Some crystals begin to organise before their order becomes visible. Cool titanium diselenide below about 200 kelvin and it settles into a charge-density wave: the electrons rearrange into a periodic pattern, and the lattice distorts with it, doubling its unit cell in all three directions. Above that temperature, the long-range pattern is gone. But for more than four decades, physicists have argued about what is rehearsing in the disordered phase above the transition — whether the electron-hole pairs that would form an excitonic insulator are already flickering into existence there, or whether the lattice is doing the organising and the electrons are merely following.
A paper published in Nature Physics on October 8, 2026, gives that argument its sharpest experimental data point yet. Alfred Zong, Sheng-Chih Lin, and colleagues in Michael Zuerch’s group at the University of California, Berkeley, with collaborators in Japan and China, used few-femtosecond extreme-ultraviolet absorption spectroscopy to watch excitons in 1T-TiSe₂ fall apart — and found that they fall apart differently near the transition than anywhere else. Short-range excitonic correlations persist above the critical temperature, and the excitonic response becomes unusually strong right at it: evidence, the authors argue, for excitonic instability in a bulk crystal.
This is a landmark result for a sixty-year-old question, and it is worth reading closely on its own terms — including the terms on which it does not settle anything. A separate study, first posted as a preprint in 2025 and since peer-reviewed in Communications Physics, reports charge-density-wave fluctuations at room temperature dominated by electron–phonon interactions. The two experiments probe different things in different ways; the honest reading is that the excitonic case is now stronger than it has ever been, while the question of what drives the transition remains open.
The sixty-year riddle
The excitonic insulator is one of the oldest unconfirmed predictions in condensed matter physics. Proposed in 1967 by Jérôme, Rice, and Kohn, it describes a material in which the energy cost of creating an electron-hole pair — an exciton — falls to zero, so that excitons form spontaneously and condense into a quantum-coherent ground state, the way Cooper pairs condense in a superconductor. The catch is that the most obvious signature of such a condensate is a periodic modulation of the charge density — which is precisely what a conventional charge-density wave produces by an entirely different route, through the lattice. An excitonic insulator and a lattice-driven charge-density wave can look identical from the outside, and telling them apart has been the central difficulty ever since.
1T-TiSe₂ has been the flagship candidate for this test since the 1970s. It is a layered transition metal dichalcogenide: sheets of titanium sandwiched between sheets of selenium, held together by weak van der Waals forces. Its band structure is nearly ideal for the excitonic scenario: the top of the selenium 4p valence band and the bottom of the titanium 3d conduction band sit close together near the Fermi level, separated by a small indirect gap, so the Coulomb attraction between a Se 4p hole and a Ti 3d electron can plausibly bind them into an exciton with little energy cost. Below roughly 200 K, the material enters a charge-density-wave phase with a 2×2×2 lattice reconstruction. The question that has survived four decades of experiments is whether that transition is driven by the excitons condensing, by the lattice distorting through electron–phonon coupling, or by some hybrid of the two.
Why static spectroscopy is nearly blind here
The Berkeley team’s starting observation is a negative one, and it is the reason their experiment exists. The static core-level absorption spectrum of 1T-TiSe₂ barely changes across the transition. The XUV absorption features that involve the Se 4p and Ti 3d states — the states that would participate in exciton formation — look almost the same at 300 K as they do deep in the ordered phase.
That blindness is not a failure of the technique. It is a statement about what the transition is: the relevant physics is not in the ground-state spectrum at all, but in the fluctuations — the short-lived, short-ranged correlations that appear and vanish above the transition temperature. To see those, you need a probe that works in the time domain, perturbs the system and watches how it responds, and does so with element and orbital specificity, separating the response of the selenium states from that of the titanium states.
Ultrafast broadband extreme-ultraviolet absorption spectroscopy — UBXAS — is that probe. It is a pump–probe technique: an ultrashort laser pulse excites the crystal, and a broadband XUV pulse, arriving a controlled delay later, measures the absorption spectrum. The XUV photon energies span the core-level edges of both elements, so one shot reads the Se 4p states through the Se M₄,₅ edges and the Ti 3d states through the Ti M₂,₃ edges. And because the pump pulse lasts just 3.4 femtoseconds, the measurement resolves dynamics on timescales shorter than the lattice’s own vibrational period — fast enough to catch electronic processes before the lattice has had time to respond.
The stopwatch trick
The result is not a picture of excitons but an inference from how quickly they break apart.
When the pump pulse hits the crystal, it creates mobile carriers. Those carriers screen the Coulomb attraction holding excitons together, and the excitons dissociate. The rate at which that dissociation happens depends on how many mobile carriers are present: more carriers mean faster screening, and faster dissociation. In the simplest picture, the initial response time should scale as one over the square root of the pump fluence. That scaling is the baseline, and the Berkeley team verified it across most of the temperature range they studied.
The interesting physics is in the deviation from that baseline. Near the transition temperature, at about 190 K, the initial response was consistently fast and roughly independent of pump fluence — even at a mild perturbation of 0.1 mJ/cm², nearly twenty times weaker than the 1.7 mJ/cm² used in the headline measurements. The excitons were dissociating as if the material were already primed to respond, as if the susceptibility of the excitonic system had become anomalously large. In the language of phase transitions, that is what an instability looks like: near the critical point, the system’s response to a small push becomes disproportionately large, because the restoring forces that normally hold it in place have weakened.
The second finding concerns what happens above the transition. The time-resolved spectra show that excitonic correlations do not switch off when the long-range order disappears. Short-range fluctuations — excitonic correlations that extend over small regions but never organise into a coherent, crystal-wide pattern — persist above T_c. The authors’ phrasing is careful: short-range excitonic fluctuations emerge before long-range-order formation in the ground state. The excitons are there, in fragments, before the order they might eventually form.
The measurement also records the lattice’s own voice: all of the absorption peaks respond within 100 femtoseconds of the pump, and coherent oscillations at 6.0 terahertz ripple across most of the spectrum — the A₁g phonon mode, in which the selenium ions displace out of plane relative to the stationary titanium ions. The lattice is present in the data; the argument is that the electronic response near T_c behaves in a way the lattice alone does not explain.
Taken together, the two observations support what the authors call a unified picture: excitonic correlations shape both the amplitude and the phase of the ordered state — the strength of the modulation and its spatial coherence — rather than being a side effect of a lattice transition. That is the paper’s central claim, and it is stated as evidence for excitonic instability in a bulk crystal, not as a proof that 1T-TiSe₂ is an excitonic insulator.
The other side
That restraint matters, because the counter-evidence is real and recent. In a study first posted to arXiv in July 2025 and now published in Communications Physics, Sotirios Fragkos and colleagues at the University of Bordeaux and collaborators used time-resolved XUV momentum microscopy — a technique that maps the electronic band structure directly in energy and momentum — to look for charge-density-wave fluctuations at room temperature, 295 K, far above the transition. They found them: the backfolded Se 4p band associated with the CDW phase remains detectable at room temperature, and when they melted those fluctuations with an infrared pump pulse, the fluctuations recovered in about 700 femtoseconds, with the recovery modulated by a coherent CDW amplitude phonon mode. Their density-functional perturbation theory calculations, which include dynamical electron–phonon interaction, reproduce the spectral features. Their conclusion is direct: the fluctuations at elevated temperatures are dominated by electron–phonon interaction, not by excitonic correlations.
The two results are not in formal contradiction — they measure different quantities with different techniques — but they pull in different directions. The Berkeley experiment measures the dissociation dynamics of excitons and finds an anomalously fast, fluence-independent response near T_c. The Bordeaux experiment measures the band-structure fingerprints of CDW fluctuations at room temperature and finds them governed by the lattice. One can imagine a material in which both are true: excitonic correlations that strengthen near the transition, and a fluctuating phase at higher temperatures in which electron–phonon coupling dominates. The authors of both papers describe a hybrid exciton–phonon mechanism as the likely reality. The dispute is over the balance, and over which interaction does the driving.
None of this diminishes the Berkeley result. It locates it. What Zong and colleagues have done is to give the excitonic side of the argument its most direct measurement yet — a probe that works above T_c, where the precursors live, with the element and orbital specificity to tie the response to the Se 4p and Ti 3d states the excitonic hypothesis names. The Fragkos result gives the lattice side a comparably direct measurement at room temperature. An earlier terahertz pump–probe study, from the Huber group, reached a similar caution years ago: excitonic correlations alone do not drive the transition. The debate is not settled by either paper. It is better posed by both.
What would close it
One route is to make the two measurements on the same footing: the dissociation-dynamics signature of excitonic susceptibility and the momentum-resolved band-structure signature of CDW fluctuations, measured as a function of temperature across the same range, ideally in the same laboratory. If the excitonic susceptibility peaks sharply at T_c while the electron–phonon-driven fluctuations persist smoothly to room temperature, that pattern would support a picture in which the two mechanisms operate in different temperature regimes — and the interesting question would become how they hand off to each other.
A second route is to test the template on other materials. The UBXAS approach is not specific to TiSe₂; it is a method for hunting excitonic instabilities in any bulk candidate where the relevant states sit at accessible core-level edges. Ta₂NiSe₅ and Ta₂Pd₃Te₅ are the obvious next targets — both are excitonic-insulator candidates in which the electronic and lattice contributions are similarly entangled. If the same fluence-independent, anomalously fast response appears near their transitions, the picture gains a generality that a single material cannot provide. If it does not, TiSe₂ starts to look like a special case.
A third route, and the hardest, is the one the field has been circling since 1967: a direct signature of the condensate itself, rather than of its precursors. Short-range fluctuations and enhanced susceptibility are evidence for instability — for a system poised to condense. They are not a measurement of a condensed state. A superconductor is not established by showing that pairing fluctuations exist above T_c; it is established by showing the coherent state below it. The excitonic insulator still awaits its equivalent demonstration.
What it is not
Three cautions, stated plainly because the framing invites overreading. First, this is not a proof that bulk 1T-TiSe₂ is an excitonic insulator. The paper’s own language is “evidence for excitonic instability,” and the distinction between instability and a realised condensate is the whole ballgame. The study probes signatures associated with excitons and their fluctuations; it does not directly image a condensate.
Second, this is not a device result, and it should not be read as one. Nothing in the paper points toward an application, and the excitonic insulator — even if confirmed in TiSe₂ — is a ground state of a crystal at cryogenic temperatures, not a technology.
Third, this is not the end of the debate, and treating it as such would misread both papers. The Fragkos result stands, the earlier terahertz work stands, and the hybrid picture — excitons and phonons intertwined — remains the consensus that any single-mechanism claim has to overcome. What has changed is the quality of the evidence on the excitonic side: the case for precursors above T_c now rests on a direct, time-resolved measurement of the fluctuations themselves, not on inference from the ordered phase.
That is a genuine advance, worth stating without overselling it. A sixty-year-old question has acquired a new instrument — a few-femtosecond clock with element and orbital resolution — and the first thing that instrument found is that the excitons are already there, flickering, before the order arrives.
Sources and further reading
Zong, A.; Lin, S.-C.; Sato, S. A.; Berger, E.; Nebgen, B. R.; Hui, M.; Lv, B. Q.; Cheng, Y.; Xia, W.; Guo, Y.; Xiang, D.; Zuerch, M. W. “Short-range excitonic correlations and enhanced excitonic susceptibility in 1T-TiSe₂.” Nature Physics (2026), published online October 8, 2026. DOI: 10.1038/s41567-026-03423-z. Received January 20, 2026; accepted July 23, 2026. Open access (CC BY 4.0). Preprint: arXiv:2407.00772. For a plain-language account of the experiment, see “Ultrafast core-level spectroscopy reveals elusive precursors of exciton condensation in quantum materials,” phys.org, October 9, 2026.
For the contrasting electron–phonon view, see Fragkos, S.; Orio, H.; Girotto Erhardt, N.; et al. “Electron-phonon-dominated charge-density-wave fluctuations in TiSe₂ accessed by ultrafast nonequilibrium dynamics.” arXiv:2507.12430 (July 2025), published in Communications Physics (2026). DOI: 10.1038/s42005-026-02521-x. For the excitonic-insulator proposal, see Jérôme, D.; Rice, T. M.; Kohn, W. “Excitonic insulator.” Physical Review 158, 462–475 (1967). For background on what an exciton is, see this site’s explainer, “What is an exciton?” This article is analysis of published research, not investment advice.