Defect Passivation in Lead‐Halide Perovskite Nanocrystals and Thin Films: Toward Efficient LEDs and Solar Cells

Abstract Lead‐halide perovskites (LHPs), in the form of both colloidal nanocrystals (NCs) and thin films, have emerged over the past decade as leading candidates for next‐generation, efficient light‐emitting diodes (LEDs) and solar cells. Owing to their high photoluminescence quantum yields (PLQYs), LHPs efficiently convert injected charge carriers into light and vice versa. However, despite the defect‐tolerance of LHPs, defects at the surface of colloidal NCs and grain boundaries in thin films play a critical role in charge‐carrier transport and nonradiative recombination, which lowers the PLQYs, device efficiency, and stability. Therefore, understanding the defects that play a key role in limiting performance, and developing effective passivation routes are critical for achieving advances in performance. This Review presents the current understanding of defects in halide perovskites and their influence on the optical and charge‐carrier transport properties. Passivation strategies toward improving the efficiencies of perovskite‐based LEDs and solar cells are also discussed.


Introduction
Defects exist in nearly all semiconductors at room temperature. [1] These defects exist at the atomic scale (e.g. vacancies) through to the macroscopic level (e.g. voids or wrinkles). [1c, 2] They play akey role in controlling the optical, electronic,a nd structural properties of semiconductors,a nd thus affect aw ide range of applications. [1a,d,f,2a, 3] Although defects give rise to numerous advantages (e.g. active sites for photocatalysis), the traps they introduce in the band gap are detrimental to the optoelectronic performance of semiconductors. [1a,g] Historically,the emphasis of material growers was to develop synthesis routes to prepare pure and perfect crystalline semiconductors which could be controllably doped with the desired elements to obtain the required electronic properties (e.g.m ajority carrier type and concentration). [1a, 4] Thep resence of defects in certain semiconductors,e ven at al evel of parts per billion, can dramatically alter their properties,s uch as conductivity,c harge-carrier mobility,a nd minority carrier lifetimes.S uch defects can, therefore,l ower the performance of the resulting optoelectronic devices.
In classical semiconductor systems,s uch as Si, GaAs, CdTe, CdSe,InP,and metal chalcogenides,asmall population of defects can be extremely detrimental, as reflected in the substantial decrease in their intrinsic photoluminescence quantum yield (PLQY), carrier mobility,a nd stability. [1f, 5] However,i nterestingly,s ome semiconductors are defecttolerant, which means that it is possible to achieve low nonradiative recombination rates despite high densities of traps in the band gap. [1a, 2c, 6] Defect-tolerant semiconductors were rare and were not widely studied until the recent serendipitous discovery of LHPs. [7] LHPs have the general formula ABX 3 ,w here A = CH 3 NH 3 + ,H C(NH 2 ) 2 + ,C s + ;B= Pb 2+ ;X= Cl À ,B r À ,a nd I À .T his class of halide perovskites shot to prominence over the past decade owing to the rapid increases in efficiency achieved in optoelectronic devices.
Originally,t his was achieved in photovoltaic devices,w hose efficiency rose from 3.8 %i n2 009 [8] to ac ertified 25.5 %i n 2020. [9] Subsequently,efficient performance was also found in LEDs (from ca. 1% external quantum efficiency in 2014 to > 23 %in2 020 [10] ), radiation detectors,and others. [11] Nevertheless,d espite their defect-tolerant nature,t he surface defects and grain boundaries (GBs) present in perovskite thin films have been found to be detrimental to Lead-halide perovskites (LHPs), in the form of both colloidal nanocrystals (NCs) and thin films,have emerged over the past decade as leading candidates for next-generation, efficient light-emitting diodes (LEDs) and solar cells.Owing to their high photoluminescence quantum yields (PLQYs), LHPs efficiently convert injected charge carriers into light and vice versa. However,despite the defect-tolerance of LHPs,defects at the surface of colloidal NCs and grain boundaries in thin films play acritical role in charge-carrier transport and nonradiative recombination, whichl owers the PLQYs,d evice efficiency,a nd stability.T herefore,u nderstanding the defects that play akey role in limiting performance,and developing effective passivation routes are critical for achieving advances in performance.T his Review presents the current understanding of defects in halide perovskites and their influence on the optical and charge-carrier transport properties.P assivation strategies toward improving the efficiencies of perovskite-based LEDs and solar cells are also discussed. the performance of the resulting optoelectronic devices. [2b,c,d,6c, 12] Thel atest advances in efficiency require careful passivation strategies. [13] Similarly,i nc olloidal perovskite NCs,t he surface defects arising from the removal of ligands and surface halides during purification significantly reduce the PLQY. [14] In particular, surface defects play acritical role in the optical properties of colloidal NCs because of their high surface area to volume ratio. [14,15] Therefore,i ti so fu tmost importance to understand the types of defects that are common at surfaces as well as their role in the optoelectronic properties of perovskite thin films and colloidal NCs to advance their applications in solar cells and LEDs. [2b, 12a, 14c, 16] Toward this goal, aw ide range of surface passivation strategies have been developed using al arge variety of organic and inorganic molecules for LHP NCs as well as thin films. [2b, 12a, 14b,c,16a, 17] In this Review,f irst we present af undamental understanding of the surface chemistry and defects in perovskite NCs and thin films.T he defect-tolerance of different compositions of LHPs is discussed, as well as the types of defects that commonly occur.W et hen provide an overview of surface passivation strategies for both colloidal NCs and thin films toward improving their optical properties and thus the efficiencies of the resulting LEDs and solar cells. Finally,w ep resent ab rief outlook, highlighting the open questions and remaining challenges in this research field.

Surface Chemistry and Defects in Thin-Film and Colloidal Perovskites
As discussed in the introduction, surface chemistry plays acritical role in the optical and electronic properties as well as the stability of LHPs.T herefore,t he surface chemistry of LHPs has been heavily investigated but is still not fully understood. [15,16,18] During the crystallization of perovskite thin films,t heir surfaces can terminate with PbX 2 or AX or both, depending on the film processing conditions.T he electronic band structure of the surface could be different from that of the bulk, and the type of surface termination can significantly influence the band alignment with other molecular systems. [19] Although AX termination is thermodynamically stable,itcan be changed to PbX 2 when organic-inorganic hybrid perovskite films are exposed to moisture because of differences in their interaction with water molecules.T his is because the AX termination can undergo fast hydration compared to the PbX 2 termination because of the strong PbÀ Xb ond.
However,t he presence of higher humidity can lead to complete hydration of the surface,r egardless of the type of surface termination. To improve the stability of LHP films by increasing the surface resistance to moisture,l ong-chain alkylammonium halide molecules have been incorporated into perovskite films,t hereby giving rise to 2D halide perovskites. [20] These molecules can also passivate the surface defects in LHP films.Like in classical semiconductors,two of the main classes of defects are point defects (atomic-scale) and structural defects.C ompared with ap erfect lattice (Figure 1a, (atoms absent from their lattice sites;F igure 1a,i i-iv), interstitial defects (atoms located between lattice sites; Figure 1a,v -vii), and anti-site defects (atoms occupying another species lattice site;F igure 1a,i x-xiv). Structural defects (extended defects) include dislocations (Figure 1a, xv) and GBs (Figure 1a,x vi). In addition, defect complexes can occur.T hese include Schottky defects (pair of anion and cation vacancies;F igure 1a,x vii)a nd Frenkel defects (a pair of an interstitial and av acancy created from the same ion; Figure 1a,x viii). Beyond these intrinsic defects,e xtrinsic impurities can also occur ( Figure 1a,v iii, xix). Unlike conventional semiconductors such as Si, CdTe, or GaAs, which require low defect concentrations to mitigate the deleterious effects of their deep traps,t he common point defects in perovskites (A-and X-site vacancies) are usually shallow-level defects (at least for Br-o rI-based perovskites) and are not detrimental to their device performance.P oint defects with deep traps such as interstitial or anti-site defects are almost absent in perovskites,s ince they have high formation energies. [22] Therefore,t he types of defect which we should consider during synthesis and passivation are more likely to be uncoordinated Pb 2+ ions and surface charged defects. [22,23] More details will be discussed in Section 2.1.
On the other hand, the surface chemistry of colloidal perovskite NCs is more complex in the presence of capping ligands on the surface (Figure 1b). [6e, 18] Most often, long-chain amines in combination with along-chain acid have been used as ligands in the synthesis of colloidal perovskite NCs.One of the most commonly used amine-acid pair is oleylamine and oleic acid. In this base-acid pair, the proton is exchanged from the acid to the amine,l eading to the formation of an ammonium cation (R-NH 3 + ), which then occupies the As ite positions on the NC surface ( Figure 1b). Theoretical studies suggest that ammonium cation ligands replace 50 %o ft he "A"c ations on the NC surface and these ligands help to maintain the colloidal stability of perovskite NCs in nonpolar solvents. [24] Furthermore,itwas shown experimentally (by Xray photoelectron spectroscopy (XPS) and 1 HNMR spectroscopy) that the R-NH 3 + binds to the surface of CsPbBr 3 NCs by replacing most of the Cs + cations and forming hydrogen bonds with Br À anions. [18b] However,interestingly,it was argued that the R-COO À ions do not bind to the NC surface,but play an important role in the colloidal stability of CsPbBr 3 NCs by forming an equilibrium with surface-bound R-NH 3 + ions and maintaining charge neutrality.
[18b] In contrast, quantitative 1 HNMR studies by Smock et al. suggest that both the oleylamine and oleic acid ligands bind to the surface of CsPbBr 3 NCs. [18a] Similarly,D eRoo et al. showed, by using 1 HNMR spectroscopy,t hat oleic acid binds to the NC surface in the form of oleylammonium oleate only when excess oleylamine is added to the purified NC solution, and this helps to improve the PLQY. [15] Another hypothesis is that the long-chain carboxylic acid molecules replace some of the surface halides to complete the octahedral coordination of lead, and thus passivate the mid-gap defect states (Figure 1b). [21] Amine-free synthesis methods were also reported, in which the nanocrystals were solely capped by oleic acid; [25] this results in colloidal CsPbBr 3 NCs with only oleateterminated surfaces.A lthough the role of oleic acid in the colloidal stability of perovskite NCs is still under debate,itis widely accepted that alkylammonium cation ligands bind to the surface of perovskite NCs.H owever,l igand binding to aperovskite NC surface is highly dynamic and, therefore,the surface is properly passivated only in the presence of excess free ligands in the solution. Because they are labile,s urface ligands often desorb from the perovskite NC surface upon ageing, dilution, or purification by washing, and these processes lead to the formation of surface defects,a nd thus ar eduction in the PLQY. [15] It is worth mentioning that asmall amount of additional ligands need to be added during the purification to maintain the colloidal stability and high PLQY of perovskite NCs. [15] To overcome these difficulties, various ligands,s uch as quaternary alkylammonium cations, alkyl phosphonates,a nd zwitterionic molecules,h ave been tested as alternatives to the oleylamine/oleic acid pair (Figures 1b). [26] It is well-established that the stability and PLQY of perovskite NCs strongly depend on the binding affinity of the capping ligands.R ecent studies suggest that ligands with multiple functional groups with opposite charges (e.g. zwitterionic molecules such as 3-(N,N-dimethyloctadecylammonio)propane sulfonate) tightly bind to the perovskite NC surface and thus significantly improve the stability.Itwas hypothesized that the ammonium cation of the zwitterionic molecules occupies the A-site position, while the carboxylate or sulfonate group simultaneously binds to uncoordinated Pb to complete its octahedral coordination (Figure 1b) thus suggesting that these types of ligands bind strongly to the perovskite NC surface. [26g] In principle,p erovskite NCs also exhibit similar types of defects as bulk thin films ( Figure 1a). Additionally,p erovskite NCs exhibit surface vacancies caused by the desorption of ligand molecules (Figure 1b). To address this,and owing to the flexible surface chemistry of perovskite NCs,awide range of post-synthetic defect-passivation strategies have been developed to improve the stability and PLQYs,w hich is discussed in detail in Section 3. Thedefects in both thin films and NCs can cause either shallow or deep states in the electronic band structure.D efects in as pecific material are generally studied by calculation of their formation energies in different charge states.Ithas been said that the vacancies (A, Pb,a nd Xp oint defects) in perovskites form shallow traps because of their low formation energy.H owever,i nterestingly,s urface halide vacancies can significantly affect nonradiative recombination rates depending on the type of halide,w hich are discussed in detail in the following section. Figure 1. Defects in lead-halide perovskites. a) Types of defects in lead-halideperovskites:i )perfect crystals, ii)A-site vacancy,i ii)X-site vacancy, iv) B-site vacancy,v)A-site interstitial,v i) X-site interstitial,v ii)B-site interstitial,v iii)impurity interstitial, ix) A B anti-sites, x) A X anti-sites, xi)X B anti-sites, xii)X A anti-sites, xiii)B A anti-sites, xiv) B x anti-sites, xv) lattice dislocations, xvi)grain boundary defects, xvii)Schottky defect (anion and cation vacancies occurring together), xviii)F renkel defect (interstitial and vacancy created from the same ion), xix) impurity substitution. b) Schematic illustration of the ligand-capped LHP nanocrystal and an enlarged view of the nanocrystal surface with defects and different types of surface ligands used for the stabilization of NCs, includinga lkylammonium ions, carboxylic acid, and zwitterions such as amino acids. The Figure is inspired by Ref. [21].

Defect Tolerance in Perovskites
Akey enabling feature of lead-halide perovskites is their defect tolerance,i nw hich low nonradiative recombination rates are achieved despite high densities of defects. [6a,e,27] Owing to their defect tolerance,l ead-halide perovskite thin films and nanocrystals that have defect densities several orders of magnitude higher than silicon or GaAs could still realize efficient solar cells and light-emitting diodes. [22,28] This is because the trap levels from the most common defects are shallow and have low capture cross-sections,a nd can be understood from the classic equation for Shockley-Read-Hall (U SRH )r ecombination [Eq. (1)], [6e] where n and p are the electron and hole carrier concentrations,respectively; n 1 and p 1 are the occupancyofthe single trap-state modelled;and t 0,h and t 0,e are the low injection hole and electron lifetimes, respectively. [6e] Based on Equation (1), there are three ways to limit U SRH ,a ss hown in Equations (2) and (3): 1) by having alow trap density (N t ); 2) by having low carrier capture crosssections (s t,h and s t,e ); and 3) by having the defect energy levels (E t )far away from the map-gap energy level (i.e.close to the band edge), thereby giving rise to alarger n 1 or p 1 value. In the third case,having atrap close to one band edge would allow it to be highly occupied by one of the carriers (hence ah igh n 1 or p 1 value) but it would be difficult for the other carrier to reach that trap state because of the large energy difference.T hus,n onradiative recombination is reduced. Tr aditionally,m aterial growers have opted for the first option of minimizing N t .Defect tolerance focuses on designing materials that have low capture cross-sections and shallow defects,e nabling low U SRH values despite high N t values,a s shown in Figure 2.
Defect tolerance in the lead-halide perovskite family was originally identified in methylammonium lead iodide (MAPbI 3 ), [29] and rationalized on the basis of the electronic structure. [6b,e] TheP b6sa nd I5po rbitals hybridize to form apair of bonding and antibonding states in the upper valence band, while the Pb 6p orbital hybridizes with the I5porbital to form ab onding state in the upper valence band and an antibonding state in the conduction band minimum. [6e] As ar esult, the original atomic orbitals are close to the band edges rather than inside the band gap (as shown in Figure 2, right). To afirst approximation, defect states form close to the original atomic orbitals,a nd the electronic structure of MAPbI 3 makes it more likely that these states are resonant within the bands or shallow within the band gap.T his contrasts with the electronic structure of traditional semiconductors (e.g. silicon; Figure 2, left), in which the bondingantibonding states form across the band gap and the original atomic orbitals are within the band gap,thereby making deep traps more likely to form. Furthermore,the presence of heavy elements in MAPbI 3 leads to strong spin-orbit coupling, which reduces the band gap,m aking it more likely that the trap states are shallow. [6b,e] Thehigh polarizability of the Pb 2+ ion also leads to ahigh dielectric constant, which favors lower capture cross-sections of charged defects. [6e, 11] Thec ombination of shallowness in the trap energy levels and low capture cross-sections leads to low trap-assisted recombination rates despite high defect densities.However,the defect tolerance of MAPbI 3 has not been found to be generalizable across the entire lead-halide perovskite family ( Figure 3a). Beyond the electronic structure,the crystal structure has also been found to play an important role.For example,CsPbI 3 can form three polymorphs (cubic a-phase,o ro rthorhombic d-a nd gphases). [30] Although the composition of the orbitals and bonding/antibonding states is qualitatively the same as that of MAPbI 3 ,the Pb-I-Pb bond angle reduces from 1808 8 for the aphase to 1558 8 (g-phase) and 958 8 (d-phase).
[30a] Whereas a-CsPbI 3 is defect-tolerant, [30b] the reduced bond angles in the alternative polymorphs lead to ar eduction in the overlap in the orbitals from Pb and I, thus reducing defect tolerance, especially in the d-phase,which forms deep traps. [30a] Indeed, the synthesis of nanocrystals is advantageous because with reductions in size,f ormation of the a-phase becomes more favorable,w hereas the defect-intolerant d-phase would form in bulk thin films or nanocrystals that are too large. [32] Another important structural factor is the lattice parameter. On changing the halide from I À to Br À to Cl À in MAPbX 3 ,t he lattice parameter decreases.T his leads to stronger interactions between Pb 2+ dangling bonds when halide vacancies form, which, when combined with the lower electron affinity,gives rise to deeper trap levels for the halide vacancy.T hus,w hile MAPbI 3 and MAPbBr 3 primarily form shallow traps,M APbCl 3 forms deep traps. [14a, 23, 34] This has also been found in their inorganic counterparts,inwhich cubic CsPbCl 3 forms deep traps and has significantly lower PLQYs than a-phase CsPbBr 3 and CsPbI 3 ,w hich are both defect-tolerant (Figures 3a,b). [6d, 30b,31] Calculations of defects in a-CsPbI 3 showed the defects with the lowest formation energies are lead vacancies (V Pb ), iodine vacancies (V I ), as well as antisite (Pb I and Cs I )a nd interstitial defects (Pb i and Cs i ). However,none of these defects have transition levels (or trap states) within the computed band gap. [30b] We note,however, that the calculated band gap of about 1eVf rom GW calculations is below the experimental band gap of 1.73 eV. [35] Band-gap corrections can lead to the transition levels also shifting in energy,b ut it is likely that the V I transition level falls within the band gap,a sh as been found in computations by other groups (Figure 3b). Similarly, calculations on defects in CsPbBr 3 showed that all defects with low formation energies form shallow transition levels. [6d] These discussions on the role of bulk point defects have been found to be applicable to surfaces,w hich play ac ritical role in nanocrystals.A binitio calculations on CsPbX 3 (X = I À ,B r À ,C l À )s labs with ah alide vacancy introduced on the surface show that the halide vacancy is shallow in CsPbI 3 , slightly deeper in CsPbBr 3 ,a nd close to mid-gap in CsPbCl 3 ( Figure 3b). This is consistent with the trend in the PLQYs decreasing as the halide is changed from ItoCl, as well as the PLQY of CsPbI 3 remaining close to unity as the halide vacancy content increased, whereas the PLQY of CsPbCl 3 sharply decreased. [31] However,t en Brinck et al. argued that understanding defects at the surfaces of nanocrystals computationally requires ad ifferent treatment to the bulk. Their work on CsPbBr 3 nanocrystals,nevertheless,still showed that the surface is defect-tolerant because the defects with the lowest formation energies remain shallow. [14a] 3. Defect Passivation in Perovskite NCs Figure 4a illustrates the formation of surface defects in perovskite NCs either by ageing or washing,a nd their passivation with ligand molecules or metal salts.L igand detachment from the NC surface leads to the formation of both A-site and X-site vacancies for charge neutralization (see Section 2f or more details). There are mainly three types of post-synthetic approaches for the surface passivation of perovskite NCs:1 )addition of metal salts along with ligand molecules,2 )addition of ligand molecules that bind strongly to the perovskite NC surface,and 3) treatment with inorganic salts.T he ligand molecules are expected to fill the A-site and X-site vacancies to recover the fluorescence of the perovskite NCs ( Figure 4a). According to covalent bond classification, these surface ligands are generally divided into L, X, and Z types depending on the number of electrons (2, 1, and 0, respectively) they donate to the metal of aN Cs urface. Because of the ionic nature of perovskite NCs,L-type ligands (cationic and anionic) have mostly been used for the passivation of the Aa nd Xsites of the NC surface.O ne of the early studies on the surface passivation of LHP NCs was reported by Pan et al. 2015; [36] they demonstrated an improvement in the photoluminescence and stability of CsPbBr 3 NC films upon treatment with didodecyldimethylammonium sulfide (S 2À -DDA + ). Thea uthors reported that passivation leads to the formation of ap rotective layer enriched with  Angewandte Chemie Reviews sulfide.However, the passivation mechanism at the molecular level was unclear. Thef irst detailed study of the surface chemistry of perovskite NCs and the highly dynamic nature of ligand binding was reported by Roo et al. in early 2016. [15] They also found that the addition of excess amine improves the binding of carboxylic acid and thus the PLQY.Later,Pan et al. found that exchanging the ligands of the classical oleylamine/oleic acid pair by alkylammonium halides could significantly improve the PLQY and the efficiency of the resulting LEDs. [37] Following these early studies,al arge variety of ligands have been reported for the electronic passivation of perovskite NCs to improve their optical properties;s ome of the ligands are illustrated in Figure 4b. [14b-d, 26d,31, 33, 38] In most studies,i tw as shown that the passivating ligands help to improve the PLQY as well as the stability of perovskite NCs.
Thec hoice of the ligands is generally based on the simultaneous refilling of A-site and halide vacancies with an ammonium cation and the same halide (or with an organic Xtype ligand), respectively [14b, 17a, 18b] This is because the healing of surface defects requires restoration of the damaged surface by filling the halide vacancies and forming acomplete ligand shell. Keeping this in mind, alkylammonium halides have been agood choice for fulfilling this criterion toward effective electronic passivation and improving the stability of perovskite NCs. [14b,17a, 36, 37, 38c, 39] Fore xample,B odnarchuk et al. demonstrated that treating the surface of aged or purified CsPbBr 3 NCs with amixture of didodecyldimethylammonium bromide and PbBr 2 leads to an improvement in the PLQY up to 90-100 % ( Figure 5a). [17a] Thea uthors proposed that the damaged NC surface is repaired by the ammonium cation occupying the A-site and the bromide anion occupying the halide vacancies.S urface passivation has resulted in stable colloidal solutions with PLQYs up to 95-98 %e ven after purifying 3or4times.Another type of ligand family that has received significant attention is zwitterionic ligands (or bidentate ligands), in which the two functional groups with positive and negative charges simultaneously bind to the Asite and X-site,r espectively. [17a, 21, 26g] Therefore,t hese ligands bind to perovskite NC surfaces much more strongly because of the chelating effect, and the resulting NCs exhibit longterm stability and high PLQYs even after washing with different solvents. [17a,26g, 38e] In contrast to using alkylammonium halides and zwitterionic ligands,A livisatos and coworkers proposed that surface treatment with X-type anionic ligands alone can improve the PLQY of CsPbBr 3 up to near unity. [31] They proposed that the X-type ligands passivate the halide vacancies based on hard-soft acid-base theory.T heir results suggest that halide vacancies are the main nonradiative traps in perovskite NCs (see Section 2a nd Figure 3b). This work has led to debate on the role of A-site vacancies on the PLQY of perovskite NCs.Inaddition, acomparison of the PLQY vs.C sPbX 3 NC concentration, that is,t he PLQY vs. density of halide vacancies,f or three different halides (Cl À , Br À ,a nd I À )r evealed that CsPbI 3 NCs are tolerant to halide vacancies,w hereas the CsPbBr 3 NCs are moderately intolerant and CsPbCl 3 NCs are highly intolerant (Figure 3a). This is consistent with the low PLQYs (1-10 %) of pristine CsPbCl 3 NCs compared to other halide compositions. [40] On the other hand, because of their defect-tolerant nature,n oi mprovement is expected in the PLQY of CsPbI 3 NCs upon surface halide passivation. In fact, most studies related to surface passivation and doping of CsPbI 3 NCs have been focused on improving their stability rather than the PLQY. [41] This is because of the structural instability of the optically active aphase cubic CsPbI 3 NCs,w hich generally changes into the nonfluorescent yellow (dÀ)phase that is less defect-tolerant, as discussed previously.However, in some studies,ithas been shown that the PLQY of CsPbI 3 NCs can be improved up to near unity upon surface passivation with bidentate ligands [38e] or by in situ passivation with inorganic salts. [42] It is still unclear whether the ligands passivate the defects or if they improve the phase stability of the CsPbI 3 NCs.H owever, despite their defect-tolerant nature,t he origin of the nonunity PLQY of CsPbI 3 NCs is quite interesting and needs further investigation.
Besides organic ligands,m etal halides [14c, 38a,b,43] and af ew other inorganic compounds [14d, 38f, 44] have also been found to be promising for the passivation of halide vacancies of perovskite NCs to improve their PLQYs and long-term stability.The halide ions of the added metal salts are expected to fill the halide vacancies and thus repair the surface of perovskite NCs.F or example,Bohn et al. [14c] found that postsynthetic treatment of CsPbBr 3 nanoplatelets (NPls) with aP bBr 2 -ligand solution leads to ad ramatic improvement in the PL intensity ( Figure 6). Controlled experiments with different metal halides suggest that both the Pb and Br vacancies are the origin of the lower PLQY of CsPbBr 3 NPls. Theaddition of asolution of PbBr 2 leads to the filling of the Pb and Br vacancies on the surface (Figure 6a). Interestingly, NPl-thickness-dependent surface passivation shows that the PL enhancement factor increases as the NPl thickness decreases (Figure 6b), which suggests ah igher density of halide vacancies with an increase in the surface to volume ratio.Infact, this is consistent with the decrease in the PLQY as the thickness of pristine CsPbBr 3 NPls decreases. [14c, 45] Removal of nonradiative traps by treatment with PbBr 2 is reflected in the increase in the PL lifetime and the shape of the decay profile ( Figure 6c). TheP Ld ecay of the treated sample is more monoexponential compared to the PL decay of the pristine NPl sample.P ost-synthetic treatment of CsPbBr 3 NC films with PbBr 2 led to aP LQY of near unity. [38a] Similarly,an ear unity PLQY of CsPbBr 3 NPls was achieved by Wu et al. by using in situ passivation with as olution of HBr during synthesis. [46] Compared with allinorganic perovskite NCs,i nw hich the surface termination could be either PbX 2 or CsX and the possible surface defects are Pb,C s, or halide vacancies, [17a] Bertolotti et al. suggested that the dominant surface termination of NPls are CsX with Cs and halide vacancies in nearly stoichiometric proportion. [47] Therefore,c ommon passivation strategies for NPls reported so far have focused on X-site vacancy passivation, such as using PbBr 2 or HBr. [14c,46] Furthermore,t reatment with metal halides was found to be very effective for improving the PLQY of weekly emissive CsPbCl 3 NCs.Since the CsPbCl 3 NCs are highly intolerant of Cl vacancies,t hey are weekly emissive.H owever,t reatment with metal chloride leads to the recovery of the damaged PbCl 6 À octahedral coordination and thus significantly improves the PLQY of CsPbCl 3 NCs.F or example,B ehera et al. systematically investigated the influence of the treating the surface of CsPbCl 3 NCs with different metal chlorides,as well as nonmetal chlorides,a nd they observed an enhancement in the PLQY up to 25-50 %. [48] Theo rigin of the PL enhancement was attributed to the presence of ah alide-rich surface after treatment with metal chlorides.M ore importantly,the purification process is very sensitive and the PL can be significantly quenched by ap hase transformation from CsPbCl 3 to CsPb 2 Cl 5 .T hese results further indicate that CsPbCl 3 NCs are highly intolerant of defects,unlike their Br and Ic ounterparts.S imilarly,A hmed et al. reported that the PLQY of CsPbCl 3 NCs could be improved up to 60 %b y surface treatment with YCl 3 . [43] They proposed that both the Y 3+ and Cl À ions contribute to the surface passivation in such away that the Cl À ions bind to under-coordinated Pb ions and the Pb-Cl ion pair vacancies will be filled by Y-Cl. This needs an in-depth investigation of how the Y 3+ ions are doped in the CsPbCl 3 NCs.S imilarly,p ost-synthetic treatment of CsPbCl 3 NCs with CdCl 2 at room temperature leads to an increase in the PLQY from 3to98%. [38d] Theorigin of the enhancement was attributed to the replacement of Pb 2+ by Cd 2+ ions during the cation-exchange process.T he author claimed that CdCl 2 in the saturated ethanol solution reacted with the CsPbCl 3 surface to bind with the surface through Cl vacancies.T hen, the smaller Cd 2+ ion partially replaced the larger Pb 2+ ions and reduced the cuboctahedral voids,t hereby leading to less tilting of the octahedral unit and an increase in the tolerance factor. [49] Therefore,with B-site doping and passivation of the Cl vacancies on the surface by Cl À ions,t he CdCl 2 -treated NCs achieved aPLQY of near unity (96 %).

Improving the Performance of LEDs through Defect Passivation
Lead-halide perovskites have attracted significant attention for applications in LEDs (PeLEDs) owing to the rapid increase in efficiency( from ca. 1% EQE in 2014 to > 20 % EQE in 2020), [50] sharp electroluminescence (EL) bands, color-tunability over the entire visible-wavelength range,and versatile,f acile processability.I np articular,t he EL bands exhibited by LHPs are narrower than those of other organic and inorganic materials,w hich give rise to high color saturation and could lead to displays that emit over aw ider color gamut. Although progress has been rapid for green-, red-, and near-infrared-emitting perovskite LEDs (all with ad emonstrated external quantum efficiency( EQE) > 20 %,  This limits the possibility of achieving all-perovskite ultrahigh-definition displays or using perovskites for solid-state white lighting. As aresult, the past couple of years have witnessed ashift in effort towards blue emitters,and progress has accelerated, with EQEs increasing from 2.12 %in2018 [51] to > 10 %t oday for both blue and sky-blue emitters (Figure 4a). [52] From adevice point of view,the prerequisite for obtaining an LED with ah igh external quantum efficiency( EQE) is governed by four factors:1)charge balance between injected electrons and holes,2)coupling between electrons and holes, 3) the fraction of all recombination processes that are radiative [PLQY; Eq. (4)],a nd 4) the fraction of photons generated that are emitted out of the device (outcoupling). [53] In this section, we will focus on factor 3, to discuss the methods used to obtain high PLQYs through defect passivation routes.D etails of the other factors can be found in Refs. [53,54].
TheP LQY is defined as the ratio between the radiative recombination (AER rad )a nd the sum of radiative recombination and nonradiative recombination (AER non ). In metal halide perovskites,radiative recombination can be from free carriers (bimolecular recombination) or excitons (excitonic recombination). Them echanism for radiative recombination is strongly dependent on the dimensionality of the perovskites and is significantly influenced by the exciton binding energy. In 3D perovskite thin films,t he exciton binding energy is below the room-temperature thermal energy, [55] thereby resulting in excitons readily dissociating into free carriers. Therole of excitons in bulk perovskites has been discussed by Marongiu et al.,w ho claimed that the majority of photoexcitations are from free carriers,with only asmall fraction of Coulombically bound electron-hole pairs under the typical working regimes of solar cells and LEDs. [56] Therefore, radiative recombination in 3D perovskites is dominated by the recombination of free charge carriers and the PLQY can be calculated by Equation (5), where a is the nonradiative trap-assisted monomolecular recombination coefficient, b is the bimolecular recombination coefficient of the radiative free charge carrier, c is the nonradiative three-body Auger recombination coefficient, and N is the excess carrier density above background under illumination of the film. [55,57] Figure 7b illustrates the carrier movement in bulk PeLEDs, where the free charge carriers can be easily trapped in material defects such as halide vacancies,A -site vacancies, anti-sites,a nd Pb precipitate clusters, [2c] because the bimolecular recombination rates are usually slow in bulk perovskites. [58] However,t he actual chemical equilibrium between free carriers and excitons in 3D PeLEDs is still under debate. Recent experiments showed that emission spectra were wellreproduced at the energy of the exciton absorption, where the contribution from band-to-band transitions is negligible,thus implying that luminescence is mainly due to the recombination of excitons.Despite,being outnumbered by free carriers in bulk perovskite films,e xcitons emit most of the light, as their faster radiative decay rate more than compensates for their lower density. [56,59] Hence,more studies are necessary to fully understand the generation of free carriers and excitons as well as the radiative recombination process in PeLEDs.
In contrast, radiative recombination in quantum-confined perovskites is dominated by excitonic monomolecular recombination due to the relatively high exciton binding energy (usually hundreds of millielectronvolts,c a. 190-500 meV). [60] Therefore,t he PLQYs for perovskite quantum dots or nanocrystal systems are strongly dependent on the competition between the first-order radiative excitonic recombination and the first-order nonradiative exciton trapping,asshown in Equation (6), where k ex, k trap ,a nd k Aug are the radiative monomolecular exciton recombination coefficient, nonradiative monomolecular exciton trapping coefficient, and nonradiative Auger recombination coefficient, respectively.F igure 7c demonstrates the exciton recombination and trapping process in perovskite nanocrystal LEDs.T he presence of surface ligands significantly reduces the number of perovskite defects compared with bulk 3D perovskites;a saresult, perovskite nanocrystals can achieve PLQYs near unity.A s discussed in the previous section, surface charged defects arising from aging or washing can act as exciton trapping centers and hinder the device performance.Another excitonloss mechanism in quantum dot systems is through excitonligand interactions.Kawano et al. confirmed that the Wannier and Frenkel exciton energies exhibit near-resonance,w hich leads to selective emission quenching in the organic ligands (naphthylmethyl moieties). They explained the selective luminescence quenching by an enhancement in the oscillator strength of near-resonant transition energies between Wannier and Frenkel excitons.T his provides potential evidence that ligands can also act like an exciton trapping center to reduce the desirable radiative recombination. [61] Theexcitonligand interactions have also been studied in atraditional PbS quantum dot system. Papagiorgis et al. studied the exciton charge transfer dynamics between PbS quantum dots and small metal chalcogenide ligands (e.g.K 4 GeS 4 ). Thea uthors suggest that the excitons can be trapped by the inter-dot ligands,which quench the photoluminescence from the corelevel excitons. [62] Additionally,R ossi et al. reported that Fçrster resonance energy transfer occurs between CsPbBr 3 nanocrystals and ligands based on perylene diimide,w hich causes emission from the ligands. [63] Based on previous studies of traditional quantum dot systems,s uch as PbS and CdS, [64] we expected to have ligand-exciton interactions or trapping phenomena. However, the ligands need to be strongly coupled to the electronic structure of the quantum dots (QDs), which can provide additional energy states to trap the excitons. [64c] Additionally,the position of the singlet or triplet states of the organic ligands relative to the conduction band minimum of the QDs are expected to be in the resonance of the incident light for sufficient energy transfer from the QDs to ligand shells. [64a,b] However,d etailed studies on perovskite NC systems are still required to understand ligand-exciton coupling and improve exciton transfer between NCs for LED and solar cell applications.
In summary,b ulk perovskite thin films have advantages for making solar cells,ascarriers can be easily extracted (low exciton binding energy and high mobility). Currently,t he record PCE of aperovskite solar cell (25.5 %) [9] was achieved with bulk perovskite thin films,w hilst the highest PCE of ap erovskite nanocrystal solar cell is 16.6 %. [65] Thep rimary reason for this considerable gap in the PCE is the capping agents of the perovskite nanocrystals,t hat is,o leic acid and oleylamine.T he insulating nature of these ligands limits charge separation and transport, thereby leading to lower carrier extraction and lower PCEs in solar cells.I nc omparison, perovskite nanocrystals have advantages for making LEDs as they can achieve unity PLQY and higher exciton binding energy through the presence of surface ligands. However,a sar esult of the insulating nature of long-chain ligands,the efficiency of nanocrystal PeLEDs can also be low even they can achieve high PLQYs (Figure 7a). Therefore, ligand engineering and defect management is key for both bulk and nanocrystal PeLEDs to balance the carrier injection efficiency and PLQYs (related to surface defects).
Theg eneral principle for designing high-performance PeLEDs is to maximize the free charge carriers or exciton radiative recombination and minimize nonradiative recombination through the Auger process and traps. [66] There have been many detailed reviews about modifying the chargetransport layer to improve the charge balance and reduce the Auger process, [67] including improving hole injection by adding an extra polymer modification layer to achieve better band alignment [68] as well as doping hole transport material to increase hole mobility [69] and decrease charge leakage. [70] Another challenge for PeLEDs and perovskite solar cells is operational instability.T he origin of this instability mainly arises from ion migration in the perovskite layers. [71] Ion migration is intrinsically ad efect migration process. [72] Thedrift-diffusion model is usually used to model the net current in PeLEDs and solar cells, [73,74] in which the drift current is induced by an external electrical bias and the diffusion current is governed by ac arrier concentration gradient. Under an applied electrical bias,t he ions (halide ions or A-site cations) can migrate via defects,t hereby causing material degradation which is irreversibly detrimental to device stability. [72a] Methods such as making diffusiondriven charge-transport LEDs [75] and effective diffusion engineering of the charge injection layers [76] would be future directions to suppress ion migration in planar LEDs.I nt his Review,wewill only focus on improving PeLED performance by mitigating the effects of defects to improve device efficiency and stability (Figure 8).
Thef irst common method to suppress nonradiative recombination is through compositional engineering,i ncluding A-site, [77] B-site, [78] and X-site [52g] tuning to improve the PLQY.T he mechanism consists of the following:1 )Increasing the defect formation energy and reducing the intrinsic defects of the materials by extra interactions between the dopants and perovskite lattice. [77,78] Fore xample,K im et al. doped guanidinium cations (GA + )into the A-site of FAPbBr 3 NCs;t he amine group of the GA cation can interact with lattice halides through hydrogen bonding between the amine group and the Pb-halide framework which improves the material stability and reduces defect formation. [77] As aresult, the PLQE increases from 79.7 %( pristine FAPbBr 3 NCs) to 93.3 %( FA 0.9 GA 0.1 PbBr 3 NCs;F igure 9a). Similarly,Z hang et al. reported that by B-site doping of CsPb(Br x Cl 1Àx ) 3 with La 3+ ions,t he density of state in the conduction band is increased, which inhibits the formation of defects at mid-band gap states (Figure 9b,c). [78a] As ar esult, the radiative recombination rate increased from 0.036 ns À1 (0 %L a 3+ )t o 0.069 ns À1 (5.43 %L a 3+ )w ith an increase in the PLQY from 3.47 %t o8 4.3 %a nd an increase in the EQE from less than 0.5 %to3.25 %at489 nm. 2) Thesurplus of the doping agent can passivate surface defects and dangling bonds to reduce nonradiative recombination sites. [77,79] Kim et al. suggested that the excess of the GA + dopant can modulate the surface of the NCs and passivate the surface defects to achieve ah ighperformance green LED with ac urrent efficiencyo f 108 cd A À1 and an EQE = 23.4 %. Similarly,L ue tal. doped Ag + ions into CsPbI 3 NCs.T he addition of the Ag + ions is believed to be beneficial for surface passivation as well as doping (Figure 9c). Thed iffusion of Ag ions into the perovskite lattice stabilizes the CsPbI 3 NCs,w hile the excess Ag ions on the NC surface reacts with iodide to form AgI, which also partially stabilizes the NCs from moisture and irradiation. Thes imultaneous silver doping and passivation lead to an increase in the device performance from 7.3 %t o 11.2 %. [79] 3) Doping elements with as maller atomic radius could increase the exciton binding energy and lead to high radiative recombination rates,a st he exciton lifetime is inversely proportional to the binding energy. [78a, 80] Furthermore,f or CsPbI 3 systems,p artially replacing the Pb 2+ by smaller Mn 2+ , [81] Sr 2+ , [82] or Zn 2+ [83] ions leads to areduction in the strain imposed by the larger iodides.T his occurs by increasing the cohesive energy among the octahedral units and, as aresult, it restricts the transition from the facile photoactive alpha-phase to the photo-inactive delta-phase and thus provides higher thermal stability. [14b] Va rious types of passivation agents have been extensively discussed in Section 3, including organic ligands (such as Lewis acids/bases, [38e,52i] amine-based passivation materials such as branched polyethyleneimine (PEI) and ethylenediamine (EDA) [84] ), inorganic salts (such as lead halides [14c, 68a] and potassium halides, [85] potassium salts (KPF 6 )), [86] and organic salts. [52g, 86, 87] Thep assivation of intrinsic material defects can be achieved by loading the passivation agents into the precursors prior to film deposition or colloidal synthesis [85,88] or during post-treatment of the crystallized perovskite films [89] or colloidal solution. [52g,l] Furthermore,t reat- ments carried out during deposition of the film can benefit the device performance.R ecently,K arlsson et al. developed av apor-assisted crystallization method to reduce the density of surface defects. [89] Directly after spin-coating,t he asdeposited perovskite thin films were placed in ap etri dish under an N,N-dimethylformamide (DMF) atmosphere.T he presence of DMF vapor allows redistribution of the surface halides,w hich leads to am ore homogenized surface halide composition with alower defect density,enabling an increase in the film PLQY from 3% (control) to 12 %(VAC-treated), as well as an increase in the PL lifetime.B yd oing so,t hey achieved am ore spectrally stable blue LED with the EQE increasing from 0.6 %(without VACtreatment) to an average of 3.8 %( with VACt reatment). [89] Another interesting con-cept for defect passivation is to design the passivation agents so that they are suitable for passivating multiple types of defects.F ang et al. reported ad ual passivation of perovskite defects using ab ifunctional molecule,4 -fluorophenylmethylammonium trifluoroacetate (FPMATFA). TheT FA anions and FPMA cations can bond with undercoordinated lead and halide ions,respectively (Figure 9e), thereby resulting in dual passivation of both lead and halide defects.T his dual passivation method results in an efficient LED device with 20.9 %E QE at l = 694 nm. [52h] Theb imolecular recombination rate for bulk perovskite LEDs is low,w hich leads to al ow PLQY,a nd the electrons and holes can be easily disassociated and trapped in defects. Therefore,a nother approach is to improve the PLQY of the  (g). It is shown that TSPO1 could passivate defects on the surface of QD films, the defects may trap carriers (e.g. holes, electrons), decrease exciton recombination, and hence degrade the device performances. h) PLQY of QD films without and with TSPO1 on the bottom, on the top, and on both sides of aQ Dfilm. Error bars represent the standard deviationo f experimental data for three measurements. Reproduced from Ref. [92] with permission under the Creative Common CC-BY license.Copyright 2020, the authors. i) PL lifetimes for films with mixed large/small grains, small-grainso nly,a nd large-grainsonly.j)EQE as afunction of time under aconstant bias for perovskiteLEDs with mixed large/small grains (black curve) and an ormal perovskitef ilm prepared by the vapor-assisted twostep method (red curve). Reproduced from Ref. [97] with permission under the Creative Common CC-BY license. Copyright 2019, the authors.

Angewandte Chemie
Reviews intrinsic material by increasing confinement effects through heterostructures,s uch as by constructing am ixed 2D/3D (quasi-2D) structure [52c, 90] or through perovskite-polymer heterostructures, [52k,91] which could prolong the carrier lifetime and improve the structural stability.T he 2D/3D passivation strategy is discussed in detail Section 5. Furthermore, the presence of polymer,s uch as poly(2-ethyl-2-oxazoline (PEOXA), can regulate the crystallization process to further reduce boundary defects. [91b] Cai et al. showed that the polymer matrix provides excess nucleation sites during the NC recrystallization process,w hich leads to more uniform distributions of NCs in the films,thereby resulting in ahigher PLQY and EQE (from 1.4 %w ith 0wt% polymer to 6.55 % with 45 wt %polymer) in thin films of the composite. [91b] Interfacial engineering is also critical to achieve highperformance PeLEDs.Itissuggested that defects can be most easily generated on the surface of as-deposited perovskites as there is asignificant number of dangling bonds at the exposed interfaces which would quench photoluminescence. [67b] Furthermore,t he wettability of the surface onto which the perovskite is grown is critical for the crystallization of the solution-deposited perovskites,a si td irectly influences the concentration of GB defects in the films. [52f,67b] As illustrated in Figures 7b and 9f,d efects between transport layers and emissive layers can trap carriers which reduce radiative recombination. Several reviews have been published that focus on interfacial layer engineering for solar cells and LEDs,s uch as improving the charge injection and charge balance,s uppressing interfacial defects,a nd exciton quenching. [2b,67b] In contrast to traditional one-sided interfacial engineering,Xuetal. reported abilateral interfacial passivation strategy by evaporating organic ligand molecules (diphenylphosphine oxide-4-(triphenylsilyl)phenyl (TSPO1)) on both the top and bottom of the perovskite emitting layer (Figure 9g). By doing so,t he film PLQY was enhanced (Figure 9h). Thei nteractions between uncoordinated Pb bonds and P=Obonds from TSPO1 can significantly suppress interfacial nonradiative recombination and improve the EQE from 7.7 %(nonpassivated) to 13.5 %(passivated only on the top) and 18.7 %( passivated on both sides). [92] Therefore,i t would be an ovel approach to incorporate interfacial passivation on both sides of perovskites in the design of future devices.Apart from adding afilm, mechanically removing the defective surface could be another interesting approach to try to modify interfacial defects in PeLEDs.Chen et al. removed the defective MAPbI 3 perovskite surface layers by using adhesive tape to mechanically peel off the top surface.B y doing so,t he PL intensity of the film increased, which indicates suppression of surface nonradiative recombination. [93] This could be au seful approach for PeLED applications.
Surface morphology and grain engineering are also used for improving PLQYs to achieve more efficient PeLEDs.The film morphology of perovskites is strongly dependent on the deposition conditions,i ncluding solution concentration, solvents,spin conditions,annealing temperature,the wettability of the substrates,a nd how supersaturation is achieved (e.g. antisolvent dropping during spin-coating or gas-phase quenching). [94] Yu et al. conducted ad etailed review on the effect of perovskite film morphology on LED device performance. [94] Both 3D and NC perovskites are strongly dependent on the surface ligands used. [95] Beyond surface passivation, appropriate ligand engineering is needed to obtain uniform and pinhole-free films. [96] Qin et al. demonstrated that the grain size of the perovskite films plays arole in carrier lifetimes and device efficiency. [97] Attaching high-band-gap small grains to the surfaces of low-band-gap large grains can form interfacial potential wells that can spatially confine GB defects and mobile ions to enable self-passivation during device operation. In addition, the small nanometer-sized grains can effectively confine injected electrons and holes and then efficiently cascade the injected carriers from high-bandgap small grains to low-band-gap large grains that function as light-emitting centers.Furthermore,attaching small grains to the surfaces of large grains leads to high-quality films for device fabrication. With this design, self-passivation can be realized when mobile ions are electrically driven to GBs to neutralize the charged defects,a nd as ar esult the carrier lifetime and device EQE is increased (Figure 9i,j). [97]

Improvement of the Performance of Solar Cells by Defect Passivation
Defect passivation strategies are also extensively studied to improve the power conversion efficiency of perovskite solar cells (PSCs). Many reviews have summarized the influence of the defects on charge-carrier recombination and transport, ion migration, and the structural stability of solar cell devices,aswell as summarizing methods to passivate defects and improve device performance. [2b,98, 99] In this section, we will focus on passivation methods for 2D/3D perovskites.

2D/3D Perovskite Passivation
Ty pically,a saresult of the polycrystalline structure and ionic nature of organic-inorganic perovskites,s tructural defects in the bulk, at GBs,a nd at the surface need to be considered and passivated to suppress the nonradiative recombination in perovskite films. [100] Fore xample,t he evaporation or migration of some volatile components such as methylammonium (MA) or to alesser extend iodine (I) can lead to the formation of cation or halide vacancies in threedimensional (3D) perovskites. [101] So far, numerous organic [102] and inorganic [103] materials have been used for passivating the different types of defects.L arge organic molecules (LOMs) with appropriate functional groups can not only passivate the defect states but also stabilize the 3D perovskite structure by forming a2Dperovskite at the GBs or the surface of the 3D films,a nd this is termed a2 D/3D structure. [104] Therefore,i n addition to suppressing the interfacial charge-carrier recombination, the moisture-entry pathways could be suppressed because of the hydrophobic nature of the LOMs.T his results in improving the efficiency and the stability of perovskite solar cells (PSCs) simultaneously ( Figure 10). However, various 2D/3D structures can be formed by applying different treatment approaches to the perovskite films.F or example, adding the LOMs to the perovskite precursor solution yields a2 D/3D mixed structure,w hile post-treatment of the perovskite films with the LOMs tends to form a2 D/3D bilayer structure ( Figure 10). Thep assivation of perovskite films with LOMs by creating different 2D/3D structures will be discussed in the following section.

2D/3D Mixed Perovskites
LOMs have been used for creating ap ure 2D Ruddlesden-Popper (RP) perovskite L 2 A nÀ1 B n X 3n+1 structure,w here Lisabulky OM spacer, Aisamonovalent cation such as Cs + , MA + ,orformamidinium (FA + ), Bisacentral metal ion (Pb 2+ or Sn 2+ ), Xisahalide (I À ,Cl À ,orBr À ), and n (1, 2, 3, 4, …. 1) is the number of inorganic perovskite layers within each quantum well. [105] Although 2D PSCs have undergone tremendous development in terms of their stability in the last few years,they still suffer from poor charge transport because of the insulating nature of the inserted LOM spacers in their structures,w hich result in lower efficiencies achieved compared to 3D PSCs. [11] To answer this challenge,combining 2D and 3D perovskites has been introduced as an efficient approach to accomplish both high stability (from the 2D structure) and high efficiency (from the 3D structure) simultaneously. [106] The2 D/3D mixed structure can be formed by adding the LOMs directly to the perovskite precursor solution. Nevertheless,v arious LOMs with different functional groups can be used to passivate different types of defects in perovskite films.T he cationic LOMs with ammonium (NH 3 + )functional groups have been used widely for defect passivation and to form a2 D/3D mixed structure. [107] Theb utylammonium (BA + ) [106a, 107,108] and phenylethylammonium (PEA + ) [109] ions are the two most widely used LOMs in this regard. TheS naith group [106a] demonstrated that adding BA + to aperovskite precursor delivers 2D platelets between highly orientated 3D perovskite grains after crystallization, thereby leading to defect passivation and an improvement in moisture stability.T hey elucidated that the 2D platelets at the GBs of the 2D/3D perovskite provide aclean electronic interface to reflect the charge carriers into 3D grains (Figure 11 a), thereby improving the carrier life-times through suppressed charge trapping and recombination. Additionally,t he 2D/3D perovskites maintained 80 %o ft heir initial PCE after 1000 hi na ir,a nd close to 4000 hw hen encapsulated.
Ho-Baillie and co-workers [109a] showed that adding phenethylammonium ions (C 6 H 5 CH 2 CH 2 NH 3 + ,P EA + ) to the perovskite precursor results in the formation of aq uasi-2D structure at the 3D perovskite GBs,t hereby leading to in an improved carrier lifetime and increased V OC from 0.99 to 1.10 Vf or the (FAPbI 3 ) 0.85 (MAPbBr 3 ) 0.15 perovskite.Indeed, the Huang group [109b] proposed amechanism for the passivation of 3D perovskites with PEA + ions.A s shown in Figure 11 b, PEA + prefers to either occupy the cation vacancies at the perovskite grains by ionic bonding or interact with undercoordinated I À ions through hydrogen bonding.M oreover,t he bulky hydrophobic aromatic ring of the attached PEA + ion at the perovskite surface or GBs can improve the moisture tolerance (Figure 11 b).
Thel ength of the alkyl chain in LOMs is critical to forming a2D/3D mixed structure.The Seok group compared the crystallization of the 2D/3D structure with BA + and PEA + ions,a st he most common 2D agent additives,a nd with octylammonium (OA + )ions having alarger organic chain. [110] In contrast to the case with BA + and PEA + ions,n oX RD peak corresponding to the 2D phase was observed for the OA-incorporated perovskite film (Figure 11 c). Thea uthors proposed that the OA + cations form strong van der Waals forces with the long alkyl chains,and ahydrogen bond forms between the ammonium and under-coordinated iodide ions, that is,cation vacancies are passivated without creating a2D/ 3D structure.T hus,t he OA + cations only encapsulated individual 3D perovskite crystal domains with ah ighly preferential orientation, thereby suppressing the nonradiative charge carrier recombination. As ar esult, the V OC and FF values of the OA-passivated PSC were increased from 1.06 V and 79.0 %t o1.11 Vand 81.5 %, respectively,t hus achieving ah igher PCE of 20.6 %t han the control PSC (18.4 %). Similarly,very recently,the Bakr group [114] observed that the 2D phase was not formed in the perovskite film with additives having very long-chain cations,s uch as OA + and oleylamine (OAm + ).
On the other hand, passivating the anion vacancies in 3D perovskites,a nother defect source that effects nonradiative recombination, is very important. One interesting strategy is the use of anionic additives with cationic LOMs.I nt his regard, Kim et al. [115] demonstrated that the coupling of PEA + cations and SCN À anions can passivate negatively and positively charged defects,that is,cation and halide vacancies, respectively.Ina ddition, the GBs were passivated efficiently by the quasi-2D perovskite (PEA) 2 Pb(I 4À2x SCN 2x ), with its larger band gaps compared to 3D perovskites.Asaresult, the

Angewandte Chemie
Reviews average PSC performance increases from 16.3 to 18.7 %f or the passivated wide-band-gap (1.68 eV) PSCs with < 4% degradation after storage for over 4000 hu nder N 2 .N evertheless,t he passivated perovskite/CIGS 4-terminal tandem solar cell showed ah igh PCE of 25.9 %. Another strategy would be to use zwitterion LOMs,w hich have both cationic and anionic functional groups in their molecular structures.5-Aminovaleric acid cations (HOOC(CH 2 ) 4 NH 3 + ,5 -AVA + hereafter) is one of the most frequently used zwitterions for the passivation of perovskite films through the formation of 2D/3D mixed perovskite structures.Zhang et al. [111] proposed that in the 5-AVA (MAPbBr 3 ) 2 perovskite structure,the NH 3 + end groups of the 5-AVA + occupy the MA + positions and the COO À group occupies the Br À sites,t hus passivating the cation and anion vacancies and forming cross-linked 2D/3D perovskites (Figure 11 d). Moreover,t he Nazeeruddin group [106b] demonstrated that adding 5-AVA + cations to the perovskite precursor can form an ultrastable 2D/3D perov- c) The XRD pattern of perovskite films without and with PEA + ,BA + ,and OA + additives. Reproducedf rom Ref. [110] with permission. Copyright 2018, the Royal Society of Chemistry.d )Schematic chemical structure of a5-AVA cross-linked perovskite structure. Reproducedfrom Ref. [111] with permission.C opyright 2017, John Wiley and Sons. e) Schematic illustration of the formation of acrosslinked 2D/3D bilayer perovskite. Reproduced from Ref. [112] with permission.C opyright 2019, American Chemical Society.f)The mechanism of forming the 2D/3D bilayer structure by treatment with BA + vapor and time-resolved PL spectra of 3D and 2D/3D bilayer perovskites. Reproduced from Ref. [113] with permission.C opyright 2019, American Chemical Society. skite film, thereby resulting in PSCs that are stable for one year under one sun AM 1.5 Gconditions at atemperature of 55 8 8C.

2D/3D Bilayer Perovskites
Apart from the 2D/3D mixed perovskites,d epositing as olution of LOMs on the top of 3D perovskite films can passivate the defects on the surface and form a2Doverlayer, that is,a2D/3D bilayer.N evertheless,i nc ontrast to 2D/3D mixed structures,this structure could create aflat or ordered band alignment distribution between the 3D and 2D perovskite phases,w hich is more favorable for charge transfer and for suppressing unfavorable charge recombination. Similar to the 2D/3D mixed structures,m any ammoniumbased cationic LOMs have so far been used for the passivation of cation vacancies on the surface of perovskite films.D ocampo and co-workers [116] developed af acile solution-based infiltration of BA + and PEA + cations to create alayered 2D/3D perovskite film. They demonstrated that, in this bilayer structure,t he 2D layer acts as as elective hole extraction layer between the 3D perovskite and spiro-OMeTAD,t hereby leading to reduced recombination losses and an enhanced V OC value from 0.99 to 1.08 Vf or both the BA + -and PEA + -passivated MAPbI 3 PSCs and with enhanced moisture stability.M any research groups tried subsequently to improve the photovoltaic parameters of PSCs by the surface passivation of perovskite films with these two LOMs. [108a, 117] Modification of LOM structures has been introduced as an efficient approach to improve their passivation effects.For example,the Sargent group [112] designed 4-vinylbenzylammonium (VBA) molecules,w ith an additional vinyl group compared to the PEA + cations,which are capable of forming ac ross-linked 2D structure by ap hotochemical reaction under ultraviolet light (Figure 11 e). Thec ross-linked 2D/3D perovskite structure showed lower nonradiative recombination compared to the pure 3D structure,w ith as ignificant improvement in the V OC value up to 1.20 Vand aPCE of up to 20.4 %w ith negligible J-V hysteresis when using (MAPbBr 3 ) 0.15 (FAPbI 3 ) 0.85 (containing 5% Cs). Moreover, the cross-linked 2D/3D PSC retained aP CE of 90 %o fi ts initial efficiency after long-term stability measurements for 2300 hina ir.
Zwitterions with their ability to react with both cation and anion vacancies can be used for the surface passivation of perovskite films and formation of a2 D/3D bilayer structure. Forexample,the Park group [118] deposited 5-AVA + cations on the perovskite surface to form an ultrathin 2D (5-AVA ) 2 PbI 4 passivation layer at the interface of the 3D perovskite and hole transporter layer. TheP CE of the passivated PSC was improved substantially from 13.72 to 16.75 %, mainly due to enhanced V OC and FF values,w hich was attributed to enhanced carrier lifetimes by effective passivation of the defect and trap states on the surface of the 3D perovskite.
In solution-processed 2D/3D bilayer passivation, however,p rotic polar solvents (e.g.i sopropyl alcohol, IPA) are frequently used to prepare the solution of LOMs.R ecently, Bawendi and co-workers [13] demonstrated that the solvent of the LOM solution, that is,I PA ,c an dissolve some of the perovskite components,s uch as FA + ions,t hereby degrading the underlying perovskite film. On the other hand, it must be considered that controlling the thicknesses and uniformity of the 2D overlayer formed through solution-processed approaches is very challenging. To answer this,t he Chen group [113] recently exposed the 3D perovskite film to BA + vapor (BAV )toform auniform 2D overlayer with atuneable thickness (Figure 11 f). Thea verage PCEs of BAV-treated PSCs were increased by 10 %, which was ascribed mainly to the higher V OC and FF values compared to control devices. Characterization by photoluminescence spectroscopy showed that the lifetime of the charge carriers was increased from 129.0 ns for the 3D film to 165.5 ns for the 2D/3D film, thus confirming the defect-passivation effect and suppression of nonradiative charge recombination.
Them ost frequently used LOMs for passivating perovskite films by creating 2D/3D mixed and bilayer perovskites are summarized in Table 1. It should be noted that besides perovskite solar cells,the 2D/3D structures have shown great potential to improve the efficiencyand stability of perovskite LEDs by tuning the radiative recombination coefficient without significant charge-carrier losses. [119,120] Fore xample, Zhang et al. introduced af acile self-organized and controllable 2D/3D mixed perovskite to improve the performance of blue perovskite LEDs. [121] They demonstrated that the 2D/3D structure increases the electron and hole densities by suppressing nonradiative recombination, thereby resulting in an enhanced external quantum efficiency of 8.2 %, from 1.5 %f or the initial device,a sw ell as a2 .6-fold improved operational stability of the device under ambient conditions. Moreover,very recently,Han et al. strategically designed 2D/ 3D hetero-phased structures on the surface of 3D bulk perovskite films that reduced the trap density and improved the long-term stability of perovskite LEDs. [122] Thefabricated LEDs with 2D/3D hetero-phased perovskite structures showed ah igher electroluminescence efficiencyo f7 .70 ph per el% and an operational lifetime of 200 hc ompared to apure 3D device with 0.46 ph per el% and 0.2 h, respectively.

Summary and Outlook
This Review provides an in-depth discussion to understand the types of defects in lead-halide perovskite thin films and NCs and their influence on defect tolerance and device performance.I mportantly,t he halide ion dependent defect tolerance/intolerance nature of LHPs is discussed in detail. In addition, the current understanding of the surface chemistry of LHP NCs,t he origin of surface defects,a nd passivation ligands that can be used to improve their photoluminescence efficiency and stability are highlighted. Furthermore,e merging defect passivation strategies are summarized for improving the performance of light-emitting diodes and solar cells. Despite significant progress in understanding the surface chemistry and the development of passivation strategies for improving the optical properties of LHP NCs and thin films, there are still an umber of questions that remain to be addressed to further improve the stability and efficiency of LHP optoelectronic devices.
Although the role of alkylamines in the stabilization of perovskite NCs is somewhat clear, the binding sites and the role of the alkyl acid are under debate.Afew studies suggest that they bind to the halide vacancies to complete the octahedral coordination of Pb,w hile af ew studies proposed that they form an equilibrium with the ammonium cation bound to the surface of NCs but do not directly interact with the surface of the NCs.T herefore,i n-depth studies focusing on the role of surface ligands are needed for ab etter understanding of their role in the surface passivation and stabilization of LHP NCs.
Cationic and anionic ligands have been used most often for the passivation;h owever, it is still unclear as to whether neutral ligands can bind and passivate LHP NC surfaces. Surface passivation is important for high PLQYs of NCs, while on the other hand, the surface ligands hinder the charge-carrier transport in NC films.T herefore,i ti si mportant to find abalance between surface passivation and charge transport to improve the efficiencyofthe resulting optoelectronic devices. [140] One alternative could be the use of small molecules or conductive organic ligand shells that could electronically couple the adjacent NCs of LHP NC films.I n addition, it is important to develop ligand-exchange strategies in both colloidal solutions and NC thin films to replace the long-chain ligands by small molecules.
It is now broadly understood that the halide vacancies on the surfaces are the main nonradiative recombination centers in LHPs.T hese halide vacancies are often filled with either the corresponding halide ions after the addition of metal halides or with softer Lewis bases.T he addition of metal halides could lead to the refilling of halide vacancies as well as the doping of the metal into the LHP lattice.Afew studies have shown that doping metal ions into aL HP lattice can significantly improve the PLQY of CsPbCl 3 NCs by eliminating nonradiative defects.However,itisstill under debate as to whether the doping,the halide passivation, or both results in the improved PLQY of CsPbCl 3 NCs.Moreover,the position of dopants in the LHP lattice is still unexplored. Therefore, more studies are required to decouple the roles of metal doping and halide passivation in the improvement of the optical performance and stability of LHP NCs.
To date,m ost studies have focused on how to eliminate identified defects in thin films and nanocrystals.T here is still alack of understanding about how defects are induced during the different synthesis methods,including precursor preparation, thin-film crystallization, and nanocrystal purification processes.M ore studies are required to understand the collective mechanism of surface defect and grain boundary defect passivation. Lastly,b oth perovskite LEDs and solar cells suffer from operation instability because of defect evolution over time because of degradation (such as phasetransformation or moisture-induced degradation). Therefore, despite the need for preparing highly efficient LEDs,such as spectra-stable blue and pure-red LEDs,a sw ell as high PCE solar cell devices,overcoming the limitations arising from the soft ionic nature of the material and improving the opera-  Angewandte Chemie Reviews tional stability would be akey focus in the future.Itshould be noted that no individual strategy can passivate all defects and prohibit device instability,i ti sc rucial to understand and utilize multiple strategies including,b ut not limited to, compositional engineering,d evice engineering,a nd defect passivation treatment to further improve perovskite-based devices.
In PSCs,itisnow well-established that the structural and point defects in the polycrystalline perovskite films must be passivated to suppress nonradiative recombination and result in the simultaneous improvement of the efficiency and stability.S imilar to perovskite LEDs,i th as been shown that large organic molecules (LOMs) with appropriate functional groups can passivate the defect states by forming a2 D perovskite at the GBs or the surface of the 3D perovskite films,t hat is,a2 D/3D structure,t hereby stabilizing the perovskite structure and preventing more defect formation. Different methods for obtaining the 2D/3D structures are summarized in this Review.T he 2D/3D mixed and 2D/3D bilayer structures have been investigated as important passivation approaches.W hereas the 2D/3D mixed structure can be formed by adding the LOMs directly to the perovskite precursor solution, the 2D/3D bilayer structures can be formed by depositing as olution of LOMs on the top of 3D perovskite films.The state-of-the-art of various 2D/3D mixed and bilayer structures have been investigated. However,w e emphasize that both types of 2D/3D structures are equally important to achieve ad efect-free perovskite film. Besides the methodology,w ee mphasize the need to develop more efficient LOMs such as zwitterions to passivate the defect states more efficiently.A lthough there have been several studies of the passivation mechanism in thin films,the binding of passivating agents to the LHP thin film surface is still not well-understood at the molecular level. More experimental and theoretical studies need to focus on the characterization of the thin-film surfaces before and after passivation to gain insight into and to rationalize the passivation mechanism. The field of PSCs is now more mature than that of perovskite LEDs.I nf act, al arge variety of passivation strategies and ahuge library of molecules have been reported for PSCs and it would be interesting to adapt such strategies to improve the EQEs of perovskite LEDs.W ethink there is alot of synergy between perovskite LEDs and solar cells and al ot can be learnt from each other for improving their efficiencies.