The term electroluminescence (EL) is often used, among others, to refer to the detection principle introduced by Conde and Policarpo in the 60s (Alves and Policarpio, 1967; Conde and Policarpo, 1967; Conde et al., 1968). In its simplest description, within this context, it might be seen as the mechanism by which to enhance the ionization response of a gaseous detector, converting ionization electrons (difficult to count) into photons (easy to count) by means of an electric field. Since the accelerated swarm of electrons experiences different energy thresholds for excitation and ionization, the net result is the appearance of an electric field regime where it is possible to have the former without the latter. In noble gases, specifically, the high VUV-scintillation probability per excited state (Serra et al., 2015; Azevedo et al., 2018) results in strong scintillation yields (Y) of up to easily 1,000s of ph/e, yet in the absence of avalanche multiplication (Monteiro et al., 2008; Freitas et al., 2010). These are the conditions under which EL exhibits its most remarkable features and finds widespread use: in the avalanche-free regime, as noble gases possess no inelastic degrees of freedom other than excitations, nearly all energy gained by electrons in the electric field (qV) turns into scintillation photons, leading to an approximately linear response (Oliveira et al., 2011). This can be expressed handily as: Y ≃ K V − V t h (1) Y z / N ≃ K E ∗ − E t h ∗ . (2)

The second expression conveys the fact that, when the density-reduced electric field (E* = E/N, a measure of the characteristic electron energy) is above a threshold value needed to excite the medium ( E t h * ) , any additional energy gained in the field goes into new excitations. In the above equations K is a proportionality constant, Y refers to the scintillation yield, Y _z to the yield per unit length and N is the number of atoms per unit volume. The high efficiency of the EL process has a second consequence, that represents perhaps one of its best known properties: the relative variance of the electroluminescence signal ( Q = ( σ Y / Y ) 2 ) is very small. Its contribution to the calorimetric response of a gaseous detector propagates into the energy resolution (FWHM) as: R = 2.355 F + Q + 1 n p e 1 + σ G 2 G 2 1 n e . (3)

In the above expression F represents the Fano factor of the medium (0.1–0.2 in noble gases (Aprile et al., 2006)), n _pe is the number of detected photons (‘photoelectrons’) per electron, and σ _G /G refers to the relative width of the single-photon distribution function of the photosensor. The last term involves the number of ionization electrons, n _e = ɛ/W _I , with ɛ being the energy deposited in the medium and W _I the average energy needed to release an electron (Blum et al., 2008). The term Q is generally so small (Escada et al., 2011; Oliveira et al., 2011) that it has eluded experimental determination except in the presence of electronegative gases such as CO₂ (Henriques et al., 2017). Due to the smallness of Q, as soon as the number of detected photons per primary electron is of the order of n _pe = 10 or more, the energy resolution will approach the intrinsic limit of the gas medium (Nygren, 2009).

The aforementioned characteristics make electroluminescence, nowadays, the workhorse of many experiments based on Time Projection Chambers (TPCs) in the field of Rare Event Searches (Irastorza et al., 2013), such as XENON (Aprile et al., 2023), LZ (Aalbers et al., 2023), PandaX (PandaX Collaboration, 2023), and DarkSide (Aalseth et al., 2020) (aimed at direct WIMP Dark Matter detection) or NEXT (Novella et al., 2022) (aimed at measuring ββ0ν decay). Its use is foreseen in upcoming experiments such as RED (Akimov et al., 2022) and GANESS (Monrabal, 2021) (aimed at precision studies of coherent neutrino scattering), and it can be found, too, in applied research for instance in proposals for Compton cameras (Azevedo et al., 2013) and Compton dispersion at next-generation Light Sources (Saá-Hernández et al., 2021). In such experiments, the single-electron detection capabilities of electroluminescence are essential to extend WIMP sensitivity down to very low masses (e.g., (Agnes et al., 2023)), while near-Fano energy resolution is required for precise calorimetry in ββ0ν searches, thus avoiding contamination from ββ2ν decays and natural radioactivity (Novella et al., 2023). High-rate environments like (Azevedo et al., 2013; Saá-Hernández et al., 2021) benefit from the absence of ion feedback and subsequent high rate capability stemming from the avalanche-free nature of the electroluminescence process.

For all their beauty, the electroluminescence process in presently running experiments is invariably implemented based on early designs relying on wires or meshes, and suffer from the inevitable scaling-up limitations due to electrostatic sagging and deformation (Rogers et al., 2018) and, chiefly, defects (Discussions at the Conference, 2023). During the operation of the NEXT-NEW detector, for instance, the pressure-reduced electric field was 1.2 kV/cm/bar at 10 bar, a mere 0.5 kV/cm/bar above the EL threshold and 3 times below the values achieved in the present work and, generally, on small R&D setups. Thus, when possible, the commissioning of the EL-region becomes a separate endeavour, requiring purposely-built chambers several square meter in size, essentially as large as the final operating conditions demand (Müller, 2022). Given that experiments keep increasing in size to achieve sensitivity, it seems very unlikely that the situation will improve except if new production techniques are introduced.

Lately, over the last 5–10 years, there has been interesting progress on the development of optically-enhanced structures, and we follow this lead. Transparent substrates like glass, for instance, have been successfully carved into gaseous electron multipliers (GEMs), be it by photolitography (Takahashi et al., 2013) or sand-blasting (Lowe et al., 2021). They are sturdy and immune to sagging or deformation, and can be easily tiled to cover large areas (Lowe et al., 2023). At the thicknesses and hole sizes needed for efficient electroluminescence (several mm), structures based on the highly VUV-reflective Teflon (Ban et al., 2017) or relying on the transparency of PMMA (González-Díaz et al., 2020) have been manufactured by CNC drilling. In order to achieve VUV-transparency, the solid wavelength-shifter polyethylene naphthalate (PEN) was proposed as a bulk material and successfully machined and operated under Xe and Ar for the first time in (Kuźniak et al., 2021). A natural continuation of the previous works was to resort to TPB coating of the (PMMA-based) GEM channels, theoretically enabling higher wavelength-shifting yields compared to PEN plates. Aiming at wide range of applications, in particular those related to Rare Event Searches, the core materials of such EL-structures are chosen to be radiopure.

Here we present a comprehensive performance study of different FAT-GEM structures under 5.9 keV X-rays, involving different fabrication procedures, different processing (adding wavelength-shifting coating or an internal reflective layer), different geometries (hole sizes), different gases (xenon and argon), and different pressures (2–10 bar). Comparisons with simulations and with a simple 2-mesh configuration are given too.

Simulations were performed according to the following steps. First, the electric field maps ensuing from an elementary amplification cell were created with Ansys (ANSYS, 2023), (Figure 7A). Field and geometry maps were then imported into Garfield++ (Schindler, 2019) and extended in the x and y axis (axes parallel to the electrode plane), building the FAT-GEM structure (Figure 7B). The position of the primary electrons was generated following an uniform x, y distribution centered at the middle of the detector, extending over a ±5 mm region and placed at 0.1 mm from the cathode plane. The impact of considering a distributed ionization source corresponding to the X-ray mean free path (around 2.7 mm/bar) was evaluated but the differences dimmed too small. In Figure 7C the path of an electron is shown: as expected, most of the collisions that produce an excited state happen inside the FAT-GEM hole.

The Garfield++ simulation provided the 3D-distribution of excited states over the entire array of holes. This was then imported into Geant4 (Agostinelli et al., 2003) and used to sample the initial positions of the scintillation photons, that were subsequently launched isotropically. ⁴ Photon ray-tracing was handled by a Geant4 model with optical properties implemented for all materials, extending the model used in Ref. (Kuźniak et al., 2021). New wavelength-dependent optical properties added to the model included the transparency of ITO (approx. 79%, based on the suppliers’ specification (Visiontek Systems Ltd., 2023)), measured reflectivity for the laminated ESR foil, and TPB properties as in (Boulay et al., 2021). This allowed us to estimate the geometrical efficiency of the experimental setup (Figure 7D). A two-mesh configuration was simulated using similar techniques, except that the mesh transparency was introduced as an effective number and the reflectivity of the aluminum cathode (Lindseth et al., 1999) was included too (yielding about a 10% increase, an effect negligible for FAT-GEMs).

Finally, photon detection efficiency (PDE) curves were considered for four illustrative sensors: Hamamatsu R7378 (hamamatsu.com, 2024) (employed in these measurements), R11410-21 (used in XENON1T (PhotoMultiplier tubes and assemblies, 2024)) PMTs, and two representative SiPMs: FBK NUV-HD-SF (Aalseth et al., 2018) and Hamamatsu VUV4 (S13370) (Gallina et al., 2019).

The best energy resolution that was measured for FAT-GEMs in this work was 17% (FWHM), at the 5.9 keV peak of ⁵⁵Fe. This represents almost a factor two increase with respect to the value expected from the Fano factor alone (De Lima et al., 1982; Henriques, 2018; Anderson et al., 1979; Biagi, 2024; Dias et al., 1991; Nygren, 2011). The measurement is not photon statistics -limited, as similar asymptotic values were obtained for different optical gains (up to a number of detected photoelectrons per electron n _pe = 10). Further, similar results were obtained both with a double-mesh and with a drift-less configuration (the latter being more immune to attachment). The deviation from the Fano limit can be qualitatively understood from the afterpulsing of this PMT model (as observed during the LED characterization): it appears with a delay of about 250 ns, and can reach about 10% of the signal; as the ionization cloud spreads over at least a few µs due to the combined effects of the photoelectron range and diffusion, afterpulsing can not be removed and its event-by-event fluctuations are not correctable.

Notwithstanding, an extrapolation of present results up to the Q _ββ -scale of ¹³⁶Xe (2.45 MeV) indicates that the achieved energy resolution is competitive with the ones reached by current experiments and other R&D efforts (Figure 13). The recent result of 0.73% at 1.836 MeV (star), reported for Teflon-based perforated structures in (Akiyama, 2023), gives further support to the notion that, in practical applications, this type of EL-structures are already within the double-mesh performance limit in terms of optical gain and energy resolution.

To better understand the present technology limits, it is relevant to estimate the wavelength-shifting-efficiency (WLSE) achieved for the TPB-coating process inside the structure channels. The ratio of the TPB-coated structure to the non-coated one has been used for this purpose, as the ratio should be more immune to mis-modeling effects (Figure 14). The uncertainty band was taken to be 15%. As shown, the estimated WLSE of the TPB lies within the 74%–128% range, centered at 105%, in line with the literature results (Kuźniak and Szelc, 2020; Araujo et al., 2022). Simulations performed for the structure with ESR produced slightly lower values of the WLSE of around 94%, that could point to a bad optical coupling (or modeling) of the ESR reflector.

The process of wavelength-shifting results in a different spectral content for the FAT-GEM and double-mesh configurations, so it is interesting to perform a comparison for different photosensors, based on the results achieved in this work. Figure 15 shows the simulated number of photoelectrons (normalized to cm⋅bar) assuming the nominal PDE of each of the four exemplary photosensors discussed here (meshes in black, FAT-GEM (present) in pink, FAT-GEM (enhanced) in green). Moreover, a flat-response light sensor with 30% PDE is also shown. The ‘enhanced’ version of the FAT-GEM assumes a diffuse reflector instead of ESR to minimize the effect of photon trapping due to internal reflection inside the PMMA. Experimental data obtained in this work are given by stars (p = 4 bar). All comparisons have been made at a pressure-reduced electric field of E _EL = 3.65 kV/cm/bar for which discharge-free operation of the FAT-GEM was comfortably achieved. Extrapolations to a half-hemisphere photosensor plane are shown by the empty bar.

The use of ‘ad hoc’ perforated structures for electroluminescence is reaching maturity. Here we present performances for the FAT-GEM technology (Field-Assisted Transparent Gaseous Electroluminescence Multipliers), that are already matching those achievable for double-meshes, our proxy for the ‘ideal’ uniform-field situation. The optical gain achieved in xenon at 4 bar, at a pressure-reduced electric field of 3.65 kV/cm/bar, is within 18% of the value achieved in a double-mesh configuration with the same photosensor coverage, a bit over the systematic uncertainty of present data (15%). Further, a good description of the observed yields can be achieved if assuming a WLSE of 105% for the TPB-coated channels. The structure is electrically stable at 3.65 kV/cm/bar (4 bar) and 2.8 kV/cm/bar (10 bar), the latter limited by the maximum power supply of our equipment. Despite being in a high-field region, no systematic TPB degradation was observed during 20 h of continuous operation. Assuming complete hemispherical coverage, the above WLSE figure extrapolates to a number of detected photoelectrons (at 4 bar) in the range of: 29 (41) for Hamamatsu R7378 PM, 43 (57) for Hamamatsu R11410-21 PM, 60 (92) for FBK NUV-HD-SF SiPM and 66 (99) for Hamamatsu VUV4 SiPM when considering the present (or enhanced) FAT-GEM structures. Higher values are anticipated at higher pressures. The observed energy resolution, on the other hand, extrapolates to 0.84% (0.91%) FWHM for 4 bar (10 bar) at the Q _ββ of ¹³⁶Xe, competitive for next-generation ββ0ν experiments. Evidence has been presented, pointing to the intrinsic resolution of the structure being better than that.

The FAT-GEM concept combines the properties of gas scintillation and light-guides, being radiopure, versatile in design and intrinsically transparent from the hard VUV up to visible and near-IR regions (1,600 nm). It is called to solve the scaling issues for large-volume chambers and perhaps offer an universal technique to implement electroluminescence readouts in noble-elements.