Zero-Dimensional Cesium Lead Halides: History, Properties, and Challenges

Abstract

Over the past decade, lead halide perovskites (LHPs) have emerged as new promising materials in the fields of photovoltaics and light emission due to their facile syntheses and exciting optical properties. The enthusiasm generated by LHPs has inspired research in perovskite-related materials, including the so-called “zero-dimensional cesium lead halides”, which will be the focus of this Perspective. The structure of these materials is formed of disconnected lead halide octahedra that are stabilized by cesium ions. Their optical properties are dominated by optical transitions that are localized within the individual octahedra, hence the title “‘zero-dimensional perovskites”. Controversial results on their physical properties have recently been reported, and the true nature of their photoluminescence is still unclear. In this Perspective, we will take a close look at these materials, both as nanocrystals and as bulk crystals/thin films, discuss the contrasting opinions on their properties, propose potential applications, and provide an outlook on future experiments.

The remarkable properties of lead halide perovskites (LHPs) directly relate to their peculiar crystal structure, the ABX₃ perovskite structure, in which corner-sharing BX₆ octahedra form a cubic framework and A cations fill the voids.¹ The first account of lead halides crystallized in the perovskite structure dates back to 1893, when different colored powders with a CsPbX₃ composition, with the color depending on the halide, could be prepared.² As was reported in that same work, cesium lead halides can also crystallize as white powders, regardless of the type of halide that is used, with a Cs₄PbX₆ stoichiometry. In this phase, the PbX₆^4– octahedra are no longer corner-shared; thus, the photoexcited carriers in these materials experience a much stronger quantum confinement than that of CsPbX₃. For this reason, the Cs₄PbX₆ phase is often called a “zero-dimensional (0D) perovskite”, although it bears no structural resemblance to the perovskite structure. The Cs₄PbX₆ phase remained almost completely forgotten during the 20th century, with only a few works investigating their optical properties (specifically by Nikl and Kondo et al. during the 1990s and 2000s).³⁻⁹ It was not until 2016 that these Cs₄PbX₆ compounds started to generate interest again, mainly as a result of the surging pursuit in LHPs and related metal halides,¹⁰ and papers began to report Cs₄PbBr₆ powders with strong and stable green photoluminescence (PL).^11,12 Soon after, the first publications on Cs₄PbX₆ nanocrystals (NCs) emerged.^13,14 Interestingly, in stark contrast to the highly emissive CsPbX₃ NCs that were first synthesized in 2015¹⁵ and the highly green emitting Cs₄PbBr₆ powders, some of these Cs₄PbX₆ NCs exhibited no PL; instead, they had optical properties that were similar to those described in the first reports on Cs₄PbX₆ in 1893 and in the 1990s.¹³ Furthermore, these optical properties were also similar to those of individual [PbBr₆]^4– clusters.¹⁶ The proposed origin of the green PL is either intrinsic to Cs₄PbBr₆ (due to the presence of defects)^{11,12,14,17−26} or from contamination by CsPbBr₃ NC-like impurities.^13,27−42

In this Perspective, we will examine the literature on 0D cesium metal halides, focusing on Cs₄PbX₆, both in their bulk form and as NCs. We will discuss both old and recent literature concerning this phase, and we will compare current attributions to the origin of the strongly downshifted PL. In addition, we will suggest further experiments that, in our opinion, will help to clarify the optical properties. Finally, we will discuss potential future applications, as well as postsynthesis transformations of Cs₄PbBr₆ NCs into highly luminescent CsPbBr₃ NCs.

We start the Perspective with a description of the cubic (3D) LHP crystal structure due to its relation with the Cs₄PbX₆ phase. In this structure, [PbX₆]⁴⁻ octahedra corner-share each X^– anion between two octahedra, resulting in a cubic (or pseudo-cubic, like orthorhombic/tetragonal, depending on the sizes of the A and X ions) Pb–X–Pb framework,⁴³ as shown in Figure 1a. To charge balance and provide geometrical stability, the lead halide framework can be stabilized by monovalent cations (often depicted as “A”), which can be in the form of Cs⁺, or organic cations like methylammonium (MA) or formamidinium (FA). The rules on whether perovskite structures are stable or not (the so-called Goldschmidt tolerance factor, t) are mainly defined by the ionic radii (r) of A, B, and X and can be determined by using the formula t = (r_A + r_X)/, with stable perovskites generally forming with t = 0.7–1.¹ Long before LHPs became popular in photovoltaics and NC chemistry, it had already been reported that cesium lead halides can crystallize in two other stoichiometries: CsPb₂X₅ and Cs₄PbX₆.² The Cs₄PbX₆ phase has a crystal structure that is very different from that of CsPbX₃; while the [PbX₆]⁴⁻ octahedra in Cs₄PbX₆ are still surrounded by eight Cs⁺ atoms, which is similar to the cubic LHP phase, the halides in Cs₄PbX₆ are no longer shared between [PbX₆]⁴⁻ octahedra. Furthermore, the Cs⁺ atoms in Cs₄PbX₆ are no longer all occupying identical crystallographic sites but actually two distinct sites.²⁹ Overall, this results in a rhombohedral (R3̅c) Cs₄PbX₆ phase, as shown in Figure 1a. An alternative way to perceive Cs₄PbX₆ is to take the cubic CsPbX₃ phase and remove 3/4 of the PbX₂ while still maintaining the cubic framework of Cs⁺ ions and keeping the remaining PbX in [PbX₆]⁴⁻ octahedra. Due to the large voids that are created by the removal of PbX₂, the system strongly compresses and distorts, resulting in a Cs₄PbX₆ phase.²⁹ In contrast to the cubic LHP phase, 0D cesium lead halides seem to be less sensitive to the ratios of the ionic radii of the A, Pb, and X components as a wide variety of A–Pb–X systems can crystallize in the rhombohedral A₄PbX₆ structure (with A = Cs, Rb, K; X = F, Cl, Br, I).⁴⁴ This general rhombohedral A₄BX₆ structure is often referred to as the “K₄CdCl₆ structure”,⁴⁴ and it can be found in a wide range of metal halides (M = Sn, Cd, Mn, Mg, Ca, etc. and with A = (NH₄), Cs, Rb, K,) with a set of formation rules that are similar to those of the Goldschmidt tolerance factor and are based on the ionic B/X and A/X radii.⁴⁵ Although it is beyond the scope of this review, the 0D A₄BX₆ crystal structure can even be found in oxides such as Sr₄IrO₆ and Ca₄RuO₆.⁴⁶

Figure 1.

Properties of Cs₄PbX₆ and CsPbBr₃. (a) Crystal structures of cubic CsPbX₃ and rhombohedral Cs₄PbX₆ (shown in its primitive cubic cell). (b) Comparison of orbital overlap between the p orbitals of each Br^– anion and the s and p orbitals of Pb²⁺ cations, showing the strong orbital overlap of the [PbBr₆]^4– clusters in CsPbBr₃ and the decoupling in Cs₄PbBr₆, courtesy of Urko Petralanda. (c) Comparison of the absorption spectra of Cs₄PbX₆ and CsPbBr₃ NCs, showing the large bandgaps and strong excitonic absorption for Cs₄PbX₆, adapted from ref (13). (d) DFT density of state calculations for CsPbBr₃ and Cs₄PbBr₆, adapted from ref (13).

Due to the (almost) perfect linear alignment of the Pb²⁺ ions and the halide ions in the LHP framework, the p orbitals of each X^– ion have a good overlap with the s and p orbitals of the two Pb²⁺ atoms with which it is shared (Figure 1b). This orbital overlap leads to a large orbital hybridization, resulting in a marked decrease in the bandgap from a single [PbX₆]^4– cluster (about 4 eV in solution for [PbBr₆]^4–)¹⁶ to CsPbBr₃ (about 2.15 eV, bulk). Due to the symmetry of the Cs₄PbX₆ phase, which is lower than the cubic/orthorhombic CsPbX₃ phase, one [PbX₆]^4– octahedron in Cs₄PbBr₆ has two different [PbX₆]^4– octahedra as its nearest neighbors, with Pb–Pb distances of 8.7 and 10 Å, respectively (compared to a 5.9 Å Pb–Pb distance between neighboring [PbX₆]^4– octahedra in orthorhombic CsPbBr₃).²⁹ As mentioned above, in Cs₄PbX₆, the [PbX₆]^4– octahedra do not share halides; thus, they exhibit negligible orbital overlap between neighboring octahedra, as shown in Figure 1b. As was reported in 1893 alongside with the first works on CsPbX₃, Cs₄PbX₆ powders are colorless, independent of the type of halide.² Indeed, the reported bandgaps are all in the UV region, with Cs₄PbCl₆ = 4.37 eV, Cs₄PbBr₆ = 3.95 eV, and Cs₄PbI₆ = 3.38 eV, as shown in Figure 1c.^3,6,8,9 The decoupled [PbX₆]^4– octahedra in Cs₄PbX₆, compared to the halide-coupled [PbX₆]^4– octahedra in LHP, cause the bandgap of Cs₄PbX₆ to shift toward values of free [PbX₆]^4– clusters in solutions and essentially lead to single molecule-like, excitonic absorption bands.^13,16,29 The fact that Cs₄PbX₆ phases have bandgap values close to their respective free [PbX₆]^4– clusters is also evident from the optical properties of a series of alkali metal halides (AX, A = Na, K, Rb, Cs; X = Cl, Br, I) doped with Pb²⁺.⁴⁷ For these materials, the spectral position of the excitonic absorption is almost independent of the alkali metal (NaBr:Pb²⁺ = 304 nm, CsBr:Pb²⁺, = 313 nm Cs₄PbBr₆ = 314 nm),^47,48 while it is strongly dependent on the halide ion. The single-cluster behavior of Pb²⁺ in Cs₄PbBr₆ is also supported by various density functional theory (DFT) calculations, which have confirmed that Cs₄PbX₆ compounds have large bandgaps (Figure 1d).^{13,17,19,29,41} That the dimensional reductions of the Pb–X–Pb framework in LHPs have a strong effect on the optical properties is also evident in the lead iodide perovskites. For instance, cubic CsPbI₃ is thermally unstable at room temperature, and it undergoes a phase transition to an orthorhombic phase. Here, the lead iodide octahedra are disconnected and reordered into linear chains so that the dimensionality in the interconnection of the [PbX₆]^4– octahedra is reduced from 3D to 1D,⁴⁹ and an increase in the bandgap from 1.73 to 2.25 eV can be observed. Similarly, cubic FAPbI₃ undergoes a transition to a more stable nonperovskite hexagonal phase at low temperatures, with the lead iodide octahedra completely disconnected, resulting in the bandgap being significantly higher than that in the cubic phase.⁵⁰ Moreover, 2D Ruddlesden–Popper phases, as well as layered CsPb₂X₅, exhibit electronic, absorption, and PL properties that are very different than those of the corresponding LHPs.^51,52

Despite the wide bandgap of Cs₄PbBr₆, several groups have recently reported strong green PL as well as absorption in the green region of the spectrum from Cs₄PbBr₆ powders, single crystals, and NCs.^11,17,53,54 This green PL from Cs₄PbBr₆ had already been observed in the early 1990s by Nikl et al., who conducted a series of experiments on CsX:Pb²⁺, Cs₄PbX₆, and CsPbX₃, with the aim of studying the optical properties of Pb²⁺ in different cesium halide matrixes.^3−6,55,56 It was noticed that the Cs₄PbBr₆ absorption at around 4 eV often was accompanied by both weak absorption and strong PL in the green region around 2.4 eV (Figure 2a,b). Nikl et al. concluded that “it is difficult to hold back the occurrence of the CsPbX₃ (X = Cl, Br) phase completely” and that “the unwanted coexistence of these two phases in the Cs₄PbX₆ crystals most probably arises because of an unavoidable incongruent melting crystal growth process”. Kondo et al. performed a similar series of experiments, and also concluded that “Cs₄PbBr₆ crystals usually coexist with the CsPbBr₃ and/or CsBr phases.”^{7−9,57−60} Nikl et al. further studied the coexistence of CsPbBr₃ and CsBr, and they hypothesized that the green luminescence that is seen in Pb-doped CsBr single crystals arose most likely from the presence of CsPbBr₃ quantum dots.⁵⁵ Similarly, CsPbCl₃ NC-like clusters with PL and absorption at around 400 nm could be crystallized inside of CsCl, and CsPbI₃ (in the yellow δ-phase) impurities were observed in Cs₄PbI₆ and CsI.^4,47 The procedure of embedding perovskite NCs in a halide-based matrix was recently also extended to hybrid organic–inorganic perovskites, in which MAPbBr₃ NCs were embedded in a MABr matrix.⁶¹ All of these works indicated that the CsX, Cs₄PbX₆, and CsPbX₃ phases are miscible and often coexist, and that the PL originates from small CsPbX₃ NC-like impurities (Figure 2c,d).

Figure 2.

Overview of studies describing the embedment of CsPbBr₃ in Cs₄PbBr₆. (a) Cs₄PbBr₆ thin films exhibiting appreciable absorption at around 520 nm, (b) accompanied by narrow PL at 550 nm originating from the presence of CsPbBr₃. Adapted from ref (6). (c) Schematic representation of a band structure and absorption/PL spectrum of CsPbBr₃ NCs embedded in a Cs₄PbBr₆ matrix. The absorption spectrum exhibits strong excitonic absorption at around 315 nm due to the Cs₄PbBr₆ host and broad absorption up to 515 nm due to the absorption of the CsPbBr₃ NCs, which emit at the band edge (around 515 nm). (d) Theoretical model for a Cs₄PbBr₆|CsPbBr₃ composite material, with various embedded cubic CsPbBr₃ NCs in a rhombic Cs₄PbBr₆ prism matrix, adapted from ref (62). (e) TEM image and electron diffractogram of a Cs₄PbBr₆|CsPbBr₃ composite, indicating the existence of small CsPbBr₃ NCs in the Cs₄PbBr₆ host, adapted from ref (30).

More recently, small CsPbBr₃ NCs encapsulated in a Cs₄PbBr₆ matrix were found to have high PL quantum yields (PLQYs) and were used to fabricate temperature-insensitive frequency-upconverted lasers, as was proven by HRTEM.^32,62 Xu et al. proposed a type-I heterostructure for such a system, and they prepared a light-emitting diode (LED) in which the carriers from the Cs₄PbBr₆ host were injected into the CsPbBr₃ NCs.³⁰ In addition, another LED based on silica encapsulation of the Cs₄PbBr₆-passsivated CsPbBr₃ NCs was also reported.³⁸ Moreover, several groups have used transmission electron microscopy (TEM) to identify the presence of green luminescent CsPbBr₃ NCs in Cs₄PbBr₆ microcrystals, as well as in Cs₄PbBr₆ NC solutions (Figure 2e).^30,31,34,62

Although all of these accounts point toward the coexistence of CsPbX₃ aggregates in CsX and Cs₄PbX₆, more recent literature has proposed alternative reasons for the green PL in Cs₄PbX₆. These recent works start with the assumption that there is no clear evidence that CsPbX₃ is present as the X-ray diffraction (XRD) patterns of Cs₄PbBr₆ show no peaks that are ascribable to CsPbBr₃ and elemental analysis yields Cs:Pb:X ratios much closer to 4:1:6 than 1:1:3.^{11,12,17,21,35,53} Many of these works provide an alternative explanation for the origin of green PL and absorption, namely, the presence of structural defects in Cs₄PbBr₆, which have been previously reported to strongly influence the PL in 3D and 2D (layered) perovskites.^17,29,63,64 De Bastiani et al. proposed that in Cs₄PbBr₆ the luminescence originates from the presence of bromide vacancies (V_Br), which form shallow or deep trap states within the bandgap (Figure 3a,b).²¹ Feng et al. proposed another type of defect, which arises from the incorporation of −OH (hydroxide) groups into Cs₄PbBr₆.²⁹ Through DFT calculations, they indicated that the incorporation of −OH can form a 2.6 eV sub-bandgap state in Cs₄PbBr₆. Recombination from midbandgap states due to crystal defects was also hypothesized as being the origin of the green PL in Cs₄PbBr₆ microdisks.²⁰

Figure 3.

Overview of proposed defect emission in Cs₄PbBr₆. (a) Sketch of a bromide vacancy in Cs₄PbBr₆. (b) Schematic representation of a band structure with a midbandgap defect state, causing midbandgap emission. (c) Defect-related PL emission in a layered 2D organic lead bromide (N1-methylethane-1,2-diammonium)PbBr₄, adapted from ref (98). (d) Absorption, PL, and PLE from green-emitting 26 nm Cs₄PbBr₆ NCs, showing a narrow PL fwhm and strong quenching of the PL at around 315 nm, adapted from ref (14). (e) Absorption and PL of a green-emitting Cs₄PbBr₆ single crystal, showing narrow band-edge-like PL, adapted from ref (21). (f) Absorption and PL from CsPbBr₃ NCs, exhibiting very similar PL and absorption, with narrow PL and a small Stokes shift, adapted from ref (13).

While the formation of defect states is a plausible explanation for the existence of intrinsic green PL in Cs₄PbBr₆, it is not in agreement with the current view of the effect of defects in perovskites. The hypothesis of V_Br as a luminescent center is mainly based on the halogen vacancies being the most prominent type of defect in LHPs that are synthesized under halogen-poor conditions. In halide-based materials, halogen vacancies are generally considered as deep traps, and this, in addition to other factors such as the high hole effective mass, accounts for their low carrier transport properties.⁶⁵ On the other hand, in LHPs, deep traps often have rather high formation energies and are not easily formed, and most shallow trap states that are formed often do not strongly effect on the optical properties.⁶⁶⁻⁶⁸ Therefore, LHPs are often described as “defect-tolerant” materials.^43,66−69 For instance, the superior properties of MAPbI₃ arise from the absence of deep traps, and iodine vacancies (V_I) only form shallow traps.^{66−68,70−74} Similarly, in its bromide counterpart, V_Br are generally described as shallow defects.⁷¹ However, V_Br and chlorine vacancies (V_Cl) in the cubic MAPbX₃ (X = Cl, Br) are significantly deeper traps than V_I in the tetragonal MAPbI₃.⁷² In the case of bromide perovskites, it was concluded that high V_Br concentrations result in a lower radiative recombination efficiency and a consequent decrease in PLQY.⁷⁵ In the case of CsPbBr₃, using a small excess of CsBr precursor (compared to PbBr₂) results in decreased V_Br concentration and, thus, an enhanced PLQY.⁷⁶ The different nature of halide vacancies in perovskites is also evident in lead-free Cs₂AgInBr₆ and Cs₂TlBiBr₆ double perovskites (elpasolites), where halide vacancies have been reported both as shallow (Cs₂AgInBr₆) and deep (Cs₂TlBiBr₆), respectively.^77,78 Furthermore, deep traps and midbandgap states in LHPs are often considered band-to-band PL quenchers.^64,79 Emissive midbandgap states are sometimes observed in LHPs, but only under vacuum, and the observed PL in these cases is weak, broad, and Stokes shifted.⁶⁴ Thus, it is evident that the nature of the halogen vacancies strongly depends on the crystal structure and the chemical nature of the B cation.⁶⁵ Hence, the marked difference in the crystal structures of CsPbBr₃ and Cs₄PbBr₆ raises the question: are defects, in particular, V_Br, in CsPbBr₃ and Cs₄PbBr₆, of a similar nature? Proper calculations on the effect of defects in Cs₄PbX₆ are still lacking, and drawing parallels between halide vacancies in CsPbX₃ and Cs₄PbX₆ should be done with caution. Also, due to the rather different types of bonding of the halides in Cs₄PbX₆, the formation energy of a V_X is likely very different than that in CsPbX₃. For instance, unlike CsPbBr₃ and its hybrid perovskite counterpart, which are usually found to be bromine-deficient,⁸⁰⁻⁸³ Cs₄PbBr₆ is often reported to have a Br:Pb ratio that is usually higher than 6, thus implying that Cs₄PbBr₆ are rather halide rich.^14,18,20,80 One could therefore expect a rather low density of V_Br.

Although it is difficult to experimentally prove or disprove the existence of vacancy-based trap emission in Cs₄PbX₆, one can observe several PL and absorption features in luminescent Cs₄PbBr₆ and compare these with defect emissions in other LHPs. For instance, several marked differences between the reported green PL in an “allegedly” pure 0D Cs₄PbBr₆ and that of a general trap and defect emission in both 2D and 1D organic lead halides, which consist of either layered PbX₆^4– clusters (2D) or linear PbX₆^4– chains (1D), can be seen in the full with at half-maximum (fwhm) and in the Stokes shift of the PL. In the 2D and 1D systems, the defect emissions exhibit a broad PL, often with fwhms larger than 100–200 nm, which is attributed to efficient exciton self-trapping that acts as an excited-state defect instead of a permanent defect (Figure 3c).^84,85 Examples of 2D and 1D organic lead halide materials exhibiting a Stokes-shifted and broad PL include (EDBE)[PbBr₄] (EDBE = 2,2′-(ethylenedioxy)bis(ethylammonium)) and C₄N₂H₁₄PbBr₄.^84,85 Another class of materials with a broad and Stokes-shifted PL are 0D (and also 1D) organic–inorganic Sn²⁺ bromides such as (C₄N₂H₁₄X)₄SnX₆ (X = Br or I).⁸⁵⁻⁸⁷ In these 0D materials, the large organic cations completely isolate each SnX₆^4– octahedron. Thus, these materials are considered to exhibit the intrinsic properties of the individual SnX₆^4– clusters, and the photoluminescent properties are explained to be not as a result of lattice defects but rather due to excited-state structural reorganization within individual SnX₆^4– clusters. This results in them having a long lifetime, in the range of a few microseconds. Hence, Zhou et al. concluded that the SnX₆^4– octahedra could be thought of as either “crystal lattice points” or “molecular species”.⁸⁷ In stark contrast with the optical behavior of these compounds, the green PL in Cs₄PbBr₆ is often reported to have a very narrow fwhm (15–25 nm), as is shown in Figure 3d,e.^12,18,21 This narrow PL matches that of CsPbX₃ NCs (12–42 nm, Figure 3f), and that of the first accounts for CsPbBr₃ NCs embedded in CsBr (0.11 eV, compared to 0.12 eV for colloidal 8 nm CsPbBr₃ NCs).^15,56,88 It is also similar to other confined NCs with band-edge PL.⁸⁹ As mentioned above, the defect emission in a LHP is often observed with a large Stokes shift as defect states are formed midbandgap (Figure 3b,c).

Although the green PL in Cs₄PbBr₆ is sometimes referred to as being Stokes-shifted or as having down-shifted PL (originating from the 3.9 eV bandgap of Cs₄PbBr₆), almost all green luminescent Cs₄PbBr₆ nano/micro/bulk crystals clearly also absorb strongly in the green region (this can be simply deduced by their green/yellow color, Figure 3d,e).^11,18,21,90 If the green PL were to originate from a midgap state or a trap state, then the absorption of the Cs₄PbBr₆ host would remain unaltered, i.e., it would have no features in the visible range. Take, for example, the case of fluorescent, low-dimensional organic–inorganic lead or tin halides,^85,91−94 in which the dimensionality (whether 2D, 1D, or 0D) is tuned by using specially designed amines, like 2,2′-(ethylenedioxy)bis(ethylamine), C₄N₂H₁₄²⁺, or C₄N₂H₁₄Br⁺. In all of these low-dimensional organic lead and tin halides, the strong dimensional confinement results in large bandgaps in the ultraviolet region of the spectrum, and powders of all of these materials are, indeed, white in color. These lower-dimensional organic lead and tin halides can also exhibit down-converted (100–200 nm) emission in the visible spectrum, and their PL excitation (PLE) matches their absorption spectra.^86,87 In the case of the green luminescent Cs₄PbBr₆, the absorption and PLE spectra show features in the visible range (around 510–520 nm), which is in the same range as the green PL. The green PL in the Cs₄PbBr₆ is therefore often reported with a very small Stokes shift of only 28 meV,^11,21 suggesting that the green PL originates from a direct band–band transition in the visible, rather than from emission from a midgap defect state (Figure 3e). Additionally, the PLE spectra from allegedly “pure” Cs₄PbBr₆ and from samples of CsPbBr₃ encased in a Cs₄PbBr₆ matrix are identical.³² In both cases, strong absorption at visible wavelengths along with a sharp absorption decrease (with most cases having no absorption at all) at shorter wavelengths (315 nm) are observed (Figure 3d). This can be explained by the presence of two band edges that correspond to two different materials (Figure 2b). Furthermore, the green PL reported for the “pure” 0D phase was actually completely quenched when excited at the bandgap of Cs₄PbBr₆ (at around 4 eV).¹¹ In the work by Wang et al., in which they reported CsPbBr₃ NCs in a Cs₄PbBr₆ matrix, excitation at around 4 eV resulted in a dwindled green emission, and the spectrum was instead dominated by broad ultraviolet emission (∼370 nm).³² This again indicates that the green PL does not originate from Cs₄PbBr₆ but rather from a lower-bandgap material (CsPbBr₃). In addition, in thin films of CsPbBr₃|Cs₄PbBr₆ composites, the intensity of the absorption in the green region decreases as the concentration of CsPbBr₃ is reduced.⁴⁰ This observation might explain the absence of an absorption peak in the visible range for the small (26 nm) green-emitting Cs₄PbBr₆, for which only an absorption tail is recorded.¹⁴ To our knowledge, no explanation has been given for the absorption in the green region for supposedly pure Cs₄PbBr₆.

The aforementioned defect emission properties (i.e., the large Stokes shift and large fwhm) are also observed in different binary and ternary chalcogenide quantum dots. For instance, selenium vacancies in CdSe NCs give rise to a broad and Stokes-shifted deep trapped emission, and a higher defect emission is observed for smaller NCs due to their larger surface/volume ratio.⁹⁵ Doping chalcogenide NCs can also cause defect emission; a typical example of this can be seen in Ag-doped CdSe NCs, which manifest enhanced band-edge emission compared to the undoped NCs. However, these Ag-doped CdSe NCs also show broad, Stokes-shifted emission, which is attributed to defect emission.⁹⁶ Another example is the donor–acceptor defect emission in highly emissive CuInS₂ NCs, which is characterized by a large Stokes shift and PL fwhm of about 100 nm.⁹⁷

In addition to the directly observable green PL and absorption, several other optical measurements on green luminescent Cs₄PbBr₆ raise questions about its origins. Recent transient absorption measurements on green luminescent Cs₄PbBr₆ films by Yin et al.¹⁹ were interpreted by assuming a short polaron lifetime of ∼2 ps, and the hypothesis of polaronic features in Cs₄PbBr₆ was confirmed by other groups.⁴¹ However, the transient absorption data strongly resemble those of colloidal CsPbBr₃ NCs, as was reported by Wu et al.⁹⁹ In both the works of Yin et al. and Wu et al., at fast delay times (0.2–0.5 ps), an exciton bleach was observed at the lowest-energy excitonic band, and an exciton absorption feature was observed 15 nm above the band edge. Moreover, in both the green luminescent Cs₄PbBr₆ films of Yin et al. and the CsPbBr₃ NCs of Wu et al., the exciton-induced shift almost completely disappeared after 2 ps, and the TA spectra only exhibited exciton bleach at the bandgap. Because Yin et al. did not consider the possibility of inclusions of CsPbBr₃ NCs in their model and they did not compare their results with Wu et al.’s data, it would be interesting to reanalyze the TA measurements taking the consideration of CsPbBr₃ NC-like impurities in mind.

The uncertainty regarding the optical properties of the green-emitting pure Cs₄PbBr₆ extends to PL lifetimes. Here, again, the short PL lifetimes of the green PL from Cs₄PbBr₆ are very similar to those of CsPbBr₃ NCs.^11,21,30,53 For instance, in the work on Cs₄PbBr₆ single crystals, De Bastiani et al. concluded that “This PL lifetime is closer to the lifetime of perovskite-quantum dots (QD) than usual perovskite single crystals.”²¹ Similarly, Saidaminov et al. reported a PL lifetime of Cs₄PbBr₆ powders that is 2 orders of magnitude faster than that of the LHP single crystals.¹¹ On the other hand, results by Cha et al. disagree with the above reports as they measured average PL lifetimes of 19.58 and 2.43 ns for Cs₄PbBr₆ and CsPbBr₃ single crystals, respectively.⁵³ Ling et al. found that the PL lifetime of their CsPbBr₃|Cs₄PbBr₆ composite increased as the Cs₄PbBr₆ ratio increased.⁴⁰ In the case of Cs₄PbBr₆ microdisks, an average PL lifetime of 11.95 ns was recorded at the center of the disks only,²⁰ whereas a shorter PL lifetime (9.26 ns) was recorded at the disks’ edges, indicating that there is a spatial inhomogeneity within the disks. Generally, PL lifetimes of the bulk (single crystals and powders) green-emitting Cs₄PbBr₆, in most reports discussed above and others, are in agreement with PL lifetimes of CsPbBr₃ NCs, suggesting that the high PLQY is due to Cs₄PbBr₆-passivated CsPbBr₃ NCs.^12,62

One final interesting observation is with regard to annealing luminescent Cs₄PbBr₆ single crystals. De Bastiani et al. observed that through annealing aggregates of CsPbBr₃ NCs were formed within the single crystals. It was observed that annealing above 250 °C results in the formation of 10 nm sized CsPbBr₃ NCs (as determined by XRD).²¹ This was explained as triggered by the V_Br defects in Cs₄PbBr₆ samples, which act as “initialization centers”. Although the PLQYs of 10 nm CsPbBr₃ NCs is often found to be in the range of 80–95%,^15,100 it was reported that the formed CsPbBr₃ NCs within the single crystal actually strongly quenches the PL. This is rather contradictory as one would expect an increase in the PLQY when CsPbBr₃ NCs are formed.

As mentioned earlier, the argument that is generally used to prove that the green emission is an intrinsic property of Cs₄PbBr₆ is the lack of CsPbBr₃ diffraction peaks. For instance, no CsPbBr₃ diffractions peaks were reported in various Cs₄PbBr₆ samples, including powders, single crystals, and NCs.^{11,14,17,18,20,21} For instance, De Bastiani et al. were able to demonstrate that the presence of any CsPbBr₃ impurity phase down to 0.5% in weight could be detected by XRD (Figure 4a).²¹ Nevertheless a similar experiment, in which nonluminescent Cs₄PbBr₆ NCs were mixed with a small amount (2% molar, 0.5 wt %) of highly luminescent CsPbBr₃ NCs, indicated that this small amount of CsPbBr₃ did not result in any clear CsPbBr₃ features in the XRD pattern (Figure 4b).¹³ Furthermore, XRD patterns of Cs₄PbBr₆|CsPbBr₃ composites also indicated that there were not any detectable CsPbBr₃ XRD diffractions.^40,62 This is expected as diffractions from CsPbBr₃ NCs should be extremely low in intensity, not only due to their low concentration but also due to the very broad and low-intensity diffraction peaks of NCs. This is clear from the XRD patterns of borosilicate glasses doped with 4 nm CsPbX₃ NCs, which hardly exhibit any perovskite XRD peaks even though the glasses are strongly absorbent and brightly luminescent, indicating that there is a high concentration of NCs.¹⁰¹

Figure 4.

XRD data. (a) XRD calibration experiment indicating that the synthesized Cs₄PbBr₆ single crystals were either pure or contained less than 0.5 wt % of CsPbBr₃, adapted from ref (21). (b) Similar experiment performed with NCs, in which Cs₄PbBr₆ NCs were mixed (2% molar, 0.5 wt %) with CsPbBr₃ NCs, indicating strong green PL after mixing but no detectable CsPbBr₃ XRD diffraction peaks.

Thus, it is very tempting to suggest that the optical properties of the green-emitting Cs₄PbBr₆ originates from CsPbBr₃ NC impurities because they have similar PLs and absorptions, PLQYs, and lifetimes. One more synthetic clue toward the formation of CsPbBr₃ NCs inside of bulk Cs₄PbBr₆ is derived from the results of the synthesis of nonluminescent Cs₄PbX₆ NCs.^13,33,37,102 These Cs₄PbX₆ NCs exhibit no PL and no absorption in the visible region, with only strong excitonic absorption in the UV, and can be synthesized in the 10–100 nm size range. These Cs₄PbX₆ NCs are spherical or hexagonally shaped for small NCs (<20 nm) and often rhombohedral for larger NCs, which strongly differs from the highly cubic CsPbX₃ NCs, and strongly reflects their hexagonal crystal structure.^13,33,37 Interestingly, these NCs are often synthesized under similar conditions to those of the highly luminescent CsPbBr₃ NCs, but they use a higher amount of alkylamine. Here, the nucleation of CsPbBr₃ can be completely suppressed because oleylamine strongly binds to Pb²⁺.^33,37,102 Even though these NCs are synthesized under stoichiometric conditions closer to CsPbBr₃ (Cs:Pb:Br = 2.2:1:2),¹³ only the Cs₄PbBr₆ phase was formed; therefore, no green PL was observed. This would then explain how both green luminescent and nonluminescent Cs₄PbBr₆ NCs could be synthesized by the simple addition of more ligands; the increased ligand amount suppresses the formation of CsPbBr₃.¹⁷ The Cs₄PbX₆ NCs were also the first to be reported with mixed halide compositions, as confirmed with both absorption and XRD.^13,37 These mixed halide Cs₄PbX₆ NCs exhibit rather broadening of their exitonic absorption compared with the single halide Cs₄PbX₆ NCs. The broadening of the absorption is a direct result of the individual decoupled, mixed halide [PbBr_nX_6–n]^4– clusters, which can only be populated with integer n = 0, 1, 2, up to 6, resulting in broad absorption due to the contributions of different [PbBr_nX_6–n]^4– octahedra within the NC.¹³

Although the green PL in Cs₄PbBr₆ can indeed be explained by the presence of small CsPbBr₃ NCs, there are still a few observations that require further scrutiny. One striking and consistent feature of the green PL in the reported luminescent Cs₄PbBr₆ is its peak position. For many Cs₄PbBr₆ single crystals, powders, and NCs, the PL is always found in a rather narrow range, from 515 to 524 nm. Although this is coincidentally in the same range as for 8–15 nm sized CsPbBr₃ NCs, it would mean that the formed CsPbBr₃ NCs would always be in the same range of sizes and would never be formed with a size smaller than ∼5 nm as this would result in blue-shifted PL due to the quantum confinement effects on CsPbBr₃. However, our group has recently demonstrated that the size of the formed CsPbBr₃ NCs is highly dependent on the synthetic procedure, including synthesis temperature and ligand ratios.¹⁰² Blue-emitting Cs₄PbBr₆ composites were only very recently reported by Chen et al., who were able to synthesize both blue- and green-emerging Cs₄PbBr₆ cm-sized single crystals embedded with CsPbBr₃ NCs.³⁴ Here, it was observed that during the crystallization first small blue-emitting NCs were formed. Interestingly, it was reported the blue PL slowly shifted toward the green, even if the single crystals were isolated from their growth media, thus indicating that small blue-emitting CsPbBr₃ NCs are thermodynamically less stable than the larger green-emitting CsPbBr₃ NCs. Although it was speculated that the CsPbBr₃ NCs were formed due to Pb²⁺ vacancies,³⁴ a deep study of the formation and growth of these embedded CsPbBr₃ NCs is still lacking.

One additional open question is whether the high PLQY in Cs₄PbBr₆|CsPbBr₃ composites is due only to the formation of CsPbBr₃ NCs or whether it also arises from a synergistic effect at the CsPbBr₃|Cs₄PBr₆ interface. Quan et al. proposed that the lattice matching at the CsPbBr₃|Cs₄PBr₆ interface improves the surface passivation of CsPbBr₃ NCs.⁶² Ling et al. suggested, however, that the shallow traps that are generated at the CsPbBr₃|Cs₄PbBr₆ interface cause an enhancement in the PL.⁴⁰ Furthermore, Xu et al. suggested that the change in dielectric constant from the CsPbBr₃ NCs to the Cs₄PbBr₆ host could result in an increase of the oscillator strength of the excitons in the imbedded nanocrystal, which increases the radiative decay.³⁰ It is vital, therefore, to understand whether the CsPbBr₃|Cs₄PbBr₆ interface further enhances the PL properties of the CsPbBr₃ NCs or not as this might lead to a better understanding of the requirements for encapsulating CsPbBr₃ NCs in other materials while maintaining high and stable PLQYs. Finally, it is interesting to note that the CsPb₂Br₅ phase, which is a layered 2D structure, is currently facing the same green PL discussion as Cs₄PbBr₆. Various different optical properties have been reported, ranging from a nonluminescent wide indirect bandgap material^52,103−105 to a green material with strong green PL (again, around 515 nm).^22,106−109

Cs₄PbX₆ phases also exhibit very different electronic properties from CsPbX₃ due to their marked structural difference from CsPbX₃. In contrast to the highly conductive perovskite phase of CsPbX₃,¹¹⁰ the decoupling of the PbX₆^4– octahedra in Cs₄PbX₆ and the large increase in distance between these octahedra carriers strongly confine the carriers to the single octahedra, meaning that Cs₄PbX₆ NCs essentially act as insulators.^53,111 A similar argument stands for Cs₄SnBr₆, which has the same crystal structure as Cs₄PbBr₆. This material was reported to have an electrical conductivity that is more than 200 times lower than that of CsSnBr₃.¹¹¹ A single crystal of Cs₄PbBr₆ was found to have ultralow photoconductivity (on the order of nA).⁵³ One work did use the luminescent Cs₄PbBr₆ (reported as a CsPbBr₃|Cs₄PbBr₆ composite, with a PLQY of 30%) for an LED.³⁰ The composites showed an increase in the external quantum efficiency (EQE) compared to that of CsPbBr₃, presumably due to the increase in the PLQY; the achieved maximum EQE was only 10^–3%, which is several orders of magnitude lower than that of LEDs using CsPbBr₃ NCs. This demonstrates that, even though CsPbBr₃|Cs₄PbBr₆ composites have a high PLQY, their lack of conductivity hinders their use in LEDs. On the other hand, Tong et al. recently demonstrated that the addition of Cs₄PbBr₆ to a MAPbI₃|CsPbBr₃ photodector led to increased and faster deep ultraviolet detection.⁴² However, it is important that the electronic and photoconductivity are first properly characterized, before Cs₄PbX₆ can be considered as an interesting material for electron injected applications such as LEDs and photodetectors. One recent work by Yin et al. indeed confirmed that “intrinsic large bandgap, heavy charge carriers, and low electrical conductivity of 0D perovskites limit their application in photovoltaic devices.”¹⁹

As a result of the stable and high PLQYs, Cs₄PbX₆|CsPbBr₃ materials are still interesting for down-converting applications like down-conversion LEDs, in which the emissions of the material are achieved via excitation with a blue LED. To this end, there have been only very few publications that have reported proper devices.^32,34 For instance, Chen et al. recently combined a highly luminescent Cs₄PbBr₆|CsPbBr₃ single crystal with a K₂SiF₆:Mn⁴⁺ phosphor as a red emitter and a blue-emitting GaN chip to create high-quality white light with luminous efficiency of ∼151 lm W^–1 and color gamut of 90.6% Rec. 2020 at 20 mA, which was reported to be “much better than that based on conventional perovskite nanocrystals.” Furthermore, Wang et al. demonstrated that the Cs₄PbX₆|CsPbBr₃ composites exhibited a temperature-insensitive gain, and they used this for a vertical cavity surface emitting laser, which could operate at temperatures as high as 100 °C.³² One other problem with having Cs₄PbX₆ as the host material for CsPbX₃ is the toxicity of the Cs₄PbX₆. If the active luminescent CsPbBr₃ is only several wt %, or even less, the majority of the host (Cs₄PbBr₆) will be inactive. Consequently, this would strongly limit the use of these materials in applications due to the Pb²⁺ toxicity. One alternative could be to find new synthesis approaches that are similar to those used for CsBr|CsPbBr₃ composites. These types of alkali metal matrixes, as discussed earlier, were already used in the 1990s in the first reports of CsPbBr₃ quantum dots, which were imbedded in CsBr, and were also recently used for organic MABr|MAPbBr₃ composites.^55,56,61,112 Although others concluded that traces of CsBr in Cs₄PbBr₆|CsPbBr₃ quenched the PL,⁴⁰ it would be of great interest to revive these types of host materials for CsPbBr₃ as it would significantly lower the required amount of lead. Recently, presynthesized CsPbX₃ NCs were, for instance, embedded in potassium halide salts, and they exhibited high PLQYs and long-term stability, which demonstrates the advantage of embedding LHP NCs in other halide-based salts.¹¹³ Novel synthesis approaches for pure CsX NCs have recently been reported and could aid in designing synthesis routes toward CsBr|CsPbBr₃ composites.¹¹⁴ Furthermore, devices have so far been limited to Cs₄PbBr₆|CsPbBr₃, and it remains to be seen whether or not other halide-based composites, like Cs₄PbI₆|CsPbI₃, are stable enough to use in devices.

Although Cs₄PbX₆|CsPbX₃ composites exhibit a low potential for applications, Cs₄PbX₆ NCs can be transformed into CsPbBr₃ NCs by various reactions (Figure 5).^13,27 This can be done via the insertion of PbBr₂, the extraction of CsBr with water or with Prussian Blue, or the extraction of Pb²⁺ with amines.^{13,27,28,33,37} Several groups have reported that the size and shape of the starting Cs₄PbBr₆ NCs can be used to tune the size of the final CsPbBr₃ NCs (Figure 5b).^{13,27,28,33,37} The transformed CsPbBr₃ NCs exhibit a strong green PL, similar to the PL of directly synthesized CsPbBr₃ NCs, and they show no excitonic absorption in the UV from their parent Cs₄PbBr₆ NCs (Figure 5c). Similarly, no Cs₄PbBr₆ diffractions are observed in XRD analyses after the transformation (Figure 5d). Interestingly, the formed CsPbBr₃ NCs can be transformed back into Cs₄PbBr₆ NCs via the addition of oleic acid (Figure 5c,d).³³ One interesting advantage of using this pathway to make CsPbX₃ NCs is that the Cs₄PbX₆ NCs can be synthesized with a large tunability over their size (from about 10 up to 50 nm), while remaining nearly monodisperse. The narrow size distribution of the Cs₄PbBr₆ NCs can be used to make large (20–50 nm), monodisperse CsPbBr₃ NCs, which currently still remains a challenge to overcome by direct CsPbBr₃ NC syntheses. Finally, Hu et al. have used a Cs₄PbBr₆ to CsPbBr₃ exchange reaction (via the extraction of Cs⁺ with water) to synthesis the CsPbX₃|SiO₂ and CsPbBr₃|Ta₂O₅ Janus NCs,³⁶ which are among the first examples of colloidal LHP NC heterostructures.

Figure 5.

Overview of transformation reactions of Cs₄PbBr₆ into CsPbBr₃ and back. (a) Proposed reaction mechanisms for the exchange reaction via the extraction of CsBr or the insertion of PbBr₂. (b) TEM image showing the preservation of NC size from Cs₄PbBr₆ NCs to CsPbBr₃ NCs, adapted from ref (13). (c) First excitonic absorption peak and (d) XRD pattern of NCs after several reversible exchanges from CsPbBr₃ to Cs₄PbBr₆ and back, adapted from ref (37).

Due to the insulator bandgap and the very low conductive properties of Cs₄PbX₆ (either luminescent or nonluminescent), its widespread use in applications remains to be seen. On the bright side, the field of zero-dimensional perovskites gives us inside knowledge on how to synthesize and stabilize CsPbX₃ NCs, without the aid of large, bulky, nonconductive ligands, directly into powders and single crystals, which still remains challenging with regard to colloidally synthesized CsPbX₃. While defects such bromine vacancies are proposed as being the origin of the green PL, additional calculations and optical studies will be required to assess the nature of defects in Cs₄PbBr₆ and of radiative decay mechanisms. Also, more extensive studies will have to correlate the properties of Cs₄PbBr₆ with those of other all-inorganic 0D phases, such as, for example, CsPb₂Br₅, Cs₂SnX₆, and Cs₃Bi₂I₉.

Biographies

Quinten Adriaan Akkerman is a Ph.D. student in the nanochemistry department at Istituto Italiano di Tecnologia (IIT), Genova, Italy. He obtained his bachelor’s degree in Chemistry (2012) and master’s degree in Chemistry and Physics (2014) at the University of Utrecht, Netherlands. Here, as part of his master thesis, he worked on the size, shape, and composition control of copper chalcogenide nanocrystals, under the supervision of Dr. Celso de Mello Donega. In 2015, he joined Liberato Manna’s group as a Ph.D. student at IIT Genova, where he currently is focused on the development of novel synthesis methods and the characterization of perovskite and perovskite-related nanocrystals.

Ahmed L. Abdelhady holds a Ph.D. in inorganic chemistry from The University of Manchester, United Kingdom (2011), under the supervision of Prof. Paul O’Brien. After he completed his Ph.D., he became a lecturer of inorganic chemistry at Mansoura University, Egypt. In 2013, he joined Prof. Osman Bakr’s Functional Nanomaterials Lab at King Abdullah University of Science and Technology (KAUST), Saudi Arabia. In 2016, he joined Prof. Liberato Manna’s group at the Italian Institute of Technology (IIT) as a postdoctoral fellow and then as a researcher in 2017. His research is focused on all-inorganic and hybrid perovskite single crystals and thin films for optoelectronic applications.

Liberato Manna received his M.Sc. in Chemistry from the University of Bari (Italy) in 1996 and his Ph.D. in Chemical Sciences from the same University in 2001. During his Ph.D. studies and later as a postdoctoral fellow, he worked at the University of California Berkeley (U.S.A.). In 2003, he moved back to Italy as staff scientist at the National Nanotechnology Lab of CNR-INFM in Lecce (Italy) where he later became responsible for the Nanochemistry Division in 2006. In April 2009, he moved to the Istituto Italiano di Tecnologia in Genova as head of the Nanochemistry Department.

The authors declare no competing financial interest.