Optical simulations

Optical simulations were conducted using a custom code based on the transfer matrix method⁴⁶ and a genetic algorithm for the Tamm plasmon structural design. After resolving the sample structure, the emission properties were analysed using a commercial finite-difference time-domain software (3D Electromagnetic Simulator from Lumerical Inc.)⁴⁷. The refractive indices of all materials used in the simulations are detailed in Supplementary Fig. 16. This methodology involved isotropically embedding emitting dipoles (averaged over three perpendicular orientations) throughout the perovskite layer to examine the luminous power emitted from the structure’s surface. Single-wavelength simulations were performed at wavelength, λ = 514 nm within a simulation box sized 2 x 2 x 1 µm³, with perfectly matched layers applied to all boundaries. The mesh grid dimensions were set to 10 nm for the x- and y-axis and 1 nm for the z-axis. A 2-dimensional frequency-domain field monitor was employed to capture the far-field projection via Fourier transform, capturing outgoing radiation across a hemisphere⁴⁸.

Fabrication of 1-dimensional photonic crystal

One-dimentional photonic crystal made of 3 pairs of alternating titanium dioxide (TiO₂) and silicon dioxide (SiO₂) were deposited on glass substrates in the Institute of Optoelectronics, Military University of Technology. Thickness of each layer was TiO₂: (70 ± 2) nm and SiO₂: (90 ± 2) nm. The photonic crystal were deposited in an e-beam evaporation system with a plasma source assistance (Syrus 710 Pro, Bühler Leybold Optics, Alzenau, Germany). The base pressure of the system was 2 × 10⁻⁶ mbar. TiO₂ and SiO₂ were deposited at 0.25 nm s⁻¹ and 0.6 nm s⁻¹ rate, respectively.

ITO sputtering

ITO electrodes were sputtered on glass substrates (for reference PeLEDs) and photonic crystal substrates (for Tamm-plasmon-driven PeLEDs) using a custom setup located in the Class 10,000 clean room in the Electrical Engineering Division, Department of Engineering, University of Cambridge. The ITO electrodes were patterned using a metal mask. The ITO sputtering utilised an In₂O₃/SnO₂ 90/10 wt% target, operated at argon flow of 20 sccm, with the pressure of 5 mTorr and power of 40 W. The sputtering rate achieved was 3.7 nm minute⁻¹, resulting in ITO conductivity of 1800 S cm⁻¹ as measured by a 4-point-probe, with an ITO thickness of 83 nm determined by AFM.

Materials

Lead (II) bromide (PbBr₂, 99.999%), phenylethylammonium bromide (PEABr, >99.5%), 1,4,7,10,13,16-hexaoxacyclooctadecane (18-crown-6, ≥99%), poly(4-butyltriphenylamine) (poly-TPD, Mw ≥20,000 g mol⁻¹), poly(9-vinylcarbazole) (PVK, MW:25,000-50,000 mg mol⁻¹), dimethyl sulfoxide (DMSO, anhydrous, 99.9%), chlorobenzene (CB, anhydrous, 99.8%) were purchased from Sigma-Aldrich. Cesium bromide (CsBr, 99.999%) was purchased from Alfra Aesar. 2,2’,2”-(1,3,5-Benzinetriyl)-tris(1-phenyl-1-H-benzimidazole) (TPBi, >99.5%), 8-Hydroxyquinolinolato-lithium (LiQ, >99%) were purchased from Ossila. All chemicals were used without any further purification.

Preparation of perovskite precursor solution

Perovskite precursor was prepared by dissolving PbBr₂, CsBr and PEABr (molar ratio of 1: 1.05: 0.4) in DMSO at 0.25 M initially and then further diluted down to concentrations between 0.13 and 0.23 M for thickness variation. 18-crown-6 was added as additive (molar ratio of 1.7% to PbBr₂) in perovskite precursors for PeLEDs to improve the PeLED performance⁴⁹.

Fabrication perovskite structures

Glass substrates and glass/photonic crystal substrates were cleaned using detergent, deionised water, acetone and isopropanol under ultrasonication for 10 minutes each, followed by a 15-min UV ozone treatment. A solution of PVK (6 mg ml⁻¹ in CB) was spin-coated onto the substrate at 4000 rpm for 30 s, then immediately annealed at 100 °C for 10 min. It is worth noting that other common hole injection layer such as Poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS), Poly(4-butyltriphenylamine) (poly-TPD), Poly[(9,9-dioctylfluorenyl-2,7-diyl)-co-(4,4′-(N-(4-sec-butylphenyl)diphenylamine)] (TFB) posses similar refractive indices, making them compatible with such optical structures. Perovskite precursors, with concentration ranging from 0.13 M to 0.25 M, were spin-coated at 6000 rpm for 90 s and immediately annealed at 70 °C for 5 min. A 100 nm thick Ag film was subsequently thermally evaporated onto the perovskite film at a rate of 0.1 nm s⁻¹.

Fabrication perovskite LEDs

Glass/ITO substrates and glass/photonic crystal/ITO substrates were cleaned in detergent, deionised water, acetone and isopropanol under ultrasonication for 10 minutes each, then treated with UV Ozone for 15 min. Poly-TPD (10 mg ml⁻¹ in CB) and PVK (6 mg ml⁻¹ in CB) is sequentially spin-coated onto the substrate at 4000 rpm for 30 s and immediately annealed at 100 °C for 10 min and 140 °C for 20 min, respectively. The perovskite precursor of concentration 0.16 M (narrow-angle-Tamm-plasmon PeLED) and 0.25 M (wide-angle-Tamm-plasmon PeLED) was spin coated at 1000 rpm for 5 s and 4000 rpm for 55 seconds and then immediately annealed at 90 °C for 10 minutes. TPBi (40 nm), LiQ (3 nm) and Ag (100 nm) were then sequentially thermal evaporated on the perovskite film.

Film thickness measurement

Film thicknesses were assessed by scanning across the depth of a scratch made on the film with a razor blade (Supplementary Fig. 4). The scan was performed on Bruker Dimension Icon atomic force microscope (AFM) with a silicon tip on Nitride lever (Bruker Scanasyst-Air cantilever, spring constant 0.4 N m⁻¹) running on peak force tapping mode. Data were analysed with WSxM 5.0 software⁵⁰.

Reflectance measurement

Tamm plasmon resonance (reflectance) was characterised by UV–visible spectrometer (Shimadzu UV-3600Plus) with an integrating sphere attachment (Shimadzu ISR-603). The total reflectance was measured at 8° offset to keep the specular reflected light within the integrating sphere. Given that the reflectance of the Tamm-plasmon-perovskite structure shown in Fig. 1d was measured at 8°, the actual Tamm plasmon resonance wavelength at 0° should be ca. 3 nm red-shifted. The baseline measurement was done with a 100 nm thick evaporated silver mirror as a reference due to high reflectivity of our samples.

Reflectance measurement was cross-checked with the Agilent Cary7000 Universal Measurement Spectrometer using the Universal Measurement Accessory. The sample is tilted at 6° and the detector at 12° to collect the specular reflectance without blocking the excitation lamp. The baseline measurement is done at 100% transmittance without any reference, thus eliminating any inaccuracy due to the reference defects.

Microscale reflectance of photonic crystal substrates was measured with hyperspectral microscope (Photon Etc. IMA). A lamp light was focussed on the sample through a condenser from below the sample and was collected by the objective lens (Olympus MPLFLN20x with NA of 0.45) and measured by a CCD camera. The measured reflectance of photonic crystal was calibrated with reflectance of a calibration mirror.

Photoluminescence quantum efficiency

PLQE of perovskite films, reference structures and Tamm plasmon structures are measured with a 405 nm continuous wave laser under excitation between 3 to 167 mW cm⁻² in an integrating sphere and the spectrum collected with Andor iDus Si detector. Calculations are based on 3 configurations of the sphere—empty sphere, sample placed in the sphere but laser beam directed on the sphere wall and laser beam directed onto the sample⁵¹.

Characterisation of perovskite LED performance

All PeLEDs were encapsulated in a N₂-filled glovebox with UV-cured resin and glass. The PeLEDs were measured under ambient condition with the LED measurement setup from the Optoelectronics group in Cavendish Laboratory. The PeLEDs were powered by a Keithley 2400 source metre as a voltage source for measuring the current density–voltage characteristics. The photon flux was simultaneously measured using a calibrated circular silicon photodiode centred over the light-emitting pixel. The luminance of the PeLEDs were calculated based on the emission function of the PeLEDs and on the known spectral response of the silicon photodiode. The EL spectra of the devices were measured using a Labsphere CDS 610 spectrometer. The current efficiency was calculated as the ratio between forward luminance and current. The resulting EQE was calculated considering the angular resolved emission using equation from Archer et al.⁴².

where \({V}_{{{{{{\rm{PD}}}}}}}\), \({R}_{{{{{{\rm{PD}}}}}}}\), \({A}_{{{{{{\rm{PD}}}}}}}\) are voltage, resistance and area of photodiode respectively. \(r\) is the distance between photodiode and PeLED. \(S\left(\lambda,\, \theta \right)\) is the measured spectral radiant intensity at angle \(\theta\). \(R\left(\lambda \right)\) is the photodiode responsivity. \({I}_{{{{{{\rm{PeLED}}}}}}}\) is the current across the PeLED. \(q\), \(h\), and \(c\) are unit charge, Planck constant and speed of light respectively.

Angular luminescence measurement

The macroscale angular PL and EL was measured on a home-built setup (Supplementary Fig. 17). For PL, a 405 nm continuous wave laser and the sample were fixed on the optical table with the excitation angle normal to sample surface in x–y direction and 20° above in the z-direction. The laser beam was focussed on the sample with a lens (f = 1000 mm, power density of 0.5 W cm⁻²). For EL, the PeLEDs were powered by the Keithley 2400 source meter at a constant current density of 0.44 mA cm⁻². Both PL and EL spectra was captured by a spectroscopy camera (Andor iDus DU420A Si detector) connected to a fibre collimator (Thorlabs F220SMA-532) through an optical fibre. The fibre collimator was fixed on an automated rotating stage (Thorlabs PRMTZ8 Motorised continuous rotation stage, Thorlabs K-Cube DC servo motor controller). The fibre collimated was fixed at 10 cm away from the sample to maximise the angular resolution while ensuring a high collection of fluorescence. The PL and EL spectra were collected at 1° interval with a scan rate of 3° s⁻¹ and 5° s⁻¹, respectively. In Supplementary Fig. 18, we showed the PL and EL stability are suitably stable over the total angular PL and EL measurement time scales for 60 s and 36 s, respectively, under the same continuous excitation and driving current of 0.5 W cm⁻² for angular PL and 0.44 mA cm⁻² for angular EL. In addition to the PL and EL stability, the PL and EL are always collected from one end to the other end, e.g. from −90° to 90°. Thus, we make sure that the angular measurements are not affected by degradation by checking that the curves are symmetrical (see Figs. 2b, d, h and 3d–f). For both PL and EL measurements, all 4 edges of the samples were masked with black tape to eliminate emission from the sample edge.

The emission across the solid angle is calculated from the angular PL and EL measurements based on the equation shown below, which is derived from Archer et al.⁴².

where \(\alpha\) is the solid angle of interest, \({I}_{{{{{{{\rm{solid}}}}}}\; {{{{{\rm{angle}}}}}}}}\) is the PL or EL intensity, \({I}_{0}\) is the forward intensity, \(S\left(\lambda,\theta \right)\) is the spectral radiant intensity.

Microscale PL was measured with hyperspectral microscope (Photon Etc. IMA), excited with a 405 nm continuous laser. The hyperspectral microscopy measurements were collected with objective lenses (Olympus MPLFLN) of 20x and NA of 0.45 (equivalent to collection angle of 26.7°).

Transient photoluminescence

Time-resolved photoluminescence was measured at fluences between 1 and 370 nJ cm⁻² pulse⁻¹ with confocal microscope (PicoQuant MicroTime 200). The samples were excited with 405 nm pulsed laser (pulse width ~100 ps, repetition rate 5 MHz) that was focussed with an 10x air objective lens.

Time-resolved photoluminescence at lower fluence between 0.01 and 5 nJ cm⁻² pulse⁻¹ were measured with photoluminescence spectrometer (Edinburgh Instruments FLS1000). A 405 nm pulsed laser (pulse width ~50 ps, repetition rate 2 MHz) was focussed on the samples at 45° and 60° and the emission was collected between 45° and 30° respectively at 1 nm bandwidth. Time resolved emission scans were done by sweeping the emission collection wavelengths between 490-520 nm with 5 nm step. Measurements are done with both unencapsulated and encapsulated samples in air and both shows similar results.

Cross-section of Tamm-Plasmon-perovskite structure

The TEM lamella cross-section was prepared with an FEI Helios Nanolab Dualbeam FIB/SEM following a standard protocol⁵². The lamella was transferred minimising air exposure into an FEI Osiris TEM operating at 200 kV and ~140 pA beam current. HAADF images were acquired using a Fischione detector at a camera length of 115 mm, with a dwell time of 1.9 µs and a spatial sampling of 0.7 nm pixel⁻¹. STEM-EDX maps were acquired using a Bruker Super-X silicon drift detector with a collection solid angle of ≈0.9 sr, a dwell time of 50 ms, a spatial sampling of 5 nm pixel⁻¹, and a spectral resolution of 5 eV channel⁻¹. STEM-EDX compositional maps were spectrally rebinned to 10 eV per channel, and denoised using PCA/NMF taking the first 8 components. Data processing was done using HyperSpy v1.6.1, a Python-based analysis suite for hyperspectral data⁵³.