Divalent NdI2 has long been a synthetic challenge but is obtainable by the reported method19,20, and was used for various Nd2+ doped MAPbI3 films (Supplementary Fig. 1) in this work. We first performed the structural and valence electronics characterization of the relevant perovskite films. Figure 1a is the X-ray diffraction (XRD) patterns of 2 mol% Nd2+-doped MAPbI3 (denoted as 2%Nd:MAPbI3) versus pure MAPbI3 films, showing the crystal structure in MAPbI3 reserved upon Nd2+ doping, thus indicating Nd2+ truly replaced Pb2+. The ratio of Nd2+ to Pb2+ is confirmed by inductively coupled plasma mass spectrometry (ICP-MS, see Supplementary Table 1). The Nd2+-to-Nd2+ distance in 2%Nd:MAPbI3 is about 5.9 nm, exceeding the distance applicable for Ruderman–Kittel–Kasuya–Yosida (RKKY) indirect exchange interaction between neighboring magnetic spins mediated by conduction electrons21,22. The slightly shifted (110) peak toward a smaller 2θ angle in the 2%Nd:MAPbI3 film is due to the larger ionic radius (1.30 Å) of Nd2+ than that of Pb2+ (1.19 Å)23,24, causing lattice expansion (see XRD refinement in Supplementary Fig. 2 and Supplementary Table 2). Further analysis by the Scherrer equation suggests that the crystallinity in both pristine MAPbI3 and 2%Nd:MAPbI3 films are nearly identical (Supplementary Table 3).Fig. 1: Characterization of pristine MAPbI3 and 2%Nd:MAPbI3 films.a XRD patterns of pristine MAPbI3 and 2%Nd:MAPbI3 thin films. The left inset is a schematic illustration of Pb2+ (green) substituted by Nd2+ (orange) in a perovskite octahedral framework, where I- is in purple and methylammonium (MA+) is in blue. The right inset is the magnified comparison for the peak at (110). b TEM images and SAED patterns of pristine MAPbI3 and 2%Nd:MAPbI3 thin films at r.t. (before cooling), 20 K and warmed back to r.t. (after cooling). c XPS comparison of Nd-3d for the Nd2+ in 2%Nd:MAPbI3 vs. Nd3+ in Nd(NO3)3 and Nd metal. Dashed lines are the fitted results. d UPS comparison of 2%Nd:MAPbI3 vs. pristine MAPbI3. e Temperature-dependent CW-EPR spectra of grinded powder of 2%Nd:MAPbI3. The dotted curves are the simulated results by Easy Spin. The intensities of all curves are offset for a better view. The inset is the proposed high spin (S = 2) electron configuration of Nd2+ (4f4) in an octahedral crystal field with I- being the weak ligand. Δ1 and Δ2 are the crystal field splitting energy between t2u and a2u energy levels, and between t1u and a2u energy levels, respectively.Film structure was further investigated by temperature-dependent selected area electron diffraction (SAED). As seen in Fig. 1b, at room temperature (r.t.), both pristine MAPbI3 and 2%Nd:MAPbI3 thin film exhibit nearly identical microscopic images and polycrystalline features in SAED pattern, belonging to the tetragonal phase except that the 2%Nd:MAPbI3 film shows a slightly increased lattice constant, in agreement with the XRD study. As temperature decreases to 20 K, the orthorhombic phase becomes predominant in both samples based on the crystal plane analysis of the SAED patterns. The (202) plane of 2%Nd:MAPbI3 shows a slight shift towards smaller reciprocal space, suggesting an increased lattice constant compared with the same plane in pure MAPbI3. Other SAED patterns at 40 K to 230 K are shown in Supplementary Fig. 3. The SAED study suggests that both pristine MAPbI3 and 2%Nd:MAPbI3 films exhibit a very similar trend in structural phase transition occurring at the same temperature range and there is no additional structural phase overserved in the 2%Nd:MAPbI3 film within the measured temperature range (r.t. to 20 K), see Supplementary Fig. 4. Therefore, XRD and SAED study clearly suggest that both pure and Nd2+ doped samples structurally resemble each other at least between r.t and 20 K and the structure is reversible for both films within at least one cooling cycle. The study of the film morphology and homogeneity of the Nd2+ doping sample is further confirmed by the scanning electron microscopy (SEM) and its associated energy dispersive X-ray (EDX) mapping of Pb, I, and Nd, showing a uniform distribution of these three elements across the film. (Supplementary Fig. 5).The oxidation state of Nd2+ is further verified by X-ray photoelectron spectroscopy (XPS) as shown in Fig. 1c. It is clear that the Nd-3d binding energy in 2%Nd:MAPbI3 (1000.3 eV for 3d3/2 and 978.9 eV for 3d5/2) is lower than that of reference Nd3+ in Nd(NO3)3 (1005.4 eV for 3d3/2 and 982.8 eV for 3d5/2)25, but higher than that of the reported Nd metal (999.1 eV for 3d3/2 and 978.0 eV for 3d5/2)26, signifying the oxidation state of Nd2+ setting between that of Nd3+ and Nd0. Moreover, due to the low electron negativity of iodine, the binding energy of Nd2+ in 2%Nd:MAPbI3 leans towards the metal Nd(0), similar to the trend when Pb2+ in MAPbI3 compared with Pb2+ in PbO27.Figure 1d is the valence band structure of pristine MAPbI3 and 2%Nd:MAPbI3 films measured by ultraviolet photoelectron spectroscopy (UPS) at r.t. Atop the VB formed mainly by 5p band of I- with minor contribution from 6 s band of Pb2+. the 4 f bands from Nd2+ dopants (the shaded area) is clearly identified, agreeing with the reported value for Nd-containing compounds28. The low binding energy cut-off of this 4 f band is approximately 2.40 eV below the VB maximum (VBM). Both samples exhibit nearly identical valence band edges in the lower binding energy region, indicating the negligible contribution from Nd2+ to the VBM at r.t. Supplementary Fig. 6 entails the analysis of relevant energy diagrams. Since the 6s5d band in metallic Nd is about 3.6 eV above the highest cut-off binding energy of the 4 f band29, the 6s5d of Nd2+ is thus estimated to be about − 4.1 eV vs vacuum. These relevant energy levels of MAPbI3 including VBM, CB maximal (CBM), and Fermi energy, and the 4 f and 6s5d orbitals of Nd2+ are illustrated in the inset of Fig. 1d. This energy diagram suggests that there is an energy crossover between the bandgap of MAPbI3 and the highest occupied (4 f) and the lowest unoccupied atomic orbitals of Nd2+, providing an energetically responsive platform for exchange interaction between the delocalized charge carriers at the band edges of the MAPbI3 host and the localized 4 f spin dopants.Figure 1e is the temperature-dependent X-band continuous-wave electron paramagnetic resonance (CW-EPR) spectra of 2%Nd:MAPbI3 powder in the dark. In general, the EPR signals are broad and do not exhibit hyperfine structures, indicating that Nd2+ is very likely a non-Kramer ion with S = 1 if in low spin configuration or S = 2 if in high spin configuration. All EPR spectra at different temperatures exhibit similar features with gx = gy = 4.26, gz = 8.93 due to the zero-field splitting (ZFS) of Nd2+. Explicitly, for Nd2+ with a 4f4 electron configuration (S = 2, L = 6, J = 4), electron dipole interaction can split the multi-fold degenerated ground state of the electron spin system in the absence of an external magnetic field. The corresponding spin Hamiltonian under ZFS is: \({{{{\mathcal{H}}}}}_{{{{\rm{ZFS}}}}}=\)D[Sz2 – S(S + 1)/3] + E(Sx2 + Sy2), where D and E are the parameters of ZFS, D describes the axial component of the magnetic dipole-dipole interaction, and E is the transversal component. At 4 K, the ZFS parameters for the simulation (the dotted line is the simulated spectra by MATLAB using the software package EasySpin) are D = 1.334 cm−1, E = 0.045 cm−1. The weak signal at g = 8.93 is associated with the transition between ms = ± 2, while the resonance at g = 4.26 is due to the transition from ms = 0 to ms = 130. The high-field signal appearing at g = 2.003 is very likely due to the lead vacancies with unpaired electron spin (note that the EPR samples are grinded powders in order to fit into the thin EPR test tube), very similar to that of the Ti4+ vacancy in BaTiO3 perovskite31. This is further confirmed by the CW-EPR test of pristine MAPbI3 powder (Supplementary Fig. 7). The EPR result suggests that there are four spins in the t2u and a2u sublevels, both of which are half-filled, hence, the Jahn-Teller effect can also be excluded. The EPR study, therefore, manifests the high-spin and low-field configuration of the 4 f electrons in Nd2+ as depicted in the inset of Fig. 1e. This configuration is also supported by the fact that crystal field splitting is weak for lanthanides due to the high angular momentum of 4 f orbital and I- being one of the weakest ligands. In contrast, orbital angular momentum can often be quenched by ligand fields in 3d metals. The temperature-dependent CW-EPR under 405 nm laser irradiation (power = 100 mW) was also conducted to investigate the interaction between the light-induced carrier and the localized spin in 2%Nd:MAPbI3. The comparison between relative EPR signal change between dark and light conditions (Supplementary Fig. 8) exhibits a clear phase change at 100 K, below which more relative percentile loss in EPR signal from dark to light conditions than above it, suggesting that there is another origin accountable for the EPR signal loss besides light-induced heat, most likely due to the AFM coupling between light-induced carries and localized impurity 4 f spins.Carrier recombination was investigated by both ss-PL and tr-PL. Figure 2a shows the temperature-dependent PL intensity ratios of 2%Nd:doped MAPbI3 to pristine MAPbI3, and the shift of PL peak wavelength referenced to that at r.t., using a 400 nm pulsed laser, which is deliberately selected to avoid exciting any PL in Nd2+ 32. (Pulse duration = 35 fs, repetition rate = 2000 pulses/s, fluence = 0.88 µJ/cm2, equivalent to intensity = 1760 µW/cm2, or photon flux of 2×1012 photons/cm2/pulse or 3.6 × 1015 photons/cm2/s). Supplementary Fig. 9 includes all full ss-PL spectra at all measured temperatures. A phase transition from tetragonal (T) to orthorhombic (O) for both pristine MAPbI3 and 2%Nd:MAPbI3 occurs between 100 K and 160 K as indicated by SAED (Supplementary Fig. 3 and 4). Phase transition of pristine MAPbI3 is associated with an emerging high-energy PL emission “shoulder” in agreement with literature33,34,35. In contrast, 2%Nd:MAPbI3 shows a blue-shifted PL peak without the “shoulder” feature (Supplementary Fig. 9). It is notable that after phase transition, the PL peak area of 2%Nd:MAPbI3 rises up from ~ 0.41 × 106 counts·nm to 1.78 × 106 counts·nm as temperature decreases from 150 K to 5.7 K, in strong contrast to the change in PL peak area of pristine MAPbI3 from 0.37 × 106 counts nm to 0.67 × 106 counts nm in the same temperature range. Such enhanced PL intensities strongly suggest a potential application of Nd2+ doping in enhancing the PL intensities of bulk perovskite films, in contrast with the conventionally adopted quantum-dot perovskite films for their large emission efficiencies.Fig. 2: Photoluminescence study of pristine MAPbI3 and 2%Nd:MAPbI3 films.a The temperature-dependent ss-PL peak wavelength, and PL intensity ratio of 2%Nd:MAPbI3 to pristine MAPbI3. b Temperature-dependent lifetime < τ > of pristine MAPbI3 vs. 2%Nd:MAPbI3 extracted from temperature-dependent time-resolved PL decays study. c Summary of lifetime < τ > at different temperatures and Nd2+-to-photon density ratios. All lines are as a guide to the eye. d Temperature-dependent lifetime < τ > of pristine MAPbI3 vs. 2%Nd:MAPbI3 under magnetic field (normal to sample surface with magnetic field strength of 1500 Gauss near sample surface) extracted from temperature-dependent time-resolved PL decays study. e Recombination of photocarriers in pristine MAPbI3. f In the absence of a magnetic field, the spin exchange interactions between the isotropic localized spins of Nd2+ cations with the electron and hole in the exciton are allowed to occur respectively, as these interactions do not break the total spin = 0 of the exciton. g In the presence of a magnetic field, the localized spins on Nd2+ cations are anisotropically polarized, preventing the coupling between Nd2+ spin with either electron or hole in the exciton as otherwise the net spin of exciton becomes non-zero.The enhanced PL emission is a clear indication of less non-radiative recombination via carrier-phonon interaction in 2%Nd:MAPbI3 than that in pristine MAPbI3. Within the O phase, pristine MAPbI3 shows distinctive high-energy emission peaks due to donor-acceptor-pair transition, as regulated by the trap states within the material36. 2%Nd:MAPbI3 in the O phase, however, still exhibits rapid redshift with decreasing temperature, in stark difference from pristine MAPbI3. Figure 2b is the comparison of temperature-dependent average PL lifetime < τ > (defined in Eq. 1) for 2%Nd:MAPbI3 and pristine MAPbI3 excited under the same conditions. The 2%Nd:MAPbI3 displays a slightly shortened lifetime compared to pristine MAPbI3 at r.t., indicating the thermal energy (kT) still outpaced the quantum spin-spin exchange interaction between the 4 f spin in Nd2+ and the photocarriers in MAPbI3. As temperature drops, 2%Nd:MAPbI3 shows a nearly monotonic increment of its < τ > that exceeds 1 μs at 5.7 K. The original tr-PL and fitting details are collected in Supplementary Fig. 10–14. This observation is strikingly discernable from the continually reduced < τ > on pristine MAPbI3 upon lowered temperature that agrees well with an earlier report33, despite a plateau between 150–200 K as related to the structural T to O phase transition37, with < τ > eventually reaching < 10 ns at 5.7 K. The temperature-dependent trends of < τ > for pristine MAPbI3 and 2%Nd:MAPbI3 films were re-confirmed by tr-PL decays obtained by using a diode laser as an excitation source (Supplementary Fig. 15). Although similar PL decay behaviors caused by magnetic polaron formation can be observed in semiconductors doped with magnetic ions38, they are still different from our finding as the former generates very short spin relaxation dynamics at the scale of picoseconds, and previous report has indicated a suppressed magnetic polaron in bulk semiconductor systems. On the other hand, as our material paradigm requires optical excitation to generate photocarriers (excitons) to imitate the conduction electrons as occurring in classic metal-based Kondo systems, and arise from the alignment of the localized dopants spins with the exciton spins, they also differ from the intrinsic magneto-optical and magneto-electric properties of diluted magnetic semiconductors. It is noteworthy that other factors could lead to elongated PL carrier lifetimes, such as Coulomb interactions, initial distribution of photocarriers in electronic bands, less defect-assisted relaxation, van der Waals structures, and transition among different quasiparticles39. But we can mostly exclude these factors because 1) perovskite materials applied in this study are three-dimensional structures with and without atomically dispersed Nd2+ dopant, and therefore do not introduce multilayered heterostructures, external dielectric environment, interlayer strain, or charge carriers; 2) identical photoexcitation and acquisition conditions are utilized, plus the Nd energy levels that mostly affect perovskite intragap and VB regions, both pristine and Nd-doped perovskite films exhibit comparable electronic band structures; 3) both samples show comparable ss-PL intensities and tr-PL decay dynamics initially before temperature reduction, and temperature-dependent TEM/SAED images and the derived diffraction patterns (Supplementary Fig. 4) indicate comparable crystal structures, thus signifying comparable densities of structural defects; 4) at cryogenic temperatures, the form of photocarriers remains excitons in the absence of electrical bias or multilayer heterostructures. On the other hand, to comprehend the relaxation of phonon-assisted photocarriers, we examined the temperature-dependent broadening of emissions and extracted the full width at half-maximum (FWHM) from the photoluminescence spectra. Through the analysis of the temperature-dependent emission broadening and extracting the full width at half-maximum (FWHM) of the PL spectra (Supplementary Fig. 16a), one can see that Nd-doped MAPbI3 film indeed exhibits a smaller change of the temperature-dependent PL linewidth compare with the pristine MAPbI3 film. For hybrid lead halide perovskite materials, charge-carrier-phonon interaction is dominated by Frohlich interaction between charge carriers and LO phonons34, the equation is expressed as:$${\varGamma }_{{LO}}={\gamma }_{{LO}} \cdot \frac{1}{\left[{e}^{{E}_{{LO}}/{k}_{B}T}-1\right]}$$
(1)
Here in Eq. 1, \({\varGamma }_{{LO}}\) is results from LO phonon scattering, \({\gamma }_{{LO}}\) is the corresponding charge-carrier-phonon coupling strength, and \({E}_{{LO}}\) is the Energy representative of the frequency for the weakly dispersive longitudinal optical (LO) phonon branch. The parameters, \({\gamma }_{{LO}}\) and \({E}_{{LO}}\) can be derived by fitting the temperature-dependent half-maximum (FWHM) of the PL spectra. Therefore, the longitudinal optical phonon energy (ΓLO) of Nd-doped MAPbI3 as compared to pristine MAPbI3 at different temperatures (Supplementary Fig. 16b) can be exported, thereby substantiating the suppressed carrier-phonon coupling mentioned in the previous context.The spin-photocarrier interaction can be further modulated by tuning the relative ratio of [Nd2+] per area (denoted as [Nd2+] area, which is 2.2×1015 Nd2+/cm2 for a 2%Nd:MAPbI3 film thickness of ~ 300 nm) to that of incident photons. Thus, a low [Nd2+] doped 20 ppm-Nd:MAPbI3 (i.e., molar ratio of Nd:Pb = 20:106) was prepared, equivalent to [Nd2+]area = 2.2×1012 Nd2+/cm2. A lower excitation intensity of fluence = 0.11 µJ/cm2, equivalent to 220 μW/cm2 (400 nm, [photon] = 2.2×1011 photons/cm2/pulse) was also adopted. The combination of two different excitation light intensities and two [Nd2+]area allow us to modulate the ratio of [Nd2+]area/[photon]. Figure 2c shows the temperature impact on the <τ> across the [Nd2+]area/[photon] range of 10−3 ~ 104. The monotonic increment of <τ> at greater [Nd2+]area/[photon] ratio below 100 K is clearly evident. At transition between 140 K − 100 K, particularly at a higher [Nd2+]area/[photon] ratio, can be clearly identified. It is known that in the O phase at a temperature below 120 K, excitons are the dominant light-induced carriers in MAPbI318. We further studied the PL properties under a fixed permanent magnetic field. First, we methodically studied the magnetization property of 2%Nd:MAPbI3. The temperature-dependent hysteresis curve of 2%Nd:MAPbI3 and pristine MAPbI3 are collected in Supplementary Fig. 17a. It indicates that there is a positive magnetic moment observed, which can be attributed to the alignment of spin moments of Nd ions within the sample. With the decrease in temperature, the magnetization curves displayed nearly S-shaped behaviors, indicating stronger magnetic interactions. In contrast, the pristine MAPbI3 shows diamagnetic behaviors with a small negative response to the external magnetic field. Furthermore, we also studied the magnetic susceptibility as a function of temperature for the 2%Nd:MAPbI3 powder, shown in Supplementary Fig. 17b. By using the Curie-Weiss (CW) law, the effective magnetic moment can be derived as μeff = 4.76 μB. This result is similar to the theoretical effective magnetic moment of Nd2+ which is μeff = 4.9046 μB. The result also indicates the bivalent state of Nd ion in the sample which corresponds to the XPS result. Figure 2d shows that the long PL lifetime in 2%Nd:MAPbI3 sample at a low-temperature range (< 120 K) intriguingly vanishes in a weak magnetic field (0.15 T) normal to sample surface provided by a permanent SmCo magnet (suitable for low-temperature)40 positioned behind the samples. See the Materials and Methods section for detailed descriptions of the magnetic field effect PL measurement procedures. The temperature-dependent PL lifetime of the 2%Nd:MAPbI3 sample resembles that of the pristine MAPbI3 sample. Both show little difference from that of the pristine MAPbI3 sample in Fig. 2b. To verify the magnetic field effect on the PL spectra of the Nd2+ doped sample, we repeated the measurement on the same 2%Nd:MAPbI3 sample. Comparison of the ss-PL spectra of this 2%Nd:MAPbI3 sample with and without magnetic field illustrates the markedly attenuated PL intensities when the magnetic field is present (Supplementary Fig. 18). Supplementary Fig. 19 verified the long PL lifetime in this 2%Nd:MAPbI3 sample in absence of magnetic field at low-temperature range (< 120 K), similar to the 2%Nd:MAPbI3 sample (no magnetic field) in Fig. 2b. However, under magnetic field, the long PL lifetime of this 2%Nd:MAPbI3 sample in the low-temperature range (< 120 K) vanished, and becomes similar to that of the pristine MAPbI3 sample (without magnetic field) as shown in Fig. 2b. On the other hand, we can rule out the possibility that Nd doping leads to lattice disorder and the potential increments in PL intensity and carrier lifetime, despite that structural distortion can in fact, enhance light emission efficiencies in a certain inorganic context such as InGaN multilayer structure41,42. The magnetic field should not be able to control the regarded lattice disorder and annihilate/generate structural defects that affect the ss-PL intensity and carrier lifetime as observed. Therefore, the exciton-spin interaction, as valved by the magnetic field and Nd impurity spins should be the responsible mechanism for the abovementioned light-induced observations at cryogenic conditions. As such, the demonstrated elongated PL lifetime as controllable upon Nd2+ doping concentration and magnetic field clearly indicates a long but manipulable spin relaxation process that are potentially useable in high-performance spintronics and quantum computing applications, where achieving long coherence time of electron spins is critical for quantum manipulation43. The original tr-PL and fitting details are collected in Supplementary Fig. 20 and 21. Figure 2e illustrates the interplay between the recombination of an exciton and its exchange interaction with localized 4 f spin, resulting in a metastable pinned exciton and the consequently retarded recombination kinetics. Note that Jf−psd is the partial intra-atomic exchange constant, while Jf−sd is the inter-atomic exchange constant. It is reported that the radii of an exciton in MAPbI3 are in the range of 3 ~ 5 nm44, which is large enough to span two neighboring Nd2+ dopants (5.9 nm apart aforementioned). Thus, we think it is likely that each exciton can interact with at least two Nd2+ in proximity. The schematic, as shown in Fig. 2f, is allowed because of the isotropic magnetic spin moments of localized 4 f spins that are randomized in the proximity of an exciton so that the photoelectrons and photoholes in the exciton have a high probability of forming antiferromagnetic exchange interaction with nearby 4 f spins. In contrast, as illustrated in Fig. 2g, when the localized 4 f spin magnetic moments are polarized by an external magnetic field, antiferromagnetic exchange interaction between the exciton and its surrounding 4 f spins becomes anisotropic with diminished probability or is replaced by weaker ferromagnetic exchange interaction3. Thus, more than the relative ratio of [Nd2+]area/[photon], a magnetic field also acts as a switch to modulate this light-induced Kondo-like coupling.Next, we investigated the local electronic properties using low-temperature scanning tunneling spectroscopy at 5 K substrate temperatures in an ultrahigh vacuum condition. For the experiments, a photon beam of 400 nm wavelength was illuminated onto the tip-sample junction. The light illumination onto the sample is required for the tip to approach the sample because the recorded tunneling current under the dark condition is significantly lower than that under illumination as shown in supporting information Supplementary Fig. 22. Scanning tunneling microscopy (STM) images of the sample (exemplified in Fig. 3a) do not show large height variations however, detailed surface features are difficult to resolve probably due to the light illumination. Therefore, the point spectroscopic measurements are performed across the surface at different locations without taking local area images. For the measurements, the STM tip is positioned at a fixed height, the bias voltage is ramped from − 1 V to + 1 V under 400 nm illumination, and the corresponding tunneling current is recorded using a lock-in amplifier. The semiconducting characteristic of the undoped sample areas is clearly observed in the tunneling spectroscopy (I–V curve) data (Fig. 3b) and simultaneously measured differential tunneling conductance (dI/dV-V) spectra (Fig. 3c). Here, the CB and VB gap is measured as ~ 1.5 eV, which agrees well with the expected gap of 1.59 eV. For Nd-doped areas of the 2%Nd:MAPbI3 sample, the dI/dV-V spectroscopy data show a smaller gap-like feature (Fig. 3d, e). Its symmetric nature and the energetic location, ± 0.28 eV, indicate that it is of a different origin. The d2I/dV2 data (Fig. 3f) clearly indicates that the step-like features observed in Fig. 3e are indeed related to the inelastic electron tunneling (IET) process. More complete tunneling spectroscopic data collected at random sample areas reveal both the presence and absence of step-like features in dI/dV and d2I/dV2 curves of Nd-doped MAPbI3 film (Supplementary Figs. 23 and 24). Supplementary Figs. 25 and 26 exhibit another local domain with a more uniform distribution of Nd2+ as evidenced by the periodical occurrence of the step-like features. In stark contrast, this step-like feature was never observed from pristine MAPbI3 film that only exhibits the typical bandgap characteristics of MAPbI3 (Supplementary Figs. 27 and 28), suggesting that the IET process only occurs in Nd-doped MAPbI3 film.Fig. 3: Local electronic structural investigations by cryogenic scanning tunneling microscopy (STM).a STM image of the sample surface acquired under 400 nm illumination (Vt = 1 V, It = 50 pA). b I–V spectroscopy, and (c) simultaneously acquired dI/dV-V spectroscopy reveal a semiconducting behavior. d I–V spectroscopy, e dI/dV-V spectroscopy, and (f) d2I/dV2-V plots associated with Nd2+ sites. The arrows indicate the change in energy at ± 0.28 V. g A demonstration of STM spectroscopy process under 400 nm illumination at the tip-sample junction. The schematic drawing at right illustrates exciton (electron-hole pair) formation upon illumination, and a trapped electron at the Nd2+ site (indicated with a dashed circle). The spectroscopy data shown in (b–f) are generated from 64 separate measurements at different locations across the sample. In each figure, the average curve appears as a single darker-color plot.It is reasonable to suggest that at the low substrate temperature of 5 K, the photoexcited electrons are being trapped or filled (as evidenced in the optic study as well) in the 6s05d0 empty orbital of the Nd2+ in agreement with the energy gap (~ 0.3 eV) between 6s05d0 orbital of the Nd2+ and the conduction band minimum of MAPbI3 as inferred by UPS results (Figs. 1d & 3g). In general, IET measures the excitation energies of atoms or molecules within the tunnel junction45,46,47. In our case, the 0.28 eV energy is attributed to the excitation of the trapped electrons in this gap (Fig. 3g).In closing, our work has demonstrated an optical Kondo-like effect in which the density of light-induced delocalized electrons is much outnumbered by the localized spins from magnetic impurity. As such, the spin-entangled electrons and holes have a high probability to respectively couple with the opposite localized impurity spin within their proximity as evidenced by the notably prolonged carrier lifetimes. Interestingly, an external magnetic field nullifies such interaction between the exciton and the localized impurity spin because of the vanishment of opposite local spins needed for the electron and hole, respectively. The formation of the exchange interaction between the delocalized electrons in MAPbI3 and localized spins in Nd2+ is antiferromagnetic by nature, a consequence of partial photoelectron trapping in the 6s5d orbitals that result in 10 folds extended carrier lifetimes of perovskite films at low temperatures. Importantly, because the concentration of magnetic spins in Nd2+ overwhelmingly exceeds the photocarriers in the perovskite host, a case impossible to be achieved in a metal-based classic Kondo process, we are able to control the coupling intensity (embodied by carrier lifetime of perovskite MAPbI3) through the amount ratio of Nd2+ to the incident photons as well as by the external magnetic field. In perspective, our work shows an approach to apply quantum interference on a pair of spin-entangled particles by localized spins within close proximity and inspires the discovery of strongly correlated light-matter interaction with evolutionary states at the localized-itinerating electron crossover regions, where mutual coupling degrees of freedom among charge, spin, orbital and lattice can lead to exotic photon-induced electronic phases and applications in spintronics and many-body entanglement-based quantum computing.
Light-induced Kondo-like exciton-spin interaction in neodymium(II) doped hybrid perovskite
