Perovskite-Quantum Dots Interface: Deciphering its Ultrafast Charge Carrier Dynamics

Understanding electron and hole (e,h) transport at semiconductor interfaces is paramount to developing efficient optoelectronic devices. Halide perovskite/semiconductor quantum dots (QDs) have emerged as smart hybrid systems with a huge potential for light emission and energy conversion. However, the dynamics of generated e-h pairs are not fully understood. Ultrafast UV-VIS transient absorption and THz spectroscopies have enabled us to unravel the processes of the e-h recombination within a hybrid film of methylammonium lead triiodide (MAPbI3) interacting with different amount of PbS/CdS core/shell QDs. To accurately analyse the complex behaviour, we applied a new model for e-h events in this hybrid material. The results obtained with sample having a high concentration of QDs (7.3 mass percentage) indicate: (i) a large population (92%) of the photogenerated charge carriers are affected by QDs presence. The main part of these carriers (85% of the total) in perovskite domain diffuse towards QDs, where they transfer to the interface (electrons) and QD ́s valence bands (holes) with rate constants of 1.2×10 s and 4.6×10 s, respectively. 7% of these affected charged entities

The use of perovskite as a passivating agent for QDs surfaces led to a material that can act as a highly conductive matrix to increase the power conversion efficiency (PCE) of photovoltaic devices based on lead trihalides [19,27], as well as the efficiency of IR LEDs [25] and photodetectors [26]. Perovskite and small core/shell QDs have been combined to prepare voltage tunable LEDs that exhibit exciplex emission, which could be used for developing advanced solar cell configurations, such as intermediate band gap ones [23]. In addition to that PbS QDs have been used as seeds to grow perovskite solar cells with larger grain sizes, significantly increasing the performance of the solar cells [19,24]. These results clearly show the enormous potential of combining perovskite and QD materials to develop highly efficient advanced optoelectronic devices. However, detailed studies of the photoinduced processes that occur in these composites are lacking. Understanding the ultrafast events in perovskite/QDs interfaces, such as charge carrier transfer and recombination dynamics, is paramount to improving the overall efficiency of optoelectronic devices based on these materials. However, the real-time observation of the interfacial processes remains unexplored.
Herein, we report on studies of thin polycrystalline films of methylammonium lead triiodide (MAPbI3) matrix embedded with different concentrations of PbS/CdS (core/shell) QDs, using ultrafast time-resolved transient visible-NIR absorption and terahertz (THz) spectroscopies (TAS and TRTS, respectively). By photogenerating charge carriers over the band gap of MAPbI3 (~1.55 eV) in the MAPbI3/QDs hybrid material, we observed fast and efficient charge carrier deactivation due to their transfer to QDs. The initial (< 1 ps) population and mobilities of electrons and holes in the perovskite domain are not affected by the presence of QDs. To accurately elucidate the complex electrons and hole events within the material, we propose a new kinetic model. We found that: (i) the majority (~90 %) of photoexcited electrons and holes diffuse towards QDs and is afterwards transferred to the interface and QDs, respectively, (ii) few percent of them fast recombine as excitons in the close vicinity of the interface, and (iii) a small part of the photoexcited charges is not affected by the presence of QDs, and therefore follows the typical perovskite deactivation channels. Lowering the QDs concentration increases the carrier diffusion time, and thus slows down their effective transfer to the interface and QDs, and decreases the concentration of excitons. These results show high collection efficiency of electrons and holes by these QDs in this hybrid material, and the relevance of MAPbI3/QDs interface states in charge carrier transition/trapping processes. We propose that further optimization of perovskite/QD system requires a reduction of the interface traps. Of not less importance, the proposed kinetic model can be used to characterize the electron-hole events in other complex hybrid structures.

PbS/CdS core-shell QD synthesis
Core/shell quantum dots (QDs) were synthesized according to a previously reported procedure [24,32]. Briefly, a mixture of 0.9 g (4 mmol) of PbO, 2.7 g (9.6 mmol) of oleic acid (OA) and 36 ml of 1-octadecene (ODE) in a three-necked round-bottom flask was heated to 150°C under N2 to form Pb-oleate moieties. The solution was degassed for 30 min under vacuum, then opened to N2, and 3 ml (6.7 mmol) of trioctylphosphine (90 %) was injected. A mixture of 0.42 ml (2 mmol) of hexamethyldisilathiane (HMDS) and 4 ml of ODE was quickly injected to the flask when the temperature dropped to 110°C. The solution was left to cool to room temperature. The reaction product was cleaned three times with ethanol/acetone (1:1, v/v), centrifuged (3000 rpm for 10 min) and dispersed in toluene (100 mg/ml).
For the growth of the CdS shell, a flask containing 0.34 g (2.6 mmol) of CdO, 1.85 g (6.5 mmol) of OA and 40 ml of ODE was heated to 220°C in air to dissolve CdO, cooled to 150°C and then degassed for 1 hour under vacuum. After that, the temperature was reduced to 70°C. 5 ml of the as-prepared PbS QD suspension (100 mg/ml in toluene) was rapidly injected to the flask. The reaction was kept at 70°C for 5 min then stopped by adding the non-solvent mixture (ethanol/acetone, 1:1, v/v). The product was washed following the same washing steps used for PbS nanocrystals and dispersed in octane for the ligand exchange. After this process, PbS/CdS QDs 3 nm in size were prepared.

Perovskite and perovskite/QDs film preparation
Glass substrates were cleaned with soap, sonicated in distilled water, ethanol and isopropanol for 15 min and then treated with an UV−O3 lamp for 15 min. MAPbI3 or MAPbI3/QD films were spin coated at 9000 rpm from 50 µl of MAPbI3 or MAPbI3/QD solution, respectively. Diethyl ether was poured on a film after the spin coater was running for 4 sec. The films were annealed for 1 min at 65°C then for 2 min at 100°C. The deposition was carried out inside a glovebox filled with N2.

SEM and XRD measurement
The morphology and structural properties of the films were analyzed using a field emission scanning electron microscope (JSM7001F, JEOL) and a Bruker AXS D4 X-ray diffractometer using Cu Kα radiation.

UV-VIS-NIR absorption spectroscopy
The UV-VIS-NIR absorption spectra of the MAPbI3, MAPbI3/QDs and QDs were measured using a standard spectrophotometer (JASCO V-670). To eliminate the contribution of the scattered light, we carried out diffuse reflectance measurements using a 60-mm integration sphere (ISN-723).

Femtosecond transient absorption spectroscopy
The used femtosecond (fs) transient UV-VIS-IR absorption setup has been described elsewhere [33]. Briefly, it consists of a Ti:sapphire oscillator (TISSA 50, CDP Systems) pumped by a 5 W diode-laser (Verdi 5, Coherent). The oscillator output pulses (30 fs, 480 mW at 86 MHz) centered at 800 nm were guided to a regenerative amplifier (Legend-USP, Coherent) and used as a seeding signal. The amplified fundamental beam (50 fs, 1 W at 1 kHz) was then directed through an optical parametric amplifier (OPA, CDP Systems) for wavelength conversion and an additional 1-mm BBO crystal for frequency doubling to obtain 400, 600, 700 and 740 nm pulses. The pump intensities ranged from ~ 40 to 250 μW. The transient absorption measurements were performed in the spectral ranges of 450 -780 nm. All the spectra analyzed in the visible-NIR region were corrected for the chirp of the white light continuum. To avoid sample degradation, the samples were moved during the measurement using the XY translational stage. The measured instrument response function (IRF) of the system was ~ 70 fs. All the experiments were performed at 293 K [34].

Sub-picosecond time-resolved terahertz spectroscopy
The terahertz (THz) experiments were done using the same laser setup described in the previous paragraph (Legend-USP regenerative amplifier seeded by the Ti:sapphire oscillator).
The amplified fundamental beam centered at 800 nm (50 fs, 1 W, 1 kHz) was divided into three parts. The first one (~ 700 mW) was directed through an OPA and additional 1-mm BBO crystal for frequency doubling to obtain 600 nm pulses. The resulting beam is sent through a long (up

Flash Photolysis Spectroscopy
The nanosecond (ns) flash photolysis experimental setup was described previously [35].

Morphology, structure and absorption spectra
The morphology and structure of MAPbI3 and MAPbI3/QDs (6.3 wt%, QDs size ~ 3nm) films were characterized by SEM ( Figure S1, for wt% calculation, see Equations S1). Both samples show comparable fine polycrystalline structure (grains size ~30 -100 nm). The thicknesses of MAPbI3 and MAPbI3/QDs films are 80±10 and 120±10 nm, respectively. Proper perovskite crystalline structure of both samples was proved also using XRD technique ( Figure   S2a). Signal peaks at 14°, 28° and 32° detected on MAPbI3 and MAPbI3/QDs correspond to {110}, {220} and {310} atomic planes of MAPbI3, respectively [19,27,36]. Steady-state absorption spectra of MAPbI3, MAPbI3/QDs (6.3 wt%) and QDs not attached to perovskite matrix are presented in Figure S2b. The spectra of MAPbI3 and MAPbI3/QD are comparable and very similar to those already published on this type of perovskite [24,37]. The presence of QDs is mainly manifested by differences in the absorption of MAPbI3 and MAPbI3/QDs films in the spectral region > 800 nm.

Transient spectra and decays
To begin with, we first show and discuss the results of neat perovskite and MAPbI3/QDs film containing 6.3 wt% of the QDs upon excitation at 600 nm, using fluence of the absorbed photons 8.210 12 ph/cm 2 . The TA spectrum of MAPbI3 film consists of two photobleaching bands (PB) with intensity maxima located at approximately 480 nm (PB1) and ~763 nm (PB2), and positive band (photoinduced absorption, PIA) in the region 520 -700 nm ( Figure 1a). This result is in agreement with previous reports on neat perovskite films [38][39][40]. The initial negative signal at 760 nm (PB2) shifts slightly towards longer wavelengths upon increasing pump-probe delay time (up to 700 fs) due to the band filling process ( Figure S3a) [41][42][43]. No further spectral changes of PB2 were recorded. The TA spectra of the MAPbI3/QDs and Consequently, the blue shift in the PB2 band of MAPbI3/QDs is due to changes in the electronic and/or crystallographic structure of MAPbI3 in presence of the QDs. Such modifications were observed after passivation by the QDs due to the lattice mismatch between the two materials [18,21,44,45]. The interpretation of the blue shift also agrees with reported band gap expansion of MAPbI3 related to structural deformation [46].  35% of its initial value at 1.5 ns. This behavior is caused by the presence of monomolecular (trap-assisted), bimolecular and Auger charge carrier deactivation processes [34,42,[48][49][50]. It is known that the contribution of each process to the overall decay depends on perovskite composition and concentration of charge carriers. We have quantified the recombination rate constants of these events adopting a kinetic model applied previously for neat perovskite materials (Equation S3, for more details see Supplementary information) [40,42,51,52]. The obtained values of the rate constants (kmono = 1.6×10 7 s -1 , kbimol = 8.1×10 -11 cm 3 s -1 and kAuger = 2.9×10 -28 cm 6 s -1 ) are comparable to previously published ones [51][52][53].
In contrast to the TA spectra, the decays of excited MAPbI3/QDs are significantly affected by the presence of QDs (Figure 1d). Except the long-lasting residual signal ~10% (after 1.2 ns), the decay is being done within the first ~200 ps (only 20% of the initial intensity is left after 100 ps). While the origin of this behavior will be discussed in more detail further, we briefly assign it to the following events. The fast decay is mainly caused by diffusion and subsequent transfer of the photoexcited holes from the perovskite valence band (VB) to the one of the QDs, and the electrons from perovskite conductive band (CB) to the MAPbI3/QDs interface. The residual signal is related to charge carriers in the perovskite domain of MAPbI3/QDs, not affected by the presence of QDs. It is evident that charge carriers diffusion and transfer from MAPbI3 to QDs dominates the dynamics in this composite material ( Figure   S3c). A new kinetic model is necessary to analyze the photoevents, and related electrons and holes dynamics in the excited MAPbI3/QDs film.
Analyzing the short-time scale (< 1ps), the TA signals of the MAPbI3 and MAPbI3/QDs (6.3 wt%) show ultrafast rising components ( Figure S3d). We were able to fit the rising part using a monoexponential function obtaining s time constant of ~150 fs for both MAPbI3 and MAPbI3/QDs films. Such behavior of the TA signal during the first 100s fs after excitation is typical for perovskite materials, and it reflects hot charge carrier cooling [41,54,55]. Since we recorded the same risetime and comparable OD changes within 1 ps after excitation of both MAPbI3/QDs and MAPbI3 films, we conclude that in the sub-ps regime, the charge carrier dynamics in the perovskite domain is not affected by the presence of QD during the initial cooling phase in the MAPbI3/QDs film.

Influence of QDs concentration
To elucidate the effect of QDs presence on charge carriers deactivation, we have recorded and analyzed the TA decays of excited MAPbI3/QDs containing different CQDs. Note that for clarity, the scheme in Figure 3b is simplified, and in D1 the photoexcited population may still follow the dynamics described in Figure 3c. to the CB of the QDs is possible. However, the occurrence of this process is unlikely due to the presence of highly competitive ultrafast electron cooling in perovskite domain, which has much shorter time (<1 ps, Figure S3d) [38,40] than that of TA decay in MAPbI3/QDs (~200 -500 ps, Figure 2). Secondly, the electrons in the CB of perovskite domain may recombine directly with the holes transferred to the QDs VB (via no vertical transitions). However, presence of such fast recombination process would be in contradiction with long and efficient PL conversion observed in this material [23]. Moreover, very similar spectral shapes and positions of the PB2 bands in the MAPbI3 and MAPbI3/QDs films suggest that the ultrafast electron-hole recombination occurs mostly in the perovskite domain. Thirdly, electrons can be transferred and subsequently trapped in perovskite/QDs interface. Indeed, it has been reported that, in the presence of QDs, the crystallographic structure of the MAPbI3 lattice might be different inside the crystallite and at the interface with the QDs, thus, new trap states will appear [18,21,44].
The presence of interfacial electron trap states in MAPbI3/PbS was also proved by the density functional theory calculations [18]. This conclusion is in agreement with our steady-state absorption measurements in the near IR region (the absorption of MAPbI3/QDs at > 800 nm, Figure S2b). It is important to note that before the carriers can be transferred, they need to be transported to MAPbI3/QDs interface first. Therefore, not only the transfer processes affect the fact deactivation of charges in D1, but also the transport time of charge carriers. We label the combination of both these effects as effective charge carrier transfer.
(ii) It is well known that the binding energy of excitons in MAPbI3 is below the thermal energy at ambient conditions (25.7 meV) that should lead to efficient formation of free charge carriers as a result of the photoexcitation [52]. However, the changes in the perovskite lattice in the close vicinity of QDs, can induce and increase localization of photoexcited charges, related excitons bounding energy and Langevin coefficient [56,57]. This may result in an initial formation of excitons able to survive several picoseconds without dissociation in the close vicinity of QD. We believe that this behavior is the origin of the observed initial fast TA decay (first few ps) of excited MAPbI3/QDs film ( Figure S3c) [34]. Appearance of the interface structural defects in MAPbI3/QDs films were already discussed in Section 3.2.1. The presence of excitons was also verified by the fluence dependence of TA decays and comparison of THz measurements of MAPbI3/QDs that show the absence of the initial fast signal decrease (vide intra, Section 3.7.). This agrees with the reports of TA signal decay in neat PbS/CdS quantum dots, interpreted on the basis of a surface exciton deactivation model [58]. Thus, the latter explanation seems to be the most plausible one (vide intra, Sections 3.4. and
In addition to the discussed processes above, the electron transfer from QDs to MAPbI3/QDs is also possible. However, taking into account the absence of TA signal of neat QDs layers, and comparable initial TA values of excited neat perovskite and MAPbI3/QDs at 760 nm, the contribution of such transfer to the signal should be negligible.

Mathematical model
To quantify the processes in D1, we propose a following set of kinetic equations based on previous information [59][60][61]: The overall decay of observed TA signal should be given by the sum of those in D1 and D2: = ( ) = 1 ( ( ) + ℎ ( ) + ( )) + 2 ( ) (2) PD1 and PD2 characterize the population fractions of charge carriers in D1 and D2 areas. We call n(t) the population of the charge carriers in D2, as described in Equation S3. To simplify the model, we did not take into account the ultrafast dynamics of charge carrier generation, their thermalization and cooling. These events happening commonly in less than 1 ps [52]. Figure 2 shows the fits of the TA decays using the kinetic model, and Table 1 gives the obtained data from the best fits (for TA decays converted to charge carrier concentration, see Figure S4). The values of kmono, kbimol and kAuger used in the fits were taken from the analysis of the photobehavior of neat perovskite and related known model (for more information about the of the model, see Supplementary information) [60,61].

Analysis of the transient absorption decays
Firstly, we discuss the observed dependence of khtrans and ketrans values on CQDs (Table   1). The dependence is visualized in the Figure

Transient absorption: fluence dependence
To support the kinetic model, we measured the TA decays of MAPbI3/QDs films containing high (6.3 wt%) and low (0.3 wt%) concentration of QDs, using different fluences of the absorbed photons. Figure 4a shows the decay of TA signal converted to the concentration of charge carriers of MAPbI3/QDs (6.3 wt%) upon excitation at 600 nm, using fluences of the absorbed photons: 1.5×10 12 , 8.2×10 12 and 16×10 12 ph/cm 2 . The similarity of the decays proves that non-monomolecular deactivation channels are significantly closed, and the electron/hole deactivation processes should be primarily driven by the interaction between the perovskite and QDs ( Figure S6a). In fact, bimolecular and especially the Auger processes are fluence dependent, and thus their contribution would cause significant shortening of the TA decays at a higher fluence of the excitation [48]. It is noteworthy that the amount of charge carriers responsible for the TA residual signal at longer times (> 400 ps) increases almost linearly with the fluence, and does not saturate even at the used highest fluences of the absorbed photons.
Thus, these carriers should not be related to the localized states in perovskite domain, but they are more likely free charge carriers not affected by QDs. In fact, concentration of trap states in perovskite domain was reported to be lower than photons absorbed per laser pulse using highest pump fluence [52]. The only difference between the recorded decays is their small shortening within the first ~5 ps, when decreasing the fluence of the absorbed photons (Figure 4a, inset).
While the influence of bimolecular or Auger processes would have exactly opposite effect, this behavior is also suggesting the suppression of higher order processes in MAPbI3/QDs [48]. We interpret this behavior in terms of rapid deactivation of charges related to limited amount of states that show saturation under stronger excitation. This is in agreement with our suggestion of excitonic states presence in perovskite domain of D1 in close vicinity to the MAPbI3/QDs interface. All the decays in Figure 4a were successfully fitted using the proposed kinetic model.
The obtained data are shown in Table S1. Taking into account that doubling the fluence of the absorbed photons from 8.2×10 12 to 16×10 12 ph/cm -2 causes only small increase of exciton amount in MAPbI3/QDs (6.3 wt%), the concentration of excitonic states in this sample can be estimated as ~ 5×10 16 cm -3 .
Comparable results were observed for MAPbI3/QDs containing low CQDs (0.30 wt%) ( Figure 4b, S6b and Table S1). However, as the inset of Figure 4b shows, there is no variation of the decays with excitation fluence at short time regime indicating a lack of excitons (low CQDs).
The risetimes of the MAPbI3/QDs (6.3 wt%) TA signal increases from 110 to 230 fs (longer than the IRF ~ 70 fs) when the fluence of the absorbed photons varies from 1.5×10 12 to 16×10 12 ph/cm 2 ( Figure S6c, Table S2). The observed behavior is not caused by QDs presence, but it is the result of state saturation and the phonon bottleneck processes happening in perovskite as it has been previously reported [41,54,55]. This explanation is also supported by the very similar behavior of the neat perovskite film ( Figure S6d, Table S2).

Transient absorption: excess energy of excitation
To  (Table S3). In addition to that, we observed an increase in the TA rising time from 70 to 420 fs when increasing the excess energy of the charge carriers ( Figure S7 inset and Table S4). This effect is due to cooling of hot charge carriers in perovskite, and it is not caused by the presence of QDs, as we explained above [38,40].

Flash photolysis
To get information on the longer-lived species, we used the nanosecond-millisecond flash photolysis technique. We excited the films at 460 nm and observed the decays at 580 nm.
For the MAPbI3 film, the fit of the decays requires two exponential components showing time constants of P = 50 ns and  P2 = 560 ns, which are attributed to the recombination of the trapped charge carriers from the surface and the internal trap states of perovskite crystals, respectively ( Figure S8). These results are in agreements with previous reports [12,40,52,64].
However, accurate multi-exponential fit of the MAPbI3/QDs (6.3 wt%) decay needs four components: two of them are similar to those of neat perovskite (P1 = 50 ns and P2 = 560 ns), and two others in the microsecond regime (P3 = 3.8 s and P4 = 45 s). The appearance of these additional components is a result of the interactions between both materials, allowing longer deactivation processes which could be attributed to the transitions involving the electrons states on the interface, holes in QDs and exciplex states [23]. The presence of long lived interface states is also in good agreement with proposed mechanisms of photoexcited electron transfer.

Time-resolved Terahertz Spectroscopy
The charge carrier mobility and its time evolution are key parameters shaping the efficiency of optoelectronic devices. Therefore, we used ultrafast time-resolved THz spectroscopy to explore the dynamics of photoconductivity in neat MAPbI3 and MAPbI3/QDs (IRF < 1 ps). To begin with, Figure S9a presents the photoconductivity decays of the MAPbI3 film upon excitation at 600 nm, using a fluence of the absorbed photons of 8.2×10 12 ph/cm 2 .
The initial mobility has s value of 16.5 cm 2 V -1 s -1 , which is in the range of previously reported

Conclusions
In summary, we have characterized the ultrafast dynamics of photoexcited charge carrier transfer and recombination processes in a MAPbI3 thin polycrystalline film containing    Solid lines are from the best fits using the kinetic model described in the text.