For years, the low-temperature glow of (PEA)2PbI4 has been an embarrassment of structure. Cool a flake of this layered perovskite — phenethylammonium lead iodide, the archetypal n = 1 Ruddlesden–Popper compound — and its photoluminescence breaks into a multiplet of sharp lines, plus a faint doublet sitting well above them. The lines were real and reproducible. What they were was not. Competing explanations invoked crystal-field effects, Rashba-type spin splitting, and exciton–polaron sidebands, and the weak high-energy doublet had no agreed assignment at all.
A paper published online in Nano Letters on October 21, 2026, by Patrick Grenzer, Fabian Lie, Klaus H. Eckstein, Tobias Hertel, and Linn Leppert — from the University of Würzburg, the University of Twente, and the University of Birmingham — now offers a cleaner account. Combining polarization-resolved low-temperature photoluminescence with first-principles G0W0 plus Bethe–Salpeter equation (G0W0+BSE) calculations, the authors argue that the low-energy multiplet is intralayer excitonic fine structure, doubled by the two inequivalent layers in the unit cell, and that the weak doublet is interlayer excitons: electron and hole sitting on adjacent inorganic sheets. Neither Rashba splitting nor exciton–polaron mechanisms is needed as the primary explanation of the low-energy structure. The claim is disciplined rather than sweeping, and the places where the theory falls short are stated plainly. That restraint is part of what makes the result worth taking seriously.
A natural quantum well, twice over
To see why (PEA)2PbI4 is such a revealing test case, start with the structure. In a Ruddlesden–Popper perovskite, sheets of corner-sharing PbI6 octahedra — the inorganic layers that carry the electronic states — are separated by bulky organic cations, here phenethylammonium (PEA). The organic spacers act as barriers, so each inorganic sheet behaves like a quantum well only one octahedron thick. In the n = 1 limit, quantum and dielectric confinement are at their strongest, and excitons are bound by several hundred meV — tight enough that their signatures survive even at room temperature, though the fine structure only sharpens out at cryogenic temperatures.
The standard fine-structure picture for such a well starts with electron–hole exchange. Exchange splits a predominantly dark singlet state, conventionally labeled D and associated with total angular momentum J = 0, from a bright triplet with J = 1. Reduced crystal symmetry then distinguishes the bright states by their polarization: the in-plane bright states X and Y from the out-of-plane bright state Z. Octahedral distortions lift the near-degeneracy of the in-plane states and set their approximately orthogonal polarization axes. That D/X/Y/Z picture has organized the field’s thinking for years, supported by time-resolved photoluminescence that put the dark–bright splitting near 10 meV and by high-field polarization spectroscopy that resolved in-plane splittings on the meV scale.
But (PEA)2PbI4 adds a complication that the textbook picture omits. Its crystallographic unit cell contains two inorganic layers, and those two layers are not symmetry-equivalent. The organic dielectric environment around each layer differs, so the band edges of the two layers sit at slightly different energies. The excitonic manifold therefore doubles: every intralayer state exists in an L1 version and an L2 version. Recent theory had already pointed to this layer degree of freedom; what was missing was an experiment that resolved it cleanly and a calculation that reproduced the pattern.
What the experiment shows
The experiment is polarization-resolved photoluminescence on mechanically exfoliated (PEA)2PbI4 flakes, measured down to 8 K. At room temperature the spectrum is largely featureless, washed out by thermal broadening. At 8 K, three emission regions appear near 2.30, 2.34, and 2.38 eV. The weak low-energy feature is attributed to a charged exciton, a trion, consistent with residual doping. The central band is the intralayer fine structure. The weak higher-energy doublet is the puzzle the paper sets out to solve.
Polarization is the decisive tool. The dominant emission between about 2.332 and 2.354 eV decomposes into two nearly orthogonal linear-polarization components, πx and πy. Polarization maps alone, however, do not assign lines to states — several candidate manifolds can produce similar patterns. The assignment comes from comparing the measured ordering and polarization pattern against the G0W0+BSE calculation, performed with the BerkeleyGW package on a fully relaxed structure including spin-orbit coupling. One methodological caveat governs how much weight this comparison can bear: one-shot G0W0+BSE calculations from semilocal density functional theory commonly underestimate absolute exciton energies in halide perovskites, so the authors align theory and experiment with a rigid blue shift of 0.491 eV for the intralayer manifold. The comparison therefore tests relative ordering and polarization character, not absolute energies — the right standard, and one the paper states explicitly.
On that standard, the agreement is good where it matters most. The calculated bright intralayer sequence reproduces the observed alternation between the dominant lines in the πx and πy channels. Fitting the polarization-resolved data with cos²-type angular dependencies yields four dominant bright transitions: X1 at 2.3402 eV, Y1 at 2.3420 eV, X2 at 2.3423 eV, and Y2 at 2.3438 eV, each with an uncertainty of ±0.2 meV. These are two closely spaced X/Y-like pairs — exactly what a doubled, layer-resolved intralayer manifold predicts. The extracted energy scales are small and precise: an L1/L2 splitting of 2.0 meV between the two bright pairs, and an intralayer Y–X anisotropy of about 1.5–1.8 meV within each pair.
The quantitative agreement is not perfect, and the paper does not pretend otherwise. The calculation overestimates the L1/L2 splitting, predicting roughly 8 meV rather than the observed 2.0 meV, and underestimates the intralayer bright-state anisotropy, giving about 0.4 meV instead of 1.5–1.8 meV. The D and Z partners, predicted to bracket each bright X–Y pair, appear only as weak shoulders, so their positions are tentative. The assignment therefore rests on the qualitative pattern — relative ordering and polarization character — rather than a numerical match of every splitting. That is an honest basis, but it is worth naming: the intralayer assignment is strongly supported, not proven to the last decimal.
Why the layer doubling happens
The most instructive part of the theory is a control experiment done in silico. To find out what drives the L1/L2 splitting, the authors compared the full (PEA)2PbI4 structure against two cesium-based models: one in which the organic PEA spacers were replaced by Cs ions while keeping the distorted inorganic framework, and one idealized Cs2PbI4 structure in which both the spacers and the octahedral distortions were removed, restoring a highly symmetric, untilted lattice.
The result isolates the cause. In the idealized symmetric limit, the L1/L2 splitting collapses to nearly zero. In the Cs-substituted model with the distorted framework retained, a small residual X–Y anisotropy remains but the layer-dependent offset is negligible. Only the full low-symmetry PEA environment produces the pronounced calculated splitting. The splitting is therefore driven primarily by a layer-dependent offset arising from the inequivalence of the organic dielectric environment around the two inorganic layers — not by strong hybridization between layer-resolved excitonic configurations. The two layers emit at slightly different energies because they live in slightly different dielectric neighborhoods, not because their excitons mix strongly.
This result also settles the question about Rashba effects and exciton–polaron mechanisms, at least as primary explanations. The relaxed reference structure is centrosymmetric, so the calculations exhibit no Rashba-type band-edge splitting, and they include no exciton–phonon coupling at all — yet they reproduce the qualitative ordering and polarization pattern of the dominant bright lines. If a mechanism is absent from a calculation that still captures the pattern, that mechanism is not required to explain the pattern. The paper keeps the claim careful: it does not say Rashba or polaron physics is absent from the material, only that it is not needed as the primary origin of the low-energy fine structure. Both effects may still show up elsewhere in the spectrum — as lattice distortions do in the second half of the paper.
The doublet, forty meV up
The weak doublet sits at 2.3794 eV and 2.3832 eV in the preprint’s reported values — about 40 meV above the bright intralayer states, with an internal splitting of roughly 4 meV. At 8 K, its two components, labeled IXL and IXU, show nearly orthogonal in-plane polarization axes and much lower intensity than the dominant intralayer bands, consistent with a weakly allowed excitonic manifold.
The G0W0+BSE calculation predicts two weak sets of interlayer excitons in about the same spectral region, lying about 58 meV above the bright intralayer excitons in the calculation compared with about 40 meV in experiment, with a larger internal splitting of 19 meV versus 4 meV observed. The quantitative mismatch is larger here than for the intralayer manifold, and the authors do not hide it. What supports the assignment is the character: the calculated states’ real-space probability distributions show clear interlayer character, with electron and hole residing on adjacent inorganic layers. The lower-energy set is dominated by L2 → L1 transitions between the inner band-edge states, the higher-energy set by L1 → L2 transitions between the outer band-edge states — a pattern that follows directly from the layer-dependent band offsets established in the intralayer analysis.
Here, though, the static picture breaks down in an instructive way. The calculated interlayer excitons with the largest oscillator strengths have nearly parallel in-plane dipole orientations, while the observed IXL and IXU peaks are nearly orthogonally polarized. A second pair of calculated interlayer states does show nearly orthogonal dipoles, but carries much weaker oscillator strength. So the reference-structure calculation gets the energetics and the real-space character right but fails on visibility and polarization — the two properties that determine whether a state actually shows up in a spectrum, and how.
The paper’s proposed resolution is lattice-coupled mixing. Phonon calculations reveal low-energy modes with strong Pb–I buckling character localized predominantly on one of the two inorganic layers. Displacing the structure along a representative such mode breaks centrosymmetry and the near-equivalence of the two layers, modifies the band-edge shape and alignment, and — in the new G0W0+BSE calculation — produces much stronger mixing between intra- and interlayer excitonic configurations. The mixed interlayer states gain oscillator strength and exhibit nearly orthogonal in-plane dipole orientations. The mixing is presented as a plausible explanation, not a demonstrated fact: the distorted-structure calculation illustrates a route by which lattice coupling could account for the observations, while the assignment itself rests on the reference-structure energetics, splitting, and real-space character.
Note the nuance here, because it is easy to misreport. The distortion calculation does produce a Rashba-type band splitting as a consequence of local symmetry breaking. That does not contradict the paper’s earlier conclusion: the Rashba-like effect appears as one consequence of the distortion that also enhances intra–interlayer mixing, not as the primary origin of the low-energy intralayer fine structure. Keeping those two roles distinct is essential to understanding what the paper actually claims.
What is settled, what is not, and what comes next
Taken together, the paper delivers a layer-resolved excitonic picture of (PEA)2PbI4 that is considerably more complete than the old D/X/Y/Z account. The low-energy multiplet is intralayer fine structure, governed by electron–hole exchange, crystal symmetry, and the inequivalence of the two layers in the unit cell. The weak doublet some 40 meV higher is, on the evidence presented, interlayer excitons — the first assignment of its kind for this much-studied compound, and one that makes the excitonic spectrum of a nominally single-layer material intrinsically a two-layer story.
The open questions are the ones the authors themselves list, and they are the right ones. The D- and Z-like partner positions remain tentatively inferred from weak shoulders, not directly resolved. The interlayer assignment still needs more direct tests — the paper singles out resonant and time-resolved spectroscopies, and temperature-, strain-, or field-dependent measurements, as particularly informative. And on the theory side, static displaced-structure calculations are a stand-in for the real physics: treatments that include dynamic lattice fluctuations and exciton–phonon coupling will be needed to determine under what conditions interlayer excitons become bright and acquire the observed polarization pattern.
There is also a broader lesson, and it is the reason this paper matters beyond one compound. Layered perovskites are often treated as stacks of independent quantum wells, with the organic spacer as a passive barrier. This work shows the spacer is not passive: the dielectric inequivalence it creates between neighboring layers is itself a spectroscopic degree of freedom, large enough to double the fine structure and to set the stage for interlayer states. In related compounds with different spacers, distortions, or n > 1 thicknesses, the same ingredients — symmetry, layer inequivalence, and lattice-coupled mixing — will rearrange into different spectra. The framework here is a way to read those spectra rather than a final answer for any one of them.
For device-minded readers, a note of proportion is in order. This is fundamental spectroscopy, not a device result. Interlayer excitons in (PEA)2PbI4 are weak, cryogenic features, and nothing in the paper claims a route to a room-temperature application. Their significance is conceptual: they show that charge-separated, layer-resolved excitonic states exist and can be identified in a material family already central to LEDs and photodetectors, and they give theorists a sharper target for the models that will eventually describe those devices’ emission. A spectrum that has been argued over for years now has a coherent reading, with its uncertainties labeled. In a field that moves as fast as this one, that kind of clarity is the real advance.
Sources and further reading
Grenzer, P.; Lie, F.; Eckstein, K. H.; Hertel, T.; Leppert, L. “Intra- and Interlayer Excitonic Fine Structure of the Two-Dimensional Perovskite (PEA)2PbI4.” Nano Letters (2026), published online October 21, 2026. DOI: 10.1021/acs.nanolett.6c01957. Received April 22, 2026; accepted September 22, 2026. The preprint version (arXiv:2604.19327, April 21, 2026) reports the doublet at 2.3794 and 2.3832 eV and describes the separation as about 45 meV above the onset of the intralayer manifold; the published abstract states the doublet as about 40 meV above the bright intralayer states, consistent with the measured bright-line energies. Energies, splittings, and polarization character quoted above follow the preprint’s reported values where the published abstract gives only the summary figure.
For background on the earlier competing interpretations, see the exciton–polaron account in Neutzner et al., Physical Review Materials (2018), and the fine-structure reviews and magneto-optical studies cited in the paper, including Fang et al., Advanced Functional Materials (2020), on band-edge exciton fine structure in layered hybrid perovskites. For the general framework of exciton fine structure in perovskites, see Sercel et al., Nano Letters (2019). This article is analysis of published research, not investment advice.