Deck LATEST

Supercomputers challenge the origin story of cosmic explosions

Los Alamos simulations reveal that some of the universe's most powerful gamma-ray bursts may forge heavy elements without neutron-star collisions

For nearly a decade, astronomers believed they had solved one of the great mysteries of cosmic alchemy, attributing the production of the universe’s heaviest elements, such as gold and platinum, primarily to kilonovae from colliding neutron stars. This consensus was further solidified by the landmark 2017 detection of gravitational waves from such a merger.
 
However, new research from Los Alamos National Laboratory, published in The Astrophysical Journal Letters, challenges this narrative. A team led by Marko Ristić demonstrates through advanced supercomputing simulations that long-duration gamma-ray bursts, some of the most energetic explosions in existence, can produce kilonova-like signatures without requiring a neutron-star merger. Instead, the team proposes that collapsing massive stars, or "collapsars," can generate these characteristic optical and infrared emissions via a previously overlooked nucleosynthesis mechanism within their relativistic jets. This discovery is more than a mere astrophysical curiosity; it highlights how modern high-performance computing is fundamentally transforming our understanding of the cosmic origins of the elements.

Rewriting the story of gamma-ray bursts

Gamma-ray bursts (GRBs) are brief flashes of extraordinarily energetic radiation capable of releasing more energy in seconds than the Sun will emit during its entire lifetime. For decades, astronomers divided GRBs into two categories: short bursts produced by compact-object mergers and long bursts generated by collapsing massive stars. Observations of GRB 211211A and GRB 230307A complicated that picture. Although both events lasted roughly 40 seconds, far longer than typical merger-driven bursts, they exhibited infrared signatures resembling kilonovae, leading many researchers to conclude they originated from neutron-star mergers. The Los Alamos team questioned that assumption.
 
Their work proposes that the observed emission can be reproduced by a collapsar, a rapidly rotating massive star collapsing into a black hole while launching relativistic jets through its interior. Rather than producing heavy elements through tidal disruption of neutron stars, the model creates neutron-rich conditions within the jet and the surrounding cocoon.

The computational challenge

Testing that idea required a computational effort spanning multiple physics domains. Researchers combined nuclear reaction networks, magnetohydrodynamic simulations, Bayesian inference techniques, radiative transfer calculations, and large-scale parameter exploration. The project united scientists from Los Alamos' Theoretical Division, Computational Division, and Center for Theoretical Astrophysics.
 
At the heart of the study was the Portable Routines for Integrated nucleoSynthesis Modeling (PRISM) framework, which simulated the creation of heavy elements through rapid neutron-capture processes. The calculations explored how intense photon fields inside collapsar jets generate neutrons that subsequently seed nucleosynthesis in the surrounding cocoon. The researchers modeled both weak and full r-process scenarios and calculated the resulting radioactive heating over time. Those outputs became inputs for another computationally intensive stage: radiation transport simulations.

Monte Carlo radiation transport at scale

To predict what astronomers should observe, the team employed Los Alamos' SuperNu code, a sophisticated Monte Carlo radiation transport framework widely used in transient astrophysics. SuperNu follows millions of photon packets as they interact with expanding ejecta, accounting for absorption, emission, scattering, radioactive heating, and detailed atomic opacities. The simulations used high-fidelity opacity tables generated by Los Alamos atomic physics codes and modeled spectra across wavelengths ranging from ultraviolet through infrared. The computational workflow produced synthetic observations that could be directly compared with telescope data from GRB 211211A and GRB 230307A.
 
Rather than relying on simplified analytical approximations, the researchers simulated the detailed microphysics governing how light emerges from expanding stellar debris. The result was remarkable. A single weak-r-process ejecta component reproduced both the optical and infrared observations associated with the two gamma-ray bursts.

Machine learning accelerates discovery

The study also demonstrates how artificial intelligence and statistical computing are becoming essential tools for modern astrophysics. The team generated dozens of high-fidelity SuperNu simulations across a wide range of ejecta masses and velocities. Because running radiation transport calculations for every possible parameter combination would be prohibitively expensive, researchers trained Gaussian Process surrogate models to emulate the simulation outputs. These surrogate models enabled rapid Bayesian inference using the RIFT parameter-estimation framework, allowing the team to explore vast parameter spaces while preserving the accuracy of the underlying simulations.
 
This combination of supercomputing and machine-learning acceleration represents an increasingly common pattern across computational science, where advanced simulations generate data and AI-driven techniques help scientists navigate the resulting complexity.

Simulating a cosmic magnetic sieve

One of the study's most innovative computational components appears in its appendix. The researchers developed a three-fluid, two-dimensional magnetohydrodynamic simulation that tracks protons, neutrons, and alpha particles inside collapsar jets. The simulation investigated a phenomenon the team calls a magnetic "sieve." Strong magnetic fields trap charged particles near the jet core while allowing neutral neutrons to migrate into the surrounding cocoon. Under sufficiently intense magnetic fields, approaching 10¹² gauss, the resulting environment becomes neutron-rich enough to sustain rapid neutron-capture nucleosynthesis.
 
The simulation solved coupled continuity, momentum, energy, transport, and magnetic induction equations using an implicit-explicit numerical approach designed for stiff plasma systems. Without modern high-performance computing resources, such calculations would be effectively impossible.

A new view of cosmic element factories

Perhaps the most surprising conclusion is that a red kilonova does not necessarily imply the creation of the heaviest r-process elements. The team's simulations show that a weak r-process producing elements only up to approximately mass number 130 can generate the observed red evolution traditionally interpreted as evidence for lanthanide production. In other words, astronomers may need to reconsider some long-standing assumptions about how heavy elements are identified in explosive transients. If confirmed, the discovery could reshape our understanding of how the universe manufactures many of its elements.

Supercomputing as a cosmic lab

The broader significance of the work extends well beyond gamma-ray bursts. The study exemplifies how supercomputing has become a primary scientific instrument. Researchers combined nuclear physics, plasma physics, radiation transport, machine learning, Bayesian statistics, and astrophysical modeling into a single computational framework capable of testing ideas that cannot be reproduced in any terrestrial laboratory. The calculations relied on resources provided through Los Alamos National Laboratory's Institutional Computing Program, underscoring the increasingly central role of advanced computing infrastructure in modern astrophysical discovery.
 
A generation ago, astronomers could only observe the universe. Today, they can recreate it. And as this Los Alamos research demonstrates, the next breakthrough in understanding the origin of the elements may emerge not from a telescope alone, but from the convergence of supercomputing, artificial intelligence, and computational physics operating at unprecedented scale. The universe still holds many secrets. Increasingly, supercomputers are becoming the tools that allow us to uncover them.

Supercomputers trace a cosmic chain reaction from primordial black holes to the elements of life

O'Neal LATEST
SUNY Poly researchers combine hydrodynamics simulations, nuclear reaction networks, and galactic chemical evolution models to investigate whether primordial black holes helped shape the chemical history of the universe.
 
The most powerful scientific discoveries often begin with an improbable question: could the universe's most significant stellar explosions be triggered not by companion stars, but by ancient black holes born moments after the Big Bang?
 
Researchers at SUNY Polytechnic Institute are using advanced computational astrophysics to investigate this provocative possibility. Their latest study examines whether primordial black holes (PBHs), hypothetical relics from the dawn of time, might trigger Type Ia supernovae, thereby offering a new explanation for the diversity of observed stellar explosions and the complex chemical evolution of galaxies.
 
This work represents a remarkable convergence of supercomputing, cosmology, nuclear physics, and observational astronomy, tracing a chain of events that links the birth of the universe to the elemental composition of stars observed today.

From dark matter candidate to cosmic trigger

Primordial black holes occupy a unique place in modern astrophysics. Unlike black holes formed from collapsing stars, PBHs may have formed directly from density fluctuations shortly after the Big Bang. In particular, asteroid-mass primordial black holes remain viable dark matter candidates because they inhabit a region of parameter space that has proven difficult to constrain through conventional observations. The SUNY Poly team investigated what happens when one of these ancient objects encounters a white dwarf, a compact stellar remnant containing roughly the mass of the Sun, compressed into a volume similar to that of Earth.
 
Their simulations show that as a primordial black hole passes through a white dwarf, tidal forces and accretion heating can create localized hotspots. Under the extreme densities inside the star, those hotspots can ignite runaway carbon fusion, transforming a quiet white dwarf into a Type Ia supernova. Testing such a scenario requires computational capabilities far beyond traditional theoretical modeling.

Supercomputers recreate stellar catastrophes

To explore these events, the researchers employed multidimensional compressible hydrodynamics simulations capable of modeling thermonuclear explosions in extraordinary detail. The simulations tracked the evolution of turbulent burning fronts, detonation transitions, and shock propagation throughout exploding white dwarfs. The computational workflow did not end there. After the hydrodynamic calculations, the team used tracer-particle techniques to follow the thermodynamic histories of material inside the explosion. Those histories were then processed through a 495-isotope nuclear reaction network spanning elements from hydrogen to technetium, enabling researchers to calculate precisely which isotopes and elements were synthesized during the explosion.
 
Such calculations are among the most demanding workloads in computational astrophysics because they require coupling fluid dynamics, nuclear reactions, gravity, and thermodynamics across enormous ranges of scale. The resulting model suite produced explosions spanning a broad range of luminosities and nickel-56 yields, from approximately 0.2 to 1.1 solar masses of radioactive nickel, matching much of the diversity observed in real Type Ia supernovae.

Matching real supernovae

A scientific hypothesis becomes powerful when it confronts observations. The team compared its simulations with well-known supernova remnants, including Tycho, Kepler, and 3C 397, as well as nearby Type Ia supernovae, including SN 2011fe, SN 2012cg, SN 2013aa, and SN 2014J. By examining isotope ratios including manganese, nickel, and iron, researchers found that several observed supernovae could be explained by PBH-triggered explosion models.
 
Particularly striking was the ability of some PBH-triggered models to reproduce observed nickel and manganese abundances in remnants such as Kepler and 3C 397. Meanwhile, isotope ratios measured from late-time supernova light curves showed consistency with several modeled PBH-triggered explosions involving white dwarfs between roughly 1.0 and 1.1 solar masses. The implication is profound: some supernovae that astronomers have already observed may carry signatures of interactions with primordial black holes.

Simulating the chemical history of a galaxy

The study's most ambitious computational achievement came after the explosions themselves. The researchers incorporated their supernova yields into a Galactic Chemical Evolution model that tracks how generations of stars enrich a galaxy with heavy elements over billions of years. The simulations followed the production and distribution of silicon, sulfur, calcium, manganese, nickel, and other elements throughout cosmic history.
 
By comparing the results against stellar abundance measurements from large astronomical surveys, the team evaluated whether the universe's observed elemental composition is consistent with a contribution from PBH-triggered supernovae. The answer appears to be yes. Across multiple parameter studies, the best-fitting models consistently favored a small but non-zero population of PBH-triggered Type Ia supernovae. Models that completely excluded the PBH channel did not provide the best agreement with observed chemical abundance trends.

A different universe in its youth

Perhaps the most intriguing conclusion concerns the early universe. The simulations suggest that primordial black hole-triggered supernovae may have been considerably more important during the universe's first epochs than they are today. Because white dwarfs could be ignited directly by PBHs without waiting for long-lived binary-star interactions, these explosions may have occurred earlier and more frequently in young galaxies rich in dark matter.
 
The researchers found evidence that the PBH channel could have been one of the dominant Type Ia supernova mechanisms during the universe's formative stages before conventional binary-star pathways became prevalent. If confirmed, this would mean that some of the iron, nickel, manganese, and other heavy elements present in galaxies today may trace their origins not only to stars, but to interactions with relic black holes formed near the beginning of time itself.

Supercomputing as a time machine

What makes this research especially compelling for the high-performance computing community is the extraordinary range of scales involved. The simulations connect physical processes occurring inside white dwarfs a few thousand kilometers across, with the chemical evolution of entire galaxies over billions of years. They bridge nuclear reactions lasting fractions of a second with cosmological questions concerning dark matter and the birth of structure in the universe. Such connections are only possible because modern supercomputing allows scientists to transform speculative ideas into testable models.
 
In this case, the computer becomes more than a calculator. It becomes a time machine, linking the universe's first moments to the elemental fingerprints found in stars today. For the supercomputing community, the message is clear: the next breakthrough in understanding dark matter may emerge not from a particle detector buried underground, but from the convergence of exascale simulation, observational astronomy, and computational astrophysics. And if SUNY Poly's results continue to withstand observational scrutiny, they may reveal that some of the universe's brightest explosions were ignited by some of its oldest objects.

The next challenge for supercomputing isn’t faster AI, it’s public trust

Tyler O'Neal, Staff Editor LATEST

As Artificial intelligence goes mainstream, Americans are demanding more human oversight, accountability

For decades, the supercomputing community has been driven by a singular mission: building faster, more powerful systems to solve increasingly complex problems. This race for performance has yielded remarkable breakthroughs, from modeling climate patterns and accelerating pharmaceutical discovery to designing next-generation aircraft. Today, these computational engines power the foundation models behind artificial intelligence, enabling machines to write code, generate creative content, and perform professional tasks once exclusive to human experts.
 
However, a new national survey from Johns Hopkins University indicates that the future of AI hinges less on raw computational speed and more on public trust. Rather than questioning whether AI should progress, Americans are focused on how it should be governed. The data reveals strong bipartisan support for robust safeguards: 75% of respondents favor mandatory disclosure when interacting with AI, 73% support restrictions on the unauthorized use of personal likenesses, and over 70% advocate for a legal right to human interaction in high-stakes fields like healthcare, education, and legal proceedings. These findings underscore a pivotal shift: the primary challenge of AI has moved beyond the technical realm and into the heart of society.

Supercomputing leaves the lab

Historically, high-performance computing operated largely behind the scenes. Supercomputers helped researchers understand hurricanes, design pharmaceuticals, and explore the origins of the universe. While these systems delivered enormous benefits, they rarely interacted directly with the public. Artificial intelligence has changed that equation.
 
The same computational infrastructure used to train large language models and multimodal AI systems is now reaching millions of people through consumer applications, enterprise software, healthcare platforms, and educational tools. For the first time, the outputs of large-scale computing are being experienced directly by society. This transition marks a fundamental shift in the role of supercomputing. No longer confined to scientific laboratories and research centers, high-performance computing has become a visible part of daily life.

The paradox of AI adoption

What makes the Johns Hopkins findings particularly interesting is that support for regulation extends even among people who regularly use AI systems. This pattern is increasingly visible across multiple surveys conducted during the past year.
 
Research from the University of Pennsylvania’s Annenberg Public Policy Center found that nearly two-thirds of Americans believe the government has done too little to regulate AI. The demand for oversight spans political affiliations, suggesting that AI governance may become one of the few technology issues capable of generating bipartisan consensus. Meanwhile, recent national polling indicates that concerns about AI’s impact on employment continue to rise. More than half of Americans worry that AI could eliminate jobs affecting themselves or members of their household.
 
This creates a fascinating paradox.
 
AI adoption is accelerating, computational capabilities continue to improve, and investment in AI infrastructure remains at record levels. Yet public enthusiasm for unchecked deployment remains limited. Americans appear willing to embrace AI’s benefits while simultaneously demanding stronger safeguards.

Why this matters to the supercomputing industry

For the high-performance computing community, the implications are profound. The next decade will likely see unprecedented investment in AI infrastructure. Hyperscale data centers, accelerated computing systems, specialized AI processors, and exaFLOPS-class architectures are becoming critical national assets. However, the long-term success of these investments may depend less on raw performance metrics and more on whether the public perceives AI systems as trustworthy.
 
History offers numerous examples of transformative technologies whose adoption depended as much on governance frameworks as on technical capability. Aviation requires safety regulations. Pharmaceutical innovation required clinical trials and oversight. Nuclear power requires extensive regulatory systems.
 
Artificial intelligence may be following a similar trajectory.
 
Rather than slowing innovation, well-designed governance structures could become a prerequisite for broader societal acceptance. Research into AI regulation increasingly suggests that standards and transparency mechanisms can support innovation by increasing trust and reducing uncertainty.

Building human-centered supercomputing

One of the survey’s most striking findings is the public’s desire for what researchers describe as a “right to a human.” Americans overwhelmingly want human involvement in medical diagnoses, legal decisions, educational guidance, and government interactions. For technologists, this should not be interpreted as resistance to AI.
 
Instead, it reflects a preference for partnership rather than replacement.
 
The most successful applications of supercomputing have often amplified human expertise rather than eliminated it. Weather forecasting combines computational models with meteorological judgment. Drug discovery combines simulation with scientific expertise. Engineering design combines computational analysis with human creativity. The future of AI may follow the same pattern. Rather than replacing professionals, advanced AI systems may become computational collaborators operating alongside physicians, teachers, scientists, engineers, and public servants.

From performance to responsibility

For much of the supercomputing era, progress was measured in FLOPS, memory bandwidth, and processor counts. Those metrics remain important. But as AI becomes the most visible manifestation of high-performance computing, a new set of measures is emerging: transparency, accountability, explainability, privacy, and trust.
 
The Johns Hopkins survey suggests that Americans are sending a clear message to the technology sector. They are not rejecting artificial intelligence. They are asking for assurances that increasingly powerful computational systems remain aligned with human values and human oversight.
 
That message may ultimately shape the next chapter of supercomputing.
 
The industry’s greatest challenge may no longer be building machines capable of thinking faster. It may be ensuring that society remains confident in how those machines are used.
 
In that sense, the future of supercomputing will not be determined solely by engineering breakthroughs. It will be determined by whether computational power and public trust can advance together.