Reversible and irreversible colossal barocaloric effects in plastic crystals

,


Introduction
Caloric effects across solid-state first-order phase transitions offer the possibility of controlling the exchange of latent heat by means of an external field. In view of the climate change emergency, caloric effects are being investigated for cooling applications as an alternative to replace current compressors that use greenhouse gases 1 . Most likely, finding appropriate solid materials has become the major hindrance for a commercial implementation. In addition to problems such as cost and toxicity, other more fundamental problems are encountered when looking for good caloric materials. Generally, this rational search is based on the knowledge of few requirements that are often accessible using widely extended conventional experimental techniques and/or available in the literature, that are a phase transition with large latent heat close to a desired operational temperature and a high sensitivity of the phase transition with the applied field 2,3 .
Despite caloric prototypes are showing high efficiencies 1,4 , the relatively small latent heats in most solid-solid transitions cannot compete against those exchanged in liquid-vapor transitions used in current compressors, which results in a refrigeration capacity significantly smaller. However, this paradigm could change due to the recent identification 5,6 of a family of materials named plastic crystals. Some compounds in this family undergo firstorder phase transitions with enormous latent heat mainly due to an unusual strong molecular orientational disordering, for which they have traditionally attracted interest for passive thermal energy storage applications 7,8 . On the other hand, their strongly non-isochoric transitional character makes the associated huge enthalpy changes be easily driven by pressure and therefore they have now been proposed for barocaloric (BC) effects. Following this advance, a list with a number of plastic crystals with similar nominal characteristics has been suggested as potential excellent BC candidates 6 . However, while the criteria of large latent heat and sensitivity with the applied field are a necessary filter to discard useless materials, they are not sufficient to determine appropriate compounds. Other not less important conditions must be fulfilled by the candidates in order to exhibit a good caloric performance as, for instance, good reversibility and maintenance of proper thermodynamic transition properties (i.e. large transition entropy and volume changes and sharp transition) at high applied fields.
Since cooling devices operate cyclically, the caloric effects used therein must be driven reversibly upon an indefinite sequence of application and removal of the external field, and therefore they may be dramatically affected by hysteresis. Near first-order phase transitions, the transition hysteresis introduces a history dependence in the system response to external stimuli such that caloric effects associated with the transition are irreversible below a certain threshold for the external field. Consequently, the necessary input work is increased and the temperature range of operation decreases so that the actual material performance decreases with respect to the ideal non-hysteretic frame. Although the key role of hysteresis for reversibility was well established almost one decade ago 9-11 , a significant number of studies on caloric effects do not address this issue yet. Promising materials have been declared without revealing the hysteresis width 6,[12][13][14] or reporting hysteresis widths that would entail manifestly excessive fields for a practical implementation 13,[15][16][17] . While these considerations do not invalidate previous works, it is undoubtedly important to have information about such features whenever possible for the purpose of practical applications.
The aim of the present study is to check these important features in four plastic crystals that were enumerated in the aforementioned published list of excellent BC candidates 6 , where reversibility and high-pressure properties were not taken into account. We show that, despite all of them exhibit nominal outstanding transition characteristics, two of them do not meet the expectations and do not demonstrate aptitude for BC applications. Our selected compounds belong to the family of neopentane [C(CH 3 ) 4 ] derivatives for which the methyl groups have been substituted by amino (NH 2 ) or hydroxymethyl (CH 2 OH) groups: [2,NPA]. Below the melting, at high temperature these molecules arrange in a cubic lattice: bcc (Im3m) for those containing amino groups, otherwise fcc (Fm3m), hereafter referred to as phase I, and exhibit dynamically disordered molecular orientations. On cooling across a first-order phase transition, they order completely accompanied by a symmetry reduction and a significant finite volume decrease. In particular, AMP, TRIS, NPA and PG adopt monoclinic 18 (space group undetermined), orthorhombic (Pna2 1 ) 19 , triclinic 20 (space group undetermined) and tetragonal (bct, I4) 18 structures respectively, hereafter referred to as phase II. This chemical and structural information is summarized in table 1.
All these transitions show entropy changes of hundreds of J K −1 kg −1 and large volume changes, which clearly point them as candidates for BC applications. However, our analysis reveals that both TRIS and AMP are unfeasible due to an insourmountable hysteresis at reasonable pressures, whereas instead PG and NPA demonstrate an excellent BC performance.

Experimental
Powdered samples of TRIS (≥99.8% purity), PG (99%) and NPA (99%) were purchased from Sigma-Aldrich and AMP (≥99.5%) was purchased from Fluka and used as received. Calorimetry at normal pressure was carried out using a conventional Differential Scanning Calorimeter Q100 from TA Instruments. Pressuredependent calorimetry was performed on two different highpressure differential thermal analyzers. One high-pressure cell is a custom-built irimo block that uses Bridgman thermal sensors and operates up to 3 kbar within a temperature range from room-temperature to 473 K controlled by a resistive heater. The other one is a high-pressure Cu-Be cell from Unipress (Poland) that uses Peltier Modules as thermal sensors and operates up to 6 kbar within a temperature range from 200 K to 393 K controlled by an external thermal bath. In both cells, few hundreds of mg of each sample were mixed with a perfluorinated inert fluid (Galden Bioblock Scientist) to remove air and encapsulated in tin capsules. The pressure-transmitting fluid was DW-Therm M90.200.02 (Huber). Calorimetry measurements coherently yielded positive dQ/|dT | ≡Q/|Ṫ | peaks corresponding to the endothermic (II→I) transitions and negative dQ/|dT | peaks corresponding to the exothermic (I→II) transitions. The respective transition temperatures, T II→I and T I→II , were determined from the onset of the calorimetric peaks. Transition enthalpy changes ∆H II→I and ∆H I→II , and transition entropy changes ∆S II→I and ∆S I→II , were determined from integrations of dQ/d|T | and of (1/T )dQ/d|T | peaks after baseline subtraction, respectively.

General aspects on reversibility and hysteresis
For a reliable estimation of reversible caloric effects, direct and quasi-direct methods are normally preferable over indirect measurements 2 as the use of the thermodynamic Maxwell relation may fail due to misleading protocols 21,22 or strong out-ofequilibrium dynamics [23][24][25] . While direct methods naturally yield reversible effects because they involve the same changing-field dynamics, quasi-direct methods are carried out by changing temperature at constant applied fields 2 so that in these cases the determination of reversible effects needs transferring the measured isofield thermally-driven hysteretic effects to isothermal fielddriven hysteretic effects via calculations. In this respect, comparison between direct and quasi-direct experimental data has shown that reversible isothermal entropy changes ∆S rev can be obtained from the overlapping between isothermal entropy changes ∆S derived from isofield measurements performed on heating and on cooling independently 11,26 (see Fig. 1a,c). For materials displaying conventional BC effects (dT II↔I /d p(p) > 0, where II↔I means the set of II→I and I→II), and assuming hereafter that BC effects will take place always with the lower pressure being atmospheric pressure, p atm , the overlapping condition entails that the minimum pressure, p rev , to achieve nonzero ∆S rev is given by the value at which the exothermic transition temperature equals the endothermic transition temperature at atmospheric pressure, i.e., T I→II (p rev ) = T II→I (p atm ). From this equality, we can write ∆T II↔I (p atm ) is the thermal hysteresis at atmospheric pressure defined as ∆T II↔I (p atm ) = T II→I (p atm ) − T I→II (p atm ). Instead, the minimum pressure to obtain reversible adiabatic temperature changes ∆T rev is usually higher than p rev . In particular, where we denote as T f I→II (p atm ) the finishing temperature on cooling across the transition I→II (see Fig. 1b,d). Notice that for ideal perfectly isothermal transitions, the ranges for ∆T rev and ∆S rev are the same because T f I→II = T I→II . Therefore, ∆T II↔I (p) and dT I→II /d p(p) represent key properties in the study of the viability of BC materials, which can be generalized to any other caloric effect. Moreover, if cycles involving lowest pressures above normal pressure are considered, then the knowledge of dT II→I /d p(p) is also required. It is worth pointing out here that for materials exhibiting inverse behavior (dT I↔II /d p(p) < 0) the same quantities need to be evaluated for the inverse transitions.
In addition, establishing the dependence of the transition thermodynamic quantities on the applied pressure 27 has been shown to become important because a possible but a priori unknown decrease of the transition entropy change with pressure might reduce substantially the caloric performance. This has been observed to occur, for instance, in some magnetostructural 28,29 , ferroelectric 30,31 and hybrid organic-inorganic systems 32 . Otherwise, omitting this feature may lead to large uncertainties or inexact conclusions such as a systematic overestimation of the caloric effects or false expectations.

Results and discussion
Calorimetry measurements performed at atmospheric pressure carried out for all compounds (not shown) yield endothermic (II→I) and exothermic (I→II) transition properties listed in Table 2. The endothermic values are in overall agreement with reported literature data 19,33 and confirm their colossal transition entropy changes. The exothermic values have not been reported so far and the differences from endothermic values arise due to the transition hysteresis.

Contributions to the transition entropy change
In the compounds under study, the total entropy change can be divided into two contributions: ∆S II→I = ∆S e II→I + ∆S c II→I where the first term arises from the lattice distortion whereas the second term originates in the release of different orientational conformations.
The strain entropy change can be calculated with respect to the unstrained cubic phase in therms of the Helmholtz free energy , where e refers to the strain. In the cubic system, F e can be written up to the harmonic approximation as F e = Ke 2 1 +C ′ (e 2 2 + e 2 3 ) +C 44 (e 2 4 + e 2 5 + e 2 6 ). Here, K = C 11 +2C 21 3 and C ′ = C 11 −C 12 2 are the bulk and deviatoric moduli respectively, with C i j being components of the stiffness tensor, and e i are the symmetry-adapted strains 34 of the cubic system corresponding to volumic (e 1 ), deviatoric (e 2 and e 3 ) and shear strains (e 4 , e 5 and e 6 ). On the other hand, the experimental measurement of elastic constants in plastic crystals is hard due to the difficulty in growing large single crystals, and available literature data are absent for the compounds under study. To estimate the order of magnitude, we have used elastic constants as a function of temperature for plastic crystal fullerene C 60 which, averaged over the full reported 35 temperature range in the ordered phase, yield ∂ K ∂ T ∼ −3.8 · 10 7 GPa K −1 , ∂C ′ ∂ T ∼ −4.7 · 10 6 GPa K −1 and ∂C 44 ∂ T ∼ −2.0 · 10 7 GPa K −1 . For PG, the lattice parameters near the fcc-bct deformation 20 are a I = 8.876 Å, a II = 6.052 Å and c II = 8.872 Å, which correspond to 36 the symmetry-adapted strains e 1 = −0.046, e 3 = 0.032 and e 2 = e 4 = e 5 = e 6 = 0. Therefore, taking a density of ρ = 1.23 g cm −3 , we obtain an entropy change associated with the deviatoric strain ∆S d II→I ≃ 4 J K −1 kg −1 and an entropy change associated with the volume change ∆S V II→I ≃ 67 J K −1 kg −1 , which are significantly smaller than the total entropy change, ∆S II→I ≃ 485 J K −1 kg −1 . For the remaining compounds, the reconstructive nature of the analyzed phase transitions in those cases prevents establishing a correspondence between the unstrained and the strained lattices and therefore the strain entropy change has no meaning.
Alternatively, the volumic entropy change can be approximately estimated as ∆S V II→I = ( α / β ) ∆V II→I , where α and β correspond to the thermal expansion and the isothermal compressibility, respectively, averaged over the two phases close to the phase transition 37,38 . Using dilatometry and x-ray diffraction data for NPA from Refs 39,40 we find ∆S V II→I ∼ 52 ± 10 J K −1 kg −1 , which is considerably smaller than the total entropy change, ∆S II→I ∼ 200 J K −1 kg −1 . Although a complete set of data is not available in the literature for any of the remaining compounds, existing data 19,41,42 strongly indicate that for PG and for AMP ∆S V II→I ∼ 100 ± 50 J K −1 kg −1 . These numbers for ∆S V II→I , even with such a prudent interval of uncertainty, lie in all cases well below half the corresponding total transition entropy changes ∆S II→I , which means that for all compounds the major contribution to ∆S II→I emerges from the orientational disordering at the II→I transition, ∆S c II→I . The contribution of the orientational disordering to these total entropy changes can be calculated as ∆S c II→I = (RM −1 )ln(N I /N II ), where R is the universal gas constant, M is the molar mass and N I and N II are the number of possible molecular conformations in phases I and II, respectively. While in general in plastic crystals phase II can still exhibit some disorder (i.e. N II > 1) [43][44][45][46][47] , for all the compounds here studied N II = 1 can be reasonably assumed because hydrogen bonds do not permit any molecular disorder 48 . Notice, however, that in other cases the presence of hydrogen bonds may not be a sufficient condition either to have N II = 1 49 . With respect to the disordered phase, for those compounds with an fcc unit cell (PG and NPA, point group T d , subgroup C 3v ) there are 10 possible orientations whereas those with a bcc unit cell (AMP and TRIS, point group D 2d ) there are 6 possible orientations 50 . For each orientation, PG, NPA, AMP and TRIS have 11, 3, 9 and 11 different conformations, respectively. Therefore, we obtain the following nominal values: For NPA, ∆S c II→I = 321 J K −1 kg −1 ; for PG, ∆S c II→I = 325 J K −1 kg −1 ; for TRIS, ∆S c II→I = 287 J K −1 kg −1 ; for AMP ∆S c II→I = 315 J K −1 kg −1 . As expected, these values are substantially larger than the values obtained for the corresponding ∆S V II→I but they are inconsistent with the values obtained for the total entropy change ∆S II→I . This means that the information for ∆S c II→I is incomplete. For instance, for NPA not all theoretically possible conformations are observed via Raman spectroscopy 51 , which explains the overestimation of the aforegiven value for ∆S c II→I . In turn, for TRIS and AMP, additional π/2 rotations of the amino groups could take place, increasing the resulting ∆S c II→I . It is thus clear that an accurate determination for ∆S c II→I requires further investigation and lies beyond the scope of this work.

Evaluation of barocaloric reversible ranges
To evaluate the minimum pressure p rev above which ∆S rev becomes nonzero, the use of eq. 1 requires the knowledge of dT I→II /d p(p) and ∆T II↔I (p atm ). In the following we analyze these two quantities and estimate p rev for the compounds under study.
Values for ∆T II↔I (p atm ) can be directly obtained from Table 2. With respect to dT I→II /d p(p), its determination via the Clausius-Clapeyron (CC) equation, dT II↔I d p (p) = ∆V II→I ∆S II→I , may be not reliable because (i) it is based on equilibrium conditions (i.e. endothermic, II→I transition) and (ii) ∆S II→I and ∆V II→I are usually available only at normal pressure. Consequently, significant differences between dT II→I /d p at normal pressure, dT II→I /d p(p atm ), (as derived from CC) and dT I→II /d p(p) at high pressure can arise. In summary, atmospheric-pressure data can be used as a rough estimate only as they may lead to large uncertainties or incorrect conclusions. Therefore, to rigorously determine the actual reversible ranges of pressure via the quasi-direct method, the knowledge of the exothermic transition temperatures as a function of pressure, T I→II (p), is imperative. To derive these values we perform pressure-dependent isobaric calorimetry on heating and cooling across the II↔I transition at temperature rates |Ṫ | ∼ 2 − 4 K min −1 (see Fig. 2a-d). The pressure-dependent transition temperatures determined from the peak onsets (see Fig. 2e-h) confirm the aforementioned discrepancies between dT II→I /d p at p atm and at p > p atm and dT I→II /d p for all pressures. For TRIS (see Fig. 2e) the hysteresis ∆T II↔I (p) increases with increasing pressure because for all pressures we obtain dT II→I /d p = 3.7 ± 0.2 K kbar −1 > dT I→II /d p = 1.5 ± 0.6 K kbar −1 . The latter magnitude is clearly insufficient to overcome ∆T II↔I (p atm ) ∼ 75 K at reasonable pressures, as eq. 1 renders p rev ∼ 50 kbar. For PG (see Fig. 2f) the hysteresis ∆T II↔I (p atm ) ∼ 3.7 K can be easily overcome given the magnitude for dT I→II /d p = 9.4 ± 0.3 K kbar −1 , which yields a low reversible threshold p rev ∼ 0.4 kbar. In this case, ∆T II↔I (p) decreases with increasing pressure because for all pressures dT I→II /d p = 9.4 ± 0.3 K kbar −1 > dT II→I /d p = 7.9 ± 0.3 K kbar −1 . Notice that the improper use of the latter value in eq. 1 would give rise to the overestimated value p rev ∼ 0.5 kbar. As for AMP (see Fig. 2g), exothermic transitions are obtained with large hysteresis close to normal pressures (∆T II↔I (p ∼ 0.5 kbar) ∼ 50 K), slightly decreasing with increasing pressure because for all pressures dT II→I /d p = 6.4 ± 0.2 K kbar −1 < dT I→II /d p = 8.5 ± 0.7 K kbar −1 , but still too large to be overcome by moderate pressures given the relatively small dT I↔II /d p. In particular we obtain p rev ∼ 6 kbar. Concerning NPA, it exhibits dT II→I /d p(p atm ) = 22 ± 1 K kbar −1 but dT II→I /d p(p ∼ 6 kbar) = 7 ± 2 K kbar −1 and, more importantly, dT I→II /d p = 11.9 ± 0.2 K kbar −1 for all pressures. Given these values (and using ∆T II↔I (p atm ) ∼ 23 K), we obtain p rev ∼ 1.9 kbar which in this case is twice larger than that expected from endothermic data or CC equation at atmospheric pressure. In fact, the out-of-equilibrium T − p diagrams shown in Figs 2e-h permit to establish graphically the reversible threshold p rev as indicated by the dashed purple lines according to the aforementioned condition T I→II (p rev ) = T II→I (p atm ). A summary of the transition characteristics at atmospheric pressure and values for p rev is given in Table 2.

Construction of isobaric entropy curves
To calculate the BC effects using the quasi-direct method, we have computed the isobaric entropy curves S ′ (T, p) at different pressures with respect to a reference value taken at normal pressure and at a temperature T 0 arbitrarily chosen well below the I→II transition temperature for each compound (see Fig. 3) so that S ′ (T, p) = S(T, p) − S(T 0 , p atm ). For this purpose we have followed the procedure described in Refs 5,52,53 according to which we need integration of our pressure-dependent calorimetric peaks after baseline subtraction and temperature-dependent heat capacity C p from literature (NPA: 54 ; AMP: 55 ; PG: 56 ; TRIS: 57,58 ).
The fact that C p data is available only at normal pressure prevents us to determine the pressure dependence of the isobaric entropy in each phase directly from this quantity. However, this lack can be compensated by the use of the Maxwell relation ∆S + (p atm → p 1 ) = − p 1 p atm (∂V /∂ T ) p d p, which establishes the entropy change when a pressure is applied in isothermal conditions outside the transition. In particular, given the conventional character of our materials, we include ∆S + in phase II 30 . To calculate ∆S + , (∂V /∂ T ) p is taken from literature (NPA: 39 , AMP: 59 ; TRIS: 42 ; PG: 60 ) and assumed to be independent of pressure within the range under study, which is a reasonable assumption as derived from Refs 5,40,41 . Nonetheless, it can be estimated that 1-9 | 5 Table 2 Thermodynamic properties for both endothermic (II→I) and exothermic (I→II) transitions measured at atmospheric pressure.  an uncertainty of 20% in (∂V /∂ T ) p would yield an error of ∼ 3% in the BC effects obtained across the transition, which is smaller than the uncertainty interval of our results.
Although in some materials such as metallic alloys neglecting ∆S + may be irrelevant because (∂V /∂ T ) p atm is small, in other materials ∆S + can influence substantially the resulting BC effects and therefore this contribution cannot be left out. For instance, ∆S + accounts for a large decrease of the BC effects in ammonium sulphate 30 whereas, more interestingly, it has been demonstrated that neglecting ∆S + in the plastic crystal Neopentylglycol ([(CH 3 ) 2 C(CH 2 OH) 2 ], NPG) as done elsewhere 6 would omit a 25% increase in ∆S at 5 kbar 5 . In the compounds analyzed here, ∆S + becomes particularly significant for NPA, as shown in Fig.  3(a,b). On the other hand, it is worth noticing here that ∆S + included in phase II, together with the knowledge of the transition entropy change ∆S II↔I (shown in Figs 2i-l) and C p , determines univocally ∆S + in phase I. Therefore, the experimental determination of ∆S + in phase I can be used to ascertain the consistency and uncertainty of different types of measurements. All the compounds studied here show 40 (∂V /∂ T ) p atm (I) ≥ (∂V /∂ T ) p atm (II), from which we can deduce that ∆S + (I) ≥ ∆S + (II). This inequality, along with the fact that in both phases the entropy varies with temperature similarly, mathematically imposes ∆S II↔I to be roughly constant or to decrease when increasing pressure. Indeed, this is consistent with pressure-dependent transition entropy change for the endothermic (∆S II→I ) and exothermic (∆S I→II ) transitions (see Fig.  2i-l) derived from integration of (1/T )dQ/|dT |. In particular, we find that ∆S II↔I for PG and AMP, and ∆S II→I for TRIS are roughly independent of pressure, but for TRIS ∆S I→II decreases according to d∆S I→II /d p = −19 ± 6 J K −1 kg −1 kbar −1 whereas for NPA d∆S II→I /d p = −4 ± 2 J K −1 kg −1 kbar −1 and d∆S I→II /d p = −10 ± 2 J K −1 kg −1 kbar −1 .

Barocaloric response
By subtracting isobaric entropies at different pressures following adiabatic and isothermal paths, we have obtained the irreversible isothermal entropy changes ∆S(p 1 → p 2 ) = S(T, p 2 ) − S(T, p 1 ) = S ′ (T, p 2 ) − S ′ (T, p 1 ) and adiabatic temperature changes ∆T (p 1 → p 2 ) = T (S, p 2 ) − T (S, p 1 ) = T (S ′ , p 2 ) − T (S ′ , p 1 ) where the lower pressure will be taken always as atmospheric pressure p atm . In materials displaying conventional BC effects, dT II↔I /d p > 0 entails that compression-induced transitions are exothermic whereas decompression-induced transitions are endothermic. Therefore, BC effects on compression are computed from cooling runs and BC effects on decompression from heating runs (see Fig. 4). Irreversible ∆S achieves colossal values in all cases at small pressures, reaching on average over compression and decompression up to 600 ± 60 J K −1 kg −1 and 8 ± 2 K under 2.5 kbar for TRIS, 510 ± 50 J K −1 kg −1 and 22 ± 2 K under 2.5 kbar for PG, 690 ± 70 J K −1 kg −1 and 15.5 ± 2.0 K under 2.5 kbar for AMP and 330 ± 30 J K −1 kg −1 and 35 ± 5 K under 2.7 kbar for NPA, that increase up to 500 ± 50 J K −1 kg −1 and 62 ± 6 K under 5.9 kbar. Although these values are impressive, they are unuseful to determine the adequacy of a material for implementation in a real device because they cannot be reproduced cyclically after the first pressure change. As discussed previously, reversible entropy changes, ∆S rev , are computed quasi-directly as the overlapping between ∆S obtained on compression and decompression 11 whereas reversible adiabatic temperature changes, ∆T rev are determined from adiabatic differences between T (S, p atm ) on heating and T (S, p) on cooling 52 . The corresponding results confirm the aforementioned conclusions: BC effects in AMP and TRIS are irreversible within the whole pressure range analyzed here. Instead, for PG reversible effects (see 5a,c) are found above p rev ∼ 0.4 kbar, displaying ∆S rev = 490 ± 50 J K −1 kg −1 and ∆T rev = 10 ± 1 K under p ∼ 2.4 kbar. In turn, reversible effects (see Fig. 5b,d) are found above p rev ∼ 1.9 kbar in NPA, displaying ∆S rev = 290 ± 30 J K −1 kg −1 and ∆T rev = 16 ± 2 K under p ∼ 2.6 kbar, and ∆S rev = 470 ± 50 J K −1 kg −1 and ∆T rev = 42 ± 4 K under p ∼ 5.8 kbar. Figure 6 shows a summary of the maximum absolute irreversible (panels a-c) and reversible (d-f) BC effects as a function of the applied pressure for the compounds under study. Data for NPG 5 belonging to the same plastic crystal family, are also displayed for comparison.
The reversible values obtained for NPA and PG are comparable to the thermal response of harmful fluids used in current refrigerators 61 yet generated by greenhouse emission-free materials. Therefore, they are excellent candidates to be employed in future more sustainable cooling devices. The low thermal conductivity of our compounds can be palliated by an appropriate design that maximizes the surface-to-volume ratio such as powdered material embedded into graphite matrices 62 or the introduction of high thermal conductive nanoparticles 63 . It must be noted here that, in terms of safety, NPA is classified 64 as a flammable solid (flash point of 301 K), which should be taken into account for applications at room temperature. Instead, for PG 65 (and AMP 66 and TRIS 67 ) no relevant hazards are identified.

Conclusions
We have studied the barocaloric response of four plastic crystals derived from neopentane across their ordered-orientationally disordered solid-state phase transition. For this purpose we have constructed isobaric entropy-temperature curves below, across and above the transition using our high-pressure calorimetry, and temperature-dependent heat capacity and volume from literature. The latter accounts for additional entropy changes due to thermal expansion of each phase, which appears to be important for NPA and negligible within error for the remaining compounds. We have shown that to provide useful information for technological applications, the characterization of the reversible BC effects from high-pressure properties must be carried out. As a result, while all compounds studied here show outstanding BC properties upon the first application or removal of pressure, those plastic crystals having a bcc structure (AMP and TRIS) display too large hysteresis to be overcome by application of reasonable pressures and therefore they can be discarded as promising BC materials. Instead, those plastic crystals with an fcc structure (PG and NPA) transform more easily to the ordered phase, and therefore do show colossal reversible effects at moderate pressures. Our study reinforces the recently demonstrated great reversible BC potential of some plastic crystals for future cooling devices and should incite future works on caloric compounds to systematically provide reversible values as a way to accelerate the development of advanced materials for environmentally-friendly solid-state refrigeration.

Conflicts of interest
The use of the compounds studied in this work and other plastic crystals for barocaloric cooling is covered in the following patent: X. Moya, A. Avramenko, L. Mañosa, J.-Ll. Tamarit and P. Lloveras, Use of barocaloric materials and barocaloric devices, PCT/EP2017/076203 (2017). The remaining authors declare no conflicts of interests.