Extrinsic nature of the broad photoluminescence in lead iodide-based Ruddlesden–Popper perovskites

Two-dimensional metal halide perovskites of Ruddlesden–Popper type have recently moved into the centre of attention of perovskite research due to their potential for light generation and for stabilisation of their 3D counterparts. It has become widespread in the field to attribute broad luminescence with a large Stokes shift to self-trapped excitons, forming due to strong carrier–phonon interactions in these compounds. Contrarily, by investigating the behaviour of two types of lead-iodide based single crystals, we here highlight the extrinsic origin of their broad band emission. As shown by below-gap excitation, in-gap states in the crystal bulk are responsible for the broad emission. With this insight, we further the understanding of the emission properties of low-dimensional perovskites and question the generality of the attribution of broad band emission in metal halide perovskite and related compounds to self-trapped excitons.

only has a small impact on its early decay dynamics. The BE exhibits a significantly longer lifetime that the initial NE decay, but as shown in (b), there is also a weak tail at long delays observable for the NE. 13   Both cases exhibit a maximum, which lies around 50 K for the green and 150 K for the red flake.   group. [1] A depletion of PEAI suppresses the formation of the broad emission band.

Supplementary Note 1
Power dependent PL spectra allow for deducing the reaction order of the radiative recombination. Supplementary Figure 2 displays the respective data for both types of crystals measured for both the NE and BE. In each case, the PL scales approximately linearly with incident laser power, indicating first order recombination. This allows, for example, for excluding free carrier recombination as the origin of the luminescence. The responsible processes are thus either due to excitons or trap-mediated recombination.
The latter is often associated with a linear behaviour that saturates under high fluence.
However, whilst the BE of PEA seems to run into saturation, it cannot be observed in the case of FPEA. The absence of saturation has led some authors to exclude trapmediated recombination as the origin of the BE.

Supplementary Note 2
Wide-field PL micrographs of the cleaved crystals show a broad variety of colours, depending on the considered flakes. Two examples or FPEA are given in Supplementary   Figure 3, for which the flake in (a) offers bright green luminescence (green flake) and the one in (b) appears much fainter and red-shifted (red flake). In both cases the edges appear brighter due to waveguiding effects, but retain the same colour.
Local probing of the PL spectra with a spot diameter of one square micrometre allows for determining whether the perceived colour of the flakes is governed by effects on the edges. As indicated by the steady state spectra in Supplementary Figure 5, the BE can be as dominant on the edge (red) as in the centre of the crystal (green), but may even be absent on the edge (black), excluding previously invoked edge states as the origin of the BE.
A simple way to increase the surface of crystal flakes and to thereby monitor whether potential surface defects are dominant in the formation of the red-shifted BE, is to physically crush the flakes. The wide-field PL micrograph in Supplementary Figure 6, however, clearly shows that the bright green or red-shifted character of the crystals is retained by their debris. Some crushed material shows distinct red-shifted emission, but other areas remain bight green.
Several reports showed that the PL characteristics of perovskites can be significantly affected by the atmosphere of their surroundings. To monitor a possible suppression of surface defects and its effect on the BE we exposed the cleaved crystals to different environments (Supplementary Figure 7). Since the BE is clearly visible in all cases and the transients of the NE remain unaffected, this serves as additional evidence to exclude surface processes.
Surface treatment of organic ligands e.g. PEA + on PEA-based 2D perovskites was previously shown to increase the luminescence yield and boost device performance through passivation. [1,2] Treatment of PEA flakes with 10 mmol of PEAI dissolved in isoprpyl alcohol does, however, not suppress the BE. In contrast, we find a much stronger BE in steady state ( Supplementary Figure 8 (a)) and a slightly reduced lifetime of the NE (b). This could be due to a large point-to-point variation and non-ideal re-alignment, but as discussed for thin films, excess PEA + can also increase the BE intensity. [1] Encapsulation of the flakes can suppress laser damage on the surface. [3] However, when the flakes are encapsulated under a thin film of PMMA and ALD-processed aluminium oxide, there is no effect on the BE beyond the light-induced healing discussed below.

Supplementary Note 3
Supplementary Figure 10 (a) displays the obtained spectra of a red flake upon continuous exposure to 3.1 eV light in the CLSM set-up. The intensity of the BE decreases over the first ten minutes and saturates afterwards. Simultaneously, the NE shows some minor variation, but by far not as pronounced as the BE. Importantly, the NE and BE behaviour do not correlate (as shown in Supplementary Figure 11). The same trend in BE intensity is observed for continuous excitation at 2.28 eV, i.e. below the absorption onset of the material (b). Given that these measurements were conducted in CLSM geometry under a tightly focused laser spot, we cannot exclude the latter to lead to two-photon absorption. Notably, as indicated by the inset of Supplementary Figure 10 (a), the red-channel PL of previously exposed areas (darker rectangles) remains lower than initially. The change in BE intensity is thus irreversible (on the timescale of the experiment), but BE suppression does not strongly affect the NE signal. Such variation underlines the BE to be of extrinsic origin.
Time-resolved photoluminescence reveals further striking differences between regions dominated by the BE and those exhibiting mostly NE. Supplementary Figure 12 (a) shows the transients of the latter on a picosecond timescale. In case of predominantly narrow emission, the PL decays much slower than in regions of pronounced BE. Conse-quently, red regions exhibit additional channels for non-radiative decay of free excitons.
Prolonged laser exposure only slightly increases the NE lifetime, which is in accordance with the negligible effect on the steady state intensity given in Supplementary Figure 11.
In other words, although the red regions coincide with additional decay channels for free excitons, the suppression of the BE is not identical with a passivation of these channels.
The lifetime of the BE lies on a longer timescale of hundreds of nanoseconds, as shown in Supplementary Figure 12 (b). In the example of FPEA, the average lifetime was determined to 82 ns. Strikingly, the NE measured in the same spot exhibits a weak, but distinct tail at long delay with a lifetime of 74 ns.

Supplementary Note 4
The complete data used for Figure 4 of the main text is given in Supplementary Figure 13. Towards room temperature, the BE vanishes for the green flake, but for the red flakes the broad emission cannot be explained by assuming solely one peak (Supplementary Figur. 14). At 5 K the BE consists of only one emissive state centred around 1.82 eV (along with I at 2.15 eV). Towards room temperature, however, the BE loses the symmetry and becomes more stretched on its high energy side.
The integrated intensity of the two main emission bands are plotted in Supplementary   Figure 14 (a) for a red and a green flake. Whilst all signals brighten upon initial temperature reduction, the trends differ significantly below 275 K. In the case of a green flake, the BE intensity increases strongly and continuously down to approximately 80 K, 25 below which the emission is reduced likely by the increasing presence of I, as discussed in the main text. The green flakes NE similar brightens over a certain range, but its intensity already diminishes below approximately 160 K. In contrast, the red flakes emission bands exhibit a comparatively small dependence on the temperature down to 80 K.
Whilst the BE brightens from RT down to 220 K it then remains virtually constant over the entire range with a small reduction below 80 K. The NE intensity slightly decrease already from 275 K, but starts to increase again below 80 K. Whilst the two different kinds of flakes thus behave differently, there seems to be a general impact on the NE and BE by the emergence of I. Plotting the BE to NE intensity ratio is often used to deduce trapping and detrapping energy barriers, especially in the framework of STEs. [4] We show the respective data in Supplementary Figure 14 (b) in a double-linear plot.
The stark contrast described above manifests in two strikingly different trends, again indicating that the observed emission cannot be of an intrinsic origin.
As noted above for low temperature spectra (e.g. Supplementary Figure 13  The NE and BE are both visible, but the laser light is clearly much stronger. In order to avoid erroneous results, two 610 nm and one 645 nm long pass filters were inserted in the collection path to suppress the detection of scattered laser light (black curves). Fluence dependent excitation at 600 nm was then successfully used to directly excite the BE in either type of crystal with, for which the obtianed curves of the PEA flakes are shown in Supplementary Figure 18 (b). The intensity used in Figure 5 of the main text was extracted at 700 nm, as indicated.
The direct excitation of the BE with photons below the band gap can be distinguished from the case of two-photon absorption by their dependence on the excitation fluence, as discussed in the main text. Given the high fluences needed in this experiment, the high quality 775 nm short pass filter does not completely suppress the detection of scattered laser light, as shown by the thin black line in Supplementary Figure 19 (a). The obtained spectra of the flakes contain both the sample PL and this scattered light. Nonetheless, the PL intensity of the NE and BE both clearly scales with a power of 2, as expected from a two-photon process, whilst the laser scales linearly (see Supplementary Figure 19)