---
title: In Situ Photoneutron Calibration
url: https://www.emergentmind.com/topics/in-situ-photoneutron-calibration
type: topic
---

# In Situ Photoneutron Calibration

In situ photoneutron calibration is the use of a deployed $\gamma$-emitting radioisotope and beryllium converter to generate low-energy neutrons inside, or immediately adjacent to, the final detector configuration, so that nuclear-recoil response is measured under operating geometry, shielding, and background conditions. In the implementations described for dark-matter and coherent elastic neutrino-nucleus scattering experiments, the underlying reaction is typically ${}^{9}{\rm Be}(\gamma,n){}^{8}{\rm Be}$, driven by sources such as ${}^{124}$Sb or ${}^{88}$Y, and the calibration target is the detector medium itself or a well-defined external beamline coupled to a backing detector [2302.03869]. The technique is used to probe sub-keV to few-keV nuclear recoils, to extract quenching factors or light and charge yields, and to validate transport and response models in the experiment’s installed configuration [2412.10451].

## 1. Nuclear basis and source classes

The central photodisintegration process is
\[
{}^{9}{\rm Be} + \gamma \;\longrightarrow\; {}^{8}{\rm Be} + n .
\]
In the SBC calibration plan, the neutron binding energy in ${}^{9}$Be is given as $Q_{\rm Be}=1.665$ MeV, while the XENONnT and Y-88/Be descriptions quote $Q_{\rm th}=1.667$ MeV or $S_n=1.667$ MeV [2511.19817; 2412.10451; 1303.2686]. In the center-of-mass frame the neutrons are essentially monoenergetic; in the laboratory frame there is a small Doppler and recoil broadening, and the SBC plan states that simulated line shapes have $\sim10\%$ width [2511.19817].

Two source classes dominate the reported implementations. The first is ${}^{124}$Sb-Be. For the portable source, the relevant $\gamma$ lines are $1691.0$ keV with $I_\gamma = 47.79\%$ and $\sigma(\gamma,n)=1.41$ mb, yielding dominant $24$ keV neutrons, and $2090.9$ keV with $I_\gamma = 5.51\%$ and $\sigma(\gamma,n)=0.412$ mb, yielding sub-dominant $379$ keV neutrons; the normalized neutron-production rate per ${}^{124}$Sb decay is $\simeq10^{-4}$ n/decay for the $24$ keV branch [2302.03869]. The second is ${}^{88}$Y-Be. In XENONnT, $1.836$ MeV $\gamma$ rays from ${}^{88}$Y decay produce quasi-monoenergetic $152$ keV neutrons on ${}^{9}$Be, with $\sigma_{\gamma n}(1.836\,{\rm MeV}) \simeq 0.65$ mb and a Monte Carlo photodisintegration probability per $\gamma$ of $p_{\gamma,n}\simeq2\times10^{-5}$ [2412.10451]. In LZ, the same source also includes the $2.7340$ MeV $\gamma$ line, so the source produces both $152$ keV and $950$ keV neutrons, with $\sigma(\gamma,n)=0.651\pm0.007$ mb and $0.592\pm0.038$ mb, respectively [2509.16281]. Collar’s earlier Y-88/Be study frames the same source class as a compact method for generating a clean, quasi-monochromatic $152$ keV neutron field for low-energy recoil measurements in NaI[Tl], bubble chambers, and noble liquids [1303.2686].

The generic neutron-yield expression used in the SBC plan is
\[
N_n
\;=\;
\Delta t\,
\int dE_\gamma\;\Phi_\gamma(E_\gamma)\;\;N_T\;\;\sigma_{\gamma n}(E_\gamma),
\]
with the monoenergetic approximation
\[
N_n
\;=\;
\Phi_\gamma\;\times N_T\;\times \sigma_{\gamma n}(E_\gamma)\;\times \Delta t .
\]
This formalism emphasizes that in situ photoneutron calibration is fundamentally a source-strength, cross-section, and geometry problem rather than a detector-only problem [2511.19817].

## 2. Shielding, transport, and deployment geometry

A defining feature of in situ photoneutron calibration is that the source is integrated into the installed apparatus rather than into a separate beam test. The portable ${}^{124}$Sb-${}^{9}$Be source realizes this explicitly as a transportable neutron beam system. Its central iron filter rod has diameter $4.47$ cm and length $20.3$ cm, and the source assembly includes successive layers of W, Pb, stainless steel, and Al in a T-slot frame, together with a detachable $2.54$ cm-thick borated-PE sleeve and a circular beam aperture of approximately $4.5$ cm aligned to the filter axis [2302.03869]. The design relies on the coincidence between the $24$ keV neutron energy and the low interaction cross-section with iron: for $E_n=24$ keV, $\sigma_{\rm tot}\simeq0.04$ barn gives a neutron mean free path of $\lambda_n\simeq29$ cm in iron, while for $1.7$ MeV $\gamma$ rays in iron, $\mu_\gamma\simeq0.7$ cm$^{-1}$ implies attenuation of approximately $10^{-6}$ over $\Delta x=20$ cm [2302.03869]. The resulting system has total shield mass $\simeq250$ kg, fits in a $30$ gal Type-A drum for road transport, and is described as suitable for existing beam ports or water-filled calibration pits [2302.03869].

The XENONnT deployment uses a different geometry but the same in situ principle. The ${}^{88}$Y disc, with activity $A_0 = 316$ kBq, is contained in a stainless-steel capsule and sandwiched between two $50$ mm $\times$ $25.4$ mm Be cylinders inside a tungsten shielding box of outer dimensions $16$ cm $\times$ $16$ cm $\times$ $16$ cm [2412.10451]. The assembly is lowered until the front face of the tungsten box is aligned with the TPC mid-cathode plane, approximately $10$ cm from the cryostat wall, and an air-filled stainless-steel water-displacement box fills the gap to reduce neutron energy degradation [2412.10451].

LZ likewise deploys a stacked YBe source inside a $20$ cm $\times$ $20$ cm tungsten block, lowered through a dedicated cut-out in the top acrylic Outer Detector into the water tank, where it rests above the outer cryostat with the ${}^{88}$Y located $86.7$ cm above the TPC gate electrode [2509.16281]. The description stresses that the calibration is fully in situ: no disassembly of the TPC is required, and source exchange is done by removing the Gd-doped liquid-scintillator plug used in normal running and inserting the YBe assembly [2509.16281].

The SBC plan makes explicit that in situ deployment need not imply collimation. There, the source-plus-Be assembly is lowered inside a $4.5$ cm ID stainless-steel calibration tube whose bottom sits $12.1$ cm above the top of the argon, and a neutron traverses $0.6$ cm stainless-steel source tube cap, $0.95$ cm stainless-steel pressure vessel wall, $\sim10$ cm liquid CF$_4$ hydraulic fluid, and $0.51$ cm fused-silica outer vessel before reaching the argon [2511.19817]. No additional moderation or collimation is used; the tube geometry defines the angular acceptance, and scattering in CF$_4$/silica slightly softens the spectrum before it reaches the argon [2511.19817].

## 3. Recoil generation and detector observables

The calibration observable is the detector response to elastic neutron scattering. The SBC plan gives the recoil-energy relation for a neutron of energy $E_n$ scattering on ${}^{40}$Ar at laboratory angle $\theta$ as
\[
E_R(\theta)
\;=\;
\frac{2\,m_n\,m_A}{(m_n + m_A)^2}\;E_n\;(1-\cos\theta),
\]
with the recoil spectrum obtained from
\[
\frac{dR}{dE_R}
\;=\;
N_{\rm Ar}\,\int dE_n\;\phi_n(E_n)\;\frac{d\sigma}{dE_R}(E_n,E_R) .
\]
This formalism makes clear that in situ photoneutron calibration links source transport to the recoil spectrum through the neutron flux spectrum at the active medium [2511.19817].

For LXe TPCs, the observables are the prompt scintillation signal $S1$ and the delayed proportional scintillation signal $S2$. In LZ, $S1$ is collected by $494$ two-inch-diameter Hamamatsu R11410 PMTs, while $S2$ is produced by electrons drifted approximately $50$ cm at $193$ V/cm from liquid into gas and extracted with an extraction field of approximately $5$ kV/cm; the instrumental gains in the YBe setting are $g_1=0.114\pm0.003$ phd/photon and $g_2=34.9\pm0.6$ phd/electron [2509.16281]. XENONnT similarly uses corrected $cS1$ and $cS2$ distributions, with an analysis region requiring $2 \le n_{\rm ph} \le 10$ observed photons and $120 \le cS2 \le 900$ PE [2412.10451].

For bubble chambers, the observable is nucleation rather than a continuous energy estimator. In the SBC plan, three cameras provide stereo reconstruction, eight piezoelectric transducers record the time and three-dimensional position of each bubble, and a pressure transducer sees the rapid drop in pressure when a bubble forms [2511.19817]. The scintillation channel, using $32$ SiPMs sensing $128$ nm LAr scintillation with wavelength shifted by $10$ ppm xenon doping, is used only to veto high-energy backgrounds or to tag neutron captures, not for bubble timing [2511.19817]. Each expansion cycle superheats the top $10$ kg of LAr, the chamber recompresses after bubble formation with approximately $30$ s dead time, and no direct energy measurement is made in each bubble event; instead bubble counts are compared with simulation-predicted recoil spectra [2511.19817]. This is a salient methodological distinction: in situ photoneutron calibration does not always mean event-by-event recoil-energy reconstruction.

The older NaI[Tl] implementation uses yet another observable, the quenching factor $Q(E_r)=E_{ee}/E_r$, extracted by subtracting a gamma-only configuration from a neutron-plus-gamma configuration and fitting the residual spectrum with MCNP-PoliMi [1303.2686]. That approach shows that photoneutron calibration can serve both threshold-setting and response-function measurement.

## 4. Measurement protocols, simulation, and inference

In situ photoneutron calibration typically combines source characterization, detailed transport simulation, and detector-response inference. The portable ${}^{124}$Sb-${}^{9}$Be source was characterized with a hydrogen-gas proportional counter and a NaI(Tl) detector [2302.03869]. The HGPC used an LND 27044 sphere with diameter $4.98$ cm, H$_2$ at $3.04$ bar and $21\,^\circ$C, an energy threshold of approximately $6$ keV, resolution $\sigma_E/E\approx10\%$, and neutron/electron pulse-shape discrimination via $10$–$90\%$ rise time with neutron selection $\tau_{\rm rise}<2.5\,\mu$s; gamma background was measured by inserting a borated-PE plug and subtracted bin by bin [2302.03869]. The NaI(Tl) detector was a $5.08$ cm $\times$ $5.08$ cm cylinder calibrated with ${}^{241}$Am, ${}^{57}$Co, ${}^{133}$Ba, ${}^{137}$Cs, ${}^{22}$Na, and ${}^{228}$Th, and the full-energy peaks at $1691$ keV and $2091$ keV matched simulation within $6\%$ [2302.03869].

The SBC plan formalizes the inference problem in terms of simulated recoil spectra $W_i(E_R)$ and a nucleation-efficiency function $\epsilon(E_R)$ parameterized piecewise with five nodes at efficiencies $0\%$, $20\%$, $50\%$, $80\%$, and $100\%$ [2511.19817]. For each source $i$, the expected bubble count is written as
\[
b_i \;=\; t_{\rm cal}\,\int_0^\infty W_i(E_R)\,\epsilon(E_R)\,dE_R\;+\;b_i^{\rm bg},
\]
and the total event count for photoneutrons as
\[
v_i \;=\; t_{\rm cal}\,\int_0^\infty W_i(E_R)\bigl[1 - (1-\epsilon(E_R))^{n(E_R)}\bigr]\,dE_R
\;+\;v_i^{\rm bg}.
\]
A global $\chi^2$ or MCMC is then used to fit the five node energies and nuisance parameters for source strengths and simulation uncertainties [2511.19817].

XENONnT adopts a forward model in Appletree, combining the NEST v2 yield model with a detector-response model, and fits the corrected $cS1$ and $cS2$ distributions with a binned Poisson likelihood
\[
\mathcal{L}(\theta) = \prod_{i=1}^{N_{\rm bins}} e^{-\mu_i(\theta)}[\mu_i(\theta)]^{n_i}/n_i!,
\qquad
\chi^2 = -2\ln\mathcal{L}(\theta) + {\rm const.}
\]
Twenty-one parameters are sampled via an affine-invariant MCMC using emcee, and the accidental-coincidence spectrum derived from the ${}^{88}$Y-PVC gamma-only run is included with a floating normalization constrained by a Gaussian prior of $55\pm12$ [2412.10451].

LZ uses NEST to convert recoil energy into expected $S1$ and $S2$, but in its first in situ YBe analysis it restricts the fit to the $152$ keV recoil population and simultaneously fits the $S1c$ and $S2c$ distributions to extract an overall $Q_y$ scale factor while leaving all other NEST parameters fixed [2509.16281]. Backgrounds from the $950$ keV neutron mode and accidentals are included as fixed-shape components, and delayed Outer Detector tagging confirms neutron capture with an $(89\pm3)\%$ veto tag efficiency [2509.16281].

Across these implementations, Geant4 appears as the standard neutron-transport engine. The portable source uses the Geant4 Shielding physics list for beam-flux prediction [2302.03869], while the SBC plan calls for a Geant4 model of the full geometry with $10^6$–$10^7$ neutrons per source [2511.19817]. A plausible implication is that in situ photoneutron calibration is as much a validation of the transport model as of the detector response model.

## 5. Representative implementations and quantitative outcomes

The portable ${}^{124}$Sb-${}^{9}$Be source is designed as a dedicated low-energy recoil calibrator. For a $1$ GBq ${}^{124}$Sb source, Geant4 predicts $\Phi_n(E>1\,{\rm keV}) \simeq 6.2$ n/cm$^2$/s at the beam exit, of which $40\%$ lies in $20$–$25$ keV, giving $\Phi_n(20$–$25\,{\rm keV}) \simeq 2.48$ n/cm$^2$/s; the measurement using the HGPC-tagged recoil endpoint yields $\Phi_n(20$–$25\,{\rm keV}) = 5.36\pm0.20$ n/cm$^2$/s [2302.03869]. The corresponding gamma flux above $100$ keV is $\Phi_\gamma = 213\pm6$ $\gamma$/cm$^2$/s, the neutron-to-gamma ratio at the exit is approximately $2.5\times10^{-2}$, and contamination from $379$ keV neutrons is stated to be $2\%$ of the total [2302.03869]. The same work reports that an Eljen-301 liquid scintillator backing detector is sensitive to incident neutrons from $8$ to $17$ keV, with tagging efficiency rising from approximately $2\%$ at $8$ keV to approximately $16\%$ at $17$ keV, and gives an expected recoil precision at the $100$ eV$_{\rm nr}$ scale of $\delta E/E\lesssim5\%$ per event in a typical dark-matter or CEnuNS detector with ring-tagging geometry [2302.03869].

XENONnT used a ${}^{88}$YBe source emitting $152$ keV neutrons to calibrate liquid-xenon light and charge yields for the first time in situ [2412.10451]. After data selection, $474$ events were accumulated from $183$ hours of exposure, while the expected background was $55\pm12$ accidental-coincidence events estimated from a dedicated $152$ hour background calibration run with a Yttrium-PVC gamma-only source and data-driven modeling [2412.10451]. The analysis extracted $L_y(E_{\rm NR})$ and $Q_y(E_{\rm NR})$ between $0.5$ keV$_{\rm NR}$ and $5.0$ keV$_{\rm NR}$ at $23$ V/cm, with the lowest observable energy set by $\Delta\chi^2(E_{\rm thr})=1$ to $E_{\rm thr}\simeq0.5$ keV$_{\rm NR}$ [2412.10451].

LZ reports its first in situ photoneutron calibration using a custom ${}^{88}$Y-${}^{9}$Be source [2509.16281]. For the initial $0.339$ MBq source, the absolute neutron rates at production are $R_{152}=(34\pm1)$ n/s and $R_{950}=(0.22\pm0.03)$ n/s [2509.16281]. Elastic scattering of the $152$ keV component gives a maximal recoil energy of $4.6$ keV$_{\rm nr}$ in xenon, and the unfolded single-scatter spectrum from this component is described as quasi-flat up to $4.6$ keV$_{\rm nr}$ with mean $3.1$ keV$_{\rm nr}$ [2509.16281]. The best-fit neutron rate at the source is $(29\pm2_{\rm stat}\pm7_{\rm sys})$ n/s, in excellent agreement with the analytic $34\pm1$ n/s; the best-fit $S1$ and $S2$ spectra match simulations with $p$-values of $0.97$ and $0.06$; and the measured $Q_y$ at $E_r\approx3.1$ keV$_{\rm nr}$ is $12\%$ below the default NEST central value but within its $\pm1\sigma$ band [2509.16281].

The SBC Collaboration’s calibration plan extends the method to a $10$ kg liquid-argon bubble chamber with $100$ eV target threshold [2511.19817]. It proposes using, for example, $5\,\mu$Ci ${}^{207}$Bi-Be and $0.5\,\mu$Ci ${}^{124}$Sb-Be sources to give approximately $50$ bubbles/hour above threshold, with $t_{\rm cal}\sim100$ h of live data at fixed thermodynamic conditions for each source [2511.19817]. Under standard systematics, the final threshold uncertainty inferred from the width of the $T_{50\%}$ pull distribution over $\mathcal{O}(50)$ mock datasets is $\sim20\%$, and under improved systematics it is $\sim5\%$ [2511.19817].

The earlier Y-88/Be study by Collar demonstrates the same logic in NaI[Tl], reporting a measured sodium quenching factor below $24$ keV$_{\rm nr}$ that falls from $\sim0.25$ at $24$ keV$_{\rm nr}$ to $\lesssim0.10$ below $\sim10$ keV$_{\rm nr}$ [1303.2686]. The abstract states that this is considerably smaller than the $30\%$ typically adopted in interpretations of the DAMA/LIBRA dark-matter experiment and results in a marked increase of its tension with other searches under the standard set of phenomenological assumptions [1303.2686].

## 6. Uncertainties, limitations, and points of interpretation

The uncertainty budget in in situ photoneutron calibration is distributed across source strength, photoneutron production, neutron transport, geometry, and detector-response inference. The SBC plan lists source-strength calibration at $\pm5\%$ near-term and $\pm1\%$ in future, photoneutron cross-section uncertainty at $\pm20\%$ background and $\pm5\%$ goal, Geant4 modeling of neutron transport at $\pm10\%$–$20\%$ common mode, geometry tolerances giving flux shifts of $\pm2\%$ for tube-position changes of a few mm, residual rock and muon neutron backgrounds stable to $\pm10\%$, and fit-model systematics from the choice of piecewise nodes [2511.19817]. These are combined via pull terms in the $\chi^2$ and by rerunning fits under extreme systematic shifts [2511.19817].

Gamma backgrounds are a recurrent practical limitation rather than a negligible detail. In the portable ${}^{124}$Sb-Be system, dedicated gamma characterization is mandatory, and the measured beam still contains $213\pm6$ gammas/cm$^2$/s above $100$ keV per $1$ GBq source activity at the exit [2302.03869]. XENONnT treats accidental-coincidence backgrounds explicitly through a gamma-only ${}^{88}$Y-PVC run and a data-driven model [2412.10451]. Collar’s Y-88/Be description states that after $15$–$20$ cm of lead shielding, the gamma-induced electron-recoil rate under the nuclear-recoil search window is less than $10\%$ of the neutron-induced signal, but emphasizes the need for Be-versus-dummy subtraction and control of laboratory-origin neutrons [1303.2686].

A common misconception is that the neutron field remains perfectly monoenergetic once deployed. The source reactions are monoenergetic or quasi-monoenergetic at production, but the flux entering the active volume is often softened or broadened by detector materials. The SBC plan states that neutron scattering in CF$_4$/silica slightly softens the spectrum before it reaches the argon [2511.19817]. LZ reports that the entering-neutron spectrum falls smoothly from $E_n$ down to zero up to each endpoint because most neutrons are moderated by the Ti cryostat and SS/PTFE structures before entering the active LXe [2509.16281]. XENONnT uses an air-filled gap specifically to reduce neutron energy degradation [2412.10451]. This suggests that “in situ” is not merely a convenience of placement; it is part of the calibration observable because the installed materials reshape the spectrum.

Another misconception is that in situ photoneutron calibration always yields direct per-event recoil energies. The portable $24$ keV beam paired with a backing detector is expressly designed for neutron scattering calibration experiments with percent-level recoil tagging [2302.03869]. By contrast, the SBC bubble-chamber calibration counts bubbles and fits $\epsilon(E_R)$ statistically, without direct event-by-event recoil measurement [2511.19817]. LXe TPC analyses lie between these extremes, inferring yield models from distributions in corrected scintillation and ionization observables rather than from uniquely determined recoil energies for every event [2412.10451; 2509.16281].

Taken together, the reported implementations show a mature calibration modality for low-threshold dark-matter and CEnuNS detectors. They establish that source physics, shielding design, transport simulation, and response inference must be treated as a coupled system, and they show that the principal value of the method is precisely that it measures nuclear-recoil response in the experiment’s operating configuration rather than in a detached proxy setup.

Source: https://www.emergentmind.com/topics/in-situ-photoneutron-calibration