B‐Site Co‐Alloying with Germanium Improves the Efficiency and Stability of All‐Inorganic Tin‐Based Perovskite Nanocrystal Solar Cells

Abstract Colloidal lead‐free perovskite nanocrystals have recently received extensive attention because of their facile synthesis, the outstanding size‐tunable optoelectronic properties, and less or no toxicity in their commercial applications. Tin (Sn) has so far led to the most efficient lead‐free solar cells, yet showing highly unstable characteristics in ambient conditions. Here, we propose the synthesis of all‐inorganic mixture Sn‐Ge perovskite nanocrystals, demonstrating the role of Ge2+ in stabilizing Sn2+ cation while enhancing the optical and photophysical properties. The partial replacement of Sn atoms by Ge atoms in the nanostructures effectively fills the high density of Sn vacancies, reducing the surface traps and leading to a longer excitonic lifetime and increased photoluminescence quantum yield. The resultant Sn‐Ge nanocrystals‐based devices show the highest efficiency of 4.9 %, enhanced by nearly 60 % compared to that of pure Sn nanocrystals‐based devices.


Introduction
Colloidal halide perovskite nanocrystals (PNCs), with their size-tunable narrow-band emission, higher photoluminescence quantum yields (PLQY) than the corresponding bulk crystals,a nd enhanced surface-to-volume ratio, [1] are promising candidates for the next-generation optoelectronics including PVs, [2,3] light-emitting diodes, [4,5] photodetectors, [6] and lasers. [7] Because of the intrinsic instability in ambient conditions of organic cations (for example,C H 3 NH 3 + )a tt he As ite of at ypical ABX 3 perovskite structure, [8] all-inorganic lead (Pb) based PNCs,such as CsPbX 3 (X = Cl, Br, or I), are preferred to pursue high stability while retaining the advanced optical properties,f or example,t he high PLQY. [9] However,t oxic Pb is still the key constituent of the most efficient halide perovskite materials,t hus impeding PNCs widespread commercialization.
Tin( Sn) is one of the most promising alternative candidates to replace Pb,b eing its closest analogue with similar structural and electronic properties. [10,11] As ar esult, the most efficient Pb-free perovskite solar cells (PSCs) to date always include Sn in the perovskite composition. [12][13][14] However,i ntegrating Sn in spatially confined nanocrystals (NCs) has been rarely successful because of its high instability.Sn 2+ (the Bs ite in the perovskite structure) is rapidly oxidized to Sn 4+ in ambient conditions. [15,16] Furthermore,the high density of surface vacancies [17] result in the low photoluminescence quantum yield (PLQY) of Sn-based PNCs.A sa nother promising Pb-free candidate,g ermanium (Ge) is ad efecttolerant and nontoxic semiconductor still highlighted by the limited studies. [18,19] Among the synthesized materials,CsGeI 3 has adirect band gap (E g )of1.6 eV,which is suitable for PVs. When compared to CsSnI 3 ,C sGeI 3 does display,h owever, inferior optoelectronic properties owing to the lone-pair effect [20] and relatively larger E g . Inspired by the heterovalent B-site co-alloying strategy for Pb and Sn, several reports have proposed to mix Sn with Ge for an ew candidate of mixture lead-free perovskite. [21][22][23][24][25][26] By first-principles computations,J u et al. predicted several potentially promising mixed Sn-Ge halide perovskites as light absorbers for solar cells. [27] Ma et al. presented ac omprehensive theoretical study on the mixed vacancy-ordered inorganic double perovskite Cs 2 Sn (1Àx) Ge x I 6 ,i ndicating that the concentration of Ge doping severely influences the thermodynamic,e lectronic, and mechanical properties. [25] Ito et al. experimentally showed an improved power conversion efficiency( PCE) of 4.48 %for FA 0.75 MA 0.25 Sn 1Àx Ge x I 3 -based solar cells with a5% Ge doping into the perovskite,compared with that (3.31 %) of pure Sn-based ones. [28] Later on, Chen et al. reported the first synthesis of CsSn 0.5 Ge 0.5 I 3 bulk crystals via av apor method, showing ar elatively high PCE of 7.11 %b ut employing ac omplex, non-scalable,a nd expensive synthetic route. [29] However,t here are no examples of mixed Sn-Ge perovskite nanocrystals.
Here,w ep ioneered the synthesis of all-inorganic Ge alloyed CsSnI 3 perovskite nanocrystals (Sn-PNCs) via afacile and low-cost solution process.T he novel CsSn 0.6 Ge 0.4 I 3 perovskite nanocrystals (SnGe-PNCs) exhibit significantly improved optical and photophysical properties compared to those of reference Sn-PNCs.T he stability of as-formed CsSn 0.6 Ge 0.4 I 3 has also been effectively enhanced upon the introduction of Ge.T he partial replacement of Sn atoms by Ge atoms in the nanostructure can effectively fill the high density of Sn vacancies and reduce the surface traps,resulting in al onger excitonic lifetime and improved PLQY.I nt his work, the PCE (4.9 %) of the PSCs employing CsSn 0.6 Ge 0.4 I 3 nanocrystals is enhanced by nearly 60 %c ompared to that (3.1 %) of CsSnI 3 -based devices.

Results and Discussion
Synthesis of CsSn x Ge 1Àx I 3 NCs Based on the previous hot-injection methods to produce Sn-based PNCs, [30,31] we developed amodified reaction route for the synthesis of CsSn x Ge 1Àx I 3 NCs.Anequal molar ratio of SnI 2 and GeI 2 has been utilized to target the half replacement of Sn atoms by Ge atoms in the expected stoichiometry (namely, x = 0.5). We found that the injection temperature is crucial to form stable SnGe-PNCs at least before the purification process.T he temperature should be above 240 8 8C. After the injection of Cs-oleate into the precursor of Sn/Ge halide source,asudden color change from transparent orange to dark brown occurred (see the appearance of CsSn x Ge 1Àx I 3 NCs suspension in the side photo in Figure 1a) and no further color change was observed after 3s reaction time,i ndicating that the formation of SnGe-PNCs features the extremely swift reaction kinetics,similarly as in the case of conventional Pb-based PNCs. [32] Thedetailed synthetic route for SnGe-PNCs and the reference Sn-PNCs has been described in Supporting Information.
To examine the crystal structure of as-synthesized PNCs, X-ray diffraction (XRD) patterns for PNCs in film state have been measured. Figure 1a shows the comparison of XRD patterns for CsSn x Ge 1Àx I 3 NCs and reference CsSnI 3 NCs with the complete assignment of all the featured peaks corresponding to the halide perovskite structure. [31,33] TheX RD scan of reference Sn-PNCs matches well with JCPDS reference 00-043-1162 from the ICDD database (Supporting Information, Figure S1a), indicating the formation of an orthorhombic crystallographic phase that belongs to the space group Pnma (62), which is consistent with previous literature findings for similar Sn-based PNCs. [31] It is noted that the XRD pattern of the newly synthesized SnGe-PNCs shows slight shifts in the peak positions compared to that of the reference,t hat is,1 4.28 8 to 14.58 8 at (110) and 29.08 8 to 29.28 8 at (220) (see inset in Figure 1a), suggesting aslight change in the size of the unit cell dimension. This change is attributed to apartial exchange of the large Sn atoms with relatively small Ge atoms,r esulting in the shrinkage of the unit cell and an upward shift in the 2q angles.T herefore,w es imilarly assign the structure of as-formed SnGe-PNCs with orthorhombic crystallographic phase (Figure 1b)a si nt he case of the reference.The surface compositions of CsSn x Ge 1Àx I 3 NCs and reference CsSnI 3 NCs in film state were analyzed by X-ray photoelectron spectroscopy (XPS) measurements (Figure 1c). Thec oncentration of iodine should be considered as al ower estimate because of the X-ray and electron beam induced desorption of Id uring XPS and EDS measurements. [34] CsSn x Ge 1Àx I 3 NCs had similar surface concentrations for Sn and Ge that support the successful substitution of Sn with Ge (Supporting Information, Table S1). Only one chemical state was resolved on CsSn x Ge 1Àx I 3 NCs for Cs (Cs 4d 5/2 at 75.2 eV), Sn (Sn 4d 5/2 at 25.2 eV), Ge (Ge 3d 5/2 at 31.6 eV), and I(I4d 5/2 at 48.8 eV), which can be attributed to Cs + ,S n 2+ ,G e 2+ ,a nd I À ,r espectively,i na ccordance with the Cs + (Sn/Ge) 2+ I 3 1À perovskite structure. [29,35] Thethorough XPS analysis confirms the stoichiometry of the PNCs films at the surface (that is,within the XPS information depth of 11 nm), to be CsSnI 3 and CsSn 0.5 Ge 0.5 I 3 ,w ith minor deviations (Supporting Information, Table S1). To further verify the stoichiometry of PNCs in bulk, we conducted energy-dispersive X-ray spectroscopy (EDS) measurements for the PNCs films.T he EDS layered images SnGe-PNCs (Supporting Information, Figure S2) for the compositional elements show the even elemental distribution on the surface of PNCs film. TheEDS analysis elucidates the stoichiometry of SnGe-PNCs bulk film (Supporting Information, Table S1). Interestingly,t he elemental ratio between Sn and Ge turns to 2:1, meaning that the actual replacement proportion of Sn atoms by Ge atoms is nearly 40 %w hile considering the results of XPS data for the EDS analysis.W ethus specify x = 0.6 in the stoichiometry,o utlining CsSn 0.6 Ge 0.4 I 3 as the final composition. TheG ea mount dependent synthesis and properties of SnGe-PNCs will be investigated in aseparate work. It is noted that the ratio of Cs:I in the bulk case is less than the nominal value (for example,1 :3;S upporting Information, Table S1), which can be attributed to the small ratio (1:4) of the injected volume (Cs-oleate) and that of SnI 2 /GeI 2 precursor,c reating the halide-rich environment to form polyhedral [SnI 6 ] 4À as nucleus for the growth of nanocrystals. [36] Themorphology of as-synthesized CsSn 0.6 Ge 0.4 I 3 NCs was investigated using transmission electron microscopy (TEM). Figure 1d reveals that the nanoparticles have ac ubic shape with an average square diameter of about 9.4 nm for CsSn 0.6 Ge 0.4 I 3 NCs,w hich is slightly larger than that (ca. 8.3 nm) of CsSnI 3 NCs (Supporting Information, Figure S3a). Since the synthesis conditions were identical for both PNCs, this may suggest that small Ge atoms can potentially assist large Sn atoms to de-focus over al onger growth duration more effectively than pure Sn-based ones according to ak inetics model for the growth of Pb-based PNCs, [37] hence resulting in the bigger size of SnGe-PNCs.B ased on the HRTEM image of as ingle CsSn 0.6 Ge 0.4 I 3 nanocube (Figure 1e), the lattice distance is 0.317 nm that corresponds to (220) facets,h ighly consistent with the calculated lattice spacing of 3.17 from the Fast Fourier Tr ansform (FFT) analysis ( Figure 1f)ofthe HRTEM image.This,together with the previous XRD data, further supports an orthorhombic crystal structure. [31] Narrow size distributions have been observed with full widths at half-maximum (FWHM) of 1.9 nm (Figure 1g)a nd 1.4 nm (Supporting Information, Figure S3b) for SnGe-PNCs and Sn-PNCs,r espectively.T his suggests that our modified hot-injection synthetic route can precisely control the nucleation and growth of nanocubes.

I 3 NCs
Theo ptical properties of as-synthesized CsSn 0.6 Ge 0.4 I 3 NCs were investigated and compared to those of reference CsSnI 3 NCs at room temperature in ambient conditions.T he comparison of absorption and photoluminescence (PL) spectra of both PNCs in suspension is presented in Figure 2a. Aclear blue shift has been observed for both absorption onset and PL peak from CsSnI 3 NCs to CsSn 0.6 Ge 0.4 I 3 NCs due to the wider E g obtained upon the introduction of Ge in the Snbased nanocrystal structure.T he PL spectrum of CsSn 0.6 Ge 0.4 I 3 NCs shows am ain peak at around 780 nm (1.6 eV) with aFWHM of 65 nm, while asmall PL sub-peak at around 725 nm has been detected that was not observed for the case of reference CsSnI 3 NCs.W ed econvoluted the emission spectrum with two contributed bands showing respective Gaussian peaks in Figure 2b,c entered at 725 and 782 nm. [38] Interestingly,wehave found that asmall amount of SnGe-based nanorods formed together with the nanocubes. Thenanorods have an average diameter of about 4.5 nm and am ean length of about 50 nm (see the corresponding TEM image in the Supporting Information, Figure S4a). This finding explains the origin of the 725 nm emission. Theband gap of perovskite nanorods is closely related to their diameter or width. [39,40] Thesmall size of the Sn-Ge nanorods (diameter of 4.5 nm) may induce aq uantum confinement effect compared to the big SnGe-nanocubes (diameter of 9.4 nm), with ac orresponding shift towards ah igher band gap when the size of PNCs is less than their Bohr radius. [39] This phenomenon has not been observed for reference Sn-PNCs, since in that case ac hange in the morphology has not been detected, as for the case of the mixed Sn 2+ and Ge 2+ cations. This point will be more thoroughly investigated in af uture work. Thea bsorption and PL spectra in film state for both PNCs (Supporting Information, Figure S4b) are nearly identical to those in suspensions,suggesting that no morphological change occurs during the formation of the PNC bulk film (see the inset photo of CsSn 0.6 Ge 0.4 I 3 NCs film in the Supporting Information, Figure S4b). TheP Lq uantum yields (PLQYs) were sequentially measured as 0.95 %a nd 0.16 %f or CsSn 0.6 Ge 0.4 I 3 NCs and reference CsSnI 3 NCs,r espectively. Although the absolute values of PLQYs are low compared to the lead analogues,adramatic increase in the PLQY of CsSn 0.6 Ge 0.4 I 3 NCs has been achieved compared to that of the Ge-free counterpart. Since the dominant defects in Sn-based PNCs are interstitial, and Sn displays low defect formation energy in terms of the metal terminate mechanism, [41] the involvement of Ge effectively fills in Sn vacancies and reduces the number of traps,resulting in the enhanced PLQY.
Photophysical Properties of CsSn 0.6 Ge 0.4 I 3 NCs We then turn to assess the PL dynamics of as-synthesized nanocrystals by conducting the time-correlated single photon counting (TCSPC) measurements in the pico-and nanosecond regime.The PL decays of nanocubes in suspension are presented in Figure 3a.A pparently,t he emission decay of CsSnI 3 NCs is faster than that of CsSn 0.6 Ge 0.4 I 3 NCs.P oisson statistics of defects distribution in PNCs have been employed to model decays and fit the experimental data (see the detail of the Poisson model in Supporting Information). [42,43] It was found that two types of defects need to be assumed to obtain ar easonable data fit. Thef itting results are shown in the Supporting Information, Table S2, where t is the quenching time constant in ananocrystal having exactly one defect and c is the relative defect concentration, or the average number of defects per nanocrystal. Theobservation of short PL lifetime for CsSnI 3 NCs makes ag ood agreement with previously reported fast decays for Sn-based nanocrystals within several hundreds of picoseconds. [30,31] In contrast to the reference,the quenching time constants of CsSn 0.6 Ge 0.4 I 3 NCs have been extended by afactor of 4to490 and 4200 ps,respectively.At the same time the relative defect concentrations are also  lower for CsSn 0.6 Ge 0.4 I 3 NCs, c 1 = 2.0, c 2 = 1.7, vs. c 1 = 2.3, c 2 = 3.3 of CsSnI 3 NCs (Table S2). This supports our hypothesis of Ge filling the Sn vacancies,w hich has the two side effects of 1) lowering the density of defects and 2) the less detrimental effect of the defects left, that is,s lower quenching.F urthermore,itisnoted that the PL decay lifetimes of both PNCs in film (Supporting Information, Figure S5) state are similar to those of their suspensions,s uggesting that the defect distribution for both PNCs is less influenced by their variable phases.
To further investigate the influence of Ge replacement on the charge carrier dynamics and the nonradiative processes, we also carried out the ultrafast transient absorption (TA) spectroscopy on the encapsulated CsSnI 3 NCs and CsSn 0.6 Ge 0.4 I 3 NCs films on glass,respectively.The TA spectra of the CsSnI 3 NCs film excited upon a500 nm pump light after different time delays are shown in the Supporting Information, Figure S6a. Theo verall TA spectra are dominated by one distinct bleach signal with ap eak at 700 nm, which are corresponding to the first (660-800 nm) excitonic absorption band, in ag ood agreement with previously reported assignments for CsSnI 3 QDs. [44] It is noted that there is ap ositive signal appearing with ap eak at 600 nm, likely attributed to the overlap of the bleach with the photoinduced absorption (PIA) between 570 and 650 nm. Thecomparative TA spectra of CsSn 0.6 Ge 0.4 I 3 NCs film measured upon the identical condition as for the case of reference Sn-PNCs film are shown in the Supporting Information, Figure S6b.I nterestingly,asecond bleached band appears at 750 nm in addition to the main bleached band below 700 nm. We conclude that the second bleach arises from the small amount of nanorods mixed into the sample.A fter assigning the origins in the TA spectra of both PNCs films,w en ow turn to assess the TA kinetics.Acomparison of TA decays for CsSnI 3 NCs and CsSn 0.6 Ge 0.4 I 3 NCs films,m onitored at 770 nm in early timescale are shown in the Supporting Information, Figure S6c.It is clearly observed that both TA profiles instantly rise within the instrument time (ca. 0.2 ps), corresponding to the formation of hot carriers after excitation well above the band gap (500 nm pump light). Theh ot carriers then relax to the band edge within acooling time of about 2.5 ps for both films, indicating that the partial replacement of Ge atoms in nanostructure does not make obvious effect on the hot carrier kinetics.W et hen compared the TA decays of both films monitored at 700 nm (Figure 3b)toevaluate the dynamics of bleaching recovery.T he TA decay curves can be well fitted using ab i-exponential function: DO.D. = A 1 exp(Àt/t 1 ) + A 2 exp(Àt/t 2 ), which provides two exponential decay constants for both films (summarized in the Supporting Information, Table S3). As discussed earlier, Ge atoms can effectively fill the high density of Sn vacancies,reducing the shallow traps and leading to amuch longer average lifetime (t avg = 259.4 ps) compared to that of reference CsSnI 3 NCs film (t avg = 10.1 ps), which is highly consistent with the trend of TRPL data in the Supporting Information, Table S2. This further suggests that, compared to reference CsSnI 3 NCs,t he CsSn 0.6 Ge 0.4 I 3 NCs possess the extended excited-state lifetime that allows efficient charge transfer,w hich is beneficial for the device performance.

I 3 NCs-based Solar Cells
Thei mproved photophysical properties of CsSn 0.6 Ge 0.4 I 3 NCs compared to the case of reference CsSnI 3 NCs has adirect effect on the performance of corresponding PSCs.T o confirm this,w ei ncorporated both types of nanocubes in standard n-i-p photovoltaic architectures.O wing to the reported high charge diffusion coefficient (> 1.3 cm 2 s À1 )f or typical bulk CsSnI 3 crystals,aplanar structure has been adopted for this type of perovskite. [45,46] Thefull PSC structure is FTO/c-TiO 2 /PNCs/spiro-OMeTAD/MoO 3 /Au, as illustrated in Figure 4a.T he devices consisted of circa 550 nm thick PNCs layer (based on the cross-sectional SEM image of at ypical CsSn 0.6 Ge 0.4 I 3 NCs-based solar cell, Figure 4b), fabricated in an itrogen-filled glovebox. Thec urrent density (J)-voltage (V)c urves of the best-performing (champion) devices based on SnGe-PNCs and reference Sn-PNCs, recorded under standard 1S un condition (100 mW cm À2 AM 1.5 Gi llumination) and in dark condition, are depicted in Figure 4c,d, respectively.I nterestingly,t he J-V curves of both PNCs-based devices display anegligible hysteresis effect between forward and backward scans (Figure 4c). This suggests that ion migration under an electric field has anearly negligible influence on the photocurrent flow within the perovskite nanocrystals layer. This,i nt urn, may lead to am ore stable power output of the corresponding PNCs devices compared to that of bulk crystals-based devices,such as the popular tri-cation perovskite solar cells in n-i-p structure. [2] Thea veraged photovoltaic performances (16 devices for each structure) along with the standard deviation, together with the photovoltaic parameters of the champion cells,a re summarized in Table 1. Theh igh reproducibility of our results is demonstrated by the small standard deviations. TheCsSn 0.6 Ge 0.4 I 3 NCs lead to PSCs with aremarkable 58 % performance enhancement (average PCE = 4.1 %) compared to devices based on reference CsSnI 3 NCs (Table 1) in this work, resulting in the highest PCE of 4.9 %for the champion cell (shown in the inset photo in Figure 4d). [a] The data are from the forward scans.
[b] Values in the brackets refer to the photovoltaic parameters of the champion cells.

Research Articles
Theimprovement in the PCE is clearly determined by the increase in all device parameters,that is, J SC , V OC ,and FF.W e have extracted the series (R S )a nd shunt (R SH )r esistances of the devices,a ccording to the 1-diode model, [47] from the J-V curves scanned in dark with the axis of current density plotted in logarithmic scale (Figure 4c and Table 1). Both R SH and R S affect the FF,with lower R S and higher R SH being required to increase the FF.T he extracted resistances in Table 1confirm this trend since CsSn 0.6 Ge 0.4 I 3 NCs-based cells have lower R S and higher R SH (nearly twice) than that of the CsSnI 3 NCsbased cells.Inaddition, the excited-state lifetime (ca. 4ns) of CsSn 0.6 Ge 0.4 I 3 NCs clarified from the TCSPC and TA measurements,w hich is nearly three times longer than that (ca. 1.5 ns) of CsSnI 3 NCs,inturn leads to amore effective charge separation at the interface between PNCs film and chargetransporting layers (TiO 2 or spiro-OMeTAD). [48] This can be one key reason for the enhanced J SC and FF in CsSn 0.6 Ge 0.4 I 3 NCs-based cells.The above-demonstrated involvement of Ge in reducing the Sn vacancies (that function as trapping sites) suggests that CsSn 0.6 Ge 0.4 I 3 NCs-based active layer possesses much fewer defect sites compared to the CsSnI 3 NCs film, leading to the hindering of charge recombination within PNCs film beneficial to the V OC . [49,50]  Finally,t he effect of Ge on the stability of the corresponding PNCs is thoroughly studied. Figure 5a shows the comparison between the time-dependent PLQYs of CsSnI 3 NCs and CsSn 0.6 Ge 0.4 I 3 NCs in suspension in air (25 8 8Ca nd 50 %RH). While the PLQY of CsSnI 3 NCs swiftly decreased to below 30 %o fi ts initial value after 30 min (Supporting Information, Figure S8a), that of CsSn 0.6 Ge 0.4 I 3 NCs was retained to above 80 %o ft he original PLQY in the same time span (Supporting Information, Figure S8b). This suggests that upon the protection of Ge atoms in the Sn-based perovskite against the oxidization, the optical properties of PNCs can be effectively maintained. Figure S9 presents the comparison of XRD patterns from unencapsulated reference CsSnI 3 NCs (Supporting Information, Figure S9a) and CsSn 0.6 Ge 0.4 I 3 NCs (Supporting Information, Figure S9b) films exposed for 10, 20, and 30 min to ambient atmosphere. Ther elative intensity of the main XRD peak can sustain above 75 %ofits initial value for the case of SnGe-PNCs after 30 min exposure while that of Sn-PNCs has dropped below 45 %ofits corresponding initial intensity,consistent with the enhancement of optical stability upon the introduction of Ge. Moreover,w em easured the temperature-dependent PL spectrum (Figure 5b,c) for both PNCs under ambient conditions to clarify the thermal stability influenced by the involvement of Ge in the perovskite structure.I nterestingly, upon temperatures above 55 8 8C, the PL spectrum of CsSnI 3 NCs exhibited ac lear decrease in the PL intensity with ad istinct blue-shift and an additional sub-peak at about 710 nm, indicating as ign of the phase transition or degradation, that is,t he formation of Sn 4+ .O nt he contrary,t he PL spectrum of CsSn 0.6 Ge 0.4 I 3 NCs retained its shape feature when temperature increased, although the peak PL intensity at 782 nm (corresponding to nanocubes) decreased, as commonly observed for other types of perovskite nanocrystals. [51,52] It is interestingly noted that the sub-peak PL intensity at 725 nm (corresponding to nanorods) effectively increased up to 70 8 8Cand then decreased with the increasing temperature (85 8 8C), indicating that some temperature-dependent morphological transition from nanocubes to nanorods may occur which will be investigated in detail in as eparate work. Theo verall study of PL stability suggests that Ge atoms can effectively stabilize the nanocrystal structures upon the thermal treatment. We then studied the stability of the corresponding photovoltaic devices by measuring the change in the short circuit current density (J SC ) under continuous 1-sun illumination in air. Figure 5d compares the decrease in J SC upon unencapsulated cells illumination for both CsSnI 3 NCs-and CsSn 0.6 Ge 0.4 I 3 NCs-based devices.Aspredicted, J SC of pristine Sn-PNCs cells decreased by more than 75 %ofits initial value after 300 sofcontinuous illumination. On the other hand, SnGe-PNCs based devices can still retain more than 80 %o fi nitial J SC ,w hich is consistent with the variation trend of the time-dependent PLQYs of both PNCs.H ence,C sSn 0.6 Ge 0.4 I 3 NCs exhibit significantly improved ambient, thermal, and photostability compared to CsSnI 3 NCs.
To investigate the underlying reason for the improved stability of CsSn 0.6 Ge 0.4 I 3 NCs,wecan further refer to the XPS analysis in Figure 5e.C learly,o nly one chemical state containing Sn could be extracted for CsSn 0.6 Ge 0.4 I 3 NCs (Sn 3d 5/2 at 486.2 eV) after 10 mins exposure to air, that is, during the sample preparation before the XPS measurement. In contrast, within the same time frame,asecond chemical state was clearly observed for CsSnI 3 NCs related to Sn 3d transition:asecond pair of doublet peaks (Sn 3d 5/2 at 487.4 eV) is found in addition to the Sn 2+ iodide 3d 5/2 at 486.2 eV.The high binding energy component corresponds to Sn 4+ oxide. [53][54][55] This was also confirmed by the O1sm etal oxide component at 530.7 eV,w hich has been observed only for the case of pure Sn-PNCs.F urthermore,w eh ave conducted the XPS analysis on CsSn 0.6 Ge 0.4 I 3 NCs after 24 hours of exposure to air (Supporting Information , Figure S10). Upon the long-time exposure to air,Sn3d/4d peaks are shifted to higher binding energies,indicating the oxidation of Sn 2+ cations into Sn 4+ state.Y et, the Ge 3d peak has shown no obvious change with the fixation of most Ge 2+ cations in the perovskite structure.I nterestingly,t he fast formation of Ge 4+ oxide that was recently observed by Chen et al. on bulk perovskite with as imilar elemental composition (bulk CsSn 0.5 Ge 0.5 I 3 )b ut prepared using am elt-crystallization method has not been observed in our case. [29] Furthermore, the enhanced stability of CsSn 0.6 Ge 0.4 I 3 NCs is in agreement with two factors demonstrated in the previous simulation results:1 )the Goldschmidt tolerance factor (t)h as been reported for the mixed Sn-Ge halide perovskite in comparison to Ge-free counterpart, [56] that is, t (MASn 0.5 Ge 0.5 I 3 ) = 0.90, and t (MASnI 3 ) = 0.84, indicating that the involvement of Ge atoms in Sn-based perovskite structure is beneficial for the structural stability;2 )the activation barrier (0.23 eV) towards water or oxygen penetration of mixed Sn-Ge halide perovskite (RbSn 0.5 Ge 0.5 I 3 )h as been reported as remarkably higher than that of pure Sn perovskite (0.17 eV for MAS-nI 3 ) [56] or even of Pb-based perovskite (0.09 eV for MAP-bI 3 ), [27] suggesting that the terminated Ge atoms can effectively protect the inner Sn atoms,m aking them less susceptible to oxidization upon the replacement and filling effect from Ge atoms (see the proposal in Figure 5f).

Conclusion
We have shown the first-ever synthesis of CsSn 0.6 Ge 0.4 I 3 nanocrystals with anarrow size distribution. ThePLQY of Sn-Ge based NCs is close to 1%,which has enhanced nearly one order magnitude than that of conventional pure Sn-based PNCs.The photophysical properties of CsSn 0.6 Ge 0.4 I 3 NCs are dramatically enhanced compared to those of pure Sn-PNCs, mainly attributed to the prolonged excited-state lifetime,a s confirmed by both transient photoluminescence and transient absorption measurements.W ea ssign the improvement of both optical and photophysical properties of CsSn 0.6 Ge 0.4 I 3 NCs to the reduced number of Sn vacancies upon their filling with Ge in the nanostructure.V acancies act as trapping sites and alternative nonradiative recombination pathways.T he overall improved charge transfer dynamics for CsSn 0.6 Ge 0.4 I 3 NCs system, that is,e fficient charge separation and suppressed charge recombination, is the key reason for the significant enhancement of solar cells performance when CsSn 0.6 Ge 0.4 I 3 NCs are used as the light absorber instead of CsSnI 3 NCs.T he highest PCE of CsSn 0.6 Ge 0.4 I 3 NCs-based solar cells in this work, 4.9 %, is one of the highest values reported for all-inorganic lead-free PNCs solar cells.T he introduction of Ge is highly beneficial for ambient, thermal, and photostability,d ue to the effective protection of Ge 2+ against Sn 2+ oxidization. We believe that, upon optimization (for example,b yi dentifying the most suitable selective contacts matching the CsSn 0.6 Ge 0.4 I 3 NCs energy levels), the solar cells based on SnGe-PNCs can outperform current generation of lead-free PNCs based solar cells with the additional benefit of outstanding stability,h ence bringing PNCs-photovoltaics astep closer to their commercialization.