Effect of Chloride Passivation on Recombination Dynamics in CdTe Colloidal Quantum Dots

Colloidal quantum dots (CQDs) can be used in conjunction with organic charge-transporting layers to produce light-emitting diodes, solar cells and other devices. The efficacy of CQDs in these applications is reduced by the non-radiative recombination associated with surface traps. Here we investigate the effect on the recombination dynamics in CdTe CQDs of the passivation of these surface traps by chloride ions. Radiative recombination dominates in these passivated CQDs, with the radiative lifetime scaling linearly with CQD volume over τr=20–55 ns. Before chloride passivation or after exposure to air, two non-radiative components are also observed in the recombination transients, with sample-dependent lifetimes typically of less than 1 ns and a few ns. The non-radiative dynamics can be explained by Auger-mediated trapping of holes and the lifetimes of this process calculated by an atomistic model are in agreement with experimental values if assuming surface oxidation of the CQDs.


Introduction
Colloidal quantum dots (CQDs)h ave applicationsa st he light emitting or absorbing speciesi narange of optoelectronic devices based on organic charge transporting layers,i ncluding light emitting diodes, [1] photovoltaic cells, [2][3][4] photodetectors, [3,5] holographic data stores and image processors. [6] CQDs are well-suited for use in combination with organic materials because they can also be synthesised and processed by costeffective and scalable solution-based methods. Moreover,t heir emission wavelength and absorption edge is size-tunable, enabling facile optimisation forp articulard evice designs, and they can also benefit from high photoluminescence (PL) quantum yield (QY) and narrow band emission. [7,8] However,t he small size (typically 2-5 nm) and consequently large surface-area-to-volume ratio of CQDs can result in ah igh concentration of trap states associated with unsaturated bonds on the surface. These surface traps can enablen on-radiative recombination pathways that compromise device efficiency by reducing the QY of radiative recombination or charge extraction,o rb yd ecreasing chargem obility. [9,10] Suppression of unwanted surface-related recombination is thuse ssentialtomaximising the performance of devices incorporating CQDs. [2,[11][12][13] CQD surface traps have previously been passivated either by organic ligandso rb yawide-band-gap semiconductor shell. [14,15] However,b oth of these techniques have their drawbacks. Ligand coverage of the surface has provedt ob ei ncomplete, [16,17] leaving somet raps unpassivated, but introducing ab arrier to charget ransfer.T he wide-band shell can prevent chargesp hotogenerated within the CQD interacting with surface traps, resulting in nearly 100 %P LQY if the CQD-shell is grown withoutd efects [18][19][20] but inhibiting charge transport. [21] Passivation of surface traps with halide ions has recently emerged [12,22] as ap romising method to significantly improve device performance, resulting, for instance, in an increase to nearly 9% in the record efficiency for aC QD-baseds olar cell. [23] These impressive results have been attributed to the ability of compacth alide ions to passivate sites on the CQD surfacet hat are inaccessible to bulkierl igands, and to do so without introducing ac harget ransport barrier. [12] Halide passivation also greatly reduces the sensitivity of CQDs to air exposure, by binding to sites on the CQD's surfaceo therwise subjectt oo xidative attack, allowing device fabrication in ambient conditions and thus reducing potential production costs. [24] Until now,t he PLQY values reported for halide-passivated CQDs, which depend on the ratio of the rates of radiative and non-radiative recombination, have all been significantly less than unity indicating that the complete passivation of surface traps has still not been achieved, despite the impressive im-Colloidal quantum dots (CQDs) can be used in conjunction with organic charge-transporting layers to produce light-emitting diodes, solar cells and other devices. The efficacy of CQDs in these applicationsi sr educed by the non-radiative recombination associated with surfacet raps. Here we investigate the effect on the recombination dynamics in CdTeC QDs of the passivation of these surface traps by chloride ions. Radiativerecombination dominates in these passivated CQDs, with the radiative lifetime scaling linearly with CQD volume over t r = 20-55 ns. Before chloride passivation or after exposure to air,t wo non-radiative components are also observed in the recombination transients, with sample-dependent lifetimes typically of less than 1nsa nd af ew ns. The non-radiative dynamics can be explained by Auger-mediated trapping of holes and the lifetimes of this process calculated by an atomistic model are in agreement with experimental values if assuming surface oxidation of the CQDs.  provements in device performance. However,w eh ave recently developed at echnique that uses chloride ions to passivate CdTeC QDs that can result in near-unity PLQY that is, almost complete surface passivation. [25] In this work, we report as tudy of the effects of chloride ion passivation on recombination dynamics in these CQDs.T he simplified recombination transient in the passivated CQDs allows the underlying charge dynamics to be studied free of the sample-specific influence of trapping, which hitherto has complicated analysis. These resultsa re also compared with the dynamics of unpassivated CQDsa nd with those exposed to air,e nabling the study of the dynamics of the trapping process itself. Recent work has shown that the Auger-mediated trapping (AMT) of holes can be used to explain the charged ynamics observed in an umber of CQD types, including CdSe and InAs/ZnSeC QDs. [26] An atomistic model of aC dTeC QD has been developed and is used to calculate AMT rates for holes, and the results are compared to the non-radiativel ifetimes found experimentally.

Radiative Recombination
Representative normalised absorption and PL spectra for as ample of CQDs before and after the surface passivation treatment are shown in Figure 1. The spectra are largely unchanged by the treatment process, except for ar ed-shift of about l = 10 nm. This spectral shift on passivation has previously been shown to depend on the treatment time and the concentration of chloride ions used. [25] Moreover,t he addition of chloride ions to the CdTes urfacew as shown to fill states near to the valance band maximum, narrowing the band gap and thusr ed-shifting the emission peak and absorption edge. [22] Notably,t he chloride treatment produces ad ramatic increasei nP LQY,f rom typically about 5% before treatment to greater than 90 %i ns ome cases afterwards. The positiono f the PL and absorbance peaks, and the QY values for all the samples used in this study,b efore and after passivation,a re given in the Supporting Information (see Ta ble S1 in S1).
As noted previously, [25] the dominance of radiativer ecombination in passivated CQDs,a se videnced by the near unity PLQY,a lso affects the form of the PL decay transients, I PL (t). As shown in Figure 2f or at ypical sample (and in the Supporting Information for the other samples reported here, see Figure S1), before passivation I PL (t)i sm ulti-exponential in form. In contrast, after passivation the transienti sw ell-described by am ono-exponential decay,a nd hence can be characterisedb y as ingle time constant, t PL .T his mono-exponential form was found to be independento fe xcitation powero ver nearly 2 orders of magnitude (see Figure S2).
The form of I PL (t)for passivated samples allows the radiative lifetime, to be estimated as t r % t PL /QY,aprocedure only possible if the transient is mono-exponential.P Li ntermittency ("blinking"),w hich can also result from trapping, [27] decreases the number of CQDs that contributet ot he PL decay,r educing QY but leaving t PL unaffected. Similarly,t rapping of hot excitons [28] also reduces the QY without affecting the rate of bandedge recombination. Thus, the true radiative lifetimel ies between t PL and the value estimated from t r % t PL /QY,with the accuracy of the estimate improving for high QY,a sf or the samples studied here. Figure 3s hows the values of t r ,w hich range between 20 ns and 55 ns, found in this way for transients at the PL peak for an umber of different samples, as af unctiono f CQD volume. The diameter of each CQD was determined from the positiono ft he absorption edge, using ap reviously reported empirical relationship, [29] and the volumec alculated assuming as pherical shape. As expected for radiative decay, [30] t r scales linearly with CQD volume. In comparison, previouss tudies [31,32] have reported radiativel ifetimes ranging between t r = 20 ns and t r = 40 ns for CdTeC QDs of similars izes. However, these works directly associated the radiativel ifetime with the observed PL decay constant that is,a re more comparable with the values of t PL found here. Thus, also shown in Figure 3a re  www.chemphyschem.org t PL and the PL lifetimes, t calc ,c alculated from the empirical relationship between the recombination rate and the peak PL photon energy reported by de Mello Donega et al. [32] The values of t PL and t calc are broadly comparable, whereas the values of t r are systematically longer than either t PL or t calc , consistentw ith t r % t PL /QY representing the upper limit for radiative lifetime.
The PL peak is widened by the size dispersion in as ample, with smaller than average CQDs contributing to the short wavelength side of the peak and larger CQDs to the long wavelength side.T hus, measuring I PL (t)a td ifferent wavelengths across the PL peak allows different diameters of CQDs to be studied, within the overall size distribution of the sample. As shown in the Supporting Information (see Figure S4), ap rogressive slowing of the PL decay is seen for longer wavelengthsw ithin the PL peak. The diameter of the CQDs corresponding to each of these decays can also be calculated from the PL wavelength used, taking into account the Stokes shift between PL peak and absorption edge. Assuming ac onstant QY value for all CQD diameters, the values of t r extracted at different wavelengths across the PL peak were extracted and are shown in Figure 4. Here, t r also scales linearly with CQD volumes howing that the observed change in decay transienta cross the PL is due to the different CQD diameters sampled at different wavelengths.

Non-radiative Recombination
I PL (t)o bserved for the untreated samples could be well-described by tri-exponential of the form [Eq. (1)]: in which t r was fixed to the value extracted for the same sample after chloride passivation; t f and t s are time constants associated with additional fast and slow non-radiative components, respectively;a nd A r , A f and A s are the amplitudes associated with the radiative, fast non-radiativea nd slow non-radiative components, respectively.T he values of t f and t s found from af it to I PL (t)f or an umber of samples of different CQD diameters are shown in Figure 5. In this case, there is no systematic variation with CQD volume, confirming that these time constantsa re not associated with radiative recombination. As discussed in Section2.1, the true value of the radiative lifetime will lie between t r % t PL /QY and t PL ,a nd so to check the sensitivity to the value used for radiative lifetime, the tri-exponential fitting procedure was repeated with the lifetimeo ft he first component fixed to t PL rather than t r .T he resulting values of t f and t s (see Figure S5) are similar thoughr educed. However, as ystematic variation with CQD volume is now evident, indicating that the values of t f and t s are somewhat influenced by radiativer ecombination if t PL is used for the first component results. That is, there is al ess reliable separation of the radiative and non-radiatived ynamics.

Effect of Exposure to Air
I PL (t)c hanges significantly for both chloride-treated and untreated samples on exposure to air,i np arallelw ith ad ecrease in QY, [25] indicating that non-radiative pathways are produced   www.chemphyschem.org or alteredb yo xidation. As shown in Figure 6, I PL (t)f or ap assivated sample becomes multi-exponential in form over an umber of hours after first contact with the air,a nd can now be well-described by Eq. (1), by using the value of t r found from the transiento btained before exposure to air.F or increasing air exposure, the time constants remainc onstant but the amplitudes of the non-radiative components increase at the expense of A r ,c onsistent with the diminishing QY,w ith A s showingt he greatest increase. For the untreated sample A r also decreases with exposure to air,a gain consistentw ith the reducing QY,b ut in this case only A f increases with A s reducing somewhat with oxidation time. The behavior of the fractional contribution of each component is shown in Figure 7f or both the case of untreated and chloride-treated CQDs.

Auger-mediated Trapping Model
If ap hotogenerated hole undergoes a( trapping) transition to as tate in the gap, itse xcesse nergy can be transferred non-radiatively to the electron, exciting it to as tate above the conductionb and edge. This AMT process, has been used recently [26,33] to explain many features observed experimentally in the charge dynamics of CQDs made of different materials in different environments. Here we extend itsa pplication to investigate PL dynamics in CdTed ots. AMT rates for the transition from the band edge Exciton, j e CBM ,h VBM > ,t oasurface-trapped Exciton, j e n ,h trap > (in which e n is an excited electron state and h trap as urfaceh ole trap state) werec alculated in CdTeC QDs with R = 2.3 nm, by using standard time-dependent perturbation theory [34] and LDA-quality wave functions obtained within the atomistic semiempirical pseudopotential method, (further details on the method can be found in Refs. [26,33]), for trap states-Teu nsaturated bonds-located at different positions on the surface. As in the case of CdSe, [26,33] the calculated AMT times were found to dependo nt he number of dangling bonds of each surfaceT ea tom:f or dots dispersed in common solvents (such as toluene), traps obtained from atoms with single dangling bonds were generally less efficient (i.e. had as lower trapping time) than those created from atoms with double dangling bonds. [35] Unlike the case of CdSe, though, the distribution of the trapping times was narrow,r anging from af ew picoseconds to af ew hundreds of picoseconds (see the green bars in Figure 8), as it was found for InAs [26,33] (which has the same crystal structure as CdTe). This result is qualitatively consistentw ith our experimental observation of as low and af ast non-radiativec omponent in the PL decay. The value of the trapping times depends on the dielectric environment of the trap. The above calculations assumed that the dots were embedded in am atrix with ar elative dielectric constant of e = 2.2, which corresponds to the toluene used as   www.chemphyschem.org as olvent. For av alue of e = 6, however,A MT becomes less efficient for all traps, with the transfer times to the efficient traps increasing by about 3o rders of magnitude,f rom af ew picoseconds to af ew nanoseconds (see the black bars in Figure 8), that is, in broad quantitative agreement with the experimentally measured time constantsf or non-radiativer ecombination. According to ab initio many-body calculations [36] the dielectric constant for cadmium oxide (CdO) has this value, which, coupled with the increasingcontribution of non-radiativerecombination for greater exposure to air noted above,i ndicates that oxidation of the CQD surface plays ak ey role in producing the nanosecondt ime scale non-radiative recombination observed. The tetradecylphosphonic acid used to passivate the untreated CQDs also bonds Cd atoms to oxygen, which,g iven that the same time constantsd escribe I PL (t)b efore and after exposure to air,s uggestst hat this too produces ad ielectric environment resultingi nn anosecond time scale non-radiative recombination. The relative contributions of the fast and slow non-radiative components, and their evolution upon exposure to air,d etailed in Sections 2.2 and 2.3, reflect the effect on the distribution of trapping times of the different ligands and chloride ions on both the passivation of dangling bonds and on the local dielectric environment.

Summary and Conclusions
The effect of chloride ion passivationont he recombination dynamics in CdTeC QDs wass tudied. The almost completep assivation of the chloride-treated CQDs produced am ono-exponentialp hotoluminescence decay transient that enabled the radiativel ifetimet ob ed eterminedf ree of the sample-specific contribution of surfacet rapping, which was found to scale linearly with CQD volumea se xpected. Before the chloride treatment and after exposure to air,t he decay transient could be well-described by at ri-exponential, characterised by the radiative lifetimea nd fast and slow time constants associated with non-radiative recombination. The contribution of these non-radiative components grew on exposure to air for both the treated and untreated samples.
These observations are consistent with am odel of non-radiative recombination based on the AMT of holes by surface dangling bonds. Calculations of trapping times by using an atomistic semiempirical pseudopotential method demonstrated that the there are two types of holes traps, corresponding to Te atoms with either one or two dangling bonds. Each trap type produced ad istribution of trapping times depending on the location of the Te atom on the CQD surface. The values of the trapping times dependedo nt he dielectric environment of the trap. For ad ielectric constant corresponding to the solvent used in the experiments,t he trapping times were in the picosecond time scale. However,u sing ad ielectric constant corresponding to CdO produced trappingt imes on the nanosecond scale, in agreement with experiment. This suggests that oxidation not only produces traps, as evidenced by the increasing non-radiative contribution to the decay transiento ne xposure to air,b ut also determines the time scale of non-radiativer ecombination.

Experimental Section
The CdTeC QDs were synthesised by using ap reviously published method [18] and well-controlled growth times to produce different diameters, capped with tetradecylphosphonic acid and trioctylphosphine ligands. The chloride treatment has been described in detail recently [25] and the same procedure was followed here. The concentration of chloride used in the treatment was equivalent to ad ensity on the CQD surface of 96 ions/nm 2 ,a nd it has been shown [25] that as imilar density of oleylamine ligands was also present on the surface. All treated and untreated samples were diluted in toluene and placed in quartz cuvettes (always sealed under N 2 atmosphere) for further investigation. The samples were stable, with unchanging QY,i fkept under an inert atmosphere. [25] PL transients were recorded by using time correlated single photon counting. The pump beam was from amode-locked Ti :sapphire laser (Mai Ta iH P, Spectra-Physics), providing t = 100 fs pulses at ar epetition rate of 80 MHz and aw avelength of l = 820 nm. With the use of an acousto-optic pulse picker (Pulse Select, APE) the rate was reduced to 2MHz, and then the wavelength was converted to l = 410 nm through second harmonic generation (APE Harmonic Generator). After excitation the PL emission was directed into am onochromator (Spex 1870c) and detected at the PL peak (or tuned at the preferred wavelength within the range of the PL spectra of each sample) by am ulti-channel plate (Hamamatsu R3809U-50). The time correlation of the detected photons was performed by using aPCc ard (TCC900, Edinburgh Instruments).
The PL emission spectra and PLQY for the CQD samples were measured by using as pectrofluorometer (FluoroLog 33-22iHR, Jobin-Yvon) with ab uilt-in integrating sphere (F-3018, Jobin-Yvon). The excitation wavelength was set to l = 450 nm with ab andwidth of 1.3 nm. Absorbance spectra were obtained by using aP erkinElmer Lambda-1050 spectrophotometer.