A study on the nature of the thermal decomposition of methylammonium lead iodide perovskite, CH₃NH₃PbI₃: an attempt to rationalise contradictory experimental results

Abstract

The nature of the gas phase product released during the thermal decomposition of CH₃NH₃PbI₃ (methylammonium lead iodide) to PbI₂ (lead diiodide) under vacuum is discussed on the basis of thermodynamic predictions, recently published experimental results, and new experiments presented here. From the limited data currently available, the nature of the main decomposition path is not clear because, both, the process releasing HI(g) + CH₃NH₂(g) (1) and that leading to NH₃(g) + CH₃I(g) (2) were observed under different conditions. Our thermodynamic analysis showed that process (2) is largely favoured for all the CH₃NH₃PbX₃ (X = Cl, Br, I) compounds. However, Knudsen effusion mass spectrometry experiments (temperature range 140–240 °C) showed that HI(g) and CH₃NH₂(g) were the predominant species in the vapor, with process (2) occurring to a much smaller extent than suggested by the thermodynamic driving force, thus being of minor importance under effusion conditions. We also found that this process was comparatively enhanced by high temperatures and low effusion rates (high impedance orifice). Our experimental evidence suggested that the thermodynamically favoured process (2) was affected by a significant kinetic hindrance. Overall, the prevailing decomposition path is likely to markedly depend on the actual operative conditions.

---

Introduction

The most investigated among light harvester materials based on organic–inorganic hybrid perovskites, is most likely methylammonium lead iodide, CH₃NH₃PbI₃, with a maximum conversion efficiency exceeding 20%.^1–3 However, a problem with large-scale applications is the low stability of this and similar compounds.^4–6 Great effort has been made in the last few years to investigate in detail the degradation of this material under several external factors usually involved in the operation of photovoltaic devices, such as humidity, irradiation, and electrical current.^3–8 With regards to the interactions with water, various degradation paths were suggested depending on the specific conditions applied, involving the formation of PbI₂, I₂, H₂, methylamine, and hydrated forms of CH₃NH₃PbI₃.^8–10 Theoretical approaches were also used to investigate the degradation mechanism at the atomic scale.^11,12 Various protection strategies were proposed based on the encapsulation and interface engineering techniques^8,13 as well as a number of chemical modifications including partial substitution of iodine with other halide species,¹⁴ longer alkylammonium chains,¹⁵ replacement of lead by other elements such as bismuth,¹⁶ and/or partial substitution of alkylammonium ions with different cations, such as Cs⁺ and formamidinium.¹⁷

Apparently, less attention was paid to the study of the intrinsic thermal instability of CH₃NH₃PbI₃, (i.e., the degradation behaviour under heating in vacuum or inert atmosphere). A few recent reports addressed this issue, both by experiments^18,19 and by theoretical calculations.²⁰ While the formation of PbI₂ as the condensed phase degradation product is well established, far less information is available about the gaseous products released upon degradation. This information is of interest to optimize the synthetic conditions and annealing of this material.^18,21,22

The thermal degradation of CH₃NH₃PbX₃ (X = Cl, Br, I) proceeds by the release of volatile products, as reported in several studies based on thermogravimetry (TGA).^23,24 This is the reason why, despite the order of enthalpies of formation, which follows an increasing trend going from Cl to Br to I,²⁵ the actual stability under real conditions is reversed, with CH₃NH₃PbI₃ being more stable than CH₃NH₃PbCl₃.²⁴ On the basis of mass loss data, Dualeh et al.²⁴ proposed a pathway based on the consecutive loss of HI(g) and CH₃NH₂(g). However, classic thermogravimetric experiments provided the total mass loss rate, with no direct information on the gaseous species released during decomposition.

To the best of our knowledge, the direct detection of the nature of the vapor phase produced from CH₃NH₃PbX₃ thermal degradation was only reported by Williams et al.²³ in a work based on TGA coupled with FTIR (Fourier transform infrared) spectroscopy. The conclusion of this study was that at 265 °C, the simple methylammonium halides CH₃NH₃Cl and CH₃NH₃I decomposed to give HCl(g) and CH₃NH₂(g) and CH₃I(g) and NH₃(g), respectively. With regards to the lead halide perovskites, IR spectra in ref. 23 were complicated by the superposition of solvent (DMF) peaks. However, at 230 °C, IR peaks of HCl(g) and CH₃NH₂(g) began to appear for CH₃NH₃PbI_3−xCl_x, whereas only a weak signal for ammonia was recognizable for CH₃NH₃PbI₃.²³

In 2016, two reports appeared in the literature specifically addressing the problem by the direct mass spectrometric observation of the evolving gases. The first contribution was provided by our group²⁶ based on evaporation experiments carried out for all three CH₃NH₃PbX₃ compounds under Knudsen effusion conditions (KEMS, Knudsen effusion mass spectrometry), with a mass spectrometric analysis of the effusing vapours in the overall temperature range of 50–166 °C (53–134 °C for CH₃NH₃PbI₃), encompassing the conditions used for the operation of the photovoltaic devices. On this basis, the following decomposition process was proposed to take place under the above mentioned experimental conditions:

CH₃NH₃PbI₃(s) = PbI₂(s) + CH₃NH₂(g) + HI(g) (1)

Similar results were found for chloride and bromide perovskites.

Shortly after, Juarez-Perez et al.²⁷ used thermogravimetry coupled with mass spectrometry to investigate the mass loss process of CH₃NH₃PbI₃ (and CH₃NH₃I) in open-crucible conditions under a helium flow at 300–420 °C. They concluded that the degradation process instead took place in the following manner:

CH₃NH₃PbI₃(s) = PbI₂(s) + CH₃I(g) + NH₃(g) (2)

It is worth noting that the decomposition path (2) had been already observed in quaternary alkylammonium halides R₄NX,^28,29 where no hydrogen atom was present in the cation, whereas process (1) was proposed for primary, secondary and tertiary alkylammonium counterparts, including the compounds RNH₃X of interest here.^28–31

In summary, the decomposition path of the CH₃NH₃PbX₃ perovskite material appears to still be controversial. Herein, we present further KEMS experiments to shed some light on the somehow contradictory results of the few experimental studies available to date. A brief thermodynamic analysis of the competition between processes (1) and (2) under equilibrium conditions is first presented.

Thermodynamic analysis

From a purely thermodynamic point of view (physicochemical equilibrium), the competition between processes (1) and (2) for the generic perovskite halide CH₃NH₃PbX₃ is governed by the K_p ratio. To evaluate the prevailing process, this quantity is calculated as follows:where Δ_rG° is the standard Gibbs energy change of the gaseous reaction (4)

CH₃NH₂(g) + HX(g) = CH₃X(g) + NH₃(g) (4)

In terms of partial pressures of the two gaseous species, p(HX) = p(CH₃NH₂) = p(1) and p(NH₃) = p(CH₃X) = p(2), eqn (3) becomes

The Δ_rG° of reaction (4) was calculated from the thermodynamic properties of the four species involved using the database IVTANTHERMO³² for all species except CH₃NH₂(g) (not included in ref. 32), for which the database by Frenkel et al.³³ was used. The derived Δ_rG° values can be approximated by the following linear relationships in the 298.15–700 K temperature range:

As expected, the entropic term has a very small influence on the Δ_rG° value in the temperature range of interest, but it governs the increasing trend of Δ_rG° with temperature.

In the temperature range of interest according to eqn (6), Δ_rG° is strongly negative. Therefore, on a thermodynamic basis, we conclude that all three lead halide perovskites CH₃NH₃PbX₃ (X = Cl, Br, I) should decompose to give ammonia and methyl halide as the predominant gaseous species under equilibrium conditions. From eqn (5) and (6), the plot reported in Fig. 1 was obtained for p(1)/p(2), showing process (2) as the major process in all cases, and particularly favoured for iodine. The contribution of process (2) decreases with increasing temperature, with a trend more pronounced for iodine.

Fig. 1 Thermodynamic competition between the decomposition processes (1) CH₃NH₃PbX₃(s) = PbX₂(s) + CH₃NH₂(g) + HX(g) and (2) CH₃NH₃PbX₃(s) = PbX₂(s) + CH₃X(g) + NH₃(g), with X = Cl, Br, I, expressed by the ratio between the corresponding equilibrium partial pressures (see eqn (3)–(5) in the text).

More insight into this thermodynamic prediction can be obtained from the data in Table 1, where the standard formation enthalpies Δ_fH° (kJ mol⁻¹) at 298 K are reported for all the species involved. Basically, the thermodynamic competition between (1) and (2) was driven by the different stabilities of CH₃NH₂(g) and NH₃(g) (the latter being more stable by 23.4 kJ mol⁻¹) and of HX(g) and CH₃X(g). As evident from Table 1, both hydrogen halides and methyl halides became more unstable going from Cl to I, but the trend for HX was more pronounced, making the formation of HI particularly unfavoured compared to that of CH₃I. As a consequence, in the case of iodine, process (2) was particularly favoured.

Table 1 Standard enthalpy of formation (T = 298 K, kJ mol⁻¹) of the gaseous species involved in processes of (1), (2) and (4)^a

X	CH3NH2(g)	HX(g)	CH3X(g)	NH3(g)
a All data are from the database IVTANTHERMO,^32 except for CH3NH2(g) (not reported in ref. 32), for which the database by Frenkel et al.^33 was used.
Cl	−22.5	−92.31	−81.87	−45.94
Br	−22.5	−36.29	−36.4	−45.94
I	−22.5	26.5	14.4	−45.94

---

As far as thermodynamic equilibrium is concerned, the above analysis and conclusions apply as such to the decomposition equilibria of the simple methylammonium halides as well. As mentioned in the introduction, these compounds are reported²⁸ to decompose releasing methylamine and hydrogen halide in the gas phase, which is in contrast with thermodynamic predictions. However, more recently, Williams et al.²³ reported spectroscopic evidence supporting process (1) for CH₃NH₃Cl and process (2) for CH₃NH₃I.

It is also interesting to note that according to eqn (6), the tendency towards decomposition would significantly increase from chloride to iodide, whereas the opposite trend was reported on the basis of TGA²⁴ and Knudsen effusion²⁶ measurements. Thus, this confirmed that the actual degradation process does not take place under thermodynamic control.

Based on eqn (3)–(5), the above reported thermodynamic estimate has the advantage of being independent from the thermodynamic properties of CH₃NH₃PbI₃, which are not well established. From heat capacity measurements,³⁴ an absolute entropy of 374.1 J K⁻¹ mol⁻¹ at 298 K was reported. With regards to the enthalpy of formation, our tensimetric value²⁶ (−375.7 ± 9.7 kJ mol⁻¹ at 298 K) is not in good agreement with the only available calorimetric result (−341.7 ± 1 kJ mol⁻¹),²⁵ whereas is in satisfactory agreement with values from first-principle calculations.²⁰ Even by assuming a more negative value of −375.7 ± 9.7 kJ mol⁻¹ for the enthalpy of formation, partial pressures of NH₃ and CH₃I of about 4 × 10⁻³ bar at 500 K were estimated – a huge volatilization was never experimentally observed.

The comparison between the thermodynamic analysis and the experimental findings on CH₃NH₃PbI₃ summarized in the introduction led us to conclude that the decomposition of this compound took place under kinetic control, regardless of the prevalence of process (1) or (2). It is interesting to note that in the previous literature^28,29 on alkylammonium halide degradation, the decomposition of quaternary salts was reported to begin at much higher temperatures than primary, secondary and ternary ones. Since quaternary salts are the only species where process (1) cannot occur due to the lack of N–H bonds, it is clear that the breaking of C–N bonds required in process (2) is kinetically slow.

In conclusion, several important questions were raised from the above analysis. In particular, the detection of the decomposition channel (1) rather than the thermodynamically favoured path (2) in the KEMS experiment²⁶ and the opposite findings resulting from the TGA-MS results²⁷ call for further experimental investigations, which were undertaken in the present study and are reported below.

Experimental details

Samples of CH₃NH₃PbI₃ were synthesized and characterized as reported elsewhere.²⁶

All the experiments were carried out by a mass spectrometry technique described in our previous study.²⁶ In a typical experiment, vapours effusing from a heated Knudsen-type source are ionized by an electron beam with an energy that could be varied from the appearance potential of a given ion up to 80 eV. The cations that are produced enter a magnetic sector zone where they were deflected according to their mass/charge ratio. The ion current of a given ion is finally measured by an electron multiplier detector.

In addition to the experiments performed under Knudsen effusion conditions with effusion orifices of different impedances (see next section), we carried out a free surface evaporation experiment (Langmuir conditions). Unlikely KEMS experiments, where the sample is placed inside a graphite Knudsen cell and the effusing flow is analysed by the mass spectrometer, in the Langmuir-type experiment, the sample was contained in an open crucible, also made of graphite, directly facing the ion source.

In all the experiments, the instrumental conditions were similar to those of ref. 26. As a rule an electron energy of 25 eV was used. The intensity of each peak was background-subtracted by shifting a movable shutter placed just above the orifice in order to measure only the portion of the signal due to the vapour species coming from evaporation.

Both the initial samples and the vaporized residues were analysed by X-ray diffraction (XRD) analysis. As expected, XRD spectra collected after vaporization experiments showed the presence of PbI₂ alone (complete decomposition) or together with residual CH₃NH₃PbI₃ (partial decomposition) in all cases. In the experiments where complete decomposition was allowed to occur, a gradual drop of all the ion intensities was observed rather than a sharp fall, similar to observations reported in TGA studies for primary, secondary and tertiary alkylammonium bromides and iodides.^30,31

Results and discussion

We previously reported CH₃NH₃PbI₃ to decompose following process (1), CH₃NH₃PbI₃(s) = PbI₂(s) + CH₃NH₂(g) + HI(g), on the basis of KEMS measurements performed at 53–134 °C, low enough to encompass the real temperatures of photovoltaic devices.²⁶ The most intense signals recorded in the mass spectra of the vapours produced by heating CH₃NH₃PbI₃ in a Knudsen source were those at m/z = 31 and 128, assigned as molecular ions of the neutral precursors CH₃NH₂(g) and HI(g). Other clear signals were detected at m/z = 30 and 127, assigned to fragment ions of the same parent species. Similar results were found for the chloride and bromide perovskites. Total pressures in the 10⁻⁹ bar to 10⁻⁵ bar range were measured within this temperature interval.

Shortly afterwards, Juarez-Perez et al.²⁷ investigated the mass loss of CH₃NH₃PbI₃ (and CH₃NH₃I) in open-crucible conditions under a helium flow using thermogravimetry coupled to mass spectrometry. On the basis of mass spectra collected at 300–420 °C, in which a significant mass loss occurred, the authors concluded that the degradation process taking place was instead CH₃NH₃PbI₃(s) = PbI₂(s) + CH₃I(g) + NH₃(g) (2). Because of detection limits, the temperature range covered in the their TGA-MS experiments²⁷ was significantly higher than ours, as well as than the temperatures of interest for practical applications of perovskite-based devices. However, the presence of the CH₃I(g) species at much lower temperatures (down to 80 °C) was also reported in the ESI of the same study, pushing the limits of the technique by a clever variation of the heating method that involved oscillating heating/cooling cycles.

The results reported in ref. 27 are in part consistent with the previous TGA-FTIR study by Williams et al.²³ In the latter, lead halide perovskites CH₃NH₃PbI_3−xCl_x and CH₃NH₃PbI₃ were subjected to evaporation in the form of dimethylformamide solutions, and the vapours were transferred from the TGA apparatus to the FTIR assembly by means of a heated (300 °C) line. Although the interpretation of IR spectra was made difficult by the superposition of solvent bands, peaks of HCl(g) and CH₃NH₂(g) were reported to appear for CH₃NH₃PbI_3−xCl_x at 230 °C, whereas only a weak signal of ammonia was recognized for CH₃NH₃PbI₃.²³ Further measurements carried out on silica-desiccated materials gave no significant mass loss signal up to 250 °C.

Possible reasons for these discrepancies are factors such as different temperature ranges, kinetic effects, heating rates, static vs. dynamic regimes, effusion vs. free-surface conditions, vacuum vs. carrier gas atmosphere, and the presence/absence of a vapor transfer line. The new KEMS experiments presented here were planned in particular to investigate the effect of temperature, kinetic hindrance and evaporation conditions on the decomposition behaviour of CH₃NH₃PbI₃. The following experiments were performed:

(1) An experiment was performed under the same conditions used in our previous study²⁶ (i.e., typical effusion cells with an orifice of 1 mm diameter and negligible thickness). In the following, these experiments will be indicated as “low impedance KEMS”, where, as usual in gas-flow terminology, the term “impedance” borrowed from electricity was used in the sense of “resistance to flow”. Measurements were done at T = 473 K and T = 514 K. These experiments were aimed at further confirming the findings of our previous paper,²⁶ paying special attention to the possible presence of the gaseous products of process (2) (m/z = 142 for CH₃I and/or 17 for NH₃) in addition to those of process (1). Indeed, our mass spectrometry measurements were done in single ion monitoring mode, and in that study,²⁶ the strong signals corresponding to HI(g) (m/z = 128) and CH₃NH₂(g) (31) were extensively monitored as a function of temperature, whereas only few scans were done in the lower and higher mass ranges.

(2) A control experiment under Langmuir conditions (see Experimental section), designed to test the hypothesis that the different evaporation modes used in our KEMS measurements²⁶ (effusion) and in Juarez-Perez's experiments (open crucible)²⁷ could strongly affect the decomposition behaviour. These measurements were performed at T = 493 K (i.e. the average temperature of the KEMS experiments).

(3) An additional Knudsen-type experiment was carried out with a cell having a smaller effusion orifice (0.5 mm diameter) and a thicker cap (0.65 mm) to create a high impedance path to the effusing vapor (“high impedance KEMS”). Since reaching thermodynamic equilibrium inside the cell was favoured in this configuration due to the lower effusion rate, this experiment aimed at giving information on the kinetic/thermodynamic competition between the two decomposition channels. Moreover, in order to investigate the effect of temperature, these experiments were performed in a larger temperature range, from 416 K to 503 K.

In all the experiments, the most intense signals observed were those at m/z = 128, 31, and 30, assigned to the neutral precursors HI(g) and CH₃NH₂(g). These results clearly confirmed the conclusion of our previous study, where the main decomposition channel was process (1). However, clear, though lower background-subtracted signals were also observed at m/z = 142, assigned to the species CH₃I(g) produced in process (2). We also observed a measurable signal at m/z = 17 and 18. The former could be assigned to NH₃⁺ as well as to OH⁺, the latter coming from fragmentation of H₂O(g) (m/z = 18). Less intense signals were measured at m/z = 127 and 254, corresponding to I and I₂: the former coming probably from the fragmentation, the latter from the thermal decomposition of HI(g).³² Interestingly, a small peak (less than 1% of the most intense signals) was finally observed at m/z = 159 due to the undissociated species CH₃NH₃I(g).

Overall, very significant differences were found in the intensity ratios of the relevant species in the three series of experiments above. To investigate the competition between the two decomposition channels, we followed the intensity ratios of HI/CH₃I (128/142) and CH₃NH₂/CH₃I (31/142) rather than the corresponding ratios involving NH₃ instead of CH₃I, due to the overlapping of the former with OH.

It should be noted that in KEMS experiments, the ion intensity ratios are proportional, not equal, to the corresponding ratios of the partial pressures inside the cell because multiplier gains and ionization cross sections are not included (see below and ref. 26).

In Table 2, the intensity ratios of 128/142 and 31/142 measured in the three experiments above are reported. The same ratios are also plotted in Fig. 2. It is worth noting that, due to the higher sensitivity of KEMS compared to TGA, our measurements were done at lower temperatures than in ref. 23 and 27, much closer to the actual operative conditions of photovoltaic devices.

Table 2 Ion intensity ratios measured in the various mass spectrometry experiments described in the text

T/K	HI/CH3I			CH3NH2/CH3I
	Low impedance KEMS	High impedance KEMS	Langmuir	Low impedance KEMS	High impedance KEMS	Langmuir
416.0		683.3			200.0
417.8		509.1			139.4
460.9		109.3			25.6
464.5		81.4			20.0
472.5	1165.4			319.2
474.0		63.3			13.6
493.0			294.9			90.4
498.0		24.6			6.1
502.6		25.7			5.3
503.0		27.1			5.6
514.1	294.1			75.9

---

Fig. 2 Logarithmic plot of the HI/CH₃I (a) and CH₃NH₂/CH₃I (b) ion intensity ratios measured under different effusive conditions.

Some important conclusions can be drawn from Table 2 and Fig. 2:

(i) Under all the explored conditions, process (1) prevailed over process (2). Since the ion intensity ratios are a good approximation to the corresponding partial pressure ratios, results in Fig. 2 are in clear contrast with the thermodynamic predictions reported in Fig. 1.

(ii) Under typical conditions for Knudsen effusion experiments, such as those used in ref. 26 and in the above described experiment 1 (“low impedance KEMS”), process (1) was overwhelmingly dominant, with a contribution from process (2) below 1%. This finding confirmed our previous results.²⁶

(iii) Under free-surface conditions, process (2) seemed slightly enhanced (see Table 2), but not to such a large extent to explain the different behaviour observed in ref. 27.

(iv) A significant increase in the contribution of process (2) was observed under conditions where the effusion rate was lower (high impedance orifice, i.e., closer to equilibrium conditions).

(v) The contribution of process (2) considerably increased with increasing temperature, again in contrast with thermodynamic predictions reported in the previous section (see Fig. 1).

Points (iv) and (v) in particular led us to conclude that the thermodynamically favoured process (2) suffered from severe kinetic limitations, which could be partly overcome at high temperatures and/or low effusion rates. This could be explained by taking into account the fact that breaking the hydrogen bond in the methylammonium cation leading to the release of methylamine and hydrogen iodide requires a much lower energy than the rupture of the N–C bond necessary to produce methyl iodide and ammonia.^30,31 Note that not only process (2) is kinetically limited, but also the gas phase equilibration according to eqn (4) was far from being attained. The theoretical analysis reported in ref. 27, based on gas phase DFT calculations, partially confirmed this view, although no transition state was found leading to the formation of HI and CH₃NH₂ as final products.

The evidence that increasing temperature enhances process (2) over process (1) was consistent with the fact that the former was observed from 300 °C to 420 °C in ref. 27. In addition, we made an attempt to convert the ion intensity ratio in Fig. 1a (high impedance data) into the corresponding partial pressure ratio, in order to estimate the temperature where p(HI) and p(CH₃I) became equal. As discussed elsewhere,²⁶ this conversion required a knowledge or estimation of the ionization cross sections (σ) of the species involved. By adopting the values σ(HI) = 6.47 Ų (ref. 12) and σ(CH₃I) = 10.3 Ų (ref. 35), and by neglecting the very small contribution of the detector gain, an extrapolated temperature of about 385 °C was determined, well within the above mentioned temperature range.

Conclusions

On the basis of the above discussion, we conclude that in the temperature range (140–240 °C) and under the various experimental conditions explored, the thermal decomposition of CH₃NH₃PbI₃ mainly takes place by the release of HI(g) and CH₃NH₂(g) as gaseous products. The process leading to NH₃(g) and CH₃I(g) was thermodynamically favoured, but kinetically hindered, most probably because of the more complex rearrangements and bond breaking necessary to form NH₃ and CH₃I in the gaseous phase from the crystal. As a consequence, the latter process remained as a minor decomposition channel despite its higher thermodynamic driving force. However, kinetic hindrance was shown to significantly diminish at higher temperatures and under low effusion rate conditions (high impedance orifice), which allowed for the thermodynamic control to be favoured. On the other hand, when carried out under free-surface (Langmuir) conditions, the decomposition behaviour did not show important changes compared to effusion conditions. Overall, our analysis showed that the decomposition path of CH₃NH₃PbI₃ and related light harvesters markedly depends on a number of experimental conditions that can favour the thermodynamic or kinetic control, with temperature, heating rate, static/dynamic regime, and vacuum vs. carrier gas atmosphere playing key roles.

Further investigations are needed for an in-depth understanding of the microscopic/mechanistic aspects of the evaporative decomposition of these compounds, to clarify the degradation pathway followed by CH₃NH₃PbI₃-based photovoltaic devices under actual operating conditions and to assess more effective protection strategies.

Acknowledgements

The idea for this study originated from a private communication between the present authors and Dr Emilio J. Juarez-Perez, first author of ref. 27. The fruitful discussion with Dr Juarez-Perez is gratefully acknowledged.

References

1. T. A. Berhe, W.-N. Su, C.-H. Chen, C.-J. Pan, J.-H. Cheng, H.-M. Chen, M.-C. Tsai, L.-Y. Chen, A. A. Dubale and B.-J. Hwang, Energy Environ. Sci., 2016, 9, 323–356.
2. L. K. Ono and Y. Qi, J. Phys. Chem. Lett., 2016, 7, 4764–4794.
3. H. Yuan, E. Debroye, K. Janssen, H. Naiki, C. Steuwe, G. Lu, M. Moris, E. Orgiu, H. Uji-i, F. De Schryver, P. Samrì, J. Hofkens and M. Roeffaers, J. Phys. Chem. Lett., 2016, 7, 561–566.
4. P. Docampo and T. Bein, Acc. Chem. Res., 2016, 49, 339–346.
5. J. S. Manser, M. I. Saidaminov, J. A. Christians, O. M. Bakr and P. V. Kamat, Acc. Chem. Res., 2016, 49, 330–338.
6. R. Gottesman and A. Zaban, Acc. Chem. Res., 2016, 49, 320–329.
7. F. Matsumoto, S. M. Vorpahl, J. Q. Banks, E. Sengupta and D. S. Ginger, J. Phys. Chem. C, 2015, 119, 20810–20816.
8. J. M. Frost, K. T. Butler, F. Brivio, C. H. Hendon, M. van Schilfgaarde and A. Walsh, Nano Lett., 2014, 14, 2584–2590.
9. N. Guangda, G. Xudong and W. Liduo, J. Mater. Chem. A, 2015, 3, 8970–8980.
10. A. M. A. Leguy, Y. Hu, M. Campoy-Quiles, M. I. Alonso, O. J. Weber, P. Azarhoosh, M. van Schilfgaarde, M. T. Weller, T. Bein and J. Nelson, Chem. Mater., 2015, 27, 3397–3407.
11. E. Mosconi, I. M. Azpiroz and F. De Angelis, Chem. Mater., 2015, 27, 4885–4892.
12. L. Zhang and P. H.-L. Sit, J. Phys. Chem. C, 2015, 119, 22370–22378.
13. D. Koushik, W. J. H. Verhees, Y. Kuang, S. Veenstra, D. Zhang, M. A. Verheijen, M. Creatore and R. E. I. Schropp, Energy Environ. Sci., 2017, 10, 91–100.
14. R. Ruess, F. Benfer, F. Böcher, M. Stumpp and D. Schlettwein, ChemPhysChem, 2016, 17, 1505–1511.
15. I. C. Smith, E. T. Hoke, D. Solis-Ibarra, M. D. McGehee and H. I. Karunadasa, Angew. Chem., Int. Ed., 2014, 53, 11232–11235.
16. R. L. Z. Hoye, R. E. Brandt, A. Osherov, V. Stevanović, S. D. Stranks, M. W. B. Wilson, H. Kim, A. J. Akey, J. D. Perkins, R. C. Kurchin, J. R. Poindexter, E. N. Wang, M. G. Bawendi, V. Bulović and T. Buonassisi, Chem.–Eur. J., 2016, 22, 2605–2610.
17. M. Saliba, T. Matsui, J.-Y. Seo, K. Domanski, J.-P. Correa-Baena, M. Khaja Nazeeruddin, S. M. Zakeeruddin, W. Tress, A. Abate, A. Hagfeldt and M. Grätzel, Energy Environ. Sci., 2016, 9, 1989–1997.
18. B. Conings, J. Drijkoningen, N. Gauquelin, A. Babayigit, J. D'Haen, L. D'Olieslaeger, A. Ethirajan, J. Verbeeck, J. Manca, E. Mosconi, F. De Angelis and H.-G. Boyen, Adv. Energy Mater., 2015, 5, 1500477.
19. A. F. Akbulatov, S. Y. Luchkin, L. A. Frolova, N. N. Dremova, K. L. Gerasimov, I. S. Zhidkov, D. V. Anokhin, A. Z. Kurmaev, K. J. Stevenson and P. A. Troshin, J. Phys. Chem. Lett., 2017, 8, 1211–1218.
20. E. Tenuta, C. Zheng and O. Rubel, Sci. Rep., 2016, 6, 37654.
21. Z. Song, S. C. Watthage, A. B. Phillips, B. L. Tompkins, R. J. Ellingson and M. J. Heben, Chem. Mater., 2015, 27, 4612–4619.
22. D. P. Nenon, J. A. Christians, L. M. Wheeler, J. L. Blackburn, E. M. Sanehira, B. Dou, M. L. Olsen, K. Zhu, J. J. Berry and J. M. Luther, Energy Environ. Sci., 2016, 9, 2072–2082.
23. A. E. Williams, P. J. Holliman, M. J. Carnie, M. L. Davies, D. A. Worsley and T. M. Watson, J. Mater. Chem. A, 2014, 2, 19338–19346.
24. A. Dualeh, P. Gao, S. I. Seok, M. K. Nazeeruddin and M. Grätzel, Chem. Mater., 2014, 26, 6160–6164.
25. G. P. Nagabhushana, R. Shivaramaiah and A. Navrotsky, Proc. Natl. Acad. Sci. U. S. A., 2016, 113, 7717–7721.
26. B. Brunetti, C. Cavallo, A. Ciccioli, G. Gigli and A. Latini, Sci. Rep., 2016, 6, 31896 . Corrigendum: Sci. Rep., in press.
27. E. J. Juarez-Perez, Z. Hawash, S. R. Raga, L. K. Ono and Y. Qi, Energy Environ. Sci., 2016, 9, 3406–3410.
28. M. Sawicka, P. Storoniak, P. Skurski, J. Błażejowski and J. Rak, Chem. Phys., 2006, 324, 425–437.
29. M. Sawicka, P. Storoniak, J. Błażejowski and J. Rak, J. Phys. Chem. A, 2006, 110, 5066–5074.
30. P. Dokurno, J. Lubkowski and J. Błażejowski, Thermochim. Acta, 1990, 165, 31–48.
31. J. Lubkowski and J. Błażejowski, Thermochim. Acta, 1990, 157, 259–277.
32. V. S. Iorish, IVTANTHERMO database-Version 3.0, Glushko Thermocenter of Russian Academy of Sciences, 2005.
33. M. Frenkel, G. J. Kabo, K. N. Marsh, G. N. Roganov and R. C. Wilhoit, Thermodynamics of Organic Compounds in the Gas State, CRC Press, 1994.
34. N. Onoda-Yamamuro, T. Matsuo and H. Suga, J. Phys. Chem. Solids, 1990, 51, 1383–1395.
35. J. N. Bull, P. W. Harland and C. Vallance, J. Phys. Chem. A, 2012, 116, 767–777.

---