Hot-Exciton Organic Glasses and the Light Lost from Low-Energy Proton Recoils

When a slow proton or carbon nucleus recoils inside an organic scintillator, it deposits its energy along a short, dense track. Most of that energy becomes spin-triplet excitations, which are dark: they do not emit light quickly. Packed closely together, they destroy each other and the singlets nearby. The result is a low "quenching factor", the fraction of light a nuclear recoil gives relative to an electron of the same energy. This loss is what keeps hydrogen-rich detectors from seeing keV-scale nuclear recoils. That matters for dark matter near 1 GeV. The kinematic estimate $E_R \approx 2\mu^2 v^2/M_N$ (reduced mass $\mu$, galactic velocity $v \sim 10^{-3}c$, nuclear mass $M_N$) gives roughly 0.5 keV on hydrogen for a 1 GeV particle, but only about 4 eV on xenon for a 0.5 GeV particle. Hydrogen matches light dark matter far better than xenon or germanium. The recoil it produces is simply too dim to see.

A separate community has spent a decade designing organic emitters that turn triplets into light within nanoseconds. These are hybridized local and charge-transfer (HLCT) molecules. Their lowest excited state mixes a bright, compact "local" excitation with a "charge-transfer" excitation, in which an electron moves from a donor part of the molecule to an acceptor part. They harvest triplets through a "hot-exciton" channel: a high-lying triplet crosses back to a singlet before it relaxes to the bottom of the triplet ladder. This proposal asks a question nobody has tested. Does that fast upper-level triplet-to-singlet conversion drain the triplet pile-up in a dense recoil track, and so raise the quenching factor for low-energy proton recoils? And does density functional theory (DFT, the standard method for computing a molecule's electronic structure from its electron density) predict which emitters will do it?

Core Claim

In low-energy proton and carbon recoil tracks, the light lost to exciton-exciton annihilation is controlled by how long triplets survive. The hot-exciton channel caps that lifetime at the nanosecond scale by intercepting triplets before they reach the lowest triplet state. We therefore predict three things. First, HLCT glass scintillators should show a higher proton quenching factor at 1–100 keV recoil energy than energy-matched emitters that lack the hot-exciton channel. Second, the size of that gain should track a DFT-computed quantity: the energy gap and spin-orbit coupling between the upper triplet and the nearest singlet. Third, the same mechanism should weaken pulse-shape discrimination, the standard trick for telling neutrons from gammas. That trick relies on the delayed light that long-lived triplets produce when they annihilate. Both the gain and its cost can be measured, and measuring both together is what would be new.

What Exists and What It Leaves Open

The closest work supplies the materials but tests only X-rays. Color-Tunable Hot-Exciton Organic Glassy Supramolecular Scintillators Enabled by Host-Guest Co-Melting melts a brominated host (TPABr) with hot-exciton guest emitters (DTPA2F, TPE4Br, BTHDMF) and quenches the melt into transparent glasses. Its best glass, G3, reaches a relative light yield of 33,763 photons per MeV, 4.22 times the commercial scintillator bismuth germanate (BGO), with a 1.69 ns lifetime and no afterglow. That paper motivates this proposal. It never measures nuclear recoils, a quenching factor, or pulse-shape discrimination, and it does no electronic-structure modeling of why triplet harvesting survives in the solid.

The mechanism comes from OLED research, in which electrical injection creates the 25:75 singlet-to-triplet mix. Highly efficient non-doped blue fluorescent OLEDs with low efficiency roll-off based on hybridized local and charge transfer excited state emitters uses DFT to find a T12→S3 crossing channel in the emitter TPAAnPI and reports 91% exciton utilization. Its low efficiency roll-off at high brightness (15% at 1000 cd m⁻²) is indirect evidence that hot excitons tolerate high excitation density. High density is exactly the regime of a recoil track. Neither paper connects the calculation to a measured density-dependent loss, which is what we do.

Two computational resources constrain the DFT component. The VERDE database supplies M06 ground- and excited-state data for more than 1,500 organic molecules, but only for S0, S1 and T1. Those are not the upper triplets we need. The Pariser–Parr–Pople screening work gets singlet-triplet gaps in milliseconds, with recall 0.56 and specificity 0.97 on 256 azaphenalenes against reference calculations. It is weak on oscillator strengths and non-planar σ–π systems, and twisted donor-acceptor HLCT emitters are precisely such systems. We therefore do not rely on it for upper-state couplings. We use it only as a cheap pre-filter.

On the detection side, the EJ-301 unfolding study shows that conventional liquid scintillators have strongly nonlinear, quenched proton response below 1–2 MeV. Its calibrated Geant4 response-matrix method is the one we reuse. The existing Migdal-effect calculations, the bremsstrahlung-like dielectric treatment and the phonon-mediated field theory, work only for silicon and diamond. We deliberately do not extend them to organics here. For an organic detector, the dominant unknown is how much light a recoil makes, not the rate of a secondary atomic process.

Research Program

The first component is a DFT map of the excited-state ladder. For about 15 candidate emitters we compute vertical and relaxed singlet and triplet energies up to roughly T12, plus the spin-orbit coupling between each upper triplet and its nearest singlet. The candidates are the three hot-exciton guests of Ye et al., TPAAnPI, and designed variants with different donor-acceptor twist angles. We use optimally tuned range-separated hybrid functionals, because conventional functionals artificially lower charge-transfer states. As a check on functional dependence, we repeat the calculations with M06, to connect to VERDE. As a sanity check, we reproduce the T12→S3 assignment for TPAAnPI. From these results we define a single descriptor per molecule: the fraction of triplet population predicted to cross to singlets before relaxing to T1. It comes from Fermi-golden-rule estimates of the h-RISC rate $k_{\mathrm{hRISC}}$, set against the internal conversion rate $k_{\mathrm{IC}}$ down the triplet ladder:

$$\eta_{\mathrm{hot}} = \frac{k_{\mathrm{hRISC}}}{k_{\mathrm{hRISC}} + k_{\mathrm{IC}}}.$$

The second component is a track-kinetics model. It turns $\eta_{\mathrm{hot}}$ into a predicted quenching factor as a function of stopping power. We extend the empirical Birks picture, in which light per unit path is $dL/dx = S\,(dE/dx)/(1 + kB\,dE/dx)$ with light yield $S$ and fitted quenching constant $kB$. The extension replaces the fitted constant with explicit singlet-triplet and triplet-triplet annihilation terms. The triplet density in these terms is depleted at rate $\eta_{\mathrm{hot}}$ at formation. The model gives two falsifiable outputs: the quenching factor as a function of proton energy, and the delayed-light fraction that pulse-shape discrimination depends on.

The third component is materials. We make co-melted glasses following Ye et al. We also make a bromine-free host variant. It removes heavy recoiling bromine nuclei from the neutron response and separates heavy-atom effects on intersystem crossing from the hot-exciton effect. Each HLCT guest gets a partner guest with matched emission color and prompt light yield but no charge-transfer mixing, and therefore no upper-level crossing channel.

Baselines

The naive reference point is the pristine guest glass and the pure host glass of Ye et al. These show what the host-guest supramolecular rigidification contributes on its own. Ye et al. report a 41% X-ray radioluminescence gain from it, and we need to know whether that gain carries over to recoils. These glasses are measured in the same holder, with the same photomultiplier and the same calibration as every other sample.

The strongest existing competitor for fast neutron and proton-recoil detection is the liquid scintillator EJ-301. It is run with the calibrated response-matrix procedure of the EJ-301 unfolding study. That study reports agreement within 1.42% of the ISO 8529-1 ²⁵²Cf reference for a 2-inch detector with the MLEM unfolding algorithm, down to a 0.35 MeV threshold. EJ-301 is the incumbent because it has mature pulse-shape discrimination and a well-characterized response. We cast our glasses at matched volume and give them the same Compton-edge calibration with ¹³⁷Cs and ²²Na. For light yield under ionizing radiation, the strongest number in the file is G3's 33,763 photons per MeV under X-rays Ye et al.. For low-energy proton-recoil quenching factors in any of these materials, no reported result was found on Lacuna. We will measure EJ-301's quenching factor ourselves in the same setup rather than borrow a value.

The ablation that isolates the key ingredient is the matched non-HLCT partner guest in the identical host. We also run a small-energy-gap thermally activated delayed fluorescence (TADF) emitter in the identical host. TADF also harvests triplets, but only from T1, on microsecond timescales. It therefore tests whether any triplet harvesting suffices or whether the fast upper-level route is what matters. Beating the baselines means two things. First, a quenching factor for 5–50 keV protons significantly above both EJ-301 and the matched partner, at equal electron-equivalent light yield. Second, a gain that ranks with $\eta_{\mathrm{hot}}$ across emitters and does not rank with prompt light yield.

Why not just…?

Why not just fit Birks' law to measured data and skip the DFT? For a single material that is the right approach, and the measurement is the core of this proposal regardless. The DFT earns its place only if the computed $\eta_{\mathrm{hot}}$ predicts the ranking of measured quenching-factor gains across roughly 15 emitters before they are measured. If the ranking correlation across emitters is weak, we report that. The practical conclusion would then be that recoil response in organic glasses is set by packing and rigidity rather than by molecular upper-state alignment, and screening should move to cheap empirical proxies.

Why not just use TADF emitters, which also harvest triplets? TADF emitters harvest triplets only after they relax to T1, over microseconds. In a dense track that leaves ample time for triplet-triplet annihilation. The TADF arm of the ablation tests this directly. If it matches the HLCT glasses, the hot-exciton hypothesis is wrong and the simpler, better-studied emitter class wins.

Why not just use semiconductors, xenon time projection chambers, or superconducting nanowires, which already reach lower thresholds? For dark matter below about 100 MeV they are better, and this proposal does not claim otherwise. QROCODILE reaches a 0.11 eV threshold, but with microgram-to-milligram target mass. PandaX-4T goes down to about 0.33 keV for nuclear recoils, but on a heavy nucleus and with a floor set by electrode micro-discharges. The niche for organics is kilogram-scale, room-temperature hydrogen targets for dark matter near 1 GeV, where proton recoils sit at 0.5–few keV. That niche exists only if the quenching factor at those energies can be raised. Neutron metrology benefits even if dark matter reach does not materialize.

Why not just compute the Migdal effect in organic glasses with DFT dielectric methods? It is a real gap, and the dielectric approach of Liang et al. could in principle be applied to molecular solids. Any Migdal signal in an organic detector would still pass through the same light-production physics measured here. Doing the quenching measurement first is the prerequisite.

Proposed Experiments and Decisive Checks

The central measurement is a tagged neutron-scattering quenching study. We scatter neutrons, starting from a ²⁵²Cf source, off each glass. The scattered neutrons are detected in EJ-301 backing detectors at fixed angles, which fixes the proton recoil energy kinematically. We scan proton recoil energies from about 5 keV to 1 MeV. We record the quenching factor versus energy, the pulse-shape discrimination figure of merit, and the prompt and delayed light fractions, with Geant4 response matrices built as in the EJ-301 study.

Three results decide the question. The hot-exciton hypothesis survives only if all three hold together. First, HLCT glasses beat the matched non-HLCT partners and the TADF arm in quenching factor at recoil energies below 50 keV. Second, the gain correlates with computed $\eta_{\mathrm{hot}}$ across the emitter set. Third, the predicted delayed-light suppression appears, and it degrades pulse-shape discrimination by the amount the kinetics model says.

A gain without the correlation means rigidity, not upper-state physics, is doing the work. The correlation without a usable pulse-shape discrimination figure of merit tells designers what they must trade. A secondary check is time-resolved photoluminescence at increasing laser fluence. It verifies that the annihilation rates fitted from the recoil data agree with those seen optically. This guards against the kinetics model absorbing unexplained losses into free parameters.