CH₃NH₃PbBr₃ is not pyroelectric, excluding ferroelectric-enhanced photovoltaic performance

Halide perovskite (HaP)-based solar cells, in particular, methyl ammonium lead iodide and bromide (MAPbI₃ and MAPbBr₃, respectively) and also the inorganic analog, CsPbBr₃, show remarkably good performance in solar cells.^1,2 They show high absorption coefficient (∼10⁵ cm⁻¹),^3,4 low exciton binding energy,⁵ and low non-radiative charge-carrier recombination, all reflected in the sharp absorption onset and small Urbach energy.³ These, together with good carrier lifetimes (∼0.1-1 μs)⁶ and reasonable mobilities (≤ ∼ 100 cm² V⁻¹ s⁻¹)⁷ for thin-absorber-architecture solar cells, make the HaP-based cells so efficient and attractive. Those properties are also reflected in their high voltage efficiency (high V_OC/E_G, where V_OC is the open-circuit voltage and E_G is the optical bandgap),⁸ which can be directly related to the low trap density and recombination rates and steep absorption onset. All of the best performing solar cells, based on the above-mentioned HaPs, show this voltage efficiency (see supplementary material, Figure S1),⁹ which entails lower voltage losses (in addition to those dictated by the Shockley-Queisser model) than for many photovoltaic materials.⁸ 

Several arguments have been forwarded to explain the remarkable photovoltaic efficiencies, including the high V_OC/E_G ratio. A high V_OC implies low charge recombination and, indeed, photovoltaic cells with very low recombination can be made with halide perovskites.¹⁰ Ferroelectricity, which is well known for oxide perovskites,^11,12 has been suggested^13,14 as a possible explanation for the low charge recombination rates (increasing charge carrier lifetimes) and because it could bring about efficient charge separation (reducing the need for high charge carrier mobilities¹⁵). In addition, it was argued that the (undesired) hysteresis in the current-voltage (I-V) characteristics of many HaP cells might be due to ferroelectric behavior.¹³ 

A necessary condition for a material to be ferroelectric is that it is non-centrosymmetric (lack of inversion symmetry) and polar, so that it can exhibit spontaneous polarization. The origin for ferroelectricity in perovskite-structured materials lies in the distortion of the perfect perovskite structure to a non-centrosymmetric one, where even very small distortions from cubic (centrosymmetric) to pseudo-cubic (potentially non-centrosymmetric) structures can, potentially, lead to an induced dipole in the crystals and to a polar structure. If such distortion occurs and the polar groups (in a sub-unit cell) cancel each other, the structure will be non-polar. If the polar sub-unit cells do not cancel each other, one should expect a non-centrosymmetric and polar structure that will exhibit second-harmonic generation (SHG), piezoelectricity and pyroelectricity. SHG, multiplication of the absorbed photon frequency, which is due to non-linear interaction of an absorbed photon with electrons, can occur only in non-centrosymmetric directions of a crystal (i.e., the crystal has to have such directions, which, ipso facto, make it non-centrosymmetric).¹⁶ Pyroelectricity, which also has a necessary condition of non-centrosymmetry of the material, is the generation of an electric charge as a result of a change in the polarization due to a temperature change of the crystal.¹⁷ If the spontaneous polarization can be switched by applying an external electric field, the material will be also ferroelectric (see supplementary material, Figure S2).⁹ 

Direct proof of ferroelectric materials usually requires applying high electric fields (at least hundreds (100s) V cm⁻¹ for soft materials¹⁸) that will result in polarization reversal of the polar domains. Based on the measurement of the polarization of the material as a function of an applied electric field, Fan et al. concluded that MAPbI₃ is not ferroelectric.¹⁹ Kutes et al., however, interpreted their piezo-force microscopy results as evidence for possible ferroelectricity of MAPbI₃.²⁰ 

A ferroelectric material has to be pyroelectric. While the opposite is not true, if a material is not pyroelectric, it cannot be ferroelectric and this led us to focus on searching for pyroelectric behavior of HaPs. The reason for this approach is that, in contrast to measuring ferroelectric behavior, measuring pyroelectricity does not require an external electric field, and any possible changes in the material due to the applied electric field that are not symmetry-related (e.g., artificially induced polarity due to ion diffusion, electrochemical reaction, or capacitance effects) can be ruled out. At the same time, pyroelectricity is measured by applying heat pulses (see the paragraph about pyroelectricity measurements below) and such measurements have to be done at temperatures that are far from any phase-transition temperature. Since MAPbI₃ has a (tetragonal → cubic) phase transition close to room temperature (∼50 °C),^21,22 we chose to study MAPbBr₃ for which all phase transitions are well below room temperature.

Although MAPbBr₃ has been reported to be centrosymmetric (cubic Pm3m),²² very small deviations from perfect cubicity might result in a non-centrosymmetric polar structure (e.g., R3). Moreover, the existence of a low temperature polar phase (orthorhombic P2₁cn)²³ emphasizes the potential of this material to become polar. We note that experimentally it is challenging to distinguish by X-ray or neutron diffraction between cubic structures and those with small distortions from cubicity. As an alternative method for determining the crystal symmetry (i.e., distinguish between centrosymmetric cubic and non-centrosymmetric pseudo-cubic), we used SHG spectroscopy at three different excitation wavelengths to probe a randomly dispersed powder of MAPbBr₃. We note that a related material (MAGeI₃)²⁴ was proven to be non-centrosymmetric by using SHG response.

To search for pyroelectric behavior, we used the Chynoweth technique^25–27 (explained below), an extremely sensitive method that requires high electric resistance and preferred orientation of the sample. Therefore, we did these experiments only on single crystal samples.

Crystal growth: For the growth of MAPbBr₃, we used slow saturation, with the vapor of an anti-solvent (ethyl acetate), of a 0.5M N,N-dimethylformamide (DMF) solution of lead bromide (PbBr₂) and methylammonium bromide (MABr). To form large single crystals, we used small crystals as seeds in the saturated solution. MAPbBr₃ single-crystal growth procedure and synthesis of MABr have been described previously.^6,28 X-ray diffraction shows that these single crystals expose the {100} faces of the cubic (Pm3m) phase. Powder XRD of the (pulverized) grown crystals matched previously reported X-ray diffraction data.²⁹ The X-ray analysis was done with an Ultima-III (sealed X-ray tube, Cu anode, 3 kW, RIGAKU, Japan) diffractometer.

Crystal preparation and electrode deposition: Clear cuboidal crystals were selected and two parallel {100} faces were first polished with fine polishing paper, and finally with a mechano-chemical etch, using a soft cloth, wetted with DMF. This procedure resulted in shiny, reflective surfaces. X-ray photoelectron spectroscopy (XPS) showed this process to be important to remove carbon and oxygen contaminations, as well as restoring the atomic ratios of Br : Pb and N : Pb to be close to the theoretical ones for MAPbBr₃–3:1 and 1:1, respectively (see supplementary material, Table SI).⁹ 

Each side was coated with symmetric electrodes, yielding metal-MAPbBr₃-metal structures. Gold and lead electrodes were thermally evaporated through a shadow mask and carbon electrodes were applied using carbon paint (Fig. 1 as well as supplementary material, Figure S3(a)).⁹ To make contact to the measuring circuit, we pressed copper wires against the deposited electrodes.

Crystal surface elemental analysis: XPS measurements were carried out with a Kratos AXIS ULTRA system using a concentric hemispherical analyzer for photo-excited electron detection. A monochromatic Al Kα X-ray source (hν = 1486.6 eV) at 75 W and detection pass energies ranging between 20 and 80 eV were used. Curve fitting analysis was based on linear or Shirley background subtraction and application of Gaussian-Lorentzian line shapes.

Work function (WF) measurements using contact potential difference: Contact potential difference measurements were done with a Kelvin probe in air,³⁰ using an Au grid as probe electrode that was calibrated before measurements against freshly peeled, highly ordered pyrolytic graphite (HOPG).

Second Harmonic Generation (SHG) spectroscopy: We used an LSM 880 upright system (Zeiss, Germany) with non-linear optics, coupled with a Chameleon MPX (Coherent, Inc., CA, USA) femtosecond pulsed, tunable Ti:sapphire laser for two-photon excitation, with a standard, spectral (Lambda mode) acquisition with 4.4 nm steps at excitation wavelengths of 900, 860, and 840 nm. Spectral analyses were performed using ZEN Imaging software from Zeiss.

Pyroelectricity measurements: The pyroelectricity of all samples (three samples for each type of electrode) was measured by the Chynoweth method, where the sample, in a given orientation, is exposed to a periodic temperature change.^25–27 The samples were placed on a copper plate (as in Fig. 1), which was placed in a Faraday cage and connected to an oscilloscope through a current-to-voltage amplifier (see supplementary material, Figure S4(a)).⁹ Then, the samples were irradiated through a small opening in the Faraday cage using an IR laser (1470 nm) that was modulated at frequencies between 1 Hz and 1 kHz. The overall heating that can be expected from the laser pulse at 1 Hz (longest heating time) is up to 10 °C, and usually less. The electric response was collected via an oscilloscope. The crystal, together with its electrical leads, was then turned over on top of the fixed copper plate, to heat the opposite side of the crystal, and the measurements were repeated. The same measurement was repeated for all three orthogonal directions of the crystals, showing similar results to those observed for the most developed face. The intensity of the pyroelectric (and pyroelectric-like) current is proportional to the contact area, A, $( i = A ⋅ α ⋅ d T d t )$⁠,²⁶ where α is the pyroelectric coefficient and $d T d t$ is the temperature difference over time. Therefore, to get the strongest effect, we show only the results that are attributed to the largest crystals that where measured, as these have the best developed faces.

Relative permittivity measurements: Relative permittivity measurements were done at 50 kHz and 0.05 V_AC using an Alpha-A impedance analyzer (Novocontrol Tech., Germany) over the temperature range of 25-95 °C.

One of the ways to distinguish between SHG emission and other possible emissions (e.g., normal fluorescence and phosphorescence) is that the emission wavelength of the SHG signal should be half of that of the excitation wavelength. We measured the emission spectra of the material at different excitation wavelengths. We used coarsely grained samples of ca. 5-50 μm diameter, i.e., much larger than what is needed for hyper-Rayleigh scattering (a scattering response from particles smaller than ∼100 nm). To test for SHG, we excited powder of MAPbBr₃ at three different excitation wavelengths, 900, 860, and 840 nm, which should result in SHG emission peaks at 450, 430, and 420 nm, respectively. No such response is seen.

The advantage of using a non-oriented powder is that these emissions should occur from at least some of the particles, if MAPbBr₃ lacks inversion symmetry. Fig. 2 shows a wide emission between 420 and 460 nm, but without correlation between the excitation and the emission wavelengths. This broad emission seems to be related to a luminescing transition. It is not due to stray light, as it is clearly above the noise level of a blank reference. Although this result strongly suggests that MAPbBr₃ is centrosymmetric, the broad emission in the relevant wavelength range prevents drawing a definitive conclusion regarding the crystal’s symmetry. Thus, we turned to the more stringent test of pyroelectricity in MAPbBr₃ (because a material can only be ferroelectric if it is pyroelectric; for clarification see supplementary material, Figure S2).⁹ 

When a sample is locally heated via illumination with an IR source, the electric current response can have several causes as follows:

• a thermoelectric effect,³¹ i.e., an electric current as a result of a temperature gradient;

• a flexoelectric effect,³² which is caused by a thermally induced stress gradient;

• release of trapped charges; and

• a pyroelectric effect, which comes from the change of the material’s polarization with temperature; this effect is proportional to the temperature change.

The most significant difference between signals that come from a pyroelectric effect and those that emanate from the thermoelectric and flexoelectric effects is that only a pyroelectric current will reverse its sign if opposite sides of the crystal are heated, as shown for a known pyroelectric material, LiTaO₃ (see supplementary material, Figure S4(b)).⁹ In the case of trapped charges, one can observe a response to a temperature change that is similar to that from pyroelectricity, i.e., inversion in the electric current, if the total charge on each side of the crystal is different in value or sign. Thus, because the currents, resulting from pyroelectricity and from emission of trapped charge signals, could be similar, one must do the following three tests to distinguish between them:³³ 

1. Measure with several types of electrodes. Since pyroelectricity is an intrinsic property of the material, a pyroelectric signal is not dependent on the electrodes. In contrast, since the trapped charge is trapped at the surface of the materials, the signal that results from emission of trapped charge will be very sensitive to the material/electrode interface, the electrode deposition process, and the electrode WF. We note that an induced field at a metal-semiconductor interface will create a symmetry-related pyroelectricity, as in the case of p-n junctions.³⁴ This surface-pyroelectric signal has a distinctive profile (i.e., the same sign for both measured sides of a crystal and a much faster decay than of a bulk-related pyroelectric signal³³).

2. Perform many consecutive cycles of measurements (i.e., induce cycles of heating and cooling of a few degrees at short circuit). Because, during the measurement, the two contacts of the crystal are shorted, a signal that stems from emission of trapped charge will decay with time due to release of the trapped charges during the temperature change. In contrast, a pyroelectric signal should not change, since it results from redistribution of the compensation charge as a response to the change in the native polarization of the material^35,36 (see supplementary material, Figure S4(c)).⁹ 

3. During periodic heating/cooling cycles, heat the sample well above the starting temperature and hold it there for a certain period of time as this should facilitate release of trapped charges. If no phase transition occurs, a pyroelectric signal should be reversible with respect to temperature, in contrast to a signal that stems from emission of trapped charge, which should change (decrease) after heating (see supplementary material, Figure S4(d)).⁹ 

The electric signals that were obtained as a response to the laser heating on the different types of electrodes, deposited on the two parallel hemihedral faces of MAPbBr₃, were different with each electrode (Fig. 3). The sample with the C electrodes (Fig. 3(a)) exhibited a slow rise and a weak electric signal (few pA) without sign reversal when heating opposite sides of the crystals.³¹ The initial response from the sample with the Pb electrodes (Fig. 3(b)) was similar to that with the C electrodes (same sign from both sides of the crystals), but in contrast to the case with C electrodes, for Pb electrodes ∼15 ms after the start of the laser pulse, the signals from opposite sides of the sample started to show opposite signs (see supplementary material, Figure S5).⁹ The electric signals from the sample with the Au electrodes (Fig. 3(c)) were much stronger (few nA) than those from samples with the other electrodes and had opposite signs from the opposite sides of the sample. The fact that we see these signals is a clear indication that the electronic conductivity of the sample is not high enough to, potentially, suppress polarity or ferroelectricity, if such exist at all.

Pyroelectricity is an intrinsic material property that should not change with the electrode type. However, our finding that the electric responses of the samples with three different electrode pairs differ from each other suggests that the different responses of the samples with the Pb or Au electrodes might not arise from crystal polarity. Since, as mentioned before, trapped charges can also induce electric signal inversion as parallel sides of a crystal are being measured, they are the most likely reason for the pyroelectric-like response that was seen with the Pb and Au electrodes. The difference between the responses with the three electrode pairs may come from different electronic energy barriers for charge transport between the electrodes and the MAPbBr₃ {100} faces. Such differences can originate from electrode-semiconductor (MAPbBr₃) WF differences. The WF differences, deduced from the measured contact potential differences (see supplementary material, Figure S3(b)),⁹ clearly show that the MAPbBr₃/Au WF difference is the largest, which correlates with the large electric signal, obtained with Au, as compared to Pb and C electrodes, but cannot explain the difference between Pb and C.

An additional explanation can be the formation of interface states³⁷ as a result of differences between the electrodes and their deposition method (i.e., temperature differences during vaporization of Pb and Au or the presence of iso-propanol thinner in the case of C). We note that charge accumulation due to induced interface states was suggested to occur between a contact and MAPbI₃³⁸ and this may also be the cause of the trapped charge we invoke to explain our results.

To check this, we measured the time dependence of the response. The time-dependent measurement (Fig. 4) shows that after 4.5 × 10⁷ cycles of the sample with the Au electrodes and 4.5 × 10⁴ cycles of the sample with the Pb electrodes, the electric signals obtained from them are significantly decreased. These results support the idea that the signal comes from the emission of trapped charge, because it is unlikely that the decay in signal stems from the degradation of the material, as the faces that are exposed to the laser are protected by the electrodes and the total heating of the sample is at most a few degrees above ambient room temperature.

Since the electrical conductivity of MAPbBr₃ increases with temperature (see supplementary material, Figure S6),⁹ the heating process increases charge recombination and, thus, will speed up the elimination of trapped charges. Heating the sample with the Pb electrodes to 52.5 °C (Fig. 5(a)) and then cooling back to room temperature (25.5 °C) make the pyroelectric signal disappear, leaving only a thermoelectric signal, as indicated by the absence of inversion of the signal’s sign after illuminating the opposite crystal face.

To exclude the possibility that the disappearance of the pyroelectric-like signal originates from spontaneous depolarization, we checked the behavior of the relative permittivity as a function of temperature (Fig. 5(b)). Spontaneous depolarization, a random reorientation of the polar domains in polar materials occurs near (below) the Curie temperature.^18,39 This means that the Chynoweth method can be used to distinguish between polar and non-polar materials only for a material at a temperature far from any phase transition that might be a Curie temperature (which presents a problem for measurements on MAPbI₃ at temperatures, relevant for its use in solar cells). Since at the Curie temperature the relative permittivity has a maximum,^40–42 the fact that we do not see any peak of the relative permittivity near the measurement temperature tells us that measurements were not done near the Curie temperature and therefore, spontaneous depolarization is highly unlikely to occur. We can see that the relative permittivity decreases after the first heating and cooling cycle, consistent with some de-trapping in the sample during the heating. We note that, although ferroelectric domains might be suppressed due to screening of charge carriers,⁴³ such ferroelectricity suppression occurs in materials with conductivity that is much higher (≥ ∼ 1 Ω⁻¹ cm⁻¹) than the MAPbBr₃ crystals studied here (∼0.1 μΩ⁻¹ × cm⁻¹) (see supplementary material, Figure S6).⁹ 

To further support our conclusion that trapped charge is responsible for the polar behavior of the MAPbBr₃ samples, we applied a bias of ±0.5 V for 30 min to the samples. Since the field that this bias creates (∼5 V cm⁻¹) is too small for spontaneous depolarization even in crystals of soft materials,³⁶ it can only affect the trapped charge. We find (Fig. 6) that the applied bias can increase or decrease the electric response depending on the bias sign.

In conclusion, our experimental results show that MAPbBr₃ is not polar. The polar behavior that this material might exhibit, like apparent pyro-, piezo-, and ferro-electricity, stems from charge that is easily trapped at the perovskite surface/interface—most prominently with gold electrodes. The claim of charge trapping is supported by a recent publication,³⁸ which suggests that the hysteresis in the I-V characteristics of planar perovskite solar cells originates from defects at the perovskite layer interfaces.

Because the dielectric properties of a material are determined by its average structure, in this paper, we do not directly address, nor are we here concerned with the local structure, but only with the average structure of MAPbBr₃. As to the possible existence of polar nanodomains, as suggested in Refs. 13 and 14, if there were nanodomains, which give local polarization and local non-centrosymmetry, we should have seen evidence for this in the SHG spectroscopy experiment as a hyper-Rayleigh scattering. We do not see any such evidence.

Although we only tested MAPbBr₃, it is clear that to be able to invoke ferroelectricity for the behavior of MAPbI₃, more, and especially more direct evidence is needed. While this is the subject of ongoing studies, these are complicated by the above-mentioned caveats, regarding phase transitions.

We thank Dr. Stanislav Kamba (Institute of Physics, Czech Republic), and Dr. Thomas M. Brenner and Lior Ne’eman (Weizmann Institute of Science, Israel) for helpful discussions. This work was supported by the Israel Science Foundation, the Israel Ministry of Science, the Israel National Nano-Initiative and by the Kimmel Centre for Nanoscience and the Grand Centre for Sensors and Security, both at the Weizmann Institute. D.C. holds the Sylvia and Rowland Schaefer Chair in Energy Research.