffusion model reconstructed high-quality, sparse-view 3D CT images, moving towards use in low radiation dose medical imaging. 🧪

The paper will be presented this afternoon at NeurIPS in Vancouver, BC.

ece.engin.umich.edu/stories/new-...

New generative AI model can reconstruct a high-quality, sparse-view 3D CT scan with a much lower X-ray dose

Improvements move the deep learning technique towards use in 3D medical imaging.

ece.engin.umich.edu

2 1 14