For decades, the discussion around autism has been a hotbed of misinformation, misinterpretation, and bad science, ranging from the long-discredited link between the neurodevelopmental condition and vaccines, to newer claims that going gluten-free and avoiding ultra-processed foods can reverse autistic traits.

On Monday night, this specter arose again in the Oval Office, as President Donald Trump announced his administration’s new push to study the causes of autism with claims that the common painkiller Tylenol, otherwise known as acetaminophen, can cause the condition. The FDA subsequently announced that the drug would be slapped with a warning label citing a “possible association.”

David Amaral, professor and director of research at the UC Davis MIND Institute, was among those watching in dismay as the president launched into a diatribe about Tylenol, repeatedly warning pregnant women not to take it, even to treat fevers.

“We heard the president say that women should tough it out,” says Amaral. “I was really taken aback by that, because we do know that prolonged fever, in particular, is a risk factor for autism. So I worry that this admonition to not take Tylenol is going to do the reverse of what they’re hoping.”

The speculation surrounding Tylenol stems from correlations drawn by some studies that have touted an association between use of the painkiller and neurodevelopmental disorders. One such analysis was published last month. The problem, says Renee Gardner, an epidemiologist at the Karolinska Institute in Sweden, is that these studies often reach this conclusion because they don’t sufficiently account for what statisticians describe as “confounding factors”—additional variables related to those being studied that might influence the relationship between them.

In particular, Gardner points out that pregnant women needing to take Tylenol are more likely to have pain, fevers, and prenatal infections, which are themselves risk factors for autism. More importantly, given the heritability of autism, many of the genetic variants that make women more likely to have impaired immunity and greater pain perception, and hence use painkillers like acetaminophen, are also linked to autism. The painkiller use, she says, is a red herring.

Last year, Gardner and other scientists published what is widely regarded within the scientific field as the most conclusive investigation so far on the subject, one which did account for confounding factors. Using health records from nearly 2.5 million children in Sweden, they reached the opposite conclusion to the president: Tylenol has no link to autism. Another major study of more than 200,000 children in Japan, published earlier this month, also found no link.

Doctors are worried that Trump’s claims will have adverse consequences. Michael Absoud, a paediatric neurodisability consultant and a researcher in pediatric neurosciences at King’s College London, says he fears that pregnant women will start using other painkillers with a less well-proven safety profile.

Gardner is concerned that it will also lead to self-blaming among parents, a flashback to the 1950s and 60s, a time when autism was wrongly attributed to emotionally cold “refrigerator mothers.” “It’s making parents of children with neurodevelopmental conditions feel responsible,” she says. “It harks back to the early dark days of psychiatry.”

A new physics-based algorithm clears a path toward nuclear microreactors that can autonomously adjust power output based on need, according to a University of Michigan-led study published in Progress in Nuclear Energy.

Easily transportable and able to generate up to 20 megawatts of thermal energy for heat or electricity, nuclear microreactors could be useful in remote locations such as rural communities, disaster zones, military bases or even cargo ships, in addition to other applications.

If integrated into an electrical grid, nuclear microreactors could provide stable, carbon-free energy, but they must be able to adjust power output to match shifting demand—a capability known as load following. In large reactors, staff make these adjustments manually, which would be cost-prohibitive in remote areas, imposing a barrier to adoption.

“Many startup and legacy companies in the U.S. are pushing toward near-term and broad deployment of nuclear microreactors, and our work establishes a clear avenue to achieve that in an economically viable way,” said Brendan Kochunas, an associate professor of nuclear engineering and radiological sciences at U-M and corresponding author of the study.

“Our method can help vendors design reactors with autonomous control systems that are safer and more secure.”

This study focused on High-Temperature Gas-Cooled Reactors (HTGR), advanced nuclear reactors that can scale from micro- to large-scale. While based on the Holos-Quad (Gen 2+) model, a HTGR-type microreactor design, the researchers outline a simplified microreactor model that preserves key parameters like power density, inlet coolant temperature, core pressure and flow velocity.

The research team leveraged model predictive control (MPC), a method that predicts future behavior, to optimize control over a defined period of time under certain constraints. Specifically, they developed an MPC controller that optimized the rotation of control drums that surround the microreactor’s central core that decreases power when facing inwards and increases power when facing outwards.

To ensure the model was based in reality and accurately represented the microreactor’s operation, the researchers integrated PROTEUS, a simulation toolset for high-fidelity reactor physics analysis.

When tasked with ramping the power up or down at 20% per minute, their control algorithm stayed within 0.234% of the target. It does all of this without AI, meaning everything about the automated control for load follow operation is grounded in physics and mathematics and readily explainable—an essential feature for passing regulatory review.

Extensive sensitivity tests confirmed their MPC controller works for a wide range of model inputs, validating feasibility for autonomous control.

“The control algorithm’s success and integration with high-fidelity simulation tools demonstrates that we can now design nuclear reactors and their instrumentation and control systems together from the ground up, rather than trying to backfit the I&C (instrumentation and control) systems to a mostly complete reactor design,” said Kochunas.

With the ability to print metal structures with complex shapes and unique mechanical properties, metal additive manufacturing (AM) could be revolutionary. However, without a better understanding of how metal AM structures behave as they are 3D printed, the technology remains too unreliable for widespread adoption in manufacturing and part quality remains a challenge.

Researchers in Lawrence Livermore National Laboratory (LLNL)’s nondestructive evaluation (NDE) group are tackling this challenge by developing first-of-their-kind approaches to look at how materials and structures evolve inside a metal AM structure during printing. These NDE techniques can become enabling technologies for metal AM, giving manufacturers the data they need to develop better simulations, processing parameters and predictive controls to ensure part quality and consistency.

“If you want people to use metal AM components out in the world, you need NDE,” said David Stobbe, group leader for NDE ultrasonics and sensors in the Materials Engineering Division (MED). “If we can prove that AM-produced parts behave as designed, it will allow them to proliferate, be used in safety-critical components in aerospace, energy and other sectors and hopefully open a new paradigm in manufacturing.”

Measuring in the middle

NDE techniques involve sending signals like X-rays, ultrasound or electrical currents through objects and observing signal changes to infer information or reconstruct an image of what’s inside. NDE is important for quality control in all manufactured parts, but for metal AM, it can also help catch printing problems before it’s too late.

Most metal AM techniques use heat to bind material together, and since metals are extremely sensitive to heat, structures can change a lot during printing. Heat diffuses from the print surface into the already-printed structure, which can affect how well the material binds, create failure-inducing defects and lead to inconsistent products.

“Evolving processes in the subsurface need to be measured and characterized if you want to have a consistent print quality,” said Saptarshi Mukherjee, a research scientist in the Lab’s Atmospheric, Earth and Energy Division (AEED). “This is very challenging because most of the current NDE technologies cannot see through heat, and even infrared cameras and antennas only detect heat at the surfaces.”

Mukherjee is part of a project to monitor internal temperature during laser powder-bed fusion (LPBF) metal AM using eddy currents, swirling loops of electrical current induced by applying magnetic fields. Eddy currents are sensitive to electrical conductivity, and since conductivity is a function of temperature, eddy current sensors provide real-time localized temperature information from inside structures.

Simulations from collaborators at Michigan State University suggested the approach was viable, and the group validated it with a simple experiment, resulting in a recent paper published in Scientific Reports.

“To our knowledge, this is the first time that eddy current sensors have been used to look at these very rapid non-equilibrium thermal processes, which are suggestive of the sort of thermal processes you would see in a metal AM process,” said MED postdoc Ethan Rosenberg.

Rosenberg is now leading the experimental testing for a follow-up study using closer to real-world conditions such as non-uniform heating and faster timescales.

Trailblazers

NDE group leader Joe Tringe launched the first Laboratory-Directed Research and Development (LDRD) project in the area in 2018 and ever since, the group has been treading new ground to keep pace with metal AM.

In their first project, the group showed that millimeter wave signatures could efficiently characterize the shape of individual droplets of liquid metal used to create structures in liquid metal jetting. They eventually collected enough data to train a machine learning algorithm to predict droplet shape.

“If we can combine that feedback with system modeling, we may be able to learn whether the print parameters are working or if they need to be changed, in real time, so that we end up with what we want when we’re done,” said Stobbe.

Follow-up projects expanded to electrical resistance tomography—which measures changes in a current’s voltage and electrical potential—X-ray computed tomography, ultrasound and neutron detection, with an emphasis on lattice structures and other complicated geometries.

The group also uses NDE to inspect processing parameters like sonication—using ultrasonic waves to create vibrations and improve homogenization—in laser-based metal AM.

In a recent study published in Communications Materials, the group and collaborators at Pennsylvania State University and Argonne National Laboratory proved they could use high-speed synchrotron X-ray imaging for these measurements. The technique is the first step toward understanding sonication’s impact on printing, which will help manufacturers optimize the process to improve part quality.

“A lot of things happen in these AM processes that affect the part, but without using NDE techniques, it’s kind of a black box,” said Rosenberg. “With ingenuity and good physical understanding, you can open that box to see what’s happening inside, and that will hopefully help you control the process.”

Enabling the future

The group plans to continue evolving, improving and generalizing a variety of NDE techniques for metal AM, since different techniques are better at measuring different types of information. They also hope to train machine learning algorithms for real-time monitoring and error correction during the print to improve success.

The information they collect along the way will be crucial to enabling widespread adoption of metal AM, and they hope that their work will also help raise awareness of the opportunities for NDE in the emerging field.

“There’s a real gold rush aspect to it,” said Stobbe. “You’re out there doing or measuring things that you know no one has ever done or measured before because this is a new technology, and that’s certainly exciting.”

Other contributors to the work include MED’s Rosa Morales, Jordan Lum, Edward Benavidez and collaborators at the University of Colorado, Boulder.