Highly confined epsilon-near-zero and surface phonon polaritons in SrTiO3 membranes

Surface phonon polaritons (SPhPs) are hybrid modes bound to interfaces in polar dielectrics, existing within Reststrahlen bands where the real permittivity is negative. In bulk crystals, SPhPs are weakly confined with wavelengths close to free space. When a polar crystal is thinned to deep-subwavelength membranes, SPhPs at the two interfaces hybridize into antisymmetric and symmetric branches. The antisymmetric branch lies at lower energy and supports propagating modes with momenta far exceeding free-space values, while the symmetric branch shifts toward the longitudinal optical (LO) phonon frequency, approaching an epsilon-near-zero (ENZ) condition where the normal electric field is strongly enhanced and the dispersion becomes nearly flat and radiative within the light cone (Berreman mode). Achieving low-loss, transferable thin films with strong phonon resonances is key for infrared nanophotonics. While van der Waals crystals (e.g., hBN, GeS, GaSe, α-MoO3) support low-loss, deeply subwavelength polaritons, their scalability and anisotropy can limit certain applications and the available Reststrahlen bands are restricted. Cubic and pseudocubic perovskite oxides, particularly SrTiO3 (STO), offer intense, low-loss phonons, isotropic response, and tunability via fields, strain, oxygen vacancies, and doping, and can now be synthesized as large-area freestanding membranes. Theory predicts highly confined SPhPs with good propagation in ultrathin STO down to monolayers, but experiments have been lacking. Here, the authors investigate 100 nm crystalline STO membranes transferred onto gold and SiO2/Si substrates. Using far-field FTIR and near-field SINS, they probe and confirm symmetric ENZ/Berreman and antisymmetric SPhP modes, revealing strong field enhancement and order-of-magnitude wavelength confinement due to the membrane geometry and demonstrating a new oxide platform for infrared photonics and polaritonics.

Prior work established SPhPs as low-loss infrared excitations in polar dielectrics and leveraged naturally anisotropic van der Waals crystals (hBN, GeS, GaSe, α-MoO3) for deeply subwavelength phonon polaritons, enabling advances in sensing, superlensing, perfect absorption, switching, nonlinear optics, and thermal management. However, vdW materials often come as small exfoliated flakes, complicating scalability, and their anisotropy can be undesirable in some applications. Perovskite oxides possess intense, tunable phonons and can support low-loss PhPs. SrTiO3 is technologically mature, exhibits incipient ferroelectricity, dilute superconductivity, and interfacial 2D electron gases, and offers multiple knobs (electrical/optical excitation, strain, stoichiometry/doping) to tune phononic and photonic properties. Advances in freestanding perovskite membranes down to the monolayer limit open opportunities for polaritonics; theory predicted highly confined, propagating SPhPs in ultrathin STO membranes, but experimental evidence had been missing prior to this study.

Synthesis and transfer: A 100 nm epitaxial SrTiO3 (STO) film was grown by pulsed-laser deposition on (001) STO substrates with a 16 nm Sr2CaAl2O6 (SCALAO) water-soluble sacrificial layer. SCALAO growth: Ar pressure 4×10^-6 Torr, 710 °C, laser fluence 1.35 J/cm^2, 1 Hz, 4.8 mm^2 spot. STO growth: O2 pressure 4×10^-4 Torr, 710 °C, fluence 0.9 J/cm^2, 3 Hz, 3.0 mm^2 spot. After growth, a 600 nm PMMA support was spin-coated and baked at 135 °C. The stack was immersed in deionized water to dissolve SCALAO, releasing a millimeter-scale STO film. The film was transferred onto a SiO2/Si substrate with half the surface coated by 50 nm Au (electron-beam evaporation) to enable direct comparison of metallic versus dielectric support. PMMA was removed in acetone at 60 °C and rinsed in isopropanol. Characterization: XRD θ–2θ scans (Bruker D8 Discover; Cu Kα1) confirmed epitaxy and high crystalline quality. Atomic-resolution STEM (ThermoFisher Titan, 300 keV) verified crystallinity post transfer. AFM (Veeco Multimode IV, tapping, MikroMasch HQ:NSC15/Al BS tips) measured atomically smooth surfaces (RMS ~850 pm) and sharp edges. Far-field FTIR: Reflectivity measured with a Bruker Hyperion 2000 microscope coupled to Bruker Vertex 70V. Objectives NA 0.4 and 0.5 (15× and 36×). Detectors: MCT (mid-IR) and liquid-He Si bolometer (far-IR). Reflectivity of STO membrane on Au and on SiO2/Si was normalized to that of the bare corresponding substrate. Bulk STO permittivity ε(ω) was obtained by fitting near-normal incidence reflectivity with a factorized multi-oscillator formula (three relevant phonons) to provide input for modeling. Near-field SINS: Performed at ALS Beamline 2.4 using a NeaSNOM s-SNOM with an asymmetric Michelson interferometer. A metal-coated AFM tip (tapping amplitude ~100 nm, Ω ~250 kHz) concentrates broadband synchrotron IR at the apex. The backscattered tip–sample near-field was demodulated at n=2 to retrieve s2(ω) and Φ2(ω) spectra; a Ge:Cu photoconductor and KRS-5 beamsplitter provided detection down to 350 cm^-1. Spectral resolution 10 cm^-1 (zero-filled by 4). Spectra normalized to a clean gold reference. Line scans across membrane edges on SiO2/Si recorded spatially resolved spectra to analyze propagation and interference fringes. Modeling: The momentum- and frequency-dependent p-polarized Fresnel reflection coefficient r_p(q,ω) for the multilayer stacks (air/STO(100 nm)/substrate) was computed via recursive methods. STO thickness fixed by AFM (100 nm); SiO2 thickness 275 nm deduced from mid-IR interference; Au (50 nm) and Si (~1 mm) treated as optically thick. Finite-dipole model (FDM) for s-SNOM spectra: tip as an ellipsoid with parameters shared across simulations (tapping amplitude 76 nm peak-to-peak, closest approach b=0, incidence α=45°, half-length L=740 nm, radius a=100 nm). FDTD (Lumerical) simulated field distributions for ENZ and antisymmetric modes under plane-wave illumination; a gold nanobeam (100 nm height, 400 nm width) on the membrane provided in-plane momentum to launch polaritons. Optical constants for STO, SiO2, and Si were taken from far-field measurements.

• Far-field reflectivity of 100 nm STO membranes on Au and on SiO2/Si shows clear dips at ~475, ~545, and ~790 cm^-1. Modeling indicates the ~545 cm^-1 feature is a TO absorption (ω_TO,3), while dips near LO frequencies (ω_LO,2 and ω_LO,3) arise from Berreman (ENZ) modes activated at finite incidence (θ ~13°), consistent with numerical aperture effects.
• Bulk STO dielectric parameters (from factorized fit): ω_LO,2 = 475 ± 0.3 cm^-1, ω_TO,3 = 546.1 ± 0.8 cm^-1, ω_LO,3 = 798.4 ± 1.7 cm^-1 (others in Supplementary). In the membrane, these modes are red-shifted by 2.8, 3.4, and 8 cm^-1, respectively, relative to bulk.
• Berreman mode linewidths in the membrane: 6.3 ± 1.7 cm^-1 (near ω_LO,2) and 33.1 ± 2.3 cm^-1 (near ω_LO,3), yielding quality factors Q ≈ 60 and ≈ 30, matching bulk LO scattering rates and reflecting low Im(ε) (<0.3) at LO frequencies.
• Near-field SINS on Au-supported membrane shows two intense asymmetric peaks just below ω_LO,2 and ω_LO,3, attributed to symmetric ENZ modes with weak negative dispersion; no feature is observed at ω_TO,3, and antisymmetric modes are absent.
• On SiO2/Si-supported membrane, SINS reveals, in addition to ENZ features, new structures: a peak at 560 cm^-1 and an upturn below ~370 cm^-1. Comparison with r_p(q,ω) dispersion maps and finite-dipole simulations identifies these as antisymmetric SPhP modes with positive group velocity; a 490 cm^-1 peak originates from SiO2 phonons.
• FDTD field maps show ENZ fields are confined within the membrane and weakly influenced by the substrate, whereas antisymmetric modes have strong fields outside the membrane and are suppressed by metallic (Au) substrates.
• Real-space SINS line scans across edges on SiO2/Si display spatially dispersing amplitude/phase features between ~550–600 cm^-1, consistent with interference fringes from propagating antisymmetric SPhPs. Fitting spatial profiles with a complex wavevector q = q1 + i q2 yields: • Confinement factor C = q1/k0 up to ~10 in the 100 nm membrane (vs. ~1 in bulk), indicating an order-of-magnitude wavelength reduction. • Propagation length L = 1/q2 decreases from ~6 μm to ~2 μm with increasing frequency (within 550–600 cm^-1). • Propagating-mode quality factor Q = q1/q2 ranges from ~2 to ~6, comparable to non-encapsulated graphene plasmons and exceeding some THz plasmonic systems, though below values for hBN-encapsulated graphene and hyperbolic PhPs in hBN or MoO3.
• Antisymmetric modes and spatial dispersion are absent in Au-supported regions, consistent with field screening by the metal and their suppression.
• Overall, experiments and modeling (FDM, FDTD, r_p dispersion) are in strong agreement, confirming symmetric-antisymmetric splitting, ENZ/Berreman modes, and highly confined, propagating antisymmetric SPhPs in ultrathin STO membranes.

The study addresses the open question of whether ultrathin perovskite oxide membranes can host low-loss, strongly confined phonon polaritons with clear symmetric (ENZ/Berreman) and antisymmetric branches. Far-field reflectivity at finite incidence reveals Berreman resonances at LO phonons, evidencing symmetric ENZ modes unlocked by the membrane geometry. Near-field SINS directly detects ENZ modes and, crucially, the antisymmetric propagating branches on dielectric support, with order-of-magnitude confinement compared to bulk. The absence of antisymmetric modes on Au corroborates their field distribution predominantly outside the membrane and sensitivity to substrate screening, while ENZ modes remain largely substrate-independent due to internal field confinement. Quantitative agreement between measured spectra, dispersion-derived r_p maps, finite-dipole spectral simulations, and FDTD field profiles builds a coherent picture: membrane thickness drives symmetric-antisymmetric splitting, ENZ field enhancement, and high in-plane momentum of antisymmetric SPhPs. These findings validate oxide membranes as a scalable, isotropic platform for infrared polaritonics, enabling enhanced light–matter interactions for sensing, perfect absorption, resonant metasurfaces, and nonlinear processes, and offering compatibility with photonic integration.

This work experimentally demonstrates symmetric–antisymmetric mode splitting of surface phonon polaritons in 100 nm SrTiO3 membranes. The symmetric branch is pushed to the ENZ regime, producing strong internal field enhancement and Berreman features at LO phonons, while the antisymmetric branch supports propagating SPhPs with up to tenfold wavelength confinement relative to bulk. Far-field FTIR, near-field SINS, and complementary modeling (finite-dipole, FDTD, Fresnel r_p) provide consistent evidence across metallic and dielectric substrates, revealing substrate sensitivity of antisymmetric modes and robustness of ENZ modes. These results establish oxide membranes as promising building blocks for infrared nanophotonics, including subwavelength resonators, metasurfaces, perfect absorbers, and hybrid polariton platforms. Future work could reduce extrinsic losses via low-loss or suspended supports, probe thickness dependence down to the monolayer limit, investigate strain/surface effects behind observed phonon red shifts, and exploit enhanced interactions to control phase transitions and realize strongly coupled hybrid polaritons.

• The antisymmetric propagating modes are suppressed on metallic (Au) substrates due to field screening, limiting substrate choices for propagation studies and devices.
• Near-field spectral modeling via the finite-dipole approximation does not capture all quantitative details of the tip–sample interaction.
• An extrinsic optical loss from the SiO2/Si substrate reduces propagation length and quality factor relative to ideal conditions; higher performance is expected on lower-loss or suspended membranes.
• Observed red shifts of phonon modes in the membrane relative to bulk (2.8, 3.4, and 8 cm^-1 for LO2, TO3, LO3) likely arise from strain or surface effects but are not fully resolved here.
• The study focuses on a single membrane thickness (100 nm); a systematic thickness dependence and broader parameter space are not explored in this work.