Science 1663

Close Encounters of the Particle Kind

Like a wise old physics guru, the Monte Carlo computer code MCNP can tell you how subatomic particles will interact with just about anything. The Los Alamos code has helped scientists find water on Mars and aided doctors battling cancer on Earth.

John Hendricks uses dice to help teach prospective MCNP users how the code works.

John Hendricks lets his studentsplay with dice. Most of us associate the little cubes with gambling, but for him they're a great teaching tool. Hendricks, a member of Los Alamos National Laboratory's Applied Physics Division, teaches people to use what he says "is one of the Lab's greatest achievements."

No, it's not "the bomb." It's MCNP—Monte Carlo N-Particle—a computer code that has indirectly touched all of our lives. The oil in our cars was likely found with help from engineers using MCNP (or its predecessor versions, or MCNPX, all of which are referred to here as MCNP). The air we breathe was possibly monitored by a system that relies on MCNP, while anyone who has undergone radiation treatment received a dose that was likely calculated and verified by the code.

The uncommon link between these widespread applications is that each requires knowing what happens to radiation—free particles such as neutrons and photons—as it makes its way through matter. Modeling the "radiation transport" of gamma-ray photons through the body can tell you how many will be absorbed by cancer cells, while following the transport of neutrons from a source to a detector, both placed underground, can tell you if the neutrons passed through oil-saturated rock. Computationally, it turns out that the most accurate way to solve a radiation transport problem is to use random numbers to decide the outcome of a particle's encounters with atoms and nuclei, and repeat the process for many particles. It's a physics-based game of chance, and MCNP is the game's most-skilled player.

MCNPX—Monte Carlo N-Particle eXtended—began in the days of the Accelerator Production of Tritium project, when highly accurate calculations were needed to simulate what happens when an accelerated beam of charged particles hits a tritium-production target. MCNPX extended the then-current MCNP code to include the interactions of virtually all elementary particles and nuclei, at all energies, and thus greatly expanded the types of problems that could be addressed. The new code was a powerful tool for researchers studying cosmic-ray interactions, could be used for medical and industrial applications, or applied to various problems related to threat reduction. In addition, MCNPX incorporated much of the physics relevant to nuclear reactors and could be used for reactor research. After more than 10 years of independent use, MCNPX and the current MCNP5 code are being formally merged into one code.

A Game of Chance

The heavenly bodies of our solar system are constantly bombarded by cosmic rays, causing elements on their surface to emit radiation. By using MCNP to simulate those emissions and their detection by a satellite system, scientists were able to identify water on Mars.

In the classroom, Hendricks poses a question to his students. "Suppose you want to design a nuclear reactor. How do you figure out whether your design produces energy safely, reliably, and sufficiently? In part, you do it by using MCNP to keep tabs on the neutrons."

When a neutron is absorbed by the oversized nucleus of a uranium atom, it can cause the bit of matter to fission, a reaction whereby the nucleus splits into two pieces (fission fragments). Fission unleashes a relatively huge amount of energy—millions of times more than is released from a chemical reaction between atoms.

Significantly, a few neutrons are also unleashed in the nuclear breakup. In what's known as a chain reaction, those freed neutrons can cause more nuclei to fission, unleashing more energy and even more neutrons, and so forth. A nuclear reactor's primary job is to achieve a steady state chain reaction: to keep fission going and the energy flowing.

A French advanced fuel assembly.

But other kinds of nuclear reactions may occur when a neutron meets up with a uranium nucleus. For example, the neutron may be captured—the uranium absorbs it but does not fission. Or the neutron may simply bounce off a nucleus and scatter. One cannot determine in advance what happens in any specific encounter, but can only determine the probability for a given reaction to occur.

MCNP knows the different probabilities from experimental data. It then uses one random number to choose between reactions (fission, capture, scattering, or other), and others to select a specific outcome: the number of neutrons and/or other particles produced, their final energies, the direction each taken by each, even how far each goes before encountering a new nucleus. The program generates its random numbers with a simple algorithm. Hendricks and his students obtain them by tossing dice. (See "A Roll of the Dice")

Making independent choices for each new encounter, MCNP is able to construct a realistic trajectory of a neutron through the reactor. It builds such a trajectory for thousands, if not millions, of neutrons, enough to create a statistically accurate picture of their fate and consequences. MCNP can then calculate the rate at which neutrons are produced by fission and the rate at which they are lost. When the production rate equals the loss rate, you get a steady-state chain reaction, which is the beating heart of a nuclear reactor.

The Best of the Best

In many ways, MCNP, with its attendant nuclear data sets, is the repository of all we know about the interactions between radiation and matter. It's not the only code that deals with those interactions; it's one of many "Monte Carlo" codes, so called for their reliance on chance (see "The Code Was in the Cards"). But if it's not unique, it is the gold standard, its superiority anchored by a complete set of physics data, efficient ways to reduce statistical uncertainty, a sophisticated graphics interface, and an uncommonly bug-free source code.

"It's the best of the best," says Hendricks, who, like Art Forster, Tom Booth, and Gregg McKinney of the Applied Physics Division and Laurie Waters of Decision Applications Division, has devoted his career to MCNP. "It accurately describes anything and everything that can happen in a particle-nucleus collision."

The Advanced Fuel Cycle

This figure shows MCNP predictions for the radiation signals emanating from a person contaminated with a known dose of the radioactive isotope cobalt-60. Note that cobalt-60 accumulates primarily in the lungs and liver.

MCNP is being used by nuclear engineers and scientists, including some at Los Alamos, to address one of the more pressing issues of our time—the production of nuclear energy. The goal is to develop new reactors, new nuclear fuels, and enhanced safeguard systems that will make nuclear energy safer, produce less waste, and be resistant to the diversion of plutonium.

In much of the world, nuclear power is produced using an "open" fuel cycle: enriched uranium fuel is burned (fissioned) in a reactor until it can no longer produce energy economically, at which point the fuel is deemed spent, removed from the reactor, and considered waste to be discarded. But spent fuel still has lots of energy-rich fissionable material, so the open cycle is analogous to a business that doesn't recycle.

The scientific understanding needed to close the fuel cycle and implement a recycle strategy already exists. The U.S. Department of Energy's Advanced Fuel Cycle Initiative (AFCI), which is the research and development arm of the Global Nuclear Energy Partnership initiative, seeks to optimize the more mature technologies and develop new ones that would enhance energy extraction from the nuclear fuel, minimize waste, and reduce proliferation risks. This includes developing and refining technologies to chemically reprocess the spent fuel, that is, remove the uranium and materials known as transuranics, and turn them into an entirely new type of fuel that can be burned for further energy production.

The transuranics, elements heavier than uranium (neptunium, plutonium, americium, etc.), are created in the fuel by nuclear processes. Plutonium, for example, is created after uranium captures a neutron and decays to neptunium, which then decays to plutonium.

Transuranics are the Methuselahs of spent fuel, taking hundreds of thousands of years to decay. That time scale underlies all efforts to stabilize the waste and develop repositories for its storage or disposition. By burning the transuranics, the advanced fuel cycle squeezes out more energy from each bit of fuel and creates a much simpler waste disposal process. Researchers are developing optimized mixtures of uranium, plutonium, and other transuranics for the new TRU fuels (so-called for their enhanced TRansUranic component).

One complication is that today's most-common commercial reactors, thermal reactors, don't burn TRU fuel efficiently. Thermal reactors depend on slow, or low-energy, neutrons for fission. The new fuels would require the use of fast-neutron reactors, designed to sustain fission with neutrons of very high energy. Through the AFCI Program, DOE is supporting research into this next generation of reactors, and MCNP is being used to help design them.

Meanwhile, work on the new fuels is underway, and once again, MCNP is involved.

A Roll of the Dice

How can choosing a random number lead to physically relevant descriptions? John Hendricks explains how to a class of young engineers:

“Suppose that in an imaginary nuclear reactor, there are only four possible outcomes for a neutron-nucleus encounter: (A) The neutron is absorbed and captured by the nucleus, so no neutrons remain after the encounter; (B) the neutron scatters off the nucleus, so one neutron remains; (C) the neutron is absorbed and the nucleus fissions, producing two neutrons, or (D) the nucleus fissions and produces three neutrons.

“Next, imagine that experiments with our nucleus reveal that for every six neutron encounters, (A) happens 3 times, and (B), (C), and (D) each happen once. How do we model this physics?

A French advanced fuel assembly.

“We’ll be rolling a die that generates a number from 1 to 6 at random and with equal probability. We have to make the six numbers correspond to the four possible outcomes in a way that matches the probabilities that were measured in our experiments. This is done by assigning the die numbers 1, 2, and 3 to (A), 4 to (B), 5 to (C) and 6 to (D). In essence, we’ve transformed the ‘real world’ into an equivalent ‘computational world,’ where each encounter now has six outcomes that result in either 0, 0, 0, 1, 2, or 3 neutrons. Because our die is unbiased, each outcome is equally likely to occur, so on average, six rolls of the die will produce a total of six neutrons.

“To find out how our reactor behaves, we do a Monte Carlo simulation. We’ll start with many neutrons in the reactor. Each neutron will encounter a nucleus, and a roll of the die will determine the outcome (how many neutrons remain). We’ll then average our results and calculate the uncertainty of our answer.

“If we run our simulation with a million neutrons, we will find, on average, that a million neutrons remain, that is, our reactor achieves a steady-state chain reaction! The relative uncertainty of our result will be about 0.1 percent. One of MCNP’s most powerful features is its use of mathematical shortcuts that reduce the uncertainty without having to ‘roll dice’ a million times. Thank goodness for that!”

MCNP and the Materials Test Station

The road to the advanced fuel cycle will pass through a new facility planned for construction at the Los Alamos Neutron Science Center (LANSCE): the Materials Test Station (MTS). Expected to open in 2012, the MTS will provide a steady stream of neutrons that approach, both in energy and in flux (neutrons per square centimeter per second) the flow expected within a fast-neutron reactor. Candidate TRU fuels for the new reactors, some of which are being developed at Los Alamos, can thus be tested in advance.

The engine powering the MTS is the LANSCE facility's 800-million-electronvolt proton beam. When the protons slam into the MTS's tungsten target, copious neutrons will be produced (about 20 per proton-tungsten nucleus collision). MCNP is critical for simulating the results of those collisions and for following the torrent of neutrons as they zip and scatter their way into the candidate fuels.

"We'll measure how much of the fuel's transuranic material fissioned during irradiation," says Eric Pitcher, the test station project manager. "MCNP will tell us the neutron flux, so we'll know how effectively each fuel burns. We'll also monitor a host of other things, including how much radiation damage is done to the fuel rod material (the cladding). The cladding maintains an all-important barrier between the fuel and the coolant and must perform with a high degree of integrity and robustness."

One Code to Model Them All

Anna Hayes of Theoretical Division heads an effort to merge MCNP with other programs to create a reactor-performance program.

While MCNP can tackle any radiation transport problem, it can't model the bulk properties of the reactor materials or predict other types of behaviors. For example, the circulation of the liquid coolant in a reactor needs to be modeled by a "hydrodynamics" code.

Anna Hayes and her colleagues in Theoretical Division are developing an "omnibus" design code that combines MCNP with a hydrodynamics code and a nuclear fuel burn-up code. It is also tied to a code that calculates the equation of state (EOS) of a material, which tells you how the material's properties change as a function of temperature and pressure.

"It's important to track the flow of energy in a fast reactor very accurately. Fission fragments don't go very far, so they dump their kinetic energy within the fuel rod," says Hayes. "We estimate that energy with a fuel burn-up code, CINDER. The neutrons travel farther and tend to transport energy from one fuel rod to another. MCNP tracks the neutrons' energy and tells us the amount dumped into the cladding. The information is handed off to the ‘hydro' and EOS codes to determine the impact of this heat on the cladding and coolant."

One concern that the group will address is the production of helium and xenon gas bubbles in the fuel and cladding. They'll be looking to see if the gases could affect the integrity of the fuel rod.

Keepin' Things Safe

Nuclear Nonproliferation’s Martyn Swinhoe holds a neutron-sensitive detector, one of many stationed vertically inside the rectangular frame beneath his hands. Fuel rods containing fresh uranium/plutonium fuel are lowered through the frame. Neutrons produced by the plutonium are detected. MCNP interprets the neutron signal and verifies the stated plutonium content.

A final area where MCNP is playing a large role is in nuclear safeguards, that is, keeping track of plutonium and other fissionable materials throughout the nuclear fuel cycle. The Los Alamos Nuclear Nonproliferation Division's Martyn Swinhoe uses the code to design radiation detectors for organizations such as the International Atomic Energy Agency that keep tabs on nuclear material worldwide. In the last year Swinhoe and his team have used MCNP to develop nine different monitoring instruments for four different organizations intent on surveying the nuclear material in the fuel reprocessing plant at Rokkasho, Japan.

MCNP determines the behavior of the neutrons that are emitted from plutonium in a fuel rod and reach the neutron-sensitive elements within the detector. The code provides Swinhoe with an estimate of the neutron detection rate (the total counting rate) and also the rate at which pairs of neutrons from a fissioning nucleus will be detected (the coincidence counting rate). The latter gives direct information on the mass of plutonium in the fuel rod.

"We use MCNP to optimize the design of the detectors before they are built," says Swinhoe. "It saves us a lot of experimental work on prototypes.

With nuclear energy making a comeback around the world, it's not surprising to see MCNP turning up as the tool of choice for designing both the nuclear fuel cycle of the future and the safeguards to control where and how nuclear fuel and its waste products are used.

The Materials Test Station will verify the performance of advanced reactor materials and fuels. The top simulation, derived from MCNP data, shows two proton beams (colors) being stopped in the twin tungsten targets. Red is highest proton flux, blue is lowest.

Fast neutrons produced by the protons permeate the fuels and materials placed within the circles and within the small rectangles surrounding the targets. Red is highest neutron flux, blue is lowest.

The Code Was in the Cards

In 1946, a game of canfield solitaire led Stanislaw Ulam, the Polish mathematician and Manhattan Project pioneer, to a brainstorm. Having returned to Los Alamos, Ulam was temporarily idled by illness and passed the time playing the game. He also pondered the odds of winning any given hand but was stymied by the combinatorial equations needed to exactly solve the probability. Instead, he hit on the idea of using statistical sampling: he would zero in on the probability of a successful outcome by laying out a large number of hands and keeping track of the winners.

Stanislaw Ulam

Statistical sampling was not original to Ulam, but his musings on cards and chance led to two key insights. He first recognized that statistical sampling was well suited to the problem of neutron diffusion and next realized that the new electronic computers could generate the potential pathways of the neutron through the material. He shared these ideas with Hungarian mathematician John von Neumann, who at the time was a consultant to both Los Alamos and the Ballistics Research Laboratory in Maryland, where the first general-purpose digital computer, the ENIAC, was being built.

Liking the idea, von Neumann in 1947 laid out the neutron diffusion problem in a 19-step computing sheet: in essence, the first Monte Carlo code. The flamboyant name was given to the method by Nicholas Metropolis, architect of Los Alamos' first computers. Ever since then, statistical-sampling computer codes have been known as Monte Carlo codes.