Freedom for Electrons in Halide Perovskites: Funnel or Get Trapped!

Halide perovskites are an exciting class of materials for the optoelectronic devices of the future. Making perovskite films in a low cost way means their structures are full of defects. How can they perform so well in spite of this? Here we use multimodal microscopy to uncover why.

Like Comment
Read the paper

The field of electronics has long been dominated by silicon. Over the past decades, the world has dedicated huge amounts of money and time to optimising silicon manufacturing processes. We have now become unbelievably proficient at making single crystalline silicon, with the concentration of harmful electronic trap states (localized states caused by defects in the material which reduce device performance) of <1011 cm-3. This means practically that there is only one harmful defect per trillion atoms in these silicon ingots! This isn’t just to show off either as silicon is very sensitive to these defects and minor increases in their concentration can be very detrimental to device performance.

Enter halide perovskites from stage left. This is a class of materials that has recently been explored in electronic devices. They can be processed from precursor inks in a very low cost manner and remarkably, are very tolerant to defects compared to silicon. The field of halide perovskite research has exploded over the past decade, with a wide and diverse array of applications having already been demonstrated in the lab: from more ‘conventional’ uses in high efficiency solar photovoltaics (PV) as well as bright and colour pure light emitting diodes (LEDs), to more exotic use cases such as X-ray detectors and retinomorphic devices (devices that mimic the response of optical cells in our eye). These devices are often collectively referred to as optoelectronic (optical+electronic) as they often turn a light input into an electrical output (PV and X-ray detectors) or the other way around (LEDs).

The early development of perovskite PV devices was driven largely by optimising devices in the lab empirically. There is a large parameter space to explore to make a good perovskite PV cell, including a considerable array of charge extraction layers and dopants thereof that one can choose from. Furthermore, the perovskite itself can be optimised, this includes methods of crystallization, thickness, annealing temperature but also the perovskite composition itself. The ABX3 perovskite crystal structure was initially populated with relatively simple compositions with one ionic species at each of the three sites, e.g. MAPbI3 or FAPbI3 (where MA and FA are methylammonium and formamidinium cations respectively). Quickly thereafter, stability concerns and the push for higher efficiency devices pushed many groups away from these simple compositions and towards ‘alloyed’ compositions where multiple distinct ionic species occupy individual sites. Arguably the most popular of these compositional families has been the aptly named ‘triple-cation’ perovskites with a general composition CsxFAyMA1-x-yPb(IzBr1-z)3. This family of compositions have been used in the highest performing solar cells to date with considerably improved stability over the simpler compositions.

Empirical device development has led us to a situation where we know what the optimal device perovskite compositions are but they do not tell us why these compositions are so good. At the Strankslab we are one of the groups trying to clean up the beautiful and extremely complicated mess left behind in the wake of this empirical progress. Perovskites crystallise into polycrystalline films with grains of a few hundred nanometres in size, but show disorder at the intragrain level and multiple length scales above. As such, we find it crucial to use microscopic techniques to understand these materials rather than getting an average of all of these different levels of disorder. We previously showed using a combination of optical microscopy, scanning electron diffraction and photoemission electron microscopy that trap states (states within the bandgap where charge carriers can be trapped or lost to non-radiative recombination) are nanoscale in size and located at grain boundaries. This confirmed the already well known presence of these trap states but visualised their spatial distribution directly for the first time. We wanted to know why devices made of these compositions are able to function very well in spite of the presence of these trap states in an attempt to understand the so-called ‘defect tolerance’ of these materials.

Figure 1. Photo of the hyperspectral microscopy setup.

Our first step was to develop a quantitative optical microscopy technique based on the hyperspectral microscope from Photon etc. in order to probe important device relevant parameters of our perovskite materials with optically diffraction limited resolution (see Figure 1). Hyperspectral microscopy techniques allow one to extract the spectrum of light from each pixel point, and with careful calibration via reference lamps and lasers, we can determine quantitatively the number of photons being emitted per unit time and per unit energy. Using a laser as an excitation source, we are able to extract absolute photoluminescence spectra, while using a white light lamp in reflection or transmission geometry allows the extraction of local reflectance, transmittance and absorptance spectra. The photoluminescence spectra on their own provide a wealth of information; by fitting them with the Generalised Planck’s law we can extract several important parameters including the quasi-Fermi level splitting (QFLS) which is the maximum voltage a solar cell can produce at that point as well as the Urbach energy (EU) a value describing the distribution of below bandgap states which is commonly attributed to disorder. Combining with the local reflectance and transmittance data, we can extract local photoluminescence quantum efficiency (PLQE) maps which tell us how many photons do we get back out for every photon absorbed, the higher this value (meaning the more light the material emits), the higher the voltage a solar cell can produce which increases the device efficiency. We found that areas in these perovskites that were the highest performing had the lowest EU (lowest disorder) but curiously had a slightly lower energy emission spectrum with a pronounced low energy feature, suggesting to us that these areas were compositionally and/or structurally distinct from the low performance areas.

To probe this disorder, we went to the I14 hard X-ray nanoprobe beamline at the Diamond Light Source synchrotron. This is a beamline which focuses a monochromatic X-ray beam down to 50 nm in lateral size so we can do nano X-ray fluorescence (nXRF) and nano X-ray diffraction (nXRD) to probe the chemical composition and crystal structure with a 50 nm resolution)! We, the self-proclaimed nXRD team joined the forces of beamline scientists Dr. Julia Parker and Dr. Paul Quinn to do correlative nanoprobe measurements on the same areas we mapped in our optical microscope. This approach is broadly known as multimodal microscopy. We marked the areas of interest beforehand with gold marker particles with recognisable, geometric shapes. Finding the areas of interest on the beamline is not easy and aligning the images afterwards is honestly a tremendous pain at first (contact us if you want tips with this!) but the resulting information is very rich. We find that the highest performance areas are strongly spatially correlated with a higher local concentrations of bromine (Figure 2) while nanoscale strain measured by nXRD does not play a large role. This is particularly curious given that bromine rich areas should shift the emission spectrum to higher energy whereas we see the opposite here! Clearly something interesting is going on.

Figure 2. (a) nXRF of a perovskite film showing the variation in bromine to lead ratio across the film. (b) and (c) show plots of the Urbach energy overlaid with the highest and lowest Br regions respectively showing that high Br correlates with lower disorder and vice versa!

To probe this apparent paradox we correlate nXRF measurements with another optical technique; transient absorption microscopy (TAM) with the help of Stuart Macpherson and Jooyoung Sung. This is a technique using two laser pulses offset from each other in time to excite charge carriers into excited state and then track their evolution in space, time and energy. This technique allows you to track the behaviour of charge carriers on the nanoscale at ultrafast timescales. We found that the areas that are highest in bromine show that the carriers initially settle into higher energy states than the bromine poor areas as we would expect. However, on timescales of hundreds of picoseconds we see that the carriers in the bromine rich areas shift towards lower energies while in the bromine poor areas they do not. This tells us that must be nanoscale iodine rich inclusions of high optoelectronic quality in the bromine rich areas. The carriers funnel down these energetic gradients into these areas, avoiding traps and emitting light. By contrasts in the bromine poor areas, there are no compositional, energetic gradients for carriers to follow and so more readily find their way to trap states and do not emit light (Figure 3).

Figure 3: Schematic of our proposed model for the nanoscale energy landscape in alloyed perovskites. Br rich regions have energy gradients which funnel carriers to high quality areas, Br poor regions lack these funnels and so the carriers can more easily be trapped.

We believe that, with these complicated compositions found experimentally, we have serendipitously created a particularly defect tolerant landscape for carriers in these perovskites, thus enabling high efficiency solar cells in the face of a huge degree of chemical and structural disorder. Additionally, understanding the complexities of this landscape opens a design strategy towards an ideal level of disorder (organised chaos), at least while there are still defects present, to optimise these perovskite solar cells and we hope to explore this more in the future.

The full paper can be accessed at the link below and we can be contacted on Twitter if you would like to reach out: @KFrohna and @MAnayaLab

Kyle Frohna

PhD Student, University of Cambridge