Characterization of the photoelectric effect in halide perovskites

Halide perovskite thin films (CsPbX₃, X = Br or I) were made by mixing precursor powders (PbX₂ and CsX) with the desired molar ratio in dimethyl sulfoxide or N, N-dimethylformamide and spin-coating on doped Si wafers. The sample preparation was made in a standard glovebox under Ar atmosphere to limit air contamination of the film’s surface. The crystal structure of CsPbX₃ along with the optical image of the films are presented in Fig. 1a. We characterized the absorbance, photoluminescence, and structure of both perovskite films in Fig. 1b, c (notice that CsPbI₃ and CsPbBr₃ respectively adopt the cubic and orthorhombic phases at room temperature) before loading them into an ultra-high vacuum chamber for photoelectric effect characterizations. The base pressure of the chamber was about 1 × 10⁻⁹ torr with contaminant gas partial pressure below 10⁻¹¹ torr, and the films were annealed at 350 °C in the chamber to clean their surface from any organic or gas contaminant. We performed Auger electron spectroscopy (AES) on CsPbBr₃ before and after annealing, as shown in Supplementary Fig. 7. The magnitudes of characteristic Auger peaks from Cs (49 eV), Br (55 eV), and Pb (97 eV) remained unchanged after UHV annealing. Moreover, we observe significant attenuation of the Auger signals associated with carbon (275 eV) and oxygen (510 eV). These results suggest that the chemistry of the perovskite thin films is not affected by UHV annealing while the latter process effectively removes oxygen adsorbed onto thin film surfaces, as well as amorphous carbon and solvent residue. The photoelectric effect was observed and studied by illuminating the sample with a monochromatic light beam and measuring the resulting photocurrent between the thin film sample and a counter anode while applying a potential difference between the two (Fig. 1d). The QE energy spectrum of the photoelectric effect was derived from the measurement of the photocurrent response as a function of the excitation light energy over the ultraviolet and visible range. The measurements were performed with and without the deposition in-situ of an ultra-thin layer (ideally one to a few monolayers) of Cs on the surface of the perovskite films.

Figure 2a illustrates the QE spectrum of the photoelectric effect for the CsPbI₃ and CsPbBr₃ thin films measured before (dashed curves) and after Cs deposition (solid curves). We observe a dramatic three to four orders of magnitude increase in the QE for all photon energies after the Cs deposition. Concomitant with the increase in the QE, we observe a lowering of the onset energy of the electron emission (i.e. the minimum photon energy necessary to observe the photoelectric effect) from 4.0 eV to about the bandgap energy of CsPbBr₃ (2.1 eV) and CsPbI₃ (1.8 eV). The maximum QE values that were measured in the CsPbBr₃ and CsPbI₃ thin films are 2.2% at 5 eV and 0.14% at 3.5 eV, respectively. In the CsPbBr₃ films, the QE value remains greater than 1% for UV light with energy above 3.9 eV, and was measured to be larger than 0.25% at 3.1 eV (400 nm). These values are one order of magnitude smaller than the QE reported in the state-of-the-art III-V semiconductor electron sources (e.g. GaN after Cs activation yields QE larger than 10% at 4.77 eV)^18,21,22, which is rather impressive given the fact that III-V semiconductor electron sources involve complex and costly fabrication processes and surface pre-treatments in order to create high-quality materials with atomically perfect surfaces. In the case of halide perovskite thin films, the fabrication is simple and low cost, with the only demanding step being the Cs deposition. We also note that the QE performances of the halide perovskite films are more than two orders of magnitude better than metallic electron sources in the UV range and at a comparable vacuum level¹⁹. The reproducibility of the high QE observed in CsPbBr₃ thin films was verified by testing 17 films yielding an average QE larger than 1.5% with two champion films exhibiting QE larger than 2% at 5 eV as illustrated in Fig. 2c.

Next, we tested whether the Cs surface activation process was applicable to the broad family of organic–inorganic (hybrid) halide perovskites. The latter materials contain extremely reactive organic salts such as methylammonium (MA), butylammonium (BA), or Formadinium (FA), which replace the Cs atoms in the perovskite lattice (Fig. 1a). Three distinct hybrid perovskites that are used for fabricating state-of-the-art perovskite-based photovoltaic cells were tested: the 3D perovskite FA_0.7MA_0.25Cs_0.05PbI₃²⁸, and the Ruddlesden-Popper layered hybrid perovskites BA₂MA_n-1Pb_nI_3n+1 with n = 2 or 5²⁹. Here, we note that the hybrid perovskites were not annealed in an ultra-high vacuum because annealing these materials above 100 °C is known to sublimate the organic cation and degrade the perovskites³⁰. Despite these constraints, all the hybrid perovskite thin films demonstrated relatively efficient photoelectric effect with a low energy onset after Cs activation (Fig. 2b and Supplementary Fig. 1). The emission onset for BA₂MA₁Pb₂I₇ is close to its intrinsic bandgap, i.e. 2.35 eV (photoelectric onset) versus 2.18 eV (optical bandgap), where the difference between these two values could be accounted for by the binding energy of electron-hole pairs (excitons)³¹. However, the onset of electron emission was blue shifted in the other two hybrid perovskite thin films as compared to their bandgap energy. We speculate that this discrepancy is partly explained by our lack of control of the Cs activation process and the presence of absorbed gases on the film’s surface (no annealing was done for hybrid perovskites), both of which can significantly affect the electronic structure on the surface of the perovskite films. The QE of the hybrid perovskites is of the order of 0.01% for photon energy between 3 and 3.5 eV, which is two-orders of magnitude higher than metallic electron sources in this energy range¹⁸. These results attest to the robustness and potential of hybrid halide perovskites as electron sources despite the presence of reactive organic cations.

Understanding the physical origin of the photoelectric effect after Cesium activation

To further understand what limits the QE and what defines the onset photon energy of the electron emission in the halide perovskite thin films, we investigated the mechanism of the photoelectric effect, which consists of three basic steps (Fig. 3a): (i) absorption of photons in the bulk of the halide perovskite films and generation of photo-excited electrons in the conduction band, (ii) transport of the photo-excited electrons to the films surface, and (iii) escape of these electrons to the vacuum by surface tunneling. We tested if the step (i) was limiting the QE of the perovskite thin films by calculating the internal quantum efficiency defined as the QE divided by the absorbance (Fig. 1c and Fig. 2a). It yielded nearly an identical response as the QE in all types of perovskites (Fig. 3b and Supplementary Fig. 2), thus implying the process of photoexcitation of electron carriers in the perovskite thin films is not the main limiting factor of the QE. The efficiency of the step (ii) is largely determined by the transport properties of the perovskite films. After photoexcitation the charge carriers in the films are subject to an electric field of about 3.6 V/cm and the efficiency of charge transport to the halide perovskite surface is quantified by the mobility-lifetime product (µτ). In halide perovskites, µτ is typically between 10⁻² and 10⁻⁴ cm²/V depending on the material composition, which is on par or better than III-V semiconductors^8,28,32,33. Moreover, transport of charge carriers in the bulk of 3D perovskite thin films is very efficient across a distance of 500 nm^10,11. Therefore, we infer that in 3D halide perovskites the QE of the photoelectric effect is mainly limited by the step (iii), which is the efficiency of ejecting the photogenerated electrons out of the surface. The efficiency of step (iii) is limited by the recombination of electrons at surface defects and surface energy barriers for electrons³⁴. The probability of surface recombination depends on the surface recombination velocity S, which needs to be minimized in order to achieve high QE—for example smaller than 10⁴ cm/s in GaAs electron sources³⁵. Previous reports have claimed S values of the order of 10³−10⁴ cm/s in CsPbBr₃ single crystal³⁶, and values as low as 4 cm/s in MAPbBr₃ single crystals under specific passivation conditions³⁷. However, polycrystalline films might have significantly higher S due to the presence of grain boundaries and a higher density of surface defect states as compared to single crystals in the absence of any surface treatment. Indeed, in the FA_0.7MA_0.25Cs_0.05PbI₃ thin films, we recently demonstrated that charge collection in solar cells is limited by interface defects, which factors into the lower QE observed in these films²⁸. In addition to these effects, in 2D Ruddlesden-Popper layered perovskites less effective transport of the photo-excited electrons to the film surface (µτ ~10⁻⁶–10⁻⁵ cm²/V) can also contribute to the observed lower QE, as explained in our recent work on 2D perovskite photovoltaic devices demonstrating field-dependent charge collection possibly limiting electron-hole separation³⁸.

a Mechanism of the photoelectric effect showing the three main steps: (i) electron photogeneration, (ii) electron transport, and (iii) electron emission. b Internal quantum efficiency of the photoelectric effect in the CsPbBr3 thin film after Cs activation. The gray dashed curve shows the corresponding (external) quantum efficiency, copied from Fig. 2a. c Schematics of the band diagram at the CsPbBr3 thin film surface before and after Cs activation. Egap, VB, CB, χ stand for the bandgap, the valence band, the conduction band, and electron affinity. Evac indicates the vacuum level where the electron can be emitted from the surface after photoexcitation (hν). d Spatial map of the quantum efficiency relative to its maximum value for the CsPbBr3 thin film after Cs activation. Scale bar is 10 mm. Dashed area mark the position of the in-situ Cs source. The quantum efficiency exhibits peak values near the center of the sample in a 10 mm² area, while other regions of the samples yield slightly smaller but uniform efficiency within 10% margin variations over a 1 cm² (indicated by dashed-green-square region).

To verify that in 3D perovskites the QE is limited by the ability of electrons to escape from the perovskite surface and into the vacuum, we elucidate the impact of the Cs deposition on the emission of free electrons. The minimum photon energy for electron emission in vacuum corresponds to the work function hν = W, where h is the Plank constant, ν is photon frequency, and W the work function. The work functions values reported in the present paper for all pristine perovskites (around 4.0 eV), including CsPbBr₃ (vide infra) are consistent with the work function of CsPbBr₃ thin films reported in a recent study at 4.1 eV³⁹. We may notice that these values differ from the bulk ionization energies IE = E_gap + χ, where E_gap is the bandgap and χ is the electron affinity, defined as the difference between vacuum level and conduction band minimum. Indeed, the ionization energy was measured to be around 5.8 eV in CsPbBr₃ thin films³⁹. Consistently, the pristine CsPbBr₃ thin films in the present work yield an emission onset of 4.0 eV, far above the expected bulk ionization energy. Another surprising observation is that the photoelectron onset at 4.0 eV is also common to CsPbI₃ and the three hybrid perovskites that were tested, before Cs activation (Fig. 2a and Supplementary Fig. 1). This result is consistent with recent reports claiming negligible dependence of the work function of perovskite thin films as a function of their compositions due to pinning of the Fermi energy, which has been explained by the similar nature of surface states in all perovskite films³⁹. The good correspondence between the CsPbBr₃ work function and the electron emission onset suggests that the activation of deep surface traps under vacuum conditions is related to the metal cation Pb at the film surface³⁷. We explain the orders of magnitude increase of the QE upon Cs surface deposition by a conditioning of the thin film surfaces resulting in a change of sign of the electron affinity (Fig. 3c)³⁴. Upon deposition of Cs on the perovskite surface, the Cs lose an electron, resulting in the formation of a surface dipole layer that lowers the electron affinity to a negative value^14,18,23,40. Correspondingly, first-principles calculations explain the increase of QE upon the addition of a Cs layer onto perovskite crystals by a decrease of the work function, which may result from changes in the Fermi energy and density of states at the conduction band minimum (see section ST1 of the supplementary materials and Supplementary Fig. 3). Even though requirements for surface conditioning are relaxed in halide perovskites in comparison with III-V semiconductors^18,24, both the inhomogeneous spatial distribution of QE (Fig. 3d) and the significantly different QE values observed in the three types of 3D perovskite films underscore the limited understanding and control of the Cs activation process in halide perovskites and is the main limiting factor for achieving higher QE.

To better understand the relationship between Cs coverage on the halide perovskite surface and QE, we performed in-situ AES measurements. The Cs coverage is quantitatively determined by calculating the ratio between Cs (49 eV) and Br (55 eV) Auger peaks (see section ST2) and plotted as a function of QE at 405 nm in Supplementary Fig. 8b. The QE continuously increased as Cs coverage increased from 0 till about 2.5 Cs/unit cell where QE reached its maximum value. This is in good agreement with our theoretical calculations, which show that two layers of Cs coating (2 Cs/unit cell) has a lower work function than one layer of Cs coating. Furthermore, deposition of more than two layers of Cs decreases the QE, similar to what has been observed on GaAs photocathodes⁴¹.

Stability of the photoelectric effect during halide perovskite electron source operation

Next, we investigated the stability of the photoelectric effect in the CsPbBr₃ thin films. Figure 4a (central panel) illustrates the evolution of the QE over time as a film is continuously illuminated with light at 3.06 eV. We observed that the QE degrades to 60% of its original value after 25 h and down to 8% after 96 h. The main degradation mechanism is attributed to the surface contamination due to the presence of oxygen-based elements inside the vacuum chamber. We quantify the stability of the perovskite films by calculating the exposure in units of Langmuir (L) equivalent to the dose of a given element for the QE to decrease by 63%. We estimated the minimum values of 2.86 L, 6.75 L, and 6.28 L for H₂O, O₂, and CO₂, respectively. Under similar conditions of vacuum pressure as used here (10⁻⁹ torr), state-of-the-art semiconductor electron sources yield lower exposures with rare exceptions such as GaN^22,25,42. For example, about 0.05 L for H₂O was found in a CsK₂Sb electron source, and exposures smaller than 0.08 L were reported in a GaAs electron source with an initial QE of 12%^18,25.

Finally, we observed that after degradation the QE of the perovskite thin films could be regenerated back to its original value (i.e. before degradation) via in-situ Cs deposition in about 1 h. Figure 4a illustrates a complete cycle starting from an initial Cs activation (left panel where the QE is continuously monitored while the films are exposed to the in-situ Cs source), followed by a decrease in the QE due to degradation during operation (central panel), and then a second Cs activation process that regenerates the perovskite thin films (right panel). The corresponding QE spectrum of the photoelectric effect taken at different time of the degradation–regeneration cycles confirms that the thin film is almost perfectly regenerated to its original QE value after the second Cs activation (Fig. 4b). It was demonstrated that the main degradation mechanism in semiconducting GaAs electron sources results from the alteration of the Cs activation layer on the surface. Indeed, exposition of the Cs surface to oxygen-based residual gas in the vacuum chamber leads to the formation of Cs–O and Cs–OH compounds, thus compromising the beneficial effect of Cs on the performances of the electron sources^26,43. Our observation indicates no appreciable degradation of the halide perovskite films as, after degradation, we can almost fully restore the QE of our electron sources through a second Cs activation (Fig. 4a). Therefore, it is reasonable to assume that the main degradation in our electron sources takes place at the Cs surface, most likely through a mechanism similar to the one described in GaAs electron sources. The in-situ AES measurements show the Cs coverage decreased from 2.5 Cs/unit cell (maximum QE) to 2.1 Cs/unit cell after degradation (Supplementary Fig. 8b) while the relative oxygen Auger peak increased by ~30% (Supplementary Fig. 9). After adding fresh Cs on the degraded photocathode, the Cs coverage increased to 2.7 Cs/unit cell and the oxygen peak decreased to a magnitude similar to that of the originally activated surface. Moreover, we observe in Fig. 4a (left and right panels) that the Cs-activation process after degradation is about four times faster than the initial Cs-activation (from the pristine films), which is another indication that the Cs surface is only partially compromised by the oxygen-based residual gas and needs to be restored with Cs atoms.

In summary, we have demonstrated perovskite-based electron sources with a few percent quantum efficiencies, with spectral tunability other than the visible spectral range, and which can be operated for tens of hours and integrally regenerated in situ after degradation. The figures of merit reported for the photoelectric effect in halide perovskite thin films offer a tremendous opportunity to develop a disruptive electron source technology that addresses the demands for low-cost fabrication and operation, high efficiency, and spectral tunability. Furthermore, future optimization of the quality of the perovskite thin films and the Cs deposition process should lead to electron sources preforming on par with those obtained using III-V semiconductors but with orders of magnitude lower cost in terms of manufacturing and operation.