Case Study 18.1 — Neutron Resonances in ${}^{238}$U: The Key to Reactor Physics

The Problem

Uranium-238 constitutes 99.3% of natural uranium. It is not fissile — a thermal neutron will not cause it to fission — but it is a voracious absorber of epithermal neutrons. The neutron cross section of ${}^{238}$U between 1 eV and 10 keV is a dense forest of resonances, some reaching peak cross sections of 20,000 barns or more. These resonances are the dominant obstacle to achieving a sustained nuclear chain reaction in a thermal reactor: neutrons that are born at MeV energies from fission must slow down (moderate) to thermal energies to efficiently fission ${}^{235}$U, and during this slowing-down process they must pass through the ${}^{238}$U resonance region without being absorbed.

Understanding, measuring, and accounting for these resonances is one of the central technical achievements of nuclear engineering. This case study traces the physics from the first resolved resonance at 6.67 eV to the design of modern reactor fuel.

The Resonance Landscape

The First Resonance: 6.67 eV

The lowest-energy s-wave neutron resonance in ${}^{238}$U occurs at $E_R = 6.674$ eV. Its parameters, measured to high precision by time-of-flight experiments at facilities such as ORELA (Oak Ridge Electron Linear Accelerator) and GELINA (Geel Linear Accelerator), are:

The compound nucleus is ${}^{239}$U$^*$ at an excitation energy of $E^* = S_n + E_R \approx 4.806$ MeV, where $S_n = 4.806$ MeV is the neutron separation energy.

The peak total cross section is approximately:

$$\sigma_{\text{tot}}(E_R) = 4\pi\lambdabar_R^2 g_J \frac{\Gamma_n}{\Gamma} \approx 23{,}000 \text{ b}$$

This is an extraordinary cross section — the effective target area presented by the ${}^{238}$U nucleus to a 6.67 eV neutron is $23{,}000 \times 10^{-24}$ cm$^2$, or about $1.5 \times 10^5$ times the geometrical cross section of the nucleus ($\pi R^2 \approx 1.7$ b). The quantum-mechanical enhancement is entirely due to the large de Broglie wavelength: $\lambdabar \approx 1.8 \times 10^{-10}$ cm at 6.67 eV, compared to the nuclear radius of $\sim 7 \times 10^{-13}$ cm.

The Resonance Forest

Above 6.67 eV, the resonances come thick and fast. The average s-wave level spacing is $D_0 \approx 20.3$ eV for ${}^{239}$U$^*$ at the neutron separation energy. Between 6 eV and 2 keV, there are roughly 100 resolved s-wave resonances. Including p-wave resonances (which become important above $\sim 500$ eV), the total count exceeds 200.

The resonance parameters vary widely. The radiation width is remarkably constant: $\Gamma_\gamma \approx 23$–$26$ meV for almost all resonances, reflecting the statistical averaging over thousands of gamma-ray transitions. The neutron width, in contrast, varies over two orders of magnitude — from less than 0.01 meV to more than 10 meV — following the Porter-Thomas distribution. The largest neutron widths produce the most prominent resonances; the smallest produce barely visible bumps on the cross section curve.

Above $\sim 4$ keV, individual resonances can no longer be resolved — they overlap. In this unresolved resonance region (URR), the cross section is described statistically using average resonance parameters and level densities, either through the Hauser-Feshbach model or through stochastic "ladders" of resonances sampled from the known statistical distributions.

Reactor Physics Consequences

The Resonance Escape Probability

In a thermal reactor, neutrons born from fission at $\sim 2$ MeV must slow down to thermal energies ($\sim 0.025$ eV) to efficiently cause fission in ${}^{235}$U. During this slowing-down process, neutrons with energies matching any ${}^{238}$U resonance have a high probability of being captured. The fraction that survives is the resonance escape probability $p$, one of the four factors in the reactor criticality condition $k_{\text{eff}} = \eta \epsilon p f$.

For a homogeneous mixture of fuel and moderator, the resonance escape probability is:

$$p = \exp\left(-\frac{N_{238} I_{\text{eff}}}{\xi \Sigma_s}\right)$$

where $I_{\text{eff}}$ is the effective resonance integral, $\xi$ is the average logarithmic energy decrement per collision in the moderator ($\xi = 1$ for hydrogen), and $\Sigma_s$ is the macroscopic scattering cross section.

The infinite-dilution resonance integral (no self-shielding) for ${}^{238}$U is $I_\gamma^{\infty} = 275$ b. The 6.67 eV resonance alone contributes $\sim 44$ b — about 16% of the total. The remaining 84% is distributed among the hundreds of higher-energy resonances.

Self-Shielding and Fuel Rod Design

In practice, the effective resonance integral is much smaller than 275 b because of self-shielding: the intense absorption at resonance energies depletes the neutron flux inside the fuel, so that interior atoms "see" a lower flux at those energies. Self-shielding is enhanced by:

• Heterogeneous fuel geometry — concentrating the fuel in discrete rods or pellets rather than mixing it uniformly with the moderator. The surface-to-volume ratio determines the degree of self-shielding.
• Temperature — higher temperatures increase Doppler broadening, which broadens the resonances and partially offsets self-shielding by exposing more of the interior to near-resonance neutrons.

For a typical PWR (pressurized water reactor) fuel rod (UO$_2$ pellet, radius $\sim 4$ mm, ${}^{235}$U enrichment $\sim 4$%), the effective resonance integral is approximately $I_{\text{eff}} \approx 25$–$30$ b at room temperature, rising to $\sim 35$–$45$ b at operating temperature ($\sim 900$ K). The resonance escape probability is $p \approx 0.87$–$0.92$, meaning 8–13% of all neutrons are captured by ${}^{238}$U resonances during slowing down.

The Doppler Coefficient: An Inherent Safety Feature

The temperature dependence of resonance absorption provides one of the most important safety features of thermal reactors: the negative Doppler temperature coefficient of reactivity. As fuel temperature increases:

1. Thermal motion of ${}^{238}$U atoms broadens the effective resonance cross section (Doppler broadening).
2. The peak cross section decreases, but the wings of the resonance extend further in energy.
3. In a self-shielded geometry, the increase in the wings (where the flux is not fully depleted) more than compensates for the decrease at the peak.
4. Net effect: more neutrons are captured by ${}^{238}$U, reducing the neutron multiplication and hence the reactor power.

This is a passive, inherent safety mechanism — it requires no operator action, no control system, no external power. It is the reason thermal reactors do not undergo runaway power excursions during normal transients. The Doppler coefficient for a typical PWR is approximately $-2$ to $-4$ pcm/K (where 1 pcm = $10^{-5}$ $\Delta k/k$).

Historical Context: Decades of Measurement

The history of ${}^{238}$U resonance measurements tracks the history of nuclear physics itself. The first resonance at 6.67 eV was identified in the 1940s during the Manhattan Project, when understanding neutron absorption in uranium was critical to the development of nuclear reactors and weapons. The earliest measurements used rudimentary mechanical velocity selectors; by the 1950s and 1960s, electron linear accelerators (linacs) at facilities like Columbia University, Harwell (UK), and Oak Ridge provided pulsed neutron sources with sufficient intensity and timing resolution to map dozens of resonances.

The push for higher-energy resonances and greater precision continued through the 1970s–2000s, with the construction of dedicated time-of-flight facilities: ORELA at Oak Ridge (186 m flight path, operated 1969–2012), GELINA at Geel, Belgium (30–400 m flight paths, still operational), and n_TOF at CERN (185 m and 20 m flight paths, operational since 2001). Each generation of measurements reduced the uncertainties on the resonance parameters and extended the resolved resonance region to higher energies.

The current ENDF/B-VIII.0 evaluation (released 2018) contains R-matrix parameters for over 200 resolved resonances below 20 keV, plus average resonance parameters for the unresolved region up to 150 keV. The uncertainties on the key parameters of the first few resonances are now at the sub-percent level — a remarkable achievement for measurements that began nearly 80 years ago.

The Data Pipeline

The resonance parameters of ${}^{238}$U have been measured and re-measured over more than 60 years. The chain from measurement to reactor design illustrates the critical infrastructure of nuclear data science:

1. Experiment — TOF transmission and capture measurements at facilities like n_TOF, GELINA, and DANCE (Detector for Advanced Neutron Capture Experiments, Los Alamos) yield raw data: counts versus time of flight, corrected for backgrounds, dead time, and detector efficiency.
2. R-matrix analysis — Codes like SAMMY (Oak Ridge) fit the multi-level R-matrix formula to the transmission and capture data simultaneously. The fitting accounts for experimental resolution, Doppler broadening, self-shielding, multiple scattering, and normalization uncertainties. A single evaluation of ${}^{238}$U may fit hundreds of resonance parameters to millions of data points.
3. Evaluation — Nuclear data evaluators at national data centers (NNDC at Brookhaven, NEA Data Bank in Paris, JAEA in Japan) assess all available experimental data, resolve discrepancies between datasets, and produce evaluated data files in ENDF-6 format.
4. Processing — Codes like NJOY (Los Alamos) process the point-wise evaluated data into the multi-group cross section libraries, probability tables, and S($\alpha$,$\beta$) thermal scattering data needed by reactor physics codes.
5. Transport calculations — Monte Carlo codes (MCNP, Serpent, OpenMC) and deterministic codes (SCALE, CASMO) use the processed nuclear data to simulate neutron transport, reaction rates, and power distributions in the reactor core.
6. Validation — Calculated results are compared to integral benchmark experiments: critical assemblies (ICSBEP handbook), reactor physics experiments, and post-irradiation examination data. Discrepancies between calculation and experiment drive improvements to the nuclear data.

Every step in this chain depends on the accurate characterization of the ${}^{238}$U resonances. A 5% error in the neutron width of the 6.67 eV resonance propagates to a measurable error in the predicted $k_{\text{eff}}$ (multiplication factor) of a pressurized water reactor. The nuclear data community tracks these sensitivities through formal uncertainty quantification (nuclear data covariances), and the resulting nuclear data uncertainties are now one of the dominant sources of uncertainty in advanced reactor design.

Discussion Questions

1. The 6.67 eV resonance has $\Gamma_n / \Gamma_\gamma \approx 0.065$. What does this ratio tell you about the competition between neutron re-emission and radiative capture at this resonance?

2. If you could magically remove all the ${}^{238}$U resonances, what would happen to the resonance escape probability? Would this make reactor design easier or harder? What other consequences would follow?

3. The negative Doppler coefficient depends on self-shielding. In a perfectly homogeneous fuel-moderator mixture (no self-shielding), would the Doppler coefficient still be negative? Explain.

4. The unresolved resonance region (above $\sim 4$ keV) requires statistical methods rather than individual resonance parameters. What physical quantity sets the boundary between the resolved and unresolved regions?