Research

Three grand challenges make the need visible

Can fusion become a reliable source of energy? What determines whether a massive star explodes and spreads newly formed elements through the universe? How can a hypersonic vehicle or atmospheric-entry system be predicted before it is built or flown? These systems differ greatly, but each becomes uncertain when small-scale interactions that cannot be computed directly control a large-scale result.

Every useful model simplifies nature

Scientists move from particle interactions to statistical descriptions, from those descriptions to fluid-scale models, and from fluid motion to affordable models of turbulence and mixing. Each step makes prediction possible, but it also removes information. Scientists often call the rule that returns the influence of the removed information a closure. TEMPEST asks which missing influence truly matters and how it should be represented.

Studying one piece in isolation can hide the real source of error

A particle-scale model may be accurate while passing incomplete information to a fluid model. A fluid model may be accurate in a simple setting while failing when turbulence, shocks, material changes, or long-lived memory become important. An application code may then inherit errors created several modeling steps earlier. Looking at only one scale can improve a local piece without improving the final prediction.

This is why TEMPEST must be a Center

The needed connections cross plasma physics, fluid dynamics, applied mathematics, experiments, scientific computing, artificial intelligence, uncertainty analysis, national laboratories, and industry. A Center can align questions, measurements, simulation outputs, mathematical requirements, software, and workforce training across those boundaries. No single project, facility, or discipline can see the entire chain from microscopic interaction to application decision.

Fusion, supernovae, and hypersonics provide independent tests

TEMPEST does not assume that the three applications are the same. Their differences are scientifically valuable. A modeling idea that remains credible across distinct regimes is stronger evidence of captured physics than a model tuned to one experiment, one code, or one dataset. When an idea fails to transfer, the failure helps identify what information is still missing.

One connected chain of models

No practical calculation can resolve every particle, collision, chemical reaction, field, instability, shock, and turbulent structure. Scientists therefore use a sequence of models, each designed for a different scale. TEMPEST studies what is lost at each transition, when that loss changes the answer, and how the missing influence can be restored without making the calculation impossible.

Fusion energy: heat flow, stability, and mixing

A fusion plasma can be compressed and heated so quickly that particle motion, magnetic fields, material response, instabilities, and turbulent mixing all interact. TEMPEST asks which of these effects most strongly change energy gain, heat loss, magnetic-field evolution, mixing, and other design-relevant outcomes. Evidence can come from detailed particle simulations, controlled plasma and turbulence experiments, extreme-matter measurements, existing fusion data, and partner tests.

Supernovae and stardust: explosion, elements, and observable signals

A core-collapse supernova combines gravity, radiation, neutrinos, magnetic fields, nuclear matter, and turbulence. The same unresolved motions that influence whether the star explodes can also affect the production and mixing of elements and the signals observed by telescopes and detectors. TEMPEST connects models to shock motion, explosion energy, element production, neutrino signals, gravitational waves, and light.

Hypersonics: shocks, heating, and survivability

At extreme speed, shock waves interact with turbulent boundary layers, surface roughness, high-temperature chemistry, flow separation, and intense wall heating. TEMPEST asks what must be represented to predict forces, heat loads, separation, transition, performance, and safety before flight. Evidence comes from matched experiments and simulations, time-resolved measurements, and carefully designed comparisons.

A credible prediction needs more than a good fit

The model must obey basic physical rules, remain stable when used inside a simulation, produce physically meaningful states, agree with measurements that can distinguish competing explanations, and state how much uncertainty remains. A model that works only on the data used to build it is not enough.

The Center creates the missing connections

Experimental teams can measure effects that modelers need; theorists can state the rules a model must obey; AI researchers can represent relationships that are too complex for a simple formula; application teams can determine which errors actually matter. TEMPEST brings those roles into one repeated cycle of question, prediction, test, revision, and transfer.

Better prediction changes how difficult science is pursued

Fusion experiments, large astronomical simulations, and hypersonic tests are expensive and cannot explore every possible condition. A trustworthy model does not replace measurement. It helps select the most informative experiment, interpret what was observed, identify the uncertainty that controls a decision, and show where new evidence is needed.

Energy

More reliable fusion models could help researchers choose experiments, compare design options, identify the transport or mixing processes that limit performance, and reduce uncertainty before costly hardware is built. The long-term opportunity is progress toward clean, firm energy and continued leadership in a strategically important technology.

Discovery

Supernovae are natural laboratories at scales no experiment can reproduce in full. Better models can connect simulations more rigorously to neutrino, gravitational-wave, and electromagnetic observations, clarify why some stars explode, and improve understanding of how elements are formed and dispersed as stardust.

Security and space

More trustworthy prediction of shocks, heat transfer, chemistry, separation, and instability could improve the interpretation of ground and flight tests, support safer atmospheric entry and access to space, and reduce reliance on trial-and-error development in high-consequence systems.

People and tools

TEMPEST will produce reusable datasets, software, benchmarks, documented models, uncertainty reports, experimental practices, and accessible explanations. It will also train scientists who can move among experiments, mathematics, computation, AI, software, and application needs.

A Center multiplies the value of each result

A dataset becomes more useful when several modeling teams use the same definitions and uncertainty information. A new method becomes more credible when independent experiments and applications test it. A trainee becomes more effective when they understand how their work connects to the larger scientific and national mission.

The missing-information hypothesis

In many extreme systems, the main limitation is not simply that a computer grid is too coarse. Important interactions, particle distributions, memory, or small-scale structures have been removed when the model was simplified. Representing the right missing influence should improve prediction more than adding cost without changing the model itself.

The connected-scales hypothesis

Errors introduced at one modeling step can pass into the next. Particle-scale assumptions can change fluid transport; fluid-scale assumptions can change instability and mixing; turbulence assumptions can change an application outcome. Treating the steps together should identify failures that remain hidden when each piece is tested alone.

The physically guided AI hypothesis

AI can help learn complicated missing relationships from data, but the result will transfer more reliably when it is required to respect known physical rules, known limits, and uncertainty. Accuracy on one training dataset is not sufficient evidence.

The adaptive-model hypothesis

A single fixed model may be unnecessary in ordinary conditions and inadequate in extreme ones. A model that can recognize when additional information is needed—and add complexity only then—may provide a better balance of speed, accuracy, and reliability.

The discriminating-evidence hypothesis

Not all data are equally useful. Measurements and simulations designed to separate competing explanations should improve models faster than large collections of data that do not directly test the missing physics.

The transfer hypothesis

A scientific idea that improves prediction in fusion, supernovae, and hypersonic flow—or in carefully chosen parts of more than one application—has passed a stronger test than an idea tuned to one case. Failure to transfer is also valuable because it reveals the limits of the idea.

Evidence determines the next step

TEMPEST will use results to refine an idea, transfer it to a new setting, or stop investing in a path that is not supported. The purpose of a hypothesis is not to protect an expectation; it is to organize learning.

What outcome must be predicted? 

The question begins with a quantity that matters: fusion energy gain or mixing, supernova explosion energy or observable signals, hypersonic heat loads or vehicle performance. Starting with the outcome prevents the research from becoming disconnected from a real scientific or engineering need.

What may the model have left out? 

The missing influence might be a particle interaction, a nonstandard distribution of particle speeds, a delayed response, transport across a large region, an instability, or unresolved turbulence. TEMPEST identifies the specific modeling step at which important information may have been removed.

What evidence can decide? 

A useful test must distinguish between competing explanations. It may be a laboratory measurement, a high-detail simulation, an astronomical observation, a comparison across codes, or a partner benchmark. The evidence should be chosen because it can change our confidence, not simply because it is available.

What physical rules and uncertainty matter? 

The proposed model must respect the rules required by the surrounding science, such as conservation of mass and energy, positive density and temperature, stable behavior, and correct results in well-understood limits. The remaining uncertainty must be stated clearly enough to support a decision.

What changes if the idea is right? 

The result might alter an experiment, a simulation, a design choice, a measurement plan, a software module, or the decision to continue a line of research. This step turns a scientific question into a traceable path toward use.

The Center keeps the question complete

Application experts define the consequential outcome; experiments and observations provide decisive evidence; theory identifies the missing information and required rules; AI and computation provide flexible and efficient models; partners help determine whether the improvement changes practice.

Why simplification is necessary

A real plasma or fluid contains an enormous number of interacting particles. Following every position, velocity, and collision over an application-sized region is far beyond practical computing. Scientists instead describe how particles are distributed and how groups of particles influence one another.

What can be lost

The average behavior may look simple even when correlations among particles, rare energetic particles, long travel distances, or delayed interactions control heat flow, diffusion, relaxation, or the formation of collective structures. Assuming those effects are unimportant can create an error before the model ever reaches the fluid scale.

What TEMPEST will do

The Center will combine detailed particle simulations, particle-resolved plasma experiments, extreme-matter measurements, and mathematical analysis to determine which interactions must be retained and which can be represented by a simpler rule.

How evidence will be used

Researchers will compare predicted transport, relaxation, distributions, and collective structures with measurements and high-detail calculations. The goal is not to reproduce every particle path; it is to reproduce the quantities that control the next level of prediction.

Why the connection to other scales matters

A particle-scale result is useful only if its influence can be carried into a fluid or application model. TEMPEST links the teams working at each step so the information passed forward is defined, tested, and usable.

Why fluid models are powerful

Instead of tracking every particle, a fluid-scale model follows a small set of average quantities across space and time. This makes it possible to simulate an entire fusion target, stellar region, or hypersonic flow.

Why averages may not be enough

Two particle populations can have the same density and temperature while carrying heat, stress, or directional motion in very different ways. Rapid compression, strong fields, shocks, or long travel distances can make those differences important.

How the missing influence can return

The model may need an additional quantity, a correction to heat or momentum transport, a history-dependent response, or a rule that activates only when the system leaves familiar conditions. TEMPEST will compare these options rather than assuming one formula works everywhere.

How TEMPEST will test the model

 Detailed particle calculations and experiments will provide reference behavior. The reduced model will then be tested for accuracy, stable operation, physically meaningful states, and correct behavior when it returns to well-understood conditions.

Why a Center is needed

The variables chosen by particle and fluid researchers determine what turbulence and application teams receive. By designing this handoff together, the Center can avoid a common failure in which each individual model looks reasonable but the combined prediction does not.

Why turbulence must be modeled

Turbulent motion spans a wide range of sizes. Resolving every motion in an application-scale calculation can be too expensive, so simulations compute the larger structures and model the influence of smaller ones.

Why extreme systems are harder

Standard approaches often assume that small-scale turbulence responds locally and quickly. Those assumptions can fail when shocks, strong material changes, particle trapping, rapidly growing instabilities, or long-lived structures create memory and transport over large distances.

What TEMPEST will measure

Shock facilities, variable-property flow experiments, instability and mixing experiments, particle tracking, velocity measurements, and high-detail simulations will reveal separation, transport, mixing, delayed response, and other quantities that a turbulence model must reproduce.

What TEMPEST will build

The Center will develop models that add memory, nonlocal influence, material changes, or learned corrections only when the evidence shows they are needed. The model must remain stable and affordable inside a large simulation.

Why the scales must stay connected

Turbulence is shaped by the fluid model it receives, and its effects feed directly into application outcomes. TEMPEST tests that full connection rather than validating a turbulence formula only in a separate, idealized flow.

AI fills a defined gap

TEMPEST will not ask a black-box system to replace all of physics. Researchers first identify where a trusted model loses information, then use AI to represent the missing relationship, correction, or fast approximation.

Physical knowledge guides learning

A learned model can be required to conserve mass or energy, respect symmetry, keep density and temperature meaningful, remain stable, and recover known behavior in familiar limits. These requirements reduce the chance that a model fits one dataset while behaving unphysically elsewhere.

Uncertainty is part of the result

The model should indicate when it is being used outside the conditions supported by data. That information can guide a new experiment, a higher-detail simulation, or a decision to use a more complete model.

AI connects experiments and theory

Experiments reveal where current models fail; theory defines the rules and possible forms of the missing influence; AI provides a flexible representation; validation determines whether it transfers to new conditions.

Why a Center matters

Useful scientific AI requires coordinated data, physical expertise, software integration, validation, and application testing. TEMPEST brings those pieces together so a learned model is judged by scientific use, not only by a training score.

Connecting scales is the central task

TEMPEST builds mathematical and computational bridges from particle interactions to statistical models, from statistical models to fluid behavior, and from fluid behavior to turbulence and application prediction.

Theory identifies the right variables

When a model fails, the answer may be another state variable, a delayed response, influence from a larger region, a better transport rule, or a recognition that the current modeling level is no longer adequate. Theory helps distinguish among these possibilities.

Credibility can be tested mathematically

Reduced and learned models should conserve the right quantities, remain stable, keep physical variables in meaningful ranges, and recover well-understood behavior when extreme effects disappear. These checks can rule out a model before it causes a failure in a large application simulation.

Simulation turns ideas into predictions

High-detail calculations create controlled reference cases; reduced calculations test speed and transfer; application simulations show whether the new model changes a consequential result.

The Center keeps theory connected to evidence and use

Mathematical ideas are developed with experimental measurements, AI representations, software requirements, and application questions in view. This helps prevent elegant theory from remaining detached from the quantities that can actually be tested.

Experiments are designed to answer model questions

The most useful measurement is not always the largest dataset. TEMPEST will identify observations that distinguish between competing explanations and show which modeling assumption needs to change.

The Center uses complementary platforms

Particle-resolved plasma experiments expose interactions and collective behavior. Shock and boundary-layer facilities reveal compression, separation, and memory. Variable-property and mixing experiments show how changing materials alter turbulence. X-ray measurements of extreme matter probe transport under conditions where ordinary assumptions may fail.

Controlled experiments isolate mechanisms

 These facilities are not small copies of a complete fusion target, star, or vehicle. They create clean tests of specific processes so researchers can determine cause and effect, measure uncertainty, and compare models on common ground.

Shared evidence links the Center

Experimental teams, simulation teams, and application teams will agree on quantities, data formats, uncertainty descriptions, and comparison procedures. That allows a result from one platform to improve more than one model or application.

The loop continues

A measurement exposes a model failure; theory and AI propose an improved representation; simulation predicts a new signature; the next experiment tests that signature.

The science requires boundary-crossing skills

A researcher studying a missing modeling influence may need to understand how it is measured, how it enters an equation, how uncertainty is estimated, how software implements it, and why it matters to an application.

Training is built into the research

Cross-institutional exchanges, facility rotations, collaboration retreats, shared mentoring, interdisciplinary courses, and the TEMPEST Turbulence Symposium will place trainees inside teams that use different methods and language.

Communication and open science are core skills

Trainees will learn to document data and software, explain assumptions, create reproducible workflows, communicate with public and technical audiences, and understand how scientific evidence supports decisions.

The Center creates a network, not isolated apprenticeships

Students and postdoctoral researchers can connect universities, laboratories, industry, experiments, and application teams, carrying methods and questions across institutional boundaries.

Extreme science offers visible entry points

Fusion energy, exploding stars, hypersonics, turbulence, AI, and experiments can help people see how fundamental science connects to energy, discovery, security, space, and technology.

Accessible explanations build trust

Images, animations, plain-language pages, captions, alternative text, and clear statements of uncertainty can show what is known, what is being tested, and what TEMPEST expects to learn.

Participation must be broader than observation

Undergraduate research, community-college pathways, teacher resources, public programs, open educational material, and partnerships with under-resourced institutions can create routes into the work.

Open resources extend the Center

Where appropriate, datasets, software, benchmarks, tutorials, and public events will allow students, researchers, educators, and partners to reuse what TEMPEST creates.

Applications identify which errors matter

A modeling difference is important when it changes an outcome such as fusion gain, supernova explosion and element production, or hypersonic heat loads and performance. Application teams help the Center focus on quantities that can affect a scientific or engineering decision.

The three applications are independent tests

Fusion is a laboratory and design challenge, supernovae are observed cosmic events, and hypersonics combines ground tests, flight, chemistry, and fluid dynamics. An idea that transfers among them has survived very different evidence and operating conditions.

Validation proceeds in stages

TEMPEST first checks basic physical and mathematical behavior, then controlled simulations, then focused experiments or observations, then established application benchmarks, and finally partner-relevant use. Claims grow only as the evidence grows.

The applications feed back into fundamental science

A failed target calculation, a mismatch with an astronomical signal, or an inaccurate heat-load prediction can reveal which scale, measurement, or assumption needs attention.

A Center makes the comparison possible

Common definitions, shared data and software, coordinated uncertainty analysis, and cross-application teams allow the Center to distinguish a transferable scientific insight from a correction that works only in one setting.